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/inet.h>
  50 #include <linux/in6.h>
  51 #include <linux/netdevice.h>
  52 #include <linux/if_addr.h>
  53 #include <linux/if_arp.h>
  54 #include <linux/if_arcnet.h>
  55 #include <linux/if_infiniband.h>
  56 #include <linux/route.h>
  57 #include <linux/inetdevice.h>
  58 #include <linux/init.h>
  59 #include <linux/slab.h>
  60 #ifdef CONFIG_SYSCTL
  61 #include <linux/sysctl.h>
  62 #endif
  63 #include <linux/capability.h>
  64 #include <linux/delay.h>
  65 #include <linux/notifier.h>
  66 #include <linux/string.h>
  67 #include <linux/hash.h>
  68
  69 #include <net/net_namespace.h>
  70 #include <net/sock.h>
  71 #include <net/snmp.h>
  72
  73 #include <net/6lowpan.h>
  74 #include <net/firewire.h>
  75 #include <net/ipv6.h>
  76 #include <net/protocol.h>
  77 #include <net/ndisc.h>
  78 #include <net/ip6_route.h>
  79 #include <net/addrconf.h>
  80 #include <net/tcp.h>
  81 #include <net/ip.h>
  82 #include <net/netlink.h>
  83 #include <net/pkt_sched.h>
  84 #include <net/l3mdev.h>
  85 #include <linux/if_tunnel.h>
  86 #include <linux/rtnetlink.h>
  87 #include <linux/netconf.h>
  88 #include <linux/random.h>
  89 #include <linux/uaccess.h>
  90 #include <asm/unaligned.h>
  91
  92 #include <linux/proc_fs.h>
  93 #include <linux/seq_file.h>
  94 #include <linux/export.h>
  95
  96 /* Set to 3 to get tracing... */
  97 #define ACONF_DEBUG 2
  98
  99 #if ACONF_DEBUG >= 3
 100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 101 #else
 102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 103 #endif
 104
 105 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 106
 107 #define IPV6_MAX_STRLEN \
 108         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 109
 110 static inline u32 cstamp_delta(unsigned long cstamp)
 111 {
 112         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 113 }
 114
 115 #ifdef CONFIG_SYSCTL
 116 static int addrconf_sysctl_register(struct inet6_dev *idev);
 117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 118 #else
 119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 120 {
 121         return 0;
 122 }
 123
 124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 125 {
 126 }
 127 #endif
 128
 129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 131 static void ipv6_regen_rndid(unsigned long data);
 132
 133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134 static int ipv6_count_addresses(struct inet6_dev *idev);
 135 static int ipv6_generate_stable_address(struct in6_addr *addr,
 136                                         u8 dad_count,
 137                                         const struct inet6_dev *idev);
 138
 139 /*
 140  *      Configured unicast address hash table
 141  */
 142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 143 static DEFINE_SPINLOCK(addrconf_hash_lock);
 144
 145 static void addrconf_verify(void);
 146 static void addrconf_verify_rtnl(void);
 147 static void addrconf_verify_work(struct work_struct *);
 148
 149 static struct workqueue_struct *addrconf_wq;
 150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 151
 152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 154
 155 static void addrconf_type_change(struct net_device *dev,
 156                                  unsigned long event);
 157 static int addrconf_ifdown(struct net_device *dev, int how);
 158
 159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 160                                                   int plen,
 161                                                   const struct net_device *dev,
 162                                                   u32 flags, u32 noflags);
 163
 164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 165 static void addrconf_dad_work(struct work_struct *w);
 166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 167 static void addrconf_dad_run(struct inet6_dev *idev);
 168 static void addrconf_rs_timer(unsigned long data);
 169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 171
 172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 173                                 struct prefix_info *pinfo);
 174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 175                                struct net_device *dev);
 176
 177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 178         .forwarding             = 0,
 179         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 180         .mtu6                   = IPV6_MIN_MTU,
 181         .accept_ra              = 1,
 182         .accept_redirects       = 1,
 183         .autoconf               = 1,
 184         .force_mld_version      = 0,
 185         .mldv1_unsolicited_report_interval = 10 * HZ,
 186         .mldv2_unsolicited_report_interval = HZ,
 187         .dad_transmits          = 1,
 188         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 189         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 190         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 191         .use_tempaddr           = 0,
 192         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 193         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 194         .regen_max_retry        = REGEN_MAX_RETRY,
 195         .max_desync_factor      = MAX_DESYNC_FACTOR,
 196         .max_addresses          = IPV6_MAX_ADDRESSES,
 197         .accept_ra_defrtr       = 1,
 198         .accept_ra_from_local   = 0,
 199         .accept_ra_min_hop_limit= 1,
 200         .accept_ra_pinfo        = 1,
 201 #ifdef CONFIG_IPV6_ROUTER_PREF
 202         .accept_ra_rtr_pref     = 1,
 203         .rtr_probe_interval     = 60 * HZ,
 204 #ifdef CONFIG_IPV6_ROUTE_INFO
 205         .accept_ra_rt_info_max_plen = 0,
 206 #endif
 207 #endif
 208         .proxy_ndp              = 0,
 209         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 210         .disable_ipv6           = 0,
 211         .accept_dad             = 1,
 212         .suppress_frag_ndisc    = 1,
 213         .accept_ra_mtu          = 1,
 214         .stable_secret          = {
 215                 .initialized = false,
 216         },
 217         .use_oif_addrs_only     = 0,
 218         .ignore_routes_with_linkdown = 0,
 219         .keep_addr_on_down      = 0,
 220 };
 221
 222 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 223         .forwarding             = 0,
 224         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 225         .mtu6                   = IPV6_MIN_MTU,
 226         .accept_ra              = 1,
 227         .accept_redirects       = 1,
 228         .autoconf               = 1,
 229         .force_mld_version      = 0,
 230         .mldv1_unsolicited_report_interval = 10 * HZ,
 231         .mldv2_unsolicited_report_interval = HZ,
 232         .dad_transmits          = 1,
 233         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 234         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 235         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 236         .use_tempaddr           = 0,
 237         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 238         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 239         .regen_max_retry        = REGEN_MAX_RETRY,
 240         .max_desync_factor      = MAX_DESYNC_FACTOR,
 241         .max_addresses          = IPV6_MAX_ADDRESSES,
 242         .accept_ra_defrtr       = 1,
 243         .accept_ra_from_local   = 0,
 244         .accept_ra_min_hop_limit= 1,
 245         .accept_ra_pinfo        = 1,
 246 #ifdef CONFIG_IPV6_ROUTER_PREF
 247         .accept_ra_rtr_pref     = 1,
 248         .rtr_probe_interval     = 60 * HZ,
 249 #ifdef CONFIG_IPV6_ROUTE_INFO
 250         .accept_ra_rt_info_max_plen = 0,
 251 #endif
 252 #endif
 253         .proxy_ndp              = 0,
 254         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 255         .disable_ipv6           = 0,
 256         .accept_dad             = 1,
 257         .suppress_frag_ndisc    = 1,
 258         .accept_ra_mtu          = 1,
 259         .stable_secret          = {
 260                 .initialized = false,
 261         },
 262         .use_oif_addrs_only     = 0,
 263         .ignore_routes_with_linkdown = 0,
 264         .keep_addr_on_down      = 0,
 265 };
 266
 267 /* Check if a valid qdisc is available */
 268 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 269 {
 270         return !qdisc_tx_is_noop(dev);
 271 }
 272
 273 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 274 {
 275         if (del_timer(&idev->rs_timer))
 276                 __in6_dev_put(idev);
 277 }
 278
 279 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 280 {
 281         if (cancel_delayed_work(&ifp->dad_work))
 282                 __in6_ifa_put(ifp);
 283 }
 284
 285 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 286                                   unsigned long when)
 287 {
 288         if (!timer_pending(&idev->rs_timer))
 289                 in6_dev_hold(idev);
 290         mod_timer(&idev->rs_timer, jiffies + when);
 291 }
 292
 293 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 294                                    unsigned long delay)
 295 {
 296         if (!delayed_work_pending(&ifp->dad_work))
 297                 in6_ifa_hold(ifp);
 298         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 299 }
 300
 301 static int snmp6_alloc_dev(struct inet6_dev *idev)
 302 {
 303         int i;
 304
 305         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 306         if (!idev->stats.ipv6)
 307                 goto err_ip;
 308
 309         for_each_possible_cpu(i) {
 310                 struct ipstats_mib *addrconf_stats;
 311                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 312                 u64_stats_init(&addrconf_stats->syncp);
 313         }
 314
 315
 316         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 317                                         GFP_KERNEL);
 318         if (!idev->stats.icmpv6dev)
 319                 goto err_icmp;
 320         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 321                                            GFP_KERNEL);
 322         if (!idev->stats.icmpv6msgdev)
 323                 goto err_icmpmsg;
 324
 325         return 0;
 326
 327 err_icmpmsg:
 328         kfree(idev->stats.icmpv6dev);
 329 err_icmp:
 330         free_percpu(idev->stats.ipv6);
 331 err_ip:
 332         return -ENOMEM;
 333 }
 334
 335 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 336 {
 337         struct inet6_dev *ndev;
 338         int err = -ENOMEM;
 339
 340         ASSERT_RTNL();
 341
 342         if (dev->mtu < IPV6_MIN_MTU)
 343                 return ERR_PTR(-EINVAL);
 344
 345         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 346         if (!ndev)
 347                 return ERR_PTR(err);
 348
 349         rwlock_init(&ndev->lock);
 350         ndev->dev = dev;
 351         INIT_LIST_HEAD(&ndev->addr_list);
 352         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 353                     (unsigned long)ndev);
 354         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 355
 356         if (ndev->cnf.stable_secret.initialized)
 357                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 358         else
 359                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
 360
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 363         if (!ndev->nd_parms) {
 364                 kfree(ndev);
 365                 return ERR_PTR(err);
 366         }
 367         if (ndev->cnf.forwarding)
 368                 dev_disable_lro(dev);
 369         /* We refer to the device */
 370         dev_hold(dev);
 371
 372         if (snmp6_alloc_dev(ndev) < 0) {
 373                 ADBG(KERN_WARNING
 374                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 375                         __func__, dev->name);
 376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 377                 dev_put(dev);
 378                 kfree(ndev);
 379                 return ERR_PTR(err);
 380         }
 381
 382         if (snmp6_register_dev(ndev) < 0) {
 383                 ADBG(KERN_WARNING
 384                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 385                         __func__, dev->name);
 386                 goto err_release;
 387         }
 388
 389         /* One reference from device.  We must do this before
 390          * we invoke __ipv6_regen_rndid().
 391          */
 392         in6_dev_hold(ndev);
 393
 394         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 395                 ndev->cnf.accept_dad = -1;
 396
 397 #if IS_ENABLED(CONFIG_IPV6_SIT)
 398         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 399                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 400                 ndev->cnf.rtr_solicits = 0;
 401         }
 402 #endif
 403
 404         INIT_LIST_HEAD(&ndev->tempaddr_list);
 405         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 406         if ((dev->flags&IFF_LOOPBACK) ||
 407             dev->type == ARPHRD_TUNNEL ||
 408             dev->type == ARPHRD_TUNNEL6 ||
 409             dev->type == ARPHRD_SIT ||
 410             dev->type == ARPHRD_NONE) {
 411                 ndev->cnf.use_tempaddr = -1;
 412         } else {
 413                 in6_dev_hold(ndev);
 414                 ipv6_regen_rndid((unsigned long) ndev);
 415         }
 416
 417         ndev->token = in6addr_any;
 418
 419         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 420                 ndev->if_flags |= IF_READY;
 421
 422         ipv6_mc_init_dev(ndev);
 423         ndev->tstamp = jiffies;
 424         err = addrconf_sysctl_register(ndev);
 425         if (err) {
 426                 ipv6_mc_destroy_dev(ndev);
 427                 del_timer(&ndev->regen_timer);
 428                 snmp6_unregister_dev(ndev);
 429                 goto err_release;
 430         }
 431         /* protected by rtnl_lock */
 432         rcu_assign_pointer(dev->ip6_ptr, ndev);
 433
 434         /* Join interface-local all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 436
 437         /* Join all-node multicast group */
 438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 439
 440         /* Join all-router multicast group if forwarding is set */
 441         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 442                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 443
 444         return ndev;
 445
 446 err_release:
 447         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 448         ndev->dead = 1;
 449         in6_dev_finish_destroy(ndev);
 450         return ERR_PTR(err);
 451 }
 452
 453 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 454 {
 455         struct inet6_dev *idev;
 456
 457         ASSERT_RTNL();
 458
 459         idev = __in6_dev_get(dev);
 460         if (!idev) {
 461                 idev = ipv6_add_dev(dev);
 462                 if (IS_ERR(idev))
 463                         return NULL;
 464         }
 465
 466         if (dev->flags&IFF_UP)
 467                 ipv6_mc_up(idev);
 468         return idev;
 469 }
 470
 471 static int inet6_netconf_msgsize_devconf(int type)
 472 {
 473         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 474                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 475         bool all = false;
 476
 477         if (type == NETCONFA_ALL)
 478                 all = true;
 479
 480         if (all || type == NETCONFA_FORWARDING)
 481                 size += nla_total_size(4);
 482 #ifdef CONFIG_IPV6_MROUTE
 483         if (all || type == NETCONFA_MC_FORWARDING)
 484                 size += nla_total_size(4);
 485 #endif
 486         if (all || type == NETCONFA_PROXY_NEIGH)
 487                 size += nla_total_size(4);
 488
 489         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 490                 size += nla_total_size(4);
 491
 492         return size;
 493 }
 494
 495 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 496                                       struct ipv6_devconf *devconf, u32 portid,
 497                                       u32 seq, int event, unsigned int flags,
 498                                       int type)
 499 {
 500         struct nlmsghdr  *nlh;
 501         struct netconfmsg *ncm;
 502         bool all = false;
 503
 504         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 505                         flags);
 506         if (!nlh)
 507                 return -EMSGSIZE;
 508
 509         if (type == NETCONFA_ALL)
 510                 all = true;
 511
 512         ncm = nlmsg_data(nlh);
 513         ncm->ncm_family = AF_INET6;
 514
 515         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 516                 goto nla_put_failure;
 517
 518         if ((all || type == NETCONFA_FORWARDING) &&
 519             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 520                 goto nla_put_failure;
 521 #ifdef CONFIG_IPV6_MROUTE
 522         if ((all || type == NETCONFA_MC_FORWARDING) &&
 523             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 524                         devconf->mc_forwarding) < 0)
 525                 goto nla_put_failure;
 526 #endif
 527         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 528             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 529                 goto nla_put_failure;
 530
 531         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 532             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 533                         devconf->ignore_routes_with_linkdown) < 0)
 534                 goto nla_put_failure;
 535
 536         nlmsg_end(skb, nlh);
 537         return 0;
 538
 539 nla_put_failure:
 540         nlmsg_cancel(skb, nlh);
 541         return -EMSGSIZE;
 542 }
 543
 544 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 545                                   struct ipv6_devconf *devconf)
 546 {
 547         struct sk_buff *skb;
 548         int err = -ENOBUFS;
 549
 550         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 551         if (!skb)
 552                 goto errout;
 553
 554         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 555                                          RTM_NEWNETCONF, 0, type);
 556         if (err < 0) {
 557                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 558                 WARN_ON(err == -EMSGSIZE);
 559                 kfree_skb(skb);
 560                 goto errout;
 561         }
 562         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 563         return;
 564 errout:
 565         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 566 }
 567
 568 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 569         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 570         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 571         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 572         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 573 };
 574
 575 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 576                                      struct nlmsghdr *nlh)
 577 {
 578         struct net *net = sock_net(in_skb->sk);
 579         struct nlattr *tb[NETCONFA_MAX+1];
 580         struct netconfmsg *ncm;
 581         struct sk_buff *skb;
 582         struct ipv6_devconf *devconf;
 583         struct inet6_dev *in6_dev;
 584         struct net_device *dev;
 585         int ifindex;
 586         int err;
 587
 588         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 589                           devconf_ipv6_policy);
 590         if (err < 0)
 591                 goto errout;
 592
 593         err = -EINVAL;
 594         if (!tb[NETCONFA_IFINDEX])
 595                 goto errout;
 596
 597         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 598         switch (ifindex) {
 599         case NETCONFA_IFINDEX_ALL:
 600                 devconf = net->ipv6.devconf_all;
 601                 break;
 602         case NETCONFA_IFINDEX_DEFAULT:
 603                 devconf = net->ipv6.devconf_dflt;
 604                 break;
 605         default:
 606                 dev = __dev_get_by_index(net, ifindex);
 607                 if (!dev)
 608                         goto errout;
 609                 in6_dev = __in6_dev_get(dev);
 610                 if (!in6_dev)
 611                         goto errout;
 612                 devconf = &in6_dev->cnf;
 613                 break;
 614         }
 615
 616         err = -ENOBUFS;
 617         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
 618         if (!skb)
 619                 goto errout;
 620
 621         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 622                                          NETLINK_CB(in_skb).portid,
 623                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 624                                          NETCONFA_ALL);
 625         if (err < 0) {
 626                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 627                 WARN_ON(err == -EMSGSIZE);
 628                 kfree_skb(skb);
 629                 goto errout;
 630         }
 631         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 632 errout:
 633         return err;
 634 }
 635
 636 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 637                                       struct netlink_callback *cb)
 638 {
 639         struct net *net = sock_net(skb->sk);
 640         int h, s_h;
 641         int idx, s_idx;
 642         struct net_device *dev;
 643         struct inet6_dev *idev;
 644         struct hlist_head *head;
 645
 646         s_h = cb->args[0];
 647         s_idx = idx = cb->args[1];
 648
 649         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 650                 idx = 0;
 651                 head = &net->dev_index_head[h];
 652                 rcu_read_lock();
 653                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 654                           net->dev_base_seq;
 655                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 656                         if (idx < s_idx)
 657                                 goto cont;
 658                         idev = __in6_dev_get(dev);
 659                         if (!idev)
 660                                 goto cont;
 661
 662                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 663                                                        &idev->cnf,
 664                                                        NETLINK_CB(cb->skb).portid,
 665                                                        cb->nlh->nlmsg_seq,
 666                                                        RTM_NEWNETCONF,
 667                                                        NLM_F_MULTI,
 668                                                        NETCONFA_ALL) < 0) {
 669                                 rcu_read_unlock();
 670                                 goto done;
 671                         }
 672                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 673 cont:
 674                         idx++;
 675                 }
 676                 rcu_read_unlock();
 677         }
 678         if (h == NETDEV_HASHENTRIES) {
 679                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 680                                                net->ipv6.devconf_all,
 681                                                NETLINK_CB(cb->skb).portid,
 682                                                cb->nlh->nlmsg_seq,
 683                                                RTM_NEWNETCONF, NLM_F_MULTI,
 684                                                NETCONFA_ALL) < 0)
 685                         goto done;
 686                 else
 687                         h++;
 688         }
 689         if (h == NETDEV_HASHENTRIES + 1) {
 690                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 691                                                net->ipv6.devconf_dflt,
 692                                                NETLINK_CB(cb->skb).portid,
 693                                                cb->nlh->nlmsg_seq,
 694                                                RTM_NEWNETCONF, NLM_F_MULTI,
 695                                                NETCONFA_ALL) < 0)
 696                         goto done;
 697                 else
 698                         h++;
 699         }
 700 done:
 701         cb->args[0] = h;
 702         cb->args[1] = idx;
 703
 704         return skb->len;
 705 }
 706
 707 #ifdef CONFIG_SYSCTL
 708 static void dev_forward_change(struct inet6_dev *idev)
 709 {
 710         struct net_device *dev;
 711         struct inet6_ifaddr *ifa;
 712
 713         if (!idev)
 714                 return;
 715         dev = idev->dev;
 716         if (idev->cnf.forwarding)
 717                 dev_disable_lro(dev);
 718         if (dev->flags & IFF_MULTICAST) {
 719                 if (idev->cnf.forwarding) {
 720                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 721                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 722                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 723                 } else {
 724                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 725                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 726                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 727                 }
 728         }
 729
 730         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 731                 if (ifa->flags&IFA_F_TENTATIVE)
 732                         continue;
 733                 if (idev->cnf.forwarding)
 734                         addrconf_join_anycast(ifa);
 735                 else
 736                         addrconf_leave_anycast(ifa);
 737         }
 738         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 739                                      dev->ifindex, &idev->cnf);
 740 }
 741
 742
 743 static void addrconf_forward_change(struct net *net, __s32 newf)
 744 {
 745         struct net_device *dev;
 746         struct inet6_dev *idev;
 747
 748         for_each_netdev(net, dev) {
 749                 idev = __in6_dev_get(dev);
 750                 if (idev) {
 751                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 752                         idev->cnf.forwarding = newf;
 753                         if (changed)
 754                                 dev_forward_change(idev);
 755                 }
 756         }
 757 }
 758
 759 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 760 {
 761         struct net *net;
 762         int old;
 763
 764         if (!rtnl_trylock())
 765                 return restart_syscall();
 766
 767         net = (struct net *)table->extra2;
 768         old = *p;
 769         *p = newf;
 770
 771         if (p == &net->ipv6.devconf_dflt->forwarding) {
 772                 if ((!newf) ^ (!old))
 773                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 774                                                      NETCONFA_IFINDEX_DEFAULT,
 775                                                      net->ipv6.devconf_dflt);
 776                 rtnl_unlock();
 777                 return 0;
 778         }
 779
 780         if (p == &net->ipv6.devconf_all->forwarding) {
 781                 net->ipv6.devconf_dflt->forwarding = newf;
 782                 addrconf_forward_change(net, newf);
 783                 if ((!newf) ^ (!old))
 784                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 785                                                      NETCONFA_IFINDEX_ALL,
 786                                                      net->ipv6.devconf_all);
 787         } else if ((!newf) ^ (!old))
 788                 dev_forward_change((struct inet6_dev *)table->extra1);
 789         rtnl_unlock();
 790
 791         if (newf)
 792                 rt6_purge_dflt_routers(net);
 793         return 1;
 794 }
 795
 796 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 797 {
 798         struct net_device *dev;
 799         struct inet6_dev *idev;
 800
 801         for_each_netdev(net, dev) {
 802                 idev = __in6_dev_get(dev);
 803                 if (idev) {
 804                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 805
 806                         idev->cnf.ignore_routes_with_linkdown = newf;
 807                         if (changed)
 808                                 inet6_netconf_notify_devconf(dev_net(dev),
 809                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 810                                                              dev->ifindex,
 811                                                              &idev->cnf);
 812                 }
 813         }
 814 }
 815
 816 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 817 {
 818         struct net *net;
 819         int old;
 820
 821         if (!rtnl_trylock())
 822                 return restart_syscall();
 823
 824         net = (struct net *)table->extra2;
 825         old = *p;
 826         *p = newf;
 827
 828         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 829                 if ((!newf) ^ (!old))
 830                         inet6_netconf_notify_devconf(net,
 831                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 832                                                      NETCONFA_IFINDEX_DEFAULT,
 833                                                      net->ipv6.devconf_dflt);
 834                 rtnl_unlock();
 835                 return 0;
 836         }
 837
 838         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 839                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 840                 addrconf_linkdown_change(net, newf);
 841                 if ((!newf) ^ (!old))
 842                         inet6_netconf_notify_devconf(net,
 843                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 844                                                      NETCONFA_IFINDEX_ALL,
 845                                                      net->ipv6.devconf_all);
 846         }
 847         rtnl_unlock();
 848
 849         return 1;
 850 }
 851
 852 #endif
 853
 854 /* Nobody refers to this ifaddr, destroy it */
 855 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 856 {
 857         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 858
 859 #ifdef NET_REFCNT_DEBUG
 860         pr_debug("%s\n", __func__);
 861 #endif
 862
 863         in6_dev_put(ifp->idev);
 864
 865         if (cancel_delayed_work(&ifp->dad_work))
 866                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 867                           ifp);
 868
 869         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 870                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 871                 return;
 872         }
 873         ip6_rt_put(ifp->rt);
 874
 875         kfree_rcu(ifp, rcu);
 876 }
 877
 878 static void
 879 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 880 {
 881         struct list_head *p;
 882         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 883
 884         /*
 885          * Each device address list is sorted in order of scope -
 886          * global before linklocal.
 887          */
 888         list_for_each(p, &idev->addr_list) {
 889                 struct inet6_ifaddr *ifa
 890                         = list_entry(p, struct inet6_ifaddr, if_list);
 891                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 892                         break;
 893         }
 894
 895         list_add_tail(&ifp->if_list, p);
 896 }
 897
 898 static u32 inet6_addr_hash(const struct in6_addr *addr)
 899 {
 900         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 901 }
 902
 903 /* On success it returns ifp with increased reference count */
 904
 905 static struct inet6_ifaddr *
 906 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 907               const struct in6_addr *peer_addr, int pfxlen,
 908               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 909 {
 910         struct inet6_ifaddr *ifa = NULL;
 911         struct rt6_info *rt;
 912         unsigned int hash;
 913         int err = 0;
 914         int addr_type = ipv6_addr_type(addr);
 915
 916         if (addr_type == IPV6_ADDR_ANY ||
 917             addr_type & IPV6_ADDR_MULTICAST ||
 918             (!(idev->dev->flags & IFF_LOOPBACK) &&
 919              addr_type & IPV6_ADDR_LOOPBACK))
 920                 return ERR_PTR(-EADDRNOTAVAIL);
 921
 922         rcu_read_lock_bh();
 923         if (idev->dead) {
 924                 err = -ENODEV;                  /*XXX*/
 925                 goto out2;
 926         }
 927
 928         if (idev->cnf.disable_ipv6) {
 929                 err = -EACCES;
 930                 goto out2;
 931         }
 932
 933         spin_lock(&addrconf_hash_lock);
 934
 935         /* Ignore adding duplicate addresses on an interface */
 936         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 937                 ADBG("ipv6_add_addr: already assigned\n");
 938                 err = -EEXIST;
 939                 goto out;
 940         }
 941
 942         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 943
 944         if (!ifa) {
 945                 ADBG("ipv6_add_addr: malloc failed\n");
 946                 err = -ENOBUFS;
 947                 goto out;
 948         }
 949
 950         rt = addrconf_dst_alloc(idev, addr, false);
 951         if (IS_ERR(rt)) {
 952                 err = PTR_ERR(rt);
 953                 goto out;
 954         }
 955
 956         neigh_parms_data_state_setall(idev->nd_parms);
 957
 958         ifa->addr = *addr;
 959         if (peer_addr)
 960                 ifa->peer_addr = *peer_addr;
 961
 962         spin_lock_init(&ifa->lock);
 963         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 964         INIT_HLIST_NODE(&ifa->addr_lst);
 965         ifa->scope = scope;
 966         ifa->prefix_len = pfxlen;
 967         ifa->flags = flags | IFA_F_TENTATIVE;
 968         ifa->valid_lft = valid_lft;
 969         ifa->prefered_lft = prefered_lft;
 970         ifa->cstamp = ifa->tstamp = jiffies;
 971         ifa->tokenized = false;
 972
 973         ifa->rt = rt;
 974
 975         ifa->idev = idev;
 976         in6_dev_hold(idev);
 977         /* For caller */
 978         in6_ifa_hold(ifa);
 979
 980         /* Add to big hash table */
 981         hash = inet6_addr_hash(addr);
 982
 983         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 984         spin_unlock(&addrconf_hash_lock);
 985
 986         write_lock(&idev->lock);
 987         /* Add to inet6_dev unicast addr list. */
 988         ipv6_link_dev_addr(idev, ifa);
 989
 990         if (ifa->flags&IFA_F_TEMPORARY) {
 991                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 992                 in6_ifa_hold(ifa);
 993         }
 994
 995         in6_ifa_hold(ifa);
 996         write_unlock(&idev->lock);
 997 out2:
 998         rcu_read_unlock_bh();
 999
1000         if (likely(err == 0))
1001                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1002         else {
1003                 kfree(ifa);
1004                 ifa = ERR_PTR(err);
1005         }
1006
1007         return ifa;
1008 out:
1009         spin_unlock(&addrconf_hash_lock);
1010         goto out2;
1011 }
1012
1013 enum cleanup_prefix_rt_t {
1014         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1015         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1016         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1017 };
1018
1019 /*
1020  * Check, whether the prefix for ifp would still need a prefix route
1021  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1022  * constants.
1023  *
1024  * 1) we don't purge prefix if address was not permanent.
1025  *    prefix is managed by its own lifetime.
1026  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1027  * 3) if there are no addresses, delete prefix.
1028  * 4) if there are still other permanent address(es),
1029  *    corresponding prefix is still permanent.
1030  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1031  *    don't purge the prefix, assume user space is managing it.
1032  * 6) otherwise, update prefix lifetime to the
1033  *    longest valid lifetime among the corresponding
1034  *    addresses on the device.
1035  *    Note: subsequent RA will update lifetime.
1036  **/
1037 static enum cleanup_prefix_rt_t
1038 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1039 {
1040         struct inet6_ifaddr *ifa;
1041         struct inet6_dev *idev = ifp->idev;
1042         unsigned long lifetime;
1043         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1044
1045         *expires = jiffies;
1046
1047         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1048                 if (ifa == ifp)
1049                         continue;
1050                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1051                                        ifp->prefix_len))
1052                         continue;
1053                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1054                         return CLEANUP_PREFIX_RT_NOP;
1055
1056                 action = CLEANUP_PREFIX_RT_EXPIRE;
1057
1058                 spin_lock(&ifa->lock);
1059
1060                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1061                 /*
1062                  * Note: Because this address is
1063                  * not permanent, lifetime <
1064                  * LONG_MAX / HZ here.
1065                  */
1066                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1067                         *expires = ifa->tstamp + lifetime * HZ;
1068                 spin_unlock(&ifa->lock);
1069         }
1070
1071         return action;
1072 }
1073
1074 static void
1075 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1076 {
1077         struct rt6_info *rt;
1078
1079         rt = addrconf_get_prefix_route(&ifp->addr,
1080                                        ifp->prefix_len,
1081                                        ifp->idev->dev,
1082                                        0, RTF_GATEWAY | RTF_DEFAULT);
1083         if (rt) {
1084                 if (del_rt)
1085                         ip6_del_rt(rt);
1086                 else {
1087                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1088                                 rt6_set_expires(rt, expires);
1089                         ip6_rt_put(rt);
1090                 }
1091         }
1092 }
1093
1094
1095 /* This function wants to get referenced ifp and releases it before return */
1096
1097 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1098 {
1099         int state;
1100         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1101         unsigned long expires;
1102
1103         ASSERT_RTNL();
1104
1105         spin_lock_bh(&ifp->lock);
1106         state = ifp->state;
1107         ifp->state = INET6_IFADDR_STATE_DEAD;
1108         spin_unlock_bh(&ifp->lock);
1109
1110         if (state == INET6_IFADDR_STATE_DEAD)
1111                 goto out;
1112
1113         spin_lock_bh(&addrconf_hash_lock);
1114         hlist_del_init_rcu(&ifp->addr_lst);
1115         spin_unlock_bh(&addrconf_hash_lock);
1116
1117         write_lock_bh(&ifp->idev->lock);
1118
1119         if (ifp->flags&IFA_F_TEMPORARY) {
1120                 list_del(&ifp->tmp_list);
1121                 if (ifp->ifpub) {
1122                         in6_ifa_put(ifp->ifpub);
1123                         ifp->ifpub = NULL;
1124                 }
1125                 __in6_ifa_put(ifp);
1126         }
1127
1128         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1129                 action = check_cleanup_prefix_route(ifp, &expires);
1130
1131         list_del_init(&ifp->if_list);
1132         __in6_ifa_put(ifp);
1133
1134         write_unlock_bh(&ifp->idev->lock);
1135
1136         addrconf_del_dad_work(ifp);
1137
1138         ipv6_ifa_notify(RTM_DELADDR, ifp);
1139
1140         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1141
1142         if (action != CLEANUP_PREFIX_RT_NOP) {
1143                 cleanup_prefix_route(ifp, expires,
1144                         action == CLEANUP_PREFIX_RT_DEL);
1145         }
1146
1147         /* clean up prefsrc entries */
1148         rt6_remove_prefsrc(ifp);
1149 out:
1150         in6_ifa_put(ifp);
1151 }
1152
1153 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1154 {
1155         struct inet6_dev *idev = ifp->idev;
1156         struct in6_addr addr, *tmpaddr;
1157         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1158         unsigned long regen_advance;
1159         int tmp_plen;
1160         int ret = 0;
1161         u32 addr_flags;
1162         unsigned long now = jiffies;
1163
1164         write_lock_bh(&idev->lock);
1165         if (ift) {
1166                 spin_lock_bh(&ift->lock);
1167                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1168                 spin_unlock_bh(&ift->lock);
1169                 tmpaddr = &addr;
1170         } else {
1171                 tmpaddr = NULL;
1172         }
1173 retry:
1174         in6_dev_hold(idev);
1175         if (idev->cnf.use_tempaddr <= 0) {
1176                 write_unlock_bh(&idev->lock);
1177                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1178                 in6_dev_put(idev);
1179                 ret = -1;
1180                 goto out;
1181         }
1182         spin_lock_bh(&ifp->lock);
1183         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1184                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1185                 spin_unlock_bh(&ifp->lock);
1186                 write_unlock_bh(&idev->lock);
1187                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1188                         __func__);
1189                 in6_dev_put(idev);
1190                 ret = -1;
1191                 goto out;
1192         }
1193         in6_ifa_hold(ifp);
1194         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1195         __ipv6_try_regen_rndid(idev, tmpaddr);
1196         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1197         age = (now - ifp->tstamp) / HZ;
1198         tmp_valid_lft = min_t(__u32,
1199                               ifp->valid_lft,
1200                               idev->cnf.temp_valid_lft + age);
1201         tmp_prefered_lft = min_t(__u32,
1202                                  ifp->prefered_lft,
1203                                  idev->cnf.temp_prefered_lft + age -
1204                                  idev->cnf.max_desync_factor);
1205         tmp_plen = ifp->prefix_len;
1206         tmp_tstamp = ifp->tstamp;
1207         spin_unlock_bh(&ifp->lock);
1208
1209         regen_advance = idev->cnf.regen_max_retry *
1210                         idev->cnf.dad_transmits *
1211                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1212         write_unlock_bh(&idev->lock);
1213
1214         /* A temporary address is created only if this calculated Preferred
1215          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1216          * an implementation must not create a temporary address with a zero
1217          * Preferred Lifetime.
1218          * Use age calculation as in addrconf_verify to avoid unnecessary
1219          * temporary addresses being generated.
1220          */
1221         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1222         if (tmp_prefered_lft <= regen_advance + age) {
1223                 in6_ifa_put(ifp);
1224                 in6_dev_put(idev);
1225                 ret = -1;
1226                 goto out;
1227         }
1228
1229         addr_flags = IFA_F_TEMPORARY;
1230         /* set in addrconf_prefix_rcv() */
1231         if (ifp->flags & IFA_F_OPTIMISTIC)
1232                 addr_flags |= IFA_F_OPTIMISTIC;
1233
1234         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1235                             ipv6_addr_scope(&addr), addr_flags,
1236                             tmp_valid_lft, tmp_prefered_lft);
1237         if (IS_ERR(ift)) {
1238                 in6_ifa_put(ifp);
1239                 in6_dev_put(idev);
1240                 pr_info("%s: retry temporary address regeneration\n", __func__);
1241                 tmpaddr = &addr;
1242                 write_lock_bh(&idev->lock);
1243                 goto retry;
1244         }
1245
1246         spin_lock_bh(&ift->lock);
1247         ift->ifpub = ifp;
1248         ift->cstamp = now;
1249         ift->tstamp = tmp_tstamp;
1250         spin_unlock_bh(&ift->lock);
1251
1252         addrconf_dad_start(ift);
1253         in6_ifa_put(ift);
1254         in6_dev_put(idev);
1255 out:
1256         return ret;
1257 }
1258
1259 /*
1260  *      Choose an appropriate source address (RFC3484)
1261  */
1262 enum {
1263         IPV6_SADDR_RULE_INIT = 0,
1264         IPV6_SADDR_RULE_LOCAL,
1265         IPV6_SADDR_RULE_SCOPE,
1266         IPV6_SADDR_RULE_PREFERRED,
1267 #ifdef CONFIG_IPV6_MIP6
1268         IPV6_SADDR_RULE_HOA,
1269 #endif
1270         IPV6_SADDR_RULE_OIF,
1271         IPV6_SADDR_RULE_LABEL,
1272         IPV6_SADDR_RULE_PRIVACY,
1273         IPV6_SADDR_RULE_ORCHID,
1274         IPV6_SADDR_RULE_PREFIX,
1275 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1276         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1277 #endif
1278         IPV6_SADDR_RULE_MAX
1279 };
1280
1281 struct ipv6_saddr_score {
1282         int                     rule;
1283         int                     addr_type;
1284         struct inet6_ifaddr     *ifa;
1285         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1286         int                     scopedist;
1287         int                     matchlen;
1288 };
1289
1290 struct ipv6_saddr_dst {
1291         const struct in6_addr *addr;
1292         int ifindex;
1293         int scope;
1294         int label;
1295         unsigned int prefs;
1296 };
1297
1298 static inline int ipv6_saddr_preferred(int type)
1299 {
1300         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1301                 return 1;
1302         return 0;
1303 }
1304
1305 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1306 {
1307 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1308         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1309 #else
1310         return false;
1311 #endif
1312 }
1313
1314 static int ipv6_get_saddr_eval(struct net *net,
1315                                struct ipv6_saddr_score *score,
1316                                struct ipv6_saddr_dst *dst,
1317                                int i)
1318 {
1319         int ret;
1320
1321         if (i <= score->rule) {
1322                 switch (i) {
1323                 case IPV6_SADDR_RULE_SCOPE:
1324                         ret = score->scopedist;
1325                         break;
1326                 case IPV6_SADDR_RULE_PREFIX:
1327                         ret = score->matchlen;
1328                         break;
1329                 default:
1330                         ret = !!test_bit(i, score->scorebits);
1331                 }
1332                 goto out;
1333         }
1334
1335         switch (i) {
1336         case IPV6_SADDR_RULE_INIT:
1337                 /* Rule 0: remember if hiscore is not ready yet */
1338                 ret = !!score->ifa;
1339                 break;
1340         case IPV6_SADDR_RULE_LOCAL:
1341                 /* Rule 1: Prefer same address */
1342                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1343                 break;
1344         case IPV6_SADDR_RULE_SCOPE:
1345                 /* Rule 2: Prefer appropriate scope
1346                  *
1347                  *      ret
1348                  *       ^
1349                  *    -1 |  d 15
1350                  *    ---+--+-+---> scope
1351                  *       |
1352                  *       |             d is scope of the destination.
1353                  *  B-d  |  \
1354                  *       |   \      <- smaller scope is better if
1355                  *  B-15 |    \        if scope is enough for destination.
1356                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1357                  * d-C-1 | /
1358                  *       |/         <- greater is better
1359                  *   -C  /             if scope is not enough for destination.
1360                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1361                  *
1362                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1363                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1364                  * Assume B = 0 and we get C > 29.
1365                  */
1366                 ret = __ipv6_addr_src_scope(score->addr_type);
1367                 if (ret >= dst->scope)
1368                         ret = -ret;
1369                 else
1370                         ret -= 128;     /* 30 is enough */
1371                 score->scopedist = ret;
1372                 break;
1373         case IPV6_SADDR_RULE_PREFERRED:
1374             {
1375                 /* Rule 3: Avoid deprecated and optimistic addresses */
1376                 u8 avoid = IFA_F_DEPRECATED;
1377
1378                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1379                         avoid |= IFA_F_OPTIMISTIC;
1380                 ret = ipv6_saddr_preferred(score->addr_type) ||
1381                       !(score->ifa->flags & avoid);
1382                 break;
1383             }
1384 #ifdef CONFIG_IPV6_MIP6
1385         case IPV6_SADDR_RULE_HOA:
1386             {
1387                 /* Rule 4: Prefer home address */
1388                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1389                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1390                 break;
1391             }
1392 #endif
1393         case IPV6_SADDR_RULE_OIF:
1394                 /* Rule 5: Prefer outgoing interface */
1395                 ret = (!dst->ifindex ||
1396                        dst->ifindex == score->ifa->idev->dev->ifindex);
1397                 break;
1398         case IPV6_SADDR_RULE_LABEL:
1399                 /* Rule 6: Prefer matching label */
1400                 ret = ipv6_addr_label(net,
1401                                       &score->ifa->addr, score->addr_type,
1402                                       score->ifa->idev->dev->ifindex) == dst->label;
1403                 break;
1404         case IPV6_SADDR_RULE_PRIVACY:
1405             {
1406                 /* Rule 7: Prefer public address
1407                  * Note: prefer temporary address if use_tempaddr >= 2
1408                  */
1409                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1410                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1411                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1412                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1413                 break;
1414             }
1415         case IPV6_SADDR_RULE_ORCHID:
1416                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1417                  *          non-ORCHID vs non-ORCHID
1418                  */
1419                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1420                         ipv6_addr_orchid(dst->addr));
1421                 break;
1422         case IPV6_SADDR_RULE_PREFIX:
1423                 /* Rule 8: Use longest matching prefix */
1424                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1425                 if (ret > score->ifa->prefix_len)
1426                         ret = score->ifa->prefix_len;
1427                 score->matchlen = ret;
1428                 break;
1429 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1430         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1431                 /* Optimistic addresses still have lower precedence than other
1432                  * preferred addresses.
1433                  */
1434                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1435                 break;
1436 #endif
1437         default:
1438                 ret = 0;
1439         }
1440
1441         if (ret)
1442                 __set_bit(i, score->scorebits);
1443         score->rule = i;
1444 out:
1445         return ret;
1446 }
1447
1448 static int __ipv6_dev_get_saddr(struct net *net,
1449                                 struct ipv6_saddr_dst *dst,
1450                                 struct inet6_dev *idev,
1451                                 struct ipv6_saddr_score *scores,
1452                                 int hiscore_idx)
1453 {
1454         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1455
1456         read_lock_bh(&idev->lock);
1457         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1458                 int i;
1459
1460                 /*
1461                  * - Tentative Address (RFC2462 section 5.4)
1462                  *  - A tentative address is not considered
1463                  *    "assigned to an interface" in the traditional
1464                  *    sense, unless it is also flagged as optimistic.
1465                  * - Candidate Source Address (section 4)
1466                  *  - In any case, anycast addresses, multicast
1467                  *    addresses, and the unspecified address MUST
1468                  *    NOT be included in a candidate set.
1469                  */
1470                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1471                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1472                         continue;
1473
1474                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1475
1476                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1477                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1478                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1479                                             idev->dev->name);
1480                         continue;
1481                 }
1482
1483                 score->rule = -1;
1484                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1485
1486                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1487                         int minihiscore, miniscore;
1488
1489                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1490                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1491
1492                         if (minihiscore > miniscore) {
1493                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1494                                     score->scopedist > 0) {
1495                                         /*
1496                                          * special case:
1497                                          * each remaining entry
1498                                          * has too small (not enough)
1499                                          * scope, because ifa entries
1500                                          * are sorted by their scope
1501                                          * values.
1502                                          */
1503                                         goto out;
1504                                 }
1505                                 break;
1506                         } else if (minihiscore < miniscore) {
1507                                 if (hiscore->ifa)
1508                                         in6_ifa_put(hiscore->ifa);
1509
1510                                 in6_ifa_hold(score->ifa);
1511
1512                                 swap(hiscore, score);
1513                                 hiscore_idx = 1 - hiscore_idx;
1514
1515                                 /* restore our iterator */
1516                                 score->ifa = hiscore->ifa;
1517
1518                                 break;
1519                         }
1520                 }
1521         }
1522 out:
1523         read_unlock_bh(&idev->lock);
1524         return hiscore_idx;
1525 }
1526
1527 static int ipv6_get_saddr_master(struct net *net,
1528                                  const struct net_device *dst_dev,
1529                                  const struct net_device *master,
1530                                  struct ipv6_saddr_dst *dst,
1531                                  struct ipv6_saddr_score *scores,
1532                                  int hiscore_idx)
1533 {
1534         struct inet6_dev *idev;
1535
1536         idev = __in6_dev_get(dst_dev);
1537         if (idev)
1538                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1539                                                    scores, hiscore_idx);
1540
1541         idev = __in6_dev_get(master);
1542         if (idev)
1543                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1544                                                    scores, hiscore_idx);
1545
1546         return hiscore_idx;
1547 }
1548
1549 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1550                        const struct in6_addr *daddr, unsigned int prefs,
1551                        struct in6_addr *saddr)
1552 {
1553         struct ipv6_saddr_score scores[2], *hiscore;
1554         struct ipv6_saddr_dst dst;
1555         struct inet6_dev *idev;
1556         struct net_device *dev;
1557         int dst_type;
1558         bool use_oif_addr = false;
1559         int hiscore_idx = 0;
1560
1561         dst_type = __ipv6_addr_type(daddr);
1562         dst.addr = daddr;
1563         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1564         dst.scope = __ipv6_addr_src_scope(dst_type);
1565         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1566         dst.prefs = prefs;
1567
1568         scores[hiscore_idx].rule = -1;
1569         scores[hiscore_idx].ifa = NULL;
1570
1571         rcu_read_lock();
1572
1573         /* Candidate Source Address (section 4)
1574          *  - multicast and link-local destination address,
1575          *    the set of candidate source address MUST only
1576          *    include addresses assigned to interfaces
1577          *    belonging to the same link as the outgoing
1578          *    interface.
1579          * (- For site-local destination addresses, the
1580          *    set of candidate source addresses MUST only
1581          *    include addresses assigned to interfaces
1582          *    belonging to the same site as the outgoing
1583          *    interface.)
1584          *  - "It is RECOMMENDED that the candidate source addresses
1585          *    be the set of unicast addresses assigned to the
1586          *    interface that will be used to send to the destination
1587          *    (the 'outgoing' interface)." (RFC 6724)
1588          */
1589         if (dst_dev) {
1590                 idev = __in6_dev_get(dst_dev);
1591                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1592                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1593                     (idev && idev->cnf.use_oif_addrs_only)) {
1594                         use_oif_addr = true;
1595                 }
1596         }
1597
1598         if (use_oif_addr) {
1599                 if (idev)
1600                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1601         } else {
1602                 const struct net_device *master;
1603                 int master_idx = 0;
1604
1605                 /* if dst_dev exists and is enslaved to an L3 device, then
1606                  * prefer addresses from dst_dev and then the master over
1607                  * any other enslaved devices in the L3 domain.
1608                  */
1609                 master = l3mdev_master_dev_rcu(dst_dev);
1610                 if (master) {
1611                         master_idx = master->ifindex;
1612
1613                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1614                                                             master, &dst,
1615                                                             scores, hiscore_idx);
1616
1617                         if (scores[hiscore_idx].ifa)
1618                                 goto out;
1619                 }
1620
1621                 for_each_netdev_rcu(net, dev) {
1622                         /* only consider addresses on devices in the
1623                          * same L3 domain
1624                          */
1625                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1626                                 continue;
1627                         idev = __in6_dev_get(dev);
1628                         if (!idev)
1629                                 continue;
1630                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1631                 }
1632         }
1633
1634 out:
1635         rcu_read_unlock();
1636
1637         hiscore = &scores[hiscore_idx];
1638         if (!hiscore->ifa)
1639                 return -EADDRNOTAVAIL;
1640
1641         *saddr = hiscore->ifa->addr;
1642         in6_ifa_put(hiscore->ifa);
1643         return 0;
1644 }
1645 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1646
1647 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1648                       u32 banned_flags)
1649 {
1650         struct inet6_ifaddr *ifp;
1651         int err = -EADDRNOTAVAIL;
1652
1653         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1654                 if (ifp->scope > IFA_LINK)
1655                         break;
1656                 if (ifp->scope == IFA_LINK &&
1657                     !(ifp->flags & banned_flags)) {
1658                         *addr = ifp->addr;
1659                         err = 0;
1660                         break;
1661                 }
1662         }
1663         return err;
1664 }
1665
1666 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1667                     u32 banned_flags)
1668 {
1669         struct inet6_dev *idev;
1670         int err = -EADDRNOTAVAIL;
1671
1672         rcu_read_lock();
1673         idev = __in6_dev_get(dev);
1674         if (idev) {
1675                 read_lock_bh(&idev->lock);
1676                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1677                 read_unlock_bh(&idev->lock);
1678         }
1679         rcu_read_unlock();
1680         return err;
1681 }
1682
1683 static int ipv6_count_addresses(struct inet6_dev *idev)
1684 {
1685         int cnt = 0;
1686         struct inet6_ifaddr *ifp;
1687
1688         read_lock_bh(&idev->lock);
1689         list_for_each_entry(ifp, &idev->addr_list, if_list)
1690                 cnt++;
1691         read_unlock_bh(&idev->lock);
1692         return cnt;
1693 }
1694
1695 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1696                   const struct net_device *dev, int strict)
1697 {
1698         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1699 }
1700 EXPORT_SYMBOL(ipv6_chk_addr);
1701
1702 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1703                             const struct net_device *dev, int strict,
1704                             u32 banned_flags)
1705 {
1706         struct inet6_ifaddr *ifp;
1707         unsigned int hash = inet6_addr_hash(addr);
1708         u32 ifp_flags;
1709
1710         rcu_read_lock_bh();
1711         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1712                 if (!net_eq(dev_net(ifp->idev->dev), net))
1713                         continue;
1714                 /* Decouple optimistic from tentative for evaluation here.
1715                  * Ban optimistic addresses explicitly, when required.
1716                  */
1717                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1718                             ? (ifp->flags&~IFA_F_TENTATIVE)
1719                             : ifp->flags;
1720                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1721                     !(ifp_flags&banned_flags) &&
1722                     (!dev || ifp->idev->dev == dev ||
1723                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1724                         rcu_read_unlock_bh();
1725                         return 1;
1726                 }
1727         }
1728
1729         rcu_read_unlock_bh();
1730         return 0;
1731 }
1732 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1733
1734 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1735                                struct net_device *dev)
1736 {
1737         unsigned int hash = inet6_addr_hash(addr);
1738         struct inet6_ifaddr *ifp;
1739
1740         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1741                 if (!net_eq(dev_net(ifp->idev->dev), net))
1742                         continue;
1743                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1744                         if (!dev || ifp->idev->dev == dev)
1745                                 return true;
1746                 }
1747         }
1748         return false;
1749 }
1750
1751 /* Compares an address/prefix_len with addresses on device @dev.
1752  * If one is found it returns true.
1753  */
1754 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1755         const unsigned int prefix_len, struct net_device *dev)
1756 {
1757         struct inet6_dev *idev;
1758         struct inet6_ifaddr *ifa;
1759         bool ret = false;
1760
1761         rcu_read_lock();
1762         idev = __in6_dev_get(dev);
1763         if (idev) {
1764                 read_lock_bh(&idev->lock);
1765                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1766                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1767                         if (ret)
1768                                 break;
1769                 }
1770                 read_unlock_bh(&idev->lock);
1771         }
1772         rcu_read_unlock();
1773
1774         return ret;
1775 }
1776 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1777
1778 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1779 {
1780         struct inet6_dev *idev;
1781         struct inet6_ifaddr *ifa;
1782         int     onlink;
1783
1784         onlink = 0;
1785         rcu_read_lock();
1786         idev = __in6_dev_get(dev);
1787         if (idev) {
1788                 read_lock_bh(&idev->lock);
1789                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1790                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1791                                                    ifa->prefix_len);
1792                         if (onlink)
1793                                 break;
1794                 }
1795                 read_unlock_bh(&idev->lock);
1796         }
1797         rcu_read_unlock();
1798         return onlink;
1799 }
1800 EXPORT_SYMBOL(ipv6_chk_prefix);
1801
1802 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1803                                      struct net_device *dev, int strict)
1804 {
1805         struct inet6_ifaddr *ifp, *result = NULL;
1806         unsigned int hash = inet6_addr_hash(addr);
1807
1808         rcu_read_lock_bh();
1809         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1810                 if (!net_eq(dev_net(ifp->idev->dev), net))
1811                         continue;
1812                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1813                         if (!dev || ifp->idev->dev == dev ||
1814                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1815                                 result = ifp;
1816                                 in6_ifa_hold(ifp);
1817                                 break;
1818                         }
1819                 }
1820         }
1821         rcu_read_unlock_bh();
1822
1823         return result;
1824 }
1825
1826 /* Gets referenced address, destroys ifaddr */
1827
1828 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1829 {
1830         if (dad_failed)
1831                 ifp->flags |= IFA_F_DADFAILED;
1832
1833         if (ifp->flags&IFA_F_PERMANENT) {
1834                 spin_lock_bh(&ifp->lock);
1835                 addrconf_del_dad_work(ifp);
1836                 ifp->flags |= IFA_F_TENTATIVE;
1837                 spin_unlock_bh(&ifp->lock);
1838                 if (dad_failed)
1839                         ipv6_ifa_notify(0, ifp);
1840                 in6_ifa_put(ifp);
1841         } else if (ifp->flags&IFA_F_TEMPORARY) {
1842                 struct inet6_ifaddr *ifpub;
1843                 spin_lock_bh(&ifp->lock);
1844                 ifpub = ifp->ifpub;
1845                 if (ifpub) {
1846                         in6_ifa_hold(ifpub);
1847                         spin_unlock_bh(&ifp->lock);
1848                         ipv6_create_tempaddr(ifpub, ifp);
1849                         in6_ifa_put(ifpub);
1850                 } else {
1851                         spin_unlock_bh(&ifp->lock);
1852                 }
1853                 ipv6_del_addr(ifp);
1854         } else {
1855                 ipv6_del_addr(ifp);
1856         }
1857 }
1858
1859 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1860 {
1861         int err = -ENOENT;
1862
1863         spin_lock_bh(&ifp->lock);
1864         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1865                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1866                 err = 0;
1867         }
1868         spin_unlock_bh(&ifp->lock);
1869
1870         return err;
1871 }
1872
1873 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1874 {
1875         struct in6_addr addr;
1876         struct inet6_dev *idev = ifp->idev;
1877         struct net *net = dev_net(ifp->idev->dev);
1878
1879         if (addrconf_dad_end(ifp)) {
1880                 in6_ifa_put(ifp);
1881                 return;
1882         }
1883
1884         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1885                              ifp->idev->dev->name, &ifp->addr);
1886
1887         spin_lock_bh(&ifp->lock);
1888
1889         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1890                 int scope = ifp->scope;
1891                 u32 flags = ifp->flags;
1892                 struct in6_addr new_addr;
1893                 struct inet6_ifaddr *ifp2;
1894                 u32 valid_lft, preferred_lft;
1895                 int pfxlen = ifp->prefix_len;
1896                 int retries = ifp->stable_privacy_retry + 1;
1897
1898                 if (retries > net->ipv6.sysctl.idgen_retries) {
1899                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1900                                              ifp->idev->dev->name);
1901                         goto errdad;
1902                 }
1903
1904                 new_addr = ifp->addr;
1905                 if (ipv6_generate_stable_address(&new_addr, retries,
1906                                                  idev))
1907                         goto errdad;
1908
1909                 valid_lft = ifp->valid_lft;
1910                 preferred_lft = ifp->prefered_lft;
1911
1912                 spin_unlock_bh(&ifp->lock);
1913
1914                 if (idev->cnf.max_addresses &&
1915                     ipv6_count_addresses(idev) >=
1916                     idev->cnf.max_addresses)
1917                         goto lock_errdad;
1918
1919                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1920                                      ifp->idev->dev->name);
1921
1922                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1923                                      scope, flags, valid_lft,
1924                                      preferred_lft);
1925                 if (IS_ERR(ifp2))
1926                         goto lock_errdad;
1927
1928                 spin_lock_bh(&ifp2->lock);
1929                 ifp2->stable_privacy_retry = retries;
1930                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1931                 spin_unlock_bh(&ifp2->lock);
1932
1933                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1934                 in6_ifa_put(ifp2);
1935 lock_errdad:
1936                 spin_lock_bh(&ifp->lock);
1937         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1938                 addr.s6_addr32[0] = htonl(0xfe800000);
1939                 addr.s6_addr32[1] = 0;
1940
1941                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1942                     ipv6_addr_equal(&ifp->addr, &addr)) {
1943                         /* DAD failed for link-local based on MAC address */
1944                         idev->cnf.disable_ipv6 = 1;
1945
1946                         pr_info("%s: IPv6 being disabled!\n",
1947                                 ifp->idev->dev->name);
1948                 }
1949         }
1950
1951 errdad:
1952         /* transition from _POSTDAD to _ERRDAD */
1953         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1954         spin_unlock_bh(&ifp->lock);
1955
1956         addrconf_mod_dad_work(ifp, 0);
1957 }
1958
1959 /* Join to solicited addr multicast group.
1960  * caller must hold RTNL */
1961 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1962 {
1963         struct in6_addr maddr;
1964
1965         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1966                 return;
1967
1968         addrconf_addr_solict_mult(addr, &maddr);
1969         ipv6_dev_mc_inc(dev, &maddr);
1970 }
1971
1972 /* caller must hold RTNL */
1973 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1974 {
1975         struct in6_addr maddr;
1976
1977         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1978                 return;
1979
1980         addrconf_addr_solict_mult(addr, &maddr);
1981         __ipv6_dev_mc_dec(idev, &maddr);
1982 }
1983
1984 /* caller must hold RTNL */
1985 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1986 {
1987         struct in6_addr addr;
1988
1989         if (ifp->prefix_len >= 127) /* RFC 6164 */
1990                 return;
1991         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1992         if (ipv6_addr_any(&addr))
1993                 return;
1994         __ipv6_dev_ac_inc(ifp->idev, &addr);
1995 }
1996
1997 /* caller must hold RTNL */
1998 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1999 {
2000         struct in6_addr addr;
2001
2002         if (ifp->prefix_len >= 127) /* RFC 6164 */
2003                 return;
2004         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2005         if (ipv6_addr_any(&addr))
2006                 return;
2007         __ipv6_dev_ac_dec(ifp->idev, &addr);
2008 }
2009
2010 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2011 {
2012         if (dev->addr_len != EUI64_ADDR_LEN)
2013                 return -1;
2014         memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2015         eui[0] ^= 2;
2016         return 0;
2017 }
2018
2019 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2020 {
2021         union fwnet_hwaddr *ha;
2022
2023         if (dev->addr_len != FWNET_ALEN)
2024                 return -1;
2025
2026         ha = (union fwnet_hwaddr *)dev->dev_addr;
2027
2028         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2029         eui[0] ^= 2;
2030         return 0;
2031 }
2032
2033 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2034 {
2035         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2036         if (dev->addr_len != ARCNET_ALEN)
2037                 return -1;
2038         memset(eui, 0, 7);
2039         eui[7] = *(u8 *)dev->dev_addr;
2040         return 0;
2041 }
2042
2043 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2044 {
2045         if (dev->addr_len != INFINIBAND_ALEN)
2046                 return -1;
2047         memcpy(eui, dev->dev_addr + 12, 8);
2048         eui[0] |= 2;
2049         return 0;
2050 }
2051
2052 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2053 {
2054         if (addr == 0)
2055                 return -1;
2056         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2057                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2058                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2059                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2060                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2061                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2062         eui[1] = 0;
2063         eui[2] = 0x5E;
2064         eui[3] = 0xFE;
2065         memcpy(eui + 4, &addr, 4);
2066         return 0;
2067 }
2068
2069 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2070 {
2071         if (dev->priv_flags & IFF_ISATAP)
2072                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2073         return -1;
2074 }
2075
2076 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2077 {
2078         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2079 }
2080
2081 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2082 {
2083         memcpy(eui, dev->perm_addr, 3);
2084         memcpy(eui + 5, dev->perm_addr + 3, 3);
2085         eui[3] = 0xFF;
2086         eui[4] = 0xFE;
2087         eui[0] ^= 2;
2088         return 0;
2089 }
2090
2091 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2092 {
2093         switch (dev->type) {
2094         case ARPHRD_ETHER:
2095         case ARPHRD_FDDI:
2096                 return addrconf_ifid_eui48(eui, dev);
2097         case ARPHRD_ARCNET:
2098                 return addrconf_ifid_arcnet(eui, dev);
2099         case ARPHRD_INFINIBAND:
2100                 return addrconf_ifid_infiniband(eui, dev);
2101         case ARPHRD_SIT:
2102                 return addrconf_ifid_sit(eui, dev);
2103         case ARPHRD_IPGRE:
2104                 return addrconf_ifid_gre(eui, dev);
2105         case ARPHRD_6LOWPAN:
2106                 return addrconf_ifid_eui64(eui, dev);
2107         case ARPHRD_IEEE1394:
2108                 return addrconf_ifid_ieee1394(eui, dev);
2109         case ARPHRD_TUNNEL6:
2110                 return addrconf_ifid_ip6tnl(eui, dev);
2111         }
2112         return -1;
2113 }
2114
2115 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2116 {
2117         int err = -1;
2118         struct inet6_ifaddr *ifp;
2119
2120         read_lock_bh(&idev->lock);
2121         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2122                 if (ifp->scope > IFA_LINK)
2123                         break;
2124                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2125                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2126                         err = 0;
2127                         break;
2128                 }
2129         }
2130         read_unlock_bh(&idev->lock);
2131         return err;
2132 }
2133
2134 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2135 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2136 {
2137 regen:
2138         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2139         idev->rndid[0] &= ~0x02;
2140
2141         /*
2142          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2143          * check if generated address is not inappropriate
2144          *
2145          *  - Reserved subnet anycast (RFC 2526)
2146          *      11111101 11....11 1xxxxxxx
2147          *  - ISATAP (RFC4214) 6.1
2148          *      00-00-5E-FE-xx-xx-xx-xx
2149          *  - value 0
2150          *  - XXX: already assigned to an address on the device
2151          */
2152         if (idev->rndid[0] == 0xfd &&
2153             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2154             (idev->rndid[7]&0x80))
2155                 goto regen;
2156         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2157                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2158                         goto regen;
2159                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2160                         goto regen;
2161         }
2162 }
2163
2164 static void ipv6_regen_rndid(unsigned long data)
2165 {
2166         struct inet6_dev *idev = (struct inet6_dev *) data;
2167         unsigned long expires;
2168
2169         rcu_read_lock_bh();
2170         write_lock_bh(&idev->lock);
2171
2172         if (idev->dead)
2173                 goto out;
2174
2175         __ipv6_regen_rndid(idev);
2176
2177         expires = jiffies +
2178                 idev->cnf.temp_prefered_lft * HZ -
2179                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2180                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2181                 idev->cnf.max_desync_factor * HZ;
2182         if (time_before(expires, jiffies)) {
2183                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2184                         __func__, idev->dev->name);
2185                 goto out;
2186         }
2187
2188         if (!mod_timer(&idev->regen_timer, expires))
2189                 in6_dev_hold(idev);
2190
2191 out:
2192         write_unlock_bh(&idev->lock);
2193         rcu_read_unlock_bh();
2194         in6_dev_put(idev);
2195 }
2196
2197 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2198 {
2199         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2200                 __ipv6_regen_rndid(idev);
2201 }
2202
2203 /*
2204  *      Add prefix route.
2205  */
2206
2207 static void
2208 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2209                       unsigned long expires, u32 flags)
2210 {
2211         struct fib6_config cfg = {
2212                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2213                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2214                 .fc_ifindex = dev->ifindex,
2215                 .fc_expires = expires,
2216                 .fc_dst_len = plen,
2217                 .fc_flags = RTF_UP | flags,
2218                 .fc_nlinfo.nl_net = dev_net(dev),
2219                 .fc_protocol = RTPROT_KERNEL,
2220         };
2221
2222         cfg.fc_dst = *pfx;
2223
2224         /* Prevent useless cloning on PtP SIT.
2225            This thing is done here expecting that the whole
2226            class of non-broadcast devices need not cloning.
2227          */
2228 #if IS_ENABLED(CONFIG_IPV6_SIT)
2229         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2230                 cfg.fc_flags |= RTF_NONEXTHOP;
2231 #endif
2232
2233         ip6_route_add(&cfg);
2234 }
2235
2236
2237 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2238                                                   int plen,
2239                                                   const struct net_device *dev,
2240                                                   u32 flags, u32 noflags)
2241 {
2242         struct fib6_node *fn;
2243         struct rt6_info *rt = NULL;
2244         struct fib6_table *table;
2245         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2246
2247         table = fib6_get_table(dev_net(dev), tb_id);
2248         if (!table)
2249                 return NULL;
2250
2251         read_lock_bh(&table->tb6_lock);
2252         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2253         if (!fn)
2254                 goto out;
2255
2256         noflags |= RTF_CACHE;
2257         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2258                 if (rt->dst.dev->ifindex != dev->ifindex)
2259                         continue;
2260                 if ((rt->rt6i_flags & flags) != flags)
2261                         continue;
2262                 if ((rt->rt6i_flags & noflags) != 0)
2263                         continue;
2264                 dst_hold(&rt->dst);
2265                 break;
2266         }
2267 out:
2268         read_unlock_bh(&table->tb6_lock);
2269         return rt;
2270 }
2271
2272
2273 /* Create "default" multicast route to the interface */
2274
2275 static void addrconf_add_mroute(struct net_device *dev)
2276 {
2277         struct fib6_config cfg = {
2278                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2279                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2280                 .fc_ifindex = dev->ifindex,
2281                 .fc_dst_len = 8,
2282                 .fc_flags = RTF_UP,
2283                 .fc_nlinfo.nl_net = dev_net(dev),
2284         };
2285
2286         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2287
2288         ip6_route_add(&cfg);
2289 }
2290
2291 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2292 {
2293         struct inet6_dev *idev;
2294
2295         ASSERT_RTNL();
2296
2297         idev = ipv6_find_idev(dev);
2298         if (!idev)
2299                 return ERR_PTR(-ENOBUFS);
2300
2301         if (idev->cnf.disable_ipv6)
2302                 return ERR_PTR(-EACCES);
2303
2304         /* Add default multicast route */
2305         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2306                 addrconf_add_mroute(dev);
2307
2308         return idev;
2309 }
2310
2311 static void manage_tempaddrs(struct inet6_dev *idev,
2312                              struct inet6_ifaddr *ifp,
2313                              __u32 valid_lft, __u32 prefered_lft,
2314                              bool create, unsigned long now)
2315 {
2316         u32 flags;
2317         struct inet6_ifaddr *ift;
2318
2319         read_lock_bh(&idev->lock);
2320         /* update all temporary addresses in the list */
2321         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2322                 int age, max_valid, max_prefered;
2323
2324                 if (ifp != ift->ifpub)
2325                         continue;
2326
2327                 /* RFC 4941 section 3.3:
2328                  * If a received option will extend the lifetime of a public
2329                  * address, the lifetimes of temporary addresses should
2330                  * be extended, subject to the overall constraint that no
2331                  * temporary addresses should ever remain "valid" or "preferred"
2332                  * for a time longer than (TEMP_VALID_LIFETIME) or
2333                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2334                  */
2335                 age = (now - ift->cstamp) / HZ;
2336                 max_valid = idev->cnf.temp_valid_lft - age;
2337                 if (max_valid < 0)
2338                         max_valid = 0;
2339
2340                 max_prefered = idev->cnf.temp_prefered_lft -
2341                                idev->cnf.max_desync_factor - age;
2342                 if (max_prefered < 0)
2343                         max_prefered = 0;
2344
2345                 if (valid_lft > max_valid)
2346                         valid_lft = max_valid;
2347
2348                 if (prefered_lft > max_prefered)
2349                         prefered_lft = max_prefered;
2350
2351                 spin_lock(&ift->lock);
2352                 flags = ift->flags;
2353                 ift->valid_lft = valid_lft;
2354                 ift->prefered_lft = prefered_lft;
2355                 ift->tstamp = now;
2356                 if (prefered_lft > 0)
2357                         ift->flags &= ~IFA_F_DEPRECATED;
2358
2359                 spin_unlock(&ift->lock);
2360                 if (!(flags&IFA_F_TENTATIVE))
2361                         ipv6_ifa_notify(0, ift);
2362         }
2363
2364         if ((create || list_empty(&idev->tempaddr_list)) &&
2365             idev->cnf.use_tempaddr > 0) {
2366                 /* When a new public address is created as described
2367                  * in [ADDRCONF], also create a new temporary address.
2368                  * Also create a temporary address if it's enabled but
2369                  * no temporary address currently exists.
2370                  */
2371                 read_unlock_bh(&idev->lock);
2372                 ipv6_create_tempaddr(ifp, NULL);
2373         } else {
2374                 read_unlock_bh(&idev->lock);
2375         }
2376 }
2377
2378 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2379 {
2380         return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2381                idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2382 }
2383
2384 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2385                                  const struct prefix_info *pinfo,
2386                                  struct inet6_dev *in6_dev,
2387                                  const struct in6_addr *addr, int addr_type,
2388                                  u32 addr_flags, bool sllao, bool tokenized,
2389                                  __u32 valid_lft, u32 prefered_lft)
2390 {
2391         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2392         int create = 0, update_lft = 0;
2393
2394         if (!ifp && valid_lft) {
2395                 int max_addresses = in6_dev->cnf.max_addresses;
2396
2397 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2398                 if (in6_dev->cnf.optimistic_dad &&
2399                     !net->ipv6.devconf_all->forwarding && sllao)
2400                         addr_flags |= IFA_F_OPTIMISTIC;
2401 #endif
2402
2403                 /* Do not allow to create too much of autoconfigured
2404                  * addresses; this would be too easy way to crash kernel.
2405                  */
2406                 if (!max_addresses ||
2407                     ipv6_count_addresses(in6_dev) < max_addresses)
2408                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2409                                             pinfo->prefix_len,
2410                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2411                                             addr_flags, valid_lft,
2412                                             prefered_lft);
2413
2414                 if (IS_ERR_OR_NULL(ifp))
2415                         return -1;
2416
2417                 update_lft = 0;
2418                 create = 1;
2419                 spin_lock_bh(&ifp->lock);
2420                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2421                 ifp->cstamp = jiffies;
2422                 ifp->tokenized = tokenized;
2423                 spin_unlock_bh(&ifp->lock);
2424                 addrconf_dad_start(ifp);
2425         }
2426
2427         if (ifp) {
2428                 u32 flags;
2429                 unsigned long now;
2430                 u32 stored_lft;
2431
2432                 /* update lifetime (RFC2462 5.5.3 e) */
2433                 spin_lock_bh(&ifp->lock);
2434                 now = jiffies;
2435                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2436                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2437                 else
2438                         stored_lft = 0;
2439                 if (!update_lft && !create && stored_lft) {
2440                         const u32 minimum_lft = min_t(u32,
2441                                 stored_lft, MIN_VALID_LIFETIME);
2442                         valid_lft = max(valid_lft, minimum_lft);
2443
2444                         /* RFC4862 Section 5.5.3e:
2445                          * "Note that the preferred lifetime of the
2446                          *  corresponding address is always reset to
2447                          *  the Preferred Lifetime in the received
2448                          *  Prefix Information option, regardless of
2449                          *  whether the valid lifetime is also reset or
2450                          *  ignored."
2451                          *
2452                          * So we should always update prefered_lft here.
2453                          */
2454                         update_lft = 1;
2455                 }
2456
2457                 if (update_lft) {
2458                         ifp->valid_lft = valid_lft;
2459                         ifp->prefered_lft = prefered_lft;
2460                         ifp->tstamp = now;
2461                         flags = ifp->flags;
2462                         ifp->flags &= ~IFA_F_DEPRECATED;
2463                         spin_unlock_bh(&ifp->lock);
2464
2465                         if (!(flags&IFA_F_TENTATIVE))
2466                                 ipv6_ifa_notify(0, ifp);
2467                 } else
2468                         spin_unlock_bh(&ifp->lock);
2469
2470                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2471                                  create, now);
2472
2473                 in6_ifa_put(ifp);
2474                 addrconf_verify();
2475         }
2476
2477         return 0;
2478 }
2479 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2480
2481 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2482 {
2483         struct prefix_info *pinfo;
2484         __u32 valid_lft;
2485         __u32 prefered_lft;
2486         int addr_type, err;
2487         u32 addr_flags = 0;
2488         struct inet6_dev *in6_dev;
2489         struct net *net = dev_net(dev);
2490
2491         pinfo = (struct prefix_info *) opt;
2492
2493         if (len < sizeof(struct prefix_info)) {
2494                 ADBG("addrconf: prefix option too short\n");
2495                 return;
2496         }
2497
2498         /*
2499          *      Validation checks ([ADDRCONF], page 19)
2500          */
2501
2502         addr_type = ipv6_addr_type(&pinfo->prefix);
2503
2504         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2505                 return;
2506
2507         valid_lft = ntohl(pinfo->valid);
2508         prefered_lft = ntohl(pinfo->prefered);
2509
2510         if (prefered_lft > valid_lft) {
2511                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2512                 return;
2513         }
2514
2515         in6_dev = in6_dev_get(dev);
2516
2517         if (!in6_dev) {
2518                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2519                                     dev->name);
2520                 return;
2521         }
2522
2523         /*
2524          *      Two things going on here:
2525          *      1) Add routes for on-link prefixes
2526          *      2) Configure prefixes with the auto flag set
2527          */
2528
2529         if (pinfo->onlink) {
2530                 struct rt6_info *rt;
2531                 unsigned long rt_expires;
2532
2533                 /* Avoid arithmetic overflow. Really, we could
2534                  * save rt_expires in seconds, likely valid_lft,
2535                  * but it would require division in fib gc, that it
2536                  * not good.
2537                  */
2538                 if (HZ > USER_HZ)
2539                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2540                 else
2541                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2542
2543                 if (addrconf_finite_timeout(rt_expires))
2544                         rt_expires *= HZ;
2545
2546                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2547                                                pinfo->prefix_len,
2548                                                dev,
2549                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2550                                                RTF_GATEWAY | RTF_DEFAULT);
2551
2552                 if (rt) {
2553                         /* Autoconf prefix route */
2554                         if (valid_lft == 0) {
2555                                 ip6_del_rt(rt);
2556                                 rt = NULL;
2557                         } else if (addrconf_finite_timeout(rt_expires)) {
2558                                 /* not infinity */
2559                                 rt6_set_expires(rt, jiffies + rt_expires);
2560                         } else {
2561                                 rt6_clean_expires(rt);
2562                         }
2563                 } else if (valid_lft) {
2564                         clock_t expires = 0;
2565                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2566                         if (addrconf_finite_timeout(rt_expires)) {
2567                                 /* not infinity */
2568                                 flags |= RTF_EXPIRES;
2569                                 expires = jiffies_to_clock_t(rt_expires);
2570                         }
2571                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2572                                               dev, expires, flags);
2573                 }
2574                 ip6_rt_put(rt);
2575         }
2576
2577         /* Try to figure out our local address for this prefix */
2578
2579         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2580                 struct in6_addr addr;
2581                 bool tokenized = false, dev_addr_generated = false;
2582
2583                 if (pinfo->prefix_len == 64) {
2584                         memcpy(&addr, &pinfo->prefix, 8);
2585
2586                         if (!ipv6_addr_any(&in6_dev->token)) {
2587                                 read_lock_bh(&in6_dev->lock);
2588                                 memcpy(addr.s6_addr + 8,
2589                                        in6_dev->token.s6_addr + 8, 8);
2590                                 read_unlock_bh(&in6_dev->lock);
2591                                 tokenized = true;
2592                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2593                                    !ipv6_generate_stable_address(&addr, 0,
2594                                                                  in6_dev)) {
2595                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2596                                 goto ok;
2597                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2598                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2599                                 goto put;
2600                         } else {
2601                                 dev_addr_generated = true;
2602                         }
2603                         goto ok;
2604                 }
2605                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2606                                     pinfo->prefix_len);
2607                 goto put;
2608
2609 ok:
2610                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2611                                                    &addr, addr_type,
2612                                                    addr_flags, sllao,
2613                                                    tokenized, valid_lft,
2614                                                    prefered_lft);
2615                 if (err)
2616                         goto put;
2617
2618                 /* Ignore error case here because previous prefix add addr was
2619                  * successful which will be notified.
2620                  */
2621                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2622                                               addr_type, addr_flags, sllao,
2623                                               tokenized, valid_lft,
2624                                               prefered_lft,
2625                                               dev_addr_generated);
2626         }
2627         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2628 put:
2629         in6_dev_put(in6_dev);
2630 }
2631
2632 /*
2633  *      Set destination address.
2634  *      Special case for SIT interfaces where we create a new "virtual"
2635  *      device.
2636  */
2637 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2638 {
2639         struct in6_ifreq ireq;
2640         struct net_device *dev;
2641         int err = -EINVAL;
2642
2643         rtnl_lock();
2644
2645         err = -EFAULT;
2646         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2647                 goto err_exit;
2648
2649         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2650
2651         err = -ENODEV;
2652         if (!dev)
2653                 goto err_exit;
2654
2655 #if IS_ENABLED(CONFIG_IPV6_SIT)
2656         if (dev->type == ARPHRD_SIT) {
2657                 const struct net_device_ops *ops = dev->netdev_ops;
2658                 struct ifreq ifr;
2659                 struct ip_tunnel_parm p;
2660
2661                 err = -EADDRNOTAVAIL;
2662                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2663                         goto err_exit;
2664
2665                 memset(&p, 0, sizeof(p));
2666                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2667                 p.iph.saddr = 0;
2668                 p.iph.version = 4;
2669                 p.iph.ihl = 5;
2670                 p.iph.protocol = IPPROTO_IPV6;
2671                 p.iph.ttl = 64;
2672                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2673
2674                 if (ops->ndo_do_ioctl) {
2675                         mm_segment_t oldfs = get_fs();
2676
2677                         set_fs(KERNEL_DS);
2678                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2679                         set_fs(oldfs);
2680                 } else
2681                         err = -EOPNOTSUPP;
2682
2683                 if (err == 0) {
2684                         err = -ENOBUFS;
2685                         dev = __dev_get_by_name(net, p.name);
2686                         if (!dev)
2687                                 goto err_exit;
2688                         err = dev_open(dev);
2689                 }
2690         }
2691 #endif
2692
2693 err_exit:
2694         rtnl_unlock();
2695         return err;
2696 }
2697
2698 static int ipv6_mc_config(struct sock *sk, bool join,
2699                           const struct in6_addr *addr, int ifindex)
2700 {
2701         int ret;
2702
2703         ASSERT_RTNL();
2704
2705         lock_sock(sk);
2706         if (join)
2707                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2708         else
2709                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2710         release_sock(sk);
2711
2712         return ret;
2713 }
2714
2715 /*
2716  *      Manual configuration of address on an interface
2717  */
2718 static int inet6_addr_add(struct net *net, int ifindex,
2719                           const struct in6_addr *pfx,
2720                           const struct in6_addr *peer_pfx,
2721                           unsigned int plen, __u32 ifa_flags,
2722                           __u32 prefered_lft, __u32 valid_lft)
2723 {
2724         struct inet6_ifaddr *ifp;
2725         struct inet6_dev *idev;
2726         struct net_device *dev;
2727         unsigned long timeout;
2728         clock_t expires;
2729         int scope;
2730         u32 flags;
2731
2732         ASSERT_RTNL();
2733
2734         if (plen > 128)
2735                 return -EINVAL;
2736
2737         /* check the lifetime */
2738         if (!valid_lft || prefered_lft > valid_lft)
2739                 return -EINVAL;
2740
2741         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2742                 return -EINVAL;
2743
2744         dev = __dev_get_by_index(net, ifindex);
2745         if (!dev)
2746                 return -ENODEV;
2747
2748         idev = addrconf_add_dev(dev);
2749         if (IS_ERR(idev))
2750                 return PTR_ERR(idev);
2751
2752         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2753                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2754                                          true, pfx, ifindex);
2755
2756                 if (ret < 0)
2757                         return ret;
2758         }
2759
2760         scope = ipv6_addr_scope(pfx);
2761
2762         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2763         if (addrconf_finite_timeout(timeout)) {
2764                 expires = jiffies_to_clock_t(timeout * HZ);
2765                 valid_lft = timeout;
2766                 flags = RTF_EXPIRES;
2767         } else {
2768                 expires = 0;
2769                 flags = 0;
2770                 ifa_flags |= IFA_F_PERMANENT;
2771         }
2772
2773         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2774         if (addrconf_finite_timeout(timeout)) {
2775                 if (timeout == 0)
2776                         ifa_flags |= IFA_F_DEPRECATED;
2777                 prefered_lft = timeout;
2778         }
2779
2780         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2781                             valid_lft, prefered_lft);
2782
2783         if (!IS_ERR(ifp)) {
2784                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2785                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2786                                               expires, flags);
2787                 }
2788
2789                 /*
2790                  * Note that section 3.1 of RFC 4429 indicates
2791                  * that the Optimistic flag should not be set for
2792                  * manually configured addresses
2793                  */
2794                 addrconf_dad_start(ifp);
2795                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2796                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2797                                          true, jiffies);
2798                 in6_ifa_put(ifp);
2799                 addrconf_verify_rtnl();
2800                 return 0;
2801         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2802                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2803                                false, pfx, ifindex);
2804         }
2805
2806         return PTR_ERR(ifp);
2807 }
2808
2809 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2810                           const struct in6_addr *pfx, unsigned int plen)
2811 {
2812         struct inet6_ifaddr *ifp;
2813         struct inet6_dev *idev;
2814         struct net_device *dev;
2815
2816         if (plen > 128)
2817                 return -EINVAL;
2818
2819         dev = __dev_get_by_index(net, ifindex);
2820         if (!dev)
2821                 return -ENODEV;
2822
2823         idev = __in6_dev_get(dev);
2824         if (!idev)
2825                 return -ENXIO;
2826
2827         read_lock_bh(&idev->lock);
2828         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2829                 if (ifp->prefix_len == plen &&
2830                     ipv6_addr_equal(pfx, &ifp->addr)) {
2831                         in6_ifa_hold(ifp);
2832                         read_unlock_bh(&idev->lock);
2833
2834                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2835                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2836                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2837                                                  jiffies);
2838                         ipv6_del_addr(ifp);
2839                         addrconf_verify_rtnl();
2840                         if (ipv6_addr_is_multicast(pfx)) {
2841                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2842                                                false, pfx, dev->ifindex);
2843                         }
2844                         return 0;
2845                 }
2846         }
2847         read_unlock_bh(&idev->lock);
2848         return -EADDRNOTAVAIL;
2849 }
2850
2851
2852 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2853 {
2854         struct in6_ifreq ireq;
2855         int err;
2856
2857         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2858                 return -EPERM;
2859
2860         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2861                 return -EFAULT;
2862
2863         rtnl_lock();
2864         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2865                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2866                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2867         rtnl_unlock();
2868         return err;
2869 }
2870
2871 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2872 {
2873         struct in6_ifreq ireq;
2874         int err;
2875
2876         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2877                 return -EPERM;
2878
2879         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2880                 return -EFAULT;
2881
2882         rtnl_lock();
2883         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2884                              ireq.ifr6_prefixlen);
2885         rtnl_unlock();
2886         return err;
2887 }
2888
2889 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2890                      int plen, int scope)
2891 {
2892         struct inet6_ifaddr *ifp;
2893
2894         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2895                             scope, IFA_F_PERMANENT,
2896                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2897         if (!IS_ERR(ifp)) {
2898                 spin_lock_bh(&ifp->lock);
2899                 ifp->flags &= ~IFA_F_TENTATIVE;
2900                 spin_unlock_bh(&ifp->lock);
2901                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2902                 in6_ifa_put(ifp);
2903         }
2904 }
2905
2906 #if IS_ENABLED(CONFIG_IPV6_SIT)
2907 static void sit_add_v4_addrs(struct inet6_dev *idev)
2908 {
2909         struct in6_addr addr;
2910         struct net_device *dev;
2911         struct net *net = dev_net(idev->dev);
2912         int scope, plen;
2913         u32 pflags = 0;
2914
2915         ASSERT_RTNL();
2916
2917         memset(&addr, 0, sizeof(struct in6_addr));
2918         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2919
2920         if (idev->dev->flags&IFF_POINTOPOINT) {
2921                 addr.s6_addr32[0] = htonl(0xfe800000);
2922                 scope = IFA_LINK;
2923                 plen = 64;
2924         } else {
2925                 scope = IPV6_ADDR_COMPATv4;
2926                 plen = 96;
2927                 pflags |= RTF_NONEXTHOP;
2928         }
2929
2930         if (addr.s6_addr32[3]) {
2931                 add_addr(idev, &addr, plen, scope);
2932                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2933                 return;
2934         }
2935
2936         for_each_netdev(net, dev) {
2937                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2938                 if (in_dev && (dev->flags & IFF_UP)) {
2939                         struct in_ifaddr *ifa;
2940
2941                         int flag = scope;
2942
2943                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2944
2945                                 addr.s6_addr32[3] = ifa->ifa_local;
2946
2947                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2948                                         continue;
2949                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2950                                         if (idev->dev->flags&IFF_POINTOPOINT)
2951                                                 continue;
2952                                         flag |= IFA_HOST;
2953                                 }
2954
2955                                 add_addr(idev, &addr, plen, flag);
2956                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2957                                                       pflags);
2958                         }
2959                 }
2960         }
2961 }
2962 #endif
2963
2964 static void init_loopback(struct net_device *dev)
2965 {
2966         struct inet6_dev  *idev;
2967         struct net_device *sp_dev;
2968         struct inet6_ifaddr *sp_ifa;
2969         struct rt6_info *sp_rt;
2970
2971         /* ::1 */
2972
2973         ASSERT_RTNL();
2974
2975         idev = ipv6_find_idev(dev);
2976         if (!idev) {
2977                 pr_debug("%s: add_dev failed\n", __func__);
2978                 return;
2979         }
2980
2981         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2982
2983         /* Add routes to other interface's IPv6 addresses */
2984         for_each_netdev(dev_net(dev), sp_dev) {
2985                 if (!strcmp(sp_dev->name, dev->name))
2986                         continue;
2987
2988                 idev = __in6_dev_get(sp_dev);
2989                 if (!idev)
2990                         continue;
2991
2992                 read_lock_bh(&idev->lock);
2993                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2994
2995                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2996                                 continue;
2997
2998                         if (sp_ifa->rt) {
2999                                 /* This dst has been added to garbage list when
3000                                  * lo device down, release this obsolete dst and
3001                                  * reallocate a new router for ifa.
3002                                  */
3003                                 if (sp_ifa->rt->dst.obsolete > 0) {
3004                                         ip6_rt_put(sp_ifa->rt);
3005                                         sp_ifa->rt = NULL;
3006                                 } else {
3007                                         continue;
3008                                 }
3009                         }
3010
3011                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3012
3013                         /* Failure cases are ignored */
3014                         if (!IS_ERR(sp_rt)) {
3015                                 sp_ifa->rt = sp_rt;
3016                                 ip6_ins_rt(sp_rt);
3017                         }
3018                 }
3019                 read_unlock_bh(&idev->lock);
3020         }
3021 }
3022
3023 void addrconf_add_linklocal(struct inet6_dev *idev,
3024                             const struct in6_addr *addr, u32 flags)
3025 {
3026         struct inet6_ifaddr *ifp;
3027         u32 addr_flags = flags | IFA_F_PERMANENT;
3028
3029 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3030         if (idev->cnf.optimistic_dad &&
3031             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3032                 addr_flags |= IFA_F_OPTIMISTIC;
3033 #endif
3034
3035         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3036                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3037         if (!IS_ERR(ifp)) {
3038                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3039                 addrconf_dad_start(ifp);
3040                 in6_ifa_put(ifp);
3041         }
3042 }
3043 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3044
3045 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3046 {
3047         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3048                 return true;
3049
3050         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3051             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3052                 return true;
3053
3054         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3055             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3056                 return true;
3057
3058         return false;
3059 }
3060
3061 static int ipv6_generate_stable_address(struct in6_addr *address,
3062                                         u8 dad_count,
3063                                         const struct inet6_dev *idev)
3064 {
3065         static DEFINE_SPINLOCK(lock);
3066         static __u32 digest[SHA_DIGEST_WORDS];
3067         static __u32 workspace[SHA_WORKSPACE_WORDS];
3068
3069         static union {
3070                 char __data[SHA_MESSAGE_BYTES];
3071                 struct {
3072                         struct in6_addr secret;
3073                         __be32 prefix[2];
3074                         unsigned char hwaddr[MAX_ADDR_LEN];
3075                         u8 dad_count;
3076                 } __packed;
3077         } data;
3078
3079         struct in6_addr secret;
3080         struct in6_addr temp;
3081         struct net *net = dev_net(idev->dev);
3082
3083         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3084
3085         if (idev->cnf.stable_secret.initialized)
3086                 secret = idev->cnf.stable_secret.secret;
3087         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3088                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3089         else
3090                 return -1;
3091
3092 retry:
3093         spin_lock_bh(&lock);
3094
3095         sha_init(digest);
3096         memset(&data, 0, sizeof(data));
3097         memset(workspace, 0, sizeof(workspace));
3098         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3099         data.prefix[0] = address->s6_addr32[0];
3100         data.prefix[1] = address->s6_addr32[1];
3101         data.secret = secret;
3102         data.dad_count = dad_count;
3103
3104         sha_transform(digest, data.__data, workspace);
3105
3106         temp = *address;
3107         temp.s6_addr32[2] = (__force __be32)digest[0];
3108         temp.s6_addr32[3] = (__force __be32)digest[1];
3109
3110         spin_unlock_bh(&lock);
3111
3112         if (ipv6_reserved_interfaceid(temp)) {
3113                 dad_count++;
3114                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3115                         return -1;
3116                 goto retry;
3117         }
3118
3119         *address = temp;
3120         return 0;
3121 }
3122
3123 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3124 {
3125         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3126
3127         if (s->initialized)
3128                 return;
3129         s = &idev->cnf.stable_secret;
3130         get_random_bytes(&s->secret, sizeof(s->secret));
3131         s->initialized = true;
3132 }
3133
3134 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3135 {
3136         struct in6_addr addr;
3137
3138         /* no link local addresses on L3 master devices */
3139         if (netif_is_l3_master(idev->dev))
3140                 return;
3141
3142         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3143
3144         switch (idev->addr_gen_mode) {
3145         case IN6_ADDR_GEN_MODE_RANDOM:
3146                 ipv6_gen_mode_random_init(idev);
3147                 /* fallthrough */
3148         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3149                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3150                         addrconf_add_linklocal(idev, &addr,
3151                                                IFA_F_STABLE_PRIVACY);
3152                 else if (prefix_route)
3153                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3154                 break;
3155         case IN6_ADDR_GEN_MODE_EUI64:
3156                 /* addrconf_add_linklocal also adds a prefix_route and we
3157                  * only need to care about prefix routes if ipv6_generate_eui64
3158                  * couldn't generate one.
3159                  */
3160                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3161                         addrconf_add_linklocal(idev, &addr, 0);
3162                 else if (prefix_route)
3163                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3164                 break;
3165         case IN6_ADDR_GEN_MODE_NONE:
3166         default:
3167                 /* will not add any link local address */
3168                 break;
3169         }
3170 }
3171
3172 static void addrconf_dev_config(struct net_device *dev)
3173 {
3174         struct inet6_dev *idev;
3175
3176         ASSERT_RTNL();
3177
3178         if ((dev->type != ARPHRD_ETHER) &&
3179             (dev->type != ARPHRD_FDDI) &&
3180             (dev->type != ARPHRD_ARCNET) &&
3181             (dev->type != ARPHRD_INFINIBAND) &&
3182             (dev->type != ARPHRD_IEEE1394) &&
3183             (dev->type != ARPHRD_TUNNEL6) &&
3184             (dev->type != ARPHRD_6LOWPAN) &&
3185             (dev->type != ARPHRD_NONE)) {
3186                 /* Alas, we support only Ethernet autoconfiguration. */
3187                 return;
3188         }
3189
3190         idev = addrconf_add_dev(dev);
3191         if (IS_ERR(idev))
3192                 return;
3193
3194         /* this device type has no EUI support */
3195         if (dev->type == ARPHRD_NONE &&
3196             idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3197                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3198
3199         addrconf_addr_gen(idev, false);
3200 }
3201
3202 #if IS_ENABLED(CONFIG_IPV6_SIT)
3203 static void addrconf_sit_config(struct net_device *dev)
3204 {
3205         struct inet6_dev *idev;
3206
3207         ASSERT_RTNL();
3208
3209         /*
3210          * Configure the tunnel with one of our IPv4
3211          * addresses... we should configure all of
3212          * our v4 addrs in the tunnel
3213          */
3214
3215         idev = ipv6_find_idev(dev);
3216         if (!idev) {
3217                 pr_debug("%s: add_dev failed\n", __func__);
3218                 return;
3219         }
3220
3221         if (dev->priv_flags & IFF_ISATAP) {
3222                 addrconf_addr_gen(idev, false);
3223                 return;
3224         }
3225
3226         sit_add_v4_addrs(idev);
3227
3228         if (dev->flags&IFF_POINTOPOINT)
3229                 addrconf_add_mroute(dev);
3230 }
3231 #endif
3232
3233 #if IS_ENABLED(CONFIG_NET_IPGRE)
3234 static void addrconf_gre_config(struct net_device *dev)
3235 {
3236         struct inet6_dev *idev;
3237
3238         ASSERT_RTNL();
3239
3240         idev = ipv6_find_idev(dev);
3241         if (!idev) {
3242                 pr_debug("%s: add_dev failed\n", __func__);
3243                 return;
3244         }
3245
3246         addrconf_addr_gen(idev, true);
3247         if (dev->flags & IFF_POINTOPOINT)
3248                 addrconf_add_mroute(dev);
3249 }
3250 #endif
3251
3252 static int fixup_permanent_addr(struct inet6_dev *idev,
3253                                 struct inet6_ifaddr *ifp)
3254 {
3255         if (!ifp->rt) {
3256                 struct rt6_info *rt;
3257
3258                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3259                 if (unlikely(IS_ERR(rt)))
3260                         return PTR_ERR(rt);
3261
3262                 ifp->rt = rt;
3263         }
3264
3265         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3266                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3267                                       idev->dev, 0, 0);
3268         }
3269
3270         addrconf_dad_start(ifp);
3271
3272         return 0;
3273 }
3274
3275 static void addrconf_permanent_addr(struct net_device *dev)
3276 {
3277         struct inet6_ifaddr *ifp, *tmp;
3278         struct inet6_dev *idev;
3279
3280         idev = __in6_dev_get(dev);
3281         if (!idev)
3282                 return;
3283
3284         write_lock_bh(&idev->lock);
3285
3286         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3287                 if ((ifp->flags & IFA_F_PERMANENT) &&
3288                     fixup_permanent_addr(idev, ifp) < 0) {
3289                         write_unlock_bh(&idev->lock);
3290                         ipv6_del_addr(ifp);
3291                         write_lock_bh(&idev->lock);
3292
3293                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3294                                              idev->dev->name, &ifp->addr);
3295                 }
3296         }
3297
3298         write_unlock_bh(&idev->lock);
3299 }
3300
3301 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3302                            void *ptr)
3303 {
3304         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3305         struct netdev_notifier_changeupper_info *info;
3306         struct inet6_dev *idev = __in6_dev_get(dev);
3307         int run_pending = 0;
3308         int err;
3309
3310         switch (event) {
3311         case NETDEV_REGISTER:
3312                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3313                         idev = ipv6_add_dev(dev);
3314                         if (IS_ERR(idev))
3315                                 return notifier_from_errno(PTR_ERR(idev));
3316                 }
3317                 break;
3318
3319         case NETDEV_CHANGEMTU:
3320                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3321                 if (dev->mtu < IPV6_MIN_MTU) {
3322                         addrconf_ifdown(dev, 1);
3323                         break;
3324                 }
3325
3326                 if (idev) {
3327                         rt6_mtu_change(dev, dev->mtu);
3328                         idev->cnf.mtu6 = dev->mtu;
3329                         break;
3330                 }
3331
3332                 /* allocate new idev */
3333                 idev = ipv6_add_dev(dev);
3334                 if (IS_ERR(idev))
3335                         break;
3336
3337                 /* device is still not ready */
3338                 if (!(idev->if_flags & IF_READY))
3339                         break;
3340
3341                 run_pending = 1;
3342
3343                 /* fall through */
3344
3345         case NETDEV_UP:
3346         case NETDEV_CHANGE:
3347                 if (dev->flags & IFF_SLAVE)
3348                         break;
3349
3350                 if (idev && idev->cnf.disable_ipv6)
3351                         break;
3352
3353                 if (event == NETDEV_UP) {
3354                         /* restore routes for permanent addresses */
3355                         addrconf_permanent_addr(dev);
3356
3357                         if (!addrconf_qdisc_ok(dev)) {
3358                                 /* device is not ready yet. */
3359                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3360                                         dev->name);
3361                                 break;
3362                         }
3363
3364                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3365                                 idev = ipv6_add_dev(dev);
3366
3367                         if (!IS_ERR_OR_NULL(idev)) {
3368                                 idev->if_flags |= IF_READY;
3369                                 run_pending = 1;
3370                         }
3371                 } else if (event == NETDEV_CHANGE) {
3372                         if (!addrconf_qdisc_ok(dev)) {
3373                                 /* device is still not ready. */
3374                                 break;
3375                         }
3376
3377                         if (idev) {
3378                                 if (idev->if_flags & IF_READY)
3379                                         /* device is already configured. */
3380                                         break;
3381                                 idev->if_flags |= IF_READY;
3382                         }
3383
3384                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3385                                 dev->name);
3386
3387                         run_pending = 1;
3388                 }
3389
3390                 switch (dev->type) {
3391 #if IS_ENABLED(CONFIG_IPV6_SIT)
3392                 case ARPHRD_SIT:
3393                         addrconf_sit_config(dev);
3394                         break;
3395 #endif
3396 #if IS_ENABLED(CONFIG_NET_IPGRE)
3397                 case ARPHRD_IPGRE:
3398                         addrconf_gre_config(dev);
3399                         break;
3400 #endif
3401                 case ARPHRD_LOOPBACK:
3402                         init_loopback(dev);
3403                         break;
3404
3405                 default:
3406                         addrconf_dev_config(dev);
3407                         break;
3408                 }
3409
3410                 if (!IS_ERR_OR_NULL(idev)) {
3411                         if (run_pending)
3412                                 addrconf_dad_run(idev);
3413
3414                         /*
3415                          * If the MTU changed during the interface down,
3416                          * when the interface up, the changed MTU must be
3417                          * reflected in the idev as well as routers.
3418                          */
3419                         if (idev->cnf.mtu6 != dev->mtu &&
3420                             dev->mtu >= IPV6_MIN_MTU) {
3421                                 rt6_mtu_change(dev, dev->mtu);
3422                                 idev->cnf.mtu6 = dev->mtu;
3423                         }
3424                         idev->tstamp = jiffies;
3425                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3426
3427                         /*
3428                          * If the changed mtu during down is lower than
3429                          * IPV6_MIN_MTU stop IPv6 on this interface.
3430                          */
3431                         if (dev->mtu < IPV6_MIN_MTU)
3432                                 addrconf_ifdown(dev, 1);
3433                 }
3434                 break;
3435
3436         case NETDEV_DOWN:
3437         case NETDEV_UNREGISTER:
3438                 /*
3439                  *      Remove all addresses from this interface.
3440                  */
3441                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3442                 break;
3443
3444         case NETDEV_CHANGENAME:
3445                 if (idev) {
3446                         snmp6_unregister_dev(idev);
3447                         addrconf_sysctl_unregister(idev);
3448                         err = addrconf_sysctl_register(idev);
3449                         if (err)
3450                                 return notifier_from_errno(err);
3451                         err = snmp6_register_dev(idev);
3452                         if (err) {
3453                                 addrconf_sysctl_unregister(idev);
3454                                 return notifier_from_errno(err);
3455                         }
3456                 }
3457                 break;
3458
3459         case NETDEV_PRE_TYPE_CHANGE:
3460         case NETDEV_POST_TYPE_CHANGE:
3461                 if (idev)
3462                         addrconf_type_change(dev, event);
3463                 break;
3464
3465         case NETDEV_CHANGEUPPER:
3466                 info = ptr;
3467
3468                 /* flush all routes if dev is linked to or unlinked from
3469                  * an L3 master device (e.g., VRF)
3470                  */
3471                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3472                         addrconf_ifdown(dev, 0);
3473         }
3474
3475         return NOTIFY_OK;
3476 }
3477
3478 /*
3479  *      addrconf module should be notified of a device going up
3480  */
3481 static struct notifier_block ipv6_dev_notf = {
3482         .notifier_call = addrconf_notify,
3483 };
3484
3485 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3486 {
3487         struct inet6_dev *idev;
3488         ASSERT_RTNL();
3489
3490         idev = __in6_dev_get(dev);
3491
3492         if (event == NETDEV_POST_TYPE_CHANGE)
3493                 ipv6_mc_remap(idev);
3494         else if (event == NETDEV_PRE_TYPE_CHANGE)
3495                 ipv6_mc_unmap(idev);
3496 }
3497
3498 static bool addr_is_local(const struct in6_addr *addr)
3499 {
3500         return ipv6_addr_type(addr) &
3501                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3502 }
3503
3504 static int addrconf_ifdown(struct net_device *dev, int how)
3505 {
3506         struct net *net = dev_net(dev);
3507         struct inet6_dev *idev;
3508         struct inet6_ifaddr *ifa, *tmp;
3509         struct list_head del_list;
3510         int _keep_addr;
3511         bool keep_addr;
3512         int state, i;
3513
3514         ASSERT_RTNL();
3515
3516         rt6_ifdown(net, dev);
3517         neigh_ifdown(&nd_tbl, dev);
3518
3519         idev = __in6_dev_get(dev);
3520         if (!idev)
3521                 return -ENODEV;
3522
3523         /*
3524          * Step 1: remove reference to ipv6 device from parent device.
3525          *         Do not dev_put!
3526          */
3527         if (how) {
3528                 idev->dead = 1;
3529
3530                 /* protected by rtnl_lock */
3531                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3532
3533                 /* Step 1.5: remove snmp6 entry */
3534                 snmp6_unregister_dev(idev);
3535
3536         }
3537
3538         /* aggregate the system setting and interface setting */
3539         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3540         if (!_keep_addr)
3541                 _keep_addr = idev->cnf.keep_addr_on_down;
3542
3543         /* combine the user config with event to determine if permanent
3544          * addresses are to be removed from address hash table
3545          */
3546         keep_addr = !(how || _keep_addr <= 0);
3547
3548         /* Step 2: clear hash table */
3549         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3550                 struct hlist_head *h = &inet6_addr_lst[i];
3551
3552                 spin_lock_bh(&addrconf_hash_lock);
3553 restart:
3554                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3555                         if (ifa->idev == idev) {
3556                                 addrconf_del_dad_work(ifa);
3557                                 /* combined flag + permanent flag decide if
3558                                  * address is retained on a down event
3559                                  */
3560                                 if (!keep_addr ||
3561                                     !(ifa->flags & IFA_F_PERMANENT) ||
3562                                     addr_is_local(&ifa->addr)) {
3563                                         hlist_del_init_rcu(&ifa->addr_lst);
3564                                         goto restart;
3565                                 }
3566                         }
3567                 }
3568                 spin_unlock_bh(&addrconf_hash_lock);
3569         }
3570
3571         write_lock_bh(&idev->lock);
3572
3573         addrconf_del_rs_timer(idev);
3574
3575         /* Step 2: clear flags for stateless addrconf */
3576         if (!how)
3577                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3578
3579         if (how && del_timer(&idev->regen_timer))
3580                 in6_dev_put(idev);
3581
3582         /* Step 3: clear tempaddr list */
3583         while (!list_empty(&idev->tempaddr_list)) {
3584                 ifa = list_first_entry(&idev->tempaddr_list,
3585                                        struct inet6_ifaddr, tmp_list);
3586                 list_del(&ifa->tmp_list);
3587                 write_unlock_bh(&idev->lock);
3588                 spin_lock_bh(&ifa->lock);
3589
3590                 if (ifa->ifpub) {
3591                         in6_ifa_put(ifa->ifpub);
3592                         ifa->ifpub = NULL;
3593                 }
3594                 spin_unlock_bh(&ifa->lock);
3595                 in6_ifa_put(ifa);
3596                 write_lock_bh(&idev->lock);
3597         }
3598
3599         /* re-combine the user config with event to determine if permanent
3600          * addresses are to be removed from the interface list
3601          */
3602         keep_addr = (!how && _keep_addr > 0);
3603
3604         INIT_LIST_HEAD(&del_list);
3605         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3606                 struct rt6_info *rt = NULL;
3607
3608                 addrconf_del_dad_work(ifa);
3609
3610                 write_unlock_bh(&idev->lock);
3611                 spin_lock_bh(&ifa->lock);
3612
3613                 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3614                     !addr_is_local(&ifa->addr)) {
3615                         /* set state to skip the notifier below */
3616                         state = INET6_IFADDR_STATE_DEAD;
3617                         ifa->state = 0;
3618                         if (!(ifa->flags & IFA_F_NODAD))
3619                                 ifa->flags |= IFA_F_TENTATIVE;
3620
3621                         rt = ifa->rt;
3622                         ifa->rt = NULL;
3623                 } else {
3624                         state = ifa->state;
3625                         ifa->state = INET6_IFADDR_STATE_DEAD;
3626
3627                         list_move(&ifa->if_list, &del_list);
3628                 }
3629
3630                 spin_unlock_bh(&ifa->lock);
3631
3632                 if (rt)
3633                         ip6_del_rt(rt);
3634
3635                 if (state != INET6_IFADDR_STATE_DEAD) {
3636                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3637                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3638                 } else {
3639                         if (idev->cnf.forwarding)
3640                                 addrconf_leave_anycast(ifa);
3641                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3642                 }
3643
3644                 write_lock_bh(&idev->lock);
3645         }
3646
3647         write_unlock_bh(&idev->lock);
3648
3649         /* now clean up addresses to be removed */
3650         while (!list_empty(&del_list)) {
3651                 ifa = list_first_entry(&del_list,
3652                                        struct inet6_ifaddr, if_list);
3653                 list_del(&ifa->if_list);
3654
3655                 in6_ifa_put(ifa);
3656         }
3657
3658         /* Step 5: Discard anycast and multicast list */
3659         if (how) {
3660                 ipv6_ac_destroy_dev(idev);
3661                 ipv6_mc_destroy_dev(idev);
3662         } else {
3663                 ipv6_mc_down(idev);
3664         }
3665
3666         idev->tstamp = jiffies;
3667
3668         /* Last: Shot the device (if unregistered) */
3669         if (how) {
3670                 addrconf_sysctl_unregister(idev);
3671                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3672                 neigh_ifdown(&nd_tbl, dev);
3673                 in6_dev_put(idev);
3674         }
3675         return 0;
3676 }
3677
3678 static void addrconf_rs_timer(unsigned long data)
3679 {
3680         struct inet6_dev *idev = (struct inet6_dev *)data;
3681         struct net_device *dev = idev->dev;
3682         struct in6_addr lladdr;
3683
3684         write_lock(&idev->lock);
3685         if (idev->dead || !(idev->if_flags & IF_READY))
3686                 goto out;
3687
3688         if (!ipv6_accept_ra(idev))
3689                 goto out;
3690
3691         /* Announcement received after solicitation was sent */
3692         if (idev->if_flags & IF_RA_RCVD)
3693                 goto out;
3694
3695         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3696                 write_unlock(&idev->lock);
3697                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3698                         ndisc_send_rs(dev, &lladdr,
3699                                       &in6addr_linklocal_allrouters);
3700                 else
3701                         goto put;
3702
3703                 write_lock(&idev->lock);
3704                 /* The wait after the last probe can be shorter */
3705                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3706                                              idev->cnf.rtr_solicits) ?
3707                                       idev->cnf.rtr_solicit_delay :
3708                                       idev->cnf.rtr_solicit_interval);
3709         } else {
3710                 /*
3711                  * Note: we do not support deprecated "all on-link"
3712                  * assumption any longer.
3713                  */
3714                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3715         }
3716
3717 out:
3718         write_unlock(&idev->lock);
3719 put:
3720         in6_dev_put(idev);
3721 }
3722
3723 /*
3724  *      Duplicate Address Detection
3725  */
3726 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3727 {
3728         unsigned long rand_num;
3729         struct inet6_dev *idev = ifp->idev;
3730
3731         if (ifp->flags & IFA_F_OPTIMISTIC)
3732                 rand_num = 0;
3733         else
3734                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3735
3736         ifp->dad_probes = idev->cnf.dad_transmits;
3737         addrconf_mod_dad_work(ifp, rand_num);
3738 }
3739
3740 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3741 {
3742         struct inet6_dev *idev = ifp->idev;
3743         struct net_device *dev = idev->dev;
3744         bool notify = false;
3745
3746         addrconf_join_solict(dev, &ifp->addr);
3747
3748         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3749
3750         read_lock_bh(&idev->lock);
3751         spin_lock(&ifp->lock);
3752         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3753                 goto out;
3754
3755         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3756             idev->cnf.accept_dad < 1 ||
3757             !(ifp->flags&IFA_F_TENTATIVE) ||
3758             ifp->flags & IFA_F_NODAD) {
3759                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3760                 spin_unlock(&ifp->lock);
3761                 read_unlock_bh(&idev->lock);
3762
3763                 addrconf_dad_completed(ifp);
3764                 return;
3765         }
3766
3767         if (!(idev->if_flags & IF_READY)) {
3768                 spin_unlock(&ifp->lock);
3769                 read_unlock_bh(&idev->lock);
3770                 /*
3771                  * If the device is not ready:
3772                  * - keep it tentative if it is a permanent address.
3773                  * - otherwise, kill it.
3774                  */
3775                 in6_ifa_hold(ifp);
3776                 addrconf_dad_stop(ifp, 0);
3777                 return;
3778         }
3779
3780         /*
3781          * Optimistic nodes can start receiving
3782          * Frames right away
3783          */
3784         if (ifp->flags & IFA_F_OPTIMISTIC) {
3785                 ip6_ins_rt(ifp->rt);
3786                 if (ipv6_use_optimistic_addr(idev)) {
3787                         /* Because optimistic nodes can use this address,
3788                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3789                          */
3790                         notify = true;
3791                 }
3792         }
3793
3794         addrconf_dad_kick(ifp);
3795 out:
3796         spin_unlock(&ifp->lock);
3797         read_unlock_bh(&idev->lock);
3798         if (notify)
3799                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3800 }
3801
3802 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3803 {
3804         bool begin_dad = false;
3805
3806         spin_lock_bh(&ifp->lock);
3807         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3808                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3809                 begin_dad = true;
3810         }
3811         spin_unlock_bh(&ifp->lock);
3812
3813         if (begin_dad)
3814                 addrconf_mod_dad_work(ifp, 0);
3815 }
3816
3817 static void addrconf_dad_work(struct work_struct *w)
3818 {
3819         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3820                                                 struct inet6_ifaddr,
3821                                                 dad_work);
3822         struct inet6_dev *idev = ifp->idev;
3823         struct in6_addr mcaddr;
3824
3825         enum {
3826                 DAD_PROCESS,
3827                 DAD_BEGIN,
3828                 DAD_ABORT,
3829         } action = DAD_PROCESS;
3830
3831         rtnl_lock();
3832
3833         spin_lock_bh(&ifp->lock);
3834         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3835                 action = DAD_BEGIN;
3836                 ifp->state = INET6_IFADDR_STATE_DAD;
3837         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3838                 action = DAD_ABORT;
3839                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3840         }
3841         spin_unlock_bh(&ifp->lock);
3842
3843         if (action == DAD_BEGIN) {
3844                 addrconf_dad_begin(ifp);
3845                 goto out;
3846         } else if (action == DAD_ABORT) {
3847                 addrconf_dad_stop(ifp, 1);
3848                 goto out;
3849         }
3850
3851         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3852                 goto out;
3853
3854         write_lock_bh(&idev->lock);
3855         if (idev->dead || !(idev->if_flags & IF_READY)) {
3856                 write_unlock_bh(&idev->lock);
3857                 goto out;
3858         }
3859
3860         spin_lock(&ifp->lock);
3861         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3862                 spin_unlock(&ifp->lock);
3863                 write_unlock_bh(&idev->lock);
3864                 goto out;
3865         }
3866
3867         if (ifp->dad_probes == 0) {
3868                 /*
3869                  * DAD was successful
3870                  */
3871
3872                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3873                 spin_unlock(&ifp->lock);
3874                 write_unlock_bh(&idev->lock);
3875
3876                 addrconf_dad_completed(ifp);
3877
3878                 goto out;
3879         }
3880
3881         ifp->dad_probes--;
3882         addrconf_mod_dad_work(ifp,
3883                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3884         spin_unlock(&ifp->lock);
3885         write_unlock_bh(&idev->lock);
3886
3887         /* send a neighbour solicitation for our addr */
3888         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3889         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3890 out:
3891         in6_ifa_put(ifp);
3892         rtnl_unlock();
3893 }
3894
3895 /* ifp->idev must be at least read locked */
3896 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3897 {
3898         struct inet6_ifaddr *ifpiter;
3899         struct inet6_dev *idev = ifp->idev;
3900
3901         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3902                 if (ifpiter->scope > IFA_LINK)
3903                         break;
3904                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3905                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3906                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3907                     IFA_F_PERMANENT)
3908                         return false;
3909         }
3910         return true;
3911 }
3912
3913 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3914 {
3915         struct net_device *dev = ifp->idev->dev;
3916         struct in6_addr lladdr;
3917         bool send_rs, send_mld;
3918
3919         addrconf_del_dad_work(ifp);
3920
3921         /*
3922          *      Configure the address for reception. Now it is valid.
3923          */
3924
3925         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3926
3927         /* If added prefix is link local and we are prepared to process
3928            router advertisements, start sending router solicitations.
3929          */
3930
3931         read_lock_bh(&ifp->idev->lock);
3932         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3933         send_rs = send_mld &&
3934                   ipv6_accept_ra(ifp->idev) &&
3935                   ifp->idev->cnf.rtr_solicits > 0 &&
3936                   (dev->flags&IFF_LOOPBACK) == 0;
3937         read_unlock_bh(&ifp->idev->lock);
3938
3939         /* While dad is in progress mld report's source address is in6_addrany.
3940          * Resend with proper ll now.
3941          */
3942         if (send_mld)
3943                 ipv6_mc_dad_complete(ifp->idev);
3944
3945         if (send_rs) {
3946                 /*
3947                  *      If a host as already performed a random delay
3948                  *      [...] as part of DAD [...] there is no need
3949                  *      to delay again before sending the first RS
3950                  */
3951                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3952                         return;
3953                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3954
3955                 write_lock_bh(&ifp->idev->lock);
3956                 spin_lock(&ifp->lock);
3957                 ifp->idev->rs_probes = 1;
3958                 ifp->idev->if_flags |= IF_RS_SENT;
3959                 addrconf_mod_rs_timer(ifp->idev,
3960                                       ifp->idev->cnf.rtr_solicit_interval);
3961                 spin_unlock(&ifp->lock);
3962                 write_unlock_bh(&ifp->idev->lock);
3963         }
3964 }
3965
3966 static void addrconf_dad_run(struct inet6_dev *idev)
3967 {
3968         struct inet6_ifaddr *ifp;
3969
3970         read_lock_bh(&idev->lock);
3971         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3972                 spin_lock(&ifp->lock);
3973                 if (ifp->flags & IFA_F_TENTATIVE &&
3974                     ifp->state == INET6_IFADDR_STATE_DAD)
3975                         addrconf_dad_kick(ifp);
3976                 spin_unlock(&ifp->lock);
3977         }
3978         read_unlock_bh(&idev->lock);
3979 }
3980
3981 #ifdef CONFIG_PROC_FS
3982 struct if6_iter_state {
3983         struct seq_net_private p;
3984         int bucket;
3985         int offset;
3986 };
3987
3988 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3989 {
3990         struct inet6_ifaddr *ifa = NULL;
3991         struct if6_iter_state *state = seq->private;
3992         struct net *net = seq_file_net(seq);
3993         int p = 0;
3994
3995         /* initial bucket if pos is 0 */
3996         if (pos == 0) {
3997                 state->bucket = 0;
3998                 state->offset = 0;
3999         }
4000
4001         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4002                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4003                                          addr_lst) {
4004                         if (!net_eq(dev_net(ifa->idev->dev), net))
4005                                 continue;
4006                         /* sync with offset */
4007                         if (p < state->offset) {
4008                                 p++;
4009                                 continue;
4010                         }
4011                         state->offset++;
4012                         return ifa;
4013                 }
4014
4015                 /* prepare for next bucket */
4016                 state->offset = 0;
4017                 p = 0;
4018         }
4019         return NULL;
4020 }
4021
4022 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4023                                          struct inet6_ifaddr *ifa)
4024 {
4025         struct if6_iter_state *state = seq->private;
4026         struct net *net = seq_file_net(seq);
4027
4028         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4029                 if (!net_eq(dev_net(ifa->idev->dev), net))
4030                         continue;
4031                 state->offset++;
4032                 return ifa;
4033         }
4034
4035         while (++state->bucket < IN6_ADDR_HSIZE) {
4036                 state->offset = 0;
4037                 hlist_for_each_entry_rcu_bh(ifa,
4038                                      &inet6_addr_lst[state->bucket], addr_lst) {
4039                         if (!net_eq(dev_net(ifa->idev->dev), net))
4040                                 continue;
4041                         state->offset++;
4042                         return ifa;
4043                 }
4044         }
4045
4046         return NULL;
4047 }
4048
4049 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4050         __acquires(rcu_bh)
4051 {
4052         rcu_read_lock_bh();
4053         return if6_get_first(seq, *pos);
4054 }
4055
4056 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4057 {
4058         struct inet6_ifaddr *ifa;
4059
4060         ifa = if6_get_next(seq, v);
4061         ++*pos;
4062         return ifa;
4063 }
4064
4065 static void if6_seq_stop(struct seq_file *seq, void *v)
4066         __releases(rcu_bh)
4067 {
4068         rcu_read_unlock_bh();
4069 }
4070
4071 static int if6_seq_show(struct seq_file *seq, void *v)
4072 {
4073         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4074         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4075                    &ifp->addr,
4076                    ifp->idev->dev->ifindex,
4077                    ifp->prefix_len,
4078                    ifp->scope,
4079                    (u8) ifp->flags,
4080                    ifp->idev->dev->name);
4081         return 0;
4082 }
4083
4084 static const struct seq_operations if6_seq_ops = {
4085         .start  = if6_seq_start,
4086         .next   = if6_seq_next,
4087         .show   = if6_seq_show,
4088         .stop   = if6_seq_stop,
4089 };
4090
4091 static int if6_seq_open(struct inode *inode, struct file *file)
4092 {
4093         return seq_open_net(inode, file, &if6_seq_ops,
4094                             sizeof(struct if6_iter_state));
4095 }
4096
4097 static const struct file_operations if6_fops = {
4098         .owner          = THIS_MODULE,
4099         .open           = if6_seq_open,
4100         .read           = seq_read,
4101         .llseek         = seq_lseek,
4102         .release        = seq_release_net,
4103 };
4104
4105 static int __net_init if6_proc_net_init(struct net *net)
4106 {
4107         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4108                 return -ENOMEM;
4109         return 0;
4110 }
4111
4112 static void __net_exit if6_proc_net_exit(struct net *net)
4113 {
4114         remove_proc_entry("if_inet6", net->proc_net);
4115 }
4116
4117 static struct pernet_operations if6_proc_net_ops = {
4118         .init = if6_proc_net_init,
4119         .exit = if6_proc_net_exit,
4120 };
4121
4122 int __init if6_proc_init(void)
4123 {
4124         return register_pernet_subsys(&if6_proc_net_ops);
4125 }
4126
4127 void if6_proc_exit(void)
4128 {
4129         unregister_pernet_subsys(&if6_proc_net_ops);
4130 }
4131 #endif  /* CONFIG_PROC_FS */
4132
4133 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4134 /* Check if address is a home address configured on any interface. */
4135 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4136 {
4137         int ret = 0;
4138         struct inet6_ifaddr *ifp = NULL;
4139         unsigned int hash = inet6_addr_hash(addr);
4140
4141         rcu_read_lock_bh();
4142         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4143                 if (!net_eq(dev_net(ifp->idev->dev), net))
4144                         continue;
4145                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4146                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4147                         ret = 1;
4148                         break;
4149                 }
4150         }
4151         rcu_read_unlock_bh();
4152         return ret;
4153 }
4154 #endif
4155
4156 /*
4157  *      Periodic address status verification
4158  */
4159
4160 static void addrconf_verify_rtnl(void)
4161 {
4162         unsigned long now, next, next_sec, next_sched;
4163         struct inet6_ifaddr *ifp;
4164         int i;
4165
4166         ASSERT_RTNL();
4167
4168         rcu_read_lock_bh();
4169         now = jiffies;
4170         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4171
4172         cancel_delayed_work(&addr_chk_work);
4173
4174         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4175 restart:
4176                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4177                         unsigned long age;
4178
4179                         /* When setting preferred_lft to a value not zero or
4180                          * infinity, while valid_lft is infinity
4181                          * IFA_F_PERMANENT has a non-infinity life time.
4182                          */
4183                         if ((ifp->flags & IFA_F_PERMANENT) &&
4184                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4185                                 continue;
4186
4187                         spin_lock(&ifp->lock);
4188                         /* We try to batch several events at once. */
4189                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4190
4191                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4192                             age >= ifp->valid_lft) {
4193                                 spin_unlock(&ifp->lock);
4194                                 in6_ifa_hold(ifp);
4195                                 ipv6_del_addr(ifp);
4196                                 goto restart;
4197                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4198                                 spin_unlock(&ifp->lock);
4199                                 continue;
4200                         } else if (age >= ifp->prefered_lft) {
4201                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4202                                 int deprecate = 0;
4203
4204                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4205                                         deprecate = 1;
4206                                         ifp->flags |= IFA_F_DEPRECATED;
4207                                 }
4208
4209                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4210                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4211                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4212
4213                                 spin_unlock(&ifp->lock);
4214
4215                                 if (deprecate) {
4216                                         in6_ifa_hold(ifp);
4217
4218                                         ipv6_ifa_notify(0, ifp);
4219                                         in6_ifa_put(ifp);
4220                                         goto restart;
4221                                 }
4222                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4223                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4224                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4225                                         ifp->idev->cnf.dad_transmits *
4226                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4227
4228                                 if (age >= ifp->prefered_lft - regen_advance) {
4229                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4230                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4231                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4232                                         if (!ifp->regen_count && ifpub) {
4233                                                 ifp->regen_count++;
4234                                                 in6_ifa_hold(ifp);
4235                                                 in6_ifa_hold(ifpub);
4236                                                 spin_unlock(&ifp->lock);
4237
4238                                                 spin_lock(&ifpub->lock);
4239                                                 ifpub->regen_count = 0;
4240                                                 spin_unlock(&ifpub->lock);
4241                                                 ipv6_create_tempaddr(ifpub, ifp);
4242                                                 in6_ifa_put(ifpub);
4243                                                 in6_ifa_put(ifp);
4244                                                 goto restart;
4245                                         }
4246                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4247                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4248                                 spin_unlock(&ifp->lock);
4249                         } else {
4250                                 /* ifp->prefered_lft <= ifp->valid_lft */
4251                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4252                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4253                                 spin_unlock(&ifp->lock);
4254                         }
4255                 }
4256         }
4257
4258         next_sec = round_jiffies_up(next);
4259         next_sched = next;
4260
4261         /* If rounded timeout is accurate enough, accept it. */
4262         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4263                 next_sched = next_sec;
4264
4265         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4266         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4267                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4268
4269         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4270               now, next, next_sec, next_sched);
4271         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4272         rcu_read_unlock_bh();
4273 }
4274
4275 static void addrconf_verify_work(struct work_struct *w)
4276 {
4277         rtnl_lock();
4278         addrconf_verify_rtnl();
4279         rtnl_unlock();
4280 }
4281
4282 static void addrconf_verify(void)
4283 {
4284         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4285 }
4286
4287 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4288                                      struct in6_addr **peer_pfx)
4289 {
4290         struct in6_addr *pfx = NULL;
4291
4292         *peer_pfx = NULL;
4293
4294         if (addr)
4295                 pfx = nla_data(addr);
4296
4297         if (local) {
4298                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4299                         *peer_pfx = pfx;
4300                 pfx = nla_data(local);
4301         }
4302
4303         return pfx;
4304 }
4305
4306 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4307         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4308         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4309         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4310         [IFA_FLAGS]             = { .len = sizeof(u32) },
4311 };
4312
4313 static int
4314 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4315 {
4316         struct net *net = sock_net(skb->sk);
4317         struct ifaddrmsg *ifm;
4318         struct nlattr *tb[IFA_MAX+1];
4319         struct in6_addr *pfx, *peer_pfx;
4320         u32 ifa_flags;
4321         int err;
4322
4323         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4324         if (err < 0)
4325                 return err;
4326
4327         ifm = nlmsg_data(nlh);
4328         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4329         if (!pfx)
4330                 return -EINVAL;
4331
4332         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4333
4334         /* We ignore other flags so far. */
4335         ifa_flags &= IFA_F_MANAGETEMPADDR;
4336
4337         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4338                               ifm->ifa_prefixlen);
4339 }
4340
4341 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4342                              u32 prefered_lft, u32 valid_lft)
4343 {
4344         u32 flags;
4345         clock_t expires;
4346         unsigned long timeout;
4347         bool was_managetempaddr;
4348         bool had_prefixroute;
4349
4350         ASSERT_RTNL();
4351
4352         if (!valid_lft || (prefered_lft > valid_lft))
4353                 return -EINVAL;
4354
4355         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4356             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4357                 return -EINVAL;
4358
4359         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4360         if (addrconf_finite_timeout(timeout)) {
4361                 expires = jiffies_to_clock_t(timeout * HZ);
4362                 valid_lft = timeout;
4363                 flags = RTF_EXPIRES;
4364         } else {
4365                 expires = 0;
4366                 flags = 0;
4367                 ifa_flags |= IFA_F_PERMANENT;
4368         }
4369
4370         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4371         if (addrconf_finite_timeout(timeout)) {
4372                 if (timeout == 0)
4373                         ifa_flags |= IFA_F_DEPRECATED;
4374                 prefered_lft = timeout;
4375         }
4376
4377         spin_lock_bh(&ifp->lock);
4378         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4379         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4380                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4381         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4382                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4383                         IFA_F_NOPREFIXROUTE);
4384         ifp->flags |= ifa_flags;
4385         ifp->tstamp = jiffies;
4386         ifp->valid_lft = valid_lft;
4387         ifp->prefered_lft = prefered_lft;
4388
4389         spin_unlock_bh(&ifp->lock);
4390         if (!(ifp->flags&IFA_F_TENTATIVE))
4391                 ipv6_ifa_notify(0, ifp);
4392
4393         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4394                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4395                                       expires, flags);
4396         } else if (had_prefixroute) {
4397                 enum cleanup_prefix_rt_t action;
4398                 unsigned long rt_expires;
4399
4400                 write_lock_bh(&ifp->idev->lock);
4401                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4402                 write_unlock_bh(&ifp->idev->lock);
4403
4404                 if (action != CLEANUP_PREFIX_RT_NOP) {
4405                         cleanup_prefix_route(ifp, rt_expires,
4406                                 action == CLEANUP_PREFIX_RT_DEL);
4407                 }
4408         }
4409
4410         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4411                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4412                         valid_lft = prefered_lft = 0;
4413                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4414                                  !was_managetempaddr, jiffies);
4415         }
4416
4417         addrconf_verify_rtnl();
4418
4419         return 0;
4420 }
4421
4422 static int
4423 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4424 {
4425         struct net *net = sock_net(skb->sk);
4426         struct ifaddrmsg *ifm;
4427         struct nlattr *tb[IFA_MAX+1];
4428         struct in6_addr *pfx, *peer_pfx;
4429         struct inet6_ifaddr *ifa;
4430         struct net_device *dev;
4431         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4432         u32 ifa_flags;
4433         int err;
4434
4435         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4436         if (err < 0)
4437                 return err;
4438
4439         ifm = nlmsg_data(nlh);
4440         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4441         if (!pfx)
4442                 return -EINVAL;
4443
4444         if (tb[IFA_CACHEINFO]) {
4445                 struct ifa_cacheinfo *ci;
4446
4447                 ci = nla_data(tb[IFA_CACHEINFO]);
4448                 valid_lft = ci->ifa_valid;
4449                 preferred_lft = ci->ifa_prefered;
4450         } else {
4451                 preferred_lft = INFINITY_LIFE_TIME;
4452                 valid_lft = INFINITY_LIFE_TIME;
4453         }
4454
4455         dev =  __dev_get_by_index(net, ifm->ifa_index);
4456         if (!dev)
4457                 return -ENODEV;
4458
4459         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4460
4461         /* We ignore other flags so far. */
4462         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4463                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4464
4465         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4466         if (!ifa) {
4467                 /*
4468                  * It would be best to check for !NLM_F_CREATE here but
4469                  * userspace already relies on not having to provide this.
4470                  */
4471                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4472                                       ifm->ifa_prefixlen, ifa_flags,
4473                                       preferred_lft, valid_lft);
4474         }
4475
4476         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4477             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4478                 err = -EEXIST;
4479         else
4480                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4481
4482         in6_ifa_put(ifa);
4483
4484         return err;
4485 }
4486
4487 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4488                           u8 scope, int ifindex)
4489 {
4490         struct ifaddrmsg *ifm;
4491
4492         ifm = nlmsg_data(nlh);
4493         ifm->ifa_family = AF_INET6;
4494         ifm->ifa_prefixlen = prefixlen;
4495         ifm->ifa_flags = flags;
4496         ifm->ifa_scope = scope;
4497         ifm->ifa_index = ifindex;
4498 }
4499
4500 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4501                          unsigned long tstamp, u32 preferred, u32 valid)
4502 {
4503         struct ifa_cacheinfo ci;
4504
4505         ci.cstamp = cstamp_delta(cstamp);
4506         ci.tstamp = cstamp_delta(tstamp);
4507         ci.ifa_prefered = preferred;
4508         ci.ifa_valid = valid;
4509
4510         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4511 }
4512
4513 static inline int rt_scope(int ifa_scope)
4514 {
4515         if (ifa_scope & IFA_HOST)
4516                 return RT_SCOPE_HOST;
4517         else if (ifa_scope & IFA_LINK)
4518                 return RT_SCOPE_LINK;
4519         else if (ifa_scope & IFA_SITE)
4520                 return RT_SCOPE_SITE;
4521         else
4522                 return RT_SCOPE_UNIVERSE;
4523 }
4524
4525 static inline int inet6_ifaddr_msgsize(void)
4526 {
4527         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4528                + nla_total_size(16) /* IFA_LOCAL */
4529                + nla_total_size(16) /* IFA_ADDRESS */
4530                + nla_total_size(sizeof(struct ifa_cacheinfo))
4531                + nla_total_size(4)  /* IFA_FLAGS */;
4532 }
4533
4534 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4535                              u32 portid, u32 seq, int event, unsigned int flags)
4536 {
4537         struct nlmsghdr  *nlh;
4538         u32 preferred, valid;
4539
4540         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4541         if (!nlh)
4542                 return -EMSGSIZE;
4543
4544         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4545                       ifa->idev->dev->ifindex);
4546
4547         if (!((ifa->flags&IFA_F_PERMANENT) &&
4548               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4549                 preferred = ifa->prefered_lft;
4550                 valid = ifa->valid_lft;
4551                 if (preferred != INFINITY_LIFE_TIME) {
4552                         long tval = (jiffies - ifa->tstamp)/HZ;
4553                         if (preferred > tval)
4554                                 preferred -= tval;
4555                         else
4556                                 preferred = 0;
4557                         if (valid != INFINITY_LIFE_TIME) {
4558                                 if (valid > tval)
4559                                         valid -= tval;
4560                                 else
4561                                         valid = 0;
4562                         }
4563                 }
4564         } else {
4565                 preferred = INFINITY_LIFE_TIME;
4566                 valid = INFINITY_LIFE_TIME;
4567         }
4568
4569         if (!ipv6_addr_any(&ifa->peer_addr)) {
4570                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4571                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4572                         goto error;
4573         } else
4574                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4575                         goto error;
4576
4577         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4578                 goto error;
4579
4580         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4581                 goto error;
4582
4583         nlmsg_end(skb, nlh);
4584         return 0;
4585
4586 error:
4587         nlmsg_cancel(skb, nlh);
4588         return -EMSGSIZE;
4589 }
4590
4591 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4592                                 u32 portid, u32 seq, int event, u16 flags)
4593 {
4594         struct nlmsghdr  *nlh;
4595         u8 scope = RT_SCOPE_UNIVERSE;
4596         int ifindex = ifmca->idev->dev->ifindex;
4597
4598         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4599                 scope = RT_SCOPE_SITE;
4600
4601         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4602         if (!nlh)
4603                 return -EMSGSIZE;
4604
4605         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4606         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4607             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4608                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4609                 nlmsg_cancel(skb, nlh);
4610                 return -EMSGSIZE;
4611         }
4612
4613         nlmsg_end(skb, nlh);
4614         return 0;
4615 }
4616
4617 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4618                                 u32 portid, u32 seq, int event, unsigned int flags)
4619 {
4620         struct nlmsghdr  *nlh;
4621         u8 scope = RT_SCOPE_UNIVERSE;
4622         int ifindex = ifaca->aca_idev->dev->ifindex;
4623
4624         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4625                 scope = RT_SCOPE_SITE;
4626
4627         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4628         if (!nlh)
4629                 return -EMSGSIZE;
4630
4631         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4632         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4633             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4634                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4635                 nlmsg_cancel(skb, nlh);
4636                 return -EMSGSIZE;
4637         }
4638
4639         nlmsg_end(skb, nlh);
4640         return 0;
4641 }
4642
4643 enum addr_type_t {
4644         UNICAST_ADDR,
4645         MULTICAST_ADDR,
4646         ANYCAST_ADDR,
4647 };
4648
4649 /* called with rcu_read_lock() */
4650 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4651                           struct netlink_callback *cb, enum addr_type_t type,
4652                           int s_ip_idx, int *p_ip_idx)
4653 {
4654         struct ifmcaddr6 *ifmca;
4655         struct ifacaddr6 *ifaca;
4656         int err = 1;
4657         int ip_idx = *p_ip_idx;
4658
4659         read_lock_bh(&idev->lock);
4660         switch (type) {
4661         case UNICAST_ADDR: {
4662                 struct inet6_ifaddr *ifa;
4663
4664                 /* unicast address incl. temp addr */
4665                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4666                         if (++ip_idx < s_ip_idx)
4667                                 continue;
4668                         err = inet6_fill_ifaddr(skb, ifa,
4669                                                 NETLINK_CB(cb->skb).portid,
4670                                                 cb->nlh->nlmsg_seq,
4671                                                 RTM_NEWADDR,
4672                                                 NLM_F_MULTI);
4673                         if (err < 0)
4674                                 break;
4675                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4676                 }
4677                 break;
4678         }
4679         case MULTICAST_ADDR:
4680                 /* multicast address */
4681                 for (ifmca = idev->mc_list; ifmca;
4682                      ifmca = ifmca->next, ip_idx++) {
4683                         if (ip_idx < s_ip_idx)
4684                                 continue;
4685                         err = inet6_fill_ifmcaddr(skb, ifmca,
4686                                                   NETLINK_CB(cb->skb).portid,
4687                                                   cb->nlh->nlmsg_seq,
4688                                                   RTM_GETMULTICAST,
4689                                                   NLM_F_MULTI);
4690                         if (err < 0)
4691                                 break;
4692                 }
4693                 break;
4694         case ANYCAST_ADDR:
4695                 /* anycast address */
4696                 for (ifaca = idev->ac_list; ifaca;
4697                      ifaca = ifaca->aca_next, ip_idx++) {
4698                         if (ip_idx < s_ip_idx)
4699                                 continue;
4700                         err = inet6_fill_ifacaddr(skb, ifaca,
4701                                                   NETLINK_CB(cb->skb).portid,
4702                                                   cb->nlh->nlmsg_seq,
4703                                                   RTM_GETANYCAST,
4704                                                   NLM_F_MULTI);
4705                         if (err < 0)
4706                                 break;
4707                 }
4708                 break;
4709         default:
4710                 break;
4711         }
4712         read_unlock_bh(&idev->lock);
4713         *p_ip_idx = ip_idx;
4714         return err;
4715 }
4716
4717 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4718                            enum addr_type_t type)
4719 {
4720         struct net *net = sock_net(skb->sk);
4721         int h, s_h;
4722         int idx, ip_idx;
4723         int s_idx, s_ip_idx;
4724         struct net_device *dev;
4725         struct inet6_dev *idev;
4726         struct hlist_head *head;
4727
4728         s_h = cb->args[0];
4729         s_idx = idx = cb->args[1];
4730         s_ip_idx = ip_idx = cb->args[2];
4731
4732         rcu_read_lock();
4733         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4734         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4735                 idx = 0;
4736                 head = &net->dev_index_head[h];
4737                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4738                         if (idx < s_idx)
4739                                 goto cont;
4740                         if (h > s_h || idx > s_idx)
4741                                 s_ip_idx = 0;
4742                         ip_idx = 0;
4743                         idev = __in6_dev_get(dev);
4744                         if (!idev)
4745                                 goto cont;
4746
4747                         if (in6_dump_addrs(idev, skb, cb, type,
4748                                            s_ip_idx, &ip_idx) < 0)
4749                                 goto done;
4750 cont:
4751                         idx++;
4752                 }
4753         }
4754 done:
4755         rcu_read_unlock();
4756         cb->args[0] = h;
4757         cb->args[1] = idx;
4758         cb->args[2] = ip_idx;
4759
4760         return skb->len;
4761 }
4762
4763 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4764 {
4765         enum addr_type_t type = UNICAST_ADDR;
4766
4767         return inet6_dump_addr(skb, cb, type);
4768 }
4769
4770 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4771 {
4772         enum addr_type_t type = MULTICAST_ADDR;
4773
4774         return inet6_dump_addr(skb, cb, type);
4775 }
4776
4777
4778 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4779 {
4780         enum addr_type_t type = ANYCAST_ADDR;
4781
4782         return inet6_dump_addr(skb, cb, type);
4783 }
4784
4785 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4786 {
4787         struct net *net = sock_net(in_skb->sk);
4788         struct ifaddrmsg *ifm;
4789         struct nlattr *tb[IFA_MAX+1];
4790         struct in6_addr *addr = NULL, *peer;
4791         struct net_device *dev = NULL;
4792         struct inet6_ifaddr *ifa;
4793         struct sk_buff *skb;
4794         int err;
4795
4796         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4797         if (err < 0)
4798                 goto errout;
4799
4800         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4801         if (!addr) {
4802                 err = -EINVAL;
4803                 goto errout;
4804         }
4805
4806         ifm = nlmsg_data(nlh);
4807         if (ifm->ifa_index)
4808                 dev = __dev_get_by_index(net, ifm->ifa_index);
4809
4810         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4811         if (!ifa) {
4812                 err = -EADDRNOTAVAIL;
4813                 goto errout;
4814         }
4815
4816         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4817         if (!skb) {
4818                 err = -ENOBUFS;
4819                 goto errout_ifa;
4820         }
4821
4822         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4823                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4824         if (err < 0) {
4825                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4826                 WARN_ON(err == -EMSGSIZE);
4827                 kfree_skb(skb);
4828                 goto errout_ifa;
4829         }
4830         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4831 errout_ifa:
4832         in6_ifa_put(ifa);
4833 errout:
4834         return err;
4835 }
4836
4837 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4838 {
4839         struct sk_buff *skb;
4840         struct net *net = dev_net(ifa->idev->dev);
4841         int err = -ENOBUFS;
4842
4843         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4844         if (!skb)
4845                 goto errout;
4846
4847         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4848         if (err < 0) {
4849                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4850                 WARN_ON(err == -EMSGSIZE);
4851                 kfree_skb(skb);
4852                 goto errout;
4853         }
4854         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4855         return;
4856 errout:
4857         if (err < 0)
4858                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4859 }
4860
4861 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4862                                 __s32 *array, int bytes)
4863 {
4864         BUG_ON(bytes < (DEVCONF_MAX * 4));
4865
4866         memset(array, 0, bytes);
4867         array[DEVCONF_FORWARDING] = cnf->forwarding;
4868         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4869         array[DEVCONF_MTU6] = cnf->mtu6;
4870         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4871         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4872         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4873         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4874         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4875         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4876                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4877         array[DEVCONF_RTR_SOLICIT_DELAY] =
4878                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4879         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4880         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4881                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4882         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4883                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4884         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4885         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4886         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4887         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4888         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4889         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4890         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4891         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4892         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4893 #ifdef CONFIG_IPV6_ROUTER_PREF
4894         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4895         array[DEVCONF_RTR_PROBE_INTERVAL] =
4896                 jiffies_to_msecs(cnf->rtr_probe_interval);
4897 #ifdef CONFIG_IPV6_ROUTE_INFO
4898         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4899 #endif
4900 #endif
4901         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4902         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4903 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4904         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4905         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4906 #endif
4907 #ifdef CONFIG_IPV6_MROUTE
4908         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4909 #endif
4910         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4911         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4912         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4913         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4914         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4915         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4916         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4917         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4918         /* we omit DEVCONF_STABLE_SECRET for now */
4919         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4920         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4921         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4922         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4923 }
4924
4925 static inline size_t inet6_ifla6_size(void)
4926 {
4927         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4928              + nla_total_size(sizeof(struct ifla_cacheinfo))
4929              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4930              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4931              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4932              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4933 }
4934
4935 static inline size_t inet6_if_nlmsg_size(void)
4936 {
4937         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4938                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4939                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4940                + nla_total_size(4) /* IFLA_MTU */
4941                + nla_total_size(4) /* IFLA_LINK */
4942                + nla_total_size(1) /* IFLA_OPERSTATE */
4943                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4944 }
4945
4946 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4947                                       int items, int bytes)
4948 {
4949         int i;
4950         int pad = bytes - sizeof(u64) * items;
4951         BUG_ON(pad < 0);
4952
4953         /* Use put_unaligned() because stats may not be aligned for u64. */
4954         put_unaligned(items, &stats[0]);
4955         for (i = 1; i < items; i++)
4956                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4957
4958         memset(&stats[items], 0, pad);
4959 }
4960
4961 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4962                                         int bytes, size_t syncpoff)
4963 {
4964         int i, c;
4965         u64 buff[IPSTATS_MIB_MAX];
4966         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4967
4968         BUG_ON(pad < 0);
4969
4970         memset(buff, 0, sizeof(buff));
4971         buff[0] = IPSTATS_MIB_MAX;
4972
4973         for_each_possible_cpu(c) {
4974                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4975                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4976         }
4977
4978         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4979         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4980 }
4981
4982 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4983                              int bytes)
4984 {
4985         switch (attrtype) {
4986         case IFLA_INET6_STATS:
4987                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4988                                      offsetof(struct ipstats_mib, syncp));
4989                 break;
4990         case IFLA_INET6_ICMP6STATS:
4991                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4992                 break;
4993         }
4994 }
4995
4996 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4997                                   u32 ext_filter_mask)
4998 {
4999         struct nlattr *nla;
5000         struct ifla_cacheinfo ci;
5001
5002         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5003                 goto nla_put_failure;
5004         ci.max_reasm_len = IPV6_MAXPLEN;
5005         ci.tstamp = cstamp_delta(idev->tstamp);
5006         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5007         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5008         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5009                 goto nla_put_failure;
5010         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5011         if (!nla)
5012                 goto nla_put_failure;
5013         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5014
5015         /* XXX - MC not implemented */
5016
5017         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5018                 return 0;
5019
5020         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5021         if (!nla)
5022                 goto nla_put_failure;
5023         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5024
5025         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5026         if (!nla)
5027                 goto nla_put_failure;
5028         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5029
5030         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5031         if (!nla)
5032                 goto nla_put_failure;
5033
5034         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5035                 goto nla_put_failure;
5036
5037         read_lock_bh(&idev->lock);
5038         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5039         read_unlock_bh(&idev->lock);
5040
5041         return 0;
5042
5043 nla_put_failure:
5044         return -EMSGSIZE;
5045 }
5046
5047 static size_t inet6_get_link_af_size(const struct net_device *dev,
5048                                      u32 ext_filter_mask)
5049 {
5050         if (!__in6_dev_get(dev))
5051                 return 0;
5052
5053         return inet6_ifla6_size();
5054 }
5055
5056 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5057                               u32 ext_filter_mask)
5058 {
5059         struct inet6_dev *idev = __in6_dev_get(dev);
5060
5061         if (!idev)
5062                 return -ENODATA;
5063
5064         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5065                 return -EMSGSIZE;
5066
5067         return 0;
5068 }
5069
5070 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5071 {
5072         struct inet6_ifaddr *ifp;
5073         struct net_device *dev = idev->dev;
5074         bool clear_token, update_rs = false;
5075         struct in6_addr ll_addr;
5076
5077         ASSERT_RTNL();
5078
5079         if (!token)
5080                 return -EINVAL;
5081         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5082                 return -EINVAL;
5083         if (!ipv6_accept_ra(idev))
5084                 return -EINVAL;
5085         if (idev->cnf.rtr_solicits <= 0)
5086                 return -EINVAL;
5087
5088         write_lock_bh(&idev->lock);
5089
5090         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5091         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5092
5093         write_unlock_bh(&idev->lock);
5094
5095         clear_token = ipv6_addr_any(token);
5096         if (clear_token)
5097                 goto update_lft;
5098
5099         if (!idev->dead && (idev->if_flags & IF_READY) &&
5100             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5101                              IFA_F_OPTIMISTIC)) {
5102                 /* If we're not ready, then normal ifup will take care
5103                  * of this. Otherwise, we need to request our rs here.
5104                  */
5105                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5106                 update_rs = true;
5107         }
5108
5109 update_lft:
5110         write_lock_bh(&idev->lock);
5111
5112         if (update_rs) {
5113                 idev->if_flags |= IF_RS_SENT;
5114                 idev->rs_probes = 1;
5115                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
5116         }
5117
5118         /* Well, that's kinda nasty ... */
5119         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5120                 spin_lock(&ifp->lock);
5121                 if (ifp->tokenized) {
5122                         ifp->valid_lft = 0;
5123                         ifp->prefered_lft = 0;
5124                 }
5125                 spin_unlock(&ifp->lock);
5126         }
5127
5128         write_unlock_bh(&idev->lock);
5129         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5130         addrconf_verify_rtnl();
5131         return 0;
5132 }
5133
5134 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5135         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5136         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5137 };
5138
5139 static int inet6_validate_link_af(const struct net_device *dev,
5140                                   const struct nlattr *nla)
5141 {
5142         struct nlattr *tb[IFLA_INET6_MAX + 1];
5143
5144         if (dev && !__in6_dev_get(dev))
5145                 return -EAFNOSUPPORT;
5146
5147         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5148 }
5149
5150 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5151 {
5152         int err = -EINVAL;
5153         struct inet6_dev *idev = __in6_dev_get(dev);
5154         struct nlattr *tb[IFLA_INET6_MAX + 1];
5155
5156         if (!idev)
5157                 return -EAFNOSUPPORT;
5158
5159         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5160                 BUG();
5161
5162         if (tb[IFLA_INET6_TOKEN]) {
5163                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5164                 if (err)
5165                         return err;
5166         }
5167
5168         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5169                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5170
5171                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5172                     mode != IN6_ADDR_GEN_MODE_NONE &&
5173                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5174                     mode != IN6_ADDR_GEN_MODE_RANDOM)
5175                         return -EINVAL;
5176
5177                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5178                     !idev->cnf.stable_secret.initialized &&
5179                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5180                         return -EINVAL;
5181
5182                 idev->addr_gen_mode = mode;
5183                 err = 0;
5184         }
5185
5186         return err;
5187 }
5188
5189 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5190                              u32 portid, u32 seq, int event, unsigned int flags)
5191 {
5192         struct net_device *dev = idev->dev;
5193         struct ifinfomsg *hdr;
5194         struct nlmsghdr *nlh;
5195         void *protoinfo;
5196
5197         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5198         if (!nlh)
5199                 return -EMSGSIZE;
5200
5201         hdr = nlmsg_data(nlh);
5202         hdr->ifi_family = AF_INET6;
5203         hdr->__ifi_pad = 0;
5204         hdr->ifi_type = dev->type;
5205         hdr->ifi_index = dev->ifindex;
5206         hdr->ifi_flags = dev_get_flags(dev);
5207         hdr->ifi_change = 0;
5208
5209         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5210             (dev->addr_len &&
5211              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5212             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5213             (dev->ifindex != dev_get_iflink(dev) &&
5214              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5215             nla_put_u8(skb, IFLA_OPERSTATE,
5216                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5217                 goto nla_put_failure;
5218         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5219         if (!protoinfo)
5220                 goto nla_put_failure;
5221
5222         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5223                 goto nla_put_failure;
5224
5225         nla_nest_end(skb, protoinfo);
5226         nlmsg_end(skb, nlh);
5227         return 0;
5228
5229 nla_put_failure:
5230         nlmsg_cancel(skb, nlh);
5231         return -EMSGSIZE;
5232 }
5233
5234 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5235 {
5236         struct net *net = sock_net(skb->sk);
5237         int h, s_h;
5238         int idx = 0, s_idx;
5239         struct net_device *dev;
5240         struct inet6_dev *idev;
5241         struct hlist_head *head;
5242
5243         s_h = cb->args[0];
5244         s_idx = cb->args[1];
5245
5246         rcu_read_lock();
5247         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5248                 idx = 0;
5249                 head = &net->dev_index_head[h];
5250                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5251                         if (idx < s_idx)
5252                                 goto cont;
5253                         idev = __in6_dev_get(dev);
5254                         if (!idev)
5255                                 goto cont;
5256                         if (inet6_fill_ifinfo(skb, idev,
5257                                               NETLINK_CB(cb->skb).portid,
5258                                               cb->nlh->nlmsg_seq,
5259                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5260                                 goto out;
5261 cont:
5262                         idx++;
5263                 }
5264         }
5265 out:
5266         rcu_read_unlock();
5267         cb->args[1] = idx;
5268         cb->args[0] = h;
5269
5270         return skb->len;
5271 }
5272
5273 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5274 {
5275         struct sk_buff *skb;
5276         struct net *net = dev_net(idev->dev);
5277         int err = -ENOBUFS;
5278
5279         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5280         if (!skb)
5281                 goto errout;
5282
5283         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5284         if (err < 0) {
5285                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5286                 WARN_ON(err == -EMSGSIZE);
5287                 kfree_skb(skb);
5288                 goto errout;
5289         }
5290         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5291         return;
5292 errout:
5293         if (err < 0)
5294                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5295 }
5296
5297 static inline size_t inet6_prefix_nlmsg_size(void)
5298 {
5299         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5300                + nla_total_size(sizeof(struct in6_addr))
5301                + nla_total_size(sizeof(struct prefix_cacheinfo));
5302 }
5303
5304 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5305                              struct prefix_info *pinfo, u32 portid, u32 seq,
5306                              int event, unsigned int flags)
5307 {
5308         struct prefixmsg *pmsg;
5309         struct nlmsghdr *nlh;
5310         struct prefix_cacheinfo ci;
5311
5312         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5313         if (!nlh)
5314                 return -EMSGSIZE;
5315
5316         pmsg = nlmsg_data(nlh);
5317         pmsg->prefix_family = AF_INET6;
5318         pmsg->prefix_pad1 = 0;
5319         pmsg->prefix_pad2 = 0;
5320         pmsg->prefix_ifindex = idev->dev->ifindex;
5321         pmsg->prefix_len = pinfo->prefix_len;
5322         pmsg->prefix_type = pinfo->type;
5323         pmsg->prefix_pad3 = 0;
5324         pmsg->prefix_flags = 0;
5325         if (pinfo->onlink)
5326                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5327         if (pinfo->autoconf)
5328                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5329
5330         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5331                 goto nla_put_failure;
5332         ci.preferred_time = ntohl(pinfo->prefered);
5333         ci.valid_time = ntohl(pinfo->valid);
5334         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5335                 goto nla_put_failure;
5336         nlmsg_end(skb, nlh);
5337         return 0;
5338
5339 nla_put_failure:
5340         nlmsg_cancel(skb, nlh);
5341         return -EMSGSIZE;
5342 }
5343
5344 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5345                          struct prefix_info *pinfo)
5346 {
5347         struct sk_buff *skb;
5348         struct net *net = dev_net(idev->dev);
5349         int err = -ENOBUFS;
5350
5351         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5352         if (!skb)
5353                 goto errout;
5354
5355         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5356         if (err < 0) {
5357                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5358                 WARN_ON(err == -EMSGSIZE);
5359                 kfree_skb(skb);
5360                 goto errout;
5361         }
5362         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5363         return;
5364 errout:
5365         if (err < 0)
5366                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5367 }
5368
5369 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5370 {
5371         struct net *net = dev_net(ifp->idev->dev);
5372
5373         if (event)
5374                 ASSERT_RTNL();
5375
5376         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5377
5378         switch (event) {
5379         case RTM_NEWADDR:
5380                 /*
5381                  * If the address was optimistic
5382                  * we inserted the route at the start of
5383                  * our DAD process, so we don't need
5384                  * to do it again
5385                  */
5386                 if (!(ifp->rt->rt6i_node))
5387                         ip6_ins_rt(ifp->rt);
5388                 if (ifp->idev->cnf.forwarding)
5389                         addrconf_join_anycast(ifp);
5390                 if (!ipv6_addr_any(&ifp->peer_addr))
5391                         addrconf_prefix_route(&ifp->peer_addr, 128,
5392                                               ifp->idev->dev, 0, 0);
5393                 break;
5394         case RTM_DELADDR:
5395                 if (ifp->idev->cnf.forwarding)
5396                         addrconf_leave_anycast(ifp);
5397                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5398                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5399                         struct rt6_info *rt;
5400
5401                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5402                                                        ifp->idev->dev, 0, 0);
5403                         if (rt)
5404                                 ip6_del_rt(rt);
5405                 }
5406                 if (ifp->rt) {
5407                         dst_hold(&ifp->rt->dst);
5408                         ip6_del_rt(ifp->rt);
5409                 }
5410                 rt_genid_bump_ipv6(net);
5411                 break;
5412         }
5413         atomic_inc(&net->ipv6.dev_addr_genid);
5414 }
5415
5416 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5417 {
5418         rcu_read_lock_bh();
5419         if (likely(ifp->idev->dead == 0))
5420                 __ipv6_ifa_notify(event, ifp);
5421         rcu_read_unlock_bh();
5422 }
5423
5424 #ifdef CONFIG_SYSCTL
5425
5426 static
5427 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5428                            void __user *buffer, size_t *lenp, loff_t *ppos)
5429 {
5430         int *valp = ctl->data;
5431         int val = *valp;
5432         loff_t pos = *ppos;
5433         struct ctl_table lctl;
5434         int ret;
5435
5436         /*
5437          * ctl->data points to idev->cnf.forwarding, we should
5438          * not modify it until we get the rtnl lock.
5439          */
5440         lctl = *ctl;
5441         lctl.data = &val;
5442
5443         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5444
5445         if (write)
5446                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5447         if (ret)
5448                 *ppos = pos;
5449         return ret;
5450 }
5451
5452 static
5453 int addrconf_sysctl_hop_limit(struct ctl_table *ctl, int write,
5454                               void __user *buffer, size_t *lenp, loff_t *ppos)
5455 {
5456         struct ctl_table lctl;
5457         int min_hl = 1, max_hl = 255;
5458
5459         lctl = *ctl;
5460         lctl.extra1 = &min_hl;
5461         lctl.extra2 = &max_hl;
5462
5463         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5464 }
5465
5466 static
5467 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5468                         void __user *buffer, size_t *lenp, loff_t *ppos)
5469 {
5470         struct inet6_dev *idev = ctl->extra1;
5471         int min_mtu = IPV6_MIN_MTU;
5472         struct ctl_table lctl;
5473
5474         lctl = *ctl;
5475         lctl.extra1 = &min_mtu;
5476         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5477
5478         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5479 }
5480
5481 static void dev_disable_change(struct inet6_dev *idev)
5482 {
5483         struct netdev_notifier_info info;
5484
5485         if (!idev || !idev->dev)
5486                 return;
5487
5488         netdev_notifier_info_init(&info, idev->dev);
5489         if (idev->cnf.disable_ipv6)
5490                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5491         else
5492                 addrconf_notify(NULL, NETDEV_UP, &info);
5493 }
5494
5495 static void addrconf_disable_change(struct net *net, __s32 newf)
5496 {
5497         struct net_device *dev;
5498         struct inet6_dev *idev;
5499
5500         rcu_read_lock();
5501         for_each_netdev_rcu(net, dev) {
5502                 idev = __in6_dev_get(dev);
5503                 if (idev) {
5504                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5505                         idev->cnf.disable_ipv6 = newf;
5506                         if (changed)
5507                                 dev_disable_change(idev);
5508                 }
5509         }
5510         rcu_read_unlock();
5511 }
5512
5513 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5514 {
5515         struct net *net;
5516         int old;
5517
5518         if (!rtnl_trylock())
5519                 return restart_syscall();
5520
5521         net = (struct net *)table->extra2;
5522         old = *p;
5523         *p = newf;
5524
5525         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5526                 rtnl_unlock();
5527                 return 0;
5528         }
5529
5530         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5531                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5532                 addrconf_disable_change(net, newf);
5533         } else if ((!newf) ^ (!old))
5534                 dev_disable_change((struct inet6_dev *)table->extra1);
5535
5536         rtnl_unlock();
5537         return 0;
5538 }
5539
5540 static
5541 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5542                             void __user *buffer, size_t *lenp, loff_t *ppos)
5543 {
5544         int *valp = ctl->data;
5545         int val = *valp;
5546         loff_t pos = *ppos;
5547         struct ctl_table lctl;
5548         int ret;
5549
5550         /*
5551          * ctl->data points to idev->cnf.disable_ipv6, we should
5552          * not modify it until we get the rtnl lock.
5553          */
5554         lctl = *ctl;
5555         lctl.data = &val;
5556
5557         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5558
5559         if (write)
5560                 ret = addrconf_disable_ipv6(ctl, valp, val);
5561         if (ret)
5562                 *ppos = pos;
5563         return ret;
5564 }
5565
5566 static
5567 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5568                               void __user *buffer, size_t *lenp, loff_t *ppos)
5569 {
5570         int *valp = ctl->data;
5571         int ret;
5572         int old, new;
5573
5574         old = *valp;
5575         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5576         new = *valp;
5577
5578         if (write && old != new) {
5579                 struct net *net = ctl->extra2;
5580
5581                 if (!rtnl_trylock())
5582                         return restart_syscall();
5583
5584                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5585                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5586                                                      NETCONFA_IFINDEX_DEFAULT,
5587                                                      net->ipv6.devconf_dflt);
5588                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5589                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5590                                                      NETCONFA_IFINDEX_ALL,
5591                                                      net->ipv6.devconf_all);
5592                 else {
5593                         struct inet6_dev *idev = ctl->extra1;
5594
5595                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5596                                                      idev->dev->ifindex,
5597                                                      &idev->cnf);
5598                 }
5599                 rtnl_unlock();
5600         }
5601
5602         return ret;
5603 }
5604
5605 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5606                                          void __user *buffer, size_t *lenp,
5607                                          loff_t *ppos)
5608 {
5609         int err;
5610         struct in6_addr addr;
5611         char str[IPV6_MAX_STRLEN];
5612         struct ctl_table lctl = *ctl;
5613         struct net *net = ctl->extra2;
5614         struct ipv6_stable_secret *secret = ctl->data;
5615
5616         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5617                 return -EIO;
5618
5619         lctl.maxlen = IPV6_MAX_STRLEN;
5620         lctl.data = str;
5621
5622         if (!rtnl_trylock())
5623                 return restart_syscall();
5624
5625         if (!write && !secret->initialized) {
5626                 err = -EIO;
5627                 goto out;
5628         }
5629
5630         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5631         if (err >= sizeof(str)) {
5632                 err = -EIO;
5633                 goto out;
5634         }
5635
5636         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5637         if (err || !write)
5638                 goto out;
5639
5640         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5641                 err = -EIO;
5642                 goto out;
5643         }
5644
5645         secret->initialized = true;
5646         secret->secret = addr;
5647
5648         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5649                 struct net_device *dev;
5650
5651                 for_each_netdev(net, dev) {
5652                         struct inet6_dev *idev = __in6_dev_get(dev);
5653
5654                         if (idev) {
5655                                 idev->addr_gen_mode =
5656                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5657                         }
5658                 }
5659         } else {
5660                 struct inet6_dev *idev = ctl->extra1;
5661
5662                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5663         }
5664
5665 out:
5666         rtnl_unlock();
5667
5668         return err;
5669 }
5670
5671 static
5672 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5673                                                 int write,
5674                                                 void __user *buffer,
5675                                                 size_t *lenp,
5676                                                 loff_t *ppos)
5677 {
5678         int *valp = ctl->data;
5679         int val = *valp;
5680         loff_t pos = *ppos;
5681         struct ctl_table lctl;
5682         int ret;
5683
5684         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5685          * we should not modify it until we get the rtnl lock.
5686          */
5687         lctl = *ctl;
5688         lctl.data = &val;
5689
5690         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5691
5692         if (write)
5693                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5694         if (ret)
5695                 *ppos = pos;
5696         return ret;
5697 }
5698
5699 static const struct ctl_table addrconf_sysctl[] = {
5700         {
5701                 .procname       = "forwarding",
5702                 .data           = &ipv6_devconf.forwarding,
5703                 .maxlen         = sizeof(int),
5704                 .mode           = 0644,
5705                 .proc_handler   = addrconf_sysctl_forward,
5706         },
5707         {
5708                 .procname       = "hop_limit",
5709                 .data           = &ipv6_devconf.hop_limit,
5710                 .maxlen         = sizeof(int),
5711                 .mode           = 0644,
5712                 .proc_handler   = addrconf_sysctl_hop_limit,
5713         },
5714         {
5715                 .procname       = "mtu",
5716                 .data           = &ipv6_devconf.mtu6,
5717                 .maxlen         = sizeof(int),
5718                 .mode           = 0644,
5719                 .proc_handler   = addrconf_sysctl_mtu,
5720         },
5721         {
5722                 .procname       = "accept_ra",
5723                 .data           = &ipv6_devconf.accept_ra,
5724                 .maxlen         = sizeof(int),
5725                 .mode           = 0644,
5726                 .proc_handler   = proc_dointvec,
5727         },
5728         {
5729                 .procname       = "accept_redirects",
5730                 .data           = &ipv6_devconf.accept_redirects,
5731                 .maxlen         = sizeof(int),
5732                 .mode           = 0644,
5733                 .proc_handler   = proc_dointvec,
5734         },
5735         {
5736                 .procname       = "autoconf",
5737                 .data           = &ipv6_devconf.autoconf,
5738                 .maxlen         = sizeof(int),
5739                 .mode           = 0644,
5740                 .proc_handler   = proc_dointvec,
5741         },
5742         {
5743                 .procname       = "dad_transmits",
5744                 .data           = &ipv6_devconf.dad_transmits,
5745                 .maxlen         = sizeof(int),
5746                 .mode           = 0644,
5747                 .proc_handler   = proc_dointvec,
5748         },
5749         {
5750                 .procname       = "router_solicitations",
5751                 .data           = &ipv6_devconf.rtr_solicits,
5752                 .maxlen         = sizeof(int),
5753                 .mode           = 0644,
5754                 .proc_handler   = proc_dointvec,
5755         },
5756         {
5757                 .procname       = "router_solicitation_interval",
5758                 .data           = &ipv6_devconf.rtr_solicit_interval,
5759                 .maxlen         = sizeof(int),
5760                 .mode           = 0644,
5761                 .proc_handler   = proc_dointvec_jiffies,
5762         },
5763         {
5764                 .procname       = "router_solicitation_delay",
5765                 .data           = &ipv6_devconf.rtr_solicit_delay,
5766                 .maxlen         = sizeof(int),
5767                 .mode           = 0644,
5768                 .proc_handler   = proc_dointvec_jiffies,
5769         },
5770         {
5771                 .procname       = "force_mld_version",
5772                 .data           = &ipv6_devconf.force_mld_version,
5773                 .maxlen         = sizeof(int),
5774                 .mode           = 0644,
5775                 .proc_handler   = proc_dointvec,
5776         },
5777         {
5778                 .procname       = "mldv1_unsolicited_report_interval",
5779                 .data           =
5780                         &ipv6_devconf.mldv1_unsolicited_report_interval,
5781                 .maxlen         = sizeof(int),
5782                 .mode           = 0644,
5783                 .proc_handler   = proc_dointvec_ms_jiffies,
5784         },
5785         {
5786                 .procname       = "mldv2_unsolicited_report_interval",
5787                 .data           =
5788                         &ipv6_devconf.mldv2_unsolicited_report_interval,
5789                 .maxlen         = sizeof(int),
5790                 .mode           = 0644,
5791                 .proc_handler   = proc_dointvec_ms_jiffies,
5792         },
5793         {
5794                 .procname       = "use_tempaddr",
5795                 .data           = &ipv6_devconf.use_tempaddr,
5796                 .maxlen         = sizeof(int),
5797                 .mode           = 0644,
5798                 .proc_handler   = proc_dointvec,
5799         },
5800         {
5801                 .procname       = "temp_valid_lft",
5802                 .data           = &ipv6_devconf.temp_valid_lft,
5803                 .maxlen         = sizeof(int),
5804                 .mode           = 0644,
5805                 .proc_handler   = proc_dointvec,
5806         },
5807         {
5808                 .procname       = "temp_prefered_lft",
5809                 .data           = &ipv6_devconf.temp_prefered_lft,
5810                 .maxlen         = sizeof(int),
5811                 .mode           = 0644,
5812                 .proc_handler   = proc_dointvec,
5813         },
5814         {
5815                 .procname       = "regen_max_retry",
5816                 .data           = &ipv6_devconf.regen_max_retry,
5817                 .maxlen         = sizeof(int),
5818                 .mode           = 0644,
5819                 .proc_handler   = proc_dointvec,
5820         },
5821         {
5822                 .procname       = "max_desync_factor",
5823                 .data           = &ipv6_devconf.max_desync_factor,
5824                 .maxlen         = sizeof(int),
5825                 .mode           = 0644,
5826                 .proc_handler   = proc_dointvec,
5827         },
5828         {
5829                 .procname       = "max_addresses",
5830                 .data           = &ipv6_devconf.max_addresses,
5831                 .maxlen         = sizeof(int),
5832                 .mode           = 0644,
5833                 .proc_handler   = proc_dointvec,
5834         },
5835         {
5836                 .procname       = "accept_ra_defrtr",
5837                 .data           = &ipv6_devconf.accept_ra_defrtr,
5838                 .maxlen         = sizeof(int),
5839                 .mode           = 0644,
5840                 .proc_handler   = proc_dointvec,
5841         },
5842         {
5843                 .procname       = "accept_ra_min_hop_limit",
5844                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5845                 .maxlen         = sizeof(int),
5846                 .mode           = 0644,
5847                 .proc_handler   = proc_dointvec,
5848         },
5849         {
5850                 .procname       = "accept_ra_pinfo",
5851                 .data           = &ipv6_devconf.accept_ra_pinfo,
5852                 .maxlen         = sizeof(int),
5853                 .mode           = 0644,
5854                 .proc_handler   = proc_dointvec,
5855         },
5856 #ifdef CONFIG_IPV6_ROUTER_PREF
5857         {
5858                 .procname       = "accept_ra_rtr_pref",
5859                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
5860                 .maxlen         = sizeof(int),
5861                 .mode           = 0644,
5862                 .proc_handler   = proc_dointvec,
5863         },
5864         {
5865                 .procname       = "router_probe_interval",
5866                 .data           = &ipv6_devconf.rtr_probe_interval,
5867                 .maxlen         = sizeof(int),
5868                 .mode           = 0644,
5869                 .proc_handler   = proc_dointvec_jiffies,
5870         },
5871 #ifdef CONFIG_IPV6_ROUTE_INFO
5872         {
5873                 .procname       = "accept_ra_rt_info_max_plen",
5874                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5875                 .maxlen         = sizeof(int),
5876                 .mode           = 0644,
5877                 .proc_handler   = proc_dointvec,
5878         },
5879 #endif
5880 #endif
5881         {
5882                 .procname       = "proxy_ndp",
5883                 .data           = &ipv6_devconf.proxy_ndp,
5884                 .maxlen         = sizeof(int),
5885                 .mode           = 0644,
5886                 .proc_handler   = addrconf_sysctl_proxy_ndp,
5887         },
5888         {
5889                 .procname       = "accept_source_route",
5890                 .data           = &ipv6_devconf.accept_source_route,
5891                 .maxlen         = sizeof(int),
5892                 .mode           = 0644,
5893                 .proc_handler   = proc_dointvec,
5894         },
5895 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5896         {
5897                 .procname       = "optimistic_dad",
5898                 .data           = &ipv6_devconf.optimistic_dad,
5899                 .maxlen         = sizeof(int),
5900                 .mode           = 0644,
5901                 .proc_handler   = proc_dointvec,
5902         },
5903         {
5904                 .procname       = "use_optimistic",
5905                 .data           = &ipv6_devconf.use_optimistic,
5906                 .maxlen         = sizeof(int),
5907                 .mode           = 0644,
5908                 .proc_handler   = proc_dointvec,
5909         },
5910 #endif
5911 #ifdef CONFIG_IPV6_MROUTE
5912         {
5913                 .procname       = "mc_forwarding",
5914                 .data           = &ipv6_devconf.mc_forwarding,
5915                 .maxlen         = sizeof(int),
5916                 .mode           = 0444,
5917                 .proc_handler   = proc_dointvec,
5918         },
5919 #endif
5920         {
5921                 .procname       = "disable_ipv6",
5922                 .data           = &ipv6_devconf.disable_ipv6,
5923                 .maxlen         = sizeof(int),
5924                 .mode           = 0644,
5925                 .proc_handler   = addrconf_sysctl_disable,
5926         },
5927         {
5928                 .procname       = "accept_dad",
5929                 .data           = &ipv6_devconf.accept_dad,
5930                 .maxlen         = sizeof(int),
5931                 .mode           = 0644,
5932                 .proc_handler   = proc_dointvec,
5933         },
5934         {
5935                 .procname       = "force_tllao",
5936                 .data           = &ipv6_devconf.force_tllao,
5937                 .maxlen         = sizeof(int),
5938                 .mode           = 0644,
5939                 .proc_handler   = proc_dointvec
5940         },
5941         {
5942                 .procname       = "ndisc_notify",
5943                 .data           = &ipv6_devconf.ndisc_notify,
5944                 .maxlen         = sizeof(int),
5945                 .mode           = 0644,
5946                 .proc_handler   = proc_dointvec
5947         },
5948         {
5949                 .procname       = "suppress_frag_ndisc",
5950                 .data           = &ipv6_devconf.suppress_frag_ndisc,
5951                 .maxlen         = sizeof(int),
5952                 .mode           = 0644,
5953                 .proc_handler   = proc_dointvec
5954         },
5955         {
5956                 .procname       = "accept_ra_from_local",
5957                 .data           = &ipv6_devconf.accept_ra_from_local,
5958                 .maxlen         = sizeof(int),
5959                 .mode           = 0644,
5960                 .proc_handler   = proc_dointvec,
5961         },
5962         {
5963                 .procname       = "accept_ra_mtu",
5964                 .data           = &ipv6_devconf.accept_ra_mtu,
5965                 .maxlen         = sizeof(int),
5966                 .mode           = 0644,
5967                 .proc_handler   = proc_dointvec,
5968         },
5969         {
5970                 .procname       = "stable_secret",
5971                 .data           = &ipv6_devconf.stable_secret,
5972                 .maxlen         = IPV6_MAX_STRLEN,
5973                 .mode           = 0600,
5974                 .proc_handler   = addrconf_sysctl_stable_secret,
5975         },
5976         {
5977                 .procname       = "use_oif_addrs_only",
5978                 .data           = &ipv6_devconf.use_oif_addrs_only,
5979                 .maxlen         = sizeof(int),
5980                 .mode           = 0644,
5981                 .proc_handler   = proc_dointvec,
5982         },
5983         {
5984                 .procname       = "ignore_routes_with_linkdown",
5985                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5986                 .maxlen         = sizeof(int),
5987                 .mode           = 0644,
5988                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5989         },
5990         {
5991                 .procname       = "drop_unicast_in_l2_multicast",
5992                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
5993                 .maxlen         = sizeof(int),
5994                 .mode           = 0644,
5995                 .proc_handler   = proc_dointvec,
5996         },
5997         {
5998                 .procname       = "drop_unsolicited_na",
5999                 .data           = &ipv6_devconf.drop_unsolicited_na,
6000                 .maxlen         = sizeof(int),
6001                 .mode           = 0644,
6002                 .proc_handler   = proc_dointvec,
6003         },
6004         {
6005                 .procname       = "keep_addr_on_down",
6006                 .data           = &ipv6_devconf.keep_addr_on_down,
6007                 .maxlen         = sizeof(int),
6008                 .mode           = 0644,
6009                 .proc_handler   = proc_dointvec,
6010
6011         },
6012         {
6013                 /* sentinel */
6014         }
6015 };
6016
6017 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6018                 struct inet6_dev *idev, struct ipv6_devconf *p)
6019 {
6020         int i;
6021         struct ctl_table *table;
6022         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6023
6024         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6025         if (!table)
6026                 goto out;
6027
6028         for (i = 0; table[i].data; i++) {
6029                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6030                 table[i].extra1 = idev; /* embedded; no ref */
6031                 table[i].extra2 = net;
6032         }
6033
6034         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6035
6036         p->sysctl_header = register_net_sysctl(net, path, table);
6037         if (!p->sysctl_header)
6038                 goto free;
6039
6040         return 0;
6041
6042 free:
6043         kfree(table);
6044 out:
6045         return -ENOBUFS;
6046 }
6047
6048 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6049 {
6050         struct ctl_table *table;
6051
6052         if (!p->sysctl_header)
6053                 return;
6054
6055         table = p->sysctl_header->ctl_table_arg;
6056         unregister_net_sysctl_table(p->sysctl_header);
6057         p->sysctl_header = NULL;
6058         kfree(table);
6059 }
6060
6061 static int addrconf_sysctl_register(struct inet6_dev *idev)
6062 {
6063         int err;
6064
6065         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6066                 return -EINVAL;
6067
6068         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6069                                     &ndisc_ifinfo_sysctl_change);
6070         if (err)
6071                 return err;
6072         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6073                                          idev, &idev->cnf);
6074         if (err)
6075                 neigh_sysctl_unregister(idev->nd_parms);
6076
6077         return err;
6078 }
6079
6080 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6081 {
6082         __addrconf_sysctl_unregister(&idev->cnf);
6083         neigh_sysctl_unregister(idev->nd_parms);
6084 }
6085
6086
6087 #endif
6088
6089 static int __net_init addrconf_init_net(struct net *net)
6090 {
6091         int err = -ENOMEM;
6092         struct ipv6_devconf *all, *dflt;
6093
6094         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6095         if (!all)
6096                 goto err_alloc_all;
6097
6098         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6099         if (!dflt)
6100                 goto err_alloc_dflt;
6101
6102         /* these will be inherited by all namespaces */
6103         dflt->autoconf = ipv6_defaults.autoconf;
6104         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6105
6106         dflt->stable_secret.initialized = false;
6107         all->stable_secret.initialized = false;
6108
6109         net->ipv6.devconf_all = all;
6110         net->ipv6.devconf_dflt = dflt;
6111
6112 #ifdef CONFIG_SYSCTL
6113         err = __addrconf_sysctl_register(net, "all", NULL, all);
6114         if (err < 0)
6115                 goto err_reg_all;
6116
6117         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6118         if (err < 0)
6119                 goto err_reg_dflt;
6120 #endif
6121         return 0;
6122
6123 #ifdef CONFIG_SYSCTL
6124 err_reg_dflt:
6125         __addrconf_sysctl_unregister(all);
6126 err_reg_all:
6127         kfree(dflt);
6128 #endif
6129 err_alloc_dflt:
6130         kfree(all);
6131 err_alloc_all:
6132         return err;
6133 }
6134
6135 static void __net_exit addrconf_exit_net(struct net *net)
6136 {
6137 #ifdef CONFIG_SYSCTL
6138         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6139         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6140 #endif
6141         kfree(net->ipv6.devconf_dflt);
6142         kfree(net->ipv6.devconf_all);
6143 }
6144
6145 static struct pernet_operations addrconf_ops = {
6146         .init = addrconf_init_net,
6147         .exit = addrconf_exit_net,
6148 };
6149
6150 static struct rtnl_af_ops inet6_ops __read_mostly = {
6151         .family           = AF_INET6,
6152         .fill_link_af     = inet6_fill_link_af,
6153         .get_link_af_size = inet6_get_link_af_size,
6154         .validate_link_af = inet6_validate_link_af,
6155         .set_link_af      = inet6_set_link_af,
6156 };
6157
6158 /*
6159  *      Init / cleanup code
6160  */
6161
6162 int __init addrconf_init(void)
6163 {
6164         struct inet6_dev *idev;
6165         int i, err;
6166
6167         err = ipv6_addr_label_init();
6168         if (err < 0) {
6169                 pr_crit("%s: cannot initialize default policy table: %d\n",
6170                         __func__, err);
6171                 goto out;
6172         }
6173
6174         err = register_pernet_subsys(&addrconf_ops);
6175         if (err < 0)
6176                 goto out_addrlabel;
6177
6178         addrconf_wq = create_workqueue("ipv6_addrconf");
6179         if (!addrconf_wq) {
6180                 err = -ENOMEM;
6181                 goto out_nowq;
6182         }
6183
6184         /* The addrconf netdev notifier requires that loopback_dev
6185          * has it's ipv6 private information allocated and setup
6186          * before it can bring up and give link-local addresses
6187          * to other devices which are up.
6188          *
6189          * Unfortunately, loopback_dev is not necessarily the first
6190          * entry in the global dev_base list of net devices.  In fact,
6191          * it is likely to be the very last entry on that list.
6192          * So this causes the notifier registry below to try and
6193          * give link-local addresses to all devices besides loopback_dev
6194          * first, then loopback_dev, which cases all the non-loopback_dev
6195          * devices to fail to get a link-local address.
6196          *
6197          * So, as a temporary fix, allocate the ipv6 structure for
6198          * loopback_dev first by hand.
6199          * Longer term, all of the dependencies ipv6 has upon the loopback
6200          * device and it being up should be removed.
6201          */
6202         rtnl_lock();
6203         idev = ipv6_add_dev(init_net.loopback_dev);
6204         rtnl_unlock();
6205         if (IS_ERR(idev)) {
6206                 err = PTR_ERR(idev);
6207                 goto errlo;
6208         }
6209
6210         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6211                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6212
6213         register_netdevice_notifier(&ipv6_dev_notf);
6214
6215         addrconf_verify();
6216
6217         rtnl_af_register(&inet6_ops);
6218
6219         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6220                               NULL);
6221         if (err < 0)
6222                 goto errout;
6223
6224         /* Only the first call to __rtnl_register can fail */
6225         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6226         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6227         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6228                         inet6_dump_ifaddr, NULL);
6229         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6230                         inet6_dump_ifmcaddr, NULL);
6231         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6232                         inet6_dump_ifacaddr, NULL);
6233         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6234                         inet6_netconf_dump_devconf, NULL);
6235
6236         ipv6_addr_label_rtnl_register();
6237
6238         return 0;
6239 errout:
6240         rtnl_af_unregister(&inet6_ops);
6241         unregister_netdevice_notifier(&ipv6_dev_notf);
6242 errlo:
6243         destroy_workqueue(addrconf_wq);
6244 out_nowq:
6245         unregister_pernet_subsys(&addrconf_ops);
6246 out_addrlabel:
6247         ipv6_addr_label_cleanup();
6248 out:
6249         return err;
6250 }
6251
6252 void addrconf_cleanup(void)
6253 {
6254         struct net_device *dev;
6255         int i;
6256
6257         unregister_netdevice_notifier(&ipv6_dev_notf);
6258         unregister_pernet_subsys(&addrconf_ops);
6259         ipv6_addr_label_cleanup();
6260
6261         rtnl_lock();
6262
6263         __rtnl_af_unregister(&inet6_ops);
6264
6265         /* clean dev list */
6266         for_each_netdev(&init_net, dev) {
6267                 if (__in6_dev_get(dev) == NULL)
6268                         continue;
6269                 addrconf_ifdown(dev, 1);
6270         }
6271         addrconf_ifdown(init_net.loopback_dev, 2);
6272
6273         /*
6274          *      Check hash table.
6275          */
6276         spin_lock_bh(&addrconf_hash_lock);
6277         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6278                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6279         spin_unlock_bh(&addrconf_hash_lock);
6280         cancel_delayed_work(&addr_chk_work);
6281         rtnl_unlock();
6282
6283         destroy_workqueue(addrconf_wq);
6284 }