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