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/sched/signal.h>
  47 #include <linux/socket.h>
  48 #include <linux/sockios.h>
  49 #include <linux/net.h>
  50 #include <linux/inet.h>
  51 #include <linux/in6.h>
  52 #include <linux/netdevice.h>
  53 #include <linux/if_addr.h>
  54 #include <linux/if_arp.h>
  55 #include <linux/if_arcnet.h>
  56 #include <linux/if_infiniband.h>
  57 #include <linux/route.h>
  58 #include <linux/inetdevice.h>
  59 #include <linux/init.h>
  60 #include <linux/slab.h>
  61 #ifdef CONFIG_SYSCTL
  62 #include <linux/sysctl.h>
  63 #endif
  64 #include <linux/capability.h>
  65 #include <linux/delay.h>
  66 #include <linux/notifier.h>
  67 #include <linux/string.h>
  68 #include <linux/hash.h>
  69
  70 #include <net/net_namespace.h>
  71 #include <net/sock.h>
  72 #include <net/snmp.h>
  73
  74 #include <net/6lowpan.h>
  75 #include <net/firewire.h>
  76 #include <net/ipv6.h>
  77 #include <net/protocol.h>
  78 #include <net/ndisc.h>
  79 #include <net/ip6_route.h>
  80 #include <net/addrconf.h>
  81 #include <net/tcp.h>
  82 #include <net/ip.h>
  83 #include <net/netlink.h>
  84 #include <net/pkt_sched.h>
  85 #include <net/l3mdev.h>
  86 #include <linux/if_tunnel.h>
  87 #include <linux/rtnetlink.h>
  88 #include <linux/netconf.h>
  89 #include <linux/random.h>
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  98
  99 #define IPV6_MAX_STRLEN \
 100         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 108 {
 109         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 110         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 111         do_div(tmp, 1000000);
 112         return (s32)tmp;
 113 }
 114
 115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 116 {
 117         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 118         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 119         do_div(tmp, 1000000);
 120         if ((s32)tmp > mrt) {
 121                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 122                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 123                 do_div(tmp, 1000000);
 124         }
 125         return (s32)tmp;
 126 }
 127
 128 #ifdef CONFIG_SYSCTL
 129 static int addrconf_sysctl_register(struct inet6_dev *idev);
 130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 131 #else
 132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 133 {
 134         return 0;
 135 }
 136
 137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 138 {
 139 }
 140 #endif
 141
 142 static void ipv6_regen_rndid(struct inet6_dev *idev);
 143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 144
 145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 146 static int ipv6_count_addresses(const struct inet6_dev *idev);
 147 static int ipv6_generate_stable_address(struct in6_addr *addr,
 148                                         u8 dad_count,
 149                                         const struct inet6_dev *idev);
 150
 151 #define IN6_ADDR_HSIZE_SHIFT    8
 152 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 153 /*
 154  *      Configured unicast address hash table
 155  */
 156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 157 static DEFINE_SPINLOCK(addrconf_hash_lock);
 158
 159 static void addrconf_verify(void);
 160 static void addrconf_verify_rtnl(void);
 161 static void addrconf_verify_work(struct work_struct *);
 162
 163 static struct workqueue_struct *addrconf_wq;
 164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 165
 166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 168
 169 static void addrconf_type_change(struct net_device *dev,
 170                                  unsigned long event);
 171 static int addrconf_ifdown(struct net_device *dev, int how);
 172
 173 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 174                                                   int plen,
 175                                                   const struct net_device *dev,
 176                                                   u32 flags, u32 noflags);
 177
 178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 179 static void addrconf_dad_work(struct work_struct *w);
 180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 181                                    bool send_na);
 182 static void addrconf_dad_run(struct inet6_dev *idev);
 183 static void addrconf_rs_timer(struct timer_list *t);
 184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 186
 187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 188                                 struct prefix_info *pinfo);
 189
 190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 191         .forwarding             = 0,
 192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 193         .mtu6                   = IPV6_MIN_MTU,
 194         .accept_ra              = 1,
 195         .accept_redirects       = 1,
 196         .autoconf               = 1,
 197         .force_mld_version      = 0,
 198         .mldv1_unsolicited_report_interval = 10 * HZ,
 199         .mldv2_unsolicited_report_interval = HZ,
 200         .dad_transmits          = 1,
 201         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 202         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 203         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205         .use_tempaddr           = 0,
 206         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 207         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 208         .regen_max_retry        = REGEN_MAX_RETRY,
 209         .max_desync_factor      = MAX_DESYNC_FACTOR,
 210         .max_addresses          = IPV6_MAX_ADDRESSES,
 211         .accept_ra_defrtr       = 1,
 212         .accept_ra_from_local   = 0,
 213         .accept_ra_min_hop_limit= 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_min_plen = 0,
 220         .accept_ra_rt_info_max_plen = 0,
 221 #endif
 222 #endif
 223         .proxy_ndp              = 0,
 224         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 225         .disable_ipv6           = 0,
 226         .accept_dad             = 0,
 227         .suppress_frag_ndisc    = 1,
 228         .accept_ra_mtu          = 1,
 229         .stable_secret          = {
 230                 .initialized = false,
 231         },
 232         .use_oif_addrs_only     = 0,
 233         .ignore_routes_with_linkdown = 0,
 234         .keep_addr_on_down      = 0,
 235         .seg6_enabled           = 0,
 236 #ifdef CONFIG_IPV6_SEG6_HMAC
 237         .seg6_require_hmac      = 0,
 238 #endif
 239         .enhanced_dad           = 1,
 240         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 241         .disable_policy         = 0,
 242 };
 243
 244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 245         .forwarding             = 0,
 246         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 247         .mtu6                   = IPV6_MIN_MTU,
 248         .accept_ra              = 1,
 249         .accept_redirects       = 1,
 250         .autoconf               = 1,
 251         .force_mld_version      = 0,
 252         .mldv1_unsolicited_report_interval = 10 * HZ,
 253         .mldv2_unsolicited_report_interval = HZ,
 254         .dad_transmits          = 1,
 255         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 256         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 257         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 258         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 259         .use_tempaddr           = 0,
 260         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 261         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 262         .regen_max_retry        = REGEN_MAX_RETRY,
 263         .max_desync_factor      = MAX_DESYNC_FACTOR,
 264         .max_addresses          = IPV6_MAX_ADDRESSES,
 265         .accept_ra_defrtr       = 1,
 266         .accept_ra_from_local   = 0,
 267         .accept_ra_min_hop_limit= 1,
 268         .accept_ra_pinfo        = 1,
 269 #ifdef CONFIG_IPV6_ROUTER_PREF
 270         .accept_ra_rtr_pref     = 1,
 271         .rtr_probe_interval     = 60 * HZ,
 272 #ifdef CONFIG_IPV6_ROUTE_INFO
 273         .accept_ra_rt_info_min_plen = 0,
 274         .accept_ra_rt_info_max_plen = 0,
 275 #endif
 276 #endif
 277         .proxy_ndp              = 0,
 278         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 279         .disable_ipv6           = 0,
 280         .accept_dad             = 1,
 281         .suppress_frag_ndisc    = 1,
 282         .accept_ra_mtu          = 1,
 283         .stable_secret          = {
 284                 .initialized = false,
 285         },
 286         .use_oif_addrs_only     = 0,
 287         .ignore_routes_with_linkdown = 0,
 288         .keep_addr_on_down      = 0,
 289         .seg6_enabled           = 0,
 290 #ifdef CONFIG_IPV6_SEG6_HMAC
 291         .seg6_require_hmac      = 0,
 292 #endif
 293         .enhanced_dad           = 1,
 294         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 295         .disable_policy         = 0,
 296 };
 297
 298 /* Check if link is ready: is it up and is a valid qdisc available */
 299 static inline bool addrconf_link_ready(const struct net_device *dev)
 300 {
 301         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 302 }
 303
 304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 305 {
 306         if (del_timer(&idev->rs_timer))
 307                 __in6_dev_put(idev);
 308 }
 309
 310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 311 {
 312         if (cancel_delayed_work(&ifp->dad_work))
 313                 __in6_ifa_put(ifp);
 314 }
 315
 316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 317                                   unsigned long when)
 318 {
 319         if (!timer_pending(&idev->rs_timer))
 320                 in6_dev_hold(idev);
 321         mod_timer(&idev->rs_timer, jiffies + when);
 322 }
 323
 324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 325                                    unsigned long delay)
 326 {
 327         in6_ifa_hold(ifp);
 328         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 329                 in6_ifa_put(ifp);
 330 }
 331
 332 static int snmp6_alloc_dev(struct inet6_dev *idev)
 333 {
 334         int i;
 335
 336         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 337         if (!idev->stats.ipv6)
 338                 goto err_ip;
 339
 340         for_each_possible_cpu(i) {
 341                 struct ipstats_mib *addrconf_stats;
 342                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 343                 u64_stats_init(&addrconf_stats->syncp);
 344         }
 345
 346
 347         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 348                                         GFP_KERNEL);
 349         if (!idev->stats.icmpv6dev)
 350                 goto err_icmp;
 351         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 352                                            GFP_KERNEL);
 353         if (!idev->stats.icmpv6msgdev)
 354                 goto err_icmpmsg;
 355
 356         return 0;
 357
 358 err_icmpmsg:
 359         kfree(idev->stats.icmpv6dev);
 360 err_icmp:
 361         free_percpu(idev->stats.ipv6);
 362 err_ip:
 363         return -ENOMEM;
 364 }
 365
 366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 367 {
 368         struct inet6_dev *ndev;
 369         int err = -ENOMEM;
 370
 371         ASSERT_RTNL();
 372
 373         if (dev->mtu < IPV6_MIN_MTU)
 374                 return ERR_PTR(-EINVAL);
 375
 376         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 377         if (!ndev)
 378                 return ERR_PTR(err);
 379
 380         rwlock_init(&ndev->lock);
 381         ndev->dev = dev;
 382         INIT_LIST_HEAD(&ndev->addr_list);
 383         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 384         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 385
 386         if (ndev->cnf.stable_secret.initialized)
 387                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 388         else
 389                 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
 390
 391         ndev->cnf.mtu6 = dev->mtu;
 392         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 393         if (!ndev->nd_parms) {
 394                 kfree(ndev);
 395                 return ERR_PTR(err);
 396         }
 397         if (ndev->cnf.forwarding)
 398                 dev_disable_lro(dev);
 399         /* We refer to the device */
 400         dev_hold(dev);
 401
 402         if (snmp6_alloc_dev(ndev) < 0) {
 403                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 404                            __func__);
 405                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 406                 dev_put(dev);
 407                 kfree(ndev);
 408                 return ERR_PTR(err);
 409         }
 410
 411         if (snmp6_register_dev(ndev) < 0) {
 412                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 413                            __func__, dev->name);
 414                 goto err_release;
 415         }
 416
 417         /* One reference from device. */
 418         refcount_set(&ndev->refcnt, 1);
 419
 420         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 421                 ndev->cnf.accept_dad = -1;
 422
 423 #if IS_ENABLED(CONFIG_IPV6_SIT)
 424         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 425                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 426                 ndev->cnf.rtr_solicits = 0;
 427         }
 428 #endif
 429
 430         INIT_LIST_HEAD(&ndev->tempaddr_list);
 431         ndev->desync_factor = U32_MAX;
 432         if ((dev->flags&IFF_LOOPBACK) ||
 433             dev->type == ARPHRD_TUNNEL ||
 434             dev->type == ARPHRD_TUNNEL6 ||
 435             dev->type == ARPHRD_SIT ||
 436             dev->type == ARPHRD_NONE) {
 437                 ndev->cnf.use_tempaddr = -1;
 438         } else
 439                 ipv6_regen_rndid(ndev);
 440
 441         ndev->token = in6addr_any;
 442
 443         if (netif_running(dev) && addrconf_link_ready(dev))
 444                 ndev->if_flags |= IF_READY;
 445
 446         ipv6_mc_init_dev(ndev);
 447         ndev->tstamp = jiffies;
 448         err = addrconf_sysctl_register(ndev);
 449         if (err) {
 450                 ipv6_mc_destroy_dev(ndev);
 451                 snmp6_unregister_dev(ndev);
 452                 goto err_release;
 453         }
 454         /* protected by rtnl_lock */
 455         rcu_assign_pointer(dev->ip6_ptr, ndev);
 456
 457         /* Join interface-local all-node multicast group */
 458         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 459
 460         /* Join all-node multicast group */
 461         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 462
 463         /* Join all-router multicast group if forwarding is set */
 464         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 465                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 466
 467         return ndev;
 468
 469 err_release:
 470         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 471         ndev->dead = 1;
 472         in6_dev_finish_destroy(ndev);
 473         return ERR_PTR(err);
 474 }
 475
 476 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 477 {
 478         struct inet6_dev *idev;
 479
 480         ASSERT_RTNL();
 481
 482         idev = __in6_dev_get(dev);
 483         if (!idev) {
 484                 idev = ipv6_add_dev(dev);
 485                 if (IS_ERR(idev))
 486                         return NULL;
 487         }
 488
 489         if (dev->flags&IFF_UP)
 490                 ipv6_mc_up(idev);
 491         return idev;
 492 }
 493
 494 static int inet6_netconf_msgsize_devconf(int type)
 495 {
 496         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 497                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 498         bool all = false;
 499
 500         if (type == NETCONFA_ALL)
 501                 all = true;
 502
 503         if (all || type == NETCONFA_FORWARDING)
 504                 size += nla_total_size(4);
 505 #ifdef CONFIG_IPV6_MROUTE
 506         if (all || type == NETCONFA_MC_FORWARDING)
 507                 size += nla_total_size(4);
 508 #endif
 509         if (all || type == NETCONFA_PROXY_NEIGH)
 510                 size += nla_total_size(4);
 511
 512         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 513                 size += nla_total_size(4);
 514
 515         return size;
 516 }
 517
 518 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 519                                       struct ipv6_devconf *devconf, u32 portid,
 520                                       u32 seq, int event, unsigned int flags,
 521                                       int type)
 522 {
 523         struct nlmsghdr  *nlh;
 524         struct netconfmsg *ncm;
 525         bool all = false;
 526
 527         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 528                         flags);
 529         if (!nlh)
 530                 return -EMSGSIZE;
 531
 532         if (type == NETCONFA_ALL)
 533                 all = true;
 534
 535         ncm = nlmsg_data(nlh);
 536         ncm->ncm_family = AF_INET6;
 537
 538         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 539                 goto nla_put_failure;
 540
 541         if (!devconf)
 542                 goto out;
 543
 544         if ((all || type == NETCONFA_FORWARDING) &&
 545             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 546                 goto nla_put_failure;
 547 #ifdef CONFIG_IPV6_MROUTE
 548         if ((all || type == NETCONFA_MC_FORWARDING) &&
 549             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 550                         devconf->mc_forwarding) < 0)
 551                 goto nla_put_failure;
 552 #endif
 553         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 554             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 555                 goto nla_put_failure;
 556
 557         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 558             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 559                         devconf->ignore_routes_with_linkdown) < 0)
 560                 goto nla_put_failure;
 561
 562 out:
 563         nlmsg_end(skb, nlh);
 564         return 0;
 565
 566 nla_put_failure:
 567         nlmsg_cancel(skb, nlh);
 568         return -EMSGSIZE;
 569 }
 570
 571 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 572                                   int ifindex, struct ipv6_devconf *devconf)
 573 {
 574         struct sk_buff *skb;
 575         int err = -ENOBUFS;
 576
 577         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 578         if (!skb)
 579                 goto errout;
 580
 581         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 582                                          event, 0, type);
 583         if (err < 0) {
 584                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 585                 WARN_ON(err == -EMSGSIZE);
 586                 kfree_skb(skb);
 587                 goto errout;
 588         }
 589         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 590         return;
 591 errout:
 592         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 593 }
 594
 595 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 596         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 597         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 598         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 599         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 600 };
 601
 602 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 603                                      struct nlmsghdr *nlh,
 604                                      struct netlink_ext_ack *extack)
 605 {
 606         struct net *net = sock_net(in_skb->sk);
 607         struct nlattr *tb[NETCONFA_MAX+1];
 608         struct inet6_dev *in6_dev = NULL;
 609         struct net_device *dev = NULL;
 610         struct netconfmsg *ncm;
 611         struct sk_buff *skb;
 612         struct ipv6_devconf *devconf;
 613         int ifindex;
 614         int err;
 615
 616         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 617                           devconf_ipv6_policy, extack);
 618         if (err < 0)
 619                 return err;
 620
 621         if (!tb[NETCONFA_IFINDEX])
 622                 return -EINVAL;
 623
 624         err = -EINVAL;
 625         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 626         switch (ifindex) {
 627         case NETCONFA_IFINDEX_ALL:
 628                 devconf = net->ipv6.devconf_all;
 629                 break;
 630         case NETCONFA_IFINDEX_DEFAULT:
 631                 devconf = net->ipv6.devconf_dflt;
 632                 break;
 633         default:
 634                 dev = dev_get_by_index(net, ifindex);
 635                 if (!dev)
 636                         return -EINVAL;
 637                 in6_dev = in6_dev_get(dev);
 638                 if (!in6_dev)
 639                         goto errout;
 640                 devconf = &in6_dev->cnf;
 641                 break;
 642         }
 643
 644         err = -ENOBUFS;
 645         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 646         if (!skb)
 647                 goto errout;
 648
 649         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 650                                          NETLINK_CB(in_skb).portid,
 651                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 652                                          NETCONFA_ALL);
 653         if (err < 0) {
 654                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 655                 WARN_ON(err == -EMSGSIZE);
 656                 kfree_skb(skb);
 657                 goto errout;
 658         }
 659         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 660 errout:
 661         if (in6_dev)
 662                 in6_dev_put(in6_dev);
 663         if (dev)
 664                 dev_put(dev);
 665         return err;
 666 }
 667
 668 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 669                                       struct netlink_callback *cb)
 670 {
 671         struct net *net = sock_net(skb->sk);
 672         int h, s_h;
 673         int idx, s_idx;
 674         struct net_device *dev;
 675         struct inet6_dev *idev;
 676         struct hlist_head *head;
 677
 678         s_h = cb->args[0];
 679         s_idx = idx = cb->args[1];
 680
 681         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 682                 idx = 0;
 683                 head = &net->dev_index_head[h];
 684                 rcu_read_lock();
 685                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 686                           net->dev_base_seq;
 687                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 688                         if (idx < s_idx)
 689                                 goto cont;
 690                         idev = __in6_dev_get(dev);
 691                         if (!idev)
 692                                 goto cont;
 693
 694                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 695                                                        &idev->cnf,
 696                                                        NETLINK_CB(cb->skb).portid,
 697                                                        cb->nlh->nlmsg_seq,
 698                                                        RTM_NEWNETCONF,
 699                                                        NLM_F_MULTI,
 700                                                        NETCONFA_ALL) < 0) {
 701                                 rcu_read_unlock();
 702                                 goto done;
 703                         }
 704                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 705 cont:
 706                         idx++;
 707                 }
 708                 rcu_read_unlock();
 709         }
 710         if (h == NETDEV_HASHENTRIES) {
 711                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 712                                                net->ipv6.devconf_all,
 713                                                NETLINK_CB(cb->skb).portid,
 714                                                cb->nlh->nlmsg_seq,
 715                                                RTM_NEWNETCONF, NLM_F_MULTI,
 716                                                NETCONFA_ALL) < 0)
 717                         goto done;
 718                 else
 719                         h++;
 720         }
 721         if (h == NETDEV_HASHENTRIES + 1) {
 722                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 723                                                net->ipv6.devconf_dflt,
 724                                                NETLINK_CB(cb->skb).portid,
 725                                                cb->nlh->nlmsg_seq,
 726                                                RTM_NEWNETCONF, NLM_F_MULTI,
 727                                                NETCONFA_ALL) < 0)
 728                         goto done;
 729                 else
 730                         h++;
 731         }
 732 done:
 733         cb->args[0] = h;
 734         cb->args[1] = idx;
 735
 736         return skb->len;
 737 }
 738
 739 #ifdef CONFIG_SYSCTL
 740 static void dev_forward_change(struct inet6_dev *idev)
 741 {
 742         struct net_device *dev;
 743         struct inet6_ifaddr *ifa;
 744
 745         if (!idev)
 746                 return;
 747         dev = idev->dev;
 748         if (idev->cnf.forwarding)
 749                 dev_disable_lro(dev);
 750         if (dev->flags & IFF_MULTICAST) {
 751                 if (idev->cnf.forwarding) {
 752                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 753                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 754                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 755                 } else {
 756                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 757                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 758                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 759                 }
 760         }
 761
 762         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 763                 if (ifa->flags&IFA_F_TENTATIVE)
 764                         continue;
 765                 if (idev->cnf.forwarding)
 766                         addrconf_join_anycast(ifa);
 767                 else
 768                         addrconf_leave_anycast(ifa);
 769         }
 770         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 771                                      NETCONFA_FORWARDING,
 772                                      dev->ifindex, &idev->cnf);
 773 }
 774
 775
 776 static void addrconf_forward_change(struct net *net, __s32 newf)
 777 {
 778         struct net_device *dev;
 779         struct inet6_dev *idev;
 780
 781         for_each_netdev(net, dev) {
 782                 idev = __in6_dev_get(dev);
 783                 if (idev) {
 784                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 785                         idev->cnf.forwarding = newf;
 786                         if (changed)
 787                                 dev_forward_change(idev);
 788                 }
 789         }
 790 }
 791
 792 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 793 {
 794         struct net *net;
 795         int old;
 796
 797         if (!rtnl_trylock())
 798                 return restart_syscall();
 799
 800         net = (struct net *)table->extra2;
 801         old = *p;
 802         *p = newf;
 803
 804         if (p == &net->ipv6.devconf_dflt->forwarding) {
 805                 if ((!newf) ^ (!old))
 806                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 807                                                      NETCONFA_FORWARDING,
 808                                                      NETCONFA_IFINDEX_DEFAULT,
 809                                                      net->ipv6.devconf_dflt);
 810                 rtnl_unlock();
 811                 return 0;
 812         }
 813
 814         if (p == &net->ipv6.devconf_all->forwarding) {
 815                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 816
 817                 net->ipv6.devconf_dflt->forwarding = newf;
 818                 if ((!newf) ^ (!old_dflt))
 819                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 820                                                      NETCONFA_FORWARDING,
 821                                                      NETCONFA_IFINDEX_DEFAULT,
 822                                                      net->ipv6.devconf_dflt);
 823
 824                 addrconf_forward_change(net, newf);
 825                 if ((!newf) ^ (!old))
 826                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 827                                                      NETCONFA_FORWARDING,
 828                                                      NETCONFA_IFINDEX_ALL,
 829                                                      net->ipv6.devconf_all);
 830         } else if ((!newf) ^ (!old))
 831                 dev_forward_change((struct inet6_dev *)table->extra1);
 832         rtnl_unlock();
 833
 834         if (newf)
 835                 rt6_purge_dflt_routers(net);
 836         return 1;
 837 }
 838
 839 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 840 {
 841         struct net_device *dev;
 842         struct inet6_dev *idev;
 843
 844         for_each_netdev(net, dev) {
 845                 idev = __in6_dev_get(dev);
 846                 if (idev) {
 847                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 848
 849                         idev->cnf.ignore_routes_with_linkdown = newf;
 850                         if (changed)
 851                                 inet6_netconf_notify_devconf(dev_net(dev),
 852                                                              RTM_NEWNETCONF,
 853                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 854                                                              dev->ifindex,
 855                                                              &idev->cnf);
 856                 }
 857         }
 858 }
 859
 860 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 861 {
 862         struct net *net;
 863         int old;
 864
 865         if (!rtnl_trylock())
 866                 return restart_syscall();
 867
 868         net = (struct net *)table->extra2;
 869         old = *p;
 870         *p = newf;
 871
 872         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 873                 if ((!newf) ^ (!old))
 874                         inet6_netconf_notify_devconf(net,
 875                                                      RTM_NEWNETCONF,
 876                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 877                                                      NETCONFA_IFINDEX_DEFAULT,
 878                                                      net->ipv6.devconf_dflt);
 879                 rtnl_unlock();
 880                 return 0;
 881         }
 882
 883         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 884                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 885                 addrconf_linkdown_change(net, newf);
 886                 if ((!newf) ^ (!old))
 887                         inet6_netconf_notify_devconf(net,
 888                                                      RTM_NEWNETCONF,
 889                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 890                                                      NETCONFA_IFINDEX_ALL,
 891                                                      net->ipv6.devconf_all);
 892         }
 893         rtnl_unlock();
 894
 895         return 1;
 896 }
 897
 898 #endif
 899
 900 /* Nobody refers to this ifaddr, destroy it */
 901 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 902 {
 903         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 904
 905 #ifdef NET_REFCNT_DEBUG
 906         pr_debug("%s\n", __func__);
 907 #endif
 908
 909         in6_dev_put(ifp->idev);
 910
 911         if (cancel_delayed_work(&ifp->dad_work))
 912                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 913                           ifp);
 914
 915         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 916                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 917                 return;
 918         }
 919
 920         kfree_rcu(ifp, rcu);
 921 }
 922
 923 static void
 924 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 925 {
 926         struct list_head *p;
 927         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 928
 929         /*
 930          * Each device address list is sorted in order of scope -
 931          * global before linklocal.
 932          */
 933         list_for_each(p, &idev->addr_list) {
 934                 struct inet6_ifaddr *ifa
 935                         = list_entry(p, struct inet6_ifaddr, if_list);
 936                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 937                         break;
 938         }
 939
 940         list_add_tail_rcu(&ifp->if_list, p);
 941 }
 942
 943 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 944 {
 945         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 946
 947         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 948 }
 949
 950 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 951                                struct net_device *dev, unsigned int hash)
 952 {
 953         struct inet6_ifaddr *ifp;
 954
 955         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
 956                 if (!net_eq(dev_net(ifp->idev->dev), net))
 957                         continue;
 958                 if (ipv6_addr_equal(&ifp->addr, addr)) {
 959                         if (!dev || ifp->idev->dev == dev)
 960                                 return true;
 961                 }
 962         }
 963         return false;
 964 }
 965
 966 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
 967 {
 968         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
 969         int err = 0;
 970
 971         spin_lock(&addrconf_hash_lock);
 972
 973         /* Ignore adding duplicate addresses on an interface */
 974         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
 975                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
 976                 err = -EEXIST;
 977         } else {
 978                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 979         }
 980
 981         spin_unlock(&addrconf_hash_lock);
 982
 983         return err;
 984 }
 985
 986 /* On success it returns ifp with increased reference count */
 987
 988 static struct inet6_ifaddr *
 989 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
 990               bool can_block, struct netlink_ext_ack *extack)
 991 {
 992         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
 993         int addr_type = ipv6_addr_type(cfg->pfx);
 994         struct net *net = dev_net(idev->dev);
 995         struct inet6_ifaddr *ifa = NULL;
 996         struct fib6_info *f6i = NULL;
 997         int err = 0;
 998
 999         if (addr_type == IPV6_ADDR_ANY ||
1000             addr_type & IPV6_ADDR_MULTICAST ||
1001             (!(idev->dev->flags & IFF_LOOPBACK) &&
1002              addr_type & IPV6_ADDR_LOOPBACK))
1003                 return ERR_PTR(-EADDRNOTAVAIL);
1004
1005         if (idev->dead) {
1006                 err = -ENODEV;                  /*XXX*/
1007                 goto out;
1008         }
1009
1010         if (idev->cnf.disable_ipv6) {
1011                 err = -EACCES;
1012                 goto out;
1013         }
1014
1015         /* validator notifier needs to be blocking;
1016          * do not call in atomic context
1017          */
1018         if (can_block) {
1019                 struct in6_validator_info i6vi = {
1020                         .i6vi_addr = *cfg->pfx,
1021                         .i6vi_dev = idev,
1022                         .extack = extack,
1023                 };
1024
1025                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1026                 err = notifier_to_errno(err);
1027                 if (err < 0)
1028                         goto out;
1029         }
1030
1031         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1032         if (!ifa) {
1033                 err = -ENOBUFS;
1034                 goto out;
1035         }
1036
1037         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1038         if (IS_ERR(f6i)) {
1039                 err = PTR_ERR(f6i);
1040                 f6i = NULL;
1041                 goto out;
1042         }
1043
1044         if (net->ipv6.devconf_all->disable_policy ||
1045             idev->cnf.disable_policy)
1046                 f6i->dst_nopolicy = true;
1047
1048         neigh_parms_data_state_setall(idev->nd_parms);
1049
1050         ifa->addr = *cfg->pfx;
1051         if (cfg->peer_pfx)
1052                 ifa->peer_addr = *cfg->peer_pfx;
1053
1054         spin_lock_init(&ifa->lock);
1055         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1056         INIT_HLIST_NODE(&ifa->addr_lst);
1057         ifa->scope = cfg->scope;
1058         ifa->prefix_len = cfg->plen;
1059         ifa->rt_priority = cfg->rt_priority;
1060         ifa->flags = cfg->ifa_flags;
1061         /* No need to add the TENTATIVE flag for addresses with NODAD */
1062         if (!(cfg->ifa_flags & IFA_F_NODAD))
1063                 ifa->flags |= IFA_F_TENTATIVE;
1064         ifa->valid_lft = cfg->valid_lft;
1065         ifa->prefered_lft = cfg->preferred_lft;
1066         ifa->cstamp = ifa->tstamp = jiffies;
1067         ifa->tokenized = false;
1068
1069         ifa->rt = f6i;
1070
1071         ifa->idev = idev;
1072         in6_dev_hold(idev);
1073
1074         /* For caller */
1075         refcount_set(&ifa->refcnt, 1);
1076
1077         rcu_read_lock_bh();
1078
1079         err = ipv6_add_addr_hash(idev->dev, ifa);
1080         if (err < 0) {
1081                 rcu_read_unlock_bh();
1082                 goto out;
1083         }
1084
1085         write_lock(&idev->lock);
1086
1087         /* Add to inet6_dev unicast addr list. */
1088         ipv6_link_dev_addr(idev, ifa);
1089
1090         if (ifa->flags&IFA_F_TEMPORARY) {
1091                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1092                 in6_ifa_hold(ifa);
1093         }
1094
1095         in6_ifa_hold(ifa);
1096         write_unlock(&idev->lock);
1097
1098         rcu_read_unlock_bh();
1099
1100         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1101 out:
1102         if (unlikely(err < 0)) {
1103                 fib6_info_release(f6i);
1104
1105                 if (ifa) {
1106                         if (ifa->idev)
1107                                 in6_dev_put(ifa->idev);
1108                         kfree(ifa);
1109                 }
1110                 ifa = ERR_PTR(err);
1111         }
1112
1113         return ifa;
1114 }
1115
1116 enum cleanup_prefix_rt_t {
1117         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1118         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1119         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1120 };
1121
1122 /*
1123  * Check, whether the prefix for ifp would still need a prefix route
1124  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1125  * constants.
1126  *
1127  * 1) we don't purge prefix if address was not permanent.
1128  *    prefix is managed by its own lifetime.
1129  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1130  * 3) if there are no addresses, delete prefix.
1131  * 4) if there are still other permanent address(es),
1132  *    corresponding prefix is still permanent.
1133  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1134  *    don't purge the prefix, assume user space is managing it.
1135  * 6) otherwise, update prefix lifetime to the
1136  *    longest valid lifetime among the corresponding
1137  *    addresses on the device.
1138  *    Note: subsequent RA will update lifetime.
1139  **/
1140 static enum cleanup_prefix_rt_t
1141 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1142 {
1143         struct inet6_ifaddr *ifa;
1144         struct inet6_dev *idev = ifp->idev;
1145         unsigned long lifetime;
1146         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1147
1148         *expires = jiffies;
1149
1150         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1151                 if (ifa == ifp)
1152                         continue;
1153                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1154                                        ifp->prefix_len))
1155                         continue;
1156                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1157                         return CLEANUP_PREFIX_RT_NOP;
1158
1159                 action = CLEANUP_PREFIX_RT_EXPIRE;
1160
1161                 spin_lock(&ifa->lock);
1162
1163                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1164                 /*
1165                  * Note: Because this address is
1166                  * not permanent, lifetime <
1167                  * LONG_MAX / HZ here.
1168                  */
1169                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1170                         *expires = ifa->tstamp + lifetime * HZ;
1171                 spin_unlock(&ifa->lock);
1172         }
1173
1174         return action;
1175 }
1176
1177 static void
1178 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1179 {
1180         struct fib6_info *f6i;
1181
1182         f6i = addrconf_get_prefix_route(&ifp->addr,
1183                                        ifp->prefix_len,
1184                                        ifp->idev->dev,
1185                                        0, RTF_GATEWAY | RTF_DEFAULT);
1186         if (f6i) {
1187                 if (del_rt)
1188                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1189                 else {
1190                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1191                                 fib6_set_expires(f6i, expires);
1192                         fib6_info_release(f6i);
1193                 }
1194         }
1195 }
1196
1197
1198 /* This function wants to get referenced ifp and releases it before return */
1199
1200 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1201 {
1202         int state;
1203         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1204         unsigned long expires;
1205
1206         ASSERT_RTNL();
1207
1208         spin_lock_bh(&ifp->lock);
1209         state = ifp->state;
1210         ifp->state = INET6_IFADDR_STATE_DEAD;
1211         spin_unlock_bh(&ifp->lock);
1212
1213         if (state == INET6_IFADDR_STATE_DEAD)
1214                 goto out;
1215
1216         spin_lock_bh(&addrconf_hash_lock);
1217         hlist_del_init_rcu(&ifp->addr_lst);
1218         spin_unlock_bh(&addrconf_hash_lock);
1219
1220         write_lock_bh(&ifp->idev->lock);
1221
1222         if (ifp->flags&IFA_F_TEMPORARY) {
1223                 list_del(&ifp->tmp_list);
1224                 if (ifp->ifpub) {
1225                         in6_ifa_put(ifp->ifpub);
1226                         ifp->ifpub = NULL;
1227                 }
1228                 __in6_ifa_put(ifp);
1229         }
1230
1231         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1232                 action = check_cleanup_prefix_route(ifp, &expires);
1233
1234         list_del_rcu(&ifp->if_list);
1235         __in6_ifa_put(ifp);
1236
1237         write_unlock_bh(&ifp->idev->lock);
1238
1239         addrconf_del_dad_work(ifp);
1240
1241         ipv6_ifa_notify(RTM_DELADDR, ifp);
1242
1243         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1244
1245         if (action != CLEANUP_PREFIX_RT_NOP) {
1246                 cleanup_prefix_route(ifp, expires,
1247                         action == CLEANUP_PREFIX_RT_DEL);
1248         }
1249
1250         /* clean up prefsrc entries */
1251         rt6_remove_prefsrc(ifp);
1252 out:
1253         in6_ifa_put(ifp);
1254 }
1255
1256 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1257                                 struct inet6_ifaddr *ift,
1258                                 bool block)
1259 {
1260         struct inet6_dev *idev = ifp->idev;
1261         struct in6_addr addr, *tmpaddr;
1262         unsigned long tmp_tstamp, age;
1263         unsigned long regen_advance;
1264         struct ifa6_config cfg;
1265         int ret = 0;
1266         unsigned long now = jiffies;
1267         long max_desync_factor;
1268         s32 cnf_temp_preferred_lft;
1269
1270         write_lock_bh(&idev->lock);
1271         if (ift) {
1272                 spin_lock_bh(&ift->lock);
1273                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1274                 spin_unlock_bh(&ift->lock);
1275                 tmpaddr = &addr;
1276         } else {
1277                 tmpaddr = NULL;
1278         }
1279 retry:
1280         in6_dev_hold(idev);
1281         if (idev->cnf.use_tempaddr <= 0) {
1282                 write_unlock_bh(&idev->lock);
1283                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1284                 in6_dev_put(idev);
1285                 ret = -1;
1286                 goto out;
1287         }
1288         spin_lock_bh(&ifp->lock);
1289         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1290                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1291                 spin_unlock_bh(&ifp->lock);
1292                 write_unlock_bh(&idev->lock);
1293                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1294                         __func__);
1295                 in6_dev_put(idev);
1296                 ret = -1;
1297                 goto out;
1298         }
1299         in6_ifa_hold(ifp);
1300         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1301         ipv6_try_regen_rndid(idev, tmpaddr);
1302         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1303         age = (now - ifp->tstamp) / HZ;
1304
1305         regen_advance = idev->cnf.regen_max_retry *
1306                         idev->cnf.dad_transmits *
1307                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1308
1309         /* recalculate max_desync_factor each time and update
1310          * idev->desync_factor if it's larger
1311          */
1312         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1313         max_desync_factor = min_t(__u32,
1314                                   idev->cnf.max_desync_factor,
1315                                   cnf_temp_preferred_lft - regen_advance);
1316
1317         if (unlikely(idev->desync_factor > max_desync_factor)) {
1318                 if (max_desync_factor > 0) {
1319                         get_random_bytes(&idev->desync_factor,
1320                                          sizeof(idev->desync_factor));
1321                         idev->desync_factor %= max_desync_factor;
1322                 } else {
1323                         idev->desync_factor = 0;
1324                 }
1325         }
1326
1327         memset(&cfg, 0, sizeof(cfg));
1328         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1329                               idev->cnf.temp_valid_lft + age);
1330         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1331         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1332
1333         cfg.plen = ifp->prefix_len;
1334         tmp_tstamp = ifp->tstamp;
1335         spin_unlock_bh(&ifp->lock);
1336
1337         write_unlock_bh(&idev->lock);
1338
1339         /* A temporary address is created only if this calculated Preferred
1340          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1341          * an implementation must not create a temporary address with a zero
1342          * Preferred Lifetime.
1343          * Use age calculation as in addrconf_verify to avoid unnecessary
1344          * temporary addresses being generated.
1345          */
1346         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1347         if (cfg.preferred_lft <= regen_advance + age) {
1348                 in6_ifa_put(ifp);
1349                 in6_dev_put(idev);
1350                 ret = -1;
1351                 goto out;
1352         }
1353
1354         cfg.ifa_flags = IFA_F_TEMPORARY;
1355         /* set in addrconf_prefix_rcv() */
1356         if (ifp->flags & IFA_F_OPTIMISTIC)
1357                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1358
1359         cfg.pfx = &addr;
1360         cfg.scope = ipv6_addr_scope(cfg.pfx);
1361
1362         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1363         if (IS_ERR(ift)) {
1364                 in6_ifa_put(ifp);
1365                 in6_dev_put(idev);
1366                 pr_info("%s: retry temporary address regeneration\n", __func__);
1367                 tmpaddr = &addr;
1368                 write_lock_bh(&idev->lock);
1369                 goto retry;
1370         }
1371
1372         spin_lock_bh(&ift->lock);
1373         ift->ifpub = ifp;
1374         ift->cstamp = now;
1375         ift->tstamp = tmp_tstamp;
1376         spin_unlock_bh(&ift->lock);
1377
1378         addrconf_dad_start(ift);
1379         in6_ifa_put(ift);
1380         in6_dev_put(idev);
1381 out:
1382         return ret;
1383 }
1384
1385 /*
1386  *      Choose an appropriate source address (RFC3484)
1387  */
1388 enum {
1389         IPV6_SADDR_RULE_INIT = 0,
1390         IPV6_SADDR_RULE_LOCAL,
1391         IPV6_SADDR_RULE_SCOPE,
1392         IPV6_SADDR_RULE_PREFERRED,
1393 #ifdef CONFIG_IPV6_MIP6
1394         IPV6_SADDR_RULE_HOA,
1395 #endif
1396         IPV6_SADDR_RULE_OIF,
1397         IPV6_SADDR_RULE_LABEL,
1398         IPV6_SADDR_RULE_PRIVACY,
1399         IPV6_SADDR_RULE_ORCHID,
1400         IPV6_SADDR_RULE_PREFIX,
1401 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1402         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1403 #endif
1404         IPV6_SADDR_RULE_MAX
1405 };
1406
1407 struct ipv6_saddr_score {
1408         int                     rule;
1409         int                     addr_type;
1410         struct inet6_ifaddr     *ifa;
1411         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1412         int                     scopedist;
1413         int                     matchlen;
1414 };
1415
1416 struct ipv6_saddr_dst {
1417         const struct in6_addr *addr;
1418         int ifindex;
1419         int scope;
1420         int label;
1421         unsigned int prefs;
1422 };
1423
1424 static inline int ipv6_saddr_preferred(int type)
1425 {
1426         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1427                 return 1;
1428         return 0;
1429 }
1430
1431 static bool ipv6_use_optimistic_addr(struct net *net,
1432                                      struct inet6_dev *idev)
1433 {
1434 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1435         if (!idev)
1436                 return false;
1437         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1438                 return false;
1439         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1440                 return false;
1441
1442         return true;
1443 #else
1444         return false;
1445 #endif
1446 }
1447
1448 static bool ipv6_allow_optimistic_dad(struct net *net,
1449                                       struct inet6_dev *idev)
1450 {
1451 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1452         if (!idev)
1453                 return false;
1454         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1455                 return false;
1456
1457         return true;
1458 #else
1459         return false;
1460 #endif
1461 }
1462
1463 static int ipv6_get_saddr_eval(struct net *net,
1464                                struct ipv6_saddr_score *score,
1465                                struct ipv6_saddr_dst *dst,
1466                                int i)
1467 {
1468         int ret;
1469
1470         if (i <= score->rule) {
1471                 switch (i) {
1472                 case IPV6_SADDR_RULE_SCOPE:
1473                         ret = score->scopedist;
1474                         break;
1475                 case IPV6_SADDR_RULE_PREFIX:
1476                         ret = score->matchlen;
1477                         break;
1478                 default:
1479                         ret = !!test_bit(i, score->scorebits);
1480                 }
1481                 goto out;
1482         }
1483
1484         switch (i) {
1485         case IPV6_SADDR_RULE_INIT:
1486                 /* Rule 0: remember if hiscore is not ready yet */
1487                 ret = !!score->ifa;
1488                 break;
1489         case IPV6_SADDR_RULE_LOCAL:
1490                 /* Rule 1: Prefer same address */
1491                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1492                 break;
1493         case IPV6_SADDR_RULE_SCOPE:
1494                 /* Rule 2: Prefer appropriate scope
1495                  *
1496                  *      ret
1497                  *       ^
1498                  *    -1 |  d 15
1499                  *    ---+--+-+---> scope
1500                  *       |
1501                  *       |             d is scope of the destination.
1502                  *  B-d  |  \
1503                  *       |   \      <- smaller scope is better if
1504                  *  B-15 |    \        if scope is enough for destination.
1505                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1506                  * d-C-1 | /
1507                  *       |/         <- greater is better
1508                  *   -C  /             if scope is not enough for destination.
1509                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1510                  *
1511                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1512                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1513                  * Assume B = 0 and we get C > 29.
1514                  */
1515                 ret = __ipv6_addr_src_scope(score->addr_type);
1516                 if (ret >= dst->scope)
1517                         ret = -ret;
1518                 else
1519                         ret -= 128;     /* 30 is enough */
1520                 score->scopedist = ret;
1521                 break;
1522         case IPV6_SADDR_RULE_PREFERRED:
1523             {
1524                 /* Rule 3: Avoid deprecated and optimistic addresses */
1525                 u8 avoid = IFA_F_DEPRECATED;
1526
1527                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1528                         avoid |= IFA_F_OPTIMISTIC;
1529                 ret = ipv6_saddr_preferred(score->addr_type) ||
1530                       !(score->ifa->flags & avoid);
1531                 break;
1532             }
1533 #ifdef CONFIG_IPV6_MIP6
1534         case IPV6_SADDR_RULE_HOA:
1535             {
1536                 /* Rule 4: Prefer home address */
1537                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1538                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1539                 break;
1540             }
1541 #endif
1542         case IPV6_SADDR_RULE_OIF:
1543                 /* Rule 5: Prefer outgoing interface */
1544                 ret = (!dst->ifindex ||
1545                        dst->ifindex == score->ifa->idev->dev->ifindex);
1546                 break;
1547         case IPV6_SADDR_RULE_LABEL:
1548                 /* Rule 6: Prefer matching label */
1549                 ret = ipv6_addr_label(net,
1550                                       &score->ifa->addr, score->addr_type,
1551                                       score->ifa->idev->dev->ifindex) == dst->label;
1552                 break;
1553         case IPV6_SADDR_RULE_PRIVACY:
1554             {
1555                 /* Rule 7: Prefer public address
1556                  * Note: prefer temporary address if use_tempaddr >= 2
1557                  */
1558                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1559                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1560                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1561                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1562                 break;
1563             }
1564         case IPV6_SADDR_RULE_ORCHID:
1565                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1566                  *          non-ORCHID vs non-ORCHID
1567                  */
1568                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1569                         ipv6_addr_orchid(dst->addr));
1570                 break;
1571         case IPV6_SADDR_RULE_PREFIX:
1572                 /* Rule 8: Use longest matching prefix */
1573                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1574                 if (ret > score->ifa->prefix_len)
1575                         ret = score->ifa->prefix_len;
1576                 score->matchlen = ret;
1577                 break;
1578 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1579         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1580                 /* Optimistic addresses still have lower precedence than other
1581                  * preferred addresses.
1582                  */
1583                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1584                 break;
1585 #endif
1586         default:
1587                 ret = 0;
1588         }
1589
1590         if (ret)
1591                 __set_bit(i, score->scorebits);
1592         score->rule = i;
1593 out:
1594         return ret;
1595 }
1596
1597 static int __ipv6_dev_get_saddr(struct net *net,
1598                                 struct ipv6_saddr_dst *dst,
1599                                 struct inet6_dev *idev,
1600                                 struct ipv6_saddr_score *scores,
1601                                 int hiscore_idx)
1602 {
1603         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1604
1605         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1606                 int i;
1607
1608                 /*
1609                  * - Tentative Address (RFC2462 section 5.4)
1610                  *  - A tentative address is not considered
1611                  *    "assigned to an interface" in the traditional
1612                  *    sense, unless it is also flagged as optimistic.
1613                  * - Candidate Source Address (section 4)
1614                  *  - In any case, anycast addresses, multicast
1615                  *    addresses, and the unspecified address MUST
1616                  *    NOT be included in a candidate set.
1617                  */
1618                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1619                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1620                         continue;
1621
1622                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1623
1624                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1625                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1626                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1627                                             idev->dev->name);
1628                         continue;
1629                 }
1630
1631                 score->rule = -1;
1632                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1633
1634                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1635                         int minihiscore, miniscore;
1636
1637                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1638                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1639
1640                         if (minihiscore > miniscore) {
1641                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1642                                     score->scopedist > 0) {
1643                                         /*
1644                                          * special case:
1645                                          * each remaining entry
1646                                          * has too small (not enough)
1647                                          * scope, because ifa entries
1648                                          * are sorted by their scope
1649                                          * values.
1650                                          */
1651                                         goto out;
1652                                 }
1653                                 break;
1654                         } else if (minihiscore < miniscore) {
1655                                 swap(hiscore, score);
1656                                 hiscore_idx = 1 - hiscore_idx;
1657
1658                                 /* restore our iterator */
1659                                 score->ifa = hiscore->ifa;
1660
1661                                 break;
1662                         }
1663                 }
1664         }
1665 out:
1666         return hiscore_idx;
1667 }
1668
1669 static int ipv6_get_saddr_master(struct net *net,
1670                                  const struct net_device *dst_dev,
1671                                  const struct net_device *master,
1672                                  struct ipv6_saddr_dst *dst,
1673                                  struct ipv6_saddr_score *scores,
1674                                  int hiscore_idx)
1675 {
1676         struct inet6_dev *idev;
1677
1678         idev = __in6_dev_get(dst_dev);
1679         if (idev)
1680                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1681                                                    scores, hiscore_idx);
1682
1683         idev = __in6_dev_get(master);
1684         if (idev)
1685                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1686                                                    scores, hiscore_idx);
1687
1688         return hiscore_idx;
1689 }
1690
1691 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1692                        const struct in6_addr *daddr, unsigned int prefs,
1693                        struct in6_addr *saddr)
1694 {
1695         struct ipv6_saddr_score scores[2], *hiscore;
1696         struct ipv6_saddr_dst dst;
1697         struct inet6_dev *idev;
1698         struct net_device *dev;
1699         int dst_type;
1700         bool use_oif_addr = false;
1701         int hiscore_idx = 0;
1702         int ret = 0;
1703
1704         dst_type = __ipv6_addr_type(daddr);
1705         dst.addr = daddr;
1706         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1707         dst.scope = __ipv6_addr_src_scope(dst_type);
1708         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1709         dst.prefs = prefs;
1710
1711         scores[hiscore_idx].rule = -1;
1712         scores[hiscore_idx].ifa = NULL;
1713
1714         rcu_read_lock();
1715
1716         /* Candidate Source Address (section 4)
1717          *  - multicast and link-local destination address,
1718          *    the set of candidate source address MUST only
1719          *    include addresses assigned to interfaces
1720          *    belonging to the same link as the outgoing
1721          *    interface.
1722          * (- For site-local destination addresses, the
1723          *    set of candidate source addresses MUST only
1724          *    include addresses assigned to interfaces
1725          *    belonging to the same site as the outgoing
1726          *    interface.)
1727          *  - "It is RECOMMENDED that the candidate source addresses
1728          *    be the set of unicast addresses assigned to the
1729          *    interface that will be used to send to the destination
1730          *    (the 'outgoing' interface)." (RFC 6724)
1731          */
1732         if (dst_dev) {
1733                 idev = __in6_dev_get(dst_dev);
1734                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1735                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1736                     (idev && idev->cnf.use_oif_addrs_only)) {
1737                         use_oif_addr = true;
1738                 }
1739         }
1740
1741         if (use_oif_addr) {
1742                 if (idev)
1743                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1744         } else {
1745                 const struct net_device *master;
1746                 int master_idx = 0;
1747
1748                 /* if dst_dev exists and is enslaved to an L3 device, then
1749                  * prefer addresses from dst_dev and then the master over
1750                  * any other enslaved devices in the L3 domain.
1751                  */
1752                 master = l3mdev_master_dev_rcu(dst_dev);
1753                 if (master) {
1754                         master_idx = master->ifindex;
1755
1756                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1757                                                             master, &dst,
1758                                                             scores, hiscore_idx);
1759
1760                         if (scores[hiscore_idx].ifa)
1761                                 goto out;
1762                 }
1763
1764                 for_each_netdev_rcu(net, dev) {
1765                         /* only consider addresses on devices in the
1766                          * same L3 domain
1767                          */
1768                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1769                                 continue;
1770                         idev = __in6_dev_get(dev);
1771                         if (!idev)
1772                                 continue;
1773                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1774                 }
1775         }
1776
1777 out:
1778         hiscore = &scores[hiscore_idx];
1779         if (!hiscore->ifa)
1780                 ret = -EADDRNOTAVAIL;
1781         else
1782                 *saddr = hiscore->ifa->addr;
1783
1784         rcu_read_unlock();
1785         return ret;
1786 }
1787 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1788
1789 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1790                       u32 banned_flags)
1791 {
1792         struct inet6_ifaddr *ifp;
1793         int err = -EADDRNOTAVAIL;
1794
1795         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1796                 if (ifp->scope > IFA_LINK)
1797                         break;
1798                 if (ifp->scope == IFA_LINK &&
1799                     !(ifp->flags & banned_flags)) {
1800                         *addr = ifp->addr;
1801                         err = 0;
1802                         break;
1803                 }
1804         }
1805         return err;
1806 }
1807
1808 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1809                     u32 banned_flags)
1810 {
1811         struct inet6_dev *idev;
1812         int err = -EADDRNOTAVAIL;
1813
1814         rcu_read_lock();
1815         idev = __in6_dev_get(dev);
1816         if (idev) {
1817                 read_lock_bh(&idev->lock);
1818                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1819                 read_unlock_bh(&idev->lock);
1820         }
1821         rcu_read_unlock();
1822         return err;
1823 }
1824
1825 static int ipv6_count_addresses(const struct inet6_dev *idev)
1826 {
1827         const struct inet6_ifaddr *ifp;
1828         int cnt = 0;
1829
1830         rcu_read_lock();
1831         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1832                 cnt++;
1833         rcu_read_unlock();
1834         return cnt;
1835 }
1836
1837 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1838                   const struct net_device *dev, int strict)
1839 {
1840         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1841                                        strict, IFA_F_TENTATIVE);
1842 }
1843 EXPORT_SYMBOL(ipv6_chk_addr);
1844
1845 /* device argument is used to find the L3 domain of interest. If
1846  * skip_dev_check is set, then the ifp device is not checked against
1847  * the passed in dev argument. So the 2 cases for addresses checks are:
1848  *   1. does the address exist in the L3 domain that dev is part of
1849  *      (skip_dev_check = true), or
1850  *
1851  *   2. does the address exist on the specific device
1852  *      (skip_dev_check = false)
1853  */
1854 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1855                             const struct net_device *dev, bool skip_dev_check,
1856                             int strict, u32 banned_flags)
1857 {
1858         unsigned int hash = inet6_addr_hash(net, addr);
1859         const struct net_device *l3mdev;
1860         struct inet6_ifaddr *ifp;
1861         u32 ifp_flags;
1862
1863         rcu_read_lock();
1864
1865         l3mdev = l3mdev_master_dev_rcu(dev);
1866         if (skip_dev_check)
1867                 dev = NULL;
1868
1869         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1870                 if (!net_eq(dev_net(ifp->idev->dev), net))
1871                         continue;
1872
1873                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1874                         continue;
1875
1876                 /* Decouple optimistic from tentative for evaluation here.
1877                  * Ban optimistic addresses explicitly, when required.
1878                  */
1879                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1880                             ? (ifp->flags&~IFA_F_TENTATIVE)
1881                             : ifp->flags;
1882                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1883                     !(ifp_flags&banned_flags) &&
1884                     (!dev || ifp->idev->dev == dev ||
1885                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1886                         rcu_read_unlock();
1887                         return 1;
1888                 }
1889         }
1890
1891         rcu_read_unlock();
1892         return 0;
1893 }
1894 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1895
1896
1897 /* Compares an address/prefix_len with addresses on device @dev.
1898  * If one is found it returns true.
1899  */
1900 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1901         const unsigned int prefix_len, struct net_device *dev)
1902 {
1903         const struct inet6_ifaddr *ifa;
1904         const struct inet6_dev *idev;
1905         bool ret = false;
1906
1907         rcu_read_lock();
1908         idev = __in6_dev_get(dev);
1909         if (idev) {
1910                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1911                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1912                         if (ret)
1913                                 break;
1914                 }
1915         }
1916         rcu_read_unlock();
1917
1918         return ret;
1919 }
1920 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1921
1922 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1923 {
1924         const struct inet6_ifaddr *ifa;
1925         const struct inet6_dev *idev;
1926         int     onlink;
1927
1928         onlink = 0;
1929         rcu_read_lock();
1930         idev = __in6_dev_get(dev);
1931         if (idev) {
1932                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1933                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1934                                                    ifa->prefix_len);
1935                         if (onlink)
1936                                 break;
1937                 }
1938         }
1939         rcu_read_unlock();
1940         return onlink;
1941 }
1942 EXPORT_SYMBOL(ipv6_chk_prefix);
1943
1944 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1945                                      struct net_device *dev, int strict)
1946 {
1947         unsigned int hash = inet6_addr_hash(net, addr);
1948         struct inet6_ifaddr *ifp, *result = NULL;
1949
1950         rcu_read_lock();
1951         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1952                 if (!net_eq(dev_net(ifp->idev->dev), net))
1953                         continue;
1954                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1955                         if (!dev || ifp->idev->dev == dev ||
1956                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1957                                 result = ifp;
1958                                 in6_ifa_hold(ifp);
1959                                 break;
1960                         }
1961                 }
1962         }
1963         rcu_read_unlock();
1964
1965         return result;
1966 }
1967
1968 /* Gets referenced address, destroys ifaddr */
1969
1970 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1971 {
1972         if (dad_failed)
1973                 ifp->flags |= IFA_F_DADFAILED;
1974
1975         if (ifp->flags&IFA_F_TEMPORARY) {
1976                 struct inet6_ifaddr *ifpub;
1977                 spin_lock_bh(&ifp->lock);
1978                 ifpub = ifp->ifpub;
1979                 if (ifpub) {
1980                         in6_ifa_hold(ifpub);
1981                         spin_unlock_bh(&ifp->lock);
1982                         ipv6_create_tempaddr(ifpub, ifp, true);
1983                         in6_ifa_put(ifpub);
1984                 } else {
1985                         spin_unlock_bh(&ifp->lock);
1986                 }
1987                 ipv6_del_addr(ifp);
1988         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1989                 spin_lock_bh(&ifp->lock);
1990                 addrconf_del_dad_work(ifp);
1991                 ifp->flags |= IFA_F_TENTATIVE;
1992                 if (dad_failed)
1993                         ifp->flags &= ~IFA_F_OPTIMISTIC;
1994                 spin_unlock_bh(&ifp->lock);
1995                 if (dad_failed)
1996                         ipv6_ifa_notify(0, ifp);
1997                 in6_ifa_put(ifp);
1998         } else {
1999                 ipv6_del_addr(ifp);
2000         }
2001 }
2002
2003 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2004 {
2005         int err = -ENOENT;
2006
2007         spin_lock_bh(&ifp->lock);
2008         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2009                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2010                 err = 0;
2011         }
2012         spin_unlock_bh(&ifp->lock);
2013
2014         return err;
2015 }
2016
2017 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2018 {
2019         struct inet6_dev *idev = ifp->idev;
2020         struct net *net = dev_net(ifp->idev->dev);
2021
2022         if (addrconf_dad_end(ifp)) {
2023                 in6_ifa_put(ifp);
2024                 return;
2025         }
2026
2027         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2028                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2029
2030         spin_lock_bh(&ifp->lock);
2031
2032         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2033                 struct in6_addr new_addr;
2034                 struct inet6_ifaddr *ifp2;
2035                 int retries = ifp->stable_privacy_retry + 1;
2036                 struct ifa6_config cfg = {
2037                         .pfx = &new_addr,
2038                         .plen = ifp->prefix_len,
2039                         .ifa_flags = ifp->flags,
2040                         .valid_lft = ifp->valid_lft,
2041                         .preferred_lft = ifp->prefered_lft,
2042                         .scope = ifp->scope,
2043                 };
2044
2045                 if (retries > net->ipv6.sysctl.idgen_retries) {
2046                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2047                                              ifp->idev->dev->name);
2048                         goto errdad;
2049                 }
2050
2051                 new_addr = ifp->addr;
2052                 if (ipv6_generate_stable_address(&new_addr, retries,
2053                                                  idev))
2054                         goto errdad;
2055
2056                 spin_unlock_bh(&ifp->lock);
2057
2058                 if (idev->cnf.max_addresses &&
2059                     ipv6_count_addresses(idev) >=
2060                     idev->cnf.max_addresses)
2061                         goto lock_errdad;
2062
2063                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2064                                      ifp->idev->dev->name);
2065
2066                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2067                 if (IS_ERR(ifp2))
2068                         goto lock_errdad;
2069
2070                 spin_lock_bh(&ifp2->lock);
2071                 ifp2->stable_privacy_retry = retries;
2072                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2073                 spin_unlock_bh(&ifp2->lock);
2074
2075                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2076                 in6_ifa_put(ifp2);
2077 lock_errdad:
2078                 spin_lock_bh(&ifp->lock);
2079         }
2080
2081 errdad:
2082         /* transition from _POSTDAD to _ERRDAD */
2083         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2084         spin_unlock_bh(&ifp->lock);
2085
2086         addrconf_mod_dad_work(ifp, 0);
2087         in6_ifa_put(ifp);
2088 }
2089
2090 /* Join to solicited addr multicast group.
2091  * caller must hold RTNL */
2092 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2093 {
2094         struct in6_addr maddr;
2095
2096         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2097                 return;
2098
2099         addrconf_addr_solict_mult(addr, &maddr);
2100         ipv6_dev_mc_inc(dev, &maddr);
2101 }
2102
2103 /* caller must hold RTNL */
2104 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2105 {
2106         struct in6_addr maddr;
2107
2108         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2109                 return;
2110
2111         addrconf_addr_solict_mult(addr, &maddr);
2112         __ipv6_dev_mc_dec(idev, &maddr);
2113 }
2114
2115 /* caller must hold RTNL */
2116 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2117 {
2118         struct in6_addr addr;
2119
2120         if (ifp->prefix_len >= 127) /* RFC 6164 */
2121                 return;
2122         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2123         if (ipv6_addr_any(&addr))
2124                 return;
2125         __ipv6_dev_ac_inc(ifp->idev, &addr);
2126 }
2127
2128 /* caller must hold RTNL */
2129 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2130 {
2131         struct in6_addr addr;
2132
2133         if (ifp->prefix_len >= 127) /* RFC 6164 */
2134                 return;
2135         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2136         if (ipv6_addr_any(&addr))
2137                 return;
2138         __ipv6_dev_ac_dec(ifp->idev, &addr);
2139 }
2140
2141 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2142 {
2143         switch (dev->addr_len) {
2144         case ETH_ALEN:
2145                 memcpy(eui, dev->dev_addr, 3);
2146                 eui[3] = 0xFF;
2147                 eui[4] = 0xFE;
2148                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2149                 break;
2150         case EUI64_ADDR_LEN:
2151                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2152                 eui[0] ^= 2;
2153                 break;
2154         default:
2155                 return -1;
2156         }
2157
2158         return 0;
2159 }
2160
2161 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2162 {
2163         union fwnet_hwaddr *ha;
2164
2165         if (dev->addr_len != FWNET_ALEN)
2166                 return -1;
2167
2168         ha = (union fwnet_hwaddr *)dev->dev_addr;
2169
2170         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2171         eui[0] ^= 2;
2172         return 0;
2173 }
2174
2175 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2176 {
2177         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2178         if (dev->addr_len != ARCNET_ALEN)
2179                 return -1;
2180         memset(eui, 0, 7);
2181         eui[7] = *(u8 *)dev->dev_addr;
2182         return 0;
2183 }
2184
2185 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2186 {
2187         if (dev->addr_len != INFINIBAND_ALEN)
2188                 return -1;
2189         memcpy(eui, dev->dev_addr + 12, 8);
2190         eui[0] |= 2;
2191         return 0;
2192 }
2193
2194 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2195 {
2196         if (addr == 0)
2197                 return -1;
2198         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2199                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2200                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2201                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2202                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2203                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2204         eui[1] = 0;
2205         eui[2] = 0x5E;
2206         eui[3] = 0xFE;
2207         memcpy(eui + 4, &addr, 4);
2208         return 0;
2209 }
2210
2211 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2212 {
2213         if (dev->priv_flags & IFF_ISATAP)
2214                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2215         return -1;
2216 }
2217
2218 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2219 {
2220         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2221 }
2222
2223 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2224 {
2225         memcpy(eui, dev->perm_addr, 3);
2226         memcpy(eui + 5, dev->perm_addr + 3, 3);
2227         eui[3] = 0xFF;
2228         eui[4] = 0xFE;
2229         eui[0] ^= 2;
2230         return 0;
2231 }
2232
2233 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2234 {
2235         switch (dev->type) {
2236         case ARPHRD_ETHER:
2237         case ARPHRD_FDDI:
2238                 return addrconf_ifid_eui48(eui, dev);
2239         case ARPHRD_ARCNET:
2240                 return addrconf_ifid_arcnet(eui, dev);
2241         case ARPHRD_INFINIBAND:
2242                 return addrconf_ifid_infiniband(eui, dev);
2243         case ARPHRD_SIT:
2244                 return addrconf_ifid_sit(eui, dev);
2245         case ARPHRD_IPGRE:
2246         case ARPHRD_TUNNEL:
2247                 return addrconf_ifid_gre(eui, dev);
2248         case ARPHRD_6LOWPAN:
2249                 return addrconf_ifid_6lowpan(eui, dev);
2250         case ARPHRD_IEEE1394:
2251                 return addrconf_ifid_ieee1394(eui, dev);
2252         case ARPHRD_TUNNEL6:
2253         case ARPHRD_IP6GRE:
2254         case ARPHRD_RAWIP:
2255                 return addrconf_ifid_ip6tnl(eui, dev);
2256         }
2257         return -1;
2258 }
2259
2260 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2261 {
2262         int err = -1;
2263         struct inet6_ifaddr *ifp;
2264
2265         read_lock_bh(&idev->lock);
2266         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2267                 if (ifp->scope > IFA_LINK)
2268                         break;
2269                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2270                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2271                         err = 0;
2272                         break;
2273                 }
2274         }
2275         read_unlock_bh(&idev->lock);
2276         return err;
2277 }
2278
2279 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2280 static void ipv6_regen_rndid(struct inet6_dev *idev)
2281 {
2282 regen:
2283         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2284         idev->rndid[0] &= ~0x02;
2285
2286         /*
2287          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2288          * check if generated address is not inappropriate
2289          *
2290          *  - Reserved subnet anycast (RFC 2526)
2291          *      11111101 11....11 1xxxxxxx
2292          *  - ISATAP (RFC4214) 6.1
2293          *      00-00-5E-FE-xx-xx-xx-xx
2294          *  - value 0
2295          *  - XXX: already assigned to an address on the device
2296          */
2297         if (idev->rndid[0] == 0xfd &&
2298             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2299             (idev->rndid[7]&0x80))
2300                 goto regen;
2301         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2302                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2303                         goto regen;
2304                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2305                         goto regen;
2306         }
2307 }
2308
2309 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2310 {
2311         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2312                 ipv6_regen_rndid(idev);
2313 }
2314
2315 /*
2316  *      Add prefix route.
2317  */
2318
2319 static void
2320 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2321                       struct net_device *dev, unsigned long expires,
2322                       u32 flags, gfp_t gfp_flags)
2323 {
2324         struct fib6_config cfg = {
2325                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2326                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2327                 .fc_ifindex = dev->ifindex,
2328                 .fc_expires = expires,
2329                 .fc_dst_len = plen,
2330                 .fc_flags = RTF_UP | flags,
2331                 .fc_nlinfo.nl_net = dev_net(dev),
2332                 .fc_protocol = RTPROT_KERNEL,
2333                 .fc_type = RTN_UNICAST,
2334         };
2335
2336         cfg.fc_dst = *pfx;
2337
2338         /* Prevent useless cloning on PtP SIT.
2339            This thing is done here expecting that the whole
2340            class of non-broadcast devices need not cloning.
2341          */
2342 #if IS_ENABLED(CONFIG_IPV6_SIT)
2343         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2344                 cfg.fc_flags |= RTF_NONEXTHOP;
2345 #endif
2346
2347         ip6_route_add(&cfg, gfp_flags, NULL);
2348 }
2349
2350
2351 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2352                                                   int plen,
2353                                                   const struct net_device *dev,
2354                                                   u32 flags, u32 noflags)
2355 {
2356         struct fib6_node *fn;
2357         struct fib6_info *rt = NULL;
2358         struct fib6_table *table;
2359         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2360
2361         table = fib6_get_table(dev_net(dev), tb_id);
2362         if (!table)
2363                 return NULL;
2364
2365         rcu_read_lock();
2366         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2367         if (!fn)
2368                 goto out;
2369
2370         for_each_fib6_node_rt_rcu(fn) {
2371                 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2372                         continue;
2373                 if ((rt->fib6_flags & flags) != flags)
2374                         continue;
2375                 if ((rt->fib6_flags & noflags) != 0)
2376                         continue;
2377                 fib6_info_hold(rt);
2378                 break;
2379         }
2380 out:
2381         rcu_read_unlock();
2382         return rt;
2383 }
2384
2385
2386 /* Create "default" multicast route to the interface */
2387
2388 static void addrconf_add_mroute(struct net_device *dev)
2389 {
2390         struct fib6_config cfg = {
2391                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2392                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2393                 .fc_ifindex = dev->ifindex,
2394                 .fc_dst_len = 8,
2395                 .fc_flags = RTF_UP,
2396                 .fc_type = RTN_UNICAST,
2397                 .fc_nlinfo.nl_net = dev_net(dev),
2398         };
2399
2400         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2401
2402         ip6_route_add(&cfg, GFP_ATOMIC, NULL);
2403 }
2404
2405 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2406 {
2407         struct inet6_dev *idev;
2408
2409         ASSERT_RTNL();
2410
2411         idev = ipv6_find_idev(dev);
2412         if (!idev)
2413                 return ERR_PTR(-ENOBUFS);
2414
2415         if (idev->cnf.disable_ipv6)
2416                 return ERR_PTR(-EACCES);
2417
2418         /* Add default multicast route */
2419         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2420                 addrconf_add_mroute(dev);
2421
2422         return idev;
2423 }
2424
2425 static void manage_tempaddrs(struct inet6_dev *idev,
2426                              struct inet6_ifaddr *ifp,
2427                              __u32 valid_lft, __u32 prefered_lft,
2428                              bool create, unsigned long now)
2429 {
2430         u32 flags;
2431         struct inet6_ifaddr *ift;
2432
2433         read_lock_bh(&idev->lock);
2434         /* update all temporary addresses in the list */
2435         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2436                 int age, max_valid, max_prefered;
2437
2438                 if (ifp != ift->ifpub)
2439                         continue;
2440
2441                 /* RFC 4941 section 3.3:
2442                  * If a received option will extend the lifetime of a public
2443                  * address, the lifetimes of temporary addresses should
2444                  * be extended, subject to the overall constraint that no
2445                  * temporary addresses should ever remain "valid" or "preferred"
2446                  * for a time longer than (TEMP_VALID_LIFETIME) or
2447                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2448                  */
2449                 age = (now - ift->cstamp) / HZ;
2450                 max_valid = idev->cnf.temp_valid_lft - age;
2451                 if (max_valid < 0)
2452                         max_valid = 0;
2453
2454                 max_prefered = idev->cnf.temp_prefered_lft -
2455                                idev->desync_factor - age;
2456                 if (max_prefered < 0)
2457                         max_prefered = 0;
2458
2459                 if (valid_lft > max_valid)
2460                         valid_lft = max_valid;
2461
2462                 if (prefered_lft > max_prefered)
2463                         prefered_lft = max_prefered;
2464
2465                 spin_lock(&ift->lock);
2466                 flags = ift->flags;
2467                 ift->valid_lft = valid_lft;
2468                 ift->prefered_lft = prefered_lft;
2469                 ift->tstamp = now;
2470                 if (prefered_lft > 0)
2471                         ift->flags &= ~IFA_F_DEPRECATED;
2472
2473                 spin_unlock(&ift->lock);
2474                 if (!(flags&IFA_F_TENTATIVE))
2475                         ipv6_ifa_notify(0, ift);
2476         }
2477
2478         if ((create || list_empty(&idev->tempaddr_list)) &&
2479             idev->cnf.use_tempaddr > 0) {
2480                 /* When a new public address is created as described
2481                  * in [ADDRCONF], also create a new temporary address.
2482                  * Also create a temporary address if it's enabled but
2483                  * no temporary address currently exists.
2484                  */
2485                 read_unlock_bh(&idev->lock);
2486                 ipv6_create_tempaddr(ifp, NULL, false);
2487         } else {
2488                 read_unlock_bh(&idev->lock);
2489         }
2490 }
2491
2492 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2493 {
2494         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2495                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2496 }
2497
2498 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2499                                  const struct prefix_info *pinfo,
2500                                  struct inet6_dev *in6_dev,
2501                                  const struct in6_addr *addr, int addr_type,
2502                                  u32 addr_flags, bool sllao, bool tokenized,
2503                                  __u32 valid_lft, u32 prefered_lft)
2504 {
2505         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2506         int create = 0, update_lft = 0;
2507
2508         if (!ifp && valid_lft) {
2509                 int max_addresses = in6_dev->cnf.max_addresses;
2510                 struct ifa6_config cfg = {
2511                         .pfx = addr,
2512                         .plen = pinfo->prefix_len,
2513                         .ifa_flags = addr_flags,
2514                         .valid_lft = valid_lft,
2515                         .preferred_lft = prefered_lft,
2516                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2517                 };
2518
2519 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2520                 if ((net->ipv6.devconf_all->optimistic_dad ||
2521                      in6_dev->cnf.optimistic_dad) &&
2522                     !net->ipv6.devconf_all->forwarding && sllao)
2523                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2524 #endif
2525
2526                 /* Do not allow to create too much of autoconfigured
2527                  * addresses; this would be too easy way to crash kernel.
2528                  */
2529                 if (!max_addresses ||
2530                     ipv6_count_addresses(in6_dev) < max_addresses)
2531                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2532
2533                 if (IS_ERR_OR_NULL(ifp))
2534                         return -1;
2535
2536                 create = 1;
2537                 spin_lock_bh(&ifp->lock);
2538                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2539                 ifp->cstamp = jiffies;
2540                 ifp->tokenized = tokenized;
2541                 spin_unlock_bh(&ifp->lock);
2542                 addrconf_dad_start(ifp);
2543         }
2544
2545         if (ifp) {
2546                 u32 flags;
2547                 unsigned long now;
2548                 u32 stored_lft;
2549
2550                 /* update lifetime (RFC2462 5.5.3 e) */
2551                 spin_lock_bh(&ifp->lock);
2552                 now = jiffies;
2553                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2554                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2555                 else
2556                         stored_lft = 0;
2557                 if (!create && stored_lft) {
2558                         const u32 minimum_lft = min_t(u32,
2559                                 stored_lft, MIN_VALID_LIFETIME);
2560                         valid_lft = max(valid_lft, minimum_lft);
2561
2562                         /* RFC4862 Section 5.5.3e:
2563                          * "Note that the preferred lifetime of the
2564                          *  corresponding address is always reset to
2565                          *  the Preferred Lifetime in the received
2566                          *  Prefix Information option, regardless of
2567                          *  whether the valid lifetime is also reset or
2568                          *  ignored."
2569                          *
2570                          * So we should always update prefered_lft here.
2571                          */
2572                         update_lft = 1;
2573                 }
2574
2575                 if (update_lft) {
2576                         ifp->valid_lft = valid_lft;
2577                         ifp->prefered_lft = prefered_lft;
2578                         ifp->tstamp = now;
2579                         flags = ifp->flags;
2580                         ifp->flags &= ~IFA_F_DEPRECATED;
2581                         spin_unlock_bh(&ifp->lock);
2582
2583                         if (!(flags&IFA_F_TENTATIVE))
2584                                 ipv6_ifa_notify(0, ifp);
2585                 } else
2586                         spin_unlock_bh(&ifp->lock);
2587
2588                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2589                                  create, now);
2590
2591                 in6_ifa_put(ifp);
2592                 addrconf_verify();
2593         }
2594
2595         return 0;
2596 }
2597 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2598
2599 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2600 {
2601         struct prefix_info *pinfo;
2602         __u32 valid_lft;
2603         __u32 prefered_lft;
2604         int addr_type, err;
2605         u32 addr_flags = 0;
2606         struct inet6_dev *in6_dev;
2607         struct net *net = dev_net(dev);
2608
2609         pinfo = (struct prefix_info *) opt;
2610
2611         if (len < sizeof(struct prefix_info)) {
2612                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2613                 return;
2614         }
2615
2616         /*
2617          *      Validation checks ([ADDRCONF], page 19)
2618          */
2619
2620         addr_type = ipv6_addr_type(&pinfo->prefix);
2621
2622         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2623                 return;
2624
2625         valid_lft = ntohl(pinfo->valid);
2626         prefered_lft = ntohl(pinfo->prefered);
2627
2628         if (prefered_lft > valid_lft) {
2629                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2630                 return;
2631         }
2632
2633         in6_dev = in6_dev_get(dev);
2634
2635         if (!in6_dev) {
2636                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2637                                     dev->name);
2638                 return;
2639         }
2640
2641         /*
2642          *      Two things going on here:
2643          *      1) Add routes for on-link prefixes
2644          *      2) Configure prefixes with the auto flag set
2645          */
2646
2647         if (pinfo->onlink) {
2648                 struct fib6_info *rt;
2649                 unsigned long rt_expires;
2650
2651                 /* Avoid arithmetic overflow. Really, we could
2652                  * save rt_expires in seconds, likely valid_lft,
2653                  * but it would require division in fib gc, that it
2654                  * not good.
2655                  */
2656                 if (HZ > USER_HZ)
2657                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2658                 else
2659                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2660
2661                 if (addrconf_finite_timeout(rt_expires))
2662                         rt_expires *= HZ;
2663
2664                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2665                                                pinfo->prefix_len,
2666                                                dev,
2667                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2668                                                RTF_GATEWAY | RTF_DEFAULT);
2669
2670                 if (rt) {
2671                         /* Autoconf prefix route */
2672                         if (valid_lft == 0) {
2673                                 ip6_del_rt(net, rt);
2674                                 rt = NULL;
2675                         } else if (addrconf_finite_timeout(rt_expires)) {
2676                                 /* not infinity */
2677                                 fib6_set_expires(rt, jiffies + rt_expires);
2678                         } else {
2679                                 fib6_clean_expires(rt);
2680                         }
2681                 } else if (valid_lft) {
2682                         clock_t expires = 0;
2683                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2684                         if (addrconf_finite_timeout(rt_expires)) {
2685                                 /* not infinity */
2686                                 flags |= RTF_EXPIRES;
2687                                 expires = jiffies_to_clock_t(rt_expires);
2688                         }
2689                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2690                                               0, dev, expires, flags,
2691                                               GFP_ATOMIC);
2692                 }
2693                 fib6_info_release(rt);
2694         }
2695
2696         /* Try to figure out our local address for this prefix */
2697
2698         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2699                 struct in6_addr addr;
2700                 bool tokenized = false, dev_addr_generated = false;
2701
2702                 if (pinfo->prefix_len == 64) {
2703                         memcpy(&addr, &pinfo->prefix, 8);
2704
2705                         if (!ipv6_addr_any(&in6_dev->token)) {
2706                                 read_lock_bh(&in6_dev->lock);
2707                                 memcpy(addr.s6_addr + 8,
2708                                        in6_dev->token.s6_addr + 8, 8);
2709                                 read_unlock_bh(&in6_dev->lock);
2710                                 tokenized = true;
2711                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2712                                    !ipv6_generate_stable_address(&addr, 0,
2713                                                                  in6_dev)) {
2714                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2715                                 goto ok;
2716                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2717                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2718                                 goto put;
2719                         } else {
2720                                 dev_addr_generated = true;
2721                         }
2722                         goto ok;
2723                 }
2724                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2725                                     pinfo->prefix_len);
2726                 goto put;
2727
2728 ok:
2729                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2730                                                    &addr, addr_type,
2731                                                    addr_flags, sllao,
2732                                                    tokenized, valid_lft,
2733                                                    prefered_lft);
2734                 if (err)
2735                         goto put;
2736
2737                 /* Ignore error case here because previous prefix add addr was
2738                  * successful which will be notified.
2739                  */
2740                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2741                                               addr_type, addr_flags, sllao,
2742                                               tokenized, valid_lft,
2743                                               prefered_lft,
2744                                               dev_addr_generated);
2745         }
2746         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2747 put:
2748         in6_dev_put(in6_dev);
2749 }
2750
2751 /*
2752  *      Set destination address.
2753  *      Special case for SIT interfaces where we create a new "virtual"
2754  *      device.
2755  */
2756 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2757 {
2758         struct in6_ifreq ireq;
2759         struct net_device *dev;
2760         int err = -EINVAL;
2761
2762         rtnl_lock();
2763
2764         err = -EFAULT;
2765         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2766                 goto err_exit;
2767
2768         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2769
2770         err = -ENODEV;
2771         if (!dev)
2772                 goto err_exit;
2773
2774 #if IS_ENABLED(CONFIG_IPV6_SIT)
2775         if (dev->type == ARPHRD_SIT) {
2776                 const struct net_device_ops *ops = dev->netdev_ops;
2777                 struct ifreq ifr;
2778                 struct ip_tunnel_parm p;
2779
2780                 err = -EADDRNOTAVAIL;
2781                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2782                         goto err_exit;
2783
2784                 memset(&p, 0, sizeof(p));
2785                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2786                 p.iph.saddr = 0;
2787                 p.iph.version = 4;
2788                 p.iph.ihl = 5;
2789                 p.iph.protocol = IPPROTO_IPV6;
2790                 p.iph.ttl = 64;
2791                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2792
2793                 if (ops->ndo_do_ioctl) {
2794                         mm_segment_t oldfs = get_fs();
2795
2796                         set_fs(KERNEL_DS);
2797                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2798                         set_fs(oldfs);
2799                 } else
2800                         err = -EOPNOTSUPP;
2801
2802                 if (err == 0) {
2803                         err = -ENOBUFS;
2804                         dev = __dev_get_by_name(net, p.name);
2805                         if (!dev)
2806                                 goto err_exit;
2807                         err = dev_open(dev);
2808                 }
2809         }
2810 #endif
2811
2812 err_exit:
2813         rtnl_unlock();
2814         return err;
2815 }
2816
2817 static int ipv6_mc_config(struct sock *sk, bool join,
2818                           const struct in6_addr *addr, int ifindex)
2819 {
2820         int ret;
2821
2822         ASSERT_RTNL();
2823
2824         lock_sock(sk);
2825         if (join)
2826                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2827         else
2828                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2829         release_sock(sk);
2830
2831         return ret;
2832 }
2833
2834 /*
2835  *      Manual configuration of address on an interface
2836  */
2837 static int inet6_addr_add(struct net *net, int ifindex,
2838                           struct ifa6_config *cfg,
2839                           struct netlink_ext_ack *extack)
2840 {
2841         struct inet6_ifaddr *ifp;
2842         struct inet6_dev *idev;
2843         struct net_device *dev;
2844         unsigned long timeout;
2845         clock_t expires;
2846         u32 flags;
2847
2848         ASSERT_RTNL();
2849
2850         if (cfg->plen > 128)
2851                 return -EINVAL;
2852
2853         /* check the lifetime */
2854         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2855                 return -EINVAL;
2856
2857         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2858                 return -EINVAL;
2859
2860         dev = __dev_get_by_index(net, ifindex);
2861         if (!dev)
2862                 return -ENODEV;
2863
2864         idev = addrconf_add_dev(dev);
2865         if (IS_ERR(idev))
2866                 return PTR_ERR(idev);
2867
2868         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2869                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2870                                          true, cfg->pfx, ifindex);
2871
2872                 if (ret < 0)
2873                         return ret;
2874         }
2875
2876         cfg->scope = ipv6_addr_scope(cfg->pfx);
2877
2878         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2879         if (addrconf_finite_timeout(timeout)) {
2880                 expires = jiffies_to_clock_t(timeout * HZ);
2881                 cfg->valid_lft = timeout;
2882                 flags = RTF_EXPIRES;
2883         } else {
2884                 expires = 0;
2885                 flags = 0;
2886                 cfg->ifa_flags |= IFA_F_PERMANENT;
2887         }
2888
2889         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2890         if (addrconf_finite_timeout(timeout)) {
2891                 if (timeout == 0)
2892                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2893                 cfg->preferred_lft = timeout;
2894         }
2895
2896         ifp = ipv6_add_addr(idev, cfg, true, extack);
2897         if (!IS_ERR(ifp)) {
2898                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2899                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2900                                               ifp->rt_priority, dev, expires,
2901                                               flags, GFP_KERNEL);
2902                 }
2903
2904                 /* Send a netlink notification if DAD is enabled and
2905                  * optimistic flag is not set
2906                  */
2907                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2908                         ipv6_ifa_notify(0, ifp);
2909                 /*
2910                  * Note that section 3.1 of RFC 4429 indicates
2911                  * that the Optimistic flag should not be set for
2912                  * manually configured addresses
2913                  */
2914                 addrconf_dad_start(ifp);
2915                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2916                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2917                                          cfg->preferred_lft, true, jiffies);
2918                 in6_ifa_put(ifp);
2919                 addrconf_verify_rtnl();
2920                 return 0;
2921         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2922                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2923                                cfg->pfx, ifindex);
2924         }
2925
2926         return PTR_ERR(ifp);
2927 }
2928
2929 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2930                           const struct in6_addr *pfx, unsigned int plen)
2931 {
2932         struct inet6_ifaddr *ifp;
2933         struct inet6_dev *idev;
2934         struct net_device *dev;
2935
2936         if (plen > 128)
2937                 return -EINVAL;
2938
2939         dev = __dev_get_by_index(net, ifindex);
2940         if (!dev)
2941                 return -ENODEV;
2942
2943         idev = __in6_dev_get(dev);
2944         if (!idev)
2945                 return -ENXIO;
2946
2947         read_lock_bh(&idev->lock);
2948         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2949                 if (ifp->prefix_len == plen &&
2950                     ipv6_addr_equal(pfx, &ifp->addr)) {
2951                         in6_ifa_hold(ifp);
2952                         read_unlock_bh(&idev->lock);
2953
2954                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2955                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2956                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2957                                                  jiffies);
2958                         ipv6_del_addr(ifp);
2959                         addrconf_verify_rtnl();
2960                         if (ipv6_addr_is_multicast(pfx)) {
2961                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2962                                                false, pfx, dev->ifindex);
2963                         }
2964                         return 0;
2965                 }
2966         }
2967         read_unlock_bh(&idev->lock);
2968         return -EADDRNOTAVAIL;
2969 }
2970
2971
2972 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2973 {
2974         struct ifa6_config cfg = {
2975                 .ifa_flags = IFA_F_PERMANENT,
2976                 .preferred_lft = INFINITY_LIFE_TIME,
2977                 .valid_lft = INFINITY_LIFE_TIME,
2978         };
2979         struct in6_ifreq ireq;
2980         int err;
2981
2982         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2983                 return -EPERM;
2984
2985         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2986                 return -EFAULT;
2987
2988         cfg.pfx = &ireq.ifr6_addr;
2989         cfg.plen = ireq.ifr6_prefixlen;
2990
2991         rtnl_lock();
2992         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
2993         rtnl_unlock();
2994         return err;
2995 }
2996
2997 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2998 {
2999         struct in6_ifreq ireq;
3000         int err;
3001
3002         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3003                 return -EPERM;
3004
3005         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3006                 return -EFAULT;
3007
3008         rtnl_lock();
3009         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3010                              ireq.ifr6_prefixlen);
3011         rtnl_unlock();
3012         return err;
3013 }
3014
3015 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3016                      int plen, int scope)
3017 {
3018         struct inet6_ifaddr *ifp;
3019         struct ifa6_config cfg = {
3020                 .pfx = addr,
3021                 .plen = plen,
3022                 .ifa_flags = IFA_F_PERMANENT,
3023                 .valid_lft = INFINITY_LIFE_TIME,
3024                 .preferred_lft = INFINITY_LIFE_TIME,
3025                 .scope = scope
3026         };
3027
3028         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3029         if (!IS_ERR(ifp)) {
3030                 spin_lock_bh(&ifp->lock);
3031                 ifp->flags &= ~IFA_F_TENTATIVE;
3032                 spin_unlock_bh(&ifp->lock);
3033                 rt_genid_bump_ipv6(dev_net(idev->dev));
3034                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3035                 in6_ifa_put(ifp);
3036         }
3037 }
3038
3039 #if IS_ENABLED(CONFIG_IPV6_SIT)
3040 static void sit_add_v4_addrs(struct inet6_dev *idev)
3041 {
3042         struct in6_addr addr;
3043         struct net_device *dev;
3044         struct net *net = dev_net(idev->dev);
3045         int scope, plen;
3046         u32 pflags = 0;
3047
3048         ASSERT_RTNL();
3049
3050         memset(&addr, 0, sizeof(struct in6_addr));
3051         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3052
3053         if (idev->dev->flags&IFF_POINTOPOINT) {
3054                 addr.s6_addr32[0] = htonl(0xfe800000);
3055                 scope = IFA_LINK;
3056                 plen = 64;
3057         } else {
3058                 scope = IPV6_ADDR_COMPATv4;
3059                 plen = 96;
3060                 pflags |= RTF_NONEXTHOP;
3061         }
3062
3063         if (addr.s6_addr32[3]) {
3064                 add_addr(idev, &addr, plen, scope);
3065                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3066                                       GFP_ATOMIC);
3067                 return;
3068         }
3069
3070         for_each_netdev(net, dev) {
3071                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3072                 if (in_dev && (dev->flags & IFF_UP)) {
3073                         struct in_ifaddr *ifa;
3074
3075                         int flag = scope;
3076
3077                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3078
3079                                 addr.s6_addr32[3] = ifa->ifa_local;
3080
3081                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3082                                         continue;
3083                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3084                                         if (idev->dev->flags&IFF_POINTOPOINT)
3085                                                 continue;
3086                                         flag |= IFA_HOST;
3087                                 }
3088
3089                                 add_addr(idev, &addr, plen, flag);
3090                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3091                                                       0, pflags, GFP_ATOMIC);
3092                         }
3093                 }
3094         }
3095 }
3096 #endif
3097
3098 static void init_loopback(struct net_device *dev)
3099 {
3100         struct inet6_dev  *idev;
3101
3102         /* ::1 */
3103
3104         ASSERT_RTNL();
3105
3106         idev = ipv6_find_idev(dev);
3107         if (!idev) {
3108                 pr_debug("%s: add_dev failed\n", __func__);
3109                 return;
3110         }
3111
3112         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3113 }
3114
3115 void addrconf_add_linklocal(struct inet6_dev *idev,
3116                             const struct in6_addr *addr, u32 flags)
3117 {
3118         struct ifa6_config cfg = {
3119                 .pfx = addr,
3120                 .plen = 64,
3121                 .ifa_flags = flags | IFA_F_PERMANENT,
3122                 .valid_lft = INFINITY_LIFE_TIME,
3123                 .preferred_lft = INFINITY_LIFE_TIME,
3124                 .scope = IFA_LINK
3125         };
3126         struct inet6_ifaddr *ifp;
3127
3128 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3129         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3130              idev->cnf.optimistic_dad) &&
3131             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3132                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3133 #endif
3134
3135         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3136         if (!IS_ERR(ifp)) {
3137                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3138                                       0, 0, GFP_ATOMIC);
3139                 addrconf_dad_start(ifp);
3140                 in6_ifa_put(ifp);
3141         }
3142 }
3143 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3144
3145 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3146 {
3147         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3148                 return true;
3149
3150         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3151             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3152                 return true;
3153
3154         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3155             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3156                 return true;
3157
3158         return false;
3159 }
3160
3161 static int ipv6_generate_stable_address(struct in6_addr *address,
3162                                         u8 dad_count,
3163                                         const struct inet6_dev *idev)
3164 {
3165         static DEFINE_SPINLOCK(lock);
3166         static __u32 digest[SHA_DIGEST_WORDS];
3167         static __u32 workspace[SHA_WORKSPACE_WORDS];
3168
3169         static union {
3170                 char __data[SHA_MESSAGE_BYTES];
3171                 struct {
3172                         struct in6_addr secret;
3173                         __be32 prefix[2];
3174                         unsigned char hwaddr[MAX_ADDR_LEN];
3175                         u8 dad_count;
3176                 } __packed;
3177         } data;
3178
3179         struct in6_addr secret;
3180         struct in6_addr temp;
3181         struct net *net = dev_net(idev->dev);
3182
3183         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3184
3185         if (idev->cnf.stable_secret.initialized)
3186                 secret = idev->cnf.stable_secret.secret;
3187         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3188                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3189         else
3190                 return -1;
3191
3192 retry:
3193         spin_lock_bh(&lock);
3194
3195         sha_init(digest);
3196         memset(&data, 0, sizeof(data));
3197         memset(workspace, 0, sizeof(workspace));
3198         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3199         data.prefix[0] = address->s6_addr32[0];
3200         data.prefix[1] = address->s6_addr32[1];
3201         data.secret = secret;
3202         data.dad_count = dad_count;
3203
3204         sha_transform(digest, data.__data, workspace);
3205
3206         temp = *address;
3207         temp.s6_addr32[2] = (__force __be32)digest[0];
3208         temp.s6_addr32[3] = (__force __be32)digest[1];
3209
3210         spin_unlock_bh(&lock);
3211
3212         if (ipv6_reserved_interfaceid(temp)) {
3213                 dad_count++;
3214                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3215                         return -1;
3216                 goto retry;
3217         }
3218
3219         *address = temp;
3220         return 0;
3221 }
3222
3223 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3224 {
3225         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3226
3227         if (s->initialized)
3228                 return;
3229         s = &idev->cnf.stable_secret;
3230         get_random_bytes(&s->secret, sizeof(s->secret));
3231         s->initialized = true;
3232 }
3233
3234 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3235 {
3236         struct in6_addr addr;
3237
3238         /* no link local addresses on L3 master devices */
3239         if (netif_is_l3_master(idev->dev))
3240                 return;
3241
3242         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3243
3244         switch (idev->cnf.addr_gen_mode) {
3245         case IN6_ADDR_GEN_MODE_RANDOM:
3246                 ipv6_gen_mode_random_init(idev);
3247                 /* fallthrough */
3248         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3249                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3250                         addrconf_add_linklocal(idev, &addr,
3251                                                IFA_F_STABLE_PRIVACY);
3252                 else if (prefix_route)
3253                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3254                                               0, 0, GFP_KERNEL);
3255                 break;
3256         case IN6_ADDR_GEN_MODE_EUI64:
3257                 /* addrconf_add_linklocal also adds a prefix_route and we
3258                  * only need to care about prefix routes if ipv6_generate_eui64
3259                  * couldn't generate one.
3260                  */
3261                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3262                         addrconf_add_linklocal(idev, &addr, 0);
3263                 else if (prefix_route)
3264                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3265                                               0, 0, GFP_KERNEL);
3266                 break;
3267         case IN6_ADDR_GEN_MODE_NONE:
3268         default:
3269                 /* will not add any link local address */
3270                 break;
3271         }
3272 }
3273
3274 static void addrconf_dev_config(struct net_device *dev)
3275 {
3276         struct inet6_dev *idev;
3277
3278         ASSERT_RTNL();
3279
3280         if ((dev->type != ARPHRD_ETHER) &&
3281             (dev->type != ARPHRD_FDDI) &&
3282             (dev->type != ARPHRD_ARCNET) &&
3283             (dev->type != ARPHRD_INFINIBAND) &&
3284             (dev->type != ARPHRD_IEEE1394) &&
3285             (dev->type != ARPHRD_TUNNEL6) &&
3286             (dev->type != ARPHRD_6LOWPAN) &&
3287             (dev->type != ARPHRD_IP6GRE) &&
3288             (dev->type != ARPHRD_IPGRE) &&
3289             (dev->type != ARPHRD_TUNNEL) &&
3290             (dev->type != ARPHRD_NONE) &&
3291             (dev->type != ARPHRD_RAWIP)) {
3292                 /* Alas, we support only Ethernet autoconfiguration. */
3293                 return;
3294         }
3295
3296         idev = addrconf_add_dev(dev);
3297         if (IS_ERR(idev))
3298                 return;
3299
3300         /* this device type has no EUI support */
3301         if (dev->type == ARPHRD_NONE &&
3302             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3303                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3304
3305         addrconf_addr_gen(idev, false);
3306 }
3307
3308 #if IS_ENABLED(CONFIG_IPV6_SIT)
3309 static void addrconf_sit_config(struct net_device *dev)
3310 {
3311         struct inet6_dev *idev;
3312
3313         ASSERT_RTNL();
3314
3315         /*
3316          * Configure the tunnel with one of our IPv4
3317          * addresses... we should configure all of
3318          * our v4 addrs in the tunnel
3319          */
3320
3321         idev = ipv6_find_idev(dev);
3322         if (!idev) {
3323                 pr_debug("%s: add_dev failed\n", __func__);
3324                 return;
3325         }
3326
3327         if (dev->priv_flags & IFF_ISATAP) {
3328                 addrconf_addr_gen(idev, false);
3329                 return;
3330         }
3331
3332         sit_add_v4_addrs(idev);
3333
3334         if (dev->flags&IFF_POINTOPOINT)
3335                 addrconf_add_mroute(dev);
3336 }
3337 #endif
3338
3339 #if IS_ENABLED(CONFIG_NET_IPGRE)
3340 static void addrconf_gre_config(struct net_device *dev)
3341 {
3342         struct inet6_dev *idev;
3343
3344         ASSERT_RTNL();
3345
3346         idev = ipv6_find_idev(dev);
3347         if (!idev) {
3348                 pr_debug("%s: add_dev failed\n", __func__);
3349                 return;
3350         }
3351
3352         addrconf_addr_gen(idev, true);
3353         if (dev->flags & IFF_POINTOPOINT)
3354                 addrconf_add_mroute(dev);
3355 }
3356 #endif
3357
3358 static int fixup_permanent_addr(struct net *net,
3359                                 struct inet6_dev *idev,
3360                                 struct inet6_ifaddr *ifp)
3361 {
3362         /* !fib6_node means the host route was removed from the
3363          * FIB, for example, if 'lo' device is taken down. In that
3364          * case regenerate the host route.
3365          */
3366         if (!ifp->rt || !ifp->rt->fib6_node) {
3367                 struct fib6_info *f6i, *prev;
3368
3369                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3370                                          GFP_ATOMIC);
3371                 if (IS_ERR(f6i))
3372                         return PTR_ERR(f6i);
3373
3374                 /* ifp->rt can be accessed outside of rtnl */
3375                 spin_lock(&ifp->lock);
3376                 prev = ifp->rt;
3377                 ifp->rt = f6i;
3378                 spin_unlock(&ifp->lock);
3379
3380                 fib6_info_release(prev);
3381         }
3382
3383         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3384                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3385                                       ifp->rt_priority, idev->dev, 0, 0,
3386                                       GFP_ATOMIC);
3387         }
3388
3389         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3390                 addrconf_dad_start(ifp);
3391
3392         return 0;
3393 }
3394
3395 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3396 {
3397         struct inet6_ifaddr *ifp, *tmp;
3398         struct inet6_dev *idev;
3399
3400         idev = __in6_dev_get(dev);
3401         if (!idev)
3402                 return;
3403
3404         write_lock_bh(&idev->lock);
3405
3406         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3407                 if ((ifp->flags & IFA_F_PERMANENT) &&
3408                     fixup_permanent_addr(net, idev, ifp) < 0) {
3409                         write_unlock_bh(&idev->lock);
3410                         in6_ifa_hold(ifp);
3411                         ipv6_del_addr(ifp);
3412                         write_lock_bh(&idev->lock);
3413
3414                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3415                                              idev->dev->name, &ifp->addr);
3416                 }
3417         }
3418
3419         write_unlock_bh(&idev->lock);
3420 }
3421
3422 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3423                            void *ptr)
3424 {
3425         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3426         struct netdev_notifier_changeupper_info *info;
3427         struct inet6_dev *idev = __in6_dev_get(dev);
3428         struct net *net = dev_net(dev);
3429         int run_pending = 0;
3430         int err;
3431
3432         switch (event) {
3433         case NETDEV_REGISTER:
3434                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3435                         idev = ipv6_add_dev(dev);
3436                         if (IS_ERR(idev))
3437                                 return notifier_from_errno(PTR_ERR(idev));
3438                 }
3439                 break;
3440
3441         case NETDEV_CHANGEMTU:
3442                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3443                 if (dev->mtu < IPV6_MIN_MTU) {
3444                         addrconf_ifdown(dev, dev != net->loopback_dev);
3445                         break;
3446                 }
3447
3448                 if (idev) {
3449                         rt6_mtu_change(dev, dev->mtu);
3450                         idev->cnf.mtu6 = dev->mtu;
3451                         break;
3452                 }
3453
3454                 /* allocate new idev */
3455                 idev = ipv6_add_dev(dev);
3456                 if (IS_ERR(idev))
3457                         break;
3458
3459                 /* device is still not ready */
3460                 if (!(idev->if_flags & IF_READY))
3461                         break;
3462
3463                 run_pending = 1;
3464
3465                 /* fall through */
3466
3467         case NETDEV_UP:
3468         case NETDEV_CHANGE:
3469                 if (dev->flags & IFF_SLAVE)
3470                         break;
3471
3472                 if (idev && idev->cnf.disable_ipv6)
3473                         break;
3474
3475                 if (event == NETDEV_UP) {
3476                         /* restore routes for permanent addresses */
3477                         addrconf_permanent_addr(net, dev);
3478
3479                         if (!addrconf_link_ready(dev)) {
3480                                 /* device is not ready yet. */
3481                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3482                                         dev->name);
3483                                 break;
3484                         }
3485
3486                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3487                                 idev = ipv6_add_dev(dev);
3488
3489                         if (!IS_ERR_OR_NULL(idev)) {
3490                                 idev->if_flags |= IF_READY;
3491                                 run_pending = 1;
3492                         }
3493                 } else if (event == NETDEV_CHANGE) {
3494                         if (!addrconf_link_ready(dev)) {
3495                                 /* device is still not ready. */
3496                                 rt6_sync_down_dev(dev, event);
3497                                 break;
3498                         }
3499
3500                         if (idev) {
3501                                 if (idev->if_flags & IF_READY) {
3502                                         /* device is already configured -
3503                                          * but resend MLD reports, we might
3504                                          * have roamed and need to update
3505                                          * multicast snooping switches
3506                                          */
3507                                         ipv6_mc_up(idev);
3508                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3509                                         break;
3510                                 }
3511                                 idev->if_flags |= IF_READY;
3512                         }
3513
3514                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3515                                 dev->name);
3516
3517                         run_pending = 1;
3518                 }
3519
3520                 switch (dev->type) {
3521 #if IS_ENABLED(CONFIG_IPV6_SIT)
3522                 case ARPHRD_SIT:
3523                         addrconf_sit_config(dev);
3524                         break;
3525 #endif
3526 #if IS_ENABLED(CONFIG_NET_IPGRE)
3527                 case ARPHRD_IPGRE:
3528                         addrconf_gre_config(dev);
3529                         break;
3530 #endif
3531                 case ARPHRD_LOOPBACK:
3532                         init_loopback(dev);
3533                         break;
3534
3535                 default:
3536                         addrconf_dev_config(dev);
3537                         break;
3538                 }
3539
3540                 if (!IS_ERR_OR_NULL(idev)) {
3541                         if (run_pending)
3542                                 addrconf_dad_run(idev);
3543
3544                         /* Device has an address by now */
3545                         rt6_sync_up(dev, RTNH_F_DEAD);
3546
3547                         /*
3548                          * If the MTU changed during the interface down,
3549                          * when the interface up, the changed MTU must be
3550                          * reflected in the idev as well as routers.
3551                          */
3552                         if (idev->cnf.mtu6 != dev->mtu &&
3553                             dev->mtu >= IPV6_MIN_MTU) {
3554                                 rt6_mtu_change(dev, dev->mtu);
3555                                 idev->cnf.mtu6 = dev->mtu;
3556                         }
3557                         idev->tstamp = jiffies;
3558                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3559
3560                         /*
3561                          * If the changed mtu during down is lower than
3562                          * IPV6_MIN_MTU stop IPv6 on this interface.
3563                          */
3564                         if (dev->mtu < IPV6_MIN_MTU)
3565                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3566                 }
3567                 break;
3568
3569         case NETDEV_DOWN:
3570         case NETDEV_UNREGISTER:
3571                 /*
3572                  *      Remove all addresses from this interface.
3573                  */
3574                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3575                 break;
3576
3577         case NETDEV_CHANGENAME:
3578                 if (idev) {
3579                         snmp6_unregister_dev(idev);
3580                         addrconf_sysctl_unregister(idev);
3581                         err = addrconf_sysctl_register(idev);
3582                         if (err)
3583                                 return notifier_from_errno(err);
3584                         err = snmp6_register_dev(idev);
3585                         if (err) {
3586                                 addrconf_sysctl_unregister(idev);
3587                                 return notifier_from_errno(err);
3588                         }
3589                 }
3590                 break;
3591
3592         case NETDEV_PRE_TYPE_CHANGE:
3593         case NETDEV_POST_TYPE_CHANGE:
3594                 if (idev)
3595                         addrconf_type_change(dev, event);
3596                 break;
3597
3598         case NETDEV_CHANGEUPPER:
3599                 info = ptr;
3600
3601                 /* flush all routes if dev is linked to or unlinked from
3602                  * an L3 master device (e.g., VRF)
3603                  */
3604                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3605                         addrconf_ifdown(dev, 0);
3606         }
3607
3608         return NOTIFY_OK;
3609 }
3610
3611 /*
3612  *      addrconf module should be notified of a device going up
3613  */
3614 static struct notifier_block ipv6_dev_notf = {
3615         .notifier_call = addrconf_notify,
3616         .priority = ADDRCONF_NOTIFY_PRIORITY,
3617 };
3618
3619 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3620 {
3621         struct inet6_dev *idev;
3622         ASSERT_RTNL();
3623
3624         idev = __in6_dev_get(dev);
3625
3626         if (event == NETDEV_POST_TYPE_CHANGE)
3627                 ipv6_mc_remap(idev);
3628         else if (event == NETDEV_PRE_TYPE_CHANGE)
3629                 ipv6_mc_unmap(idev);
3630 }
3631
3632 static bool addr_is_local(const struct in6_addr *addr)
3633 {
3634         return ipv6_addr_type(addr) &
3635                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3636 }
3637
3638 static int addrconf_ifdown(struct net_device *dev, int how)
3639 {
3640         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3641         struct net *net = dev_net(dev);
3642         struct inet6_dev *idev;
3643         struct inet6_ifaddr *ifa, *tmp;
3644         bool keep_addr = false;
3645         int state, i;
3646
3647         ASSERT_RTNL();
3648
3649         rt6_disable_ip(dev, event);
3650
3651         idev = __in6_dev_get(dev);
3652         if (!idev)
3653                 return -ENODEV;
3654
3655         /*
3656          * Step 1: remove reference to ipv6 device from parent device.
3657          *         Do not dev_put!
3658          */
3659         if (how) {
3660                 idev->dead = 1;
3661
3662                 /* protected by rtnl_lock */
3663                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3664
3665                 /* Step 1.5: remove snmp6 entry */
3666                 snmp6_unregister_dev(idev);
3667
3668         }
3669
3670         /* combine the user config with event to determine if permanent
3671          * addresses are to be removed from address hash table
3672          */
3673         if (!how && !idev->cnf.disable_ipv6) {
3674                 /* aggregate the system setting and interface setting */
3675                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3676
3677                 if (!_keep_addr)
3678                         _keep_addr = idev->cnf.keep_addr_on_down;
3679
3680                 keep_addr = (_keep_addr > 0);
3681         }
3682
3683         /* Step 2: clear hash table */
3684         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3685                 struct hlist_head *h = &inet6_addr_lst[i];
3686
3687                 spin_lock_bh(&addrconf_hash_lock);
3688 restart:
3689                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3690                         if (ifa->idev == idev) {
3691                                 addrconf_del_dad_work(ifa);
3692                                 /* combined flag + permanent flag decide if
3693                                  * address is retained on a down event
3694                                  */
3695                                 if (!keep_addr ||
3696                                     !(ifa->flags & IFA_F_PERMANENT) ||
3697                                     addr_is_local(&ifa->addr)) {
3698                                         hlist_del_init_rcu(&ifa->addr_lst);
3699                                         goto restart;
3700                                 }
3701                         }
3702                 }
3703                 spin_unlock_bh(&addrconf_hash_lock);
3704         }
3705
3706         write_lock_bh(&idev->lock);
3707
3708         addrconf_del_rs_timer(idev);
3709
3710         /* Step 2: clear flags for stateless addrconf */
3711         if (!how)
3712                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3713
3714         /* Step 3: clear tempaddr list */
3715         while (!list_empty(&idev->tempaddr_list)) {
3716                 ifa = list_first_entry(&idev->tempaddr_list,
3717                                        struct inet6_ifaddr, tmp_list);
3718                 list_del(&ifa->tmp_list);
3719                 write_unlock_bh(&idev->lock);
3720                 spin_lock_bh(&ifa->lock);
3721
3722                 if (ifa->ifpub) {
3723                         in6_ifa_put(ifa->ifpub);
3724                         ifa->ifpub = NULL;
3725                 }
3726                 spin_unlock_bh(&ifa->lock);
3727                 in6_ifa_put(ifa);
3728                 write_lock_bh(&idev->lock);
3729         }
3730
3731         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3732                 struct fib6_info *rt = NULL;
3733                 bool keep;
3734
3735                 addrconf_del_dad_work(ifa);
3736
3737                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3738                         !addr_is_local(&ifa->addr);
3739
3740                 write_unlock_bh(&idev->lock);
3741                 spin_lock_bh(&ifa->lock);
3742
3743                 if (keep) {
3744                         /* set state to skip the notifier below */
3745                         state = INET6_IFADDR_STATE_DEAD;
3746                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3747                         if (!(ifa->flags & IFA_F_NODAD))
3748                                 ifa->flags |= IFA_F_TENTATIVE;
3749
3750                         rt = ifa->rt;
3751                         ifa->rt = NULL;
3752                 } else {
3753                         state = ifa->state;
3754                         ifa->state = INET6_IFADDR_STATE_DEAD;
3755                 }
3756
3757                 spin_unlock_bh(&ifa->lock);
3758
3759                 if (rt)
3760                         ip6_del_rt(net, rt);
3761
3762                 if (state != INET6_IFADDR_STATE_DEAD) {
3763                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3764                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3765                 } else {
3766                         if (idev->cnf.forwarding)
3767                                 addrconf_leave_anycast(ifa);
3768                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3769                 }
3770
3771                 write_lock_bh(&idev->lock);
3772                 if (!keep) {
3773                         list_del_rcu(&ifa->if_list);
3774                         in6_ifa_put(ifa);
3775                 }
3776         }
3777
3778         write_unlock_bh(&idev->lock);
3779
3780         /* Step 5: Discard anycast and multicast list */
3781         if (how) {
3782                 ipv6_ac_destroy_dev(idev);
3783                 ipv6_mc_destroy_dev(idev);
3784         } else {
3785                 ipv6_mc_down(idev);
3786         }
3787
3788         idev->tstamp = jiffies;
3789
3790         /* Last: Shot the device (if unregistered) */
3791         if (how) {
3792                 addrconf_sysctl_unregister(idev);
3793                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3794                 neigh_ifdown(&nd_tbl, dev);
3795                 in6_dev_put(idev);
3796         }
3797         return 0;
3798 }
3799
3800 static void addrconf_rs_timer(struct timer_list *t)
3801 {
3802         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3803         struct net_device *dev = idev->dev;
3804         struct in6_addr lladdr;
3805
3806         write_lock(&idev->lock);
3807         if (idev->dead || !(idev->if_flags & IF_READY))
3808                 goto out;
3809
3810         if (!ipv6_accept_ra(idev))
3811                 goto out;
3812
3813         /* Announcement received after solicitation was sent */
3814         if (idev->if_flags & IF_RA_RCVD)
3815                 goto out;
3816
3817         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3818                 write_unlock(&idev->lock);
3819                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3820                         ndisc_send_rs(dev, &lladdr,
3821                                       &in6addr_linklocal_allrouters);
3822                 else
3823                         goto put;
3824
3825                 write_lock(&idev->lock);
3826                 idev->rs_interval = rfc3315_s14_backoff_update(
3827                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3828                 /* The wait after the last probe can be shorter */
3829                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3830                                              idev->cnf.rtr_solicits) ?
3831                                       idev->cnf.rtr_solicit_delay :
3832                                       idev->rs_interval);
3833         } else {
3834                 /*
3835                  * Note: we do not support deprecated "all on-link"
3836                  * assumption any longer.
3837                  */
3838                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3839         }
3840
3841 out:
3842         write_unlock(&idev->lock);
3843 put:
3844         in6_dev_put(idev);
3845 }
3846
3847 /*
3848  *      Duplicate Address Detection
3849  */
3850 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3851 {
3852         unsigned long rand_num;
3853         struct inet6_dev *idev = ifp->idev;
3854         u64 nonce;
3855
3856         if (ifp->flags & IFA_F_OPTIMISTIC)
3857                 rand_num = 0;
3858         else
3859                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3860
3861         nonce = 0;
3862         if (idev->cnf.enhanced_dad ||
3863             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3864                 do
3865                         get_random_bytes(&nonce, 6);
3866                 while (nonce == 0);
3867         }
3868         ifp->dad_nonce = nonce;
3869         ifp->dad_probes = idev->cnf.dad_transmits;
3870         addrconf_mod_dad_work(ifp, rand_num);
3871 }
3872
3873 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3874 {
3875         struct inet6_dev *idev = ifp->idev;
3876         struct net_device *dev = idev->dev;
3877         bool bump_id, notify = false;
3878         struct net *net;
3879
3880         addrconf_join_solict(dev, &ifp->addr);
3881
3882         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3883
3884         read_lock_bh(&idev->lock);
3885         spin_lock(&ifp->lock);
3886         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3887                 goto out;
3888
3889         net = dev_net(dev);
3890         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3891             (net->ipv6.devconf_all->accept_dad < 1 &&
3892              idev->cnf.accept_dad < 1) ||
3893             !(ifp->flags&IFA_F_TENTATIVE) ||
3894             ifp->flags & IFA_F_NODAD) {
3895                 bool send_na = false;
3896
3897                 if (ifp->flags & IFA_F_TENTATIVE &&
3898                     !(ifp->flags & IFA_F_OPTIMISTIC))
3899                         send_na = true;
3900                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3901                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3902                 spin_unlock(&ifp->lock);
3903                 read_unlock_bh(&idev->lock);
3904
3905                 addrconf_dad_completed(ifp, bump_id, send_na);
3906                 return;
3907         }
3908
3909         if (!(idev->if_flags & IF_READY)) {
3910                 spin_unlock(&ifp->lock);
3911                 read_unlock_bh(&idev->lock);
3912                 /*
3913                  * If the device is not ready:
3914                  * - keep it tentative if it is a permanent address.
3915                  * - otherwise, kill it.
3916                  */
3917                 in6_ifa_hold(ifp);
3918                 addrconf_dad_stop(ifp, 0);
3919                 return;
3920         }
3921
3922         /*
3923          * Optimistic nodes can start receiving
3924          * Frames right away
3925          */
3926         if (ifp->flags & IFA_F_OPTIMISTIC) {
3927                 ip6_ins_rt(net, ifp->rt);
3928                 if (ipv6_use_optimistic_addr(net, idev)) {
3929                         /* Because optimistic nodes can use this address,
3930                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3931                          */
3932                         notify = true;
3933                 }
3934         }
3935
3936         addrconf_dad_kick(ifp);
3937 out:
3938         spin_unlock(&ifp->lock);
3939         read_unlock_bh(&idev->lock);
3940         if (notify)
3941                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3942 }
3943
3944 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3945 {
3946         bool begin_dad = false;
3947
3948         spin_lock_bh(&ifp->lock);
3949         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3950                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3951                 begin_dad = true;
3952         }
3953         spin_unlock_bh(&ifp->lock);
3954
3955         if (begin_dad)
3956                 addrconf_mod_dad_work(ifp, 0);
3957 }
3958
3959 static void addrconf_dad_work(struct work_struct *w)
3960 {
3961         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3962                                                 struct inet6_ifaddr,
3963                                                 dad_work);
3964         struct inet6_dev *idev = ifp->idev;
3965         bool bump_id, disable_ipv6 = false;
3966         struct in6_addr mcaddr;
3967
3968         enum {
3969                 DAD_PROCESS,
3970                 DAD_BEGIN,
3971                 DAD_ABORT,
3972         } action = DAD_PROCESS;
3973
3974         rtnl_lock();
3975
3976         spin_lock_bh(&ifp->lock);
3977         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3978                 action = DAD_BEGIN;
3979                 ifp->state = INET6_IFADDR_STATE_DAD;
3980         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3981                 action = DAD_ABORT;
3982                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3983
3984                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3985                      idev->cnf.accept_dad > 1) &&
3986                     !idev->cnf.disable_ipv6 &&
3987                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3988                         struct in6_addr addr;
3989
3990                         addr.s6_addr32[0] = htonl(0xfe800000);
3991                         addr.s6_addr32[1] = 0;
3992
3993                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3994                             ipv6_addr_equal(&ifp->addr, &addr)) {
3995                                 /* DAD failed for link-local based on MAC */
3996                                 idev->cnf.disable_ipv6 = 1;
3997
3998                                 pr_info("%s: IPv6 being disabled!\n",
3999                                         ifp->idev->dev->name);
4000                                 disable_ipv6 = true;
4001                         }
4002                 }
4003         }
4004         spin_unlock_bh(&ifp->lock);
4005
4006         if (action == DAD_BEGIN) {
4007                 addrconf_dad_begin(ifp);
4008                 goto out;
4009         } else if (action == DAD_ABORT) {
4010                 in6_ifa_hold(ifp);
4011                 addrconf_dad_stop(ifp, 1);
4012                 if (disable_ipv6)
4013                         addrconf_ifdown(idev->dev, 0);
4014                 goto out;
4015         }
4016
4017         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4018                 goto out;
4019
4020         write_lock_bh(&idev->lock);
4021         if (idev->dead || !(idev->if_flags & IF_READY)) {
4022                 write_unlock_bh(&idev->lock);
4023                 goto out;
4024         }
4025
4026         spin_lock(&ifp->lock);
4027         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4028                 spin_unlock(&ifp->lock);
4029                 write_unlock_bh(&idev->lock);
4030                 goto out;
4031         }
4032
4033         if (ifp->dad_probes == 0) {
4034                 bool send_na = false;
4035
4036                 /*
4037                  * DAD was successful
4038                  */
4039
4040                 if (ifp->flags & IFA_F_TENTATIVE &&
4041                     !(ifp->flags & IFA_F_OPTIMISTIC))
4042                         send_na = true;
4043                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4044                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4045                 spin_unlock(&ifp->lock);
4046                 write_unlock_bh(&idev->lock);
4047
4048                 addrconf_dad_completed(ifp, bump_id, send_na);
4049
4050                 goto out;
4051         }
4052
4053         ifp->dad_probes--;
4054         addrconf_mod_dad_work(ifp,
4055                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4056         spin_unlock(&ifp->lock);
4057         write_unlock_bh(&idev->lock);
4058
4059         /* send a neighbour solicitation for our addr */
4060         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4061         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4062                       ifp->dad_nonce);
4063 out:
4064         in6_ifa_put(ifp);
4065         rtnl_unlock();
4066 }
4067
4068 /* ifp->idev must be at least read locked */
4069 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4070 {
4071         struct inet6_ifaddr *ifpiter;
4072         struct inet6_dev *idev = ifp->idev;
4073
4074         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4075                 if (ifpiter->scope > IFA_LINK)
4076                         break;
4077                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4078                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4079                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4080                     IFA_F_PERMANENT)
4081                         return false;
4082         }
4083         return true;
4084 }
4085
4086 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4087                                    bool send_na)
4088 {
4089         struct net_device *dev = ifp->idev->dev;
4090         struct in6_addr lladdr;
4091         bool send_rs, send_mld;
4092
4093         addrconf_del_dad_work(ifp);
4094
4095         /*
4096          *      Configure the address for reception. Now it is valid.
4097          */
4098
4099         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4100
4101         /* If added prefix is link local and we are prepared to process
4102            router advertisements, start sending router solicitations.
4103          */
4104
4105         read_lock_bh(&ifp->idev->lock);
4106         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4107         send_rs = send_mld &&
4108                   ipv6_accept_ra(ifp->idev) &&
4109                   ifp->idev->cnf.rtr_solicits != 0 &&
4110                   (dev->flags&IFF_LOOPBACK) == 0;
4111         read_unlock_bh(&ifp->idev->lock);
4112
4113         /* While dad is in progress mld report's source address is in6_addrany.
4114          * Resend with proper ll now.
4115          */
4116         if (send_mld)
4117                 ipv6_mc_dad_complete(ifp->idev);
4118
4119         /* send unsolicited NA if enabled */
4120         if (send_na &&
4121             (ifp->idev->cnf.ndisc_notify ||
4122              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4123                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4124                               /*router=*/ !!ifp->idev->cnf.forwarding,
4125                               /*solicited=*/ false, /*override=*/ true,
4126                               /*inc_opt=*/ true);
4127         }
4128
4129         if (send_rs) {
4130                 /*
4131                  *      If a host as already performed a random delay
4132                  *      [...] as part of DAD [...] there is no need
4133                  *      to delay again before sending the first RS
4134                  */
4135                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4136                         return;
4137                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4138
4139                 write_lock_bh(&ifp->idev->lock);
4140                 spin_lock(&ifp->lock);
4141                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4142                         ifp->idev->cnf.rtr_solicit_interval);
4143                 ifp->idev->rs_probes = 1;
4144                 ifp->idev->if_flags |= IF_RS_SENT;
4145                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4146                 spin_unlock(&ifp->lock);
4147                 write_unlock_bh(&ifp->idev->lock);
4148         }
4149
4150         if (bump_id)
4151                 rt_genid_bump_ipv6(dev_net(dev));
4152
4153         /* Make sure that a new temporary address will be created
4154          * before this temporary address becomes deprecated.
4155          */
4156         if (ifp->flags & IFA_F_TEMPORARY)
4157                 addrconf_verify_rtnl();
4158 }
4159
4160 static void addrconf_dad_run(struct inet6_dev *idev)
4161 {
4162         struct inet6_ifaddr *ifp;
4163
4164         read_lock_bh(&idev->lock);
4165         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4166                 spin_lock(&ifp->lock);
4167                 if (ifp->flags & IFA_F_TENTATIVE &&
4168                     ifp->state == INET6_IFADDR_STATE_DAD)
4169                         addrconf_dad_kick(ifp);
4170                 spin_unlock(&ifp->lock);
4171         }
4172         read_unlock_bh(&idev->lock);
4173 }
4174
4175 #ifdef CONFIG_PROC_FS
4176 struct if6_iter_state {
4177         struct seq_net_private p;
4178         int bucket;
4179         int offset;
4180 };
4181
4182 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4183 {
4184         struct if6_iter_state *state = seq->private;
4185         struct net *net = seq_file_net(seq);
4186         struct inet6_ifaddr *ifa = NULL;
4187         int p = 0;
4188
4189         /* initial bucket if pos is 0 */
4190         if (pos == 0) {
4191                 state->bucket = 0;
4192                 state->offset = 0;
4193         }
4194
4195         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4196                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4197                                          addr_lst) {
4198                         if (!net_eq(dev_net(ifa->idev->dev), net))
4199                                 continue;
4200                         /* sync with offset */
4201                         if (p < state->offset) {
4202                                 p++;
4203                                 continue;
4204                         }
4205                         state->offset++;
4206                         return ifa;
4207                 }
4208
4209                 /* prepare for next bucket */
4210                 state->offset = 0;
4211                 p = 0;
4212         }
4213         return NULL;
4214 }
4215
4216 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4217                                          struct inet6_ifaddr *ifa)
4218 {
4219         struct if6_iter_state *state = seq->private;
4220         struct net *net = seq_file_net(seq);
4221
4222         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4223                 if (!net_eq(dev_net(ifa->idev->dev), net))
4224                         continue;
4225                 state->offset++;
4226                 return ifa;
4227         }
4228
4229         while (++state->bucket < IN6_ADDR_HSIZE) {
4230                 state->offset = 0;
4231                 hlist_for_each_entry_rcu(ifa,
4232                                      &inet6_addr_lst[state->bucket], addr_lst) {
4233                         if (!net_eq(dev_net(ifa->idev->dev), net))
4234                                 continue;
4235                         state->offset++;
4236                         return ifa;
4237                 }
4238         }
4239
4240         return NULL;
4241 }
4242
4243 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4244         __acquires(rcu)
4245 {
4246         rcu_read_lock();
4247         return if6_get_first(seq, *pos);
4248 }
4249
4250 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4251 {
4252         struct inet6_ifaddr *ifa;
4253
4254         ifa = if6_get_next(seq, v);
4255         ++*pos;
4256         return ifa;
4257 }
4258
4259 static void if6_seq_stop(struct seq_file *seq, void *v)
4260         __releases(rcu)
4261 {
4262         rcu_read_unlock();
4263 }
4264
4265 static int if6_seq_show(struct seq_file *seq, void *v)
4266 {
4267         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4268         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4269                    &ifp->addr,
4270                    ifp->idev->dev->ifindex,
4271                    ifp->prefix_len,
4272                    ifp->scope,
4273                    (u8) ifp->flags,
4274                    ifp->idev->dev->name);
4275         return 0;
4276 }
4277
4278 static const struct seq_operations if6_seq_ops = {
4279         .start  = if6_seq_start,
4280         .next   = if6_seq_next,
4281         .show   = if6_seq_show,
4282         .stop   = if6_seq_stop,
4283 };
4284
4285 static int __net_init if6_proc_net_init(struct net *net)
4286 {
4287         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4288                         sizeof(struct if6_iter_state)))
4289                 return -ENOMEM;
4290         return 0;
4291 }
4292
4293 static void __net_exit if6_proc_net_exit(struct net *net)
4294 {
4295         remove_proc_entry("if_inet6", net->proc_net);
4296 }
4297
4298 static struct pernet_operations if6_proc_net_ops = {
4299         .init = if6_proc_net_init,
4300         .exit = if6_proc_net_exit,
4301 };
4302
4303 int __init if6_proc_init(void)
4304 {
4305         return register_pernet_subsys(&if6_proc_net_ops);
4306 }
4307
4308 void if6_proc_exit(void)
4309 {
4310         unregister_pernet_subsys(&if6_proc_net_ops);
4311 }
4312 #endif  /* CONFIG_PROC_FS */
4313
4314 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4315 /* Check if address is a home address configured on any interface. */
4316 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4317 {
4318         unsigned int hash = inet6_addr_hash(net, addr);
4319         struct inet6_ifaddr *ifp = NULL;
4320         int ret = 0;
4321
4322         rcu_read_lock();
4323         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4324                 if (!net_eq(dev_net(ifp->idev->dev), net))
4325                         continue;
4326                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4327                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4328                         ret = 1;
4329                         break;
4330                 }
4331         }
4332         rcu_read_unlock();
4333         return ret;
4334 }
4335 #endif
4336
4337 /*
4338  *      Periodic address status verification
4339  */
4340
4341 static void addrconf_verify_rtnl(void)
4342 {
4343         unsigned long now, next, next_sec, next_sched;
4344         struct inet6_ifaddr *ifp;
4345         int i;
4346
4347         ASSERT_RTNL();
4348
4349         rcu_read_lock_bh();
4350         now = jiffies;
4351         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4352
4353         cancel_delayed_work(&addr_chk_work);
4354
4355         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4356 restart:
4357                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4358                         unsigned long age;
4359
4360                         /* When setting preferred_lft to a value not zero or
4361                          * infinity, while valid_lft is infinity
4362                          * IFA_F_PERMANENT has a non-infinity life time.
4363                          */
4364                         if ((ifp->flags & IFA_F_PERMANENT) &&
4365                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4366                                 continue;
4367
4368                         spin_lock(&ifp->lock);
4369                         /* We try to batch several events at once. */
4370                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4371
4372                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4373                             age >= ifp->valid_lft) {
4374                                 spin_unlock(&ifp->lock);
4375                                 in6_ifa_hold(ifp);
4376                                 ipv6_del_addr(ifp);
4377                                 goto restart;
4378                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4379                                 spin_unlock(&ifp->lock);
4380                                 continue;
4381                         } else if (age >= ifp->prefered_lft) {
4382                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4383                                 int deprecate = 0;
4384
4385                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4386                                         deprecate = 1;
4387                                         ifp->flags |= IFA_F_DEPRECATED;
4388                                 }
4389
4390                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4391                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4392                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4393
4394                                 spin_unlock(&ifp->lock);
4395
4396                                 if (deprecate) {
4397                                         in6_ifa_hold(ifp);
4398
4399                                         ipv6_ifa_notify(0, ifp);
4400                                         in6_ifa_put(ifp);
4401                                         goto restart;
4402                                 }
4403                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4404                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4405                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4406                                         ifp->idev->cnf.dad_transmits *
4407                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4408
4409                                 if (age >= ifp->prefered_lft - regen_advance) {
4410                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4411                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4412                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4413                                         if (!ifp->regen_count && ifpub) {
4414                                                 ifp->regen_count++;
4415                                                 in6_ifa_hold(ifp);
4416                                                 in6_ifa_hold(ifpub);
4417                                                 spin_unlock(&ifp->lock);
4418
4419                                                 spin_lock(&ifpub->lock);
4420                                                 ifpub->regen_count = 0;
4421                                                 spin_unlock(&ifpub->lock);
4422                                                 rcu_read_unlock_bh();
4423                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4424                                                 in6_ifa_put(ifpub);
4425                                                 in6_ifa_put(ifp);
4426                                                 rcu_read_lock_bh();
4427                                                 goto restart;
4428                                         }
4429                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4430                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4431                                 spin_unlock(&ifp->lock);
4432                         } else {
4433                                 /* ifp->prefered_lft <= ifp->valid_lft */
4434                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4435                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4436                                 spin_unlock(&ifp->lock);
4437                         }
4438                 }
4439         }
4440
4441         next_sec = round_jiffies_up(next);
4442         next_sched = next;
4443
4444         /* If rounded timeout is accurate enough, accept it. */
4445         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4446                 next_sched = next_sec;
4447
4448         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4449         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4450                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4451
4452         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4453                  now, next, next_sec, next_sched);
4454         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4455         rcu_read_unlock_bh();
4456 }
4457
4458 static void addrconf_verify_work(struct work_struct *w)
4459 {
4460         rtnl_lock();
4461         addrconf_verify_rtnl();
4462         rtnl_unlock();
4463 }
4464
4465 static void addrconf_verify(void)
4466 {
4467         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4468 }
4469
4470 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4471                                      struct in6_addr **peer_pfx)
4472 {
4473         struct in6_addr *pfx = NULL;
4474
4475         *peer_pfx = NULL;
4476
4477         if (addr)
4478                 pfx = nla_data(addr);
4479
4480         if (local) {
4481                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4482                         *peer_pfx = pfx;
4483                 pfx = nla_data(local);
4484         }
4485
4486         return pfx;
4487 }
4488
4489 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4490         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4491         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4492         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4493         [IFA_FLAGS]             = { .len = sizeof(u32) },
4494         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4495 };
4496
4497 static int
4498 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4499                   struct netlink_ext_ack *extack)
4500 {
4501         struct net *net = sock_net(skb->sk);
4502         struct ifaddrmsg *ifm;
4503         struct nlattr *tb[IFA_MAX+1];
4504         struct in6_addr *pfx, *peer_pfx;
4505         u32 ifa_flags;
4506         int err;
4507
4508         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4509                           extack);
4510         if (err < 0)
4511                 return err;
4512
4513         ifm = nlmsg_data(nlh);
4514         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4515         if (!pfx)
4516                 return -EINVAL;
4517
4518         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4519
4520         /* We ignore other flags so far. */
4521         ifa_flags &= IFA_F_MANAGETEMPADDR;
4522
4523         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4524                               ifm->ifa_prefixlen);
4525 }
4526
4527 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4528                                unsigned long expires, u32 flags)
4529 {
4530         struct fib6_info *f6i;
4531
4532         f6i = addrconf_get_prefix_route(&ifp->addr,
4533                                         ifp->prefix_len,
4534                                         ifp->idev->dev,
4535                                         0, RTF_GATEWAY | RTF_DEFAULT);
4536         if (!f6i)
4537                 return -ENOENT;
4538
4539         if (f6i->fib6_metric != ifp->rt_priority) {
4540                 /* add new one */
4541                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4542                                       ifp->rt_priority, ifp->idev->dev,
4543                                       expires, flags, GFP_KERNEL);
4544                 /* delete old one */
4545                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4546         } else {
4547                 if (!expires)
4548                         fib6_clean_expires(f6i);
4549                 else
4550                         fib6_set_expires(f6i, expires);
4551
4552                 fib6_info_release(f6i);
4553         }
4554
4555         return 0;
4556 }
4557
4558 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4559 {
4560         u32 flags;
4561         clock_t expires;
4562         unsigned long timeout;
4563         bool was_managetempaddr;
4564         bool had_prefixroute;
4565
4566         ASSERT_RTNL();
4567
4568         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4569                 return -EINVAL;
4570
4571         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4572             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4573                 return -EINVAL;
4574
4575         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4576                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4577
4578         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4579         if (addrconf_finite_timeout(timeout)) {
4580                 expires = jiffies_to_clock_t(timeout * HZ);
4581                 cfg->valid_lft = timeout;
4582                 flags = RTF_EXPIRES;
4583         } else {
4584                 expires = 0;
4585                 flags = 0;
4586                 cfg->ifa_flags |= IFA_F_PERMANENT;
4587         }
4588
4589         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4590         if (addrconf_finite_timeout(timeout)) {
4591                 if (timeout == 0)
4592                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4593                 cfg->preferred_lft = timeout;
4594         }
4595
4596         spin_lock_bh(&ifp->lock);
4597         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4598         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4599                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4600         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4601                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4602                         IFA_F_NOPREFIXROUTE);
4603         ifp->flags |= cfg->ifa_flags;
4604         ifp->tstamp = jiffies;
4605         ifp->valid_lft = cfg->valid_lft;
4606         ifp->prefered_lft = cfg->preferred_lft;
4607
4608         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4609                 ifp->rt_priority = cfg->rt_priority;
4610
4611         spin_unlock_bh(&ifp->lock);
4612         if (!(ifp->flags&IFA_F_TENTATIVE))
4613                 ipv6_ifa_notify(0, ifp);
4614
4615         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4616                 int rc = -ENOENT;
4617
4618                 if (had_prefixroute)
4619                         rc = modify_prefix_route(ifp, expires, flags);
4620
4621                 /* prefix route could have been deleted; if so restore it */
4622                 if (rc == -ENOENT) {
4623                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4624                                               ifp->rt_priority, ifp->idev->dev,
4625                                               expires, flags, GFP_KERNEL);
4626                 }
4627         } else if (had_prefixroute) {
4628                 enum cleanup_prefix_rt_t action;
4629                 unsigned long rt_expires;
4630
4631                 write_lock_bh(&ifp->idev->lock);
4632                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4633                 write_unlock_bh(&ifp->idev->lock);
4634
4635                 if (action != CLEANUP_PREFIX_RT_NOP) {
4636                         cleanup_prefix_route(ifp, rt_expires,
4637                                 action == CLEANUP_PREFIX_RT_DEL);
4638                 }
4639         }
4640
4641         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4642                 if (was_managetempaddr &&
4643                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4644                         cfg->valid_lft = 0;
4645                         cfg->preferred_lft = 0;
4646                 }
4647                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4648                                  cfg->preferred_lft, !was_managetempaddr,
4649                                  jiffies);
4650         }
4651
4652         addrconf_verify_rtnl();
4653
4654         return 0;
4655 }
4656
4657 static int
4658 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4659                   struct netlink_ext_ack *extack)
4660 {
4661         struct net *net = sock_net(skb->sk);
4662         struct ifaddrmsg *ifm;
4663         struct nlattr *tb[IFA_MAX+1];
4664         struct in6_addr *peer_pfx;
4665         struct inet6_ifaddr *ifa;
4666         struct net_device *dev;
4667         struct inet6_dev *idev;
4668         struct ifa6_config cfg;
4669         int err;
4670
4671         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4672                           extack);
4673         if (err < 0)
4674                 return err;
4675
4676         memset(&cfg, 0, sizeof(cfg));
4677
4678         ifm = nlmsg_data(nlh);
4679         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4680         if (!cfg.pfx)
4681                 return -EINVAL;
4682
4683         cfg.peer_pfx = peer_pfx;
4684         cfg.plen = ifm->ifa_prefixlen;
4685         if (tb[IFA_RT_PRIORITY])
4686                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4687
4688         cfg.valid_lft = INFINITY_LIFE_TIME;
4689         cfg.preferred_lft = INFINITY_LIFE_TIME;
4690
4691         if (tb[IFA_CACHEINFO]) {
4692                 struct ifa_cacheinfo *ci;
4693
4694                 ci = nla_data(tb[IFA_CACHEINFO]);
4695                 cfg.valid_lft = ci->ifa_valid;
4696                 cfg.preferred_lft = ci->ifa_prefered;
4697         }
4698
4699         dev =  __dev_get_by_index(net, ifm->ifa_index);
4700         if (!dev)
4701                 return -ENODEV;
4702
4703         if (tb[IFA_FLAGS])
4704                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4705         else
4706                 cfg.ifa_flags = ifm->ifa_flags;
4707
4708         /* We ignore other flags so far. */
4709         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4710                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4711                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4712
4713         idev = ipv6_find_idev(dev);
4714         if (IS_ERR(idev))
4715                 return PTR_ERR(idev);
4716
4717         if (!ipv6_allow_optimistic_dad(net, idev))
4718                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4719
4720         if (cfg.ifa_flags & IFA_F_NODAD &&
4721             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4722                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4723                 return -EINVAL;
4724         }
4725
4726         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4727         if (!ifa) {
4728                 /*
4729                  * It would be best to check for !NLM_F_CREATE here but
4730                  * userspace already relies on not having to provide this.
4731                  */
4732                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4733         }
4734
4735         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4736             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4737                 err = -EEXIST;
4738         else
4739                 err = inet6_addr_modify(ifa, &cfg);
4740
4741         in6_ifa_put(ifa);
4742
4743         return err;
4744 }
4745
4746 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4747                           u8 scope, int ifindex)
4748 {
4749         struct ifaddrmsg *ifm;
4750
4751         ifm = nlmsg_data(nlh);
4752         ifm->ifa_family = AF_INET6;
4753         ifm->ifa_prefixlen = prefixlen;
4754         ifm->ifa_flags = flags;
4755         ifm->ifa_scope = scope;
4756         ifm->ifa_index = ifindex;
4757 }
4758
4759 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4760                          unsigned long tstamp, u32 preferred, u32 valid)
4761 {
4762         struct ifa_cacheinfo ci;
4763
4764         ci.cstamp = cstamp_delta(cstamp);
4765         ci.tstamp = cstamp_delta(tstamp);
4766         ci.ifa_prefered = preferred;
4767         ci.ifa_valid = valid;
4768
4769         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4770 }
4771
4772 static inline int rt_scope(int ifa_scope)
4773 {
4774         if (ifa_scope & IFA_HOST)
4775                 return RT_SCOPE_HOST;
4776         else if (ifa_scope & IFA_LINK)
4777                 return RT_SCOPE_LINK;
4778         else if (ifa_scope & IFA_SITE)
4779                 return RT_SCOPE_SITE;
4780         else
4781                 return RT_SCOPE_UNIVERSE;
4782 }
4783
4784 static inline int inet6_ifaddr_msgsize(void)
4785 {
4786         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4787                + nla_total_size(16) /* IFA_LOCAL */
4788                + nla_total_size(16) /* IFA_ADDRESS */
4789                + nla_total_size(sizeof(struct ifa_cacheinfo))
4790                + nla_total_size(4)  /* IFA_FLAGS */
4791                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4792 }
4793
4794 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4795                              u32 portid, u32 seq, int event, unsigned int flags)
4796 {
4797         struct nlmsghdr  *nlh;
4798         u32 preferred, valid;
4799
4800         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4801         if (!nlh)
4802                 return -EMSGSIZE;
4803
4804         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4805                       ifa->idev->dev->ifindex);
4806
4807         if (!((ifa->flags&IFA_F_PERMANENT) &&
4808               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4809                 preferred = ifa->prefered_lft;
4810                 valid = ifa->valid_lft;
4811                 if (preferred != INFINITY_LIFE_TIME) {
4812                         long tval = (jiffies - ifa->tstamp)/HZ;
4813                         if (preferred > tval)
4814                                 preferred -= tval;
4815                         else
4816                                 preferred = 0;
4817                         if (valid != INFINITY_LIFE_TIME) {
4818                                 if (valid > tval)
4819                                         valid -= tval;
4820                                 else
4821                                         valid = 0;
4822                         }
4823                 }
4824         } else {
4825                 preferred = INFINITY_LIFE_TIME;
4826                 valid = INFINITY_LIFE_TIME;
4827         }
4828
4829         if (!ipv6_addr_any(&ifa->peer_addr)) {
4830                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4831                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4832                         goto error;
4833         } else
4834                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4835                         goto error;
4836
4837         if (ifa->rt_priority &&
4838             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4839                 goto error;
4840
4841         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4842                 goto error;
4843
4844         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4845                 goto error;
4846
4847         nlmsg_end(skb, nlh);
4848         return 0;
4849
4850 error:
4851         nlmsg_cancel(skb, nlh);
4852         return -EMSGSIZE;
4853 }
4854
4855 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4856                                 u32 portid, u32 seq, int event, u16 flags)
4857 {
4858         struct nlmsghdr  *nlh;
4859         u8 scope = RT_SCOPE_UNIVERSE;
4860         int ifindex = ifmca->idev->dev->ifindex;
4861
4862         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4863                 scope = RT_SCOPE_SITE;
4864
4865         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4866         if (!nlh)
4867                 return -EMSGSIZE;
4868
4869         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4870         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4871             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4872                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4873                 nlmsg_cancel(skb, nlh);
4874                 return -EMSGSIZE;
4875         }
4876
4877         nlmsg_end(skb, nlh);
4878         return 0;
4879 }
4880
4881 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4882                                 u32 portid, u32 seq, int event, unsigned int flags)
4883 {
4884         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4885         int ifindex = dev ? dev->ifindex : 1;
4886         struct nlmsghdr  *nlh;
4887         u8 scope = RT_SCOPE_UNIVERSE;
4888
4889         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4890                 scope = RT_SCOPE_SITE;
4891
4892         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4893         if (!nlh)
4894                 return -EMSGSIZE;
4895
4896         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4897         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4898             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4899                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4900                 nlmsg_cancel(skb, nlh);
4901                 return -EMSGSIZE;
4902         }
4903
4904         nlmsg_end(skb, nlh);
4905         return 0;
4906 }
4907
4908 enum addr_type_t {
4909         UNICAST_ADDR,
4910         MULTICAST_ADDR,
4911         ANYCAST_ADDR,
4912 };
4913
4914 /* called with rcu_read_lock() */
4915 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4916                           struct netlink_callback *cb, enum addr_type_t type,
4917                           int s_ip_idx, int *p_ip_idx)
4918 {
4919         struct ifmcaddr6 *ifmca;
4920         struct ifacaddr6 *ifaca;
4921         int err = 1;
4922         int ip_idx = *p_ip_idx;
4923
4924         read_lock_bh(&idev->lock);
4925         switch (type) {
4926         case UNICAST_ADDR: {
4927                 struct inet6_ifaddr *ifa;
4928
4929                 /* unicast address incl. temp addr */
4930                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4931                         if (++ip_idx < s_ip_idx)
4932                                 continue;
4933                         err = inet6_fill_ifaddr(skb, ifa,
4934                                                 NETLINK_CB(cb->skb).portid,
4935                                                 cb->nlh->nlmsg_seq,
4936                                                 RTM_NEWADDR,
4937                                                 NLM_F_MULTI);
4938                         if (err < 0)
4939                                 break;
4940                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4941                 }
4942                 break;
4943         }
4944         case MULTICAST_ADDR:
4945                 /* multicast address */
4946                 for (ifmca = idev->mc_list; ifmca;
4947                      ifmca = ifmca->next, ip_idx++) {
4948                         if (ip_idx < s_ip_idx)
4949                                 continue;
4950                         err = inet6_fill_ifmcaddr(skb, ifmca,
4951                                                   NETLINK_CB(cb->skb).portid,
4952                                                   cb->nlh->nlmsg_seq,
4953                                                   RTM_GETMULTICAST,
4954                                                   NLM_F_MULTI);
4955                         if (err < 0)
4956                                 break;
4957                 }
4958                 break;
4959         case ANYCAST_ADDR:
4960                 /* anycast address */
4961                 for (ifaca = idev->ac_list; ifaca;
4962                      ifaca = ifaca->aca_next, ip_idx++) {
4963                         if (ip_idx < s_ip_idx)
4964                                 continue;
4965                         err = inet6_fill_ifacaddr(skb, ifaca,
4966                                                   NETLINK_CB(cb->skb).portid,
4967                                                   cb->nlh->nlmsg_seq,
4968                                                   RTM_GETANYCAST,
4969                                                   NLM_F_MULTI);
4970                         if (err < 0)
4971                                 break;
4972                 }
4973                 break;
4974         default:
4975                 break;
4976         }
4977         read_unlock_bh(&idev->lock);
4978         *p_ip_idx = ip_idx;
4979         return err;
4980 }
4981
4982 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4983                            enum addr_type_t type)
4984 {
4985         struct net *net = sock_net(skb->sk);
4986         int h, s_h;
4987         int idx, ip_idx;
4988         int s_idx, s_ip_idx;
4989         struct net_device *dev;
4990         struct inet6_dev *idev;
4991         struct hlist_head *head;
4992
4993         s_h = cb->args[0];
4994         s_idx = idx = cb->args[1];
4995         s_ip_idx = ip_idx = cb->args[2];
4996
4997         rcu_read_lock();
4998         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4999         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5000                 idx = 0;
5001                 head = &net->dev_index_head[h];
5002                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5003                         if (idx < s_idx)
5004                                 goto cont;
5005                         if (h > s_h || idx > s_idx)
5006                                 s_ip_idx = 0;
5007                         ip_idx = 0;
5008                         idev = __in6_dev_get(dev);
5009                         if (!idev)
5010                                 goto cont;
5011
5012                         if (in6_dump_addrs(idev, skb, cb, type,
5013                                            s_ip_idx, &ip_idx) < 0)
5014                                 goto done;
5015 cont:
5016                         idx++;
5017                 }
5018         }
5019 done:
5020         rcu_read_unlock();
5021         cb->args[0] = h;
5022         cb->args[1] = idx;
5023         cb->args[2] = ip_idx;
5024
5025         return skb->len;
5026 }
5027
5028 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5029 {
5030         enum addr_type_t type = UNICAST_ADDR;
5031
5032         return inet6_dump_addr(skb, cb, type);
5033 }
5034
5035 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5036 {
5037         enum addr_type_t type = MULTICAST_ADDR;
5038
5039         return inet6_dump_addr(skb, cb, type);
5040 }
5041
5042
5043 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5044 {
5045         enum addr_type_t type = ANYCAST_ADDR;
5046
5047         return inet6_dump_addr(skb, cb, type);
5048 }
5049
5050 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5051                              struct netlink_ext_ack *extack)
5052 {
5053         struct net *net = sock_net(in_skb->sk);
5054         struct ifaddrmsg *ifm;
5055         struct nlattr *tb[IFA_MAX+1];
5056         struct in6_addr *addr = NULL, *peer;
5057         struct net_device *dev = NULL;
5058         struct inet6_ifaddr *ifa;
5059         struct sk_buff *skb;
5060         int err;
5061
5062         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5063                           extack);
5064         if (err < 0)
5065                 return err;
5066
5067         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5068         if (!addr)
5069                 return -EINVAL;
5070
5071         ifm = nlmsg_data(nlh);
5072         if (ifm->ifa_index)
5073                 dev = dev_get_by_index(net, ifm->ifa_index);
5074
5075         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5076         if (!ifa) {
5077                 err = -EADDRNOTAVAIL;
5078                 goto errout;
5079         }
5080
5081         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5082         if (!skb) {
5083                 err = -ENOBUFS;
5084                 goto errout_ifa;
5085         }
5086
5087         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5088                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5089         if (err < 0) {
5090                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5091                 WARN_ON(err == -EMSGSIZE);
5092                 kfree_skb(skb);
5093                 goto errout_ifa;
5094         }
5095         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5096 errout_ifa:
5097         in6_ifa_put(ifa);
5098 errout:
5099         if (dev)
5100                 dev_put(dev);
5101         return err;
5102 }
5103
5104 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5105 {
5106         struct sk_buff *skb;
5107         struct net *net = dev_net(ifa->idev->dev);
5108         int err = -ENOBUFS;
5109
5110         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5111         if (!skb)
5112                 goto errout;
5113
5114         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5115         if (err < 0) {
5116                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5117                 WARN_ON(err == -EMSGSIZE);
5118                 kfree_skb(skb);
5119                 goto errout;
5120         }
5121         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5122         return;
5123 errout:
5124         if (err < 0)
5125                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5126 }
5127
5128 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5129                                 __s32 *array, int bytes)
5130 {
5131         BUG_ON(bytes < (DEVCONF_MAX * 4));
5132
5133         memset(array, 0, bytes);
5134         array[DEVCONF_FORWARDING] = cnf->forwarding;
5135         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5136         array[DEVCONF_MTU6] = cnf->mtu6;
5137         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5138         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5139         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5140         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5141         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5142         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5143                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5144         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5145                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5146         array[DEVCONF_RTR_SOLICIT_DELAY] =
5147                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5148         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5149         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5150                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5151         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5152                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5153         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5154         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5155         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5156         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5157         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5158         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5159         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5160         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5161         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5162 #ifdef CONFIG_IPV6_ROUTER_PREF
5163         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5164         array[DEVCONF_RTR_PROBE_INTERVAL] =
5165                 jiffies_to_msecs(cnf->rtr_probe_interval);
5166 #ifdef CONFIG_IPV6_ROUTE_INFO
5167         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5168         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5169 #endif
5170 #endif
5171         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5172         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5173 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5174         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5175         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5176 #endif
5177 #ifdef CONFIG_IPV6_MROUTE
5178         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5179 #endif
5180         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5181         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5182         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5183         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5184         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5185         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5186         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5187         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5188         /* we omit DEVCONF_STABLE_SECRET for now */
5189         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5190         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5191         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5192         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5193         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5194 #ifdef CONFIG_IPV6_SEG6_HMAC
5195         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5196 #endif
5197         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5198         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5199         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5200         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5201 }
5202
5203 static inline size_t inet6_ifla6_size(void)
5204 {
5205         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5206              + nla_total_size(sizeof(struct ifla_cacheinfo))
5207              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5208              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5209              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5210              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5211 }
5212
5213 static inline size_t inet6_if_nlmsg_size(void)
5214 {
5215         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5216                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5217                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5218                + nla_total_size(4) /* IFLA_MTU */
5219                + nla_total_size(4) /* IFLA_LINK */
5220                + nla_total_size(1) /* IFLA_OPERSTATE */
5221                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5222 }
5223
5224 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5225                                         int bytes)
5226 {
5227         int i;
5228         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5229         BUG_ON(pad < 0);
5230
5231         /* Use put_unaligned() because stats may not be aligned for u64. */
5232         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5233         for (i = 1; i < ICMP6_MIB_MAX; i++)
5234                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5235
5236         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5237 }
5238
5239 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5240                                         int bytes, size_t syncpoff)
5241 {
5242         int i, c;
5243         u64 buff[IPSTATS_MIB_MAX];
5244         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5245
5246         BUG_ON(pad < 0);
5247
5248         memset(buff, 0, sizeof(buff));
5249         buff[0] = IPSTATS_MIB_MAX;
5250
5251         for_each_possible_cpu(c) {
5252                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5253                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5254         }
5255
5256         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5257         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5258 }
5259
5260 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5261                              int bytes)
5262 {
5263         switch (attrtype) {
5264         case IFLA_INET6_STATS:
5265                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5266                                      offsetof(struct ipstats_mib, syncp));
5267                 break;
5268         case IFLA_INET6_ICMP6STATS:
5269                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5270                 break;
5271         }
5272 }
5273
5274 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5275                                   u32 ext_filter_mask)
5276 {
5277         struct nlattr *nla;
5278         struct ifla_cacheinfo ci;
5279
5280         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5281                 goto nla_put_failure;
5282         ci.max_reasm_len = IPV6_MAXPLEN;
5283         ci.tstamp = cstamp_delta(idev->tstamp);
5284         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5285         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5286         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5287                 goto nla_put_failure;
5288         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5289         if (!nla)
5290                 goto nla_put_failure;
5291         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5292
5293         /* XXX - MC not implemented */
5294
5295         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5296                 return 0;
5297
5298         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5299         if (!nla)
5300                 goto nla_put_failure;
5301         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5302
5303         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5304         if (!nla)
5305                 goto nla_put_failure;
5306         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5307
5308         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5309         if (!nla)
5310                 goto nla_put_failure;
5311
5312         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5313                 goto nla_put_failure;
5314
5315         read_lock_bh(&idev->lock);
5316         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5317         read_unlock_bh(&idev->lock);
5318
5319         return 0;
5320
5321 nla_put_failure:
5322         return -EMSGSIZE;
5323 }
5324
5325 static size_t inet6_get_link_af_size(const struct net_device *dev,
5326                                      u32 ext_filter_mask)
5327 {
5328         if (!__in6_dev_get(dev))
5329                 return 0;
5330
5331         return inet6_ifla6_size();
5332 }
5333
5334 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5335                               u32 ext_filter_mask)
5336 {
5337         struct inet6_dev *idev = __in6_dev_get(dev);
5338
5339         if (!idev)
5340                 return -ENODATA;
5341
5342         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5343                 return -EMSGSIZE;
5344
5345         return 0;
5346 }
5347
5348 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5349 {
5350         struct inet6_ifaddr *ifp;
5351         struct net_device *dev = idev->dev;
5352         bool clear_token, update_rs = false;
5353         struct in6_addr ll_addr;
5354
5355         ASSERT_RTNL();
5356
5357         if (!token)
5358                 return -EINVAL;
5359         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5360                 return -EINVAL;
5361         if (!ipv6_accept_ra(idev))
5362                 return -EINVAL;
5363         if (idev->cnf.rtr_solicits == 0)
5364                 return -EINVAL;
5365
5366         write_lock_bh(&idev->lock);
5367
5368         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5369         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5370
5371         write_unlock_bh(&idev->lock);
5372
5373         clear_token = ipv6_addr_any(token);
5374         if (clear_token)
5375                 goto update_lft;
5376
5377         if (!idev->dead && (idev->if_flags & IF_READY) &&
5378             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5379                              IFA_F_OPTIMISTIC)) {
5380                 /* If we're not ready, then normal ifup will take care
5381                  * of this. Otherwise, we need to request our rs here.
5382                  */
5383                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5384                 update_rs = true;
5385         }
5386
5387 update_lft:
5388         write_lock_bh(&idev->lock);
5389
5390         if (update_rs) {
5391                 idev->if_flags |= IF_RS_SENT;
5392                 idev->rs_interval = rfc3315_s14_backoff_init(
5393                         idev->cnf.rtr_solicit_interval);
5394                 idev->rs_probes = 1;
5395                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5396         }
5397
5398         /* Well, that's kinda nasty ... */
5399         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5400                 spin_lock(&ifp->lock);
5401                 if (ifp->tokenized) {
5402                         ifp->valid_lft = 0;
5403                         ifp->prefered_lft = 0;
5404                 }
5405                 spin_unlock(&ifp->lock);
5406         }
5407
5408         write_unlock_bh(&idev->lock);
5409         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5410         addrconf_verify_rtnl();
5411         return 0;
5412 }
5413
5414 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5415         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5416         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5417 };
5418
5419 static int inet6_validate_link_af(const struct net_device *dev,
5420                                   const struct nlattr *nla)
5421 {
5422         struct nlattr *tb[IFLA_INET6_MAX + 1];
5423
5424         if (dev && !__in6_dev_get(dev))
5425                 return -EAFNOSUPPORT;
5426
5427         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5428                                 NULL);
5429 }
5430
5431 static int check_addr_gen_mode(int mode)
5432 {
5433         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5434             mode != IN6_ADDR_GEN_MODE_NONE &&
5435             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5436             mode != IN6_ADDR_GEN_MODE_RANDOM)
5437                 return -EINVAL;
5438         return 1;
5439 }
5440
5441 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5442                                 int mode)
5443 {
5444         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5445             !idev->cnf.stable_secret.initialized &&
5446             !net->ipv6.devconf_dflt->stable_secret.initialized)
5447                 return -EINVAL;
5448         return 1;
5449 }
5450
5451 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5452 {
5453         int err = -EINVAL;
5454         struct inet6_dev *idev = __in6_dev_get(dev);
5455         struct nlattr *tb[IFLA_INET6_MAX + 1];
5456
5457         if (!idev)
5458                 return -EAFNOSUPPORT;
5459
5460         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5461                 BUG();
5462
5463         if (tb[IFLA_INET6_TOKEN]) {
5464                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5465                 if (err)
5466                         return err;
5467         }
5468
5469         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5470                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5471
5472                 if (check_addr_gen_mode(mode) < 0 ||
5473                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5474                         return -EINVAL;
5475
5476                 idev->cnf.addr_gen_mode = mode;
5477                 err = 0;
5478         }
5479
5480         return err;
5481 }
5482
5483 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5484                              u32 portid, u32 seq, int event, unsigned int flags)
5485 {
5486         struct net_device *dev = idev->dev;
5487         struct ifinfomsg *hdr;
5488         struct nlmsghdr *nlh;
5489         void *protoinfo;
5490
5491         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5492         if (!nlh)
5493                 return -EMSGSIZE;
5494
5495         hdr = nlmsg_data(nlh);
5496         hdr->ifi_family = AF_INET6;
5497         hdr->__ifi_pad = 0;
5498         hdr->ifi_type = dev->type;
5499         hdr->ifi_index = dev->ifindex;
5500         hdr->ifi_flags = dev_get_flags(dev);
5501         hdr->ifi_change = 0;
5502
5503         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5504             (dev->addr_len &&
5505              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5506             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5507             (dev->ifindex != dev_get_iflink(dev) &&
5508              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5509             nla_put_u8(skb, IFLA_OPERSTATE,
5510                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5511                 goto nla_put_failure;
5512         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5513         if (!protoinfo)
5514                 goto nla_put_failure;
5515
5516         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5517                 goto nla_put_failure;
5518
5519         nla_nest_end(skb, protoinfo);
5520         nlmsg_end(skb, nlh);
5521         return 0;
5522
5523 nla_put_failure:
5524         nlmsg_cancel(skb, nlh);
5525         return -EMSGSIZE;
5526 }
5527
5528 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5529 {
5530         struct net *net = sock_net(skb->sk);
5531         int h, s_h;
5532         int idx = 0, s_idx;
5533         struct net_device *dev;
5534         struct inet6_dev *idev;
5535         struct hlist_head *head;
5536
5537         s_h = cb->args[0];
5538         s_idx = cb->args[1];
5539
5540         rcu_read_lock();
5541         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5542                 idx = 0;
5543                 head = &net->dev_index_head[h];
5544                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5545                         if (idx < s_idx)
5546                                 goto cont;
5547                         idev = __in6_dev_get(dev);
5548                         if (!idev)
5549                                 goto cont;
5550                         if (inet6_fill_ifinfo(skb, idev,
5551                                               NETLINK_CB(cb->skb).portid,
5552                                               cb->nlh->nlmsg_seq,
5553                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5554                                 goto out;
5555 cont:
5556                         idx++;
5557                 }
5558         }
5559 out:
5560         rcu_read_unlock();
5561         cb->args[1] = idx;
5562         cb->args[0] = h;
5563
5564         return skb->len;
5565 }
5566
5567 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5568 {
5569         struct sk_buff *skb;
5570         struct net *net = dev_net(idev->dev);
5571         int err = -ENOBUFS;
5572
5573         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5574         if (!skb)
5575                 goto errout;
5576
5577         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5578         if (err < 0) {
5579                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5580                 WARN_ON(err == -EMSGSIZE);
5581                 kfree_skb(skb);
5582                 goto errout;
5583         }
5584         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5585         return;
5586 errout:
5587         if (err < 0)
5588                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5589 }
5590
5591 static inline size_t inet6_prefix_nlmsg_size(void)
5592 {
5593         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5594                + nla_total_size(sizeof(struct in6_addr))
5595                + nla_total_size(sizeof(struct prefix_cacheinfo));
5596 }
5597
5598 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5599                              struct prefix_info *pinfo, u32 portid, u32 seq,
5600                              int event, unsigned int flags)
5601 {
5602         struct prefixmsg *pmsg;
5603         struct nlmsghdr *nlh;
5604         struct prefix_cacheinfo ci;
5605
5606         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5607         if (!nlh)
5608                 return -EMSGSIZE;
5609
5610         pmsg = nlmsg_data(nlh);
5611         pmsg->prefix_family = AF_INET6;
5612         pmsg->prefix_pad1 = 0;
5613         pmsg->prefix_pad2 = 0;
5614         pmsg->prefix_ifindex = idev->dev->ifindex;
5615         pmsg->prefix_len = pinfo->prefix_len;
5616         pmsg->prefix_type = pinfo->type;
5617         pmsg->prefix_pad3 = 0;
5618         pmsg->prefix_flags = 0;
5619         if (pinfo->onlink)
5620                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5621         if (pinfo->autoconf)
5622                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5623
5624         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5625                 goto nla_put_failure;
5626         ci.preferred_time = ntohl(pinfo->prefered);
5627         ci.valid_time = ntohl(pinfo->valid);
5628         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5629                 goto nla_put_failure;
5630         nlmsg_end(skb, nlh);
5631         return 0;
5632
5633 nla_put_failure:
5634         nlmsg_cancel(skb, nlh);
5635         return -EMSGSIZE;
5636 }
5637
5638 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5639                          struct prefix_info *pinfo)
5640 {
5641         struct sk_buff *skb;
5642         struct net *net = dev_net(idev->dev);
5643         int err = -ENOBUFS;
5644
5645         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5646         if (!skb)
5647                 goto errout;
5648
5649         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5650         if (err < 0) {
5651                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5652                 WARN_ON(err == -EMSGSIZE);
5653                 kfree_skb(skb);
5654                 goto errout;
5655         }
5656         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5657         return;
5658 errout:
5659         if (err < 0)
5660                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5661 }
5662
5663 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5664 {
5665         struct net *net = dev_net(ifp->idev->dev);
5666
5667         if (event)
5668                 ASSERT_RTNL();
5669
5670         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5671
5672         switch (event) {
5673         case RTM_NEWADDR:
5674                 /*
5675                  * If the address was optimistic
5676                  * we inserted the route at the start of
5677                  * our DAD process, so we don't need
5678                  * to do it again
5679                  */
5680                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5681                         ip6_ins_rt(net, ifp->rt);
5682                 if (ifp->idev->cnf.forwarding)
5683                         addrconf_join_anycast(ifp);
5684                 if (!ipv6_addr_any(&ifp->peer_addr))
5685                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5686                                               ifp->idev->dev, 0, 0,
5687                                               GFP_ATOMIC);
5688                 break;
5689         case RTM_DELADDR:
5690                 if (ifp->idev->cnf.forwarding)
5691                         addrconf_leave_anycast(ifp);
5692                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5693                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5694                         struct fib6_info *rt;
5695
5696                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5697                                                        ifp->idev->dev, 0, 0);
5698                         if (rt)
5699                                 ip6_del_rt(net, rt);
5700                 }
5701                 if (ifp->rt) {
5702                         ip6_del_rt(net, ifp->rt);
5703                         ifp->rt = NULL;
5704                 }
5705                 rt_genid_bump_ipv6(net);
5706                 break;
5707         }
5708         atomic_inc(&net->ipv6.dev_addr_genid);
5709 }
5710
5711 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5712 {
5713         rcu_read_lock_bh();
5714         if (likely(ifp->idev->dead == 0))
5715                 __ipv6_ifa_notify(event, ifp);
5716         rcu_read_unlock_bh();
5717 }
5718
5719 #ifdef CONFIG_SYSCTL
5720
5721 static
5722 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5723                            void __user *buffer, size_t *lenp, loff_t *ppos)
5724 {
5725         int *valp = ctl->data;
5726         int val = *valp;
5727         loff_t pos = *ppos;
5728         struct ctl_table lctl;
5729         int ret;
5730
5731         /*
5732          * ctl->data points to idev->cnf.forwarding, we should
5733          * not modify it until we get the rtnl lock.
5734          */
5735         lctl = *ctl;
5736         lctl.data = &val;
5737
5738         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5739
5740         if (write)
5741                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5742         if (ret)
5743                 *ppos = pos;
5744         return ret;
5745 }
5746
5747 static
5748 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5749                         void __user *buffer, size_t *lenp, loff_t *ppos)
5750 {
5751         struct inet6_dev *idev = ctl->extra1;
5752         int min_mtu = IPV6_MIN_MTU;
5753         struct ctl_table lctl;
5754
5755         lctl = *ctl;
5756         lctl.extra1 = &min_mtu;
5757         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5758
5759         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5760 }
5761
5762 static void dev_disable_change(struct inet6_dev *idev)
5763 {
5764         struct netdev_notifier_info info;
5765
5766         if (!idev || !idev->dev)
5767                 return;
5768
5769         netdev_notifier_info_init(&info, idev->dev);
5770         if (idev->cnf.disable_ipv6)
5771                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5772         else
5773                 addrconf_notify(NULL, NETDEV_UP, &info);
5774 }
5775
5776 static void addrconf_disable_change(struct net *net, __s32 newf)
5777 {
5778         struct net_device *dev;
5779         struct inet6_dev *idev;
5780
5781         for_each_netdev(net, dev) {
5782                 idev = __in6_dev_get(dev);
5783                 if (idev) {
5784                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5785                         idev->cnf.disable_ipv6 = newf;
5786                         if (changed)
5787                                 dev_disable_change(idev);
5788                 }
5789         }
5790 }
5791
5792 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5793 {
5794         struct net *net;
5795         int old;
5796
5797         if (!rtnl_trylock())
5798                 return restart_syscall();
5799
5800         net = (struct net *)table->extra2;
5801         old = *p;
5802         *p = newf;
5803
5804         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5805                 rtnl_unlock();
5806                 return 0;
5807         }
5808
5809         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5810                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5811                 addrconf_disable_change(net, newf);
5812         } else if ((!newf) ^ (!old))
5813                 dev_disable_change((struct inet6_dev *)table->extra1);
5814
5815         rtnl_unlock();
5816         return 0;
5817 }
5818
5819 static
5820 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5821                             void __user *buffer, size_t *lenp, loff_t *ppos)
5822 {
5823         int *valp = ctl->data;
5824         int val = *valp;
5825         loff_t pos = *ppos;
5826         struct ctl_table lctl;
5827         int ret;
5828
5829         /*
5830          * ctl->data points to idev->cnf.disable_ipv6, we should
5831          * not modify it until we get the rtnl lock.
5832          */
5833         lctl = *ctl;
5834         lctl.data = &val;
5835
5836         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5837
5838         if (write)
5839                 ret = addrconf_disable_ipv6(ctl, valp, val);
5840         if (ret)
5841                 *ppos = pos;
5842         return ret;
5843 }
5844
5845 static
5846 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5847                               void __user *buffer, size_t *lenp, loff_t *ppos)
5848 {
5849         int *valp = ctl->data;
5850         int ret;
5851         int old, new;
5852
5853         old = *valp;
5854         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5855         new = *valp;
5856
5857         if (write && old != new) {
5858                 struct net *net = ctl->extra2;
5859
5860                 if (!rtnl_trylock())
5861                         return restart_syscall();
5862
5863                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5864                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5865                                                      NETCONFA_PROXY_NEIGH,
5866                                                      NETCONFA_IFINDEX_DEFAULT,
5867                                                      net->ipv6.devconf_dflt);
5868                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5869                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5870                                                      NETCONFA_PROXY_NEIGH,
5871                                                      NETCONFA_IFINDEX_ALL,
5872                                                      net->ipv6.devconf_all);
5873                 else {
5874                         struct inet6_dev *idev = ctl->extra1;
5875
5876                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5877                                                      NETCONFA_PROXY_NEIGH,
5878                                                      idev->dev->ifindex,
5879                                                      &idev->cnf);
5880                 }
5881                 rtnl_unlock();
5882         }
5883
5884         return ret;
5885 }
5886
5887 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5888                                          void __user *buffer, size_t *lenp,
5889                                          loff_t *ppos)
5890 {
5891         int ret = 0;
5892         int new_val;
5893         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5894         struct net *net = (struct net *)ctl->extra2;
5895
5896         if (!rtnl_trylock())
5897                 return restart_syscall();
5898
5899         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5900
5901         if (write) {
5902                 new_val = *((int *)ctl->data);
5903
5904                 if (check_addr_gen_mode(new_val) < 0) {
5905                         ret = -EINVAL;
5906                         goto out;
5907                 }
5908
5909                 /* request for default */
5910                 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5911                         ipv6_devconf_dflt.addr_gen_mode = new_val;
5912
5913                 /* request for individual net device */
5914                 } else {
5915                         if (!idev)
5916                                 goto out;
5917
5918                         if (check_stable_privacy(idev, net, new_val) < 0) {
5919                                 ret = -EINVAL;
5920                                 goto out;
5921                         }
5922
5923                         if (idev->cnf.addr_gen_mode != new_val) {
5924                                 idev->cnf.addr_gen_mode = new_val;
5925                                 addrconf_dev_config(idev->dev);
5926                         }
5927                 }
5928         }
5929
5930 out:
5931         rtnl_unlock();
5932
5933         return ret;
5934 }
5935
5936 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5937                                          void __user *buffer, size_t *lenp,
5938                                          loff_t *ppos)
5939 {
5940         int err;
5941         struct in6_addr addr;
5942         char str[IPV6_MAX_STRLEN];
5943         struct ctl_table lctl = *ctl;
5944         struct net *net = ctl->extra2;
5945         struct ipv6_stable_secret *secret = ctl->data;
5946
5947         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5948                 return -EIO;
5949
5950         lctl.maxlen = IPV6_MAX_STRLEN;
5951         lctl.data = str;
5952
5953         if (!rtnl_trylock())
5954                 return restart_syscall();
5955
5956         if (!write && !secret->initialized) {
5957                 err = -EIO;
5958                 goto out;
5959         }
5960
5961         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5962         if (err >= sizeof(str)) {
5963                 err = -EIO;
5964                 goto out;
5965         }
5966
5967         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5968         if (err || !write)
5969                 goto out;
5970
5971         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5972                 err = -EIO;
5973                 goto out;
5974         }
5975
5976         secret->initialized = true;
5977         secret->secret = addr;
5978
5979         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5980                 struct net_device *dev;
5981
5982                 for_each_netdev(net, dev) {
5983                         struct inet6_dev *idev = __in6_dev_get(dev);
5984
5985                         if (idev) {
5986                                 idev->cnf.addr_gen_mode =
5987                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5988                         }
5989                 }
5990         } else {
5991                 struct inet6_dev *idev = ctl->extra1;
5992
5993                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5994         }
5995
5996 out:
5997         rtnl_unlock();
5998
5999         return err;
6000 }
6001
6002 static
6003 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6004                                                 int write,
6005                                                 void __user *buffer,
6006                                                 size_t *lenp,
6007                                                 loff_t *ppos)
6008 {
6009         int *valp = ctl->data;
6010         int val = *valp;
6011         loff_t pos = *ppos;
6012         struct ctl_table lctl;
6013         int ret;
6014
6015         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6016          * we should not modify it until we get the rtnl lock.
6017          */
6018         lctl = *ctl;
6019         lctl.data = &val;
6020
6021         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6022
6023         if (write)
6024                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6025         if (ret)
6026                 *ppos = pos;
6027         return ret;
6028 }
6029
6030 static
6031 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6032 {
6033         if (rt) {
6034                 if (action)
6035                         rt->dst.flags |= DST_NOPOLICY;
6036                 else
6037                         rt->dst.flags &= ~DST_NOPOLICY;
6038         }
6039 }
6040
6041 static
6042 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6043 {
6044         struct inet6_ifaddr *ifa;
6045
6046         read_lock_bh(&idev->lock);
6047         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6048                 spin_lock(&ifa->lock);
6049                 if (ifa->rt) {
6050                         struct fib6_info *rt = ifa->rt;
6051                         int cpu;
6052
6053                         rcu_read_lock();
6054                         ifa->rt->dst_nopolicy = val ? true : false;
6055                         if (rt->rt6i_pcpu) {
6056                                 for_each_possible_cpu(cpu) {
6057                                         struct rt6_info **rtp;
6058
6059                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6060                                         addrconf_set_nopolicy(*rtp, val);
6061                                 }
6062                         }
6063                         rcu_read_unlock();
6064                 }
6065                 spin_unlock(&ifa->lock);
6066         }
6067         read_unlock_bh(&idev->lock);
6068 }
6069
6070 static
6071 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6072 {
6073         struct inet6_dev *idev;
6074         struct net *net;
6075
6076         if (!rtnl_trylock())
6077                 return restart_syscall();
6078
6079         *valp = val;
6080
6081         net = (struct net *)ctl->extra2;
6082         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6083                 rtnl_unlock();
6084                 return 0;
6085         }
6086
6087         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6088                 struct net_device *dev;
6089
6090                 for_each_netdev(net, dev) {
6091                         idev = __in6_dev_get(dev);
6092                         if (idev)
6093                                 addrconf_disable_policy_idev(idev, val);
6094                 }
6095         } else {
6096                 idev = (struct inet6_dev *)ctl->extra1;
6097                 addrconf_disable_policy_idev(idev, val);
6098         }
6099
6100         rtnl_unlock();
6101         return 0;
6102 }
6103
6104 static
6105 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6106                                    void __user *buffer, size_t *lenp,
6107                                    loff_t *ppos)
6108 {
6109         int *valp = ctl->data;
6110         int val = *valp;
6111         loff_t pos = *ppos;
6112         struct ctl_table lctl;
6113         int ret;
6114
6115         lctl = *ctl;
6116         lctl.data = &val;
6117         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6118
6119         if (write && (*valp != val))
6120                 ret = addrconf_disable_policy(ctl, valp, val);
6121
6122         if (ret)
6123                 *ppos = pos;
6124
6125         return ret;
6126 }
6127
6128 static int minus_one = -1;
6129 static const int zero = 0;
6130 static const int one = 1;
6131 static const int two_five_five = 255;
6132
6133 static const struct ctl_table addrconf_sysctl[] = {
6134         {
6135                 .procname       = "forwarding",
6136                 .data           = &ipv6_devconf.forwarding,
6137                 .maxlen         = sizeof(int),
6138                 .mode           = 0644,
6139                 .proc_handler   = addrconf_sysctl_forward,
6140         },
6141         {
6142                 .procname       = "hop_limit",
6143                 .data           = &ipv6_devconf.hop_limit,
6144                 .maxlen         = sizeof(int),
6145                 .mode           = 0644,
6146                 .proc_handler   = proc_dointvec_minmax,
6147                 .extra1         = (void *)&one,
6148                 .extra2         = (void *)&two_five_five,
6149         },
6150         {
6151                 .procname       = "mtu",
6152                 .data           = &ipv6_devconf.mtu6,
6153                 .maxlen         = sizeof(int),
6154                 .mode           = 0644,
6155                 .proc_handler   = addrconf_sysctl_mtu,
6156         },
6157         {
6158                 .procname       = "accept_ra",
6159                 .data           = &ipv6_devconf.accept_ra,
6160                 .maxlen         = sizeof(int),
6161                 .mode           = 0644,
6162                 .proc_handler   = proc_dointvec,
6163         },
6164         {
6165                 .procname       = "accept_redirects",
6166                 .data           = &ipv6_devconf.accept_redirects,
6167                 .maxlen         = sizeof(int),
6168                 .mode           = 0644,
6169                 .proc_handler   = proc_dointvec,
6170         },
6171         {
6172                 .procname       = "autoconf",
6173                 .data           = &ipv6_devconf.autoconf,
6174                 .maxlen         = sizeof(int),
6175                 .mode           = 0644,
6176                 .proc_handler   = proc_dointvec,
6177         },
6178         {
6179                 .procname       = "dad_transmits",
6180                 .data           = &ipv6_devconf.dad_transmits,
6181                 .maxlen         = sizeof(int),
6182                 .mode           = 0644,
6183                 .proc_handler   = proc_dointvec,
6184         },
6185         {
6186                 .procname       = "router_solicitations",
6187                 .data           = &ipv6_devconf.rtr_solicits,
6188                 .maxlen         = sizeof(int),
6189                 .mode           = 0644,
6190                 .proc_handler   = proc_dointvec_minmax,
6191                 .extra1         = &minus_one,
6192         },
6193         {
6194                 .procname       = "router_solicitation_interval",
6195                 .data           = &ipv6_devconf.rtr_solicit_interval,
6196                 .maxlen         = sizeof(int),
6197                 .mode           = 0644,
6198                 .proc_handler   = proc_dointvec_jiffies,
6199         },
6200         {
6201                 .procname       = "router_solicitation_max_interval",
6202                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6203                 .maxlen         = sizeof(int),
6204                 .mode           = 0644,
6205                 .proc_handler   = proc_dointvec_jiffies,
6206         },
6207         {
6208                 .procname       = "router_solicitation_delay",
6209                 .data           = &ipv6_devconf.rtr_solicit_delay,
6210                 .maxlen         = sizeof(int),
6211                 .mode           = 0644,
6212                 .proc_handler   = proc_dointvec_jiffies,
6213         },
6214         {
6215                 .procname       = "force_mld_version",
6216                 .data           = &ipv6_devconf.force_mld_version,
6217                 .maxlen         = sizeof(int),
6218                 .mode           = 0644,
6219                 .proc_handler   = proc_dointvec,
6220         },
6221         {
6222                 .procname       = "mldv1_unsolicited_report_interval",
6223                 .data           =
6224                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6225                 .maxlen         = sizeof(int),
6226                 .mode           = 0644,
6227                 .proc_handler   = proc_dointvec_ms_jiffies,
6228         },
6229         {
6230                 .procname       = "mldv2_unsolicited_report_interval",
6231                 .data           =
6232                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6233                 .maxlen         = sizeof(int),
6234                 .mode           = 0644,
6235                 .proc_handler   = proc_dointvec_ms_jiffies,
6236         },
6237         {
6238                 .procname       = "use_tempaddr",
6239                 .data           = &ipv6_devconf.use_tempaddr,
6240                 .maxlen         = sizeof(int),
6241                 .mode           = 0644,
6242                 .proc_handler   = proc_dointvec,
6243         },
6244         {
6245                 .procname       = "temp_valid_lft",
6246                 .data           = &ipv6_devconf.temp_valid_lft,
6247                 .maxlen         = sizeof(int),
6248                 .mode           = 0644,
6249                 .proc_handler   = proc_dointvec,
6250         },
6251         {
6252                 .procname       = "temp_prefered_lft",
6253                 .data           = &ipv6_devconf.temp_prefered_lft,
6254                 .maxlen         = sizeof(int),
6255                 .mode           = 0644,
6256                 .proc_handler   = proc_dointvec,
6257         },
6258         {
6259                 .procname       = "regen_max_retry",
6260                 .data           = &ipv6_devconf.regen_max_retry,
6261                 .maxlen         = sizeof(int),
6262                 .mode           = 0644,
6263                 .proc_handler   = proc_dointvec,
6264         },
6265         {
6266                 .procname       = "max_desync_factor",
6267                 .data           = &ipv6_devconf.max_desync_factor,
6268                 .maxlen         = sizeof(int),
6269                 .mode           = 0644,
6270                 .proc_handler   = proc_dointvec,
6271         },
6272         {
6273                 .procname       = "max_addresses",
6274                 .data           = &ipv6_devconf.max_addresses,
6275                 .maxlen         = sizeof(int),
6276                 .mode           = 0644,
6277                 .proc_handler   = proc_dointvec,
6278         },
6279         {
6280                 .procname       = "accept_ra_defrtr",
6281                 .data           = &ipv6_devconf.accept_ra_defrtr,
6282                 .maxlen         = sizeof(int),
6283                 .mode           = 0644,
6284                 .proc_handler   = proc_dointvec,
6285         },
6286         {
6287                 .procname       = "accept_ra_min_hop_limit",
6288                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6289                 .maxlen         = sizeof(int),
6290                 .mode           = 0644,
6291                 .proc_handler   = proc_dointvec,
6292         },
6293         {
6294                 .procname       = "accept_ra_pinfo",
6295                 .data           = &ipv6_devconf.accept_ra_pinfo,
6296                 .maxlen         = sizeof(int),
6297                 .mode           = 0644,
6298                 .proc_handler   = proc_dointvec,
6299         },
6300 #ifdef CONFIG_IPV6_ROUTER_PREF
6301         {
6302                 .procname       = "accept_ra_rtr_pref",
6303                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6304                 .maxlen         = sizeof(int),
6305                 .mode           = 0644,
6306                 .proc_handler   = proc_dointvec,
6307         },
6308         {
6309                 .procname       = "router_probe_interval",
6310                 .data           = &ipv6_devconf.rtr_probe_interval,
6311                 .maxlen         = sizeof(int),
6312                 .mode           = 0644,
6313                 .proc_handler   = proc_dointvec_jiffies,
6314         },
6315 #ifdef CONFIG_IPV6_ROUTE_INFO
6316         {
6317                 .procname       = "accept_ra_rt_info_min_plen",
6318                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6319                 .maxlen         = sizeof(int),
6320                 .mode           = 0644,
6321                 .proc_handler   = proc_dointvec,
6322         },
6323         {
6324                 .procname       = "accept_ra_rt_info_max_plen",
6325                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6326                 .maxlen         = sizeof(int),
6327                 .mode           = 0644,
6328                 .proc_handler   = proc_dointvec,
6329         },
6330 #endif
6331 #endif
6332         {
6333                 .procname       = "proxy_ndp",
6334                 .data           = &ipv6_devconf.proxy_ndp,
6335                 .maxlen         = sizeof(int),
6336                 .mode           = 0644,
6337                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6338         },
6339         {
6340                 .procname       = "accept_source_route",
6341                 .data           = &ipv6_devconf.accept_source_route,
6342                 .maxlen         = sizeof(int),
6343                 .mode           = 0644,
6344                 .proc_handler   = proc_dointvec,
6345         },
6346 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6347         {
6348                 .procname       = "optimistic_dad",
6349                 .data           = &ipv6_devconf.optimistic_dad,
6350                 .maxlen         = sizeof(int),
6351                 .mode           = 0644,
6352                 .proc_handler   = proc_dointvec,
6353         },
6354         {
6355                 .procname       = "use_optimistic",
6356                 .data           = &ipv6_devconf.use_optimistic,
6357                 .maxlen         = sizeof(int),
6358                 .mode           = 0644,
6359                 .proc_handler   = proc_dointvec,
6360         },
6361 #endif
6362 #ifdef CONFIG_IPV6_MROUTE
6363         {
6364                 .procname       = "mc_forwarding",
6365                 .data           = &ipv6_devconf.mc_forwarding,
6366                 .maxlen         = sizeof(int),
6367                 .mode           = 0444,
6368                 .proc_handler   = proc_dointvec,
6369         },
6370 #endif
6371         {
6372                 .procname       = "disable_ipv6",
6373                 .data           = &ipv6_devconf.disable_ipv6,
6374                 .maxlen         = sizeof(int),
6375                 .mode           = 0644,
6376                 .proc_handler   = addrconf_sysctl_disable,
6377         },
6378         {
6379                 .procname       = "accept_dad",
6380                 .data           = &ipv6_devconf.accept_dad,
6381                 .maxlen         = sizeof(int),
6382                 .mode           = 0644,
6383                 .proc_handler   = proc_dointvec,
6384         },
6385         {
6386                 .procname       = "force_tllao",
6387                 .data           = &ipv6_devconf.force_tllao,
6388                 .maxlen         = sizeof(int),
6389                 .mode           = 0644,
6390                 .proc_handler   = proc_dointvec
6391         },
6392         {
6393                 .procname       = "ndisc_notify",
6394                 .data           = &ipv6_devconf.ndisc_notify,
6395                 .maxlen         = sizeof(int),
6396                 .mode           = 0644,
6397                 .proc_handler   = proc_dointvec
6398         },
6399         {
6400                 .procname       = "suppress_frag_ndisc",
6401                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6402                 .maxlen         = sizeof(int),
6403                 .mode           = 0644,
6404                 .proc_handler   = proc_dointvec
6405         },
6406         {
6407                 .procname       = "accept_ra_from_local",
6408                 .data           = &ipv6_devconf.accept_ra_from_local,
6409                 .maxlen         = sizeof(int),
6410                 .mode           = 0644,
6411                 .proc_handler   = proc_dointvec,
6412         },
6413         {
6414                 .procname       = "accept_ra_mtu",
6415                 .data           = &ipv6_devconf.accept_ra_mtu,
6416                 .maxlen         = sizeof(int),
6417                 .mode           = 0644,
6418                 .proc_handler   = proc_dointvec,
6419         },
6420         {
6421                 .procname       = "stable_secret",
6422                 .data           = &ipv6_devconf.stable_secret,
6423                 .maxlen         = IPV6_MAX_STRLEN,
6424                 .mode           = 0600,
6425                 .proc_handler   = addrconf_sysctl_stable_secret,
6426         },
6427         {
6428                 .procname       = "use_oif_addrs_only",
6429                 .data           = &ipv6_devconf.use_oif_addrs_only,
6430                 .maxlen         = sizeof(int),
6431                 .mode           = 0644,
6432                 .proc_handler   = proc_dointvec,
6433         },
6434         {
6435                 .procname       = "ignore_routes_with_linkdown",
6436                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6437                 .maxlen         = sizeof(int),
6438                 .mode           = 0644,
6439                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6440         },
6441         {
6442                 .procname       = "drop_unicast_in_l2_multicast",
6443                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6444                 .maxlen         = sizeof(int),
6445                 .mode           = 0644,
6446                 .proc_handler   = proc_dointvec,
6447         },
6448         {
6449                 .procname       = "drop_unsolicited_na",
6450                 .data           = &ipv6_devconf.drop_unsolicited_na,
6451                 .maxlen         = sizeof(int),
6452                 .mode           = 0644,
6453                 .proc_handler   = proc_dointvec,
6454         },
6455         {
6456                 .procname       = "keep_addr_on_down",
6457                 .data           = &ipv6_devconf.keep_addr_on_down,
6458                 .maxlen         = sizeof(int),
6459                 .mode           = 0644,
6460                 .proc_handler   = proc_dointvec,
6461
6462         },
6463         {
6464                 .procname       = "seg6_enabled",
6465                 .data           = &ipv6_devconf.seg6_enabled,
6466                 .maxlen         = sizeof(int),
6467                 .mode           = 0644,
6468                 .proc_handler   = proc_dointvec,
6469         },
6470 #ifdef CONFIG_IPV6_SEG6_HMAC
6471         {
6472                 .procname       = "seg6_require_hmac",
6473                 .data           = &ipv6_devconf.seg6_require_hmac,
6474                 .maxlen         = sizeof(int),
6475                 .mode           = 0644,
6476                 .proc_handler   = proc_dointvec,
6477         },
6478 #endif
6479         {
6480                 .procname       = "enhanced_dad",
6481                 .data           = &ipv6_devconf.enhanced_dad,
6482                 .maxlen         = sizeof(int),
6483                 .mode           = 0644,
6484                 .proc_handler   = proc_dointvec,
6485         },
6486         {
6487                 .procname               = "addr_gen_mode",
6488                 .data                   = &ipv6_devconf.addr_gen_mode,
6489                 .maxlen                 = sizeof(int),
6490                 .mode                   = 0644,
6491                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6492         },
6493         {
6494                 .procname       = "disable_policy",
6495                 .data           = &ipv6_devconf.disable_policy,
6496                 .maxlen         = sizeof(int),
6497                 .mode           = 0644,
6498                 .proc_handler   = addrconf_sysctl_disable_policy,
6499         },
6500         {
6501                 .procname       = "ndisc_tclass",
6502                 .data           = &ipv6_devconf.ndisc_tclass,
6503                 .maxlen         = sizeof(int),
6504                 .mode           = 0644,
6505                 .proc_handler   = proc_dointvec_minmax,
6506                 .extra1         = (void *)&zero,
6507                 .extra2         = (void *)&two_five_five,
6508         },
6509         {
6510                 /* sentinel */
6511         }
6512 };
6513
6514 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6515                 struct inet6_dev *idev, struct ipv6_devconf *p)
6516 {
6517         int i, ifindex;
6518         struct ctl_table *table;
6519         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6520
6521         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6522         if (!table)
6523                 goto out;
6524
6525         for (i = 0; table[i].data; i++) {
6526                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6527                 /* If one of these is already set, then it is not safe to
6528                  * overwrite either of them: this makes proc_dointvec_minmax
6529                  * usable.
6530                  */
6531                 if (!table[i].extra1 && !table[i].extra2) {
6532                         table[i].extra1 = idev; /* embedded; no ref */
6533                         table[i].extra2 = net;
6534                 }
6535         }
6536
6537         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6538
6539         p->sysctl_header = register_net_sysctl(net, path, table);
6540         if (!p->sysctl_header)
6541                 goto free;
6542
6543         if (!strcmp(dev_name, "all"))
6544                 ifindex = NETCONFA_IFINDEX_ALL;
6545         else if (!strcmp(dev_name, "default"))
6546                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6547         else
6548                 ifindex = idev->dev->ifindex;
6549         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6550                                      ifindex, p);
6551         return 0;
6552
6553 free:
6554         kfree(table);
6555 out:
6556         return -ENOBUFS;
6557 }
6558
6559 static void __addrconf_sysctl_unregister(struct net *net,
6560                                          struct ipv6_devconf *p, int ifindex)
6561 {
6562         struct ctl_table *table;
6563
6564         if (!p->sysctl_header)
6565                 return;
6566
6567         table = p->sysctl_header->ctl_table_arg;
6568         unregister_net_sysctl_table(p->sysctl_header);
6569         p->sysctl_header = NULL;
6570         kfree(table);
6571
6572         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6573 }
6574
6575 static int addrconf_sysctl_register(struct inet6_dev *idev)
6576 {
6577         int err;
6578
6579         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6580                 return -EINVAL;
6581
6582         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6583                                     &ndisc_ifinfo_sysctl_change);
6584         if (err)
6585                 return err;
6586         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6587                                          idev, &idev->cnf);
6588         if (err)
6589                 neigh_sysctl_unregister(idev->nd_parms);
6590
6591         return err;
6592 }
6593
6594 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6595 {
6596         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6597                                      idev->dev->ifindex);
6598         neigh_sysctl_unregister(idev->nd_parms);
6599 }
6600
6601
6602 #endif
6603
6604 static int __net_init addrconf_init_net(struct net *net)
6605 {
6606         int err = -ENOMEM;
6607         struct ipv6_devconf *all, *dflt;
6608
6609         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6610         if (!all)
6611                 goto err_alloc_all;
6612
6613         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6614         if (!dflt)
6615                 goto err_alloc_dflt;
6616
6617         /* these will be inherited by all namespaces */
6618         dflt->autoconf = ipv6_defaults.autoconf;
6619         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6620
6621         dflt->stable_secret.initialized = false;
6622         all->stable_secret.initialized = false;
6623
6624         net->ipv6.devconf_all = all;
6625         net->ipv6.devconf_dflt = dflt;
6626
6627 #ifdef CONFIG_SYSCTL
6628         err = __addrconf_sysctl_register(net, "all", NULL, all);
6629         if (err < 0)
6630                 goto err_reg_all;
6631
6632         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6633         if (err < 0)
6634                 goto err_reg_dflt;
6635 #endif
6636         return 0;
6637
6638 #ifdef CONFIG_SYSCTL
6639 err_reg_dflt:
6640         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6641 err_reg_all:
6642         kfree(dflt);
6643 #endif
6644 err_alloc_dflt:
6645         kfree(all);
6646 err_alloc_all:
6647         return err;
6648 }
6649
6650 static void __net_exit addrconf_exit_net(struct net *net)
6651 {
6652 #ifdef CONFIG_SYSCTL
6653         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6654                                      NETCONFA_IFINDEX_DEFAULT);
6655         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6656                                      NETCONFA_IFINDEX_ALL);
6657 #endif
6658         kfree(net->ipv6.devconf_dflt);
6659         kfree(net->ipv6.devconf_all);
6660 }
6661
6662 static struct pernet_operations addrconf_ops = {
6663         .init = addrconf_init_net,
6664         .exit = addrconf_exit_net,
6665 };
6666
6667 static struct rtnl_af_ops inet6_ops __read_mostly = {
6668         .family           = AF_INET6,
6669         .fill_link_af     = inet6_fill_link_af,
6670         .get_link_af_size = inet6_get_link_af_size,
6671         .validate_link_af = inet6_validate_link_af,
6672         .set_link_af      = inet6_set_link_af,
6673 };
6674
6675 /*
6676  *      Init / cleanup code
6677  */
6678
6679 int __init addrconf_init(void)
6680 {
6681         struct inet6_dev *idev;
6682         int i, err;
6683
6684         err = ipv6_addr_label_init();
6685         if (err < 0) {
6686                 pr_crit("%s: cannot initialize default policy table: %d\n",
6687                         __func__, err);
6688                 goto out;
6689         }
6690
6691         err = register_pernet_subsys(&addrconf_ops);
6692         if (err < 0)
6693                 goto out_addrlabel;
6694
6695         addrconf_wq = create_workqueue("ipv6_addrconf");
6696         if (!addrconf_wq) {
6697                 err = -ENOMEM;
6698                 goto out_nowq;
6699         }
6700
6701         /* The addrconf netdev notifier requires that loopback_dev
6702          * has it's ipv6 private information allocated and setup
6703          * before it can bring up and give link-local addresses
6704          * to other devices which are up.
6705          *
6706          * Unfortunately, loopback_dev is not necessarily the first
6707          * entry in the global dev_base list of net devices.  In fact,
6708          * it is likely to be the very last entry on that list.
6709          * So this causes the notifier registry below to try and
6710          * give link-local addresses to all devices besides loopback_dev
6711          * first, then loopback_dev, which cases all the non-loopback_dev
6712          * devices to fail to get a link-local address.
6713          *
6714          * So, as a temporary fix, allocate the ipv6 structure for
6715          * loopback_dev first by hand.
6716          * Longer term, all of the dependencies ipv6 has upon the loopback
6717          * device and it being up should be removed.
6718          */
6719         rtnl_lock();
6720         idev = ipv6_add_dev(init_net.loopback_dev);
6721         rtnl_unlock();
6722         if (IS_ERR(idev)) {
6723                 err = PTR_ERR(idev);
6724                 goto errlo;
6725         }
6726
6727         ip6_route_init_special_entries();
6728
6729         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6730                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6731
6732         register_netdevice_notifier(&ipv6_dev_notf);
6733
6734         addrconf_verify();
6735
6736         rtnl_af_register(&inet6_ops);
6737
6738         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6739                                    NULL, inet6_dump_ifinfo, 0);
6740         if (err < 0)
6741                 goto errout;
6742
6743         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6744                                    inet6_rtm_newaddr, NULL, 0);
6745         if (err < 0)
6746                 goto errout;
6747         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6748                                    inet6_rtm_deladdr, NULL, 0);
6749         if (err < 0)
6750                 goto errout;
6751         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6752                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
6753                                    RTNL_FLAG_DOIT_UNLOCKED);
6754         if (err < 0)
6755                 goto errout;
6756         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6757                                    NULL, inet6_dump_ifmcaddr, 0);
6758         if (err < 0)
6759                 goto errout;
6760         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6761                                    NULL, inet6_dump_ifacaddr, 0);
6762         if (err < 0)
6763                 goto errout;
6764         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6765                                    inet6_netconf_get_devconf,
6766                                    inet6_netconf_dump_devconf,
6767                                    RTNL_FLAG_DOIT_UNLOCKED);
6768         if (err < 0)
6769                 goto errout;
6770         err = ipv6_addr_label_rtnl_register();
6771         if (err < 0)
6772                 goto errout;
6773
6774         return 0;
6775 errout:
6776         rtnl_unregister_all(PF_INET6);
6777         rtnl_af_unregister(&inet6_ops);
6778         unregister_netdevice_notifier(&ipv6_dev_notf);
6779 errlo:
6780         destroy_workqueue(addrconf_wq);
6781 out_nowq:
6782         unregister_pernet_subsys(&addrconf_ops);
6783 out_addrlabel:
6784         ipv6_addr_label_cleanup();
6785 out:
6786         return err;
6787 }
6788
6789 void addrconf_cleanup(void)
6790 {
6791         struct net_device *dev;
6792         int i;
6793
6794         unregister_netdevice_notifier(&ipv6_dev_notf);
6795         unregister_pernet_subsys(&addrconf_ops);
6796         ipv6_addr_label_cleanup();
6797
6798         rtnl_af_unregister(&inet6_ops);
6799
6800         rtnl_lock();
6801
6802         /* clean dev list */
6803         for_each_netdev(&init_net, dev) {
6804                 if (__in6_dev_get(dev) == NULL)
6805                         continue;
6806                 addrconf_ifdown(dev, 1);
6807         }
6808         addrconf_ifdown(init_net.loopback_dev, 2);
6809
6810         /*
6811          *      Check hash table.
6812          */
6813         spin_lock_bh(&addrconf_hash_lock);
6814         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6815                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6816         spin_unlock_bh(&addrconf_hash_lock);
6817         cancel_delayed_work(&addr_chk_work);
6818         rtnl_unlock();
6819
6820         destroy_workqueue(addrconf_wq);
6821 }