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

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66 #include <linux/hash.h>
  67
  68 #include <net/net_namespace.h>
  69 #include <net/sock.h>
  70 #include <net/snmp.h>
  71
  72 #include <net/af_ieee802154.h>
  73 #include <net/firewire.h>
  74 #include <net/ipv6.h>
  75 #include <net/protocol.h>
  76 #include <net/ndisc.h>
  77 #include <net/ip6_route.h>
  78 #include <net/addrconf.h>
  79 #include <net/tcp.h>
  80 #include <net/ip.h>
  81 #include <net/netlink.h>
  82 #include <net/pkt_sched.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86
  87 #ifdef CONFIG_IPV6_PRIVACY
  88 #include <linux/random.h>
  89 #endif
  90
  91 #include <linux/uaccess.h>
  92 #include <asm/unaligned.h>
  93
  94 #include <linux/proc_fs.h>
  95 #include <linux/seq_file.h>
  96 #include <linux/export.h>
  97
  98 /* Set to 3 to get tracing... */
  99 #define ACONF_DEBUG 2
 100
 101 #if ACONF_DEBUG >= 3
 102 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 103 #else
 104 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 105 #endif
 106
 107 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 108
 109 static inline u32 cstamp_delta(unsigned long cstamp)
 110 {
 111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112 }
 113
 114 #ifdef CONFIG_SYSCTL
 115 static void addrconf_sysctl_register(struct inet6_dev *idev);
 116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117 #else
 118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 119 {
 120 }
 121
 122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 123 {
 124 }
 125 #endif
 126
 127 #ifdef CONFIG_IPV6_PRIVACY
 128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130 static void ipv6_regen_rndid(unsigned long data);
 131 #endif
 132
 133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134 static int ipv6_count_addresses(struct inet6_dev *idev);
 135
 136 /*
 137  *      Configured unicast address hash table
 138  */
 139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 140 static DEFINE_SPINLOCK(addrconf_hash_lock);
 141
 142 static void addrconf_verify(unsigned long);
 143
 144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 145 static DEFINE_SPINLOCK(addrconf_verify_lock);
 146
 147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 149
 150 static void addrconf_type_change(struct net_device *dev,
 151                                  unsigned long event);
 152 static int addrconf_ifdown(struct net_device *dev, int how);
 153
 154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 155                                                   int plen,
 156                                                   const struct net_device *dev,
 157                                                   u32 flags, u32 noflags);
 158
 159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 160 static void addrconf_dad_timer(unsigned long data);
 161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 162 static void addrconf_dad_run(struct inet6_dev *idev);
 163 static void addrconf_rs_timer(unsigned long data);
 164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 166
 167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 168                                 struct prefix_info *pinfo);
 169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 170                                struct net_device *dev);
 171
 172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 173         .forwarding             = 0,
 174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 175         .mtu6                   = IPV6_MIN_MTU,
 176         .accept_ra              = 1,
 177         .accept_redirects       = 1,
 178         .autoconf               = 1,
 179         .force_mld_version      = 0,
 180         .mldv1_unsolicited_report_interval = 10 * HZ,
 181         .mldv2_unsolicited_report_interval = HZ,
 182         .dad_transmits          = 1,
 183         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 184         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 185         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 186 #ifdef CONFIG_IPV6_PRIVACY
 187         .use_tempaddr           = 0,
 188         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 189         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 190         .regen_max_retry        = REGEN_MAX_RETRY,
 191         .max_desync_factor      = MAX_DESYNC_FACTOR,
 192 #endif
 193         .max_addresses          = IPV6_MAX_ADDRESSES,
 194         .accept_ra_defrtr       = 1,
 195         .accept_ra_pinfo        = 1,
 196 #ifdef CONFIG_IPV6_ROUTER_PREF
 197         .accept_ra_rtr_pref     = 1,
 198         .rtr_probe_interval     = 60 * HZ,
 199 #ifdef CONFIG_IPV6_ROUTE_INFO
 200         .accept_ra_rt_info_max_plen = 0,
 201 #endif
 202 #endif
 203         .proxy_ndp              = 0,
 204         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 205         .disable_ipv6           = 0,
 206         .accept_dad             = 1,
 207         .suppress_frag_ndisc    = 1,
 208 };
 209
 210 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 211         .forwarding             = 0,
 212         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 213         .mtu6                   = IPV6_MIN_MTU,
 214         .accept_ra              = 1,
 215         .accept_redirects       = 1,
 216         .autoconf               = 1,
 217         .force_mld_version      = 0,
 218         .mldv1_unsolicited_report_interval = 10 * HZ,
 219         .mldv2_unsolicited_report_interval = HZ,
 220         .dad_transmits          = 1,
 221         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 222         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 223         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 224 #ifdef CONFIG_IPV6_PRIVACY
 225         .use_tempaddr           = 0,
 226         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 227         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 228         .regen_max_retry        = REGEN_MAX_RETRY,
 229         .max_desync_factor      = MAX_DESYNC_FACTOR,
 230 #endif
 231         .max_addresses          = IPV6_MAX_ADDRESSES,
 232         .accept_ra_defrtr       = 1,
 233         .accept_ra_pinfo        = 1,
 234 #ifdef CONFIG_IPV6_ROUTER_PREF
 235         .accept_ra_rtr_pref     = 1,
 236         .rtr_probe_interval     = 60 * HZ,
 237 #ifdef CONFIG_IPV6_ROUTE_INFO
 238         .accept_ra_rt_info_max_plen = 0,
 239 #endif
 240 #endif
 241         .proxy_ndp              = 0,
 242         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 243         .disable_ipv6           = 0,
 244         .accept_dad             = 1,
 245         .suppress_frag_ndisc    = 1,
 246 };
 247
 248 /* Check if a valid qdisc is available */
 249 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 250 {
 251         return !qdisc_tx_is_noop(dev);
 252 }
 253
 254 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 255 {
 256         if (del_timer(&idev->rs_timer))
 257                 __in6_dev_put(idev);
 258 }
 259
 260 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
 261 {
 262         if (del_timer(&ifp->dad_timer))
 263                 __in6_ifa_put(ifp);
 264 }
 265
 266 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 267                                   unsigned long when)
 268 {
 269         if (!timer_pending(&idev->rs_timer))
 270                 in6_dev_hold(idev);
 271         mod_timer(&idev->rs_timer, jiffies + when);
 272 }
 273
 274 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
 275                                    unsigned long when)
 276 {
 277         if (!timer_pending(&ifp->dad_timer))
 278                 in6_ifa_hold(ifp);
 279         mod_timer(&ifp->dad_timer, jiffies + when);
 280 }
 281
 282 static int snmp6_alloc_dev(struct inet6_dev *idev)
 283 {
 284         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 285                           sizeof(struct ipstats_mib),
 286                           __alignof__(struct ipstats_mib)) < 0)
 287                 goto err_ip;
 288         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 289                                         GFP_KERNEL);
 290         if (!idev->stats.icmpv6dev)
 291                 goto err_icmp;
 292         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 293                                            GFP_KERNEL);
 294         if (!idev->stats.icmpv6msgdev)
 295                 goto err_icmpmsg;
 296
 297         return 0;
 298
 299 err_icmpmsg:
 300         kfree(idev->stats.icmpv6dev);
 301 err_icmp:
 302         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 303 err_ip:
 304         return -ENOMEM;
 305 }
 306
 307 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 308 {
 309         struct inet6_dev *ndev;
 310
 311         ASSERT_RTNL();
 312
 313         if (dev->mtu < IPV6_MIN_MTU)
 314                 return NULL;
 315
 316         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 317
 318         if (ndev == NULL)
 319                 return NULL;
 320
 321         rwlock_init(&ndev->lock);
 322         ndev->dev = dev;
 323         INIT_LIST_HEAD(&ndev->addr_list);
 324         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 325                     (unsigned long)ndev);
 326         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 327         ndev->cnf.mtu6 = dev->mtu;
 328         ndev->cnf.sysctl = NULL;
 329         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 330         if (ndev->nd_parms == NULL) {
 331                 kfree(ndev);
 332                 return NULL;
 333         }
 334         if (ndev->cnf.forwarding)
 335                 dev_disable_lro(dev);
 336         /* We refer to the device */
 337         dev_hold(dev);
 338
 339         if (snmp6_alloc_dev(ndev) < 0) {
 340                 ADBG(KERN_WARNING
 341                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 342                         __func__, dev->name);
 343                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 344                 dev_put(dev);
 345                 kfree(ndev);
 346                 return NULL;
 347         }
 348
 349         if (snmp6_register_dev(ndev) < 0) {
 350                 ADBG(KERN_WARNING
 351                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 352                         __func__, dev->name);
 353                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 354                 ndev->dead = 1;
 355                 in6_dev_finish_destroy(ndev);
 356                 return NULL;
 357         }
 358
 359         /* One reference from device.  We must do this before
 360          * we invoke __ipv6_regen_rndid().
 361          */
 362         in6_dev_hold(ndev);
 363
 364         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 365                 ndev->cnf.accept_dad = -1;
 366
 367 #if IS_ENABLED(CONFIG_IPV6_SIT)
 368         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 369                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 370                 ndev->cnf.rtr_solicits = 0;
 371         }
 372 #endif
 373
 374 #ifdef CONFIG_IPV6_PRIVACY
 375         INIT_LIST_HEAD(&ndev->tempaddr_list);
 376         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 377         if ((dev->flags&IFF_LOOPBACK) ||
 378             dev->type == ARPHRD_TUNNEL ||
 379             dev->type == ARPHRD_TUNNEL6 ||
 380             dev->type == ARPHRD_SIT ||
 381             dev->type == ARPHRD_NONE) {
 382                 ndev->cnf.use_tempaddr = -1;
 383         } else {
 384                 in6_dev_hold(ndev);
 385                 ipv6_regen_rndid((unsigned long) ndev);
 386         }
 387 #endif
 388         ndev->token = in6addr_any;
 389
 390         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 391                 ndev->if_flags |= IF_READY;
 392
 393         ipv6_mc_init_dev(ndev);
 394         ndev->tstamp = jiffies;
 395         addrconf_sysctl_register(ndev);
 396         /* protected by rtnl_lock */
 397         rcu_assign_pointer(dev->ip6_ptr, ndev);
 398
 399         /* Join interface-local all-node multicast group */
 400         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 401
 402         /* Join all-node multicast group */
 403         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 404
 405         /* Join all-router multicast group if forwarding is set */
 406         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 407                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 408
 409         return ndev;
 410 }
 411
 412 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 413 {
 414         struct inet6_dev *idev;
 415
 416         ASSERT_RTNL();
 417
 418         idev = __in6_dev_get(dev);
 419         if (!idev) {
 420                 idev = ipv6_add_dev(dev);
 421                 if (!idev)
 422                         return NULL;
 423         }
 424
 425         if (dev->flags&IFF_UP)
 426                 ipv6_mc_up(idev);
 427         return idev;
 428 }
 429
 430 static int inet6_netconf_msgsize_devconf(int type)
 431 {
 432         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 433                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 434
 435         /* type -1 is used for ALL */
 436         if (type == -1 || type == NETCONFA_FORWARDING)
 437                 size += nla_total_size(4);
 438 #ifdef CONFIG_IPV6_MROUTE
 439         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 440                 size += nla_total_size(4);
 441 #endif
 442
 443         return size;
 444 }
 445
 446 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 447                                       struct ipv6_devconf *devconf, u32 portid,
 448                                       u32 seq, int event, unsigned int flags,
 449                                       int type)
 450 {
 451         struct nlmsghdr  *nlh;
 452         struct netconfmsg *ncm;
 453
 454         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 455                         flags);
 456         if (nlh == NULL)
 457                 return -EMSGSIZE;
 458
 459         ncm = nlmsg_data(nlh);
 460         ncm->ncm_family = AF_INET6;
 461
 462         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 463                 goto nla_put_failure;
 464
 465         /* type -1 is used for ALL */
 466         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 467             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 468                 goto nla_put_failure;
 469 #ifdef CONFIG_IPV6_MROUTE
 470         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 471             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 472                         devconf->mc_forwarding) < 0)
 473                 goto nla_put_failure;
 474 #endif
 475         return nlmsg_end(skb, nlh);
 476
 477 nla_put_failure:
 478         nlmsg_cancel(skb, nlh);
 479         return -EMSGSIZE;
 480 }
 481
 482 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 483                                   struct ipv6_devconf *devconf)
 484 {
 485         struct sk_buff *skb;
 486         int err = -ENOBUFS;
 487
 488         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 489         if (skb == NULL)
 490                 goto errout;
 491
 492         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 493                                          RTM_NEWNETCONF, 0, type);
 494         if (err < 0) {
 495                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 496                 WARN_ON(err == -EMSGSIZE);
 497                 kfree_skb(skb);
 498                 goto errout;
 499         }
 500         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 501         return;
 502 errout:
 503         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 504 }
 505
 506 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 507         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 508         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 509 };
 510
 511 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 512                                      struct nlmsghdr *nlh)
 513 {
 514         struct net *net = sock_net(in_skb->sk);
 515         struct nlattr *tb[NETCONFA_MAX+1];
 516         struct netconfmsg *ncm;
 517         struct sk_buff *skb;
 518         struct ipv6_devconf *devconf;
 519         struct inet6_dev *in6_dev;
 520         struct net_device *dev;
 521         int ifindex;
 522         int err;
 523
 524         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 525                           devconf_ipv6_policy);
 526         if (err < 0)
 527                 goto errout;
 528
 529         err = EINVAL;
 530         if (!tb[NETCONFA_IFINDEX])
 531                 goto errout;
 532
 533         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 534         switch (ifindex) {
 535         case NETCONFA_IFINDEX_ALL:
 536                 devconf = net->ipv6.devconf_all;
 537                 break;
 538         case NETCONFA_IFINDEX_DEFAULT:
 539                 devconf = net->ipv6.devconf_dflt;
 540                 break;
 541         default:
 542                 dev = __dev_get_by_index(net, ifindex);
 543                 if (dev == NULL)
 544                         goto errout;
 545                 in6_dev = __in6_dev_get(dev);
 546                 if (in6_dev == NULL)
 547                         goto errout;
 548                 devconf = &in6_dev->cnf;
 549                 break;
 550         }
 551
 552         err = -ENOBUFS;
 553         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 554         if (skb == NULL)
 555                 goto errout;
 556
 557         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 558                                          NETLINK_CB(in_skb).portid,
 559                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 560                                          -1);
 561         if (err < 0) {
 562                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 563                 WARN_ON(err == -EMSGSIZE);
 564                 kfree_skb(skb);
 565                 goto errout;
 566         }
 567         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 568 errout:
 569         return err;
 570 }
 571
 572 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 573                                       struct netlink_callback *cb)
 574 {
 575         struct net *net = sock_net(skb->sk);
 576         int h, s_h;
 577         int idx, s_idx;
 578         struct net_device *dev;
 579         struct inet6_dev *idev;
 580         struct hlist_head *head;
 581
 582         s_h = cb->args[0];
 583         s_idx = idx = cb->args[1];
 584
 585         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 586                 idx = 0;
 587                 head = &net->dev_index_head[h];
 588                 rcu_read_lock();
 589                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 590                           net->dev_base_seq;
 591                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 592                         if (idx < s_idx)
 593                                 goto cont;
 594                         idev = __in6_dev_get(dev);
 595                         if (!idev)
 596                                 goto cont;
 597
 598                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 599                                                        &idev->cnf,
 600                                                        NETLINK_CB(cb->skb).portid,
 601                                                        cb->nlh->nlmsg_seq,
 602                                                        RTM_NEWNETCONF,
 603                                                        NLM_F_MULTI,
 604                                                        -1) <= 0) {
 605                                 rcu_read_unlock();
 606                                 goto done;
 607                         }
 608                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 609 cont:
 610                         idx++;
 611                 }
 612                 rcu_read_unlock();
 613         }
 614         if (h == NETDEV_HASHENTRIES) {
 615                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 616                                                net->ipv6.devconf_all,
 617                                                NETLINK_CB(cb->skb).portid,
 618                                                cb->nlh->nlmsg_seq,
 619                                                RTM_NEWNETCONF, NLM_F_MULTI,
 620                                                -1) <= 0)
 621                         goto done;
 622                 else
 623                         h++;
 624         }
 625         if (h == NETDEV_HASHENTRIES + 1) {
 626                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 627                                                net->ipv6.devconf_dflt,
 628                                                NETLINK_CB(cb->skb).portid,
 629                                                cb->nlh->nlmsg_seq,
 630                                                RTM_NEWNETCONF, NLM_F_MULTI,
 631                                                -1) <= 0)
 632                         goto done;
 633                 else
 634                         h++;
 635         }
 636 done:
 637         cb->args[0] = h;
 638         cb->args[1] = idx;
 639
 640         return skb->len;
 641 }
 642
 643 #ifdef CONFIG_SYSCTL
 644 static void dev_forward_change(struct inet6_dev *idev)
 645 {
 646         struct net_device *dev;
 647         struct inet6_ifaddr *ifa;
 648
 649         if (!idev)
 650                 return;
 651         dev = idev->dev;
 652         if (idev->cnf.forwarding)
 653                 dev_disable_lro(dev);
 654         if (dev->flags & IFF_MULTICAST) {
 655                 if (idev->cnf.forwarding) {
 656                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 657                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 658                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 659                 } else {
 660                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 661                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 662                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 663                 }
 664         }
 665
 666         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 667                 if (ifa->flags&IFA_F_TENTATIVE)
 668                         continue;
 669                 if (idev->cnf.forwarding)
 670                         addrconf_join_anycast(ifa);
 671                 else
 672                         addrconf_leave_anycast(ifa);
 673         }
 674         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 675                                      dev->ifindex, &idev->cnf);
 676 }
 677
 678
 679 static void addrconf_forward_change(struct net *net, __s32 newf)
 680 {
 681         struct net_device *dev;
 682         struct inet6_dev *idev;
 683
 684         for_each_netdev(net, dev) {
 685                 idev = __in6_dev_get(dev);
 686                 if (idev) {
 687                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 688                         idev->cnf.forwarding = newf;
 689                         if (changed)
 690                                 dev_forward_change(idev);
 691                 }
 692         }
 693 }
 694
 695 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 696 {
 697         struct net *net;
 698         int old;
 699
 700         if (!rtnl_trylock())
 701                 return restart_syscall();
 702
 703         net = (struct net *)table->extra2;
 704         old = *p;
 705         *p = newf;
 706
 707         if (p == &net->ipv6.devconf_dflt->forwarding) {
 708                 if ((!newf) ^ (!old))
 709                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 710                                                      NETCONFA_IFINDEX_DEFAULT,
 711                                                      net->ipv6.devconf_dflt);
 712                 rtnl_unlock();
 713                 return 0;
 714         }
 715
 716         if (p == &net->ipv6.devconf_all->forwarding) {
 717                 net->ipv6.devconf_dflt->forwarding = newf;
 718                 addrconf_forward_change(net, newf);
 719                 if ((!newf) ^ (!old))
 720                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 721                                                      NETCONFA_IFINDEX_ALL,
 722                                                      net->ipv6.devconf_all);
 723         } else if ((!newf) ^ (!old))
 724                 dev_forward_change((struct inet6_dev *)table->extra1);
 725         rtnl_unlock();
 726
 727         if (newf)
 728                 rt6_purge_dflt_routers(net);
 729         return 1;
 730 }
 731 #endif
 732
 733 /* Nobody refers to this ifaddr, destroy it */
 734 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 735 {
 736         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 737
 738 #ifdef NET_REFCNT_DEBUG
 739         pr_debug("%s\n", __func__);
 740 #endif
 741
 742         in6_dev_put(ifp->idev);
 743
 744         if (del_timer(&ifp->dad_timer))
 745                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 746
 747         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 748                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 749                 return;
 750         }
 751         ip6_rt_put(ifp->rt);
 752
 753         kfree_rcu(ifp, rcu);
 754 }
 755
 756 static void
 757 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 758 {
 759         struct list_head *p;
 760         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 761
 762         /*
 763          * Each device address list is sorted in order of scope -
 764          * global before linklocal.
 765          */
 766         list_for_each(p, &idev->addr_list) {
 767                 struct inet6_ifaddr *ifa
 768                         = list_entry(p, struct inet6_ifaddr, if_list);
 769                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 770                         break;
 771         }
 772
 773         list_add_tail(&ifp->if_list, p);
 774 }
 775
 776 static u32 inet6_addr_hash(const struct in6_addr *addr)
 777 {
 778         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 779 }
 780
 781 /* On success it returns ifp with increased reference count */
 782
 783 static struct inet6_ifaddr *
 784 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 785               const struct in6_addr *peer_addr, int pfxlen,
 786               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 787 {
 788         struct inet6_ifaddr *ifa = NULL;
 789         struct rt6_info *rt;
 790         unsigned int hash;
 791         int err = 0;
 792         int addr_type = ipv6_addr_type(addr);
 793
 794         if (addr_type == IPV6_ADDR_ANY ||
 795             addr_type & IPV6_ADDR_MULTICAST ||
 796             (!(idev->dev->flags & IFF_LOOPBACK) &&
 797              addr_type & IPV6_ADDR_LOOPBACK))
 798                 return ERR_PTR(-EADDRNOTAVAIL);
 799
 800         rcu_read_lock_bh();
 801         if (idev->dead) {
 802                 err = -ENODEV;                  /*XXX*/
 803                 goto out2;
 804         }
 805
 806         if (idev->cnf.disable_ipv6) {
 807                 err = -EACCES;
 808                 goto out2;
 809         }
 810
 811         spin_lock(&addrconf_hash_lock);
 812
 813         /* Ignore adding duplicate addresses on an interface */
 814         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 815                 ADBG("ipv6_add_addr: already assigned\n");
 816                 err = -EEXIST;
 817                 goto out;
 818         }
 819
 820         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 821
 822         if (ifa == NULL) {
 823                 ADBG("ipv6_add_addr: malloc failed\n");
 824                 err = -ENOBUFS;
 825                 goto out;
 826         }
 827
 828         rt = addrconf_dst_alloc(idev, addr, false);
 829         if (IS_ERR(rt)) {
 830                 err = PTR_ERR(rt);
 831                 goto out;
 832         }
 833
 834         ifa->addr = *addr;
 835         if (peer_addr)
 836                 ifa->peer_addr = *peer_addr;
 837
 838         spin_lock_init(&ifa->lock);
 839         spin_lock_init(&ifa->state_lock);
 840         setup_timer(&ifa->dad_timer, addrconf_dad_timer,
 841                     (unsigned long)ifa);
 842         INIT_HLIST_NODE(&ifa->addr_lst);
 843         ifa->scope = scope;
 844         ifa->prefix_len = pfxlen;
 845         ifa->flags = flags | IFA_F_TENTATIVE;
 846         ifa->valid_lft = valid_lft;
 847         ifa->prefered_lft = prefered_lft;
 848         ifa->cstamp = ifa->tstamp = jiffies;
 849         ifa->tokenized = false;
 850
 851         ifa->rt = rt;
 852
 853         ifa->idev = idev;
 854         in6_dev_hold(idev);
 855         /* For caller */
 856         in6_ifa_hold(ifa);
 857
 858         /* Add to big hash table */
 859         hash = inet6_addr_hash(addr);
 860
 861         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 862         spin_unlock(&addrconf_hash_lock);
 863
 864         write_lock(&idev->lock);
 865         /* Add to inet6_dev unicast addr list. */
 866         ipv6_link_dev_addr(idev, ifa);
 867
 868 #ifdef CONFIG_IPV6_PRIVACY
 869         if (ifa->flags&IFA_F_TEMPORARY) {
 870                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 871                 in6_ifa_hold(ifa);
 872         }
 873 #endif
 874
 875         in6_ifa_hold(ifa);
 876         write_unlock(&idev->lock);
 877 out2:
 878         rcu_read_unlock_bh();
 879
 880         if (likely(err == 0))
 881                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 882         else {
 883                 kfree(ifa);
 884                 ifa = ERR_PTR(err);
 885         }
 886
 887         return ifa;
 888 out:
 889         spin_unlock(&addrconf_hash_lock);
 890         goto out2;
 891 }
 892
 893 /* This function wants to get referenced ifp and releases it before return */
 894
 895 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 896 {
 897         struct inet6_ifaddr *ifa, *ifn;
 898         struct inet6_dev *idev = ifp->idev;
 899         int state;
 900         int deleted = 0, onlink = 0;
 901         unsigned long expires = jiffies;
 902
 903         spin_lock_bh(&ifp->state_lock);
 904         state = ifp->state;
 905         ifp->state = INET6_IFADDR_STATE_DEAD;
 906         spin_unlock_bh(&ifp->state_lock);
 907
 908         if (state == INET6_IFADDR_STATE_DEAD)
 909                 goto out;
 910
 911         spin_lock_bh(&addrconf_hash_lock);
 912         hlist_del_init_rcu(&ifp->addr_lst);
 913         spin_unlock_bh(&addrconf_hash_lock);
 914
 915         write_lock_bh(&idev->lock);
 916 #ifdef CONFIG_IPV6_PRIVACY
 917         if (ifp->flags&IFA_F_TEMPORARY) {
 918                 list_del(&ifp->tmp_list);
 919                 if (ifp->ifpub) {
 920                         in6_ifa_put(ifp->ifpub);
 921                         ifp->ifpub = NULL;
 922                 }
 923                 __in6_ifa_put(ifp);
 924         }
 925 #endif
 926
 927         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 928                 if (ifa == ifp) {
 929                         list_del_init(&ifp->if_list);
 930                         __in6_ifa_put(ifp);
 931
 932                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 933                                 break;
 934                         deleted = 1;
 935                         continue;
 936                 } else if (ifp->flags & IFA_F_PERMANENT) {
 937                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 938                                               ifp->prefix_len)) {
 939                                 if (ifa->flags & IFA_F_PERMANENT) {
 940                                         onlink = 1;
 941                                         if (deleted)
 942                                                 break;
 943                                 } else {
 944                                         unsigned long lifetime;
 945
 946                                         if (!onlink)
 947                                                 onlink = -1;
 948
 949                                         spin_lock(&ifa->lock);
 950
 951                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 952                                         /*
 953                                          * Note: Because this address is
 954                                          * not permanent, lifetime <
 955                                          * LONG_MAX / HZ here.
 956                                          */
 957                                         if (time_before(expires,
 958                                                         ifa->tstamp + lifetime * HZ))
 959                                                 expires = ifa->tstamp + lifetime * HZ;
 960                                         spin_unlock(&ifa->lock);
 961                                 }
 962                         }
 963                 }
 964         }
 965         write_unlock_bh(&idev->lock);
 966
 967         addrconf_del_dad_timer(ifp);
 968
 969         ipv6_ifa_notify(RTM_DELADDR, ifp);
 970
 971         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
 972
 973         /*
 974          * Purge or update corresponding prefix
 975          *
 976          * 1) we don't purge prefix here if address was not permanent.
 977          *    prefix is managed by its own lifetime.
 978          * 2) if there're no addresses, delete prefix.
 979          * 3) if there're still other permanent address(es),
 980          *    corresponding prefix is still permanent.
 981          * 4) otherwise, update prefix lifetime to the
 982          *    longest valid lifetime among the corresponding
 983          *    addresses on the device.
 984          *    Note: subsequent RA will update lifetime.
 985          *
 986          * --yoshfuji
 987          */
 988         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 989                 struct in6_addr prefix;
 990                 struct rt6_info *rt;
 991
 992                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 993
 994                 rt = addrconf_get_prefix_route(&prefix,
 995                                                ifp->prefix_len,
 996                                                ifp->idev->dev,
 997                                                0, RTF_GATEWAY | RTF_DEFAULT);
 998
 999                 if (rt) {
1000                         if (onlink == 0) {
1001                                 ip6_del_rt(rt);
1002                                 rt = NULL;
1003                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1004                                 rt6_set_expires(rt, expires);
1005                         }
1006                 }
1007                 ip6_rt_put(rt);
1008         }
1009
1010         /* clean up prefsrc entries */
1011         rt6_remove_prefsrc(ifp);
1012 out:
1013         in6_ifa_put(ifp);
1014 }
1015
1016 #ifdef CONFIG_IPV6_PRIVACY
1017 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1018 {
1019         struct inet6_dev *idev = ifp->idev;
1020         struct in6_addr addr, *tmpaddr;
1021         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1022         unsigned long regen_advance;
1023         int tmp_plen;
1024         int ret = 0;
1025         u32 addr_flags;
1026         unsigned long now = jiffies;
1027
1028         write_lock(&idev->lock);
1029         if (ift) {
1030                 spin_lock_bh(&ift->lock);
1031                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1032                 spin_unlock_bh(&ift->lock);
1033                 tmpaddr = &addr;
1034         } else {
1035                 tmpaddr = NULL;
1036         }
1037 retry:
1038         in6_dev_hold(idev);
1039         if (idev->cnf.use_tempaddr <= 0) {
1040                 write_unlock(&idev->lock);
1041                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1042                 in6_dev_put(idev);
1043                 ret = -1;
1044                 goto out;
1045         }
1046         spin_lock_bh(&ifp->lock);
1047         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1048                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1049                 spin_unlock_bh(&ifp->lock);
1050                 write_unlock(&idev->lock);
1051                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1052                         __func__);
1053                 in6_dev_put(idev);
1054                 ret = -1;
1055                 goto out;
1056         }
1057         in6_ifa_hold(ifp);
1058         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1059         __ipv6_try_regen_rndid(idev, tmpaddr);
1060         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1061         age = (now - ifp->tstamp) / HZ;
1062         tmp_valid_lft = min_t(__u32,
1063                               ifp->valid_lft,
1064                               idev->cnf.temp_valid_lft + age);
1065         tmp_prefered_lft = min_t(__u32,
1066                                  ifp->prefered_lft,
1067                                  idev->cnf.temp_prefered_lft + age -
1068                                  idev->cnf.max_desync_factor);
1069         tmp_plen = ifp->prefix_len;
1070         tmp_tstamp = ifp->tstamp;
1071         spin_unlock_bh(&ifp->lock);
1072
1073         regen_advance = idev->cnf.regen_max_retry *
1074                         idev->cnf.dad_transmits *
1075                         idev->nd_parms->retrans_time / HZ;
1076         write_unlock(&idev->lock);
1077
1078         /* A temporary address is created only if this calculated Preferred
1079          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1080          * an implementation must not create a temporary address with a zero
1081          * Preferred Lifetime.
1082          */
1083         if (tmp_prefered_lft <= regen_advance) {
1084                 in6_ifa_put(ifp);
1085                 in6_dev_put(idev);
1086                 ret = -1;
1087                 goto out;
1088         }
1089
1090         addr_flags = IFA_F_TEMPORARY;
1091         /* set in addrconf_prefix_rcv() */
1092         if (ifp->flags & IFA_F_OPTIMISTIC)
1093                 addr_flags |= IFA_F_OPTIMISTIC;
1094
1095         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1096                             ipv6_addr_scope(&addr), addr_flags,
1097                             tmp_valid_lft, tmp_prefered_lft);
1098         if (IS_ERR(ift)) {
1099                 in6_ifa_put(ifp);
1100                 in6_dev_put(idev);
1101                 pr_info("%s: retry temporary address regeneration\n", __func__);
1102                 tmpaddr = &addr;
1103                 write_lock(&idev->lock);
1104                 goto retry;
1105         }
1106
1107         spin_lock_bh(&ift->lock);
1108         ift->ifpub = ifp;
1109         ift->cstamp = now;
1110         ift->tstamp = tmp_tstamp;
1111         spin_unlock_bh(&ift->lock);
1112
1113         addrconf_dad_start(ift);
1114         in6_ifa_put(ift);
1115         in6_dev_put(idev);
1116 out:
1117         return ret;
1118 }
1119 #endif
1120
1121 /*
1122  *      Choose an appropriate source address (RFC3484)
1123  */
1124 enum {
1125         IPV6_SADDR_RULE_INIT = 0,
1126         IPV6_SADDR_RULE_LOCAL,
1127         IPV6_SADDR_RULE_SCOPE,
1128         IPV6_SADDR_RULE_PREFERRED,
1129 #ifdef CONFIG_IPV6_MIP6
1130         IPV6_SADDR_RULE_HOA,
1131 #endif
1132         IPV6_SADDR_RULE_OIF,
1133         IPV6_SADDR_RULE_LABEL,
1134 #ifdef CONFIG_IPV6_PRIVACY
1135         IPV6_SADDR_RULE_PRIVACY,
1136 #endif
1137         IPV6_SADDR_RULE_ORCHID,
1138         IPV6_SADDR_RULE_PREFIX,
1139         IPV6_SADDR_RULE_MAX
1140 };
1141
1142 struct ipv6_saddr_score {
1143         int                     rule;
1144         int                     addr_type;
1145         struct inet6_ifaddr     *ifa;
1146         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1147         int                     scopedist;
1148         int                     matchlen;
1149 };
1150
1151 struct ipv6_saddr_dst {
1152         const struct in6_addr *addr;
1153         int ifindex;
1154         int scope;
1155         int label;
1156         unsigned int prefs;
1157 };
1158
1159 static inline int ipv6_saddr_preferred(int type)
1160 {
1161         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1162                 return 1;
1163         return 0;
1164 }
1165
1166 static int ipv6_get_saddr_eval(struct net *net,
1167                                struct ipv6_saddr_score *score,
1168                                struct ipv6_saddr_dst *dst,
1169                                int i)
1170 {
1171         int ret;
1172
1173         if (i <= score->rule) {
1174                 switch (i) {
1175                 case IPV6_SADDR_RULE_SCOPE:
1176                         ret = score->scopedist;
1177                         break;
1178                 case IPV6_SADDR_RULE_PREFIX:
1179                         ret = score->matchlen;
1180                         break;
1181                 default:
1182                         ret = !!test_bit(i, score->scorebits);
1183                 }
1184                 goto out;
1185         }
1186
1187         switch (i) {
1188         case IPV6_SADDR_RULE_INIT:
1189                 /* Rule 0: remember if hiscore is not ready yet */
1190                 ret = !!score->ifa;
1191                 break;
1192         case IPV6_SADDR_RULE_LOCAL:
1193                 /* Rule 1: Prefer same address */
1194                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1195                 break;
1196         case IPV6_SADDR_RULE_SCOPE:
1197                 /* Rule 2: Prefer appropriate scope
1198                  *
1199                  *      ret
1200                  *       ^
1201                  *    -1 |  d 15
1202                  *    ---+--+-+---> scope
1203                  *       |
1204                  *       |             d is scope of the destination.
1205                  *  B-d  |  \
1206                  *       |   \      <- smaller scope is better if
1207                  *  B-15 |    \        if scope is enough for destinaion.
1208                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1209                  * d-C-1 | /
1210                  *       |/         <- greater is better
1211                  *   -C  /             if scope is not enough for destination.
1212                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1213                  *
1214                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1215                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1216                  * Assume B = 0 and we get C > 29.
1217                  */
1218                 ret = __ipv6_addr_src_scope(score->addr_type);
1219                 if (ret >= dst->scope)
1220                         ret = -ret;
1221                 else
1222                         ret -= 128;     /* 30 is enough */
1223                 score->scopedist = ret;
1224                 break;
1225         case IPV6_SADDR_RULE_PREFERRED:
1226                 /* Rule 3: Avoid deprecated and optimistic addresses */
1227                 ret = ipv6_saddr_preferred(score->addr_type) ||
1228                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1229                 break;
1230 #ifdef CONFIG_IPV6_MIP6
1231         case IPV6_SADDR_RULE_HOA:
1232             {
1233                 /* Rule 4: Prefer home address */
1234                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1235                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1236                 break;
1237             }
1238 #endif
1239         case IPV6_SADDR_RULE_OIF:
1240                 /* Rule 5: Prefer outgoing interface */
1241                 ret = (!dst->ifindex ||
1242                        dst->ifindex == score->ifa->idev->dev->ifindex);
1243                 break;
1244         case IPV6_SADDR_RULE_LABEL:
1245                 /* Rule 6: Prefer matching label */
1246                 ret = ipv6_addr_label(net,
1247                                       &score->ifa->addr, score->addr_type,
1248                                       score->ifa->idev->dev->ifindex) == dst->label;
1249                 break;
1250 #ifdef CONFIG_IPV6_PRIVACY
1251         case IPV6_SADDR_RULE_PRIVACY:
1252             {
1253                 /* Rule 7: Prefer public address
1254                  * Note: prefer temporary address if use_tempaddr >= 2
1255                  */
1256                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1257                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1258                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1259                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1260                 break;
1261             }
1262 #endif
1263         case IPV6_SADDR_RULE_ORCHID:
1264                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1265                  *          non-ORCHID vs non-ORCHID
1266                  */
1267                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1268                         ipv6_addr_orchid(dst->addr));
1269                 break;
1270         case IPV6_SADDR_RULE_PREFIX:
1271                 /* Rule 8: Use longest matching prefix */
1272                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1273                 if (ret > score->ifa->prefix_len)
1274                         ret = score->ifa->prefix_len;
1275                 score->matchlen = ret;
1276                 break;
1277         default:
1278                 ret = 0;
1279         }
1280
1281         if (ret)
1282                 __set_bit(i, score->scorebits);
1283         score->rule = i;
1284 out:
1285         return ret;
1286 }
1287
1288 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1289                        const struct in6_addr *daddr, unsigned int prefs,
1290                        struct in6_addr *saddr)
1291 {
1292         struct ipv6_saddr_score scores[2],
1293                                 *score = &scores[0], *hiscore = &scores[1];
1294         struct ipv6_saddr_dst dst;
1295         struct net_device *dev;
1296         int dst_type;
1297
1298         dst_type = __ipv6_addr_type(daddr);
1299         dst.addr = daddr;
1300         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1301         dst.scope = __ipv6_addr_src_scope(dst_type);
1302         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1303         dst.prefs = prefs;
1304
1305         hiscore->rule = -1;
1306         hiscore->ifa = NULL;
1307
1308         rcu_read_lock();
1309
1310         for_each_netdev_rcu(net, dev) {
1311                 struct inet6_dev *idev;
1312
1313                 /* Candidate Source Address (section 4)
1314                  *  - multicast and link-local destination address,
1315                  *    the set of candidate source address MUST only
1316                  *    include addresses assigned to interfaces
1317                  *    belonging to the same link as the outgoing
1318                  *    interface.
1319                  * (- For site-local destination addresses, the
1320                  *    set of candidate source addresses MUST only
1321                  *    include addresses assigned to interfaces
1322                  *    belonging to the same site as the outgoing
1323                  *    interface.)
1324                  */
1325                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1326                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1327                     dst.ifindex && dev->ifindex != dst.ifindex)
1328                         continue;
1329
1330                 idev = __in6_dev_get(dev);
1331                 if (!idev)
1332                         continue;
1333
1334                 read_lock_bh(&idev->lock);
1335                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1336                         int i;
1337
1338                         /*
1339                          * - Tentative Address (RFC2462 section 5.4)
1340                          *  - A tentative address is not considered
1341                          *    "assigned to an interface" in the traditional
1342                          *    sense, unless it is also flagged as optimistic.
1343                          * - Candidate Source Address (section 4)
1344                          *  - In any case, anycast addresses, multicast
1345                          *    addresses, and the unspecified address MUST
1346                          *    NOT be included in a candidate set.
1347                          */
1348                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1349                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1350                                 continue;
1351
1352                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1353
1354                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1355                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1356                                 LIMIT_NETDEBUG(KERN_DEBUG
1357                                                "ADDRCONF: unspecified / multicast address "
1358                                                "assigned as unicast address on %s",
1359                                                dev->name);
1360                                 continue;
1361                         }
1362
1363                         score->rule = -1;
1364                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1365
1366                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1367                                 int minihiscore, miniscore;
1368
1369                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1370                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1371
1372                                 if (minihiscore > miniscore) {
1373                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1374                                             score->scopedist > 0) {
1375                                                 /*
1376                                                  * special case:
1377                                                  * each remaining entry
1378                                                  * has too small (not enough)
1379                                                  * scope, because ifa entries
1380                                                  * are sorted by their scope
1381                                                  * values.
1382                                                  */
1383                                                 goto try_nextdev;
1384                                         }
1385                                         break;
1386                                 } else if (minihiscore < miniscore) {
1387                                         if (hiscore->ifa)
1388                                                 in6_ifa_put(hiscore->ifa);
1389
1390                                         in6_ifa_hold(score->ifa);
1391
1392                                         swap(hiscore, score);
1393
1394                                         /* restore our iterator */
1395                                         score->ifa = hiscore->ifa;
1396
1397                                         break;
1398                                 }
1399                         }
1400                 }
1401 try_nextdev:
1402                 read_unlock_bh(&idev->lock);
1403         }
1404         rcu_read_unlock();
1405
1406         if (!hiscore->ifa)
1407                 return -EADDRNOTAVAIL;
1408
1409         *saddr = hiscore->ifa->addr;
1410         in6_ifa_put(hiscore->ifa);
1411         return 0;
1412 }
1413 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1414
1415 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1416                       unsigned char banned_flags)
1417 {
1418         struct inet6_ifaddr *ifp;
1419         int err = -EADDRNOTAVAIL;
1420
1421         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1422                 if (ifp->scope == IFA_LINK &&
1423                     !(ifp->flags & banned_flags)) {
1424                         *addr = ifp->addr;
1425                         err = 0;
1426                         break;
1427                 }
1428         }
1429         return err;
1430 }
1431
1432 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1433                     unsigned char banned_flags)
1434 {
1435         struct inet6_dev *idev;
1436         int err = -EADDRNOTAVAIL;
1437
1438         rcu_read_lock();
1439         idev = __in6_dev_get(dev);
1440         if (idev) {
1441                 read_lock_bh(&idev->lock);
1442                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1443                 read_unlock_bh(&idev->lock);
1444         }
1445         rcu_read_unlock();
1446         return err;
1447 }
1448
1449 static int ipv6_count_addresses(struct inet6_dev *idev)
1450 {
1451         int cnt = 0;
1452         struct inet6_ifaddr *ifp;
1453
1454         read_lock_bh(&idev->lock);
1455         list_for_each_entry(ifp, &idev->addr_list, if_list)
1456                 cnt++;
1457         read_unlock_bh(&idev->lock);
1458         return cnt;
1459 }
1460
1461 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1462                   const struct net_device *dev, int strict)
1463 {
1464         struct inet6_ifaddr *ifp;
1465         unsigned int hash = inet6_addr_hash(addr);
1466
1467         rcu_read_lock_bh();
1468         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1469                 if (!net_eq(dev_net(ifp->idev->dev), net))
1470                         continue;
1471                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1472                     !(ifp->flags&IFA_F_TENTATIVE) &&
1473                     (dev == NULL || ifp->idev->dev == dev ||
1474                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1475                         rcu_read_unlock_bh();
1476                         return 1;
1477                 }
1478         }
1479
1480         rcu_read_unlock_bh();
1481         return 0;
1482 }
1483 EXPORT_SYMBOL(ipv6_chk_addr);
1484
1485 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1486                                struct net_device *dev)
1487 {
1488         unsigned int hash = inet6_addr_hash(addr);
1489         struct inet6_ifaddr *ifp;
1490
1491         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1492                 if (!net_eq(dev_net(ifp->idev->dev), net))
1493                         continue;
1494                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1495                         if (dev == NULL || ifp->idev->dev == dev)
1496                                 return true;
1497                 }
1498         }
1499         return false;
1500 }
1501
1502 /* Compares an address/prefix_len with addresses on device @dev.
1503  * If one is found it returns true.
1504  */
1505 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1506         const unsigned int prefix_len, struct net_device *dev)
1507 {
1508         struct inet6_dev *idev;
1509         struct inet6_ifaddr *ifa;
1510         bool ret = false;
1511
1512         rcu_read_lock();
1513         idev = __in6_dev_get(dev);
1514         if (idev) {
1515                 read_lock_bh(&idev->lock);
1516                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1517                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1518                         if (ret)
1519                                 break;
1520                 }
1521                 read_unlock_bh(&idev->lock);
1522         }
1523         rcu_read_unlock();
1524
1525         return ret;
1526 }
1527 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1528
1529 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1530 {
1531         struct inet6_dev *idev;
1532         struct inet6_ifaddr *ifa;
1533         int     onlink;
1534
1535         onlink = 0;
1536         rcu_read_lock();
1537         idev = __in6_dev_get(dev);
1538         if (idev) {
1539                 read_lock_bh(&idev->lock);
1540                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1541                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1542                                                    ifa->prefix_len);
1543                         if (onlink)
1544                                 break;
1545                 }
1546                 read_unlock_bh(&idev->lock);
1547         }
1548         rcu_read_unlock();
1549         return onlink;
1550 }
1551 EXPORT_SYMBOL(ipv6_chk_prefix);
1552
1553 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1554                                      struct net_device *dev, int strict)
1555 {
1556         struct inet6_ifaddr *ifp, *result = NULL;
1557         unsigned int hash = inet6_addr_hash(addr);
1558
1559         rcu_read_lock_bh();
1560         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1561                 if (!net_eq(dev_net(ifp->idev->dev), net))
1562                         continue;
1563                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1564                         if (dev == NULL || ifp->idev->dev == dev ||
1565                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1566                                 result = ifp;
1567                                 in6_ifa_hold(ifp);
1568                                 break;
1569                         }
1570                 }
1571         }
1572         rcu_read_unlock_bh();
1573
1574         return result;
1575 }
1576
1577 /* Gets referenced address, destroys ifaddr */
1578
1579 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1580 {
1581         if (ifp->flags&IFA_F_PERMANENT) {
1582                 spin_lock_bh(&ifp->lock);
1583                 addrconf_del_dad_timer(ifp);
1584                 ifp->flags |= IFA_F_TENTATIVE;
1585                 if (dad_failed)
1586                         ifp->flags |= IFA_F_DADFAILED;
1587                 spin_unlock_bh(&ifp->lock);
1588                 if (dad_failed)
1589                         ipv6_ifa_notify(0, ifp);
1590                 in6_ifa_put(ifp);
1591 #ifdef CONFIG_IPV6_PRIVACY
1592         } else if (ifp->flags&IFA_F_TEMPORARY) {
1593                 struct inet6_ifaddr *ifpub;
1594                 spin_lock_bh(&ifp->lock);
1595                 ifpub = ifp->ifpub;
1596                 if (ifpub) {
1597                         in6_ifa_hold(ifpub);
1598                         spin_unlock_bh(&ifp->lock);
1599                         ipv6_create_tempaddr(ifpub, ifp);
1600                         in6_ifa_put(ifpub);
1601                 } else {
1602                         spin_unlock_bh(&ifp->lock);
1603                 }
1604                 ipv6_del_addr(ifp);
1605 #endif
1606         } else
1607                 ipv6_del_addr(ifp);
1608 }
1609
1610 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1611 {
1612         int err = -ENOENT;
1613
1614         spin_lock(&ifp->state_lock);
1615         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1616                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1617                 err = 0;
1618         }
1619         spin_unlock(&ifp->state_lock);
1620
1621         return err;
1622 }
1623
1624 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1625 {
1626         struct inet6_dev *idev = ifp->idev;
1627
1628         if (addrconf_dad_end(ifp)) {
1629                 in6_ifa_put(ifp);
1630                 return;
1631         }
1632
1633         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1634                              ifp->idev->dev->name, &ifp->addr);
1635
1636         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1637                 struct in6_addr addr;
1638
1639                 addr.s6_addr32[0] = htonl(0xfe800000);
1640                 addr.s6_addr32[1] = 0;
1641
1642                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1643                     ipv6_addr_equal(&ifp->addr, &addr)) {
1644                         /* DAD failed for link-local based on MAC address */
1645                         idev->cnf.disable_ipv6 = 1;
1646
1647                         pr_info("%s: IPv6 being disabled!\n",
1648                                 ifp->idev->dev->name);
1649                 }
1650         }
1651
1652         addrconf_dad_stop(ifp, 1);
1653 }
1654
1655 /* Join to solicited addr multicast group. */
1656
1657 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1658 {
1659         struct in6_addr maddr;
1660
1661         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1662                 return;
1663
1664         addrconf_addr_solict_mult(addr, &maddr);
1665         ipv6_dev_mc_inc(dev, &maddr);
1666 }
1667
1668 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1669 {
1670         struct in6_addr maddr;
1671
1672         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1673                 return;
1674
1675         addrconf_addr_solict_mult(addr, &maddr);
1676         __ipv6_dev_mc_dec(idev, &maddr);
1677 }
1678
1679 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1680 {
1681         struct in6_addr addr;
1682         if (ifp->prefix_len == 127) /* RFC 6164 */
1683                 return;
1684         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1685         if (ipv6_addr_any(&addr))
1686                 return;
1687         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1688 }
1689
1690 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1691 {
1692         struct in6_addr addr;
1693         if (ifp->prefix_len == 127) /* RFC 6164 */
1694                 return;
1695         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1696         if (ipv6_addr_any(&addr))
1697                 return;
1698         __ipv6_dev_ac_dec(ifp->idev, &addr);
1699 }
1700
1701 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1702 {
1703         if (dev->addr_len != ETH_ALEN)
1704                 return -1;
1705         memcpy(eui, dev->dev_addr, 3);
1706         memcpy(eui + 5, dev->dev_addr + 3, 3);
1707
1708         /*
1709          * The zSeries OSA network cards can be shared among various
1710          * OS instances, but the OSA cards have only one MAC address.
1711          * This leads to duplicate address conflicts in conjunction
1712          * with IPv6 if more than one instance uses the same card.
1713          *
1714          * The driver for these cards can deliver a unique 16-bit
1715          * identifier for each instance sharing the same card.  It is
1716          * placed instead of 0xFFFE in the interface identifier.  The
1717          * "u" bit of the interface identifier is not inverted in this
1718          * case.  Hence the resulting interface identifier has local
1719          * scope according to RFC2373.
1720          */
1721         if (dev->dev_id) {
1722                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1723                 eui[4] = dev->dev_id & 0xFF;
1724         } else {
1725                 eui[3] = 0xFF;
1726                 eui[4] = 0xFE;
1727                 eui[0] ^= 2;
1728         }
1729         return 0;
1730 }
1731
1732 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1733 {
1734         if (dev->addr_len != IEEE802154_ADDR_LEN)
1735                 return -1;
1736         memcpy(eui, dev->dev_addr, 8);
1737         eui[0] ^= 2;
1738         return 0;
1739 }
1740
1741 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1742 {
1743         union fwnet_hwaddr *ha;
1744
1745         if (dev->addr_len != FWNET_ALEN)
1746                 return -1;
1747
1748         ha = (union fwnet_hwaddr *)dev->dev_addr;
1749
1750         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1751         eui[0] ^= 2;
1752         return 0;
1753 }
1754
1755 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1756 {
1757         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1758         if (dev->addr_len != ARCNET_ALEN)
1759                 return -1;
1760         memset(eui, 0, 7);
1761         eui[7] = *(u8 *)dev->dev_addr;
1762         return 0;
1763 }
1764
1765 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1766 {
1767         if (dev->addr_len != INFINIBAND_ALEN)
1768                 return -1;
1769         memcpy(eui, dev->dev_addr + 12, 8);
1770         eui[0] |= 2;
1771         return 0;
1772 }
1773
1774 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1775 {
1776         if (addr == 0)
1777                 return -1;
1778         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1779                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1780                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1781                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1782                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1783                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1784         eui[1] = 0;
1785         eui[2] = 0x5E;
1786         eui[3] = 0xFE;
1787         memcpy(eui + 4, &addr, 4);
1788         return 0;
1789 }
1790
1791 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1792 {
1793         if (dev->priv_flags & IFF_ISATAP)
1794                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1795         return -1;
1796 }
1797
1798 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1799 {
1800         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1801 }
1802
1803 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1804 {
1805         memcpy(eui, dev->perm_addr, 3);
1806         memcpy(eui + 5, dev->perm_addr + 3, 3);
1807         eui[3] = 0xFF;
1808         eui[4] = 0xFE;
1809         eui[0] ^= 2;
1810         return 0;
1811 }
1812
1813 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1814 {
1815         switch (dev->type) {
1816         case ARPHRD_ETHER:
1817         case ARPHRD_FDDI:
1818                 return addrconf_ifid_eui48(eui, dev);
1819         case ARPHRD_ARCNET:
1820                 return addrconf_ifid_arcnet(eui, dev);
1821         case ARPHRD_INFINIBAND:
1822                 return addrconf_ifid_infiniband(eui, dev);
1823         case ARPHRD_SIT:
1824                 return addrconf_ifid_sit(eui, dev);
1825         case ARPHRD_IPGRE:
1826                 return addrconf_ifid_gre(eui, dev);
1827         case ARPHRD_IEEE802154:
1828                 return addrconf_ifid_eui64(eui, dev);
1829         case ARPHRD_IEEE1394:
1830                 return addrconf_ifid_ieee1394(eui, dev);
1831         case ARPHRD_TUNNEL6:
1832                 return addrconf_ifid_ip6tnl(eui, dev);
1833         }
1834         return -1;
1835 }
1836
1837 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1838 {
1839         int err = -1;
1840         struct inet6_ifaddr *ifp;
1841
1842         read_lock_bh(&idev->lock);
1843         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1844                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1845                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1846                         err = 0;
1847                         break;
1848                 }
1849         }
1850         read_unlock_bh(&idev->lock);
1851         return err;
1852 }
1853
1854 #ifdef CONFIG_IPV6_PRIVACY
1855 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1856 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1857 {
1858 regen:
1859         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1860         idev->rndid[0] &= ~0x02;
1861
1862         /*
1863          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1864          * check if generated address is not inappropriate
1865          *
1866          *  - Reserved subnet anycast (RFC 2526)
1867          *      11111101 11....11 1xxxxxxx
1868          *  - ISATAP (RFC4214) 6.1
1869          *      00-00-5E-FE-xx-xx-xx-xx
1870          *  - value 0
1871          *  - XXX: already assigned to an address on the device
1872          */
1873         if (idev->rndid[0] == 0xfd &&
1874             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1875             (idev->rndid[7]&0x80))
1876                 goto regen;
1877         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1878                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1879                         goto regen;
1880                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1881                         goto regen;
1882         }
1883 }
1884
1885 static void ipv6_regen_rndid(unsigned long data)
1886 {
1887         struct inet6_dev *idev = (struct inet6_dev *) data;
1888         unsigned long expires;
1889
1890         rcu_read_lock_bh();
1891         write_lock_bh(&idev->lock);
1892
1893         if (idev->dead)
1894                 goto out;
1895
1896         __ipv6_regen_rndid(idev);
1897
1898         expires = jiffies +
1899                 idev->cnf.temp_prefered_lft * HZ -
1900                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1901                 idev->cnf.max_desync_factor * HZ;
1902         if (time_before(expires, jiffies)) {
1903                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1904                         __func__, idev->dev->name);
1905                 goto out;
1906         }
1907
1908         if (!mod_timer(&idev->regen_timer, expires))
1909                 in6_dev_hold(idev);
1910
1911 out:
1912         write_unlock_bh(&idev->lock);
1913         rcu_read_unlock_bh();
1914         in6_dev_put(idev);
1915 }
1916
1917 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1918 {
1919         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1920                 __ipv6_regen_rndid(idev);
1921 }
1922 #endif
1923
1924 /*
1925  *      Add prefix route.
1926  */
1927
1928 static void
1929 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1930                       unsigned long expires, u32 flags)
1931 {
1932         struct fib6_config cfg = {
1933                 .fc_table = RT6_TABLE_PREFIX,
1934                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1935                 .fc_ifindex = dev->ifindex,
1936                 .fc_expires = expires,
1937                 .fc_dst_len = plen,
1938                 .fc_flags = RTF_UP | flags,
1939                 .fc_nlinfo.nl_net = dev_net(dev),
1940                 .fc_protocol = RTPROT_KERNEL,
1941         };
1942
1943         cfg.fc_dst = *pfx;
1944
1945         /* Prevent useless cloning on PtP SIT.
1946            This thing is done here expecting that the whole
1947            class of non-broadcast devices need not cloning.
1948          */
1949 #if IS_ENABLED(CONFIG_IPV6_SIT)
1950         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1951                 cfg.fc_flags |= RTF_NONEXTHOP;
1952 #endif
1953
1954         ip6_route_add(&cfg);
1955 }
1956
1957
1958 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1959                                                   int plen,
1960                                                   const struct net_device *dev,
1961                                                   u32 flags, u32 noflags)
1962 {
1963         struct fib6_node *fn;
1964         struct rt6_info *rt = NULL;
1965         struct fib6_table *table;
1966
1967         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1968         if (table == NULL)
1969                 return NULL;
1970
1971         read_lock_bh(&table->tb6_lock);
1972         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1973         if (!fn)
1974                 goto out;
1975         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1976                 if (rt->dst.dev->ifindex != dev->ifindex)
1977                         continue;
1978                 if ((rt->rt6i_flags & flags) != flags)
1979                         continue;
1980                 if ((rt->rt6i_flags & noflags) != 0)
1981                         continue;
1982                 dst_hold(&rt->dst);
1983                 break;
1984         }
1985 out:
1986         read_unlock_bh(&table->tb6_lock);
1987         return rt;
1988 }
1989
1990
1991 /* Create "default" multicast route to the interface */
1992
1993 static void addrconf_add_mroute(struct net_device *dev)
1994 {
1995         struct fib6_config cfg = {
1996                 .fc_table = RT6_TABLE_LOCAL,
1997                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1998                 .fc_ifindex = dev->ifindex,
1999                 .fc_dst_len = 8,
2000                 .fc_flags = RTF_UP,
2001                 .fc_nlinfo.nl_net = dev_net(dev),
2002         };
2003
2004         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2005
2006         ip6_route_add(&cfg);
2007 }
2008
2009 #if IS_ENABLED(CONFIG_IPV6_SIT)
2010 static void sit_route_add(struct net_device *dev)
2011 {
2012         struct fib6_config cfg = {
2013                 .fc_table = RT6_TABLE_MAIN,
2014                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2015                 .fc_ifindex = dev->ifindex,
2016                 .fc_dst_len = 96,
2017                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2018                 .fc_nlinfo.nl_net = dev_net(dev),
2019         };
2020
2021         /* prefix length - 96 bits "::d.d.d.d" */
2022         ip6_route_add(&cfg);
2023 }
2024 #endif
2025
2026 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2027 {
2028         struct inet6_dev *idev;
2029
2030         ASSERT_RTNL();
2031
2032         idev = ipv6_find_idev(dev);
2033         if (!idev)
2034                 return ERR_PTR(-ENOBUFS);
2035
2036         if (idev->cnf.disable_ipv6)
2037                 return ERR_PTR(-EACCES);
2038
2039         /* Add default multicast route */
2040         if (!(dev->flags & IFF_LOOPBACK))
2041                 addrconf_add_mroute(dev);
2042
2043         return idev;
2044 }
2045
2046 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2047 {
2048         struct prefix_info *pinfo;
2049         __u32 valid_lft;
2050         __u32 prefered_lft;
2051         int addr_type;
2052         struct inet6_dev *in6_dev;
2053         struct net *net = dev_net(dev);
2054
2055         pinfo = (struct prefix_info *) opt;
2056
2057         if (len < sizeof(struct prefix_info)) {
2058                 ADBG("addrconf: prefix option too short\n");
2059                 return;
2060         }
2061
2062         /*
2063          *      Validation checks ([ADDRCONF], page 19)
2064          */
2065
2066         addr_type = ipv6_addr_type(&pinfo->prefix);
2067
2068         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2069                 return;
2070
2071         valid_lft = ntohl(pinfo->valid);
2072         prefered_lft = ntohl(pinfo->prefered);
2073
2074         if (prefered_lft > valid_lft) {
2075                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2076                 return;
2077         }
2078
2079         in6_dev = in6_dev_get(dev);
2080
2081         if (in6_dev == NULL) {
2082                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2083                                     dev->name);
2084                 return;
2085         }
2086
2087         /*
2088          *      Two things going on here:
2089          *      1) Add routes for on-link prefixes
2090          *      2) Configure prefixes with the auto flag set
2091          */
2092
2093         if (pinfo->onlink) {
2094                 struct rt6_info *rt;
2095                 unsigned long rt_expires;
2096
2097                 /* Avoid arithmetic overflow. Really, we could
2098                  * save rt_expires in seconds, likely valid_lft,
2099                  * but it would require division in fib gc, that it
2100                  * not good.
2101                  */
2102                 if (HZ > USER_HZ)
2103                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2104                 else
2105                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2106
2107                 if (addrconf_finite_timeout(rt_expires))
2108                         rt_expires *= HZ;
2109
2110                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2111                                                pinfo->prefix_len,
2112                                                dev,
2113                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2114                                                RTF_GATEWAY | RTF_DEFAULT);
2115
2116                 if (rt) {
2117                         /* Autoconf prefix route */
2118                         if (valid_lft == 0) {
2119                                 ip6_del_rt(rt);
2120                                 rt = NULL;
2121                         } else if (addrconf_finite_timeout(rt_expires)) {
2122                                 /* not infinity */
2123                                 rt6_set_expires(rt, jiffies + rt_expires);
2124                         } else {
2125                                 rt6_clean_expires(rt);
2126                         }
2127                 } else if (valid_lft) {
2128                         clock_t expires = 0;
2129                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2130                         if (addrconf_finite_timeout(rt_expires)) {
2131                                 /* not infinity */
2132                                 flags |= RTF_EXPIRES;
2133                                 expires = jiffies_to_clock_t(rt_expires);
2134                         }
2135                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2136                                               dev, expires, flags);
2137                 }
2138                 ip6_rt_put(rt);
2139         }
2140
2141         /* Try to figure out our local address for this prefix */
2142
2143         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2144                 struct inet6_ifaddr *ifp;
2145                 struct in6_addr addr;
2146                 int create = 0, update_lft = 0;
2147                 bool tokenized = false;
2148
2149                 if (pinfo->prefix_len == 64) {
2150                         memcpy(&addr, &pinfo->prefix, 8);
2151
2152                         if (!ipv6_addr_any(&in6_dev->token)) {
2153                                 read_lock_bh(&in6_dev->lock);
2154                                 memcpy(addr.s6_addr + 8,
2155                                        in6_dev->token.s6_addr + 8, 8);
2156                                 read_unlock_bh(&in6_dev->lock);
2157                                 tokenized = true;
2158                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2159                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2160                                 in6_dev_put(in6_dev);
2161                                 return;
2162                         }
2163                         goto ok;
2164                 }
2165                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2166                                     pinfo->prefix_len);
2167                 in6_dev_put(in6_dev);
2168                 return;
2169
2170 ok:
2171
2172                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2173
2174                 if (ifp == NULL && valid_lft) {
2175                         int max_addresses = in6_dev->cnf.max_addresses;
2176                         u32 addr_flags = 0;
2177
2178 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2179                         if (in6_dev->cnf.optimistic_dad &&
2180                             !net->ipv6.devconf_all->forwarding && sllao)
2181                                 addr_flags = IFA_F_OPTIMISTIC;
2182 #endif
2183
2184                         /* Do not allow to create too much of autoconfigured
2185                          * addresses; this would be too easy way to crash kernel.
2186                          */
2187                         if (!max_addresses ||
2188                             ipv6_count_addresses(in6_dev) < max_addresses)
2189                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2190                                                     pinfo->prefix_len,
2191                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2192                                                     addr_flags, valid_lft,
2193                                                     prefered_lft);
2194
2195                         if (IS_ERR_OR_NULL(ifp)) {
2196                                 in6_dev_put(in6_dev);
2197                                 return;
2198                         }
2199
2200                         update_lft = 0;
2201                         create = 1;
2202                         ifp->cstamp = jiffies;
2203                         ifp->tokenized = tokenized;
2204                         addrconf_dad_start(ifp);
2205                 }
2206
2207                 if (ifp) {
2208                         int flags;
2209                         unsigned long now;
2210 #ifdef CONFIG_IPV6_PRIVACY
2211                         struct inet6_ifaddr *ift;
2212 #endif
2213                         u32 stored_lft;
2214
2215                         /* update lifetime (RFC2462 5.5.3 e) */
2216                         spin_lock(&ifp->lock);
2217                         now = jiffies;
2218                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2219                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2220                         else
2221                                 stored_lft = 0;
2222                         if (!update_lft && !create && stored_lft) {
2223                                 if (valid_lft > MIN_VALID_LIFETIME ||
2224                                     valid_lft > stored_lft)
2225                                         update_lft = 1;
2226                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2227                                         /* valid_lft <= stored_lft is always true */
2228                                         /*
2229                                          * RFC 4862 Section 5.5.3e:
2230                                          * "Note that the preferred lifetime of
2231                                          *  the corresponding address is always
2232                                          *  reset to the Preferred Lifetime in
2233                                          *  the received Prefix Information
2234                                          *  option, regardless of whether the
2235                                          *  valid lifetime is also reset or
2236                                          *  ignored."
2237                                          *
2238                                          *  So if the preferred lifetime in
2239                                          *  this advertisement is different
2240                                          *  than what we have stored, but the
2241                                          *  valid lifetime is invalid, just
2242                                          *  reset prefered_lft.
2243                                          *
2244                                          *  We must set the valid lifetime
2245                                          *  to the stored lifetime since we'll
2246                                          *  be updating the timestamp below,
2247                                          *  else we'll set it back to the
2248                                          *  minimum.
2249                                          */
2250                                         if (prefered_lft != ifp->prefered_lft) {
2251                                                 valid_lft = stored_lft;
2252                                                 update_lft = 1;
2253                                         }
2254                                 } else {
2255                                         valid_lft = MIN_VALID_LIFETIME;
2256                                         if (valid_lft < prefered_lft)
2257                                                 prefered_lft = valid_lft;
2258                                         update_lft = 1;
2259                                 }
2260                         }
2261
2262                         if (update_lft) {
2263                                 ifp->valid_lft = valid_lft;
2264                                 ifp->prefered_lft = prefered_lft;
2265                                 ifp->tstamp = now;
2266                                 flags = ifp->flags;
2267                                 ifp->flags &= ~IFA_F_DEPRECATED;
2268                                 spin_unlock(&ifp->lock);
2269
2270                                 if (!(flags&IFA_F_TENTATIVE))
2271                                         ipv6_ifa_notify(0, ifp);
2272                         } else
2273                                 spin_unlock(&ifp->lock);
2274
2275 #ifdef CONFIG_IPV6_PRIVACY
2276                         read_lock_bh(&in6_dev->lock);
2277                         /* update all temporary addresses in the list */
2278                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2279                                             tmp_list) {
2280                                 int age, max_valid, max_prefered;
2281
2282                                 if (ifp != ift->ifpub)
2283                                         continue;
2284
2285                                 /*
2286                                  * RFC 4941 section 3.3:
2287                                  * If a received option will extend the lifetime
2288                                  * of a public address, the lifetimes of
2289                                  * temporary addresses should be extended,
2290                                  * subject to the overall constraint that no
2291                                  * temporary addresses should ever remain
2292                                  * "valid" or "preferred" for a time longer than
2293                                  * (TEMP_VALID_LIFETIME) or
2294                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2295                                  * respectively.
2296                                  */
2297                                 age = (now - ift->cstamp) / HZ;
2298                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2299                                 if (max_valid < 0)
2300                                         max_valid = 0;
2301
2302                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2303                                                in6_dev->cnf.max_desync_factor -
2304                                                age;
2305                                 if (max_prefered < 0)
2306                                         max_prefered = 0;
2307
2308                                 if (valid_lft > max_valid)
2309                                         valid_lft = max_valid;
2310
2311                                 if (prefered_lft > max_prefered)
2312                                         prefered_lft = max_prefered;
2313
2314                                 spin_lock(&ift->lock);
2315                                 flags = ift->flags;
2316                                 ift->valid_lft = valid_lft;
2317                                 ift->prefered_lft = prefered_lft;
2318                                 ift->tstamp = now;
2319                                 if (prefered_lft > 0)
2320                                         ift->flags &= ~IFA_F_DEPRECATED;
2321
2322                                 spin_unlock(&ift->lock);
2323                                 if (!(flags&IFA_F_TENTATIVE))
2324                                         ipv6_ifa_notify(0, ift);
2325                         }
2326
2327                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2328                                 /*
2329                                  * When a new public address is created as
2330                                  * described in [ADDRCONF], also create a new
2331                                  * temporary address. Also create a temporary
2332                                  * address if it's enabled but no temporary
2333                                  * address currently exists.
2334                                  */
2335                                 read_unlock_bh(&in6_dev->lock);
2336                                 ipv6_create_tempaddr(ifp, NULL);
2337                         } else {
2338                                 read_unlock_bh(&in6_dev->lock);
2339                         }
2340 #endif
2341                         in6_ifa_put(ifp);
2342                         addrconf_verify(0);
2343                 }
2344         }
2345         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2346         in6_dev_put(in6_dev);
2347 }
2348
2349 /*
2350  *      Set destination address.
2351  *      Special case for SIT interfaces where we create a new "virtual"
2352  *      device.
2353  */
2354 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2355 {
2356         struct in6_ifreq ireq;
2357         struct net_device *dev;
2358         int err = -EINVAL;
2359
2360         rtnl_lock();
2361
2362         err = -EFAULT;
2363         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2364                 goto err_exit;
2365
2366         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2367
2368         err = -ENODEV;
2369         if (dev == NULL)
2370                 goto err_exit;
2371
2372 #if IS_ENABLED(CONFIG_IPV6_SIT)
2373         if (dev->type == ARPHRD_SIT) {
2374                 const struct net_device_ops *ops = dev->netdev_ops;
2375                 struct ifreq ifr;
2376                 struct ip_tunnel_parm p;
2377
2378                 err = -EADDRNOTAVAIL;
2379                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2380                         goto err_exit;
2381
2382                 memset(&p, 0, sizeof(p));
2383                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2384                 p.iph.saddr = 0;
2385                 p.iph.version = 4;
2386                 p.iph.ihl = 5;
2387                 p.iph.protocol = IPPROTO_IPV6;
2388                 p.iph.ttl = 64;
2389                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2390
2391                 if (ops->ndo_do_ioctl) {
2392                         mm_segment_t oldfs = get_fs();
2393
2394                         set_fs(KERNEL_DS);
2395                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2396                         set_fs(oldfs);
2397                 } else
2398                         err = -EOPNOTSUPP;
2399
2400                 if (err == 0) {
2401                         err = -ENOBUFS;
2402                         dev = __dev_get_by_name(net, p.name);
2403                         if (!dev)
2404                                 goto err_exit;
2405                         err = dev_open(dev);
2406                 }
2407         }
2408 #endif
2409
2410 err_exit:
2411         rtnl_unlock();
2412         return err;
2413 }
2414
2415 /*
2416  *      Manual configuration of address on an interface
2417  */
2418 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2419                           const struct in6_addr *peer_pfx,
2420                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2421                           __u32 valid_lft)
2422 {
2423         struct inet6_ifaddr *ifp;
2424         struct inet6_dev *idev;
2425         struct net_device *dev;
2426         int scope;
2427         u32 flags;
2428         clock_t expires;
2429         unsigned long timeout;
2430
2431         ASSERT_RTNL();
2432
2433         if (plen > 128)
2434                 return -EINVAL;
2435
2436         /* check the lifetime */
2437         if (!valid_lft || prefered_lft > valid_lft)
2438                 return -EINVAL;
2439
2440         dev = __dev_get_by_index(net, ifindex);
2441         if (!dev)
2442                 return -ENODEV;
2443
2444         idev = addrconf_add_dev(dev);
2445         if (IS_ERR(idev))
2446                 return PTR_ERR(idev);
2447
2448         scope = ipv6_addr_scope(pfx);
2449
2450         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2451         if (addrconf_finite_timeout(timeout)) {
2452                 expires = jiffies_to_clock_t(timeout * HZ);
2453                 valid_lft = timeout;
2454                 flags = RTF_EXPIRES;
2455         } else {
2456                 expires = 0;
2457                 flags = 0;
2458                 ifa_flags |= IFA_F_PERMANENT;
2459         }
2460
2461         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2462         if (addrconf_finite_timeout(timeout)) {
2463                 if (timeout == 0)
2464                         ifa_flags |= IFA_F_DEPRECATED;
2465                 prefered_lft = timeout;
2466         }
2467
2468         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2469                             valid_lft, prefered_lft);
2470
2471         if (!IS_ERR(ifp)) {
2472                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2473                                       expires, flags);
2474                 /*
2475                  * Note that section 3.1 of RFC 4429 indicates
2476                  * that the Optimistic flag should not be set for
2477                  * manually configured addresses
2478                  */
2479                 addrconf_dad_start(ifp);
2480                 in6_ifa_put(ifp);
2481                 addrconf_verify(0);
2482                 return 0;
2483         }
2484
2485         return PTR_ERR(ifp);
2486 }
2487
2488 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2489                           unsigned int plen)
2490 {
2491         struct inet6_ifaddr *ifp;
2492         struct inet6_dev *idev;
2493         struct net_device *dev;
2494
2495         if (plen > 128)
2496                 return -EINVAL;
2497
2498         dev = __dev_get_by_index(net, ifindex);
2499         if (!dev)
2500                 return -ENODEV;
2501
2502         if ((idev = __in6_dev_get(dev)) == NULL)
2503                 return -ENXIO;
2504
2505         read_lock_bh(&idev->lock);
2506         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2507                 if (ifp->prefix_len == plen &&
2508                     ipv6_addr_equal(pfx, &ifp->addr)) {
2509                         in6_ifa_hold(ifp);
2510                         read_unlock_bh(&idev->lock);
2511
2512                         ipv6_del_addr(ifp);
2513                         return 0;
2514                 }
2515         }
2516         read_unlock_bh(&idev->lock);
2517         return -EADDRNOTAVAIL;
2518 }
2519
2520
2521 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2522 {
2523         struct in6_ifreq ireq;
2524         int err;
2525
2526         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2527                 return -EPERM;
2528
2529         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2530                 return -EFAULT;
2531
2532         rtnl_lock();
2533         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2534                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2535                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2536         rtnl_unlock();
2537         return err;
2538 }
2539
2540 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2541 {
2542         struct in6_ifreq ireq;
2543         int err;
2544
2545         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2546                 return -EPERM;
2547
2548         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2549                 return -EFAULT;
2550
2551         rtnl_lock();
2552         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2553                              ireq.ifr6_prefixlen);
2554         rtnl_unlock();
2555         return err;
2556 }
2557
2558 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2559                      int plen, int scope)
2560 {
2561         struct inet6_ifaddr *ifp;
2562
2563         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2564                             scope, IFA_F_PERMANENT, 0, 0);
2565         if (!IS_ERR(ifp)) {
2566                 spin_lock_bh(&ifp->lock);
2567                 ifp->flags &= ~IFA_F_TENTATIVE;
2568                 spin_unlock_bh(&ifp->lock);
2569                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2570                 in6_ifa_put(ifp);
2571         }
2572 }
2573
2574 #if IS_ENABLED(CONFIG_IPV6_SIT)
2575 static void sit_add_v4_addrs(struct inet6_dev *idev)
2576 {
2577         struct in6_addr addr;
2578         struct net_device *dev;
2579         struct net *net = dev_net(idev->dev);
2580         int scope;
2581
2582         ASSERT_RTNL();
2583
2584         memset(&addr, 0, sizeof(struct in6_addr));
2585         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2586
2587         if (idev->dev->flags&IFF_POINTOPOINT) {
2588                 addr.s6_addr32[0] = htonl(0xfe800000);
2589                 scope = IFA_LINK;
2590         } else {
2591                 scope = IPV6_ADDR_COMPATv4;
2592         }
2593
2594         if (addr.s6_addr32[3]) {
2595                 add_addr(idev, &addr, 128, scope);
2596                 return;
2597         }
2598
2599         for_each_netdev(net, dev) {
2600                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2601                 if (in_dev && (dev->flags & IFF_UP)) {
2602                         struct in_ifaddr *ifa;
2603
2604                         int flag = scope;
2605
2606                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2607                                 int plen;
2608
2609                                 addr.s6_addr32[3] = ifa->ifa_local;
2610
2611                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2612                                         continue;
2613                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2614                                         if (idev->dev->flags&IFF_POINTOPOINT)
2615                                                 continue;
2616                                         flag |= IFA_HOST;
2617                                 }
2618                                 if (idev->dev->flags&IFF_POINTOPOINT)
2619                                         plen = 64;
2620                                 else
2621                                         plen = 96;
2622
2623                                 add_addr(idev, &addr, plen, flag);
2624                         }
2625                 }
2626         }
2627 }
2628 #endif
2629
2630 static void init_loopback(struct net_device *dev)
2631 {
2632         struct inet6_dev  *idev;
2633         struct net_device *sp_dev;
2634         struct inet6_ifaddr *sp_ifa;
2635         struct rt6_info *sp_rt;
2636
2637         /* ::1 */
2638
2639         ASSERT_RTNL();
2640
2641         if ((idev = ipv6_find_idev(dev)) == NULL) {
2642                 pr_debug("%s: add_dev failed\n", __func__);
2643                 return;
2644         }
2645
2646         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2647
2648         /* Add routes to other interface's IPv6 addresses */
2649         for_each_netdev(dev_net(dev), sp_dev) {
2650                 if (!strcmp(sp_dev->name, dev->name))
2651                         continue;
2652
2653                 idev = __in6_dev_get(sp_dev);
2654                 if (!idev)
2655                         continue;
2656
2657                 read_lock_bh(&idev->lock);
2658                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2659
2660                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2661                                 continue;
2662
2663                         if (sp_ifa->rt)
2664                                 continue;
2665
2666                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2667
2668                         /* Failure cases are ignored */
2669                         if (!IS_ERR(sp_rt)) {
2670                                 sp_ifa->rt = sp_rt;
2671                                 ip6_ins_rt(sp_rt);
2672                         }
2673                 }
2674                 read_unlock_bh(&idev->lock);
2675         }
2676 }
2677
2678 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2679 {
2680         struct inet6_ifaddr *ifp;
2681         u32 addr_flags = IFA_F_PERMANENT;
2682
2683 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2684         if (idev->cnf.optimistic_dad &&
2685             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2686                 addr_flags |= IFA_F_OPTIMISTIC;
2687 #endif
2688
2689
2690         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2691         if (!IS_ERR(ifp)) {
2692                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2693                 addrconf_dad_start(ifp);
2694                 in6_ifa_put(ifp);
2695         }
2696 }
2697
2698 static void addrconf_dev_config(struct net_device *dev)
2699 {
2700         struct in6_addr addr;
2701         struct inet6_dev *idev;
2702
2703         ASSERT_RTNL();
2704
2705         if ((dev->type != ARPHRD_ETHER) &&
2706             (dev->type != ARPHRD_FDDI) &&
2707             (dev->type != ARPHRD_ARCNET) &&
2708             (dev->type != ARPHRD_INFINIBAND) &&
2709             (dev->type != ARPHRD_IEEE802154) &&
2710             (dev->type != ARPHRD_IEEE1394) &&
2711             (dev->type != ARPHRD_TUNNEL6)) {
2712                 /* Alas, we support only Ethernet autoconfiguration. */
2713                 return;
2714         }
2715
2716         idev = addrconf_add_dev(dev);
2717         if (IS_ERR(idev))
2718                 return;
2719
2720         memset(&addr, 0, sizeof(struct in6_addr));
2721         addr.s6_addr32[0] = htonl(0xFE800000);
2722
2723         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2724                 addrconf_add_linklocal(idev, &addr);
2725 }
2726
2727 #if IS_ENABLED(CONFIG_IPV6_SIT)
2728 static void addrconf_sit_config(struct net_device *dev)
2729 {
2730         struct inet6_dev *idev;
2731
2732         ASSERT_RTNL();
2733
2734         /*
2735          * Configure the tunnel with one of our IPv4
2736          * addresses... we should configure all of
2737          * our v4 addrs in the tunnel
2738          */
2739
2740         if ((idev = ipv6_find_idev(dev)) == NULL) {
2741                 pr_debug("%s: add_dev failed\n", __func__);
2742                 return;
2743         }
2744
2745         if (dev->priv_flags & IFF_ISATAP) {
2746                 struct in6_addr addr;
2747
2748                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2749                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2750                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2751                         addrconf_add_linklocal(idev, &addr);
2752                 return;
2753         }
2754
2755         sit_add_v4_addrs(idev);
2756
2757         if (dev->flags&IFF_POINTOPOINT)
2758                 addrconf_add_mroute(dev);
2759         else
2760                 sit_route_add(dev);
2761 }
2762 #endif
2763
2764 #if IS_ENABLED(CONFIG_NET_IPGRE)
2765 static void addrconf_gre_config(struct net_device *dev)
2766 {
2767         struct inet6_dev *idev;
2768         struct in6_addr addr;
2769
2770         ASSERT_RTNL();
2771
2772         if ((idev = ipv6_find_idev(dev)) == NULL) {
2773                 pr_debug("%s: add_dev failed\n", __func__);
2774                 return;
2775         }
2776
2777         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2778         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2779
2780         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2781                 addrconf_add_linklocal(idev, &addr);
2782 }
2783 #endif
2784
2785 static inline int
2786 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2787 {
2788         struct in6_addr lladdr;
2789
2790         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2791                 addrconf_add_linklocal(idev, &lladdr);
2792                 return 0;
2793         }
2794         return -1;
2795 }
2796
2797 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2798                            void *ptr)
2799 {
2800         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2801         struct inet6_dev *idev = __in6_dev_get(dev);
2802         int run_pending = 0;
2803         int err;
2804
2805         switch (event) {
2806         case NETDEV_REGISTER:
2807                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2808                         idev = ipv6_add_dev(dev);
2809                         if (!idev)
2810                                 return notifier_from_errno(-ENOMEM);
2811                 }
2812                 break;
2813
2814         case NETDEV_UP:
2815         case NETDEV_CHANGE:
2816                 if (dev->flags & IFF_SLAVE)
2817                         break;
2818
2819                 if (event == NETDEV_UP) {
2820                         if (!addrconf_qdisc_ok(dev)) {
2821                                 /* device is not ready yet. */
2822                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2823                                         dev->name);
2824                                 break;
2825                         }
2826
2827                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2828                                 idev = ipv6_add_dev(dev);
2829
2830                         if (idev) {
2831                                 idev->if_flags |= IF_READY;
2832                                 run_pending = 1;
2833                         }
2834                 } else {
2835                         if (!addrconf_qdisc_ok(dev)) {
2836                                 /* device is still not ready. */
2837                                 break;
2838                         }
2839
2840                         if (idev) {
2841                                 if (idev->if_flags & IF_READY)
2842                                         /* device is already configured. */
2843                                         break;
2844                                 idev->if_flags |= IF_READY;
2845                         }
2846
2847                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2848                                 dev->name);
2849
2850                         run_pending = 1;
2851                 }
2852
2853                 switch (dev->type) {
2854 #if IS_ENABLED(CONFIG_IPV6_SIT)
2855                 case ARPHRD_SIT:
2856                         addrconf_sit_config(dev);
2857                         break;
2858 #endif
2859 #if IS_ENABLED(CONFIG_NET_IPGRE)
2860                 case ARPHRD_IPGRE:
2861                         addrconf_gre_config(dev);
2862                         break;
2863 #endif
2864                 case ARPHRD_LOOPBACK:
2865                         init_loopback(dev);
2866                         break;
2867
2868                 default:
2869                         addrconf_dev_config(dev);
2870                         break;
2871                 }
2872
2873                 if (idev) {
2874                         if (run_pending)
2875                                 addrconf_dad_run(idev);
2876
2877                         /*
2878                          * If the MTU changed during the interface down,
2879                          * when the interface up, the changed MTU must be
2880                          * reflected in the idev as well as routers.
2881                          */
2882                         if (idev->cnf.mtu6 != dev->mtu &&
2883                             dev->mtu >= IPV6_MIN_MTU) {
2884                                 rt6_mtu_change(dev, dev->mtu);
2885                                 idev->cnf.mtu6 = dev->mtu;
2886                         }
2887                         idev->tstamp = jiffies;
2888                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2889
2890                         /*
2891                          * If the changed mtu during down is lower than
2892                          * IPV6_MIN_MTU stop IPv6 on this interface.
2893                          */
2894                         if (dev->mtu < IPV6_MIN_MTU)
2895                                 addrconf_ifdown(dev, 1);
2896                 }
2897                 break;
2898
2899         case NETDEV_CHANGEMTU:
2900                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2901                         rt6_mtu_change(dev, dev->mtu);
2902                         idev->cnf.mtu6 = dev->mtu;
2903                         break;
2904                 }
2905
2906                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2907                         idev = ipv6_add_dev(dev);
2908                         if (idev)
2909                                 break;
2910                 }
2911
2912                 /*
2913                  * MTU falled under IPV6_MIN_MTU.
2914                  * Stop IPv6 on this interface.
2915                  */
2916
2917         case NETDEV_DOWN:
2918         case NETDEV_UNREGISTER:
2919                 /*
2920                  *      Remove all addresses from this interface.
2921                  */
2922                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2923                 break;
2924
2925         case NETDEV_CHANGENAME:
2926                 if (idev) {
2927                         snmp6_unregister_dev(idev);
2928                         addrconf_sysctl_unregister(idev);
2929                         addrconf_sysctl_register(idev);
2930                         err = snmp6_register_dev(idev);
2931                         if (err)
2932                                 return notifier_from_errno(err);
2933                 }
2934                 break;
2935
2936         case NETDEV_PRE_TYPE_CHANGE:
2937         case NETDEV_POST_TYPE_CHANGE:
2938                 addrconf_type_change(dev, event);
2939                 break;
2940         }
2941
2942         return NOTIFY_OK;
2943 }
2944
2945 /*
2946  *      addrconf module should be notified of a device going up
2947  */
2948 static struct notifier_block ipv6_dev_notf = {
2949         .notifier_call = addrconf_notify,
2950 };
2951
2952 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2953 {
2954         struct inet6_dev *idev;
2955         ASSERT_RTNL();
2956
2957         idev = __in6_dev_get(dev);
2958
2959         if (event == NETDEV_POST_TYPE_CHANGE)
2960                 ipv6_mc_remap(idev);
2961         else if (event == NETDEV_PRE_TYPE_CHANGE)
2962                 ipv6_mc_unmap(idev);
2963 }
2964
2965 static int addrconf_ifdown(struct net_device *dev, int how)
2966 {
2967         struct net *net = dev_net(dev);
2968         struct inet6_dev *idev;
2969         struct inet6_ifaddr *ifa;
2970         int state, i;
2971
2972         ASSERT_RTNL();
2973
2974         rt6_ifdown(net, dev);
2975         neigh_ifdown(&nd_tbl, dev);
2976
2977         idev = __in6_dev_get(dev);
2978         if (idev == NULL)
2979                 return -ENODEV;
2980
2981         /*
2982          * Step 1: remove reference to ipv6 device from parent device.
2983          *         Do not dev_put!
2984          */
2985         if (how) {
2986                 idev->dead = 1;
2987
2988                 /* protected by rtnl_lock */
2989                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2990
2991                 /* Step 1.5: remove snmp6 entry */
2992                 snmp6_unregister_dev(idev);
2993
2994         }
2995
2996         /* Step 2: clear hash table */
2997         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2998                 struct hlist_head *h = &inet6_addr_lst[i];
2999
3000                 spin_lock_bh(&addrconf_hash_lock);
3001         restart:
3002                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3003                         if (ifa->idev == idev) {
3004                                 hlist_del_init_rcu(&ifa->addr_lst);
3005                                 addrconf_del_dad_timer(ifa);
3006                                 goto restart;
3007                         }
3008                 }
3009                 spin_unlock_bh(&addrconf_hash_lock);
3010         }
3011
3012         write_lock_bh(&idev->lock);
3013
3014         addrconf_del_rs_timer(idev);
3015
3016         /* Step 2: clear flags for stateless addrconf */
3017         if (!how)
3018                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3019
3020 #ifdef CONFIG_IPV6_PRIVACY
3021         if (how && del_timer(&idev->regen_timer))
3022                 in6_dev_put(idev);
3023
3024         /* Step 3: clear tempaddr list */
3025         while (!list_empty(&idev->tempaddr_list)) {
3026                 ifa = list_first_entry(&idev->tempaddr_list,
3027                                        struct inet6_ifaddr, tmp_list);
3028                 list_del(&ifa->tmp_list);
3029                 write_unlock_bh(&idev->lock);
3030                 spin_lock_bh(&ifa->lock);
3031
3032                 if (ifa->ifpub) {
3033                         in6_ifa_put(ifa->ifpub);
3034                         ifa->ifpub = NULL;
3035                 }
3036                 spin_unlock_bh(&ifa->lock);
3037                 in6_ifa_put(ifa);
3038                 write_lock_bh(&idev->lock);
3039         }
3040 #endif
3041
3042         while (!list_empty(&idev->addr_list)) {
3043                 ifa = list_first_entry(&idev->addr_list,
3044                                        struct inet6_ifaddr, if_list);
3045                 addrconf_del_dad_timer(ifa);
3046
3047                 list_del(&ifa->if_list);
3048
3049                 write_unlock_bh(&idev->lock);
3050
3051                 spin_lock_bh(&ifa->state_lock);
3052                 state = ifa->state;
3053                 ifa->state = INET6_IFADDR_STATE_DEAD;
3054                 spin_unlock_bh(&ifa->state_lock);
3055
3056                 if (state != INET6_IFADDR_STATE_DEAD) {
3057                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3058                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3059                 }
3060                 in6_ifa_put(ifa);
3061
3062                 write_lock_bh(&idev->lock);
3063         }
3064
3065         write_unlock_bh(&idev->lock);
3066
3067         /* Step 5: Discard multicast list */
3068         if (how)
3069                 ipv6_mc_destroy_dev(idev);
3070         else
3071                 ipv6_mc_down(idev);
3072
3073         idev->tstamp = jiffies;
3074
3075         /* Last: Shot the device (if unregistered) */
3076         if (how) {
3077                 addrconf_sysctl_unregister(idev);
3078                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3079                 neigh_ifdown(&nd_tbl, dev);
3080                 in6_dev_put(idev);
3081         }
3082         return 0;
3083 }
3084
3085 static void addrconf_rs_timer(unsigned long data)
3086 {
3087         struct inet6_dev *idev = (struct inet6_dev *)data;
3088         struct net_device *dev = idev->dev;
3089         struct in6_addr lladdr;
3090
3091         write_lock(&idev->lock);
3092         if (idev->dead || !(idev->if_flags & IF_READY))
3093                 goto out;
3094
3095         if (!ipv6_accept_ra(idev))
3096                 goto out;
3097
3098         /* Announcement received after solicitation was sent */
3099         if (idev->if_flags & IF_RA_RCVD)
3100                 goto out;
3101
3102         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3103                 write_unlock(&idev->lock);
3104                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3105                         ndisc_send_rs(dev, &lladdr,
3106                                       &in6addr_linklocal_allrouters);
3107                 else
3108                         goto put;
3109
3110                 write_lock(&idev->lock);
3111                 /* The wait after the last probe can be shorter */
3112                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3113                                              idev->cnf.rtr_solicits) ?
3114                                       idev->cnf.rtr_solicit_delay :
3115                                       idev->cnf.rtr_solicit_interval);
3116         } else {
3117                 /*
3118                  * Note: we do not support deprecated "all on-link"
3119                  * assumption any longer.
3120                  */
3121                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3122         }
3123
3124 out:
3125         write_unlock(&idev->lock);
3126 put:
3127         in6_dev_put(idev);
3128 }
3129
3130 /*
3131  *      Duplicate Address Detection
3132  */
3133 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3134 {
3135         unsigned long rand_num;
3136         struct inet6_dev *idev = ifp->idev;
3137
3138         if (ifp->flags & IFA_F_OPTIMISTIC)
3139                 rand_num = 0;
3140         else
3141                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3142
3143         ifp->dad_probes = idev->cnf.dad_transmits;
3144         addrconf_mod_dad_timer(ifp, rand_num);
3145 }
3146
3147 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3148 {
3149         struct inet6_dev *idev = ifp->idev;
3150         struct net_device *dev = idev->dev;
3151
3152         addrconf_join_solict(dev, &ifp->addr);
3153
3154         net_srandom(ifp->addr.s6_addr32[3]);
3155
3156         read_lock_bh(&idev->lock);
3157         spin_lock(&ifp->lock);
3158         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3159                 goto out;
3160
3161         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3162             idev->cnf.accept_dad < 1 ||
3163             !(ifp->flags&IFA_F_TENTATIVE) ||
3164             ifp->flags & IFA_F_NODAD) {
3165                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3166                 spin_unlock(&ifp->lock);
3167                 read_unlock_bh(&idev->lock);
3168
3169                 addrconf_dad_completed(ifp);
3170                 return;
3171         }
3172
3173         if (!(idev->if_flags & IF_READY)) {
3174                 spin_unlock(&ifp->lock);
3175                 read_unlock_bh(&idev->lock);
3176                 /*
3177                  * If the device is not ready:
3178                  * - keep it tentative if it is a permanent address.
3179                  * - otherwise, kill it.
3180                  */
3181                 in6_ifa_hold(ifp);
3182                 addrconf_dad_stop(ifp, 0);
3183                 return;
3184         }
3185
3186         /*
3187          * Optimistic nodes can start receiving
3188          * Frames right away
3189          */
3190         if (ifp->flags & IFA_F_OPTIMISTIC)
3191                 ip6_ins_rt(ifp->rt);
3192
3193         addrconf_dad_kick(ifp);
3194 out:
3195         spin_unlock(&ifp->lock);
3196         read_unlock_bh(&idev->lock);
3197 }
3198
3199 static void addrconf_dad_timer(unsigned long data)
3200 {
3201         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3202         struct inet6_dev *idev = ifp->idev;
3203         struct in6_addr mcaddr;
3204
3205         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3206                 goto out;
3207
3208         write_lock(&idev->lock);
3209         if (idev->dead || !(idev->if_flags & IF_READY)) {
3210                 write_unlock(&idev->lock);
3211                 goto out;
3212         }
3213
3214         spin_lock(&ifp->lock);
3215         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3216                 spin_unlock(&ifp->lock);
3217                 write_unlock(&idev->lock);
3218                 goto out;
3219         }
3220
3221         if (ifp->dad_probes == 0) {
3222                 /*
3223                  * DAD was successful
3224                  */
3225
3226                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3227                 spin_unlock(&ifp->lock);
3228                 write_unlock(&idev->lock);
3229
3230                 addrconf_dad_completed(ifp);
3231
3232                 goto out;
3233         }
3234
3235         ifp->dad_probes--;
3236         addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3237         spin_unlock(&ifp->lock);
3238         write_unlock(&idev->lock);
3239
3240         /* send a neighbour solicitation for our addr */
3241         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3242         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3243 out:
3244         in6_ifa_put(ifp);
3245 }
3246
3247 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3248 {
3249         struct net_device *dev = ifp->idev->dev;
3250         struct in6_addr lladdr;
3251         bool send_rs, send_mld;
3252
3253         addrconf_del_dad_timer(ifp);
3254
3255         /*
3256          *      Configure the address for reception. Now it is valid.
3257          */
3258
3259         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3260
3261         /* If added prefix is link local and we are prepared to process
3262            router advertisements, start sending router solicitations.
3263          */
3264
3265         read_lock_bh(&ifp->idev->lock);
3266         spin_lock(&ifp->lock);
3267         send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3268                    ifp->idev->valid_ll_addr_cnt == 1;
3269         send_rs = send_mld &&
3270                   ipv6_accept_ra(ifp->idev) &&
3271                   ifp->idev->cnf.rtr_solicits > 0 &&
3272                   (dev->flags&IFF_LOOPBACK) == 0;
3273         spin_unlock(&ifp->lock);
3274         read_unlock_bh(&ifp->idev->lock);
3275
3276         /* While dad is in progress mld report's source address is in6_addrany.
3277          * Resend with proper ll now.
3278          */
3279         if (send_mld)
3280                 ipv6_mc_dad_complete(ifp->idev);
3281
3282         if (send_rs) {
3283                 /*
3284                  *      If a host as already performed a random delay
3285                  *      [...] as part of DAD [...] there is no need
3286                  *      to delay again before sending the first RS
3287                  */
3288                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3289                         return;
3290                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3291
3292                 write_lock_bh(&ifp->idev->lock);
3293                 spin_lock(&ifp->lock);
3294                 ifp->idev->rs_probes = 1;
3295                 ifp->idev->if_flags |= IF_RS_SENT;
3296                 addrconf_mod_rs_timer(ifp->idev,
3297                                       ifp->idev->cnf.rtr_solicit_interval);
3298                 spin_unlock(&ifp->lock);
3299                 write_unlock_bh(&ifp->idev->lock);
3300         }
3301 }
3302
3303 static void addrconf_dad_run(struct inet6_dev *idev)
3304 {
3305         struct inet6_ifaddr *ifp;
3306
3307         read_lock_bh(&idev->lock);
3308         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3309                 spin_lock(&ifp->lock);
3310                 if (ifp->flags & IFA_F_TENTATIVE &&
3311                     ifp->state == INET6_IFADDR_STATE_DAD)
3312                         addrconf_dad_kick(ifp);
3313                 spin_unlock(&ifp->lock);
3314         }
3315         read_unlock_bh(&idev->lock);
3316 }
3317
3318 #ifdef CONFIG_PROC_FS
3319 struct if6_iter_state {
3320         struct seq_net_private p;
3321         int bucket;
3322         int offset;
3323 };
3324
3325 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3326 {
3327         struct inet6_ifaddr *ifa = NULL;
3328         struct if6_iter_state *state = seq->private;
3329         struct net *net = seq_file_net(seq);
3330         int p = 0;
3331
3332         /* initial bucket if pos is 0 */
3333         if (pos == 0) {
3334                 state->bucket = 0;
3335                 state->offset = 0;
3336         }
3337
3338         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3339                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3340                                          addr_lst) {
3341                         if (!net_eq(dev_net(ifa->idev->dev), net))
3342                                 continue;
3343                         /* sync with offset */
3344                         if (p < state->offset) {
3345                                 p++;
3346                                 continue;
3347                         }
3348                         state->offset++;
3349                         return ifa;
3350                 }
3351
3352                 /* prepare for next bucket */
3353                 state->offset = 0;
3354                 p = 0;
3355         }
3356         return NULL;
3357 }
3358
3359 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3360                                          struct inet6_ifaddr *ifa)
3361 {
3362         struct if6_iter_state *state = seq->private;
3363         struct net *net = seq_file_net(seq);
3364
3365         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3366                 if (!net_eq(dev_net(ifa->idev->dev), net))
3367                         continue;
3368                 state->offset++;
3369                 return ifa;
3370         }
3371
3372         while (++state->bucket < IN6_ADDR_HSIZE) {
3373                 state->offset = 0;
3374                 hlist_for_each_entry_rcu_bh(ifa,
3375                                      &inet6_addr_lst[state->bucket], addr_lst) {
3376                         if (!net_eq(dev_net(ifa->idev->dev), net))
3377                                 continue;
3378                         state->offset++;
3379                         return ifa;
3380                 }
3381         }
3382
3383         return NULL;
3384 }
3385
3386 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3387         __acquires(rcu_bh)
3388 {
3389         rcu_read_lock_bh();
3390         return if6_get_first(seq, *pos);
3391 }
3392
3393 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3394 {
3395         struct inet6_ifaddr *ifa;
3396
3397         ifa = if6_get_next(seq, v);
3398         ++*pos;
3399         return ifa;
3400 }
3401
3402 static void if6_seq_stop(struct seq_file *seq, void *v)
3403         __releases(rcu_bh)
3404 {
3405         rcu_read_unlock_bh();
3406 }
3407
3408 static int if6_seq_show(struct seq_file *seq, void *v)
3409 {
3410         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3411         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3412                    &ifp->addr,
3413                    ifp->idev->dev->ifindex,
3414                    ifp->prefix_len,
3415                    ifp->scope,
3416                    ifp->flags,
3417                    ifp->idev->dev->name);
3418         return 0;
3419 }
3420
3421 static const struct seq_operations if6_seq_ops = {
3422         .start  = if6_seq_start,
3423         .next   = if6_seq_next,
3424         .show   = if6_seq_show,
3425         .stop   = if6_seq_stop,
3426 };
3427
3428 static int if6_seq_open(struct inode *inode, struct file *file)
3429 {
3430         return seq_open_net(inode, file, &if6_seq_ops,
3431                             sizeof(struct if6_iter_state));
3432 }
3433
3434 static const struct file_operations if6_fops = {
3435         .owner          = THIS_MODULE,
3436         .open           = if6_seq_open,
3437         .read           = seq_read,
3438         .llseek         = seq_lseek,
3439         .release        = seq_release_net,
3440 };
3441
3442 static int __net_init if6_proc_net_init(struct net *net)
3443 {
3444         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3445                 return -ENOMEM;
3446         return 0;
3447 }
3448
3449 static void __net_exit if6_proc_net_exit(struct net *net)
3450 {
3451         remove_proc_entry("if_inet6", net->proc_net);
3452 }
3453
3454 static struct pernet_operations if6_proc_net_ops = {
3455        .init = if6_proc_net_init,
3456        .exit = if6_proc_net_exit,
3457 };
3458
3459 int __init if6_proc_init(void)
3460 {
3461         return register_pernet_subsys(&if6_proc_net_ops);
3462 }
3463
3464 void if6_proc_exit(void)
3465 {
3466         unregister_pernet_subsys(&if6_proc_net_ops);
3467 }
3468 #endif  /* CONFIG_PROC_FS */
3469
3470 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3471 /* Check if address is a home address configured on any interface. */
3472 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3473 {
3474         int ret = 0;
3475         struct inet6_ifaddr *ifp = NULL;
3476         unsigned int hash = inet6_addr_hash(addr);
3477
3478         rcu_read_lock_bh();
3479         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3480                 if (!net_eq(dev_net(ifp->idev->dev), net))
3481                         continue;
3482                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3483                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3484                         ret = 1;
3485                         break;
3486                 }
3487         }
3488         rcu_read_unlock_bh();
3489         return ret;
3490 }
3491 #endif
3492
3493 /*
3494  *      Periodic address status verification
3495  */
3496
3497 static void addrconf_verify(unsigned long foo)
3498 {
3499         unsigned long now, next, next_sec, next_sched;
3500         struct inet6_ifaddr *ifp;
3501         int i;
3502
3503         rcu_read_lock_bh();
3504         spin_lock(&addrconf_verify_lock);
3505         now = jiffies;
3506         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3507
3508         del_timer(&addr_chk_timer);
3509
3510         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3511 restart:
3512                 hlist_for_each_entry_rcu_bh(ifp,
3513                                          &inet6_addr_lst[i], addr_lst) {
3514                         unsigned long age;
3515
3516                         if (ifp->flags & IFA_F_PERMANENT)
3517                                 continue;
3518
3519                         spin_lock(&ifp->lock);
3520                         /* We try to batch several events at once. */
3521                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3522
3523                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3524                             age >= ifp->valid_lft) {
3525                                 spin_unlock(&ifp->lock);
3526                                 in6_ifa_hold(ifp);
3527                                 ipv6_del_addr(ifp);
3528                                 goto restart;
3529                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3530                                 spin_unlock(&ifp->lock);
3531                                 continue;
3532                         } else if (age >= ifp->prefered_lft) {
3533                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3534                                 int deprecate = 0;
3535
3536                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3537                                         deprecate = 1;
3538                                         ifp->flags |= IFA_F_DEPRECATED;
3539                                 }
3540
3541                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3542                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3543
3544                                 spin_unlock(&ifp->lock);
3545
3546                                 if (deprecate) {
3547                                         in6_ifa_hold(ifp);
3548
3549                                         ipv6_ifa_notify(0, ifp);
3550                                         in6_ifa_put(ifp);
3551                                         goto restart;
3552                                 }
3553 #ifdef CONFIG_IPV6_PRIVACY
3554                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3555                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3556                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3557                                         ifp->idev->cnf.dad_transmits *
3558                                         ifp->idev->nd_parms->retrans_time / HZ;
3559
3560                                 if (age >= ifp->prefered_lft - regen_advance) {
3561                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3562                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3563                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3564                                         if (!ifp->regen_count && ifpub) {
3565                                                 ifp->regen_count++;
3566                                                 in6_ifa_hold(ifp);
3567                                                 in6_ifa_hold(ifpub);
3568                                                 spin_unlock(&ifp->lock);
3569
3570                                                 spin_lock(&ifpub->lock);
3571                                                 ifpub->regen_count = 0;
3572                                                 spin_unlock(&ifpub->lock);
3573                                                 ipv6_create_tempaddr(ifpub, ifp);
3574                                                 in6_ifa_put(ifpub);
3575                                                 in6_ifa_put(ifp);
3576                                                 goto restart;
3577                                         }
3578                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3579                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3580                                 spin_unlock(&ifp->lock);
3581 #endif
3582                         } else {
3583                                 /* ifp->prefered_lft <= ifp->valid_lft */
3584                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3585                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3586                                 spin_unlock(&ifp->lock);
3587                         }
3588                 }
3589         }
3590
3591         next_sec = round_jiffies_up(next);
3592         next_sched = next;
3593
3594         /* If rounded timeout is accurate enough, accept it. */
3595         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3596                 next_sched = next_sec;
3597
3598         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3599         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3600                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3601
3602         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3603               now, next, next_sec, next_sched);
3604
3605         addr_chk_timer.expires = next_sched;
3606         add_timer(&addr_chk_timer);
3607         spin_unlock(&addrconf_verify_lock);
3608         rcu_read_unlock_bh();
3609 }
3610
3611 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3612                                      struct in6_addr **peer_pfx)
3613 {
3614         struct in6_addr *pfx = NULL;
3615
3616         *peer_pfx = NULL;
3617
3618         if (addr)
3619                 pfx = nla_data(addr);
3620
3621         if (local) {
3622                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3623                         *peer_pfx = pfx;
3624                 pfx = nla_data(local);
3625         }
3626
3627         return pfx;
3628 }
3629
3630 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3631         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3632         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3633         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3634 };
3635
3636 static int
3637 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3638 {
3639         struct net *net = sock_net(skb->sk);
3640         struct ifaddrmsg *ifm;
3641         struct nlattr *tb[IFA_MAX+1];
3642         struct in6_addr *pfx, *peer_pfx;
3643         int err;
3644
3645         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3646         if (err < 0)
3647                 return err;
3648
3649         ifm = nlmsg_data(nlh);
3650         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3651         if (pfx == NULL)
3652                 return -EINVAL;
3653
3654         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3655 }
3656
3657 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3658                              u32 prefered_lft, u32 valid_lft)
3659 {
3660         u32 flags;
3661         clock_t expires;
3662         unsigned long timeout;
3663
3664         if (!valid_lft || (prefered_lft > valid_lft))
3665                 return -EINVAL;
3666
3667         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3668         if (addrconf_finite_timeout(timeout)) {
3669                 expires = jiffies_to_clock_t(timeout * HZ);
3670                 valid_lft = timeout;
3671                 flags = RTF_EXPIRES;
3672         } else {
3673                 expires = 0;
3674                 flags = 0;
3675                 ifa_flags |= IFA_F_PERMANENT;
3676         }
3677
3678         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3679         if (addrconf_finite_timeout(timeout)) {
3680                 if (timeout == 0)
3681                         ifa_flags |= IFA_F_DEPRECATED;
3682                 prefered_lft = timeout;
3683         }
3684
3685         spin_lock_bh(&ifp->lock);
3686         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3687         ifp->tstamp = jiffies;
3688         ifp->valid_lft = valid_lft;
3689         ifp->prefered_lft = prefered_lft;
3690
3691         spin_unlock_bh(&ifp->lock);
3692         if (!(ifp->flags&IFA_F_TENTATIVE))
3693                 ipv6_ifa_notify(0, ifp);
3694
3695         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3696                               expires, flags);
3697         addrconf_verify(0);
3698
3699         return 0;
3700 }
3701
3702 static int
3703 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3704 {
3705         struct net *net = sock_net(skb->sk);
3706         struct ifaddrmsg *ifm;
3707         struct nlattr *tb[IFA_MAX+1];
3708         struct in6_addr *pfx, *peer_pfx;
3709         struct inet6_ifaddr *ifa;
3710         struct net_device *dev;
3711         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3712         u8 ifa_flags;
3713         int err;
3714
3715         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3716         if (err < 0)
3717                 return err;
3718
3719         ifm = nlmsg_data(nlh);
3720         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3721         if (pfx == NULL)
3722                 return -EINVAL;
3723
3724         if (tb[IFA_CACHEINFO]) {
3725                 struct ifa_cacheinfo *ci;
3726
3727                 ci = nla_data(tb[IFA_CACHEINFO]);
3728                 valid_lft = ci->ifa_valid;
3729                 preferred_lft = ci->ifa_prefered;
3730         } else {
3731                 preferred_lft = INFINITY_LIFE_TIME;
3732                 valid_lft = INFINITY_LIFE_TIME;
3733         }
3734
3735         dev =  __dev_get_by_index(net, ifm->ifa_index);
3736         if (dev == NULL)
3737                 return -ENODEV;
3738
3739         /* We ignore other flags so far. */
3740         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3741
3742         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3743         if (ifa == NULL) {
3744                 /*
3745                  * It would be best to check for !NLM_F_CREATE here but
3746                  * userspace alreay relies on not having to provide this.
3747                  */
3748                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3749                                       ifm->ifa_prefixlen, ifa_flags,
3750                                       preferred_lft, valid_lft);
3751         }
3752
3753         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3754             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3755                 err = -EEXIST;
3756         else
3757                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3758
3759         in6_ifa_put(ifa);
3760
3761         return err;
3762 }
3763
3764 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3765                           u8 scope, int ifindex)
3766 {
3767         struct ifaddrmsg *ifm;
3768
3769         ifm = nlmsg_data(nlh);
3770         ifm->ifa_family = AF_INET6;
3771         ifm->ifa_prefixlen = prefixlen;
3772         ifm->ifa_flags = flags;
3773         ifm->ifa_scope = scope;
3774         ifm->ifa_index = ifindex;
3775 }
3776
3777 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3778                          unsigned long tstamp, u32 preferred, u32 valid)
3779 {
3780         struct ifa_cacheinfo ci;
3781
3782         ci.cstamp = cstamp_delta(cstamp);
3783         ci.tstamp = cstamp_delta(tstamp);
3784         ci.ifa_prefered = preferred;
3785         ci.ifa_valid = valid;
3786
3787         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3788 }
3789
3790 static inline int rt_scope(int ifa_scope)
3791 {
3792         if (ifa_scope & IFA_HOST)
3793                 return RT_SCOPE_HOST;
3794         else if (ifa_scope & IFA_LINK)
3795                 return RT_SCOPE_LINK;
3796         else if (ifa_scope & IFA_SITE)
3797                 return RT_SCOPE_SITE;
3798         else
3799                 return RT_SCOPE_UNIVERSE;
3800 }
3801
3802 static inline int inet6_ifaddr_msgsize(void)
3803 {
3804         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3805                + nla_total_size(16) /* IFA_LOCAL */
3806                + nla_total_size(16) /* IFA_ADDRESS */
3807                + nla_total_size(sizeof(struct ifa_cacheinfo));
3808 }
3809
3810 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3811                              u32 portid, u32 seq, int event, unsigned int flags)
3812 {
3813         struct nlmsghdr  *nlh;
3814         u32 preferred, valid;
3815
3816         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3817         if (nlh == NULL)
3818                 return -EMSGSIZE;
3819
3820         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3821                       ifa->idev->dev->ifindex);
3822
3823         if (!(ifa->flags&IFA_F_PERMANENT)) {
3824                 preferred = ifa->prefered_lft;
3825                 valid = ifa->valid_lft;
3826                 if (preferred != INFINITY_LIFE_TIME) {
3827                         long tval = (jiffies - ifa->tstamp)/HZ;
3828                         if (preferred > tval)
3829                                 preferred -= tval;
3830                         else
3831                                 preferred = 0;
3832                         if (valid != INFINITY_LIFE_TIME) {
3833                                 if (valid > tval)
3834                                         valid -= tval;
3835                                 else
3836                                         valid = 0;
3837                         }
3838                 }
3839         } else {
3840                 preferred = INFINITY_LIFE_TIME;
3841                 valid = INFINITY_LIFE_TIME;
3842         }
3843
3844         if (!ipv6_addr_any(&ifa->peer_addr)) {
3845                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3846                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3847                         goto error;
3848         } else
3849                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3850                         goto error;
3851
3852         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3853                 goto error;
3854
3855         return nlmsg_end(skb, nlh);
3856
3857 error:
3858         nlmsg_cancel(skb, nlh);
3859         return -EMSGSIZE;
3860 }
3861
3862 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3863                                 u32 portid, u32 seq, int event, u16 flags)
3864 {
3865         struct nlmsghdr  *nlh;
3866         u8 scope = RT_SCOPE_UNIVERSE;
3867         int ifindex = ifmca->idev->dev->ifindex;
3868
3869         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3870                 scope = RT_SCOPE_SITE;
3871
3872         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3873         if (nlh == NULL)
3874                 return -EMSGSIZE;
3875
3876         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3877         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3878             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3879                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3880                 nlmsg_cancel(skb, nlh);
3881                 return -EMSGSIZE;
3882         }
3883
3884         return nlmsg_end(skb, nlh);
3885 }
3886
3887 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3888                                 u32 portid, u32 seq, int event, unsigned int flags)
3889 {
3890         struct nlmsghdr  *nlh;
3891         u8 scope = RT_SCOPE_UNIVERSE;
3892         int ifindex = ifaca->aca_idev->dev->ifindex;
3893
3894         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3895                 scope = RT_SCOPE_SITE;
3896
3897         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3898         if (nlh == NULL)
3899                 return -EMSGSIZE;
3900
3901         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3902         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3903             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3904                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3905                 nlmsg_cancel(skb, nlh);
3906                 return -EMSGSIZE;
3907         }
3908
3909         return nlmsg_end(skb, nlh);
3910 }
3911
3912 enum addr_type_t {
3913         UNICAST_ADDR,
3914         MULTICAST_ADDR,
3915         ANYCAST_ADDR,
3916 };
3917
3918 /* called with rcu_read_lock() */
3919 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3920                           struct netlink_callback *cb, enum addr_type_t type,
3921                           int s_ip_idx, int *p_ip_idx)
3922 {
3923         struct ifmcaddr6 *ifmca;
3924         struct ifacaddr6 *ifaca;
3925         int err = 1;
3926         int ip_idx = *p_ip_idx;
3927
3928         read_lock_bh(&idev->lock);
3929         switch (type) {
3930         case UNICAST_ADDR: {
3931                 struct inet6_ifaddr *ifa;
3932
3933                 /* unicast address incl. temp addr */
3934                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3935                         if (++ip_idx < s_ip_idx)
3936                                 continue;
3937                         err = inet6_fill_ifaddr(skb, ifa,
3938                                                 NETLINK_CB(cb->skb).portid,
3939                                                 cb->nlh->nlmsg_seq,
3940                                                 RTM_NEWADDR,
3941                                                 NLM_F_MULTI);
3942                         if (err <= 0)
3943                                 break;
3944                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3945                 }
3946                 break;
3947         }
3948         case MULTICAST_ADDR:
3949                 /* multicast address */
3950                 for (ifmca = idev->mc_list; ifmca;
3951                      ifmca = ifmca->next, ip_idx++) {
3952                         if (ip_idx < s_ip_idx)
3953                                 continue;
3954                         err = inet6_fill_ifmcaddr(skb, ifmca,
3955                                                   NETLINK_CB(cb->skb).portid,
3956                                                   cb->nlh->nlmsg_seq,
3957                                                   RTM_GETMULTICAST,
3958                                                   NLM_F_MULTI);
3959                         if (err <= 0)
3960                                 break;
3961                 }
3962                 break;
3963         case ANYCAST_ADDR:
3964                 /* anycast address */
3965                 for (ifaca = idev->ac_list; ifaca;
3966                      ifaca = ifaca->aca_next, ip_idx++) {
3967                         if (ip_idx < s_ip_idx)
3968                                 continue;
3969                         err = inet6_fill_ifacaddr(skb, ifaca,
3970                                                   NETLINK_CB(cb->skb).portid,
3971                                                   cb->nlh->nlmsg_seq,
3972                                                   RTM_GETANYCAST,
3973                                                   NLM_F_MULTI);
3974                         if (err <= 0)
3975                                 break;
3976                 }
3977                 break;
3978         default:
3979                 break;
3980         }
3981         read_unlock_bh(&idev->lock);
3982         *p_ip_idx = ip_idx;
3983         return err;
3984 }
3985
3986 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3987                            enum addr_type_t type)
3988 {
3989         struct net *net = sock_net(skb->sk);
3990         int h, s_h;
3991         int idx, ip_idx;
3992         int s_idx, s_ip_idx;
3993         struct net_device *dev;
3994         struct inet6_dev *idev;
3995         struct hlist_head *head;
3996
3997         s_h = cb->args[0];
3998         s_idx = idx = cb->args[1];
3999         s_ip_idx = ip_idx = cb->args[2];
4000
4001         rcu_read_lock();
4002         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4003         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4004                 idx = 0;
4005                 head = &net->dev_index_head[h];
4006                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4007                         if (idx < s_idx)
4008                                 goto cont;
4009                         if (h > s_h || idx > s_idx)
4010                                 s_ip_idx = 0;
4011                         ip_idx = 0;
4012                         idev = __in6_dev_get(dev);
4013                         if (!idev)
4014                                 goto cont;
4015
4016                         if (in6_dump_addrs(idev, skb, cb, type,
4017                                            s_ip_idx, &ip_idx) <= 0)
4018                                 goto done;
4019 cont:
4020                         idx++;
4021                 }
4022         }
4023 done:
4024         rcu_read_unlock();
4025         cb->args[0] = h;
4026         cb->args[1] = idx;
4027         cb->args[2] = ip_idx;
4028
4029         return skb->len;
4030 }
4031
4032 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4033 {
4034         enum addr_type_t type = UNICAST_ADDR;
4035
4036         return inet6_dump_addr(skb, cb, type);
4037 }
4038
4039 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4040 {
4041         enum addr_type_t type = MULTICAST_ADDR;
4042
4043         return inet6_dump_addr(skb, cb, type);
4044 }
4045
4046
4047 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4048 {
4049         enum addr_type_t type = ANYCAST_ADDR;
4050
4051         return inet6_dump_addr(skb, cb, type);
4052 }
4053
4054 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4055 {
4056         struct net *net = sock_net(in_skb->sk);
4057         struct ifaddrmsg *ifm;
4058         struct nlattr *tb[IFA_MAX+1];
4059         struct in6_addr *addr = NULL, *peer;
4060         struct net_device *dev = NULL;
4061         struct inet6_ifaddr *ifa;
4062         struct sk_buff *skb;
4063         int err;
4064
4065         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4066         if (err < 0)
4067                 goto errout;
4068
4069         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4070         if (addr == NULL) {
4071                 err = -EINVAL;
4072                 goto errout;
4073         }
4074
4075         ifm = nlmsg_data(nlh);
4076         if (ifm->ifa_index)
4077                 dev = __dev_get_by_index(net, ifm->ifa_index);
4078
4079         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4080         if (!ifa) {
4081                 err = -EADDRNOTAVAIL;
4082                 goto errout;
4083         }
4084
4085         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4086         if (!skb) {
4087                 err = -ENOBUFS;
4088                 goto errout_ifa;
4089         }
4090
4091         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4092                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4093         if (err < 0) {
4094                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4095                 WARN_ON(err == -EMSGSIZE);
4096                 kfree_skb(skb);
4097                 goto errout_ifa;
4098         }
4099         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4100 errout_ifa:
4101         in6_ifa_put(ifa);
4102 errout:
4103         return err;
4104 }
4105
4106 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4107 {
4108         struct sk_buff *skb;
4109         struct net *net = dev_net(ifa->idev->dev);
4110         int err = -ENOBUFS;
4111
4112         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4113         if (skb == NULL)
4114                 goto errout;
4115
4116         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4117         if (err < 0) {
4118                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4119                 WARN_ON(err == -EMSGSIZE);
4120                 kfree_skb(skb);
4121                 goto errout;
4122         }
4123         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4124         return;
4125 errout:
4126         if (err < 0)
4127                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4128 }
4129
4130 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4131                                 __s32 *array, int bytes)
4132 {
4133         BUG_ON(bytes < (DEVCONF_MAX * 4));
4134
4135         memset(array, 0, bytes);
4136         array[DEVCONF_FORWARDING] = cnf->forwarding;
4137         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4138         array[DEVCONF_MTU6] = cnf->mtu6;
4139         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4140         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4141         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4142         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4143         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4144         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4145                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4146         array[DEVCONF_RTR_SOLICIT_DELAY] =
4147                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4148         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4149         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4150                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4151         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4152                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4153 #ifdef CONFIG_IPV6_PRIVACY
4154         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4155         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4156         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4157         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4158         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4159 #endif
4160         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4161         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4162         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4163 #ifdef CONFIG_IPV6_ROUTER_PREF
4164         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4165         array[DEVCONF_RTR_PROBE_INTERVAL] =
4166                 jiffies_to_msecs(cnf->rtr_probe_interval);
4167 #ifdef CONFIG_IPV6_ROUTE_INFO
4168         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4169 #endif
4170 #endif
4171         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4172         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4173 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4174         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4175 #endif
4176 #ifdef CONFIG_IPV6_MROUTE
4177         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4178 #endif
4179         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4180         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4181         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4182         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4183         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4184 }
4185
4186 static inline size_t inet6_ifla6_size(void)
4187 {
4188         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4189              + nla_total_size(sizeof(struct ifla_cacheinfo))
4190              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4191              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4192              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4193              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4194 }
4195
4196 static inline size_t inet6_if_nlmsg_size(void)
4197 {
4198         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4199                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4200                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4201                + nla_total_size(4) /* IFLA_MTU */
4202                + nla_total_size(4) /* IFLA_LINK */
4203                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4204 }
4205
4206 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4207                                       int items, int bytes)
4208 {
4209         int i;
4210         int pad = bytes - sizeof(u64) * items;
4211         BUG_ON(pad < 0);
4212
4213         /* Use put_unaligned() because stats may not be aligned for u64. */
4214         put_unaligned(items, &stats[0]);
4215         for (i = 1; i < items; i++)
4216                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4217
4218         memset(&stats[items], 0, pad);
4219 }
4220
4221 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4222                                       int items, int bytes, size_t syncpoff)
4223 {
4224         int i;
4225         int pad = bytes - sizeof(u64) * items;
4226         BUG_ON(pad < 0);
4227
4228         /* Use put_unaligned() because stats may not be aligned for u64. */
4229         put_unaligned(items, &stats[0]);
4230         for (i = 1; i < items; i++)
4231                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4232
4233         memset(&stats[items], 0, pad);
4234 }
4235
4236 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4237                              int bytes)
4238 {
4239         switch (attrtype) {
4240         case IFLA_INET6_STATS:
4241                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4242                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4243                 break;
4244         case IFLA_INET6_ICMP6STATS:
4245                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4246                 break;
4247         }
4248 }
4249
4250 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4251 {
4252         struct nlattr *nla;
4253         struct ifla_cacheinfo ci;
4254
4255         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4256                 goto nla_put_failure;
4257         ci.max_reasm_len = IPV6_MAXPLEN;
4258         ci.tstamp = cstamp_delta(idev->tstamp);
4259         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4260         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4261         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4262                 goto nla_put_failure;
4263         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4264         if (nla == NULL)
4265                 goto nla_put_failure;
4266         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4267
4268         /* XXX - MC not implemented */
4269
4270         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4271         if (nla == NULL)
4272                 goto nla_put_failure;
4273         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4274
4275         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4276         if (nla == NULL)
4277                 goto nla_put_failure;
4278         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4279
4280         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4281         if (nla == NULL)
4282                 goto nla_put_failure;
4283         read_lock_bh(&idev->lock);
4284         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4285         read_unlock_bh(&idev->lock);
4286
4287         return 0;
4288
4289 nla_put_failure:
4290         return -EMSGSIZE;
4291 }
4292
4293 static size_t inet6_get_link_af_size(const struct net_device *dev)
4294 {
4295         if (!__in6_dev_get(dev))
4296                 return 0;
4297
4298         return inet6_ifla6_size();
4299 }
4300
4301 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4302 {
4303         struct inet6_dev *idev = __in6_dev_get(dev);
4304
4305         if (!idev)
4306                 return -ENODATA;
4307
4308         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4309                 return -EMSGSIZE;
4310
4311         return 0;
4312 }
4313
4314 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4315 {
4316         struct inet6_ifaddr *ifp;
4317         struct net_device *dev = idev->dev;
4318         bool update_rs = false;
4319         struct in6_addr ll_addr;
4320
4321         if (token == NULL)
4322                 return -EINVAL;
4323         if (ipv6_addr_any(token))
4324                 return -EINVAL;
4325         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4326                 return -EINVAL;
4327         if (!ipv6_accept_ra(idev))
4328                 return -EINVAL;
4329         if (idev->cnf.rtr_solicits <= 0)
4330                 return -EINVAL;
4331
4332         write_lock_bh(&idev->lock);
4333
4334         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4335         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4336
4337         write_unlock_bh(&idev->lock);
4338
4339         if (!idev->dead && (idev->if_flags & IF_READY) &&
4340             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4341                              IFA_F_OPTIMISTIC)) {
4342
4343                 /* If we're not ready, then normal ifup will take care
4344                  * of this. Otherwise, we need to request our rs here.
4345                  */
4346                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4347                 update_rs = true;
4348         }
4349
4350         write_lock_bh(&idev->lock);
4351
4352         if (update_rs) {
4353                 idev->if_flags |= IF_RS_SENT;
4354                 idev->rs_probes = 1;
4355                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4356         }
4357
4358         /* Well, that's kinda nasty ... */
4359         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4360                 spin_lock(&ifp->lock);
4361                 if (ifp->tokenized) {
4362                         ifp->valid_lft = 0;
4363                         ifp->prefered_lft = 0;
4364                 }
4365                 spin_unlock(&ifp->lock);
4366         }
4367
4368         write_unlock_bh(&idev->lock);
4369         addrconf_verify(0);
4370         return 0;
4371 }
4372
4373 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4374 {
4375         int err = -EINVAL;
4376         struct inet6_dev *idev = __in6_dev_get(dev);
4377         struct nlattr *tb[IFLA_INET6_MAX + 1];
4378
4379         if (!idev)
4380                 return -EAFNOSUPPORT;
4381
4382         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4383                 BUG();
4384
4385         if (tb[IFLA_INET6_TOKEN])
4386                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4387
4388         return err;
4389 }
4390
4391 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4392                              u32 portid, u32 seq, int event, unsigned int flags)
4393 {
4394         struct net_device *dev = idev->dev;
4395         struct ifinfomsg *hdr;
4396         struct nlmsghdr *nlh;
4397         void *protoinfo;
4398
4399         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4400         if (nlh == NULL)
4401                 return -EMSGSIZE;
4402
4403         hdr = nlmsg_data(nlh);
4404         hdr->ifi_family = AF_INET6;
4405         hdr->__ifi_pad = 0;
4406         hdr->ifi_type = dev->type;
4407         hdr->ifi_index = dev->ifindex;
4408         hdr->ifi_flags = dev_get_flags(dev);
4409         hdr->ifi_change = 0;
4410
4411         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4412             (dev->addr_len &&
4413              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4414             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4415             (dev->ifindex != dev->iflink &&
4416              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4417                 goto nla_put_failure;
4418         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4419         if (protoinfo == NULL)
4420                 goto nla_put_failure;
4421
4422         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4423                 goto nla_put_failure;
4424
4425         nla_nest_end(skb, protoinfo);
4426         return nlmsg_end(skb, nlh);
4427
4428 nla_put_failure:
4429         nlmsg_cancel(skb, nlh);
4430         return -EMSGSIZE;
4431 }
4432
4433 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4434 {
4435         struct net *net = sock_net(skb->sk);
4436         int h, s_h;
4437         int idx = 0, s_idx;
4438         struct net_device *dev;
4439         struct inet6_dev *idev;
4440         struct hlist_head *head;
4441
4442         s_h = cb->args[0];
4443         s_idx = cb->args[1];
4444
4445         rcu_read_lock();
4446         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4447                 idx = 0;
4448                 head = &net->dev_index_head[h];
4449                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4450                         if (idx < s_idx)
4451                                 goto cont;
4452                         idev = __in6_dev_get(dev);
4453                         if (!idev)
4454                                 goto cont;
4455                         if (inet6_fill_ifinfo(skb, idev,
4456                                               NETLINK_CB(cb->skb).portid,
4457                                               cb->nlh->nlmsg_seq,
4458                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4459                                 goto out;
4460 cont:
4461                         idx++;
4462                 }
4463         }
4464 out:
4465         rcu_read_unlock();
4466         cb->args[1] = idx;
4467         cb->args[0] = h;
4468
4469         return skb->len;
4470 }
4471
4472 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4473 {
4474         struct sk_buff *skb;
4475         struct net *net = dev_net(idev->dev);
4476         int err = -ENOBUFS;
4477
4478         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4479         if (skb == NULL)
4480                 goto errout;
4481
4482         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4483         if (err < 0) {
4484                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4485                 WARN_ON(err == -EMSGSIZE);
4486                 kfree_skb(skb);
4487                 goto errout;
4488         }
4489         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4490         return;
4491 errout:
4492         if (err < 0)
4493                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4494 }
4495
4496 static inline size_t inet6_prefix_nlmsg_size(void)
4497 {
4498         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4499                + nla_total_size(sizeof(struct in6_addr))
4500                + nla_total_size(sizeof(struct prefix_cacheinfo));
4501 }
4502
4503 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4504                              struct prefix_info *pinfo, u32 portid, u32 seq,
4505                              int event, unsigned int flags)
4506 {
4507         struct prefixmsg *pmsg;
4508         struct nlmsghdr *nlh;
4509         struct prefix_cacheinfo ci;
4510
4511         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4512         if (nlh == NULL)
4513                 return -EMSGSIZE;
4514
4515         pmsg = nlmsg_data(nlh);
4516         pmsg->prefix_family = AF_INET6;
4517         pmsg->prefix_pad1 = 0;
4518         pmsg->prefix_pad2 = 0;
4519         pmsg->prefix_ifindex = idev->dev->ifindex;
4520         pmsg->prefix_len = pinfo->prefix_len;
4521         pmsg->prefix_type = pinfo->type;
4522         pmsg->prefix_pad3 = 0;
4523         pmsg->prefix_flags = 0;
4524         if (pinfo->onlink)
4525                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4526         if (pinfo->autoconf)
4527                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4528
4529         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4530                 goto nla_put_failure;
4531         ci.preferred_time = ntohl(pinfo->prefered);
4532         ci.valid_time = ntohl(pinfo->valid);
4533         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4534                 goto nla_put_failure;
4535         return nlmsg_end(skb, nlh);
4536
4537 nla_put_failure:
4538         nlmsg_cancel(skb, nlh);
4539         return -EMSGSIZE;
4540 }
4541
4542 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4543                          struct prefix_info *pinfo)
4544 {
4545         struct sk_buff *skb;
4546         struct net *net = dev_net(idev->dev);
4547         int err = -ENOBUFS;
4548
4549         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4550         if (skb == NULL)
4551                 goto errout;
4552
4553         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4554         if (err < 0) {
4555                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4556                 WARN_ON(err == -EMSGSIZE);
4557                 kfree_skb(skb);
4558                 goto errout;
4559         }
4560         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4561         return;
4562 errout:
4563         if (err < 0)
4564                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4565 }
4566
4567 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4568 {
4569         write_lock_bh(&ifp->idev->lock);
4570         spin_lock(&ifp->lock);
4571         if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4572                             IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4573             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4574                 ifp->idev->valid_ll_addr_cnt += count;
4575         WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4576         spin_unlock(&ifp->lock);
4577         write_unlock_bh(&ifp->idev->lock);
4578 }
4579
4580 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4581 {
4582         struct net *net = dev_net(ifp->idev->dev);
4583
4584         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4585
4586         switch (event) {
4587         case RTM_NEWADDR:
4588                 update_valid_ll_addr_cnt(ifp, 1);
4589
4590                 /*
4591                  * If the address was optimistic
4592                  * we inserted the route at the start of
4593                  * our DAD process, so we don't need
4594                  * to do it again
4595                  */
4596                 if (!(ifp->rt->rt6i_node))
4597                         ip6_ins_rt(ifp->rt);
4598                 if (ifp->idev->cnf.forwarding)
4599                         addrconf_join_anycast(ifp);
4600                 if (!ipv6_addr_any(&ifp->peer_addr))
4601                         addrconf_prefix_route(&ifp->peer_addr, 128,
4602                                               ifp->idev->dev, 0, 0);
4603                 break;
4604         case RTM_DELADDR:
4605                 update_valid_ll_addr_cnt(ifp, -1);
4606
4607                 if (ifp->idev->cnf.forwarding)
4608                         addrconf_leave_anycast(ifp);
4609                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4610                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4611                         struct rt6_info *rt;
4612                         struct net_device *dev = ifp->idev->dev;
4613
4614                         rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4615                                         dev->ifindex, 1);
4616                         if (rt) {
4617                                 dst_hold(&rt->dst);
4618                                 if (ip6_del_rt(rt))
4619                                         dst_free(&rt->dst);
4620                         }
4621                 }
4622                 dst_hold(&ifp->rt->dst);
4623
4624                 if (ip6_del_rt(ifp->rt))
4625                         dst_free(&ifp->rt->dst);
4626                 break;
4627         }
4628         atomic_inc(&net->ipv6.dev_addr_genid);
4629         rt_genid_bump_ipv6(net);
4630 }
4631
4632 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4633 {
4634         rcu_read_lock_bh();
4635         if (likely(ifp->idev->dead == 0))
4636                 __ipv6_ifa_notify(event, ifp);
4637         rcu_read_unlock_bh();
4638 }
4639
4640 #ifdef CONFIG_SYSCTL
4641
4642 static
4643 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4644                            void __user *buffer, size_t *lenp, loff_t *ppos)
4645 {
4646         int *valp = ctl->data;
4647         int val = *valp;
4648         loff_t pos = *ppos;
4649         struct ctl_table lctl;
4650         int ret;
4651
4652         /*
4653          * ctl->data points to idev->cnf.forwarding, we should
4654          * not modify it until we get the rtnl lock.
4655          */
4656         lctl = *ctl;
4657         lctl.data = &val;
4658
4659         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4660
4661         if (write)
4662                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4663         if (ret)
4664                 *ppos = pos;
4665         return ret;
4666 }
4667
4668 static void dev_disable_change(struct inet6_dev *idev)
4669 {
4670         struct netdev_notifier_info info;
4671
4672         if (!idev || !idev->dev)
4673                 return;
4674
4675         netdev_notifier_info_init(&info, idev->dev);
4676         if (idev->cnf.disable_ipv6)
4677                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4678         else
4679                 addrconf_notify(NULL, NETDEV_UP, &info);
4680 }
4681
4682 static void addrconf_disable_change(struct net *net, __s32 newf)
4683 {
4684         struct net_device *dev;
4685         struct inet6_dev *idev;
4686
4687         rcu_read_lock();
4688         for_each_netdev_rcu(net, dev) {
4689                 idev = __in6_dev_get(dev);
4690                 if (idev) {
4691                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4692                         idev->cnf.disable_ipv6 = newf;
4693                         if (changed)
4694                                 dev_disable_change(idev);
4695                 }
4696         }
4697         rcu_read_unlock();
4698 }
4699
4700 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4701 {
4702         struct net *net;
4703         int old;
4704
4705         if (!rtnl_trylock())
4706                 return restart_syscall();
4707
4708         net = (struct net *)table->extra2;
4709         old = *p;
4710         *p = newf;
4711
4712         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4713                 rtnl_unlock();
4714                 return 0;
4715         }
4716
4717         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4718                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4719                 addrconf_disable_change(net, newf);
4720         } else if ((!newf) ^ (!old))
4721                 dev_disable_change((struct inet6_dev *)table->extra1);
4722
4723         rtnl_unlock();
4724         return 0;
4725 }
4726
4727 static
4728 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4729                             void __user *buffer, size_t *lenp, loff_t *ppos)
4730 {
4731         int *valp = ctl->data;
4732         int val = *valp;
4733         loff_t pos = *ppos;
4734         struct ctl_table lctl;
4735         int ret;
4736
4737         /*
4738          * ctl->data points to idev->cnf.disable_ipv6, we should
4739          * not modify it until we get the rtnl lock.
4740          */
4741         lctl = *ctl;
4742         lctl.data = &val;
4743
4744         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4745
4746         if (write)
4747                 ret = addrconf_disable_ipv6(ctl, valp, val);
4748         if (ret)
4749                 *ppos = pos;
4750         return ret;
4751 }
4752
4753 static struct addrconf_sysctl_table
4754 {
4755         struct ctl_table_header *sysctl_header;
4756         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4757 } addrconf_sysctl __read_mostly = {
4758         .sysctl_header = NULL,
4759         .addrconf_vars = {
4760                 {
4761                         .procname       = "forwarding",
4762                         .data           = &ipv6_devconf.forwarding,
4763                         .maxlen         = sizeof(int),
4764                         .mode           = 0644,
4765                         .proc_handler   = addrconf_sysctl_forward,
4766                 },
4767                 {
4768                         .procname       = "hop_limit",
4769                         .data           = &ipv6_devconf.hop_limit,
4770                         .maxlen         = sizeof(int),
4771                         .mode           = 0644,
4772                         .proc_handler   = proc_dointvec,
4773                 },
4774                 {
4775                         .procname       = "mtu",
4776                         .data           = &ipv6_devconf.mtu6,
4777                         .maxlen         = sizeof(int),
4778                         .mode           = 0644,
4779                         .proc_handler   = proc_dointvec,
4780                 },
4781                 {
4782                         .procname       = "accept_ra",
4783                         .data           = &ipv6_devconf.accept_ra,
4784                         .maxlen         = sizeof(int),
4785                         .mode           = 0644,
4786                         .proc_handler   = proc_dointvec,
4787                 },
4788                 {
4789                         .procname       = "accept_redirects",
4790                         .data           = &ipv6_devconf.accept_redirects,
4791                         .maxlen         = sizeof(int),
4792                         .mode           = 0644,
4793                         .proc_handler   = proc_dointvec,
4794                 },
4795                 {
4796                         .procname       = "autoconf",
4797                         .data           = &ipv6_devconf.autoconf,
4798                         .maxlen         = sizeof(int),
4799                         .mode           = 0644,
4800                         .proc_handler   = proc_dointvec,
4801                 },
4802                 {
4803                         .procname       = "dad_transmits",
4804                         .data           = &ipv6_devconf.dad_transmits,
4805                         .maxlen         = sizeof(int),
4806                         .mode           = 0644,
4807                         .proc_handler   = proc_dointvec,
4808                 },
4809                 {
4810                         .procname       = "router_solicitations",
4811                         .data           = &ipv6_devconf.rtr_solicits,
4812                         .maxlen         = sizeof(int),
4813                         .mode           = 0644,
4814                         .proc_handler   = proc_dointvec,
4815                 },
4816                 {
4817                         .procname       = "router_solicitation_interval",
4818                         .data           = &ipv6_devconf.rtr_solicit_interval,
4819                         .maxlen         = sizeof(int),
4820                         .mode           = 0644,
4821                         .proc_handler   = proc_dointvec_jiffies,
4822                 },
4823                 {
4824                         .procname       = "router_solicitation_delay",
4825                         .data           = &ipv6_devconf.rtr_solicit_delay,
4826                         .maxlen         = sizeof(int),
4827                         .mode           = 0644,
4828                         .proc_handler   = proc_dointvec_jiffies,
4829                 },
4830                 {
4831                         .procname       = "force_mld_version",
4832                         .data           = &ipv6_devconf.force_mld_version,
4833                         .maxlen         = sizeof(int),
4834                         .mode           = 0644,
4835                         .proc_handler   = proc_dointvec,
4836                 },
4837                 {
4838                         .procname       = "mldv1_unsolicited_report_interval",
4839                         .data           =
4840                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
4841                         .maxlen         = sizeof(int),
4842                         .mode           = 0644,
4843                         .proc_handler   = proc_dointvec_ms_jiffies,
4844                 },
4845                 {
4846                         .procname       = "mldv2_unsolicited_report_interval",
4847                         .data           =
4848                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
4849                         .maxlen         = sizeof(int),
4850                         .mode           = 0644,
4851                         .proc_handler   = proc_dointvec_ms_jiffies,
4852                 },
4853 #ifdef CONFIG_IPV6_PRIVACY
4854                 {
4855                         .procname       = "use_tempaddr",
4856                         .data           = &ipv6_devconf.use_tempaddr,
4857                         .maxlen         = sizeof(int),
4858                         .mode           = 0644,
4859                         .proc_handler   = proc_dointvec,
4860                 },
4861                 {
4862                         .procname       = "temp_valid_lft",
4863                         .data           = &ipv6_devconf.temp_valid_lft,
4864                         .maxlen         = sizeof(int),
4865                         .mode           = 0644,
4866                         .proc_handler   = proc_dointvec,
4867                 },
4868                 {
4869                         .procname       = "temp_prefered_lft",
4870                         .data           = &ipv6_devconf.temp_prefered_lft,
4871                         .maxlen         = sizeof(int),
4872                         .mode           = 0644,
4873                         .proc_handler   = proc_dointvec,
4874                 },
4875                 {
4876                         .procname       = "regen_max_retry",
4877                         .data           = &ipv6_devconf.regen_max_retry,
4878                         .maxlen         = sizeof(int),
4879                         .mode           = 0644,
4880                         .proc_handler   = proc_dointvec,
4881                 },
4882                 {
4883                         .procname       = "max_desync_factor",
4884                         .data           = &ipv6_devconf.max_desync_factor,
4885                         .maxlen         = sizeof(int),
4886                         .mode           = 0644,
4887                         .proc_handler   = proc_dointvec,
4888                 },
4889 #endif
4890                 {
4891                         .procname       = "max_addresses",
4892                         .data           = &ipv6_devconf.max_addresses,
4893                         .maxlen         = sizeof(int),
4894                         .mode           = 0644,
4895                         .proc_handler   = proc_dointvec,
4896                 },
4897                 {
4898                         .procname       = "accept_ra_defrtr",
4899                         .data           = &ipv6_devconf.accept_ra_defrtr,
4900                         .maxlen         = sizeof(int),
4901                         .mode           = 0644,
4902                         .proc_handler   = proc_dointvec,
4903                 },
4904                 {
4905                         .procname       = "accept_ra_pinfo",
4906                         .data           = &ipv6_devconf.accept_ra_pinfo,
4907                         .maxlen         = sizeof(int),
4908                         .mode           = 0644,
4909                         .proc_handler   = proc_dointvec,
4910                 },
4911 #ifdef CONFIG_IPV6_ROUTER_PREF
4912                 {
4913                         .procname       = "accept_ra_rtr_pref",
4914                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4915                         .maxlen         = sizeof(int),
4916                         .mode           = 0644,
4917                         .proc_handler   = proc_dointvec,
4918                 },
4919                 {
4920                         .procname       = "router_probe_interval",
4921                         .data           = &ipv6_devconf.rtr_probe_interval,
4922                         .maxlen         = sizeof(int),
4923                         .mode           = 0644,
4924                         .proc_handler   = proc_dointvec_jiffies,
4925                 },
4926 #ifdef CONFIG_IPV6_ROUTE_INFO
4927                 {
4928                         .procname       = "accept_ra_rt_info_max_plen",
4929                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4930                         .maxlen         = sizeof(int),
4931                         .mode           = 0644,
4932                         .proc_handler   = proc_dointvec,
4933                 },
4934 #endif
4935 #endif
4936                 {
4937                         .procname       = "proxy_ndp",
4938                         .data           = &ipv6_devconf.proxy_ndp,
4939                         .maxlen         = sizeof(int),
4940                         .mode           = 0644,
4941                         .proc_handler   = proc_dointvec,
4942                 },
4943                 {
4944                         .procname       = "accept_source_route",
4945                         .data           = &ipv6_devconf.accept_source_route,
4946                         .maxlen         = sizeof(int),
4947                         .mode           = 0644,
4948                         .proc_handler   = proc_dointvec,
4949                 },
4950 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4951                 {
4952                         .procname       = "optimistic_dad",
4953                         .data           = &ipv6_devconf.optimistic_dad,
4954                         .maxlen         = sizeof(int),
4955                         .mode           = 0644,
4956                         .proc_handler   = proc_dointvec,
4957
4958                 },
4959 #endif
4960 #ifdef CONFIG_IPV6_MROUTE
4961                 {
4962                         .procname       = "mc_forwarding",
4963                         .data           = &ipv6_devconf.mc_forwarding,
4964                         .maxlen         = sizeof(int),
4965                         .mode           = 0444,
4966                         .proc_handler   = proc_dointvec,
4967                 },
4968 #endif
4969                 {
4970                         .procname       = "disable_ipv6",
4971                         .data           = &ipv6_devconf.disable_ipv6,
4972                         .maxlen         = sizeof(int),
4973                         .mode           = 0644,
4974                         .proc_handler   = addrconf_sysctl_disable,
4975                 },
4976                 {
4977                         .procname       = "accept_dad",
4978                         .data           = &ipv6_devconf.accept_dad,
4979                         .maxlen         = sizeof(int),
4980                         .mode           = 0644,
4981                         .proc_handler   = proc_dointvec,
4982                 },
4983                 {
4984                         .procname       = "force_tllao",
4985                         .data           = &ipv6_devconf.force_tllao,
4986                         .maxlen         = sizeof(int),
4987                         .mode           = 0644,
4988                         .proc_handler   = proc_dointvec
4989                 },
4990                 {
4991                         .procname       = "ndisc_notify",
4992                         .data           = &ipv6_devconf.ndisc_notify,
4993                         .maxlen         = sizeof(int),
4994                         .mode           = 0644,
4995                         .proc_handler   = proc_dointvec
4996                 },
4997                 {
4998                         .procname       = "suppress_frag_ndisc",
4999                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5000                         .maxlen         = sizeof(int),
5001                         .mode           = 0644,
5002                         .proc_handler   = proc_dointvec
5003                 },
5004                 {
5005                         /* sentinel */
5006                 }
5007         },
5008 };
5009
5010 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5011                 struct inet6_dev *idev, struct ipv6_devconf *p)
5012 {
5013         int i;
5014         struct addrconf_sysctl_table *t;
5015         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5016
5017         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5018         if (t == NULL)
5019                 goto out;
5020
5021         for (i = 0; t->addrconf_vars[i].data; i++) {
5022                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5023                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5024                 t->addrconf_vars[i].extra2 = net;
5025         }
5026
5027         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5028
5029         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5030         if (t->sysctl_header == NULL)
5031                 goto free;
5032
5033         p->sysctl = t;
5034         return 0;
5035
5036 free:
5037         kfree(t);
5038 out:
5039         return -ENOBUFS;
5040 }
5041
5042 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5043 {
5044         struct addrconf_sysctl_table *t;
5045
5046         if (p->sysctl == NULL)
5047                 return;
5048
5049         t = p->sysctl;
5050         p->sysctl = NULL;
5051         unregister_net_sysctl_table(t->sysctl_header);
5052         kfree(t);
5053 }
5054
5055 static void addrconf_sysctl_register(struct inet6_dev *idev)
5056 {
5057         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5058                               &ndisc_ifinfo_sysctl_change);
5059         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5060                                         idev, &idev->cnf);
5061 }
5062
5063 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5064 {
5065         __addrconf_sysctl_unregister(&idev->cnf);
5066         neigh_sysctl_unregister(idev->nd_parms);
5067 }
5068
5069
5070 #endif
5071
5072 static int __net_init addrconf_init_net(struct net *net)
5073 {
5074         int err = -ENOMEM;
5075         struct ipv6_devconf *all, *dflt;
5076
5077         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5078         if (all == NULL)
5079                 goto err_alloc_all;
5080
5081         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5082         if (dflt == NULL)
5083                 goto err_alloc_dflt;
5084
5085         /* these will be inherited by all namespaces */
5086         dflt->autoconf = ipv6_defaults.autoconf;
5087         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5088
5089         net->ipv6.devconf_all = all;
5090         net->ipv6.devconf_dflt = dflt;
5091
5092 #ifdef CONFIG_SYSCTL
5093         err = __addrconf_sysctl_register(net, "all", NULL, all);
5094         if (err < 0)
5095                 goto err_reg_all;
5096
5097         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5098         if (err < 0)
5099                 goto err_reg_dflt;
5100 #endif
5101         return 0;
5102
5103 #ifdef CONFIG_SYSCTL
5104 err_reg_dflt:
5105         __addrconf_sysctl_unregister(all);
5106 err_reg_all:
5107         kfree(dflt);
5108 #endif
5109 err_alloc_dflt:
5110         kfree(all);
5111 err_alloc_all:
5112         return err;
5113 }
5114
5115 static void __net_exit addrconf_exit_net(struct net *net)
5116 {
5117 #ifdef CONFIG_SYSCTL
5118         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5119         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5120 #endif
5121         if (!net_eq(net, &init_net)) {
5122                 kfree(net->ipv6.devconf_dflt);
5123                 kfree(net->ipv6.devconf_all);
5124         }
5125 }
5126
5127 static struct pernet_operations addrconf_ops = {
5128         .init = addrconf_init_net,
5129         .exit = addrconf_exit_net,
5130 };
5131
5132 static struct rtnl_af_ops inet6_ops = {
5133         .family           = AF_INET6,
5134         .fill_link_af     = inet6_fill_link_af,
5135         .get_link_af_size = inet6_get_link_af_size,
5136         .set_link_af      = inet6_set_link_af,
5137 };
5138
5139 /*
5140  *      Init / cleanup code
5141  */
5142
5143 int __init addrconf_init(void)
5144 {
5145         int i, err;
5146
5147         err = ipv6_addr_label_init();
5148         if (err < 0) {
5149                 pr_crit("%s: cannot initialize default policy table: %d\n",
5150                         __func__, err);
5151                 goto out;
5152         }
5153
5154         err = register_pernet_subsys(&addrconf_ops);
5155         if (err < 0)
5156                 goto out_addrlabel;
5157
5158         /* The addrconf netdev notifier requires that loopback_dev
5159          * has it's ipv6 private information allocated and setup
5160          * before it can bring up and give link-local addresses
5161          * to other devices which are up.
5162          *
5163          * Unfortunately, loopback_dev is not necessarily the first
5164          * entry in the global dev_base list of net devices.  In fact,
5165          * it is likely to be the very last entry on that list.
5166          * So this causes the notifier registry below to try and
5167          * give link-local addresses to all devices besides loopback_dev
5168          * first, then loopback_dev, which cases all the non-loopback_dev
5169          * devices to fail to get a link-local address.
5170          *
5171          * So, as a temporary fix, allocate the ipv6 structure for
5172          * loopback_dev first by hand.
5173          * Longer term, all of the dependencies ipv6 has upon the loopback
5174          * device and it being up should be removed.
5175          */
5176         rtnl_lock();
5177         if (!ipv6_add_dev(init_net.loopback_dev))
5178                 err = -ENOMEM;
5179         rtnl_unlock();
5180         if (err)
5181                 goto errlo;
5182
5183         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5184                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5185
5186         register_netdevice_notifier(&ipv6_dev_notf);
5187
5188         addrconf_verify(0);
5189
5190         err = rtnl_af_register(&inet6_ops);
5191         if (err < 0)
5192                 goto errout_af;
5193
5194         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5195                               NULL);
5196         if (err < 0)
5197                 goto errout;
5198
5199         /* Only the first call to __rtnl_register can fail */
5200         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5201         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5202         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5203                         inet6_dump_ifaddr, NULL);
5204         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5205                         inet6_dump_ifmcaddr, NULL);
5206         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5207                         inet6_dump_ifacaddr, NULL);
5208         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5209                         inet6_netconf_dump_devconf, NULL);
5210
5211         ipv6_addr_label_rtnl_register();
5212
5213         return 0;
5214 errout:
5215         rtnl_af_unregister(&inet6_ops);
5216 errout_af:
5217         unregister_netdevice_notifier(&ipv6_dev_notf);
5218 errlo:
5219         unregister_pernet_subsys(&addrconf_ops);
5220 out_addrlabel:
5221         ipv6_addr_label_cleanup();
5222 out:
5223         return err;
5224 }
5225
5226 void addrconf_cleanup(void)
5227 {
5228         struct net_device *dev;
5229         int i;
5230
5231         unregister_netdevice_notifier(&ipv6_dev_notf);
5232         unregister_pernet_subsys(&addrconf_ops);
5233         ipv6_addr_label_cleanup();
5234
5235         rtnl_lock();
5236
5237         __rtnl_af_unregister(&inet6_ops);
5238
5239         /* clean dev list */
5240         for_each_netdev(&init_net, dev) {
5241                 if (__in6_dev_get(dev) == NULL)
5242                         continue;
5243                 addrconf_ifdown(dev, 1);
5244         }
5245         addrconf_ifdown(init_net.loopback_dev, 2);
5246
5247         /*
5248          *      Check hash table.
5249          */
5250         spin_lock_bh(&addrconf_hash_lock);
5251         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5252                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5253         spin_unlock_bh(&addrconf_hash_lock);
5254
5255         del_timer(&addr_chk_timer);
5256         rtnl_unlock();
5257 }