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

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86
  87 #ifdef CONFIG_IPV6_PRIVACY
  88 #include <linux/random.h>
  89 #endif
  90
  91 #include <linux/uaccess.h>
  92 #include <asm/unaligned.h>
  93
  94 #include <linux/proc_fs.h>
  95 #include <linux/seq_file.h>
  96 #include <linux/export.h>
  97
  98 /* Set to 3 to get tracing... */
  99 #define ACONF_DEBUG 2
 100
 101 #if ACONF_DEBUG >= 3
 102 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 103 #else
 104 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 105 #endif
 106
 107 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 108
 109 static inline u32 cstamp_delta(unsigned long cstamp)
 110 {
 111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112 }
 113
 114 #ifdef CONFIG_SYSCTL
 115 static void addrconf_sysctl_register(struct inet6_dev *idev);
 116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117 #else
 118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 119 {
 120 }
 121
 122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 123 {
 124 }
 125 #endif
 126
 127 #ifdef CONFIG_IPV6_PRIVACY
 128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130 static void ipv6_regen_rndid(unsigned long data);
 131 #endif
 132
 133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134 static int ipv6_count_addresses(struct inet6_dev *idev);
 135
 136 /*
 137  *      Configured unicast address hash table
 138  */
 139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 140 static DEFINE_SPINLOCK(addrconf_hash_lock);
 141
 142 static void addrconf_verify(unsigned long);
 143
 144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 145 static DEFINE_SPINLOCK(addrconf_verify_lock);
 146
 147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 149
 150 static void addrconf_type_change(struct net_device *dev,
 151                                  unsigned long event);
 152 static int addrconf_ifdown(struct net_device *dev, int how);
 153
 154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 155                                                   int plen,
 156                                                   const struct net_device *dev,
 157                                                   u32 flags, u32 noflags);
 158
 159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 160 static void addrconf_dad_timer(unsigned long data);
 161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 162 static void addrconf_dad_run(struct inet6_dev *idev);
 163 static void addrconf_rs_timer(unsigned long data);
 164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 166
 167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 168                                 struct prefix_info *pinfo);
 169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 170                                struct net_device *dev);
 171
 172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 173         .forwarding             = 0,
 174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 175         .mtu6                   = IPV6_MIN_MTU,
 176         .accept_ra              = 1,
 177         .accept_redirects       = 1,
 178         .autoconf               = 1,
 179         .force_mld_version      = 0,
 180         .mldv1_unsolicited_report_interval = 10 * HZ,
 181         .mldv2_unsolicited_report_interval = HZ,
 182         .dad_transmits          = 1,
 183         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 184         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 185         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 186 #ifdef CONFIG_IPV6_PRIVACY
 187         .use_tempaddr           = 0,
 188         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 189         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 190         .regen_max_retry        = REGEN_MAX_RETRY,
 191         .max_desync_factor      = MAX_DESYNC_FACTOR,
 192 #endif
 193         .max_addresses          = IPV6_MAX_ADDRESSES,
 194         .accept_ra_defrtr       = 1,
 195         .accept_ra_pinfo        = 1,
 196 #ifdef CONFIG_IPV6_ROUTER_PREF
 197         .accept_ra_rtr_pref     = 1,
 198         .rtr_probe_interval     = 60 * HZ,
 199 #ifdef CONFIG_IPV6_ROUTE_INFO
 200         .accept_ra_rt_info_max_plen = 0,
 201 #endif
 202 #endif
 203         .proxy_ndp              = 0,
 204         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 205         .disable_ipv6           = 0,
 206         .accept_dad             = 1,
 207         .suppress_frag_ndisc    = 1,
 208 };
 209
 210 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 211         .forwarding             = 0,
 212         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 213         .mtu6                   = IPV6_MIN_MTU,
 214         .accept_ra              = 1,
 215         .accept_redirects       = 1,
 216         .autoconf               = 1,
 217         .force_mld_version      = 0,
 218         .mldv1_unsolicited_report_interval = 10 * HZ,
 219         .mldv2_unsolicited_report_interval = HZ,
 220         .dad_transmits          = 1,
 221         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 222         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 223         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 224 #ifdef CONFIG_IPV6_PRIVACY
 225         .use_tempaddr           = 0,
 226         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 227         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 228         .regen_max_retry        = REGEN_MAX_RETRY,
 229         .max_desync_factor      = MAX_DESYNC_FACTOR,
 230 #endif
 231         .max_addresses          = IPV6_MAX_ADDRESSES,
 232         .accept_ra_defrtr       = 1,
 233         .accept_ra_pinfo        = 1,
 234 #ifdef CONFIG_IPV6_ROUTER_PREF
 235         .accept_ra_rtr_pref     = 1,
 236         .rtr_probe_interval     = 60 * HZ,
 237 #ifdef CONFIG_IPV6_ROUTE_INFO
 238         .accept_ra_rt_info_max_plen = 0,
 239 #endif
 240 #endif
 241         .proxy_ndp              = 0,
 242         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 243         .disable_ipv6           = 0,
 244         .accept_dad             = 1,
 245         .suppress_frag_ndisc    = 1,
 246 };
 247
 248 /* Check if a valid qdisc is available */
 249 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 250 {
 251         return !qdisc_tx_is_noop(dev);
 252 }
 253
 254 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 255 {
 256         if (del_timer(&idev->rs_timer))
 257                 __in6_dev_put(idev);
 258 }
 259
 260 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 261 {
 262         if (del_timer(&ifp->dad_timer))
 263                 __in6_ifa_put(ifp);
 264 }
 265
 266 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 267                                   unsigned long when)
 268 {
 269         if (!timer_pending(&idev->rs_timer))
 270                 in6_dev_hold(idev);
 271         mod_timer(&idev->rs_timer, jiffies + when);
 272 }
 273
 274 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 275                                    unsigned long when)
 276 {
 277         if (!timer_pending(&ifp->dad_timer))
 278                 in6_ifa_hold(ifp);
 279         mod_timer(&ifp->dad_timer, jiffies + when);
 280 }
 281
 282 static int snmp6_alloc_dev(struct inet6_dev *idev)
 283 {
 284         int i;
 285
 286         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 287                           sizeof(struct ipstats_mib),
 288                           __alignof__(struct ipstats_mib)) < 0)
 289                 goto err_ip;
 290
 291         for_each_possible_cpu(i) {
 292                 struct ipstats_mib *addrconf_stats;
 293                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i);
 294                 u64_stats_init(&addrconf_stats->syncp);
 295 #if SNMP_ARRAY_SZ == 2
 296                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i);
 297                 u64_stats_init(&addrconf_stats->syncp);
 298 #endif
 299         }
 300
 301
 302         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 303                                         GFP_KERNEL);
 304         if (!idev->stats.icmpv6dev)
 305                 goto err_icmp;
 306         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 307                                            GFP_KERNEL);
 308         if (!idev->stats.icmpv6msgdev)
 309                 goto err_icmpmsg;
 310
 311         return 0;
 312
 313 err_icmpmsg:
 314         kfree(idev->stats.icmpv6dev);
 315 err_icmp:
 316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 317 err_ip:
 318         return -ENOMEM;
 319 }
 320
 321 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 322 {
 323         struct inet6_dev *ndev;
 324
 325         ASSERT_RTNL();
 326
 327         if (dev->mtu < IPV6_MIN_MTU)
 328                 return NULL;
 329
 330         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 331
 332         if (ndev == NULL)
 333                 return NULL;
 334
 335         rwlock_init(&ndev->lock);
 336         ndev->dev = dev;
 337         INIT_LIST_HEAD(&ndev->addr_list);
 338         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 339                     (unsigned long)ndev);
 340         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 341         ndev->cnf.mtu6 = dev->mtu;
 342         ndev->cnf.sysctl = NULL;
 343         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 344         if (ndev->nd_parms == NULL) {
 345                 kfree(ndev);
 346                 return NULL;
 347         }
 348         if (ndev->cnf.forwarding)
 349                 dev_disable_lro(dev);
 350         /* We refer to the device */
 351         dev_hold(dev);
 352
 353         if (snmp6_alloc_dev(ndev) < 0) {
 354                 ADBG(KERN_WARNING
 355                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 356                         __func__, dev->name);
 357                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 358                 dev_put(dev);
 359                 kfree(ndev);
 360                 return NULL;
 361         }
 362
 363         if (snmp6_register_dev(ndev) < 0) {
 364                 ADBG(KERN_WARNING
 365                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 366                         __func__, dev->name);
 367                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 368                 ndev->dead = 1;
 369                 in6_dev_finish_destroy(ndev);
 370                 return NULL;
 371         }
 372
 373         /* One reference from device.  We must do this before
 374          * we invoke __ipv6_regen_rndid().
 375          */
 376         in6_dev_hold(ndev);
 377
 378         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 379                 ndev->cnf.accept_dad = -1;
 380
 381 #if IS_ENABLED(CONFIG_IPV6_SIT)
 382         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 383                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 384                 ndev->cnf.rtr_solicits = 0;
 385         }
 386 #endif
 387
 388 #ifdef CONFIG_IPV6_PRIVACY
 389         INIT_LIST_HEAD(&ndev->tempaddr_list);
 390         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 391         if ((dev->flags&IFF_LOOPBACK) ||
 392             dev->type == ARPHRD_TUNNEL ||
 393             dev->type == ARPHRD_TUNNEL6 ||
 394             dev->type == ARPHRD_SIT ||
 395             dev->type == ARPHRD_NONE) {
 396                 ndev->cnf.use_tempaddr = -1;
 397         } else {
 398                 in6_dev_hold(ndev);
 399                 ipv6_regen_rndid((unsigned long) ndev);
 400         }
 401 #endif
 402         ndev->token = in6addr_any;
 403
 404         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 405                 ndev->if_flags |= IF_READY;
 406
 407         ipv6_mc_init_dev(ndev);
 408         ndev->tstamp = jiffies;
 409         addrconf_sysctl_register(ndev);
 410         /* protected by rtnl_lock */
 411         rcu_assign_pointer(dev->ip6_ptr, ndev);
 412
 413         /* Join interface-local all-node multicast group */
 414         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 415
 416         /* Join all-node multicast group */
 417         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 418
 419         /* Join all-router multicast group if forwarding is set */
 420         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 421                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 422
 423         return ndev;
 424 }
 425
 426 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 427 {
 428         struct inet6_dev *idev;
 429
 430         ASSERT_RTNL();
 431
 432         idev = __in6_dev_get(dev);
 433         if (!idev) {
 434                 idev = ipv6_add_dev(dev);
 435                 if (!idev)
 436                         return NULL;
 437         }
 438
 439         if (dev->flags&IFF_UP)
 440                 ipv6_mc_up(idev);
 441         return idev;
 442 }
 443
 444 static int inet6_netconf_msgsize_devconf(int type)
 445 {
 446         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 447                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 448
 449         /* type -1 is used for ALL */
 450         if (type == -1 || type == NETCONFA_FORWARDING)
 451                 size += nla_total_size(4);
 452 #ifdef CONFIG_IPV6_MROUTE
 453         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 454                 size += nla_total_size(4);
 455 #endif
 456
 457         return size;
 458 }
 459
 460 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 461                                       struct ipv6_devconf *devconf, u32 portid,
 462                                       u32 seq, int event, unsigned int flags,
 463                                       int type)
 464 {
 465         struct nlmsghdr  *nlh;
 466         struct netconfmsg *ncm;
 467
 468         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 469                         flags);
 470         if (nlh == NULL)
 471                 return -EMSGSIZE;
 472
 473         ncm = nlmsg_data(nlh);
 474         ncm->ncm_family = AF_INET6;
 475
 476         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 477                 goto nla_put_failure;
 478
 479         /* type -1 is used for ALL */
 480         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 481             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 482                 goto nla_put_failure;
 483 #ifdef CONFIG_IPV6_MROUTE
 484         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 485             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 486                         devconf->mc_forwarding) < 0)
 487                 goto nla_put_failure;
 488 #endif
 489         return nlmsg_end(skb, nlh);
 490
 491 nla_put_failure:
 492         nlmsg_cancel(skb, nlh);
 493         return -EMSGSIZE;
 494 }
 495
 496 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 497                                   struct ipv6_devconf *devconf)
 498 {
 499         struct sk_buff *skb;
 500         int err = -ENOBUFS;
 501
 502         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 503         if (skb == NULL)
 504                 goto errout;
 505
 506         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 507                                          RTM_NEWNETCONF, 0, type);
 508         if (err < 0) {
 509                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 510                 WARN_ON(err == -EMSGSIZE);
 511                 kfree_skb(skb);
 512                 goto errout;
 513         }
 514         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 515         return;
 516 errout:
 517         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 518 }
 519
 520 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 521         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 522         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 523 };
 524
 525 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 526                                      struct nlmsghdr *nlh)
 527 {
 528         struct net *net = sock_net(in_skb->sk);
 529         struct nlattr *tb[NETCONFA_MAX+1];
 530         struct netconfmsg *ncm;
 531         struct sk_buff *skb;
 532         struct ipv6_devconf *devconf;
 533         struct inet6_dev *in6_dev;
 534         struct net_device *dev;
 535         int ifindex;
 536         int err;
 537
 538         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 539                           devconf_ipv6_policy);
 540         if (err < 0)
 541                 goto errout;
 542
 543         err = EINVAL;
 544         if (!tb[NETCONFA_IFINDEX])
 545                 goto errout;
 546
 547         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 548         switch (ifindex) {
 549         case NETCONFA_IFINDEX_ALL:
 550                 devconf = net->ipv6.devconf_all;
 551                 break;
 552         case NETCONFA_IFINDEX_DEFAULT:
 553                 devconf = net->ipv6.devconf_dflt;
 554                 break;
 555         default:
 556                 dev = __dev_get_by_index(net, ifindex);
 557                 if (dev == NULL)
 558                         goto errout;
 559                 in6_dev = __in6_dev_get(dev);
 560                 if (in6_dev == NULL)
 561                         goto errout;
 562                 devconf = &in6_dev->cnf;
 563                 break;
 564         }
 565
 566         err = -ENOBUFS;
 567         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 568         if (skb == NULL)
 569                 goto errout;
 570
 571         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 572                                          NETLINK_CB(in_skb).portid,
 573                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 574                                          -1);
 575         if (err < 0) {
 576                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 577                 WARN_ON(err == -EMSGSIZE);
 578                 kfree_skb(skb);
 579                 goto errout;
 580         }
 581         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 582 errout:
 583         return err;
 584 }
 585
 586 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 587                                       struct netlink_callback *cb)
 588 {
 589         struct net *net = sock_net(skb->sk);
 590         int h, s_h;
 591         int idx, s_idx;
 592         struct net_device *dev;
 593         struct inet6_dev *idev;
 594         struct hlist_head *head;
 595
 596         s_h = cb->args[0];
 597         s_idx = idx = cb->args[1];
 598
 599         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 600                 idx = 0;
 601                 head = &net->dev_index_head[h];
 602                 rcu_read_lock();
 603                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 604                           net->dev_base_seq;
 605                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 606                         if (idx < s_idx)
 607                                 goto cont;
 608                         idev = __in6_dev_get(dev);
 609                         if (!idev)
 610                                 goto cont;
 611
 612                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 613                                                        &idev->cnf,
 614                                                        NETLINK_CB(cb->skb).portid,
 615                                                        cb->nlh->nlmsg_seq,
 616                                                        RTM_NEWNETCONF,
 617                                                        NLM_F_MULTI,
 618                                                        -1) <= 0) {
 619                                 rcu_read_unlock();
 620                                 goto done;
 621                         }
 622                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 623 cont:
 624                         idx++;
 625                 }
 626                 rcu_read_unlock();
 627         }
 628         if (h == NETDEV_HASHENTRIES) {
 629                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 630                                                net->ipv6.devconf_all,
 631                                                NETLINK_CB(cb->skb).portid,
 632                                                cb->nlh->nlmsg_seq,
 633                                                RTM_NEWNETCONF, NLM_F_MULTI,
 634                                                -1) <= 0)
 635                         goto done;
 636                 else
 637                         h++;
 638         }
 639         if (h == NETDEV_HASHENTRIES + 1) {
 640                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 641                                                net->ipv6.devconf_dflt,
 642                                                NETLINK_CB(cb->skb).portid,
 643                                                cb->nlh->nlmsg_seq,
 644                                                RTM_NEWNETCONF, NLM_F_MULTI,
 645                                                -1) <= 0)
 646                         goto done;
 647                 else
 648                         h++;
 649         }
 650 done:
 651         cb->args[0] = h;
 652         cb->args[1] = idx;
 653
 654         return skb->len;
 655 }
 656
 657 #ifdef CONFIG_SYSCTL
 658 static void dev_forward_change(struct inet6_dev *idev)
 659 {
 660         struct net_device *dev;
 661         struct inet6_ifaddr *ifa;
 662
 663         if (!idev)
 664                 return;
 665         dev = idev->dev;
 666         if (idev->cnf.forwarding)
 667                 dev_disable_lro(dev);
 668         if (dev->flags & IFF_MULTICAST) {
 669                 if (idev->cnf.forwarding) {
 670                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 671                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 672                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 673                 } else {
 674                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 675                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 676                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 677                 }
 678         }
 679
 680         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 681                 if (ifa->flags&IFA_F_TENTATIVE)
 682                         continue;
 683                 if (idev->cnf.forwarding)
 684                         addrconf_join_anycast(ifa);
 685                 else
 686                         addrconf_leave_anycast(ifa);
 687         }
 688         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 689                                      dev->ifindex, &idev->cnf);
 690 }
 691
 692
 693 static void addrconf_forward_change(struct net *net, __s32 newf)
 694 {
 695         struct net_device *dev;
 696         struct inet6_dev *idev;
 697
 698         for_each_netdev(net, dev) {
 699                 idev = __in6_dev_get(dev);
 700                 if (idev) {
 701                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 702                         idev->cnf.forwarding = newf;
 703                         if (changed)
 704                                 dev_forward_change(idev);
 705                 }
 706         }
 707 }
 708
 709 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 710 {
 711         struct net *net;
 712         int old;
 713
 714         if (!rtnl_trylock())
 715                 return restart_syscall();
 716
 717         net = (struct net *)table->extra2;
 718         old = *p;
 719         *p = newf;
 720
 721         if (p == &net->ipv6.devconf_dflt->forwarding) {
 722                 if ((!newf) ^ (!old))
 723                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 724                                                      NETCONFA_IFINDEX_DEFAULT,
 725                                                      net->ipv6.devconf_dflt);
 726                 rtnl_unlock();
 727                 return 0;
 728         }
 729
 730         if (p == &net->ipv6.devconf_all->forwarding) {
 731                 net->ipv6.devconf_dflt->forwarding = newf;
 732                 addrconf_forward_change(net, newf);
 733                 if ((!newf) ^ (!old))
 734                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 735                                                      NETCONFA_IFINDEX_ALL,
 736                                                      net->ipv6.devconf_all);
 737         } else if ((!newf) ^ (!old))
 738                 dev_forward_change((struct inet6_dev *)table->extra1);
 739         rtnl_unlock();
 740
 741         if (newf)
 742                 rt6_purge_dflt_routers(net);
 743         return 1;
 744 }
 745 #endif
 746
 747 /* Nobody refers to this ifaddr, destroy it */
 748 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 749 {
 750         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 751
 752 #ifdef NET_REFCNT_DEBUG
 753         pr_debug("%s\n", __func__);
 754 #endif
 755
 756         in6_dev_put(ifp->idev);
 757
 758         if (del_timer(&ifp->dad_timer))
 759                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 760
 761         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 762                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 763                 return;
 764         }
 765         ip6_rt_put(ifp->rt);
 766
 767         kfree_rcu(ifp, rcu);
 768 }
 769
 770 static void
 771 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 772 {
 773         struct list_head *p;
 774         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 775
 776         /*
 777          * Each device address list is sorted in order of scope -
 778          * global before linklocal.
 779          */
 780         list_for_each(p, &idev->addr_list) {
 781                 struct inet6_ifaddr *ifa
 782                         = list_entry(p, struct inet6_ifaddr, if_list);
 783                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 784                         break;
 785         }
 786
 787         list_add_tail(&ifp->if_list, p);
 788 }
 789
 790 static u32 inet6_addr_hash(const struct in6_addr *addr)
 791 {
 792         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 793 }
 794
 795 /* On success it returns ifp with increased reference count */
 796
 797 static struct inet6_ifaddr *
 798 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 799               const struct in6_addr *peer_addr, int pfxlen,
 800               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 801 {
 802         struct inet6_ifaddr *ifa = NULL;
 803         struct rt6_info *rt;
 804         unsigned int hash;
 805         int err = 0;
 806         int addr_type = ipv6_addr_type(addr);
 807
 808         if (addr_type == IPV6_ADDR_ANY ||
 809             addr_type & IPV6_ADDR_MULTICAST ||
 810             (!(idev->dev->flags & IFF_LOOPBACK) &&
 811              addr_type & IPV6_ADDR_LOOPBACK))
 812                 return ERR_PTR(-EADDRNOTAVAIL);
 813
 814         rcu_read_lock_bh();
 815         if (idev->dead) {
 816                 err = -ENODEV;                  /*XXX*/
 817                 goto out2;
 818         }
 819
 820         if (idev->cnf.disable_ipv6) {
 821                 err = -EACCES;
 822                 goto out2;
 823         }
 824
 825         spin_lock(&addrconf_hash_lock);
 826
 827         /* Ignore adding duplicate addresses on an interface */
 828         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 829                 ADBG("ipv6_add_addr: already assigned\n");
 830                 err = -EEXIST;
 831                 goto out;
 832         }
 833
 834         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 835
 836         if (ifa == NULL) {
 837                 ADBG("ipv6_add_addr: malloc failed\n");
 838                 err = -ENOBUFS;
 839                 goto out;
 840         }
 841
 842         rt = addrconf_dst_alloc(idev, addr, false);
 843         if (IS_ERR(rt)) {
 844                 err = PTR_ERR(rt);
 845                 goto out;
 846         }
 847
 848         ifa->addr = *addr;
 849         if (peer_addr)
 850                 ifa->peer_addr = *peer_addr;
 851
 852         spin_lock_init(&ifa->lock);
 853         spin_lock_init(&ifa->state_lock);
 854         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 855                     (unsigned long)ifa);
 856         INIT_HLIST_NODE(&ifa->addr_lst);
 857         ifa->scope = scope;
 858         ifa->prefix_len = pfxlen;
 859         ifa->flags = flags | IFA_F_TENTATIVE;
 860         ifa->valid_lft = valid_lft;
 861         ifa->prefered_lft = prefered_lft;
 862         ifa->cstamp = ifa->tstamp = jiffies;
 863         ifa->tokenized = false;
 864
 865         ifa->rt = rt;
 866
 867         ifa->idev = idev;
 868         in6_dev_hold(idev);
 869         /* For caller */
 870         in6_ifa_hold(ifa);
 871
 872         /* Add to big hash table */
 873         hash = inet6_addr_hash(addr);
 874
 875         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 876         spin_unlock(&addrconf_hash_lock);
 877
 878         write_lock(&idev->lock);
 879         /* Add to inet6_dev unicast addr list. */
 880         ipv6_link_dev_addr(idev, ifa);
 881
 882 #ifdef CONFIG_IPV6_PRIVACY
 883         if (ifa->flags&IFA_F_TEMPORARY) {
 884                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 885                 in6_ifa_hold(ifa);
 886         }
 887 #endif
 888
 889         in6_ifa_hold(ifa);
 890         write_unlock(&idev->lock);
 891 out2:
 892         rcu_read_unlock_bh();
 893
 894         if (likely(err == 0))
 895                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 896         else {
 897                 kfree(ifa);
 898                 ifa = ERR_PTR(err);
 899         }
 900
 901         return ifa;
 902 out:
 903         spin_unlock(&addrconf_hash_lock);
 904         goto out2;
 905 }
 906
 907 /* This function wants to get referenced ifp and releases it before return */
 908
 909 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 910 {
 911         struct inet6_ifaddr *ifa, *ifn;
 912         struct inet6_dev *idev = ifp->idev;
 913         int state;
 914         int deleted = 0, onlink = 0;
 915         unsigned long expires = jiffies;
 916
 917         spin_lock_bh(&ifp->state_lock);
 918         state = ifp->state;
 919         ifp->state = INET6_IFADDR_STATE_DEAD;
 920         spin_unlock_bh(&ifp->state_lock);
 921
 922         if (state == INET6_IFADDR_STATE_DEAD)
 923                 goto out;
 924
 925         spin_lock_bh(&addrconf_hash_lock);
 926         hlist_del_init_rcu(&ifp->addr_lst);
 927         spin_unlock_bh(&addrconf_hash_lock);
 928
 929         write_lock_bh(&idev->lock);
 930 #ifdef CONFIG_IPV6_PRIVACY
 931         if (ifp->flags&IFA_F_TEMPORARY) {
 932                 list_del(&ifp->tmp_list);
 933                 if (ifp->ifpub) {
 934                         in6_ifa_put(ifp->ifpub);
 935                         ifp->ifpub = NULL;
 936                 }
 937                 __in6_ifa_put(ifp);
 938         }
 939 #endif
 940
 941         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 942                 if (ifa == ifp) {
 943                         list_del_init(&ifp->if_list);
 944                         __in6_ifa_put(ifp);
 945
 946                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 947                                 break;
 948                         deleted = 1;
 949                         continue;
 950                 } else if (ifp->flags & IFA_F_PERMANENT) {
 951                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 952                                               ifp->prefix_len)) {
 953                                 if (ifa->flags & IFA_F_PERMANENT) {
 954                                         onlink = 1;
 955                                         if (deleted)
 956                                                 break;
 957                                 } else {
 958                                         unsigned long lifetime;
 959
 960                                         if (!onlink)
 961                                                 onlink = -1;
 962
 963                                         spin_lock(&ifa->lock);
 964
 965                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 966                                         /*
 967                                          * Note: Because this address is
 968                                          * not permanent, lifetime <
 969                                          * LONG_MAX / HZ here.
 970                                          */
 971                                         if (time_before(expires,
 972                                                         ifa->tstamp + lifetime * HZ))
 973                                                 expires = ifa->tstamp + lifetime * HZ;
 974                                         spin_unlock(&ifa->lock);
 975                                 }
 976                         }
 977                 }
 978         }
 979         write_unlock_bh(&idev->lock);
 980
 981         addrconf_del_dad_timer(ifp);
 982
 983         ipv6_ifa_notify(RTM_DELADDR, ifp);
 984
 985         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
 986
 987         /*
 988          * Purge or update corresponding prefix
 989          *
 990          * 1) we don't purge prefix here if address was not permanent.
 991          *    prefix is managed by its own lifetime.
 992          * 2) if there're no addresses, delete prefix.
 993          * 3) if there're still other permanent address(es),
 994          *    corresponding prefix is still permanent.
 995          * 4) otherwise, update prefix lifetime to the
 996          *    longest valid lifetime among the corresponding
 997          *    addresses on the device.
 998          *    Note: subsequent RA will update lifetime.
 999          *
1000          * --yoshfuji
1001          */
1002         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1003                 struct in6_addr prefix;
1004                 struct rt6_info *rt;
1005
1006                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1007
1008                 rt = addrconf_get_prefix_route(&prefix,
1009                                                ifp->prefix_len,
1010                                                ifp->idev->dev,
1011                                                0, RTF_GATEWAY | RTF_DEFAULT);
1012
1013                 if (rt) {
1014                         if (onlink == 0) {
1015                                 ip6_del_rt(rt);
1016                                 rt = NULL;
1017                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1018                                 rt6_set_expires(rt, expires);
1019                         }
1020                 }
1021                 ip6_rt_put(rt);
1022         }
1023
1024         /* clean up prefsrc entries */
1025         rt6_remove_prefsrc(ifp);
1026 out:
1027         in6_ifa_put(ifp);
1028 }
1029
1030 #ifdef CONFIG_IPV6_PRIVACY
1031 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1032 {
1033         struct inet6_dev *idev = ifp->idev;
1034         struct in6_addr addr, *tmpaddr;
1035         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1036         unsigned long regen_advance;
1037         int tmp_plen;
1038         int ret = 0;
1039         u32 addr_flags;
1040         unsigned long now = jiffies;
1041
1042         write_lock(&idev->lock);
1043         if (ift) {
1044                 spin_lock_bh(&ift->lock);
1045                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1046                 spin_unlock_bh(&ift->lock);
1047                 tmpaddr = &addr;
1048         } else {
1049                 tmpaddr = NULL;
1050         }
1051 retry:
1052         in6_dev_hold(idev);
1053         if (idev->cnf.use_tempaddr <= 0) {
1054                 write_unlock(&idev->lock);
1055                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1056                 in6_dev_put(idev);
1057                 ret = -1;
1058                 goto out;
1059         }
1060         spin_lock_bh(&ifp->lock);
1061         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1062                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1063                 spin_unlock_bh(&ifp->lock);
1064                 write_unlock(&idev->lock);
1065                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1066                         __func__);
1067                 in6_dev_put(idev);
1068                 ret = -1;
1069                 goto out;
1070         }
1071         in6_ifa_hold(ifp);
1072         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1073         __ipv6_try_regen_rndid(idev, tmpaddr);
1074         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1075         age = (now - ifp->tstamp) / HZ;
1076         tmp_valid_lft = min_t(__u32,
1077                               ifp->valid_lft,
1078                               idev->cnf.temp_valid_lft + age);
1079         tmp_prefered_lft = min_t(__u32,
1080                                  ifp->prefered_lft,
1081                                  idev->cnf.temp_prefered_lft + age -
1082                                  idev->cnf.max_desync_factor);
1083         tmp_plen = ifp->prefix_len;
1084         tmp_tstamp = ifp->tstamp;
1085         spin_unlock_bh(&ifp->lock);
1086
1087         regen_advance = idev->cnf.regen_max_retry *
1088                         idev->cnf.dad_transmits *
1089                         idev->nd_parms->retrans_time / HZ;
1090         write_unlock(&idev->lock);
1091
1092         /* A temporary address is created only if this calculated Preferred
1093          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1094          * an implementation must not create a temporary address with a zero
1095          * Preferred Lifetime.
1096          */
1097         if (tmp_prefered_lft <= regen_advance) {
1098                 in6_ifa_put(ifp);
1099                 in6_dev_put(idev);
1100                 ret = -1;
1101                 goto out;
1102         }
1103
1104         addr_flags = IFA_F_TEMPORARY;
1105         /* set in addrconf_prefix_rcv() */
1106         if (ifp->flags & IFA_F_OPTIMISTIC)
1107                 addr_flags |= IFA_F_OPTIMISTIC;
1108
1109         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1110                             ipv6_addr_scope(&addr), addr_flags,
1111                             tmp_valid_lft, tmp_prefered_lft);
1112         if (IS_ERR(ift)) {
1113                 in6_ifa_put(ifp);
1114                 in6_dev_put(idev);
1115                 pr_info("%s: retry temporary address regeneration\n", __func__);
1116                 tmpaddr = &addr;
1117                 write_lock(&idev->lock);
1118                 goto retry;
1119         }
1120
1121         spin_lock_bh(&ift->lock);
1122         ift->ifpub = ifp;
1123         ift->cstamp = now;
1124         ift->tstamp = tmp_tstamp;
1125         spin_unlock_bh(&ift->lock);
1126
1127         addrconf_dad_start(ift);
1128         in6_ifa_put(ift);
1129         in6_dev_put(idev);
1130 out:
1131         return ret;
1132 }
1133 #endif
1134
1135 /*
1136  *      Choose an appropriate source address (RFC3484)
1137  */
1138 enum {
1139         IPV6_SADDR_RULE_INIT = 0,
1140         IPV6_SADDR_RULE_LOCAL,
1141         IPV6_SADDR_RULE_SCOPE,
1142         IPV6_SADDR_RULE_PREFERRED,
1143 #ifdef CONFIG_IPV6_MIP6
1144         IPV6_SADDR_RULE_HOA,
1145 #endif
1146         IPV6_SADDR_RULE_OIF,
1147         IPV6_SADDR_RULE_LABEL,
1148 #ifdef CONFIG_IPV6_PRIVACY
1149         IPV6_SADDR_RULE_PRIVACY,
1150 #endif
1151         IPV6_SADDR_RULE_ORCHID,
1152         IPV6_SADDR_RULE_PREFIX,
1153         IPV6_SADDR_RULE_MAX
1154 };
1155
1156 struct ipv6_saddr_score {
1157         int                     rule;
1158         int                     addr_type;
1159         struct inet6_ifaddr     *ifa;
1160         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1161         int                     scopedist;
1162         int                     matchlen;
1163 };
1164
1165 struct ipv6_saddr_dst {
1166         const struct in6_addr *addr;
1167         int ifindex;
1168         int scope;
1169         int label;
1170         unsigned int prefs;
1171 };
1172
1173 static inline int ipv6_saddr_preferred(int type)
1174 {
1175         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1176                 return 1;
1177         return 0;
1178 }
1179
1180 static int ipv6_get_saddr_eval(struct net *net,
1181                                struct ipv6_saddr_score *score,
1182                                struct ipv6_saddr_dst *dst,
1183                                int i)
1184 {
1185         int ret;
1186
1187         if (i <= score->rule) {
1188                 switch (i) {
1189                 case IPV6_SADDR_RULE_SCOPE:
1190                         ret = score->scopedist;
1191                         break;
1192                 case IPV6_SADDR_RULE_PREFIX:
1193                         ret = score->matchlen;
1194                         break;
1195                 default:
1196                         ret = !!test_bit(i, score->scorebits);
1197                 }
1198                 goto out;
1199         }
1200
1201         switch (i) {
1202         case IPV6_SADDR_RULE_INIT:
1203                 /* Rule 0: remember if hiscore is not ready yet */
1204                 ret = !!score->ifa;
1205                 break;
1206         case IPV6_SADDR_RULE_LOCAL:
1207                 /* Rule 1: Prefer same address */
1208                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1209                 break;
1210         case IPV6_SADDR_RULE_SCOPE:
1211                 /* Rule 2: Prefer appropriate scope
1212                  *
1213                  *      ret
1214                  *       ^
1215                  *    -1 |  d 15
1216                  *    ---+--+-+---> scope
1217                  *       |
1218                  *       |             d is scope of the destination.
1219                  *  B-d  |  \
1220                  *       |   \      <- smaller scope is better if
1221                  *  B-15 |    \        if scope is enough for destinaion.
1222                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1223                  * d-C-1 | /
1224                  *       |/         <- greater is better
1225                  *   -C  /             if scope is not enough for destination.
1226                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1227                  *
1228                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1229                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1230                  * Assume B = 0 and we get C > 29.
1231                  */
1232                 ret = __ipv6_addr_src_scope(score->addr_type);
1233                 if (ret >= dst->scope)
1234                         ret = -ret;
1235                 else
1236                         ret -= 128;     /* 30 is enough */
1237                 score->scopedist = ret;
1238                 break;
1239         case IPV6_SADDR_RULE_PREFERRED:
1240                 /* Rule 3: Avoid deprecated and optimistic addresses */
1241                 ret = ipv6_saddr_preferred(score->addr_type) ||
1242                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1243                 break;
1244 #ifdef CONFIG_IPV6_MIP6
1245         case IPV6_SADDR_RULE_HOA:
1246             {
1247                 /* Rule 4: Prefer home address */
1248                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1249                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1250                 break;
1251             }
1252 #endif
1253         case IPV6_SADDR_RULE_OIF:
1254                 /* Rule 5: Prefer outgoing interface */
1255                 ret = (!dst->ifindex ||
1256                        dst->ifindex == score->ifa->idev->dev->ifindex);
1257                 break;
1258         case IPV6_SADDR_RULE_LABEL:
1259                 /* Rule 6: Prefer matching label */
1260                 ret = ipv6_addr_label(net,
1261                                       &score->ifa->addr, score->addr_type,
1262                                       score->ifa->idev->dev->ifindex) == dst->label;
1263                 break;
1264 #ifdef CONFIG_IPV6_PRIVACY
1265         case IPV6_SADDR_RULE_PRIVACY:
1266             {
1267                 /* Rule 7: Prefer public address
1268                  * Note: prefer temporary address if use_tempaddr >= 2
1269                  */
1270                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1271                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1272                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1273                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1274                 break;
1275             }
1276 #endif
1277         case IPV6_SADDR_RULE_ORCHID:
1278                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1279                  *          non-ORCHID vs non-ORCHID
1280                  */
1281                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1282                         ipv6_addr_orchid(dst->addr));
1283                 break;
1284         case IPV6_SADDR_RULE_PREFIX:
1285                 /* Rule 8: Use longest matching prefix */
1286                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1287                 if (ret > score->ifa->prefix_len)
1288                         ret = score->ifa->prefix_len;
1289                 score->matchlen = ret;
1290                 break;
1291         default:
1292                 ret = 0;
1293         }
1294
1295         if (ret)
1296                 __set_bit(i, score->scorebits);
1297         score->rule = i;
1298 out:
1299         return ret;
1300 }
1301
1302 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1303                        const struct in6_addr *daddr, unsigned int prefs,
1304                        struct in6_addr *saddr)
1305 {
1306         struct ipv6_saddr_score scores[2],
1307                                 *score = &scores[0], *hiscore = &scores[1];
1308         struct ipv6_saddr_dst dst;
1309         struct net_device *dev;
1310         int dst_type;
1311
1312         dst_type = __ipv6_addr_type(daddr);
1313         dst.addr = daddr;
1314         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1315         dst.scope = __ipv6_addr_src_scope(dst_type);
1316         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1317         dst.prefs = prefs;
1318
1319         hiscore->rule = -1;
1320         hiscore->ifa = NULL;
1321
1322         rcu_read_lock();
1323
1324         for_each_netdev_rcu(net, dev) {
1325                 struct inet6_dev *idev;
1326
1327                 /* Candidate Source Address (section 4)
1328                  *  - multicast and link-local destination address,
1329                  *    the set of candidate source address MUST only
1330                  *    include addresses assigned to interfaces
1331                  *    belonging to the same link as the outgoing
1332                  *    interface.
1333                  * (- For site-local destination addresses, the
1334                  *    set of candidate source addresses MUST only
1335                  *    include addresses assigned to interfaces
1336                  *    belonging to the same site as the outgoing
1337                  *    interface.)
1338                  */
1339                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1340                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1341                     dst.ifindex && dev->ifindex != dst.ifindex)
1342                         continue;
1343
1344                 idev = __in6_dev_get(dev);
1345                 if (!idev)
1346                         continue;
1347
1348                 read_lock_bh(&idev->lock);
1349                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1350                         int i;
1351
1352                         /*
1353                          * - Tentative Address (RFC2462 section 5.4)
1354                          *  - A tentative address is not considered
1355                          *    "assigned to an interface" in the traditional
1356                          *    sense, unless it is also flagged as optimistic.
1357                          * - Candidate Source Address (section 4)
1358                          *  - In any case, anycast addresses, multicast
1359                          *    addresses, and the unspecified address MUST
1360                          *    NOT be included in a candidate set.
1361                          */
1362                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1363                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1364                                 continue;
1365
1366                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1367
1368                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1369                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1370                                 LIMIT_NETDEBUG(KERN_DEBUG
1371                                                "ADDRCONF: unspecified / multicast address "
1372                                                "assigned as unicast address on %s",
1373                                                dev->name);
1374                                 continue;
1375                         }
1376
1377                         score->rule = -1;
1378                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1379
1380                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1381                                 int minihiscore, miniscore;
1382
1383                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1384                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1385
1386                                 if (minihiscore > miniscore) {
1387                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1388                                             score->scopedist > 0) {
1389                                                 /*
1390                                                  * special case:
1391                                                  * each remaining entry
1392                                                  * has too small (not enough)
1393                                                  * scope, because ifa entries
1394                                                  * are sorted by their scope
1395                                                  * values.
1396                                                  */
1397                                                 goto try_nextdev;
1398                                         }
1399                                         break;
1400                                 } else if (minihiscore < miniscore) {
1401                                         if (hiscore->ifa)
1402                                                 in6_ifa_put(hiscore->ifa);
1403
1404                                         in6_ifa_hold(score->ifa);
1405
1406                                         swap(hiscore, score);
1407
1408                                         /* restore our iterator */
1409                                         score->ifa = hiscore->ifa;
1410
1411                                         break;
1412                                 }
1413                         }
1414                 }
1415 try_nextdev:
1416                 read_unlock_bh(&idev->lock);
1417         }
1418         rcu_read_unlock();
1419
1420         if (!hiscore->ifa)
1421                 return -EADDRNOTAVAIL;
1422
1423         *saddr = hiscore->ifa->addr;
1424         in6_ifa_put(hiscore->ifa);
1425         return 0;
1426 }
1427 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1428
1429 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1430                       unsigned char banned_flags)
1431 {
1432         struct inet6_ifaddr *ifp;
1433         int err = -EADDRNOTAVAIL;
1434
1435         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1436                 if (ifp->scope == IFA_LINK &&
1437                     !(ifp->flags & banned_flags)) {
1438                         *addr = ifp->addr;
1439                         err = 0;
1440                         break;
1441                 }
1442         }
1443         return err;
1444 }
1445
1446 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1447                     unsigned char banned_flags)
1448 {
1449         struct inet6_dev *idev;
1450         int err = -EADDRNOTAVAIL;
1451
1452         rcu_read_lock();
1453         idev = __in6_dev_get(dev);
1454         if (idev) {
1455                 read_lock_bh(&idev->lock);
1456                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1457                 read_unlock_bh(&idev->lock);
1458         }
1459         rcu_read_unlock();
1460         return err;
1461 }
1462
1463 static int ipv6_count_addresses(struct inet6_dev *idev)
1464 {
1465         int cnt = 0;
1466         struct inet6_ifaddr *ifp;
1467
1468         read_lock_bh(&idev->lock);
1469         list_for_each_entry(ifp, &idev->addr_list, if_list)
1470                 cnt++;
1471         read_unlock_bh(&idev->lock);
1472         return cnt;
1473 }
1474
1475 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1476                   const struct net_device *dev, int strict)
1477 {
1478         struct inet6_ifaddr *ifp;
1479         unsigned int hash = inet6_addr_hash(addr);
1480
1481         rcu_read_lock_bh();
1482         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1483                 if (!net_eq(dev_net(ifp->idev->dev), net))
1484                         continue;
1485                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1486                     !(ifp->flags&IFA_F_TENTATIVE) &&
1487                     (dev == NULL || ifp->idev->dev == dev ||
1488                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1489                         rcu_read_unlock_bh();
1490                         return 1;
1491                 }
1492         }
1493
1494         rcu_read_unlock_bh();
1495         return 0;
1496 }
1497 EXPORT_SYMBOL(ipv6_chk_addr);
1498
1499 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1500                                struct net_device *dev)
1501 {
1502         unsigned int hash = inet6_addr_hash(addr);
1503         struct inet6_ifaddr *ifp;
1504
1505         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1506                 if (!net_eq(dev_net(ifp->idev->dev), net))
1507                         continue;
1508                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1509                         if (dev == NULL || ifp->idev->dev == dev)
1510                                 return true;
1511                 }
1512         }
1513         return false;
1514 }
1515
1516 /* Compares an address/prefix_len with addresses on device @dev.
1517  * If one is found it returns true.
1518  */
1519 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1520         const unsigned int prefix_len, struct net_device *dev)
1521 {
1522         struct inet6_dev *idev;
1523         struct inet6_ifaddr *ifa;
1524         bool ret = false;
1525
1526         rcu_read_lock();
1527         idev = __in6_dev_get(dev);
1528         if (idev) {
1529                 read_lock_bh(&idev->lock);
1530                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1531                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1532                         if (ret)
1533                                 break;
1534                 }
1535                 read_unlock_bh(&idev->lock);
1536         }
1537         rcu_read_unlock();
1538
1539         return ret;
1540 }
1541 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1542
1543 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1544 {
1545         struct inet6_dev *idev;
1546         struct inet6_ifaddr *ifa;
1547         int     onlink;
1548
1549         onlink = 0;
1550         rcu_read_lock();
1551         idev = __in6_dev_get(dev);
1552         if (idev) {
1553                 read_lock_bh(&idev->lock);
1554                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1555                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1556                                                    ifa->prefix_len);
1557                         if (onlink)
1558                                 break;
1559                 }
1560                 read_unlock_bh(&idev->lock);
1561         }
1562         rcu_read_unlock();
1563         return onlink;
1564 }
1565 EXPORT_SYMBOL(ipv6_chk_prefix);
1566
1567 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1568                                      struct net_device *dev, int strict)
1569 {
1570         struct inet6_ifaddr *ifp, *result = NULL;
1571         unsigned int hash = inet6_addr_hash(addr);
1572
1573         rcu_read_lock_bh();
1574         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1575                 if (!net_eq(dev_net(ifp->idev->dev), net))
1576                         continue;
1577                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1578                         if (dev == NULL || ifp->idev->dev == dev ||
1579                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1580                                 result = ifp;
1581                                 in6_ifa_hold(ifp);
1582                                 break;
1583                         }
1584                 }
1585         }
1586         rcu_read_unlock_bh();
1587
1588         return result;
1589 }
1590
1591 /* Gets referenced address, destroys ifaddr */
1592
1593 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1594 {
1595         if (ifp->flags&IFA_F_PERMANENT) {
1596                 spin_lock_bh(&ifp->lock);
1597                 addrconf_del_dad_timer(ifp);
1598                 ifp->flags |= IFA_F_TENTATIVE;
1599                 if (dad_failed)
1600                         ifp->flags |= IFA_F_DADFAILED;
1601                 spin_unlock_bh(&ifp->lock);
1602                 if (dad_failed)
1603                         ipv6_ifa_notify(0, ifp);
1604                 in6_ifa_put(ifp);
1605 #ifdef CONFIG_IPV6_PRIVACY
1606         } else if (ifp->flags&IFA_F_TEMPORARY) {
1607                 struct inet6_ifaddr *ifpub;
1608                 spin_lock_bh(&ifp->lock);
1609                 ifpub = ifp->ifpub;
1610                 if (ifpub) {
1611                         in6_ifa_hold(ifpub);
1612                         spin_unlock_bh(&ifp->lock);
1613                         ipv6_create_tempaddr(ifpub, ifp);
1614                         in6_ifa_put(ifpub);
1615                 } else {
1616                         spin_unlock_bh(&ifp->lock);
1617                 }
1618                 ipv6_del_addr(ifp);
1619 #endif
1620         } else
1621                 ipv6_del_addr(ifp);
1622 }
1623
1624 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1625 {
1626         int err = -ENOENT;
1627
1628         spin_lock(&ifp->state_lock);
1629         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1630                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1631                 err = 0;
1632         }
1633         spin_unlock(&ifp->state_lock);
1634
1635         return err;
1636 }
1637
1638 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1639 {
1640         struct inet6_dev *idev = ifp->idev;
1641
1642         if (addrconf_dad_end(ifp)) {
1643                 in6_ifa_put(ifp);
1644                 return;
1645         }
1646
1647         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1648                              ifp->idev->dev->name, &ifp->addr);
1649
1650         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1651                 struct in6_addr addr;
1652
1653                 addr.s6_addr32[0] = htonl(0xfe800000);
1654                 addr.s6_addr32[1] = 0;
1655
1656                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1657                     ipv6_addr_equal(&ifp->addr, &addr)) {
1658                         /* DAD failed for link-local based on MAC address */
1659                         idev->cnf.disable_ipv6 = 1;
1660
1661                         pr_info("%s: IPv6 being disabled!\n",
1662                                 ifp->idev->dev->name);
1663                 }
1664         }
1665
1666         addrconf_dad_stop(ifp, 1);
1667 }
1668
1669 /* Join to solicited addr multicast group. */
1670
1671 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1672 {
1673         struct in6_addr maddr;
1674
1675         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1676                 return;
1677
1678         addrconf_addr_solict_mult(addr, &maddr);
1679         ipv6_dev_mc_inc(dev, &maddr);
1680 }
1681
1682 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1683 {
1684         struct in6_addr maddr;
1685
1686         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1687                 return;
1688
1689         addrconf_addr_solict_mult(addr, &maddr);
1690         __ipv6_dev_mc_dec(idev, &maddr);
1691 }
1692
1693 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1694 {
1695         struct in6_addr addr;
1696         if (ifp->prefix_len == 127) /* RFC 6164 */
1697                 return;
1698         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1699         if (ipv6_addr_any(&addr))
1700                 return;
1701         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1702 }
1703
1704 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1705 {
1706         struct in6_addr addr;
1707         if (ifp->prefix_len == 127) /* RFC 6164 */
1708                 return;
1709         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1710         if (ipv6_addr_any(&addr))
1711                 return;
1712         __ipv6_dev_ac_dec(ifp->idev, &addr);
1713 }
1714
1715 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1716 {
1717         if (dev->addr_len != ETH_ALEN)
1718                 return -1;
1719         memcpy(eui, dev->dev_addr, 3);
1720         memcpy(eui + 5, dev->dev_addr + 3, 3);
1721
1722         /*
1723          * The zSeries OSA network cards can be shared among various
1724          * OS instances, but the OSA cards have only one MAC address.
1725          * This leads to duplicate address conflicts in conjunction
1726          * with IPv6 if more than one instance uses the same card.
1727          *
1728          * The driver for these cards can deliver a unique 16-bit
1729          * identifier for each instance sharing the same card.  It is
1730          * placed instead of 0xFFFE in the interface identifier.  The
1731          * "u" bit of the interface identifier is not inverted in this
1732          * case.  Hence the resulting interface identifier has local
1733          * scope according to RFC2373.
1734          */
1735         if (dev->dev_id) {
1736                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1737                 eui[4] = dev->dev_id & 0xFF;
1738         } else {
1739                 eui[3] = 0xFF;
1740                 eui[4] = 0xFE;
1741                 eui[0] ^= 2;
1742         }
1743         return 0;
1744 }
1745
1746 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1747 {
1748         if (dev->addr_len != IEEE802154_ADDR_LEN)
1749                 return -1;
1750         memcpy(eui, dev->dev_addr, 8);
1751         eui[0] ^= 2;
1752         return 0;
1753 }
1754
1755 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1756 {
1757         union fwnet_hwaddr *ha;
1758
1759         if (dev->addr_len != FWNET_ALEN)
1760                 return -1;
1761
1762         ha = (union fwnet_hwaddr *)dev->dev_addr;
1763
1764         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1765         eui[0] ^= 2;
1766         return 0;
1767 }
1768
1769 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1770 {
1771         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1772         if (dev->addr_len != ARCNET_ALEN)
1773                 return -1;
1774         memset(eui, 0, 7);
1775         eui[7] = *(u8 *)dev->dev_addr;
1776         return 0;
1777 }
1778
1779 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1780 {
1781         if (dev->addr_len != INFINIBAND_ALEN)
1782                 return -1;
1783         memcpy(eui, dev->dev_addr + 12, 8);
1784         eui[0] |= 2;
1785         return 0;
1786 }
1787
1788 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1789 {
1790         if (addr == 0)
1791                 return -1;
1792         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1793                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1794                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1795                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1796                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1797                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1798         eui[1] = 0;
1799         eui[2] = 0x5E;
1800         eui[3] = 0xFE;
1801         memcpy(eui + 4, &addr, 4);
1802         return 0;
1803 }
1804
1805 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1806 {
1807         if (dev->priv_flags & IFF_ISATAP)
1808                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1809         return -1;
1810 }
1811
1812 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1813 {
1814         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1815 }
1816
1817 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1818 {
1819         memcpy(eui, dev->perm_addr, 3);
1820         memcpy(eui + 5, dev->perm_addr + 3, 3);
1821         eui[3] = 0xFF;
1822         eui[4] = 0xFE;
1823         eui[0] ^= 2;
1824         return 0;
1825 }
1826
1827 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1828 {
1829         switch (dev->type) {
1830         case ARPHRD_ETHER:
1831         case ARPHRD_FDDI:
1832                 return addrconf_ifid_eui48(eui, dev);
1833         case ARPHRD_ARCNET:
1834                 return addrconf_ifid_arcnet(eui, dev);
1835         case ARPHRD_INFINIBAND:
1836                 return addrconf_ifid_infiniband(eui, dev);
1837         case ARPHRD_SIT:
1838                 return addrconf_ifid_sit(eui, dev);
1839         case ARPHRD_IPGRE:
1840                 return addrconf_ifid_gre(eui, dev);
1841         case ARPHRD_IEEE802154:
1842                 return addrconf_ifid_eui64(eui, dev);
1843         case ARPHRD_IEEE1394:
1844                 return addrconf_ifid_ieee1394(eui, dev);
1845         case ARPHRD_TUNNEL6:
1846                 return addrconf_ifid_ip6tnl(eui, dev);
1847         }
1848         return -1;
1849 }
1850
1851 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1852 {
1853         int err = -1;
1854         struct inet6_ifaddr *ifp;
1855
1856         read_lock_bh(&idev->lock);
1857         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1858                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1859                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1860                         err = 0;
1861                         break;
1862                 }
1863         }
1864         read_unlock_bh(&idev->lock);
1865         return err;
1866 }
1867
1868 #ifdef CONFIG_IPV6_PRIVACY
1869 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1870 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1871 {
1872 regen:
1873         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1874         idev->rndid[0] &= ~0x02;
1875
1876         /*
1877          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1878          * check if generated address is not inappropriate
1879          *
1880          *  - Reserved subnet anycast (RFC 2526)
1881          *      11111101 11....11 1xxxxxxx
1882          *  - ISATAP (RFC4214) 6.1
1883          *      00-00-5E-FE-xx-xx-xx-xx
1884          *  - value 0
1885          *  - XXX: already assigned to an address on the device
1886          */
1887         if (idev->rndid[0] == 0xfd &&
1888             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1889             (idev->rndid[7]&0x80))
1890                 goto regen;
1891         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1892                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1893                         goto regen;
1894                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1895                         goto regen;
1896         }
1897 }
1898
1899 static void ipv6_regen_rndid(unsigned long data)
1900 {
1901         struct inet6_dev *idev = (struct inet6_dev *) data;
1902         unsigned long expires;
1903
1904         rcu_read_lock_bh();
1905         write_lock_bh(&idev->lock);
1906
1907         if (idev->dead)
1908                 goto out;
1909
1910         __ipv6_regen_rndid(idev);
1911
1912         expires = jiffies +
1913                 idev->cnf.temp_prefered_lft * HZ -
1914                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1915                 idev->cnf.max_desync_factor * HZ;
1916         if (time_before(expires, jiffies)) {
1917                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1918                         __func__, idev->dev->name);
1919                 goto out;
1920         }
1921
1922         if (!mod_timer(&idev->regen_timer, expires))
1923                 in6_dev_hold(idev);
1924
1925 out:
1926         write_unlock_bh(&idev->lock);
1927         rcu_read_unlock_bh();
1928         in6_dev_put(idev);
1929 }
1930
1931 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1932 {
1933         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1934                 __ipv6_regen_rndid(idev);
1935 }
1936 #endif
1937
1938 /*
1939  *      Add prefix route.
1940  */
1941
1942 static void
1943 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1944                       unsigned long expires, u32 flags)
1945 {
1946         struct fib6_config cfg = {
1947                 .fc_table = RT6_TABLE_PREFIX,
1948                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1949                 .fc_ifindex = dev->ifindex,
1950                 .fc_expires = expires,
1951                 .fc_dst_len = plen,
1952                 .fc_flags = RTF_UP | flags,
1953                 .fc_nlinfo.nl_net = dev_net(dev),
1954                 .fc_protocol = RTPROT_KERNEL,
1955         };
1956
1957         cfg.fc_dst = *pfx;
1958
1959         /* Prevent useless cloning on PtP SIT.
1960            This thing is done here expecting that the whole
1961            class of non-broadcast devices need not cloning.
1962          */
1963 #if IS_ENABLED(CONFIG_IPV6_SIT)
1964         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1965                 cfg.fc_flags |= RTF_NONEXTHOP;
1966 #endif
1967
1968         ip6_route_add(&cfg);
1969 }
1970
1971
1972 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1973                                                   int plen,
1974                                                   const struct net_device *dev,
1975                                                   u32 flags, u32 noflags)
1976 {
1977         struct fib6_node *fn;
1978         struct rt6_info *rt = NULL;
1979         struct fib6_table *table;
1980
1981         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1982         if (table == NULL)
1983                 return NULL;
1984
1985         read_lock_bh(&table->tb6_lock);
1986         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1987         if (!fn)
1988                 goto out;
1989         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1990                 if (rt->dst.dev->ifindex != dev->ifindex)
1991                         continue;
1992                 if ((rt->rt6i_flags & flags) != flags)
1993                         continue;
1994                 if ((rt->rt6i_flags & noflags) != 0)
1995                         continue;
1996                 dst_hold(&rt->dst);
1997                 break;
1998         }
1999 out:
2000         read_unlock_bh(&table->tb6_lock);
2001         return rt;
2002 }
2003
2004
2005 /* Create "default" multicast route to the interface */
2006
2007 static void addrconf_add_mroute(struct net_device *dev)
2008 {
2009         struct fib6_config cfg = {
2010                 .fc_table = RT6_TABLE_LOCAL,
2011                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2012                 .fc_ifindex = dev->ifindex,
2013                 .fc_dst_len = 8,
2014                 .fc_flags = RTF_UP,
2015                 .fc_nlinfo.nl_net = dev_net(dev),
2016         };
2017
2018         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2019
2020         ip6_route_add(&cfg);
2021 }
2022
2023 #if IS_ENABLED(CONFIG_IPV6_SIT)
2024 static void sit_route_add(struct net_device *dev)
2025 {
2026         struct fib6_config cfg = {
2027                 .fc_table = RT6_TABLE_MAIN,
2028                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2029                 .fc_ifindex = dev->ifindex,
2030                 .fc_dst_len = 96,
2031                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2032                 .fc_nlinfo.nl_net = dev_net(dev),
2033         };
2034
2035         /* prefix length - 96 bits "::d.d.d.d" */
2036         ip6_route_add(&cfg);
2037 }
2038 #endif
2039
2040 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2041 {
2042         struct inet6_dev *idev;
2043
2044         ASSERT_RTNL();
2045
2046         idev = ipv6_find_idev(dev);
2047         if (!idev)
2048                 return ERR_PTR(-ENOBUFS);
2049
2050         if (idev->cnf.disable_ipv6)
2051                 return ERR_PTR(-EACCES);
2052
2053         /* Add default multicast route */
2054         if (!(dev->flags & IFF_LOOPBACK))
2055                 addrconf_add_mroute(dev);
2056
2057         return idev;
2058 }
2059
2060 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2061 {
2062         struct prefix_info *pinfo;
2063         __u32 valid_lft;
2064         __u32 prefered_lft;
2065         int addr_type;
2066         struct inet6_dev *in6_dev;
2067         struct net *net = dev_net(dev);
2068
2069         pinfo = (struct prefix_info *) opt;
2070
2071         if (len < sizeof(struct prefix_info)) {
2072                 ADBG("addrconf: prefix option too short\n");
2073                 return;
2074         }
2075
2076         /*
2077          *      Validation checks ([ADDRCONF], page 19)
2078          */
2079
2080         addr_type = ipv6_addr_type(&pinfo->prefix);
2081
2082         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2083                 return;
2084
2085         valid_lft = ntohl(pinfo->valid);
2086         prefered_lft = ntohl(pinfo->prefered);
2087
2088         if (prefered_lft > valid_lft) {
2089                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2090                 return;
2091         }
2092
2093         in6_dev = in6_dev_get(dev);
2094
2095         if (in6_dev == NULL) {
2096                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2097                                     dev->name);
2098                 return;
2099         }
2100
2101         /*
2102          *      Two things going on here:
2103          *      1) Add routes for on-link prefixes
2104          *      2) Configure prefixes with the auto flag set
2105          */
2106
2107         if (pinfo->onlink) {
2108                 struct rt6_info *rt;
2109                 unsigned long rt_expires;
2110
2111                 /* Avoid arithmetic overflow. Really, we could
2112                  * save rt_expires in seconds, likely valid_lft,
2113                  * but it would require division in fib gc, that it
2114                  * not good.
2115                  */
2116                 if (HZ > USER_HZ)
2117                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2118                 else
2119                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2120
2121                 if (addrconf_finite_timeout(rt_expires))
2122                         rt_expires *= HZ;
2123
2124                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2125                                                pinfo->prefix_len,
2126                                                dev,
2127                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2128                                                RTF_GATEWAY | RTF_DEFAULT);
2129
2130                 if (rt) {
2131                         /* Autoconf prefix route */
2132                         if (valid_lft == 0) {
2133                                 ip6_del_rt(rt);
2134                                 rt = NULL;
2135                         } else if (addrconf_finite_timeout(rt_expires)) {
2136                                 /* not infinity */
2137                                 rt6_set_expires(rt, jiffies + rt_expires);
2138                         } else {
2139                                 rt6_clean_expires(rt);
2140                         }
2141                 } else if (valid_lft) {
2142                         clock_t expires = 0;
2143                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2144                         if (addrconf_finite_timeout(rt_expires)) {
2145                                 /* not infinity */
2146                                 flags |= RTF_EXPIRES;
2147                                 expires = jiffies_to_clock_t(rt_expires);
2148                         }
2149                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2150                                               dev, expires, flags);
2151                 }
2152                 ip6_rt_put(rt);
2153         }
2154
2155         /* Try to figure out our local address for this prefix */
2156
2157         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2158                 struct inet6_ifaddr *ifp;
2159                 struct in6_addr addr;
2160                 int create = 0, update_lft = 0;
2161                 bool tokenized = false;
2162
2163                 if (pinfo->prefix_len == 64) {
2164                         memcpy(&addr, &pinfo->prefix, 8);
2165
2166                         if (!ipv6_addr_any(&in6_dev->token)) {
2167                                 read_lock_bh(&in6_dev->lock);
2168                                 memcpy(addr.s6_addr + 8,
2169                                        in6_dev->token.s6_addr + 8, 8);
2170                                 read_unlock_bh(&in6_dev->lock);
2171                                 tokenized = true;
2172                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2173                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2174                                 in6_dev_put(in6_dev);
2175                                 return;
2176                         }
2177                         goto ok;
2178                 }
2179                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2180                                     pinfo->prefix_len);
2181                 in6_dev_put(in6_dev);
2182                 return;
2183
2184 ok:
2185
2186                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2187
2188                 if (ifp == NULL && valid_lft) {
2189                         int max_addresses = in6_dev->cnf.max_addresses;
2190                         u32 addr_flags = 0;
2191
2192 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2193                         if (in6_dev->cnf.optimistic_dad &&
2194                             !net->ipv6.devconf_all->forwarding && sllao)
2195                                 addr_flags = IFA_F_OPTIMISTIC;
2196 #endif
2197
2198                         /* Do not allow to create too much of autoconfigured
2199                          * addresses; this would be too easy way to crash kernel.
2200                          */
2201                         if (!max_addresses ||
2202                             ipv6_count_addresses(in6_dev) < max_addresses)
2203                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2204                                                     pinfo->prefix_len,
2205                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2206                                                     addr_flags, valid_lft,
2207                                                     prefered_lft);
2208
2209                         if (IS_ERR_OR_NULL(ifp)) {
2210                                 in6_dev_put(in6_dev);
2211                                 return;
2212                         }
2213
2214                         update_lft = 0;
2215                         create = 1;
2216                         ifp->cstamp = jiffies;
2217                         ifp->tokenized = tokenized;
2218                         addrconf_dad_start(ifp);
2219                 }
2220
2221                 if (ifp) {
2222                         int flags;
2223                         unsigned long now;
2224 #ifdef CONFIG_IPV6_PRIVACY
2225                         struct inet6_ifaddr *ift;
2226 #endif
2227                         u32 stored_lft;
2228
2229                         /* update lifetime (RFC2462 5.5.3 e) */
2230                         spin_lock(&ifp->lock);
2231                         now = jiffies;
2232                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2233                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2234                         else
2235                                 stored_lft = 0;
2236                         if (!update_lft && !create && stored_lft) {
2237                                 const u32 minimum_lft = min(
2238                                         stored_lft, (u32)MIN_VALID_LIFETIME);
2239                                 valid_lft = max(valid_lft, minimum_lft);
2240
2241                                 /* RFC4862 Section 5.5.3e:
2242                                  * "Note that the preferred lifetime of the
2243                                  *  corresponding address is always reset to
2244                                  *  the Preferred Lifetime in the received
2245                                  *  Prefix Information option, regardless of
2246                                  *  whether the valid lifetime is also reset or
2247                                  *  ignored."
2248                                  *
2249                                  * So we should always update prefered_lft here.
2250                                  */
2251                                 update_lft = 1;
2252                         }
2253
2254                         if (update_lft) {
2255                                 ifp->valid_lft = valid_lft;
2256                                 ifp->prefered_lft = prefered_lft;
2257                                 ifp->tstamp = now;
2258                                 flags = ifp->flags;
2259                                 ifp->flags &= ~IFA_F_DEPRECATED;
2260                                 spin_unlock(&ifp->lock);
2261
2262                                 if (!(flags&IFA_F_TENTATIVE))
2263                                         ipv6_ifa_notify(0, ifp);
2264                         } else
2265                                 spin_unlock(&ifp->lock);
2266
2267 #ifdef CONFIG_IPV6_PRIVACY
2268                         read_lock_bh(&in6_dev->lock);
2269                         /* update all temporary addresses in the list */
2270                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2271                                             tmp_list) {
2272                                 int age, max_valid, max_prefered;
2273
2274                                 if (ifp != ift->ifpub)
2275                                         continue;
2276
2277                                 /*
2278                                  * RFC 4941 section 3.3:
2279                                  * If a received option will extend the lifetime
2280                                  * of a public address, the lifetimes of
2281                                  * temporary addresses should be extended,
2282                                  * subject to the overall constraint that no
2283                                  * temporary addresses should ever remain
2284                                  * "valid" or "preferred" for a time longer than
2285                                  * (TEMP_VALID_LIFETIME) or
2286                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2287                                  * respectively.
2288                                  */
2289                                 age = (now - ift->cstamp) / HZ;
2290                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2291                                 if (max_valid < 0)
2292                                         max_valid = 0;
2293
2294                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2295                                                in6_dev->cnf.max_desync_factor -
2296                                                age;
2297                                 if (max_prefered < 0)
2298                                         max_prefered = 0;
2299
2300                                 if (valid_lft > max_valid)
2301                                         valid_lft = max_valid;
2302
2303                                 if (prefered_lft > max_prefered)
2304                                         prefered_lft = max_prefered;
2305
2306                                 spin_lock(&ift->lock);
2307                                 flags = ift->flags;
2308                                 ift->valid_lft = valid_lft;
2309                                 ift->prefered_lft = prefered_lft;
2310                                 ift->tstamp = now;
2311                                 if (prefered_lft > 0)
2312                                         ift->flags &= ~IFA_F_DEPRECATED;
2313
2314                                 spin_unlock(&ift->lock);
2315                                 if (!(flags&IFA_F_TENTATIVE))
2316                                         ipv6_ifa_notify(0, ift);
2317                         }
2318
2319                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2320                                 /*
2321                                  * When a new public address is created as
2322                                  * described in [ADDRCONF], also create a new
2323                                  * temporary address. Also create a temporary
2324                                  * address if it's enabled but no temporary
2325                                  * address currently exists.
2326                                  */
2327                                 read_unlock_bh(&in6_dev->lock);
2328                                 ipv6_create_tempaddr(ifp, NULL);
2329                         } else {
2330                                 read_unlock_bh(&in6_dev->lock);
2331                         }
2332 #endif
2333                         in6_ifa_put(ifp);
2334                         addrconf_verify(0);
2335                 }
2336         }
2337         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2338         in6_dev_put(in6_dev);
2339 }
2340
2341 /*
2342  *      Set destination address.
2343  *      Special case for SIT interfaces where we create a new "virtual"
2344  *      device.
2345  */
2346 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2347 {
2348         struct in6_ifreq ireq;
2349         struct net_device *dev;
2350         int err = -EINVAL;
2351
2352         rtnl_lock();
2353
2354         err = -EFAULT;
2355         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2356                 goto err_exit;
2357
2358         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2359
2360         err = -ENODEV;
2361         if (dev == NULL)
2362                 goto err_exit;
2363
2364 #if IS_ENABLED(CONFIG_IPV6_SIT)
2365         if (dev->type == ARPHRD_SIT) {
2366                 const struct net_device_ops *ops = dev->netdev_ops;
2367                 struct ifreq ifr;
2368                 struct ip_tunnel_parm p;
2369
2370                 err = -EADDRNOTAVAIL;
2371                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2372                         goto err_exit;
2373
2374                 memset(&p, 0, sizeof(p));
2375                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2376                 p.iph.saddr = 0;
2377                 p.iph.version = 4;
2378                 p.iph.ihl = 5;
2379                 p.iph.protocol = IPPROTO_IPV6;
2380                 p.iph.ttl = 64;
2381                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2382
2383                 if (ops->ndo_do_ioctl) {
2384                         mm_segment_t oldfs = get_fs();
2385
2386                         set_fs(KERNEL_DS);
2387                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2388                         set_fs(oldfs);
2389                 } else
2390                         err = -EOPNOTSUPP;
2391
2392                 if (err == 0) {
2393                         err = -ENOBUFS;
2394                         dev = __dev_get_by_name(net, p.name);
2395                         if (!dev)
2396                                 goto err_exit;
2397                         err = dev_open(dev);
2398                 }
2399         }
2400 #endif
2401
2402 err_exit:
2403         rtnl_unlock();
2404         return err;
2405 }
2406
2407 /*
2408  *      Manual configuration of address on an interface
2409  */
2410 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2411                           const struct in6_addr *peer_pfx,
2412                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2413                           __u32 valid_lft)
2414 {
2415         struct inet6_ifaddr *ifp;
2416         struct inet6_dev *idev;
2417         struct net_device *dev;
2418         int scope;
2419         u32 flags;
2420         clock_t expires;
2421         unsigned long timeout;
2422
2423         ASSERT_RTNL();
2424
2425         if (plen > 128)
2426                 return -EINVAL;
2427
2428         /* check the lifetime */
2429         if (!valid_lft || prefered_lft > valid_lft)
2430                 return -EINVAL;
2431
2432         dev = __dev_get_by_index(net, ifindex);
2433         if (!dev)
2434                 return -ENODEV;
2435
2436         idev = addrconf_add_dev(dev);
2437         if (IS_ERR(idev))
2438                 return PTR_ERR(idev);
2439
2440         scope = ipv6_addr_scope(pfx);
2441
2442         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2443         if (addrconf_finite_timeout(timeout)) {
2444                 expires = jiffies_to_clock_t(timeout * HZ);
2445                 valid_lft = timeout;
2446                 flags = RTF_EXPIRES;
2447         } else {
2448                 expires = 0;
2449                 flags = 0;
2450                 ifa_flags |= IFA_F_PERMANENT;
2451         }
2452
2453         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2454         if (addrconf_finite_timeout(timeout)) {
2455                 if (timeout == 0)
2456                         ifa_flags |= IFA_F_DEPRECATED;
2457                 prefered_lft = timeout;
2458         }
2459
2460         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2461                             valid_lft, prefered_lft);
2462
2463         if (!IS_ERR(ifp)) {
2464                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2465                                       expires, flags);
2466                 /*
2467                  * Note that section 3.1 of RFC 4429 indicates
2468                  * that the Optimistic flag should not be set for
2469                  * manually configured addresses
2470                  */
2471                 addrconf_dad_start(ifp);
2472                 in6_ifa_put(ifp);
2473                 addrconf_verify(0);
2474                 return 0;
2475         }
2476
2477         return PTR_ERR(ifp);
2478 }
2479
2480 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2481                           unsigned int plen)
2482 {
2483         struct inet6_ifaddr *ifp;
2484         struct inet6_dev *idev;
2485         struct net_device *dev;
2486
2487         if (plen > 128)
2488                 return -EINVAL;
2489
2490         dev = __dev_get_by_index(net, ifindex);
2491         if (!dev)
2492                 return -ENODEV;
2493
2494         if ((idev = __in6_dev_get(dev)) == NULL)
2495                 return -ENXIO;
2496
2497         read_lock_bh(&idev->lock);
2498         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2499                 if (ifp->prefix_len == plen &&
2500                     ipv6_addr_equal(pfx, &ifp->addr)) {
2501                         in6_ifa_hold(ifp);
2502                         read_unlock_bh(&idev->lock);
2503
2504                         ipv6_del_addr(ifp);
2505                         return 0;
2506                 }
2507         }
2508         read_unlock_bh(&idev->lock);
2509         return -EADDRNOTAVAIL;
2510 }
2511
2512
2513 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2514 {
2515         struct in6_ifreq ireq;
2516         int err;
2517
2518         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2519                 return -EPERM;
2520
2521         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2522                 return -EFAULT;
2523
2524         rtnl_lock();
2525         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2526                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2527                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2528         rtnl_unlock();
2529         return err;
2530 }
2531
2532 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2533 {
2534         struct in6_ifreq ireq;
2535         int err;
2536
2537         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2538                 return -EPERM;
2539
2540         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2541                 return -EFAULT;
2542
2543         rtnl_lock();
2544         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2545                              ireq.ifr6_prefixlen);
2546         rtnl_unlock();
2547         return err;
2548 }
2549
2550 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2551                      int plen, int scope)
2552 {
2553         struct inet6_ifaddr *ifp;
2554
2555         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2556                             scope, IFA_F_PERMANENT, 0, 0);
2557         if (!IS_ERR(ifp)) {
2558                 spin_lock_bh(&ifp->lock);
2559                 ifp->flags &= ~IFA_F_TENTATIVE;
2560                 spin_unlock_bh(&ifp->lock);
2561                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2562                 in6_ifa_put(ifp);
2563         }
2564 }
2565
2566 #if IS_ENABLED(CONFIG_IPV6_SIT)
2567 static void sit_add_v4_addrs(struct inet6_dev *idev)
2568 {
2569         struct in6_addr addr;
2570         struct net_device *dev;
2571         struct net *net = dev_net(idev->dev);
2572         int scope;
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         } else {
2583                 scope = IPV6_ADDR_COMPATv4;
2584         }
2585
2586         if (addr.s6_addr32[3]) {
2587                 add_addr(idev, &addr, 128, scope);
2588                 return;
2589         }
2590
2591         for_each_netdev(net, dev) {
2592                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2593                 if (in_dev && (dev->flags & IFF_UP)) {
2594                         struct in_ifaddr *ifa;
2595
2596                         int flag = scope;
2597
2598                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2599                                 int plen;
2600
2601                                 addr.s6_addr32[3] = ifa->ifa_local;
2602
2603                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2604                                         continue;
2605                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2606                                         if (idev->dev->flags&IFF_POINTOPOINT)
2607                                                 continue;
2608                                         flag |= IFA_HOST;
2609                                 }
2610                                 if (idev->dev->flags&IFF_POINTOPOINT)
2611                                         plen = 64;
2612                                 else
2613                                         plen = 96;
2614
2615                                 add_addr(idev, &addr, plen, flag);
2616                         }
2617                 }
2618         }
2619 }
2620 #endif
2621
2622 static void init_loopback(struct net_device *dev)
2623 {
2624         struct inet6_dev  *idev;
2625         struct net_device *sp_dev;
2626         struct inet6_ifaddr *sp_ifa;
2627         struct rt6_info *sp_rt;
2628
2629         /* ::1 */
2630
2631         ASSERT_RTNL();
2632
2633         if ((idev = ipv6_find_idev(dev)) == NULL) {
2634                 pr_debug("%s: add_dev failed\n", __func__);
2635                 return;
2636         }
2637
2638         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2639
2640         /* Add routes to other interface's IPv6 addresses */
2641         for_each_netdev(dev_net(dev), sp_dev) {
2642                 if (!strcmp(sp_dev->name, dev->name))
2643                         continue;
2644
2645                 idev = __in6_dev_get(sp_dev);
2646                 if (!idev)
2647                         continue;
2648
2649                 read_lock_bh(&idev->lock);
2650                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2651
2652                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2653                                 continue;
2654
2655                         if (sp_ifa->rt)
2656                                 continue;
2657
2658                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2659
2660                         /* Failure cases are ignored */
2661                         if (!IS_ERR(sp_rt)) {
2662                                 sp_ifa->rt = sp_rt;
2663                                 ip6_ins_rt(sp_rt);
2664                         }
2665                 }
2666                 read_unlock_bh(&idev->lock);
2667         }
2668 }
2669
2670 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2671 {
2672         struct inet6_ifaddr *ifp;
2673         u32 addr_flags = IFA_F_PERMANENT;
2674
2675 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2676         if (idev->cnf.optimistic_dad &&
2677             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2678                 addr_flags |= IFA_F_OPTIMISTIC;
2679 #endif
2680
2681
2682         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2683         if (!IS_ERR(ifp)) {
2684                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2685                 addrconf_dad_start(ifp);
2686                 in6_ifa_put(ifp);
2687         }
2688 }
2689
2690 static void addrconf_dev_config(struct net_device *dev)
2691 {
2692         struct in6_addr addr;
2693         struct inet6_dev *idev;
2694
2695         ASSERT_RTNL();
2696
2697         if ((dev->type != ARPHRD_ETHER) &&
2698             (dev->type != ARPHRD_FDDI) &&
2699             (dev->type != ARPHRD_ARCNET) &&
2700             (dev->type != ARPHRD_INFINIBAND) &&
2701             (dev->type != ARPHRD_IEEE802154) &&
2702             (dev->type != ARPHRD_IEEE1394) &&
2703             (dev->type != ARPHRD_TUNNEL6)) {
2704                 /* Alas, we support only Ethernet autoconfiguration. */
2705                 return;
2706         }
2707
2708         idev = addrconf_add_dev(dev);
2709         if (IS_ERR(idev))
2710                 return;
2711
2712         memset(&addr, 0, sizeof(struct in6_addr));
2713         addr.s6_addr32[0] = htonl(0xFE800000);
2714
2715         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2716                 addrconf_add_linklocal(idev, &addr);
2717 }
2718
2719 #if IS_ENABLED(CONFIG_IPV6_SIT)
2720 static void addrconf_sit_config(struct net_device *dev)
2721 {
2722         struct inet6_dev *idev;
2723
2724         ASSERT_RTNL();
2725
2726         /*
2727          * Configure the tunnel with one of our IPv4
2728          * addresses... we should configure all of
2729          * our v4 addrs in the tunnel
2730          */
2731
2732         if ((idev = ipv6_find_idev(dev)) == NULL) {
2733                 pr_debug("%s: add_dev failed\n", __func__);
2734                 return;
2735         }
2736
2737         if (dev->priv_flags & IFF_ISATAP) {
2738                 struct in6_addr addr;
2739
2740                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2741                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2742                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2743                         addrconf_add_linklocal(idev, &addr);
2744                 return;
2745         }
2746
2747         sit_add_v4_addrs(idev);
2748
2749         if (dev->flags&IFF_POINTOPOINT)
2750                 addrconf_add_mroute(dev);
2751         else
2752                 sit_route_add(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         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2771
2772         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2773                 addrconf_add_linklocal(idev, &addr);
2774 }
2775 #endif
2776
2777 static inline int
2778 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2779 {
2780         struct in6_addr lladdr;
2781
2782         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2783                 addrconf_add_linklocal(idev, &lladdr);
2784                 return 0;
2785         }
2786         return -1;
2787 }
2788
2789 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2790                            void *ptr)
2791 {
2792         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2793         struct inet6_dev *idev = __in6_dev_get(dev);
2794         int run_pending = 0;
2795         int err;
2796
2797         switch (event) {
2798         case NETDEV_REGISTER:
2799                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2800                         idev = ipv6_add_dev(dev);
2801                         if (!idev)
2802                                 return notifier_from_errno(-ENOMEM);
2803                 }
2804                 break;
2805
2806         case NETDEV_UP:
2807         case NETDEV_CHANGE:
2808                 if (dev->flags & IFF_SLAVE)
2809                         break;
2810
2811                 if (event == NETDEV_UP) {
2812                         if (!addrconf_qdisc_ok(dev)) {
2813                                 /* device is not ready yet. */
2814                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2815                                         dev->name);
2816                                 break;
2817                         }
2818
2819                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2820                                 idev = ipv6_add_dev(dev);
2821
2822                         if (idev) {
2823                                 idev->if_flags |= IF_READY;
2824                                 run_pending = 1;
2825                         }
2826                 } else {
2827                         if (!addrconf_qdisc_ok(dev)) {
2828                                 /* device is still not ready. */
2829                                 break;
2830                         }
2831
2832                         if (idev) {
2833                                 if (idev->if_flags & IF_READY)
2834                                         /* device is already configured. */
2835                                         break;
2836                                 idev->if_flags |= IF_READY;
2837                         }
2838
2839                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2840                                 dev->name);
2841
2842                         run_pending = 1;
2843                 }
2844
2845                 switch (dev->type) {
2846 #if IS_ENABLED(CONFIG_IPV6_SIT)
2847                 case ARPHRD_SIT:
2848                         addrconf_sit_config(dev);
2849                         break;
2850 #endif
2851 #if IS_ENABLED(CONFIG_NET_IPGRE)
2852                 case ARPHRD_IPGRE:
2853                         addrconf_gre_config(dev);
2854                         break;
2855 #endif
2856                 case ARPHRD_LOOPBACK:
2857                         init_loopback(dev);
2858                         break;
2859
2860                 default:
2861                         addrconf_dev_config(dev);
2862                         break;
2863                 }
2864
2865                 if (idev) {
2866                         if (run_pending)
2867                                 addrconf_dad_run(idev);
2868
2869                         /*
2870                          * If the MTU changed during the interface down,
2871                          * when the interface up, the changed MTU must be
2872                          * reflected in the idev as well as routers.
2873                          */
2874                         if (idev->cnf.mtu6 != dev->mtu &&
2875                             dev->mtu >= IPV6_MIN_MTU) {
2876                                 rt6_mtu_change(dev, dev->mtu);
2877                                 idev->cnf.mtu6 = dev->mtu;
2878                         }
2879                         idev->tstamp = jiffies;
2880                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2881
2882                         /*
2883                          * If the changed mtu during down is lower than
2884                          * IPV6_MIN_MTU stop IPv6 on this interface.
2885                          */
2886                         if (dev->mtu < IPV6_MIN_MTU)
2887                                 addrconf_ifdown(dev, 1);
2888                 }
2889                 break;
2890
2891         case NETDEV_CHANGEMTU:
2892                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2893                         rt6_mtu_change(dev, dev->mtu);
2894                         idev->cnf.mtu6 = dev->mtu;
2895                         break;
2896                 }
2897
2898                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2899                         idev = ipv6_add_dev(dev);
2900                         if (idev)
2901                                 break;
2902                 }
2903
2904                 /*
2905                  * MTU falled under IPV6_MIN_MTU.
2906                  * Stop IPv6 on this interface.
2907                  */
2908
2909         case NETDEV_DOWN:
2910         case NETDEV_UNREGISTER:
2911                 /*
2912                  *      Remove all addresses from this interface.
2913                  */
2914                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2915                 break;
2916
2917         case NETDEV_CHANGENAME:
2918                 if (idev) {
2919                         snmp6_unregister_dev(idev);
2920                         addrconf_sysctl_unregister(idev);
2921                         addrconf_sysctl_register(idev);
2922                         err = snmp6_register_dev(idev);
2923                         if (err)
2924                                 return notifier_from_errno(err);
2925                 }
2926                 break;
2927
2928         case NETDEV_PRE_TYPE_CHANGE:
2929         case NETDEV_POST_TYPE_CHANGE:
2930                 addrconf_type_change(dev, event);
2931                 break;
2932         }
2933
2934         return NOTIFY_OK;
2935 }
2936
2937 /*
2938  *      addrconf module should be notified of a device going up
2939  */
2940 static struct notifier_block ipv6_dev_notf = {
2941         .notifier_call = addrconf_notify,
2942 };
2943
2944 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2945 {
2946         struct inet6_dev *idev;
2947         ASSERT_RTNL();
2948
2949         idev = __in6_dev_get(dev);
2950
2951         if (event == NETDEV_POST_TYPE_CHANGE)
2952                 ipv6_mc_remap(idev);
2953         else if (event == NETDEV_PRE_TYPE_CHANGE)
2954                 ipv6_mc_unmap(idev);
2955 }
2956
2957 static int addrconf_ifdown(struct net_device *dev, int how)
2958 {
2959         struct net *net = dev_net(dev);
2960         struct inet6_dev *idev;
2961         struct inet6_ifaddr *ifa;
2962         int state, i;
2963
2964         ASSERT_RTNL();
2965
2966         rt6_ifdown(net, dev);
2967         neigh_ifdown(&nd_tbl, dev);
2968
2969         idev = __in6_dev_get(dev);
2970         if (idev == NULL)
2971                 return -ENODEV;
2972
2973         /*
2974          * Step 1: remove reference to ipv6 device from parent device.
2975          *         Do not dev_put!
2976          */
2977         if (how) {
2978                 idev->dead = 1;
2979
2980                 /* protected by rtnl_lock */
2981                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2982
2983                 /* Step 1.5: remove snmp6 entry */
2984                 snmp6_unregister_dev(idev);
2985
2986         }
2987
2988         /* Step 2: clear hash table */
2989         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2990                 struct hlist_head *h = &inet6_addr_lst[i];
2991
2992                 spin_lock_bh(&addrconf_hash_lock);
2993         restart:
2994                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2995                         if (ifa->idev == idev) {
2996                                 hlist_del_init_rcu(&ifa->addr_lst);
2997                                 addrconf_del_dad_timer(ifa);
2998                                 goto restart;
2999                         }
3000                 }
3001                 spin_unlock_bh(&addrconf_hash_lock);
3002         }
3003
3004         write_lock_bh(&idev->lock);
3005
3006         addrconf_del_rs_timer(idev);
3007
3008         /* Step 2: clear flags for stateless addrconf */
3009         if (!how)
3010                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3011
3012 #ifdef CONFIG_IPV6_PRIVACY
3013         if (how && del_timer(&idev->regen_timer))
3014                 in6_dev_put(idev);
3015
3016         /* Step 3: clear tempaddr list */
3017         while (!list_empty(&idev->tempaddr_list)) {
3018                 ifa = list_first_entry(&idev->tempaddr_list,
3019                                        struct inet6_ifaddr, tmp_list);
3020                 list_del(&ifa->tmp_list);
3021                 write_unlock_bh(&idev->lock);
3022                 spin_lock_bh(&ifa->lock);
3023
3024                 if (ifa->ifpub) {
3025                         in6_ifa_put(ifa->ifpub);
3026                         ifa->ifpub = NULL;
3027                 }
3028                 spin_unlock_bh(&ifa->lock);
3029                 in6_ifa_put(ifa);
3030                 write_lock_bh(&idev->lock);
3031         }
3032 #endif
3033
3034         while (!list_empty(&idev->addr_list)) {
3035                 ifa = list_first_entry(&idev->addr_list,
3036                                        struct inet6_ifaddr, if_list);
3037                 addrconf_del_dad_timer(ifa);
3038
3039                 list_del(&ifa->if_list);
3040
3041                 write_unlock_bh(&idev->lock);
3042
3043                 spin_lock_bh(&ifa->state_lock);
3044                 state = ifa->state;
3045                 ifa->state = INET6_IFADDR_STATE_DEAD;
3046                 spin_unlock_bh(&ifa->state_lock);
3047
3048                 if (state != INET6_IFADDR_STATE_DEAD) {
3049                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3050                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3051                 }
3052                 in6_ifa_put(ifa);
3053
3054                 write_lock_bh(&idev->lock);
3055         }
3056
3057         write_unlock_bh(&idev->lock);
3058
3059         /* Step 5: Discard multicast list */
3060         if (how)
3061                 ipv6_mc_destroy_dev(idev);
3062         else
3063                 ipv6_mc_down(idev);
3064
3065         idev->tstamp = jiffies;
3066
3067         /* Last: Shot the device (if unregistered) */
3068         if (how) {
3069                 addrconf_sysctl_unregister(idev);
3070                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3071                 neigh_ifdown(&nd_tbl, dev);
3072                 in6_dev_put(idev);
3073         }
3074         return 0;
3075 }
3076
3077 static void addrconf_rs_timer(unsigned long data)
3078 {
3079         struct inet6_dev *idev = (struct inet6_dev *)data;
3080         struct net_device *dev = idev->dev;
3081         struct in6_addr lladdr;
3082
3083         write_lock(&idev->lock);
3084         if (idev->dead || !(idev->if_flags & IF_READY))
3085                 goto out;
3086
3087         if (!ipv6_accept_ra(idev))
3088                 goto out;
3089
3090         /* Announcement received after solicitation was sent */
3091         if (idev->if_flags & IF_RA_RCVD)
3092                 goto out;
3093
3094         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3095                 write_unlock(&idev->lock);
3096                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3097                         ndisc_send_rs(dev, &lladdr,
3098                                       &in6addr_linklocal_allrouters);
3099                 else
3100                         goto put;
3101
3102                 write_lock(&idev->lock);
3103                 /* The wait after the last probe can be shorter */
3104                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3105                                              idev->cnf.rtr_solicits) ?
3106                                       idev->cnf.rtr_solicit_delay :
3107                                       idev->cnf.rtr_solicit_interval);
3108         } else {
3109                 /*
3110                  * Note: we do not support deprecated "all on-link"
3111                  * assumption any longer.
3112                  */
3113                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3114         }
3115
3116 out:
3117         write_unlock(&idev->lock);
3118 put:
3119         in6_dev_put(idev);
3120 }
3121
3122 /*
3123  *      Duplicate Address Detection
3124  */
3125 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3126 {
3127         unsigned long rand_num;
3128         struct inet6_dev *idev = ifp->idev;
3129
3130         if (ifp->flags & IFA_F_OPTIMISTIC)
3131                 rand_num = 0;
3132         else
3133                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3134
3135         ifp->dad_probes = idev->cnf.dad_transmits;
3136         addrconf_mod_dad_timer(ifp, rand_num);
3137 }
3138
3139 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3140 {
3141         struct inet6_dev *idev = ifp->idev;
3142         struct net_device *dev = idev->dev;
3143
3144         addrconf_join_solict(dev, &ifp->addr);
3145
3146         net_srandom(ifp->addr.s6_addr32[3]);
3147
3148         read_lock_bh(&idev->lock);
3149         spin_lock(&ifp->lock);
3150         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3151                 goto out;
3152
3153         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3154             idev->cnf.accept_dad < 1 ||
3155             !(ifp->flags&IFA_F_TENTATIVE) ||
3156             ifp->flags & IFA_F_NODAD) {
3157                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3158                 spin_unlock(&ifp->lock);
3159                 read_unlock_bh(&idev->lock);
3160
3161                 addrconf_dad_completed(ifp);
3162                 return;
3163         }
3164
3165         if (!(idev->if_flags & IF_READY)) {
3166                 spin_unlock(&ifp->lock);
3167                 read_unlock_bh(&idev->lock);
3168                 /*
3169                  * If the device is not ready:
3170                  * - keep it tentative if it is a permanent address.
3171                  * - otherwise, kill it.
3172                  */
3173                 in6_ifa_hold(ifp);
3174                 addrconf_dad_stop(ifp, 0);
3175                 return;
3176         }
3177
3178         /*
3179          * Optimistic nodes can start receiving
3180          * Frames right away
3181          */
3182         if (ifp->flags & IFA_F_OPTIMISTIC)
3183                 ip6_ins_rt(ifp->rt);
3184
3185         addrconf_dad_kick(ifp);
3186 out:
3187         spin_unlock(&ifp->lock);
3188         read_unlock_bh(&idev->lock);
3189 }
3190
3191 static void addrconf_dad_timer(unsigned long data)
3192 {
3193         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3194         struct inet6_dev *idev = ifp->idev;
3195         struct in6_addr mcaddr;
3196
3197         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3198                 goto out;
3199
3200         write_lock(&idev->lock);
3201         if (idev->dead || !(idev->if_flags & IF_READY)) {
3202                 write_unlock(&idev->lock);
3203                 goto out;
3204         }
3205
3206         spin_lock(&ifp->lock);
3207         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3208                 spin_unlock(&ifp->lock);
3209                 write_unlock(&idev->lock);
3210                 goto out;
3211         }
3212
3213         if (ifp->dad_probes == 0) {
3214                 /*
3215                  * DAD was successful
3216                  */
3217
3218                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3219                 spin_unlock(&ifp->lock);
3220                 write_unlock(&idev->lock);
3221
3222                 addrconf_dad_completed(ifp);
3223
3224                 goto out;
3225         }
3226
3227         ifp->dad_probes--;
3228         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3229         spin_unlock(&ifp->lock);
3230         write_unlock(&idev->lock);
3231
3232         /* send a neighbour solicitation for our addr */
3233         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3234         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3235 out:
3236         in6_ifa_put(ifp);
3237 }
3238
3239 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3240 {
3241         struct net_device *dev = ifp->idev->dev;
3242         struct in6_addr lladdr;
3243         bool send_rs, send_mld;
3244
3245         addrconf_del_dad_timer(ifp);
3246
3247         /*
3248          *      Configure the address for reception. Now it is valid.
3249          */
3250
3251         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3252
3253         /* If added prefix is link local and we are prepared to process
3254            router advertisements, start sending router solicitations.
3255          */
3256
3257         read_lock_bh(&ifp->idev->lock);
3258         spin_lock(&ifp->lock);
3259         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3260                    ifp->idev->valid_ll_addr_cnt == 1;
3261         send_rs = send_mld &&
3262                   ipv6_accept_ra(ifp->idev) &&
3263                   ifp->idev->cnf.rtr_solicits > 0 &&
3264                   (dev->flags&IFF_LOOPBACK) == 0;
3265         spin_unlock(&ifp->lock);
3266         read_unlock_bh(&ifp->idev->lock);
3267
3268         /* While dad is in progress mld report's source address is in6_addrany.
3269          * Resend with proper ll now.
3270          */
3271         if (send_mld)
3272                 ipv6_mc_dad_complete(ifp->idev);
3273
3274         if (send_rs) {
3275                 /*
3276                  *      If a host as already performed a random delay
3277                  *      [...] as part of DAD [...] there is no need
3278                  *      to delay again before sending the first RS
3279                  */
3280                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3281                         return;
3282                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3283
3284                 write_lock_bh(&ifp->idev->lock);
3285                 spin_lock(&ifp->lock);
3286                 ifp->idev->rs_probes = 1;
3287                 ifp->idev->if_flags |= IF_RS_SENT;
3288                 addrconf_mod_rs_timer(ifp->idev,
3289                                       ifp->idev->cnf.rtr_solicit_interval);
3290                 spin_unlock(&ifp->lock);
3291                 write_unlock_bh(&ifp->idev->lock);
3292         }
3293 }
3294
3295 static void addrconf_dad_run(struct inet6_dev *idev)
3296 {
3297         struct inet6_ifaddr *ifp;
3298
3299         read_lock_bh(&idev->lock);
3300         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3301                 spin_lock(&ifp->lock);
3302                 if (ifp->flags & IFA_F_TENTATIVE &&
3303                     ifp->state == INET6_IFADDR_STATE_DAD)
3304                         addrconf_dad_kick(ifp);
3305                 spin_unlock(&ifp->lock);
3306         }
3307         read_unlock_bh(&idev->lock);
3308 }
3309
3310 #ifdef CONFIG_PROC_FS
3311 struct if6_iter_state {
3312         struct seq_net_private p;
3313         int bucket;
3314         int offset;
3315 };
3316
3317 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3318 {
3319         struct inet6_ifaddr *ifa = NULL;
3320         struct if6_iter_state *state = seq->private;
3321         struct net *net = seq_file_net(seq);
3322         int p = 0;
3323
3324         /* initial bucket if pos is 0 */
3325         if (pos == 0) {
3326                 state->bucket = 0;
3327                 state->offset = 0;
3328         }
3329
3330         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3331                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3332                                          addr_lst) {
3333                         if (!net_eq(dev_net(ifa->idev->dev), net))
3334                                 continue;
3335                         /* sync with offset */
3336                         if (p < state->offset) {
3337                                 p++;
3338                                 continue;
3339                         }
3340                         state->offset++;
3341                         return ifa;
3342                 }
3343
3344                 /* prepare for next bucket */
3345                 state->offset = 0;
3346                 p = 0;
3347         }
3348         return NULL;
3349 }
3350
3351 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3352                                          struct inet6_ifaddr *ifa)
3353 {
3354         struct if6_iter_state *state = seq->private;
3355         struct net *net = seq_file_net(seq);
3356
3357         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3358                 if (!net_eq(dev_net(ifa->idev->dev), net))
3359                         continue;
3360                 state->offset++;
3361                 return ifa;
3362         }
3363
3364         while (++state->bucket < IN6_ADDR_HSIZE) {
3365                 state->offset = 0;
3366                 hlist_for_each_entry_rcu_bh(ifa,
3367                                      &inet6_addr_lst[state->bucket], addr_lst) {
3368                         if (!net_eq(dev_net(ifa->idev->dev), net))
3369                                 continue;
3370                         state->offset++;
3371                         return ifa;
3372                 }
3373         }
3374
3375         return NULL;
3376 }
3377
3378 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3379         __acquires(rcu_bh)
3380 {
3381         rcu_read_lock_bh();
3382         return if6_get_first(seq, *pos);
3383 }
3384
3385 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3386 {
3387         struct inet6_ifaddr *ifa;
3388
3389         ifa = if6_get_next(seq, v);
3390         ++*pos;
3391         return ifa;
3392 }
3393
3394 static void if6_seq_stop(struct seq_file *seq, void *v)
3395         __releases(rcu_bh)
3396 {
3397         rcu_read_unlock_bh();
3398 }
3399
3400 static int if6_seq_show(struct seq_file *seq, void *v)
3401 {
3402         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3403         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3404                    &ifp->addr,
3405                    ifp->idev->dev->ifindex,
3406                    ifp->prefix_len,
3407                    ifp->scope,
3408                    ifp->flags,
3409                    ifp->idev->dev->name);
3410         return 0;
3411 }
3412
3413 static const struct seq_operations if6_seq_ops = {
3414         .start  = if6_seq_start,
3415         .next   = if6_seq_next,
3416         .show   = if6_seq_show,
3417         .stop   = if6_seq_stop,
3418 };
3419
3420 static int if6_seq_open(struct inode *inode, struct file *file)
3421 {
3422         return seq_open_net(inode, file, &if6_seq_ops,
3423                             sizeof(struct if6_iter_state));
3424 }
3425
3426 static const struct file_operations if6_fops = {
3427         .owner          = THIS_MODULE,
3428         .open           = if6_seq_open,
3429         .read           = seq_read,
3430         .llseek         = seq_lseek,
3431         .release        = seq_release_net,
3432 };
3433
3434 static int __net_init if6_proc_net_init(struct net *net)
3435 {
3436         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3437                 return -ENOMEM;
3438         return 0;
3439 }
3440
3441 static void __net_exit if6_proc_net_exit(struct net *net)
3442 {
3443         remove_proc_entry("if_inet6", net->proc_net);
3444 }
3445
3446 static struct pernet_operations if6_proc_net_ops = {
3447        .init = if6_proc_net_init,
3448        .exit = if6_proc_net_exit,
3449 };
3450
3451 int __init if6_proc_init(void)
3452 {
3453         return register_pernet_subsys(&if6_proc_net_ops);
3454 }
3455
3456 void if6_proc_exit(void)
3457 {
3458         unregister_pernet_subsys(&if6_proc_net_ops);
3459 }
3460 #endif  /* CONFIG_PROC_FS */
3461
3462 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3463 /* Check if address is a home address configured on any interface. */
3464 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3465 {
3466         int ret = 0;
3467         struct inet6_ifaddr *ifp = NULL;
3468         unsigned int hash = inet6_addr_hash(addr);
3469
3470         rcu_read_lock_bh();
3471         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3472                 if (!net_eq(dev_net(ifp->idev->dev), net))
3473                         continue;
3474                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3475                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3476                         ret = 1;
3477                         break;
3478                 }
3479         }
3480         rcu_read_unlock_bh();
3481         return ret;
3482 }
3483 #endif
3484
3485 /*
3486  *      Periodic address status verification
3487  */
3488
3489 static void addrconf_verify(unsigned long foo)
3490 {
3491         unsigned long now, next, next_sec, next_sched;
3492         struct inet6_ifaddr *ifp;
3493         int i;
3494
3495         rcu_read_lock_bh();
3496         spin_lock(&addrconf_verify_lock);
3497         now = jiffies;
3498         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3499
3500         del_timer(&addr_chk_timer);
3501
3502         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3503 restart:
3504                 hlist_for_each_entry_rcu_bh(ifp,
3505                                          &inet6_addr_lst[i], addr_lst) {
3506                         unsigned long age;
3507
3508                         if (ifp->flags & IFA_F_PERMANENT)
3509                                 continue;
3510
3511                         spin_lock(&ifp->lock);
3512                         /* We try to batch several events at once. */
3513                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3514
3515                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3516                             age >= ifp->valid_lft) {
3517                                 spin_unlock(&ifp->lock);
3518                                 in6_ifa_hold(ifp);
3519                                 ipv6_del_addr(ifp);
3520                                 goto restart;
3521                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3522                                 spin_unlock(&ifp->lock);
3523                                 continue;
3524                         } else if (age >= ifp->prefered_lft) {
3525                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3526                                 int deprecate = 0;
3527
3528                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3529                                         deprecate = 1;
3530                                         ifp->flags |= IFA_F_DEPRECATED;
3531                                 }
3532
3533                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3534                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3535
3536                                 spin_unlock(&ifp->lock);
3537
3538                                 if (deprecate) {
3539                                         in6_ifa_hold(ifp);
3540
3541                                         ipv6_ifa_notify(0, ifp);
3542                                         in6_ifa_put(ifp);
3543                                         goto restart;
3544                                 }
3545 #ifdef CONFIG_IPV6_PRIVACY
3546                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3547                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3548                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3549                                         ifp->idev->cnf.dad_transmits *
3550                                         ifp->idev->nd_parms->retrans_time / HZ;
3551
3552                                 if (age >= ifp->prefered_lft - regen_advance) {
3553                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3554                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3555                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3556                                         if (!ifp->regen_count && ifpub) {
3557                                                 ifp->regen_count++;
3558                                                 in6_ifa_hold(ifp);
3559                                                 in6_ifa_hold(ifpub);
3560                                                 spin_unlock(&ifp->lock);
3561
3562                                                 spin_lock(&ifpub->lock);
3563                                                 ifpub->regen_count = 0;
3564                                                 spin_unlock(&ifpub->lock);
3565                                                 ipv6_create_tempaddr(ifpub, ifp);
3566                                                 in6_ifa_put(ifpub);
3567                                                 in6_ifa_put(ifp);
3568                                                 goto restart;
3569                                         }
3570                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3571                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3572                                 spin_unlock(&ifp->lock);
3573 #endif
3574                         } else {
3575                                 /* ifp->prefered_lft <= ifp->valid_lft */
3576                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3577                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3578                                 spin_unlock(&ifp->lock);
3579                         }
3580                 }
3581         }
3582
3583         next_sec = round_jiffies_up(next);
3584         next_sched = next;
3585
3586         /* If rounded timeout is accurate enough, accept it. */
3587         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3588                 next_sched = next_sec;
3589
3590         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3591         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3592                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3593
3594         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3595               now, next, next_sec, next_sched);
3596
3597         addr_chk_timer.expires = next_sched;
3598         add_timer(&addr_chk_timer);
3599         spin_unlock(&addrconf_verify_lock);
3600         rcu_read_unlock_bh();
3601 }
3602
3603 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3604                                      struct in6_addr **peer_pfx)
3605 {
3606         struct in6_addr *pfx = NULL;
3607
3608         *peer_pfx = NULL;
3609
3610         if (addr)
3611                 pfx = nla_data(addr);
3612
3613         if (local) {
3614                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3615                         *peer_pfx = pfx;
3616                 pfx = nla_data(local);
3617         }
3618
3619         return pfx;
3620 }
3621
3622 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3623         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3624         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3625         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3626 };
3627
3628 static int
3629 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3630 {
3631         struct net *net = sock_net(skb->sk);
3632         struct ifaddrmsg *ifm;
3633         struct nlattr *tb[IFA_MAX+1];
3634         struct in6_addr *pfx, *peer_pfx;
3635         int err;
3636
3637         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3638         if (err < 0)
3639                 return err;
3640
3641         ifm = nlmsg_data(nlh);
3642         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3643         if (pfx == NULL)
3644                 return -EINVAL;
3645
3646         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3647 }
3648
3649 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3650                              u32 prefered_lft, u32 valid_lft)
3651 {
3652         u32 flags;
3653         clock_t expires;
3654         unsigned long timeout;
3655
3656         if (!valid_lft || (prefered_lft > valid_lft))
3657                 return -EINVAL;
3658
3659         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3660         if (addrconf_finite_timeout(timeout)) {
3661                 expires = jiffies_to_clock_t(timeout * HZ);
3662                 valid_lft = timeout;
3663                 flags = RTF_EXPIRES;
3664         } else {
3665                 expires = 0;
3666                 flags = 0;
3667                 ifa_flags |= IFA_F_PERMANENT;
3668         }
3669
3670         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3671         if (addrconf_finite_timeout(timeout)) {
3672                 if (timeout == 0)
3673                         ifa_flags |= IFA_F_DEPRECATED;
3674                 prefered_lft = timeout;
3675         }
3676
3677         spin_lock_bh(&ifp->lock);
3678         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3679         ifp->tstamp = jiffies;
3680         ifp->valid_lft = valid_lft;
3681         ifp->prefered_lft = prefered_lft;
3682
3683         spin_unlock_bh(&ifp->lock);
3684         if (!(ifp->flags&IFA_F_TENTATIVE))
3685                 ipv6_ifa_notify(0, ifp);
3686
3687         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3688                               expires, flags);
3689         addrconf_verify(0);
3690
3691         return 0;
3692 }
3693
3694 static int
3695 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3696 {
3697         struct net *net = sock_net(skb->sk);
3698         struct ifaddrmsg *ifm;
3699         struct nlattr *tb[IFA_MAX+1];
3700         struct in6_addr *pfx, *peer_pfx;
3701         struct inet6_ifaddr *ifa;
3702         struct net_device *dev;
3703         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3704         u8 ifa_flags;
3705         int err;
3706
3707         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3708         if (err < 0)
3709                 return err;
3710
3711         ifm = nlmsg_data(nlh);
3712         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3713         if (pfx == NULL)
3714                 return -EINVAL;
3715
3716         if (tb[IFA_CACHEINFO]) {
3717                 struct ifa_cacheinfo *ci;
3718
3719                 ci = nla_data(tb[IFA_CACHEINFO]);
3720                 valid_lft = ci->ifa_valid;
3721                 preferred_lft = ci->ifa_prefered;
3722         } else {
3723                 preferred_lft = INFINITY_LIFE_TIME;
3724                 valid_lft = INFINITY_LIFE_TIME;
3725         }
3726
3727         dev =  __dev_get_by_index(net, ifm->ifa_index);
3728         if (dev == NULL)
3729                 return -ENODEV;
3730
3731         /* We ignore other flags so far. */
3732         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3733
3734         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3735         if (ifa == NULL) {
3736                 /*
3737                  * It would be best to check for !NLM_F_CREATE here but
3738                  * userspace alreay relies on not having to provide this.
3739                  */
3740                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3741                                       ifm->ifa_prefixlen, ifa_flags,
3742                                       preferred_lft, valid_lft);
3743         }
3744
3745         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3746             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3747                 err = -EEXIST;
3748         else
3749                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3750
3751         in6_ifa_put(ifa);
3752
3753         return err;
3754 }
3755
3756 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3757                           u8 scope, int ifindex)
3758 {
3759         struct ifaddrmsg *ifm;
3760
3761         ifm = nlmsg_data(nlh);
3762         ifm->ifa_family = AF_INET6;
3763         ifm->ifa_prefixlen = prefixlen;
3764         ifm->ifa_flags = flags;
3765         ifm->ifa_scope = scope;
3766         ifm->ifa_index = ifindex;
3767 }
3768
3769 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3770                          unsigned long tstamp, u32 preferred, u32 valid)
3771 {
3772         struct ifa_cacheinfo ci;
3773
3774         ci.cstamp = cstamp_delta(cstamp);
3775         ci.tstamp = cstamp_delta(tstamp);
3776         ci.ifa_prefered = preferred;
3777         ci.ifa_valid = valid;
3778
3779         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3780 }
3781
3782 static inline int rt_scope(int ifa_scope)
3783 {
3784         if (ifa_scope & IFA_HOST)
3785                 return RT_SCOPE_HOST;
3786         else if (ifa_scope & IFA_LINK)
3787                 return RT_SCOPE_LINK;
3788         else if (ifa_scope & IFA_SITE)
3789                 return RT_SCOPE_SITE;
3790         else
3791                 return RT_SCOPE_UNIVERSE;
3792 }
3793
3794 static inline int inet6_ifaddr_msgsize(void)
3795 {
3796         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3797                + nla_total_size(16) /* IFA_LOCAL */
3798                + nla_total_size(16) /* IFA_ADDRESS */
3799                + nla_total_size(sizeof(struct ifa_cacheinfo));
3800 }
3801
3802 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3803                              u32 portid, u32 seq, int event, unsigned int flags)
3804 {
3805         struct nlmsghdr  *nlh;
3806         u32 preferred, valid;
3807
3808         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3809         if (nlh == NULL)
3810                 return -EMSGSIZE;
3811
3812         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3813                       ifa->idev->dev->ifindex);
3814
3815         if (!(ifa->flags&IFA_F_PERMANENT)) {
3816                 preferred = ifa->prefered_lft;
3817                 valid = ifa->valid_lft;
3818                 if (preferred != INFINITY_LIFE_TIME) {
3819                         long tval = (jiffies - ifa->tstamp)/HZ;
3820                         if (preferred > tval)
3821                                 preferred -= tval;
3822                         else
3823                                 preferred = 0;
3824                         if (valid != INFINITY_LIFE_TIME) {
3825                                 if (valid > tval)
3826                                         valid -= tval;
3827                                 else
3828                                         valid = 0;
3829                         }
3830                 }
3831         } else {
3832                 preferred = INFINITY_LIFE_TIME;
3833                 valid = INFINITY_LIFE_TIME;
3834         }
3835
3836         if (!ipv6_addr_any(&ifa->peer_addr)) {
3837                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3838                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3839                         goto error;
3840         } else
3841                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3842                         goto error;
3843
3844         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3845                 goto error;
3846
3847         return nlmsg_end(skb, nlh);
3848
3849 error:
3850         nlmsg_cancel(skb, nlh);
3851         return -EMSGSIZE;
3852 }
3853
3854 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3855                                 u32 portid, u32 seq, int event, u16 flags)
3856 {
3857         struct nlmsghdr  *nlh;
3858         u8 scope = RT_SCOPE_UNIVERSE;
3859         int ifindex = ifmca->idev->dev->ifindex;
3860
3861         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3862                 scope = RT_SCOPE_SITE;
3863
3864         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3865         if (nlh == NULL)
3866                 return -EMSGSIZE;
3867
3868         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3869         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3870             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3871                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3872                 nlmsg_cancel(skb, nlh);
3873                 return -EMSGSIZE;
3874         }
3875
3876         return nlmsg_end(skb, nlh);
3877 }
3878
3879 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3880                                 u32 portid, u32 seq, int event, unsigned int flags)
3881 {
3882         struct nlmsghdr  *nlh;
3883         u8 scope = RT_SCOPE_UNIVERSE;
3884         int ifindex = ifaca->aca_idev->dev->ifindex;
3885
3886         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3887                 scope = RT_SCOPE_SITE;
3888
3889         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3890         if (nlh == NULL)
3891                 return -EMSGSIZE;
3892
3893         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3894         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3895             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3896                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3897                 nlmsg_cancel(skb, nlh);
3898                 return -EMSGSIZE;
3899         }
3900
3901         return nlmsg_end(skb, nlh);
3902 }
3903
3904 enum addr_type_t {
3905         UNICAST_ADDR,
3906         MULTICAST_ADDR,
3907         ANYCAST_ADDR,
3908 };
3909
3910 /* called with rcu_read_lock() */
3911 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3912                           struct netlink_callback *cb, enum addr_type_t type,
3913                           int s_ip_idx, int *p_ip_idx)
3914 {
3915         struct ifmcaddr6 *ifmca;
3916         struct ifacaddr6 *ifaca;
3917         int err = 1;
3918         int ip_idx = *p_ip_idx;
3919
3920         read_lock_bh(&idev->lock);
3921         switch (type) {
3922         case UNICAST_ADDR: {
3923                 struct inet6_ifaddr *ifa;
3924
3925                 /* unicast address incl. temp addr */
3926                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3927                         if (++ip_idx < s_ip_idx)
3928                                 continue;
3929                         err = inet6_fill_ifaddr(skb, ifa,
3930                                                 NETLINK_CB(cb->skb).portid,
3931                                                 cb->nlh->nlmsg_seq,
3932                                                 RTM_NEWADDR,
3933                                                 NLM_F_MULTI);
3934                         if (err <= 0)
3935                                 break;
3936                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3937                 }
3938                 break;
3939         }
3940         case MULTICAST_ADDR:
3941                 /* multicast address */
3942                 for (ifmca = idev->mc_list; ifmca;
3943                      ifmca = ifmca->next, ip_idx++) {
3944                         if (ip_idx < s_ip_idx)
3945                                 continue;
3946                         err = inet6_fill_ifmcaddr(skb, ifmca,
3947                                                   NETLINK_CB(cb->skb).portid,
3948                                                   cb->nlh->nlmsg_seq,
3949                                                   RTM_GETMULTICAST,
3950                                                   NLM_F_MULTI);
3951                         if (err <= 0)
3952                                 break;
3953                 }
3954                 break;
3955         case ANYCAST_ADDR:
3956                 /* anycast address */
3957                 for (ifaca = idev->ac_list; ifaca;
3958                      ifaca = ifaca->aca_next, ip_idx++) {
3959                         if (ip_idx < s_ip_idx)
3960                                 continue;
3961                         err = inet6_fill_ifacaddr(skb, ifaca,
3962                                                   NETLINK_CB(cb->skb).portid,
3963                                                   cb->nlh->nlmsg_seq,
3964                                                   RTM_GETANYCAST,
3965                                                   NLM_F_MULTI);
3966                         if (err <= 0)
3967                                 break;
3968                 }
3969                 break;
3970         default:
3971                 break;
3972         }
3973         read_unlock_bh(&idev->lock);
3974         *p_ip_idx = ip_idx;
3975         return err;
3976 }
3977
3978 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3979                            enum addr_type_t type)
3980 {
3981         struct net *net = sock_net(skb->sk);
3982         int h, s_h;
3983         int idx, ip_idx;
3984         int s_idx, s_ip_idx;
3985         struct net_device *dev;
3986         struct inet6_dev *idev;
3987         struct hlist_head *head;
3988
3989         s_h = cb->args[0];
3990         s_idx = idx = cb->args[1];
3991         s_ip_idx = ip_idx = cb->args[2];
3992
3993         rcu_read_lock();
3994         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3995         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3996                 idx = 0;
3997                 head = &net->dev_index_head[h];
3998                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3999                         if (idx < s_idx)
4000                                 goto cont;
4001                         if (h > s_h || idx > s_idx)
4002                                 s_ip_idx = 0;
4003                         ip_idx = 0;
4004                         idev = __in6_dev_get(dev);
4005                         if (!idev)
4006                                 goto cont;
4007
4008                         if (in6_dump_addrs(idev, skb, cb, type,
4009                                            s_ip_idx, &ip_idx) <= 0)
4010                                 goto done;
4011 cont:
4012                         idx++;
4013                 }
4014         }
4015 done:
4016         rcu_read_unlock();
4017         cb->args[0] = h;
4018         cb->args[1] = idx;
4019         cb->args[2] = ip_idx;
4020
4021         return skb->len;
4022 }
4023
4024 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4025 {
4026         enum addr_type_t type = UNICAST_ADDR;
4027
4028         return inet6_dump_addr(skb, cb, type);
4029 }
4030
4031 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4032 {
4033         enum addr_type_t type = MULTICAST_ADDR;
4034
4035         return inet6_dump_addr(skb, cb, type);
4036 }
4037
4038
4039 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4040 {
4041         enum addr_type_t type = ANYCAST_ADDR;
4042
4043         return inet6_dump_addr(skb, cb, type);
4044 }
4045
4046 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4047 {
4048         struct net *net = sock_net(in_skb->sk);
4049         struct ifaddrmsg *ifm;
4050         struct nlattr *tb[IFA_MAX+1];
4051         struct in6_addr *addr = NULL, *peer;
4052         struct net_device *dev = NULL;
4053         struct inet6_ifaddr *ifa;
4054         struct sk_buff *skb;
4055         int err;
4056
4057         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4058         if (err < 0)
4059                 goto errout;
4060
4061         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4062         if (addr == NULL) {
4063                 err = -EINVAL;
4064                 goto errout;
4065         }
4066
4067         ifm = nlmsg_data(nlh);
4068         if (ifm->ifa_index)
4069                 dev = __dev_get_by_index(net, ifm->ifa_index);
4070
4071         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4072         if (!ifa) {
4073                 err = -EADDRNOTAVAIL;
4074                 goto errout;
4075         }
4076
4077         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4078         if (!skb) {
4079                 err = -ENOBUFS;
4080                 goto errout_ifa;
4081         }
4082
4083         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4084                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4085         if (err < 0) {
4086                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4087                 WARN_ON(err == -EMSGSIZE);
4088                 kfree_skb(skb);
4089                 goto errout_ifa;
4090         }
4091         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4092 errout_ifa:
4093         in6_ifa_put(ifa);
4094 errout:
4095         return err;
4096 }
4097
4098 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4099 {
4100         struct sk_buff *skb;
4101         struct net *net = dev_net(ifa->idev->dev);
4102         int err = -ENOBUFS;
4103
4104         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4105         if (skb == NULL)
4106                 goto errout;
4107
4108         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4109         if (err < 0) {
4110                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4111                 WARN_ON(err == -EMSGSIZE);
4112                 kfree_skb(skb);
4113                 goto errout;
4114         }
4115         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4116         return;
4117 errout:
4118         if (err < 0)
4119                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4120 }
4121
4122 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4123                                 __s32 *array, int bytes)
4124 {
4125         BUG_ON(bytes < (DEVCONF_MAX * 4));
4126
4127         memset(array, 0, bytes);
4128         array[DEVCONF_FORWARDING] = cnf->forwarding;
4129         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4130         array[DEVCONF_MTU6] = cnf->mtu6;
4131         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4132         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4133         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4134         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4135         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4136         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4137                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4138         array[DEVCONF_RTR_SOLICIT_DELAY] =
4139                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4140         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4141         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4142                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4143         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4144                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4145 #ifdef CONFIG_IPV6_PRIVACY
4146         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4147         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4148         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4149         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4150         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4151 #endif
4152         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4153         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4154         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4155 #ifdef CONFIG_IPV6_ROUTER_PREF
4156         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4157         array[DEVCONF_RTR_PROBE_INTERVAL] =
4158                 jiffies_to_msecs(cnf->rtr_probe_interval);
4159 #ifdef CONFIG_IPV6_ROUTE_INFO
4160         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4161 #endif
4162 #endif
4163         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4164         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4165 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4166         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4167 #endif
4168 #ifdef CONFIG_IPV6_MROUTE
4169         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4170 #endif
4171         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4172         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4173         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4174         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4175         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4176 }
4177
4178 static inline size_t inet6_ifla6_size(void)
4179 {
4180         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4181              + nla_total_size(sizeof(struct ifla_cacheinfo))
4182              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4183              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4184              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4185              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4186 }
4187
4188 static inline size_t inet6_if_nlmsg_size(void)
4189 {
4190         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4191                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4192                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4193                + nla_total_size(4) /* IFLA_MTU */
4194                + nla_total_size(4) /* IFLA_LINK */
4195                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4196 }
4197
4198 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4199                                       int items, int bytes)
4200 {
4201         int i;
4202         int pad = bytes - sizeof(u64) * items;
4203         BUG_ON(pad < 0);
4204
4205         /* Use put_unaligned() because stats may not be aligned for u64. */
4206         put_unaligned(items, &stats[0]);
4207         for (i = 1; i < items; i++)
4208                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4209
4210         memset(&stats[items], 0, pad);
4211 }
4212
4213 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4214                                       int items, int bytes, size_t syncpoff)
4215 {
4216         int i;
4217         int pad = bytes - sizeof(u64) * items;
4218         BUG_ON(pad < 0);
4219
4220         /* Use put_unaligned() because stats may not be aligned for u64. */
4221         put_unaligned(items, &stats[0]);
4222         for (i = 1; i < items; i++)
4223                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4224
4225         memset(&stats[items], 0, pad);
4226 }
4227
4228 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4229                              int bytes)
4230 {
4231         switch (attrtype) {
4232         case IFLA_INET6_STATS:
4233                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4234                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4235                 break;
4236         case IFLA_INET6_ICMP6STATS:
4237                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4238                 break;
4239         }
4240 }
4241
4242 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4243 {
4244         struct nlattr *nla;
4245         struct ifla_cacheinfo ci;
4246
4247         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4248                 goto nla_put_failure;
4249         ci.max_reasm_len = IPV6_MAXPLEN;
4250         ci.tstamp = cstamp_delta(idev->tstamp);
4251         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4252         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4253         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4254                 goto nla_put_failure;
4255         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4256         if (nla == NULL)
4257                 goto nla_put_failure;
4258         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4259
4260         /* XXX - MC not implemented */
4261
4262         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4263         if (nla == NULL)
4264                 goto nla_put_failure;
4265         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4266
4267         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4268         if (nla == NULL)
4269                 goto nla_put_failure;
4270         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4271
4272         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4273         if (nla == NULL)
4274                 goto nla_put_failure;
4275         read_lock_bh(&idev->lock);
4276         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4277         read_unlock_bh(&idev->lock);
4278
4279         return 0;
4280
4281 nla_put_failure:
4282         return -EMSGSIZE;
4283 }
4284
4285 static size_t inet6_get_link_af_size(const struct net_device *dev)
4286 {
4287         if (!__in6_dev_get(dev))
4288                 return 0;
4289
4290         return inet6_ifla6_size();
4291 }
4292
4293 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4294 {
4295         struct inet6_dev *idev = __in6_dev_get(dev);
4296
4297         if (!idev)
4298                 return -ENODATA;
4299
4300         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4301                 return -EMSGSIZE;
4302
4303         return 0;
4304 }
4305
4306 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4307 {
4308         struct inet6_ifaddr *ifp;
4309         struct net_device *dev = idev->dev;
4310         bool update_rs = false;
4311         struct in6_addr ll_addr;
4312
4313         if (token == NULL)
4314                 return -EINVAL;
4315         if (ipv6_addr_any(token))
4316                 return -EINVAL;
4317         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4318                 return -EINVAL;
4319         if (!ipv6_accept_ra(idev))
4320                 return -EINVAL;
4321         if (idev->cnf.rtr_solicits <= 0)
4322                 return -EINVAL;
4323
4324         write_lock_bh(&idev->lock);
4325
4326         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4327         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4328
4329         write_unlock_bh(&idev->lock);
4330
4331         if (!idev->dead && (idev->if_flags & IF_READY) &&
4332             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4333                              IFA_F_OPTIMISTIC)) {
4334
4335                 /* If we're not ready, then normal ifup will take care
4336                  * of this. Otherwise, we need to request our rs here.
4337                  */
4338                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4339                 update_rs = true;
4340         }
4341
4342         write_lock_bh(&idev->lock);
4343
4344         if (update_rs) {
4345                 idev->if_flags |= IF_RS_SENT;
4346                 idev->rs_probes = 1;
4347                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4348         }
4349
4350         /* Well, that's kinda nasty ... */
4351         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4352                 spin_lock(&ifp->lock);
4353                 if (ifp->tokenized) {
4354                         ifp->valid_lft = 0;
4355                         ifp->prefered_lft = 0;
4356                 }
4357                 spin_unlock(&ifp->lock);
4358         }
4359
4360         write_unlock_bh(&idev->lock);
4361         addrconf_verify(0);
4362         return 0;
4363 }
4364
4365 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4366 {
4367         int err = -EINVAL;
4368         struct inet6_dev *idev = __in6_dev_get(dev);
4369         struct nlattr *tb[IFLA_INET6_MAX + 1];
4370
4371         if (!idev)
4372                 return -EAFNOSUPPORT;
4373
4374         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4375                 BUG();
4376
4377         if (tb[IFLA_INET6_TOKEN])
4378                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4379
4380         return err;
4381 }
4382
4383 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4384                              u32 portid, u32 seq, int event, unsigned int flags)
4385 {
4386         struct net_device *dev = idev->dev;
4387         struct ifinfomsg *hdr;
4388         struct nlmsghdr *nlh;
4389         void *protoinfo;
4390
4391         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4392         if (nlh == NULL)
4393                 return -EMSGSIZE;
4394
4395         hdr = nlmsg_data(nlh);
4396         hdr->ifi_family = AF_INET6;
4397         hdr->__ifi_pad = 0;
4398         hdr->ifi_type = dev->type;
4399         hdr->ifi_index = dev->ifindex;
4400         hdr->ifi_flags = dev_get_flags(dev);
4401         hdr->ifi_change = 0;
4402
4403         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4404             (dev->addr_len &&
4405              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4406             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4407             (dev->ifindex != dev->iflink &&
4408              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4409                 goto nla_put_failure;
4410         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4411         if (protoinfo == NULL)
4412                 goto nla_put_failure;
4413
4414         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4415                 goto nla_put_failure;
4416
4417         nla_nest_end(skb, protoinfo);
4418         return nlmsg_end(skb, nlh);
4419
4420 nla_put_failure:
4421         nlmsg_cancel(skb, nlh);
4422         return -EMSGSIZE;
4423 }
4424
4425 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4426 {
4427         struct net *net = sock_net(skb->sk);
4428         int h, s_h;
4429         int idx = 0, s_idx;
4430         struct net_device *dev;
4431         struct inet6_dev *idev;
4432         struct hlist_head *head;
4433
4434         s_h = cb->args[0];
4435         s_idx = cb->args[1];
4436
4437         rcu_read_lock();
4438         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4439                 idx = 0;
4440                 head = &net->dev_index_head[h];
4441                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4442                         if (idx < s_idx)
4443                                 goto cont;
4444                         idev = __in6_dev_get(dev);
4445                         if (!idev)
4446                                 goto cont;
4447                         if (inet6_fill_ifinfo(skb, idev,
4448                                               NETLINK_CB(cb->skb).portid,
4449                                               cb->nlh->nlmsg_seq,
4450                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4451                                 goto out;
4452 cont:
4453                         idx++;
4454                 }
4455         }
4456 out:
4457         rcu_read_unlock();
4458         cb->args[1] = idx;
4459         cb->args[0] = h;
4460
4461         return skb->len;
4462 }
4463
4464 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4465 {
4466         struct sk_buff *skb;
4467         struct net *net = dev_net(idev->dev);
4468         int err = -ENOBUFS;
4469
4470         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4471         if (skb == NULL)
4472                 goto errout;
4473
4474         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4475         if (err < 0) {
4476                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4477                 WARN_ON(err == -EMSGSIZE);
4478                 kfree_skb(skb);
4479                 goto errout;
4480         }
4481         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4482         return;
4483 errout:
4484         if (err < 0)
4485                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4486 }
4487
4488 static inline size_t inet6_prefix_nlmsg_size(void)
4489 {
4490         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4491                + nla_total_size(sizeof(struct in6_addr))
4492                + nla_total_size(sizeof(struct prefix_cacheinfo));
4493 }
4494
4495 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4496                              struct prefix_info *pinfo, u32 portid, u32 seq,
4497                              int event, unsigned int flags)
4498 {
4499         struct prefixmsg *pmsg;
4500         struct nlmsghdr *nlh;
4501         struct prefix_cacheinfo ci;
4502
4503         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4504         if (nlh == NULL)
4505                 return -EMSGSIZE;
4506
4507         pmsg = nlmsg_data(nlh);
4508         pmsg->prefix_family = AF_INET6;
4509         pmsg->prefix_pad1 = 0;
4510         pmsg->prefix_pad2 = 0;
4511         pmsg->prefix_ifindex = idev->dev->ifindex;
4512         pmsg->prefix_len = pinfo->prefix_len;
4513         pmsg->prefix_type = pinfo->type;
4514         pmsg->prefix_pad3 = 0;
4515         pmsg->prefix_flags = 0;
4516         if (pinfo->onlink)
4517                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4518         if (pinfo->autoconf)
4519                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4520
4521         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4522                 goto nla_put_failure;
4523         ci.preferred_time = ntohl(pinfo->prefered);
4524         ci.valid_time = ntohl(pinfo->valid);
4525         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4526                 goto nla_put_failure;
4527         return nlmsg_end(skb, nlh);
4528
4529 nla_put_failure:
4530         nlmsg_cancel(skb, nlh);
4531         return -EMSGSIZE;
4532 }
4533
4534 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4535                          struct prefix_info *pinfo)
4536 {
4537         struct sk_buff *skb;
4538         struct net *net = dev_net(idev->dev);
4539         int err = -ENOBUFS;
4540
4541         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4542         if (skb == NULL)
4543                 goto errout;
4544
4545         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4546         if (err < 0) {
4547                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4548                 WARN_ON(err == -EMSGSIZE);
4549                 kfree_skb(skb);
4550                 goto errout;
4551         }
4552         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4553         return;
4554 errout:
4555         if (err < 0)
4556                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4557 }
4558
4559 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4560 {
4561         write_lock_bh(&ifp->idev->lock);
4562         spin_lock(&ifp->lock);
4563         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4564                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4565             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4566                 ifp->idev->valid_ll_addr_cnt += count;
4567         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4568         spin_unlock(&ifp->lock);
4569         write_unlock_bh(&ifp->idev->lock);
4570 }
4571
4572 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4573 {
4574         struct net *net = dev_net(ifp->idev->dev);
4575
4576         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4577
4578         switch (event) {
4579         case RTM_NEWADDR:
4580                 update_valid_ll_addr_cnt(ifp, 1);
4581
4582                 /*
4583                  * If the address was optimistic
4584                  * we inserted the route at the start of
4585                  * our DAD process, so we don't need
4586                  * to do it again
4587                  */
4588                 if (!(ifp->rt->rt6i_node))
4589                         ip6_ins_rt(ifp->rt);
4590                 if (ifp->idev->cnf.forwarding)
4591                         addrconf_join_anycast(ifp);
4592                 if (!ipv6_addr_any(&ifp->peer_addr))
4593                         addrconf_prefix_route(&ifp->peer_addr, 128,
4594                                               ifp->idev->dev, 0, 0);
4595                 break;
4596         case RTM_DELADDR:
4597                 update_valid_ll_addr_cnt(ifp, -1);
4598
4599                 if (ifp->idev->cnf.forwarding)
4600                         addrconf_leave_anycast(ifp);
4601                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4602                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4603                         struct rt6_info *rt;
4604                         struct net_device *dev = ifp->idev->dev;
4605
4606                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4607                                         dev->ifindex, 1);
4608                         if (rt) {
4609                                 dst_hold(&rt->dst);
4610                                 if (ip6_del_rt(rt))
4611                                         dst_free(&rt->dst);
4612                         }
4613                 }
4614                 dst_hold(&ifp->rt->dst);
4615
4616                 if (ip6_del_rt(ifp->rt))
4617                         dst_free(&ifp->rt->dst);
4618                 break;
4619         }
4620         atomic_inc(&net->ipv6.dev_addr_genid);
4621         rt_genid_bump_ipv6(net);
4622 }
4623
4624 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4625 {
4626         rcu_read_lock_bh();
4627         if (likely(ifp->idev->dead == 0))
4628                 __ipv6_ifa_notify(event, ifp);
4629         rcu_read_unlock_bh();
4630 }
4631
4632 #ifdef CONFIG_SYSCTL
4633
4634 static
4635 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4636                            void __user *buffer, size_t *lenp, loff_t *ppos)
4637 {
4638         int *valp = ctl->data;
4639         int val = *valp;
4640         loff_t pos = *ppos;
4641         struct ctl_table lctl;
4642         int ret;
4643
4644         /*
4645          * ctl->data points to idev->cnf.forwarding, we should
4646          * not modify it until we get the rtnl lock.
4647          */
4648         lctl = *ctl;
4649         lctl.data = &val;
4650
4651         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4652
4653         if (write)
4654                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4655         if (ret)
4656                 *ppos = pos;
4657         return ret;
4658 }
4659
4660 static void dev_disable_change(struct inet6_dev *idev)
4661 {
4662         struct netdev_notifier_info info;
4663
4664         if (!idev || !idev->dev)
4665                 return;
4666
4667         netdev_notifier_info_init(&info, idev->dev);
4668         if (idev->cnf.disable_ipv6)
4669                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4670         else
4671                 addrconf_notify(NULL, NETDEV_UP, &info);
4672 }
4673
4674 static void addrconf_disable_change(struct net *net, __s32 newf)
4675 {
4676         struct net_device *dev;
4677         struct inet6_dev *idev;
4678
4679         rcu_read_lock();
4680         for_each_netdev_rcu(net, dev) {
4681                 idev = __in6_dev_get(dev);
4682                 if (idev) {
4683                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4684                         idev->cnf.disable_ipv6 = newf;
4685                         if (changed)
4686                                 dev_disable_change(idev);
4687                 }
4688         }
4689         rcu_read_unlock();
4690 }
4691
4692 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4693 {
4694         struct net *net;
4695         int old;
4696
4697         if (!rtnl_trylock())
4698                 return restart_syscall();
4699
4700         net = (struct net *)table->extra2;
4701         old = *p;
4702         *p = newf;
4703
4704         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4705                 rtnl_unlock();
4706                 return 0;
4707         }
4708
4709         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4710                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4711                 addrconf_disable_change(net, newf);
4712         } else if ((!newf) ^ (!old))
4713                 dev_disable_change((struct inet6_dev *)table->extra1);
4714
4715         rtnl_unlock();
4716         return 0;
4717 }
4718
4719 static
4720 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4721                             void __user *buffer, size_t *lenp, loff_t *ppos)
4722 {
4723         int *valp = ctl->data;
4724         int val = *valp;
4725         loff_t pos = *ppos;
4726         struct ctl_table lctl;
4727         int ret;
4728
4729         /*
4730          * ctl->data points to idev->cnf.disable_ipv6, we should
4731          * not modify it until we get the rtnl lock.
4732          */
4733         lctl = *ctl;
4734         lctl.data = &val;
4735
4736         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4737
4738         if (write)
4739                 ret = addrconf_disable_ipv6(ctl, valp, val);
4740         if (ret)
4741                 *ppos = pos;
4742         return ret;
4743 }
4744
4745 static struct addrconf_sysctl_table
4746 {
4747         struct ctl_table_header *sysctl_header;
4748         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4749 } addrconf_sysctl __read_mostly = {
4750         .sysctl_header = NULL,
4751         .addrconf_vars = {
4752                 {
4753                         .procname       = "forwarding",
4754                         .data           = &ipv6_devconf.forwarding,
4755                         .maxlen         = sizeof(int),
4756                         .mode           = 0644,
4757                         .proc_handler   = addrconf_sysctl_forward,
4758                 },
4759                 {
4760                         .procname       = "hop_limit",
4761                         .data           = &ipv6_devconf.hop_limit,
4762                         .maxlen         = sizeof(int),
4763                         .mode           = 0644,
4764                         .proc_handler   = proc_dointvec,
4765                 },
4766                 {
4767                         .procname       = "mtu",
4768                         .data           = &ipv6_devconf.mtu6,
4769                         .maxlen         = sizeof(int),
4770                         .mode           = 0644,
4771                         .proc_handler   = proc_dointvec,
4772                 },
4773                 {
4774                         .procname       = "accept_ra",
4775                         .data           = &ipv6_devconf.accept_ra,
4776                         .maxlen         = sizeof(int),
4777                         .mode           = 0644,
4778                         .proc_handler   = proc_dointvec,
4779                 },
4780                 {
4781                         .procname       = "accept_redirects",
4782                         .data           = &ipv6_devconf.accept_redirects,
4783                         .maxlen         = sizeof(int),
4784                         .mode           = 0644,
4785                         .proc_handler   = proc_dointvec,
4786                 },
4787                 {
4788                         .procname       = "autoconf",
4789                         .data           = &ipv6_devconf.autoconf,
4790                         .maxlen         = sizeof(int),
4791                         .mode           = 0644,
4792                         .proc_handler   = proc_dointvec,
4793                 },
4794                 {
4795                         .procname       = "dad_transmits",
4796                         .data           = &ipv6_devconf.dad_transmits,
4797                         .maxlen         = sizeof(int),
4798                         .mode           = 0644,
4799                         .proc_handler   = proc_dointvec,
4800                 },
4801                 {
4802                         .procname       = "router_solicitations",
4803                         .data           = &ipv6_devconf.rtr_solicits,
4804                         .maxlen         = sizeof(int),
4805                         .mode           = 0644,
4806                         .proc_handler   = proc_dointvec,
4807                 },
4808                 {
4809                         .procname       = "router_solicitation_interval",
4810                         .data           = &ipv6_devconf.rtr_solicit_interval,
4811                         .maxlen         = sizeof(int),
4812                         .mode           = 0644,
4813                         .proc_handler   = proc_dointvec_jiffies,
4814                 },
4815                 {
4816                         .procname       = "router_solicitation_delay",
4817                         .data           = &ipv6_devconf.rtr_solicit_delay,
4818                         .maxlen         = sizeof(int),
4819                         .mode           = 0644,
4820                         .proc_handler   = proc_dointvec_jiffies,
4821                 },
4822                 {
4823                         .procname       = "force_mld_version",
4824                         .data           = &ipv6_devconf.force_mld_version,
4825                         .maxlen         = sizeof(int),
4826                         .mode           = 0644,
4827                         .proc_handler   = proc_dointvec,
4828                 },
4829                 {
4830                         .procname       = "mldv1_unsolicited_report_interval",
4831                         .data           =
4832                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4833                         .maxlen         = sizeof(int),
4834                         .mode           = 0644,
4835                         .proc_handler   = proc_dointvec_ms_jiffies,
4836                 },
4837                 {
4838                         .procname       = "mldv2_unsolicited_report_interval",
4839                         .data           =
4840                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4841                         .maxlen         = sizeof(int),
4842                         .mode           = 0644,
4843                         .proc_handler   = proc_dointvec_ms_jiffies,
4844                 },
4845 #ifdef CONFIG_IPV6_PRIVACY
4846                 {
4847                         .procname       = "use_tempaddr",
4848                         .data           = &ipv6_devconf.use_tempaddr,
4849                         .maxlen         = sizeof(int),
4850                         .mode           = 0644,
4851                         .proc_handler   = proc_dointvec,
4852                 },
4853                 {
4854                         .procname       = "temp_valid_lft",
4855                         .data           = &ipv6_devconf.temp_valid_lft,
4856                         .maxlen         = sizeof(int),
4857                         .mode           = 0644,
4858                         .proc_handler   = proc_dointvec,
4859                 },
4860                 {
4861                         .procname       = "temp_prefered_lft",
4862                         .data           = &ipv6_devconf.temp_prefered_lft,
4863                         .maxlen         = sizeof(int),
4864                         .mode           = 0644,
4865                         .proc_handler   = proc_dointvec,
4866                 },
4867                 {
4868                         .procname       = "regen_max_retry",
4869                         .data           = &ipv6_devconf.regen_max_retry,
4870                         .maxlen         = sizeof(int),
4871                         .mode           = 0644,
4872                         .proc_handler   = proc_dointvec,
4873                 },
4874                 {
4875                         .procname       = "max_desync_factor",
4876                         .data           = &ipv6_devconf.max_desync_factor,
4877                         .maxlen         = sizeof(int),
4878                         .mode           = 0644,
4879                         .proc_handler   = proc_dointvec,
4880                 },
4881 #endif
4882                 {
4883                         .procname       = "max_addresses",
4884                         .data           = &ipv6_devconf.max_addresses,
4885                         .maxlen         = sizeof(int),
4886                         .mode           = 0644,
4887                         .proc_handler   = proc_dointvec,
4888                 },
4889                 {
4890                         .procname       = "accept_ra_defrtr",
4891                         .data           = &ipv6_devconf.accept_ra_defrtr,
4892                         .maxlen         = sizeof(int),
4893                         .mode           = 0644,
4894                         .proc_handler   = proc_dointvec,
4895                 },
4896                 {
4897                         .procname       = "accept_ra_pinfo",
4898                         .data           = &ipv6_devconf.accept_ra_pinfo,
4899                         .maxlen         = sizeof(int),
4900                         .mode           = 0644,
4901                         .proc_handler   = proc_dointvec,
4902                 },
4903 #ifdef CONFIG_IPV6_ROUTER_PREF
4904                 {
4905                         .procname       = "accept_ra_rtr_pref",
4906                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4907                         .maxlen         = sizeof(int),
4908                         .mode           = 0644,
4909                         .proc_handler   = proc_dointvec,
4910                 },
4911                 {
4912                         .procname       = "router_probe_interval",
4913                         .data           = &ipv6_devconf.rtr_probe_interval,
4914                         .maxlen         = sizeof(int),
4915                         .mode           = 0644,
4916                         .proc_handler   = proc_dointvec_jiffies,
4917                 },
4918 #ifdef CONFIG_IPV6_ROUTE_INFO
4919                 {
4920                         .procname       = "accept_ra_rt_info_max_plen",
4921                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4922                         .maxlen         = sizeof(int),
4923                         .mode           = 0644,
4924                         .proc_handler   = proc_dointvec,
4925                 },
4926 #endif
4927 #endif
4928                 {
4929                         .procname       = "proxy_ndp",
4930                         .data           = &ipv6_devconf.proxy_ndp,
4931                         .maxlen         = sizeof(int),
4932                         .mode           = 0644,
4933                         .proc_handler   = proc_dointvec,
4934                 },
4935                 {
4936                         .procname       = "accept_source_route",
4937                         .data           = &ipv6_devconf.accept_source_route,
4938                         .maxlen         = sizeof(int),
4939                         .mode           = 0644,
4940                         .proc_handler   = proc_dointvec,
4941                 },
4942 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4943                 {
4944                         .procname       = "optimistic_dad",
4945                         .data           = &ipv6_devconf.optimistic_dad,
4946                         .maxlen         = sizeof(int),
4947                         .mode           = 0644,
4948                         .proc_handler   = proc_dointvec,
4949
4950                 },
4951 #endif
4952 #ifdef CONFIG_IPV6_MROUTE
4953                 {
4954                         .procname       = "mc_forwarding",
4955                         .data           = &ipv6_devconf.mc_forwarding,
4956                         .maxlen         = sizeof(int),
4957                         .mode           = 0444,
4958                         .proc_handler   = proc_dointvec,
4959                 },
4960 #endif
4961                 {
4962                         .procname       = "disable_ipv6",
4963                         .data           = &ipv6_devconf.disable_ipv6,
4964                         .maxlen         = sizeof(int),
4965                         .mode           = 0644,
4966                         .proc_handler   = addrconf_sysctl_disable,
4967                 },
4968                 {
4969                         .procname       = "accept_dad",
4970                         .data           = &ipv6_devconf.accept_dad,
4971                         .maxlen         = sizeof(int),
4972                         .mode           = 0644,
4973                         .proc_handler   = proc_dointvec,
4974                 },
4975                 {
4976                         .procname       = "force_tllao",
4977                         .data           = &ipv6_devconf.force_tllao,
4978                         .maxlen         = sizeof(int),
4979                         .mode           = 0644,
4980                         .proc_handler   = proc_dointvec
4981                 },
4982                 {
4983                         .procname       = "ndisc_notify",
4984                         .data           = &ipv6_devconf.ndisc_notify,
4985                         .maxlen         = sizeof(int),
4986                         .mode           = 0644,
4987                         .proc_handler   = proc_dointvec
4988                 },
4989                 {
4990                         .procname       = "suppress_frag_ndisc",
4991                         .data           = &ipv6_devconf.suppress_frag_ndisc,
4992                         .maxlen         = sizeof(int),
4993                         .mode           = 0644,
4994                         .proc_handler   = proc_dointvec
4995                 },
4996                 {
4997                         /* sentinel */
4998                 }
4999         },
5000 };
5001
5002 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5003                 struct inet6_dev *idev, struct ipv6_devconf *p)
5004 {
5005         int i;
5006         struct addrconf_sysctl_table *t;
5007         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5008
5009         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5010         if (t == NULL)
5011                 goto out;
5012
5013         for (i = 0; t->addrconf_vars[i].data; i++) {
5014                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5015                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5016                 t->addrconf_vars[i].extra2 = net;
5017         }
5018
5019         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5020
5021         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5022         if (t->sysctl_header == NULL)
5023                 goto free;
5024
5025         p->sysctl = t;
5026         return 0;
5027
5028 free:
5029         kfree(t);
5030 out:
5031         return -ENOBUFS;
5032 }
5033
5034 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5035 {
5036         struct addrconf_sysctl_table *t;
5037
5038         if (p->sysctl == NULL)
5039                 return;
5040
5041         t = p->sysctl;
5042         p->sysctl = NULL;
5043         unregister_net_sysctl_table(t->sysctl_header);
5044         kfree(t);
5045 }
5046
5047 static void addrconf_sysctl_register(struct inet6_dev *idev)
5048 {
5049         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5050                               &ndisc_ifinfo_sysctl_change);
5051         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5052                                         idev, &idev->cnf);
5053 }
5054
5055 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5056 {
5057         __addrconf_sysctl_unregister(&idev->cnf);
5058         neigh_sysctl_unregister(idev->nd_parms);
5059 }
5060
5061
5062 #endif
5063
5064 static int __net_init addrconf_init_net(struct net *net)
5065 {
5066         int err = -ENOMEM;
5067         struct ipv6_devconf *all, *dflt;
5068
5069         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5070         if (all == NULL)
5071                 goto err_alloc_all;
5072
5073         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5074         if (dflt == NULL)
5075                 goto err_alloc_dflt;
5076
5077         /* these will be inherited by all namespaces */
5078         dflt->autoconf = ipv6_defaults.autoconf;
5079         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5080
5081         net->ipv6.devconf_all = all;
5082         net->ipv6.devconf_dflt = dflt;
5083
5084 #ifdef CONFIG_SYSCTL
5085         err = __addrconf_sysctl_register(net, "all", NULL, all);
5086         if (err < 0)
5087                 goto err_reg_all;
5088
5089         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5090         if (err < 0)
5091                 goto err_reg_dflt;
5092 #endif
5093         return 0;
5094
5095 #ifdef CONFIG_SYSCTL
5096 err_reg_dflt:
5097         __addrconf_sysctl_unregister(all);
5098 err_reg_all:
5099         kfree(dflt);
5100 #endif
5101 err_alloc_dflt:
5102         kfree(all);
5103 err_alloc_all:
5104         return err;
5105 }
5106
5107 static void __net_exit addrconf_exit_net(struct net *net)
5108 {
5109 #ifdef CONFIG_SYSCTL
5110         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5111         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5112 #endif
5113         if (!net_eq(net, &init_net)) {
5114                 kfree(net->ipv6.devconf_dflt);
5115                 kfree(net->ipv6.devconf_all);
5116         }
5117 }
5118
5119 static struct pernet_operations addrconf_ops = {
5120         .init = addrconf_init_net,
5121         .exit = addrconf_exit_net,
5122 };
5123
5124 static struct rtnl_af_ops inet6_ops = {
5125         .family           = AF_INET6,
5126         .fill_link_af     = inet6_fill_link_af,
5127         .get_link_af_size = inet6_get_link_af_size,
5128         .set_link_af      = inet6_set_link_af,
5129 };
5130
5131 /*
5132  *      Init / cleanup code
5133  */
5134
5135 int __init addrconf_init(void)
5136 {
5137         int i, err;
5138
5139         err = ipv6_addr_label_init();
5140         if (err < 0) {
5141                 pr_crit("%s: cannot initialize default policy table: %d\n",
5142                         __func__, err);
5143                 goto out;
5144         }
5145
5146         err = register_pernet_subsys(&addrconf_ops);
5147         if (err < 0)
5148                 goto out_addrlabel;
5149
5150         /* The addrconf netdev notifier requires that loopback_dev
5151          * has it's ipv6 private information allocated and setup
5152          * before it can bring up and give link-local addresses
5153          * to other devices which are up.
5154          *
5155          * Unfortunately, loopback_dev is not necessarily the first
5156          * entry in the global dev_base list of net devices.  In fact,
5157          * it is likely to be the very last entry on that list.
5158          * So this causes the notifier registry below to try and
5159          * give link-local addresses to all devices besides loopback_dev
5160          * first, then loopback_dev, which cases all the non-loopback_dev
5161          * devices to fail to get a link-local address.
5162          *
5163          * So, as a temporary fix, allocate the ipv6 structure for
5164          * loopback_dev first by hand.
5165          * Longer term, all of the dependencies ipv6 has upon the loopback
5166          * device and it being up should be removed.
5167          */
5168         rtnl_lock();
5169         if (!ipv6_add_dev(init_net.loopback_dev))
5170                 err = -ENOMEM;
5171         rtnl_unlock();
5172         if (err)
5173                 goto errlo;
5174
5175         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5176                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5177
5178         register_netdevice_notifier(&ipv6_dev_notf);
5179
5180         addrconf_verify(0);
5181
5182         err = rtnl_af_register(&inet6_ops);
5183         if (err < 0)
5184                 goto errout_af;
5185
5186         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5187                               NULL);
5188         if (err < 0)
5189                 goto errout;
5190
5191         /* Only the first call to __rtnl_register can fail */
5192         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5193         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5194         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5195                         inet6_dump_ifaddr, NULL);
5196         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5197                         inet6_dump_ifmcaddr, NULL);
5198         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5199                         inet6_dump_ifacaddr, NULL);
5200         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5201                         inet6_netconf_dump_devconf, NULL);
5202
5203         ipv6_addr_label_rtnl_register();
5204
5205         return 0;
5206 errout:
5207         rtnl_af_unregister(&inet6_ops);
5208 errout_af:
5209         unregister_netdevice_notifier(&ipv6_dev_notf);
5210 errlo:
5211         unregister_pernet_subsys(&addrconf_ops);
5212 out_addrlabel:
5213         ipv6_addr_label_cleanup();
5214 out:
5215         return err;
5216 }
5217
5218 void addrconf_cleanup(void)
5219 {
5220         struct net_device *dev;
5221         int i;
5222
5223         unregister_netdevice_notifier(&ipv6_dev_notf);
5224         unregister_pernet_subsys(&addrconf_ops);
5225         ipv6_addr_label_cleanup();
5226
5227         rtnl_lock();
5228
5229         __rtnl_af_unregister(&inet6_ops);
5230
5231         /* clean dev list */
5232         for_each_netdev(&init_net, dev) {
5233                 if (__in6_dev_get(dev) == NULL)
5234                         continue;
5235                 addrconf_ifdown(dev, 1);
5236         }
5237         addrconf_ifdown(init_net.loopback_dev, 2);
5238
5239         /*
5240          *      Check hash table.
5241          */
5242         spin_lock_bh(&addrconf_hash_lock);
5243         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5244                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5245         spin_unlock_bh(&addrconf_hash_lock);
5246
5247         del_timer(&addr_chk_timer);
5248         rtnl_unlock();
5249 }