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
 389         ndev->cnf.mtu6 = dev->mtu;
 390         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 391         if (!ndev->nd_parms) {
 392                 kfree(ndev);
 393                 return ERR_PTR(err);
 394         }
 395         if (ndev->cnf.forwarding)
 396                 dev_disable_lro(dev);
 397         /* We refer to the device */
 398         dev_hold(dev);
 399
 400         if (snmp6_alloc_dev(ndev) < 0) {
 401                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 402                            __func__);
 403                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 404                 dev_put(dev);
 405                 kfree(ndev);
 406                 return ERR_PTR(err);
 407         }
 408
 409         if (snmp6_register_dev(ndev) < 0) {
 410                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 411                            __func__, dev->name);
 412                 goto err_release;
 413         }
 414
 415         /* One reference from device. */
 416         refcount_set(&ndev->refcnt, 1);
 417
 418         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 419                 ndev->cnf.accept_dad = -1;
 420
 421 #if IS_ENABLED(CONFIG_IPV6_SIT)
 422         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 423                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 424                 ndev->cnf.rtr_solicits = 0;
 425         }
 426 #endif
 427
 428         INIT_LIST_HEAD(&ndev->tempaddr_list);
 429         ndev->desync_factor = U32_MAX;
 430         if ((dev->flags&IFF_LOOPBACK) ||
 431             dev->type == ARPHRD_TUNNEL ||
 432             dev->type == ARPHRD_TUNNEL6 ||
 433             dev->type == ARPHRD_SIT ||
 434             dev->type == ARPHRD_NONE) {
 435                 ndev->cnf.use_tempaddr = -1;
 436         } else
 437                 ipv6_regen_rndid(ndev);
 438
 439         ndev->token = in6addr_any;
 440
 441         if (netif_running(dev) && addrconf_link_ready(dev))
 442                 ndev->if_flags |= IF_READY;
 443
 444         ipv6_mc_init_dev(ndev);
 445         ndev->tstamp = jiffies;
 446         err = addrconf_sysctl_register(ndev);
 447         if (err) {
 448                 ipv6_mc_destroy_dev(ndev);
 449                 snmp6_unregister_dev(ndev);
 450                 goto err_release;
 451         }
 452         /* protected by rtnl_lock */
 453         rcu_assign_pointer(dev->ip6_ptr, ndev);
 454
 455         /* Join interface-local all-node multicast group */
 456         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 457
 458         /* Join all-node multicast group */
 459         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 460
 461         /* Join all-router multicast group if forwarding is set */
 462         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 463                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 464
 465         return ndev;
 466
 467 err_release:
 468         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 469         ndev->dead = 1;
 470         in6_dev_finish_destroy(ndev);
 471         return ERR_PTR(err);
 472 }
 473
 474 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 475 {
 476         struct inet6_dev *idev;
 477
 478         ASSERT_RTNL();
 479
 480         idev = __in6_dev_get(dev);
 481         if (!idev) {
 482                 idev = ipv6_add_dev(dev);
 483                 if (IS_ERR(idev))
 484                         return NULL;
 485         }
 486
 487         if (dev->flags&IFF_UP)
 488                 ipv6_mc_up(idev);
 489         return idev;
 490 }
 491
 492 static int inet6_netconf_msgsize_devconf(int type)
 493 {
 494         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 495                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 496         bool all = false;
 497
 498         if (type == NETCONFA_ALL)
 499                 all = true;
 500
 501         if (all || type == NETCONFA_FORWARDING)
 502                 size += nla_total_size(4);
 503 #ifdef CONFIG_IPV6_MROUTE
 504         if (all || type == NETCONFA_MC_FORWARDING)
 505                 size += nla_total_size(4);
 506 #endif
 507         if (all || type == NETCONFA_PROXY_NEIGH)
 508                 size += nla_total_size(4);
 509
 510         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 511                 size += nla_total_size(4);
 512
 513         return size;
 514 }
 515
 516 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 517                                       struct ipv6_devconf *devconf, u32 portid,
 518                                       u32 seq, int event, unsigned int flags,
 519                                       int type)
 520 {
 521         struct nlmsghdr  *nlh;
 522         struct netconfmsg *ncm;
 523         bool all = false;
 524
 525         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 526                         flags);
 527         if (!nlh)
 528                 return -EMSGSIZE;
 529
 530         if (type == NETCONFA_ALL)
 531                 all = true;
 532
 533         ncm = nlmsg_data(nlh);
 534         ncm->ncm_family = AF_INET6;
 535
 536         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 537                 goto nla_put_failure;
 538
 539         if (!devconf)
 540                 goto out;
 541
 542         if ((all || type == NETCONFA_FORWARDING) &&
 543             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 544                 goto nla_put_failure;
 545 #ifdef CONFIG_IPV6_MROUTE
 546         if ((all || type == NETCONFA_MC_FORWARDING) &&
 547             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 548                         devconf->mc_forwarding) < 0)
 549                 goto nla_put_failure;
 550 #endif
 551         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 552             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 553                 goto nla_put_failure;
 554
 555         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 556             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 557                         devconf->ignore_routes_with_linkdown) < 0)
 558                 goto nla_put_failure;
 559
 560 out:
 561         nlmsg_end(skb, nlh);
 562         return 0;
 563
 564 nla_put_failure:
 565         nlmsg_cancel(skb, nlh);
 566         return -EMSGSIZE;
 567 }
 568
 569 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 570                                   int ifindex, struct ipv6_devconf *devconf)
 571 {
 572         struct sk_buff *skb;
 573         int err = -ENOBUFS;
 574
 575         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 576         if (!skb)
 577                 goto errout;
 578
 579         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 580                                          event, 0, type);
 581         if (err < 0) {
 582                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 583                 WARN_ON(err == -EMSGSIZE);
 584                 kfree_skb(skb);
 585                 goto errout;
 586         }
 587         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 588         return;
 589 errout:
 590         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 591 }
 592
 593 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 594         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 595         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 596         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 597         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 598 };
 599
 600 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 601                                      struct nlmsghdr *nlh,
 602                                      struct netlink_ext_ack *extack)
 603 {
 604         struct net *net = sock_net(in_skb->sk);
 605         struct nlattr *tb[NETCONFA_MAX+1];
 606         struct inet6_dev *in6_dev = NULL;
 607         struct net_device *dev = NULL;
 608         struct netconfmsg *ncm;
 609         struct sk_buff *skb;
 610         struct ipv6_devconf *devconf;
 611         int ifindex;
 612         int err;
 613
 614         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 615                           devconf_ipv6_policy, extack);
 616         if (err < 0)
 617                 return err;
 618
 619         if (!tb[NETCONFA_IFINDEX])
 620                 return -EINVAL;
 621
 622         err = -EINVAL;
 623         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 624         switch (ifindex) {
 625         case NETCONFA_IFINDEX_ALL:
 626                 devconf = net->ipv6.devconf_all;
 627                 break;
 628         case NETCONFA_IFINDEX_DEFAULT:
 629                 devconf = net->ipv6.devconf_dflt;
 630                 break;
 631         default:
 632                 dev = dev_get_by_index(net, ifindex);
 633                 if (!dev)
 634                         return -EINVAL;
 635                 in6_dev = in6_dev_get(dev);
 636                 if (!in6_dev)
 637                         goto errout;
 638                 devconf = &in6_dev->cnf;
 639                 break;
 640         }
 641
 642         err = -ENOBUFS;
 643         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 644         if (!skb)
 645                 goto errout;
 646
 647         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 648                                          NETLINK_CB(in_skb).portid,
 649                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 650                                          NETCONFA_ALL);
 651         if (err < 0) {
 652                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 653                 WARN_ON(err == -EMSGSIZE);
 654                 kfree_skb(skb);
 655                 goto errout;
 656         }
 657         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 658 errout:
 659         if (in6_dev)
 660                 in6_dev_put(in6_dev);
 661         if (dev)
 662                 dev_put(dev);
 663         return err;
 664 }
 665
 666 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 667                                       struct netlink_callback *cb)
 668 {
 669         struct net *net = sock_net(skb->sk);
 670         int h, s_h;
 671         int idx, s_idx;
 672         struct net_device *dev;
 673         struct inet6_dev *idev;
 674         struct hlist_head *head;
 675
 676         s_h = cb->args[0];
 677         s_idx = idx = cb->args[1];
 678
 679         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 680                 idx = 0;
 681                 head = &net->dev_index_head[h];
 682                 rcu_read_lock();
 683                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 684                           net->dev_base_seq;
 685                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 686                         if (idx < s_idx)
 687                                 goto cont;
 688                         idev = __in6_dev_get(dev);
 689                         if (!idev)
 690                                 goto cont;
 691
 692                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 693                                                        &idev->cnf,
 694                                                        NETLINK_CB(cb->skb).portid,
 695                                                        cb->nlh->nlmsg_seq,
 696                                                        RTM_NEWNETCONF,
 697                                                        NLM_F_MULTI,
 698                                                        NETCONFA_ALL) < 0) {
 699                                 rcu_read_unlock();
 700                                 goto done;
 701                         }
 702                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 703 cont:
 704                         idx++;
 705                 }
 706                 rcu_read_unlock();
 707         }
 708         if (h == NETDEV_HASHENTRIES) {
 709                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 710                                                net->ipv6.devconf_all,
 711                                                NETLINK_CB(cb->skb).portid,
 712                                                cb->nlh->nlmsg_seq,
 713                                                RTM_NEWNETCONF, NLM_F_MULTI,
 714                                                NETCONFA_ALL) < 0)
 715                         goto done;
 716                 else
 717                         h++;
 718         }
 719         if (h == NETDEV_HASHENTRIES + 1) {
 720                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 721                                                net->ipv6.devconf_dflt,
 722                                                NETLINK_CB(cb->skb).portid,
 723                                                cb->nlh->nlmsg_seq,
 724                                                RTM_NEWNETCONF, NLM_F_MULTI,
 725                                                NETCONFA_ALL) < 0)
 726                         goto done;
 727                 else
 728                         h++;
 729         }
 730 done:
 731         cb->args[0] = h;
 732         cb->args[1] = idx;
 733
 734         return skb->len;
 735 }
 736
 737 #ifdef CONFIG_SYSCTL
 738 static void dev_forward_change(struct inet6_dev *idev)
 739 {
 740         struct net_device *dev;
 741         struct inet6_ifaddr *ifa;
 742
 743         if (!idev)
 744                 return;
 745         dev = idev->dev;
 746         if (idev->cnf.forwarding)
 747                 dev_disable_lro(dev);
 748         if (dev->flags & IFF_MULTICAST) {
 749                 if (idev->cnf.forwarding) {
 750                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 751                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 752                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 753                 } else {
 754                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 755                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 756                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 757                 }
 758         }
 759
 760         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 761                 if (ifa->flags&IFA_F_TENTATIVE)
 762                         continue;
 763                 if (idev->cnf.forwarding)
 764                         addrconf_join_anycast(ifa);
 765                 else
 766                         addrconf_leave_anycast(ifa);
 767         }
 768         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 769                                      NETCONFA_FORWARDING,
 770                                      dev->ifindex, &idev->cnf);
 771 }
 772
 773
 774 static void addrconf_forward_change(struct net *net, __s32 newf)
 775 {
 776         struct net_device *dev;
 777         struct inet6_dev *idev;
 778
 779         for_each_netdev(net, dev) {
 780                 idev = __in6_dev_get(dev);
 781                 if (idev) {
 782                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 783                         idev->cnf.forwarding = newf;
 784                         if (changed)
 785                                 dev_forward_change(idev);
 786                 }
 787         }
 788 }
 789
 790 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 791 {
 792         struct net *net;
 793         int old;
 794
 795         if (!rtnl_trylock())
 796                 return restart_syscall();
 797
 798         net = (struct net *)table->extra2;
 799         old = *p;
 800         *p = newf;
 801
 802         if (p == &net->ipv6.devconf_dflt->forwarding) {
 803                 if ((!newf) ^ (!old))
 804                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 805                                                      NETCONFA_FORWARDING,
 806                                                      NETCONFA_IFINDEX_DEFAULT,
 807                                                      net->ipv6.devconf_dflt);
 808                 rtnl_unlock();
 809                 return 0;
 810         }
 811
 812         if (p == &net->ipv6.devconf_all->forwarding) {
 813                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 814
 815                 net->ipv6.devconf_dflt->forwarding = newf;
 816                 if ((!newf) ^ (!old_dflt))
 817                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 818                                                      NETCONFA_FORWARDING,
 819                                                      NETCONFA_IFINDEX_DEFAULT,
 820                                                      net->ipv6.devconf_dflt);
 821
 822                 addrconf_forward_change(net, newf);
 823                 if ((!newf) ^ (!old))
 824                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 825                                                      NETCONFA_FORWARDING,
 826                                                      NETCONFA_IFINDEX_ALL,
 827                                                      net->ipv6.devconf_all);
 828         } else if ((!newf) ^ (!old))
 829                 dev_forward_change((struct inet6_dev *)table->extra1);
 830         rtnl_unlock();
 831
 832         if (newf)
 833                 rt6_purge_dflt_routers(net);
 834         return 1;
 835 }
 836
 837 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 838 {
 839         struct net_device *dev;
 840         struct inet6_dev *idev;
 841
 842         for_each_netdev(net, dev) {
 843                 idev = __in6_dev_get(dev);
 844                 if (idev) {
 845                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 846
 847                         idev->cnf.ignore_routes_with_linkdown = newf;
 848                         if (changed)
 849                                 inet6_netconf_notify_devconf(dev_net(dev),
 850                                                              RTM_NEWNETCONF,
 851                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 852                                                              dev->ifindex,
 853                                                              &idev->cnf);
 854                 }
 855         }
 856 }
 857
 858 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 859 {
 860         struct net *net;
 861         int old;
 862
 863         if (!rtnl_trylock())
 864                 return restart_syscall();
 865
 866         net = (struct net *)table->extra2;
 867         old = *p;
 868         *p = newf;
 869
 870         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 871                 if ((!newf) ^ (!old))
 872                         inet6_netconf_notify_devconf(net,
 873                                                      RTM_NEWNETCONF,
 874                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 875                                                      NETCONFA_IFINDEX_DEFAULT,
 876                                                      net->ipv6.devconf_dflt);
 877                 rtnl_unlock();
 878                 return 0;
 879         }
 880
 881         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 882                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 883                 addrconf_linkdown_change(net, newf);
 884                 if ((!newf) ^ (!old))
 885                         inet6_netconf_notify_devconf(net,
 886                                                      RTM_NEWNETCONF,
 887                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 888                                                      NETCONFA_IFINDEX_ALL,
 889                                                      net->ipv6.devconf_all);
 890         }
 891         rtnl_unlock();
 892
 893         return 1;
 894 }
 895
 896 #endif
 897
 898 /* Nobody refers to this ifaddr, destroy it */
 899 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 900 {
 901         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 902
 903 #ifdef NET_REFCNT_DEBUG
 904         pr_debug("%s\n", __func__);
 905 #endif
 906
 907         in6_dev_put(ifp->idev);
 908
 909         if (cancel_delayed_work(&ifp->dad_work))
 910                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 911                           ifp);
 912
 913         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 914                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 915                 return;
 916         }
 917
 918         kfree_rcu(ifp, rcu);
 919 }
 920
 921 static void
 922 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 923 {
 924         struct list_head *p;
 925         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 926
 927         /*
 928          * Each device address list is sorted in order of scope -
 929          * global before linklocal.
 930          */
 931         list_for_each(p, &idev->addr_list) {
 932                 struct inet6_ifaddr *ifa
 933                         = list_entry(p, struct inet6_ifaddr, if_list);
 934                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 935                         break;
 936         }
 937
 938         list_add_tail_rcu(&ifp->if_list, p);
 939 }
 940
 941 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 942 {
 943         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 944
 945         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 946 }
 947
 948 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 949                                struct net_device *dev, unsigned int hash)
 950 {
 951         struct inet6_ifaddr *ifp;
 952
 953         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
 954                 if (!net_eq(dev_net(ifp->idev->dev), net))
 955                         continue;
 956                 if (ipv6_addr_equal(&ifp->addr, addr)) {
 957                         if (!dev || ifp->idev->dev == dev)
 958                                 return true;
 959                 }
 960         }
 961         return false;
 962 }
 963
 964 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
 965 {
 966         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
 967         int err = 0;
 968
 969         spin_lock(&addrconf_hash_lock);
 970
 971         /* Ignore adding duplicate addresses on an interface */
 972         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
 973                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
 974                 err = -EEXIST;
 975         } else {
 976                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 977         }
 978
 979         spin_unlock(&addrconf_hash_lock);
 980
 981         return err;
 982 }
 983
 984 /* On success it returns ifp with increased reference count */
 985
 986 static struct inet6_ifaddr *
 987 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
 988               bool can_block, struct netlink_ext_ack *extack)
 989 {
 990         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
 991         int addr_type = ipv6_addr_type(cfg->pfx);
 992         struct net *net = dev_net(idev->dev);
 993         struct inet6_ifaddr *ifa = NULL;
 994         struct fib6_info *f6i = NULL;
 995         int err = 0;
 996
 997         if (addr_type == IPV6_ADDR_ANY ||
 998             addr_type & IPV6_ADDR_MULTICAST ||
 999             (!(idev->dev->flags & IFF_LOOPBACK) &&
1000              addr_type & IPV6_ADDR_LOOPBACK))
1001                 return ERR_PTR(-EADDRNOTAVAIL);
1002
1003         if (idev->dead) {
1004                 err = -ENODEV;                  /*XXX*/
1005                 goto out;
1006         }
1007
1008         if (idev->cnf.disable_ipv6) {
1009                 err = -EACCES;
1010                 goto out;
1011         }
1012
1013         /* validator notifier needs to be blocking;
1014          * do not call in atomic context
1015          */
1016         if (can_block) {
1017                 struct in6_validator_info i6vi = {
1018                         .i6vi_addr = *cfg->pfx,
1019                         .i6vi_dev = idev,
1020                         .extack = extack,
1021                 };
1022
1023                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1024                 err = notifier_to_errno(err);
1025                 if (err < 0)
1026                         goto out;
1027         }
1028
1029         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1030         if (!ifa) {
1031                 err = -ENOBUFS;
1032                 goto out;
1033         }
1034
1035         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1036         if (IS_ERR(f6i)) {
1037                 err = PTR_ERR(f6i);
1038                 f6i = NULL;
1039                 goto out;
1040         }
1041
1042         if (net->ipv6.devconf_all->disable_policy ||
1043             idev->cnf.disable_policy)
1044                 f6i->dst_nopolicy = true;
1045
1046         neigh_parms_data_state_setall(idev->nd_parms);
1047
1048         ifa->addr = *cfg->pfx;
1049         if (cfg->peer_pfx)
1050                 ifa->peer_addr = *cfg->peer_pfx;
1051
1052         spin_lock_init(&ifa->lock);
1053         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1054         INIT_HLIST_NODE(&ifa->addr_lst);
1055         ifa->scope = cfg->scope;
1056         ifa->prefix_len = cfg->plen;
1057         ifa->rt_priority = cfg->rt_priority;
1058         ifa->flags = cfg->ifa_flags;
1059         /* No need to add the TENTATIVE flag for addresses with NODAD */
1060         if (!(cfg->ifa_flags & IFA_F_NODAD))
1061                 ifa->flags |= IFA_F_TENTATIVE;
1062         ifa->valid_lft = cfg->valid_lft;
1063         ifa->prefered_lft = cfg->preferred_lft;
1064         ifa->cstamp = ifa->tstamp = jiffies;
1065         ifa->tokenized = false;
1066
1067         ifa->rt = f6i;
1068
1069         ifa->idev = idev;
1070         in6_dev_hold(idev);
1071
1072         /* For caller */
1073         refcount_set(&ifa->refcnt, 1);
1074
1075         rcu_read_lock_bh();
1076
1077         err = ipv6_add_addr_hash(idev->dev, ifa);
1078         if (err < 0) {
1079                 rcu_read_unlock_bh();
1080                 goto out;
1081         }
1082
1083         write_lock(&idev->lock);
1084
1085         /* Add to inet6_dev unicast addr list. */
1086         ipv6_link_dev_addr(idev, ifa);
1087
1088         if (ifa->flags&IFA_F_TEMPORARY) {
1089                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1090                 in6_ifa_hold(ifa);
1091         }
1092
1093         in6_ifa_hold(ifa);
1094         write_unlock(&idev->lock);
1095
1096         rcu_read_unlock_bh();
1097
1098         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1099 out:
1100         if (unlikely(err < 0)) {
1101                 fib6_info_release(f6i);
1102
1103                 if (ifa) {
1104                         if (ifa->idev)
1105                                 in6_dev_put(ifa->idev);
1106                         kfree(ifa);
1107                 }
1108                 ifa = ERR_PTR(err);
1109         }
1110
1111         return ifa;
1112 }
1113
1114 enum cleanup_prefix_rt_t {
1115         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1116         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1117         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1118 };
1119
1120 /*
1121  * Check, whether the prefix for ifp would still need a prefix route
1122  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1123  * constants.
1124  *
1125  * 1) we don't purge prefix if address was not permanent.
1126  *    prefix is managed by its own lifetime.
1127  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1128  * 3) if there are no addresses, delete prefix.
1129  * 4) if there are still other permanent address(es),
1130  *    corresponding prefix is still permanent.
1131  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1132  *    don't purge the prefix, assume user space is managing it.
1133  * 6) otherwise, update prefix lifetime to the
1134  *    longest valid lifetime among the corresponding
1135  *    addresses on the device.
1136  *    Note: subsequent RA will update lifetime.
1137  **/
1138 static enum cleanup_prefix_rt_t
1139 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1140 {
1141         struct inet6_ifaddr *ifa;
1142         struct inet6_dev *idev = ifp->idev;
1143         unsigned long lifetime;
1144         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1145
1146         *expires = jiffies;
1147
1148         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1149                 if (ifa == ifp)
1150                         continue;
1151                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1152                                        ifp->prefix_len))
1153                         continue;
1154                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1155                         return CLEANUP_PREFIX_RT_NOP;
1156
1157                 action = CLEANUP_PREFIX_RT_EXPIRE;
1158
1159                 spin_lock(&ifa->lock);
1160
1161                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1162                 /*
1163                  * Note: Because this address is
1164                  * not permanent, lifetime <
1165                  * LONG_MAX / HZ here.
1166                  */
1167                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1168                         *expires = ifa->tstamp + lifetime * HZ;
1169                 spin_unlock(&ifa->lock);
1170         }
1171
1172         return action;
1173 }
1174
1175 static void
1176 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1177 {
1178         struct fib6_info *f6i;
1179
1180         f6i = addrconf_get_prefix_route(&ifp->addr,
1181                                        ifp->prefix_len,
1182                                        ifp->idev->dev,
1183                                        0, RTF_GATEWAY | RTF_DEFAULT);
1184         if (f6i) {
1185                 if (del_rt)
1186                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1187                 else {
1188                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1189                                 fib6_set_expires(f6i, expires);
1190                         fib6_info_release(f6i);
1191                 }
1192         }
1193 }
1194
1195
1196 /* This function wants to get referenced ifp and releases it before return */
1197
1198 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1199 {
1200         int state;
1201         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1202         unsigned long expires;
1203
1204         ASSERT_RTNL();
1205
1206         spin_lock_bh(&ifp->lock);
1207         state = ifp->state;
1208         ifp->state = INET6_IFADDR_STATE_DEAD;
1209         spin_unlock_bh(&ifp->lock);
1210
1211         if (state == INET6_IFADDR_STATE_DEAD)
1212                 goto out;
1213
1214         spin_lock_bh(&addrconf_hash_lock);
1215         hlist_del_init_rcu(&ifp->addr_lst);
1216         spin_unlock_bh(&addrconf_hash_lock);
1217
1218         write_lock_bh(&ifp->idev->lock);
1219
1220         if (ifp->flags&IFA_F_TEMPORARY) {
1221                 list_del(&ifp->tmp_list);
1222                 if (ifp->ifpub) {
1223                         in6_ifa_put(ifp->ifpub);
1224                         ifp->ifpub = NULL;
1225                 }
1226                 __in6_ifa_put(ifp);
1227         }
1228
1229         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1230                 action = check_cleanup_prefix_route(ifp, &expires);
1231
1232         list_del_rcu(&ifp->if_list);
1233         __in6_ifa_put(ifp);
1234
1235         write_unlock_bh(&ifp->idev->lock);
1236
1237         addrconf_del_dad_work(ifp);
1238
1239         ipv6_ifa_notify(RTM_DELADDR, ifp);
1240
1241         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1242
1243         if (action != CLEANUP_PREFIX_RT_NOP) {
1244                 cleanup_prefix_route(ifp, expires,
1245                         action == CLEANUP_PREFIX_RT_DEL);
1246         }
1247
1248         /* clean up prefsrc entries */
1249         rt6_remove_prefsrc(ifp);
1250 out:
1251         in6_ifa_put(ifp);
1252 }
1253
1254 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1255                                 struct inet6_ifaddr *ift,
1256                                 bool block)
1257 {
1258         struct inet6_dev *idev = ifp->idev;
1259         struct in6_addr addr, *tmpaddr;
1260         unsigned long tmp_tstamp, age;
1261         unsigned long regen_advance;
1262         struct ifa6_config cfg;
1263         int ret = 0;
1264         unsigned long now = jiffies;
1265         long max_desync_factor;
1266         s32 cnf_temp_preferred_lft;
1267
1268         write_lock_bh(&idev->lock);
1269         if (ift) {
1270                 spin_lock_bh(&ift->lock);
1271                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1272                 spin_unlock_bh(&ift->lock);
1273                 tmpaddr = &addr;
1274         } else {
1275                 tmpaddr = NULL;
1276         }
1277 retry:
1278         in6_dev_hold(idev);
1279         if (idev->cnf.use_tempaddr <= 0) {
1280                 write_unlock_bh(&idev->lock);
1281                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1282                 in6_dev_put(idev);
1283                 ret = -1;
1284                 goto out;
1285         }
1286         spin_lock_bh(&ifp->lock);
1287         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1288                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1289                 spin_unlock_bh(&ifp->lock);
1290                 write_unlock_bh(&idev->lock);
1291                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1292                         __func__);
1293                 in6_dev_put(idev);
1294                 ret = -1;
1295                 goto out;
1296         }
1297         in6_ifa_hold(ifp);
1298         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1299         ipv6_try_regen_rndid(idev, tmpaddr);
1300         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1301         age = (now - ifp->tstamp) / HZ;
1302
1303         regen_advance = idev->cnf.regen_max_retry *
1304                         idev->cnf.dad_transmits *
1305                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1306
1307         /* recalculate max_desync_factor each time and update
1308          * idev->desync_factor if it's larger
1309          */
1310         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1311         max_desync_factor = min_t(__u32,
1312                                   idev->cnf.max_desync_factor,
1313                                   cnf_temp_preferred_lft - regen_advance);
1314
1315         if (unlikely(idev->desync_factor > max_desync_factor)) {
1316                 if (max_desync_factor > 0) {
1317                         get_random_bytes(&idev->desync_factor,
1318                                          sizeof(idev->desync_factor));
1319                         idev->desync_factor %= max_desync_factor;
1320                 } else {
1321                         idev->desync_factor = 0;
1322                 }
1323         }
1324
1325         memset(&cfg, 0, sizeof(cfg));
1326         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1327                               idev->cnf.temp_valid_lft + age);
1328         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1329         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1330
1331         cfg.plen = ifp->prefix_len;
1332         tmp_tstamp = ifp->tstamp;
1333         spin_unlock_bh(&ifp->lock);
1334
1335         write_unlock_bh(&idev->lock);
1336
1337         /* A temporary address is created only if this calculated Preferred
1338          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1339          * an implementation must not create a temporary address with a zero
1340          * Preferred Lifetime.
1341          * Use age calculation as in addrconf_verify to avoid unnecessary
1342          * temporary addresses being generated.
1343          */
1344         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1345         if (cfg.preferred_lft <= regen_advance + age) {
1346                 in6_ifa_put(ifp);
1347                 in6_dev_put(idev);
1348                 ret = -1;
1349                 goto out;
1350         }
1351
1352         cfg.ifa_flags = IFA_F_TEMPORARY;
1353         /* set in addrconf_prefix_rcv() */
1354         if (ifp->flags & IFA_F_OPTIMISTIC)
1355                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1356
1357         cfg.pfx = &addr;
1358         cfg.scope = ipv6_addr_scope(cfg.pfx);
1359
1360         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1361         if (IS_ERR(ift)) {
1362                 in6_ifa_put(ifp);
1363                 in6_dev_put(idev);
1364                 pr_info("%s: retry temporary address regeneration\n", __func__);
1365                 tmpaddr = &addr;
1366                 write_lock_bh(&idev->lock);
1367                 goto retry;
1368         }
1369
1370         spin_lock_bh(&ift->lock);
1371         ift->ifpub = ifp;
1372         ift->cstamp = now;
1373         ift->tstamp = tmp_tstamp;
1374         spin_unlock_bh(&ift->lock);
1375
1376         addrconf_dad_start(ift);
1377         in6_ifa_put(ift);
1378         in6_dev_put(idev);
1379 out:
1380         return ret;
1381 }
1382
1383 /*
1384  *      Choose an appropriate source address (RFC3484)
1385  */
1386 enum {
1387         IPV6_SADDR_RULE_INIT = 0,
1388         IPV6_SADDR_RULE_LOCAL,
1389         IPV6_SADDR_RULE_SCOPE,
1390         IPV6_SADDR_RULE_PREFERRED,
1391 #ifdef CONFIG_IPV6_MIP6
1392         IPV6_SADDR_RULE_HOA,
1393 #endif
1394         IPV6_SADDR_RULE_OIF,
1395         IPV6_SADDR_RULE_LABEL,
1396         IPV6_SADDR_RULE_PRIVACY,
1397         IPV6_SADDR_RULE_ORCHID,
1398         IPV6_SADDR_RULE_PREFIX,
1399 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1400         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1401 #endif
1402         IPV6_SADDR_RULE_MAX
1403 };
1404
1405 struct ipv6_saddr_score {
1406         int                     rule;
1407         int                     addr_type;
1408         struct inet6_ifaddr     *ifa;
1409         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1410         int                     scopedist;
1411         int                     matchlen;
1412 };
1413
1414 struct ipv6_saddr_dst {
1415         const struct in6_addr *addr;
1416         int ifindex;
1417         int scope;
1418         int label;
1419         unsigned int prefs;
1420 };
1421
1422 static inline int ipv6_saddr_preferred(int type)
1423 {
1424         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1425                 return 1;
1426         return 0;
1427 }
1428
1429 static bool ipv6_use_optimistic_addr(struct net *net,
1430                                      struct inet6_dev *idev)
1431 {
1432 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1433         if (!idev)
1434                 return false;
1435         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1436                 return false;
1437         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1438                 return false;
1439
1440         return true;
1441 #else
1442         return false;
1443 #endif
1444 }
1445
1446 static bool ipv6_allow_optimistic_dad(struct net *net,
1447                                       struct inet6_dev *idev)
1448 {
1449 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1450         if (!idev)
1451                 return false;
1452         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1453                 return false;
1454
1455         return true;
1456 #else
1457         return false;
1458 #endif
1459 }
1460
1461 static int ipv6_get_saddr_eval(struct net *net,
1462                                struct ipv6_saddr_score *score,
1463                                struct ipv6_saddr_dst *dst,
1464                                int i)
1465 {
1466         int ret;
1467
1468         if (i <= score->rule) {
1469                 switch (i) {
1470                 case IPV6_SADDR_RULE_SCOPE:
1471                         ret = score->scopedist;
1472                         break;
1473                 case IPV6_SADDR_RULE_PREFIX:
1474                         ret = score->matchlen;
1475                         break;
1476                 default:
1477                         ret = !!test_bit(i, score->scorebits);
1478                 }
1479                 goto out;
1480         }
1481
1482         switch (i) {
1483         case IPV6_SADDR_RULE_INIT:
1484                 /* Rule 0: remember if hiscore is not ready yet */
1485                 ret = !!score->ifa;
1486                 break;
1487         case IPV6_SADDR_RULE_LOCAL:
1488                 /* Rule 1: Prefer same address */
1489                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1490                 break;
1491         case IPV6_SADDR_RULE_SCOPE:
1492                 /* Rule 2: Prefer appropriate scope
1493                  *
1494                  *      ret
1495                  *       ^
1496                  *    -1 |  d 15
1497                  *    ---+--+-+---> scope
1498                  *       |
1499                  *       |             d is scope of the destination.
1500                  *  B-d  |  \
1501                  *       |   \      <- smaller scope is better if
1502                  *  B-15 |    \        if scope is enough for destination.
1503                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1504                  * d-C-1 | /
1505                  *       |/         <- greater is better
1506                  *   -C  /             if scope is not enough for destination.
1507                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1508                  *
1509                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1510                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1511                  * Assume B = 0 and we get C > 29.
1512                  */
1513                 ret = __ipv6_addr_src_scope(score->addr_type);
1514                 if (ret >= dst->scope)
1515                         ret = -ret;
1516                 else
1517                         ret -= 128;     /* 30 is enough */
1518                 score->scopedist = ret;
1519                 break;
1520         case IPV6_SADDR_RULE_PREFERRED:
1521             {
1522                 /* Rule 3: Avoid deprecated and optimistic addresses */
1523                 u8 avoid = IFA_F_DEPRECATED;
1524
1525                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1526                         avoid |= IFA_F_OPTIMISTIC;
1527                 ret = ipv6_saddr_preferred(score->addr_type) ||
1528                       !(score->ifa->flags & avoid);
1529                 break;
1530             }
1531 #ifdef CONFIG_IPV6_MIP6
1532         case IPV6_SADDR_RULE_HOA:
1533             {
1534                 /* Rule 4: Prefer home address */
1535                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1536                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1537                 break;
1538             }
1539 #endif
1540         case IPV6_SADDR_RULE_OIF:
1541                 /* Rule 5: Prefer outgoing interface */
1542                 ret = (!dst->ifindex ||
1543                        dst->ifindex == score->ifa->idev->dev->ifindex);
1544                 break;
1545         case IPV6_SADDR_RULE_LABEL:
1546                 /* Rule 6: Prefer matching label */
1547                 ret = ipv6_addr_label(net,
1548                                       &score->ifa->addr, score->addr_type,
1549                                       score->ifa->idev->dev->ifindex) == dst->label;
1550                 break;
1551         case IPV6_SADDR_RULE_PRIVACY:
1552             {
1553                 /* Rule 7: Prefer public address
1554                  * Note: prefer temporary address if use_tempaddr >= 2
1555                  */
1556                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1557                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1558                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1559                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1560                 break;
1561             }
1562         case IPV6_SADDR_RULE_ORCHID:
1563                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1564                  *          non-ORCHID vs non-ORCHID
1565                  */
1566                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1567                         ipv6_addr_orchid(dst->addr));
1568                 break;
1569         case IPV6_SADDR_RULE_PREFIX:
1570                 /* Rule 8: Use longest matching prefix */
1571                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1572                 if (ret > score->ifa->prefix_len)
1573                         ret = score->ifa->prefix_len;
1574                 score->matchlen = ret;
1575                 break;
1576 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1577         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1578                 /* Optimistic addresses still have lower precedence than other
1579                  * preferred addresses.
1580                  */
1581                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1582                 break;
1583 #endif
1584         default:
1585                 ret = 0;
1586         }
1587
1588         if (ret)
1589                 __set_bit(i, score->scorebits);
1590         score->rule = i;
1591 out:
1592         return ret;
1593 }
1594
1595 static int __ipv6_dev_get_saddr(struct net *net,
1596                                 struct ipv6_saddr_dst *dst,
1597                                 struct inet6_dev *idev,
1598                                 struct ipv6_saddr_score *scores,
1599                                 int hiscore_idx)
1600 {
1601         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1602
1603         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1604                 int i;
1605
1606                 /*
1607                  * - Tentative Address (RFC2462 section 5.4)
1608                  *  - A tentative address is not considered
1609                  *    "assigned to an interface" in the traditional
1610                  *    sense, unless it is also flagged as optimistic.
1611                  * - Candidate Source Address (section 4)
1612                  *  - In any case, anycast addresses, multicast
1613                  *    addresses, and the unspecified address MUST
1614                  *    NOT be included in a candidate set.
1615                  */
1616                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1617                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1618                         continue;
1619
1620                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1621
1622                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1623                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1624                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1625                                             idev->dev->name);
1626                         continue;
1627                 }
1628
1629                 score->rule = -1;
1630                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1631
1632                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1633                         int minihiscore, miniscore;
1634
1635                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1636                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1637
1638                         if (minihiscore > miniscore) {
1639                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1640                                     score->scopedist > 0) {
1641                                         /*
1642                                          * special case:
1643                                          * each remaining entry
1644                                          * has too small (not enough)
1645                                          * scope, because ifa entries
1646                                          * are sorted by their scope
1647                                          * values.
1648                                          */
1649                                         goto out;
1650                                 }
1651                                 break;
1652                         } else if (minihiscore < miniscore) {
1653                                 swap(hiscore, score);
1654                                 hiscore_idx = 1 - hiscore_idx;
1655
1656                                 /* restore our iterator */
1657                                 score->ifa = hiscore->ifa;
1658
1659                                 break;
1660                         }
1661                 }
1662         }
1663 out:
1664         return hiscore_idx;
1665 }
1666
1667 static int ipv6_get_saddr_master(struct net *net,
1668                                  const struct net_device *dst_dev,
1669                                  const struct net_device *master,
1670                                  struct ipv6_saddr_dst *dst,
1671                                  struct ipv6_saddr_score *scores,
1672                                  int hiscore_idx)
1673 {
1674         struct inet6_dev *idev;
1675
1676         idev = __in6_dev_get(dst_dev);
1677         if (idev)
1678                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1679                                                    scores, hiscore_idx);
1680
1681         idev = __in6_dev_get(master);
1682         if (idev)
1683                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1684                                                    scores, hiscore_idx);
1685
1686         return hiscore_idx;
1687 }
1688
1689 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1690                        const struct in6_addr *daddr, unsigned int prefs,
1691                        struct in6_addr *saddr)
1692 {
1693         struct ipv6_saddr_score scores[2], *hiscore;
1694         struct ipv6_saddr_dst dst;
1695         struct inet6_dev *idev;
1696         struct net_device *dev;
1697         int dst_type;
1698         bool use_oif_addr = false;
1699         int hiscore_idx = 0;
1700         int ret = 0;
1701
1702         dst_type = __ipv6_addr_type(daddr);
1703         dst.addr = daddr;
1704         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1705         dst.scope = __ipv6_addr_src_scope(dst_type);
1706         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1707         dst.prefs = prefs;
1708
1709         scores[hiscore_idx].rule = -1;
1710         scores[hiscore_idx].ifa = NULL;
1711
1712         rcu_read_lock();
1713
1714         /* Candidate Source Address (section 4)
1715          *  - multicast and link-local destination address,
1716          *    the set of candidate source address MUST only
1717          *    include addresses assigned to interfaces
1718          *    belonging to the same link as the outgoing
1719          *    interface.
1720          * (- For site-local destination addresses, the
1721          *    set of candidate source addresses MUST only
1722          *    include addresses assigned to interfaces
1723          *    belonging to the same site as the outgoing
1724          *    interface.)
1725          *  - "It is RECOMMENDED that the candidate source addresses
1726          *    be the set of unicast addresses assigned to the
1727          *    interface that will be used to send to the destination
1728          *    (the 'outgoing' interface)." (RFC 6724)
1729          */
1730         if (dst_dev) {
1731                 idev = __in6_dev_get(dst_dev);
1732                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1733                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1734                     (idev && idev->cnf.use_oif_addrs_only)) {
1735                         use_oif_addr = true;
1736                 }
1737         }
1738
1739         if (use_oif_addr) {
1740                 if (idev)
1741                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1742         } else {
1743                 const struct net_device *master;
1744                 int master_idx = 0;
1745
1746                 /* if dst_dev exists and is enslaved to an L3 device, then
1747                  * prefer addresses from dst_dev and then the master over
1748                  * any other enslaved devices in the L3 domain.
1749                  */
1750                 master = l3mdev_master_dev_rcu(dst_dev);
1751                 if (master) {
1752                         master_idx = master->ifindex;
1753
1754                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1755                                                             master, &dst,
1756                                                             scores, hiscore_idx);
1757
1758                         if (scores[hiscore_idx].ifa)
1759                                 goto out;
1760                 }
1761
1762                 for_each_netdev_rcu(net, dev) {
1763                         /* only consider addresses on devices in the
1764                          * same L3 domain
1765                          */
1766                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1767                                 continue;
1768                         idev = __in6_dev_get(dev);
1769                         if (!idev)
1770                                 continue;
1771                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1772                 }
1773         }
1774
1775 out:
1776         hiscore = &scores[hiscore_idx];
1777         if (!hiscore->ifa)
1778                 ret = -EADDRNOTAVAIL;
1779         else
1780                 *saddr = hiscore->ifa->addr;
1781
1782         rcu_read_unlock();
1783         return ret;
1784 }
1785 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1786
1787 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1788                       u32 banned_flags)
1789 {
1790         struct inet6_ifaddr *ifp;
1791         int err = -EADDRNOTAVAIL;
1792
1793         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1794                 if (ifp->scope > IFA_LINK)
1795                         break;
1796                 if (ifp->scope == IFA_LINK &&
1797                     !(ifp->flags & banned_flags)) {
1798                         *addr = ifp->addr;
1799                         err = 0;
1800                         break;
1801                 }
1802         }
1803         return err;
1804 }
1805
1806 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1807                     u32 banned_flags)
1808 {
1809         struct inet6_dev *idev;
1810         int err = -EADDRNOTAVAIL;
1811
1812         rcu_read_lock();
1813         idev = __in6_dev_get(dev);
1814         if (idev) {
1815                 read_lock_bh(&idev->lock);
1816                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1817                 read_unlock_bh(&idev->lock);
1818         }
1819         rcu_read_unlock();
1820         return err;
1821 }
1822
1823 static int ipv6_count_addresses(const struct inet6_dev *idev)
1824 {
1825         const struct inet6_ifaddr *ifp;
1826         int cnt = 0;
1827
1828         rcu_read_lock();
1829         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1830                 cnt++;
1831         rcu_read_unlock();
1832         return cnt;
1833 }
1834
1835 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1836                   const struct net_device *dev, int strict)
1837 {
1838         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1839                                        strict, IFA_F_TENTATIVE);
1840 }
1841 EXPORT_SYMBOL(ipv6_chk_addr);
1842
1843 /* device argument is used to find the L3 domain of interest. If
1844  * skip_dev_check is set, then the ifp device is not checked against
1845  * the passed in dev argument. So the 2 cases for addresses checks are:
1846  *   1. does the address exist in the L3 domain that dev is part of
1847  *      (skip_dev_check = true), or
1848  *
1849  *   2. does the address exist on the specific device
1850  *      (skip_dev_check = false)
1851  */
1852 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1853                             const struct net_device *dev, bool skip_dev_check,
1854                             int strict, u32 banned_flags)
1855 {
1856         unsigned int hash = inet6_addr_hash(net, addr);
1857         const struct net_device *l3mdev;
1858         struct inet6_ifaddr *ifp;
1859         u32 ifp_flags;
1860
1861         rcu_read_lock();
1862
1863         l3mdev = l3mdev_master_dev_rcu(dev);
1864         if (skip_dev_check)
1865                 dev = NULL;
1866
1867         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1868                 if (!net_eq(dev_net(ifp->idev->dev), net))
1869                         continue;
1870
1871                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1872                         continue;
1873
1874                 /* Decouple optimistic from tentative for evaluation here.
1875                  * Ban optimistic addresses explicitly, when required.
1876                  */
1877                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1878                             ? (ifp->flags&~IFA_F_TENTATIVE)
1879                             : ifp->flags;
1880                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1881                     !(ifp_flags&banned_flags) &&
1882                     (!dev || ifp->idev->dev == dev ||
1883                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1884                         rcu_read_unlock();
1885                         return 1;
1886                 }
1887         }
1888
1889         rcu_read_unlock();
1890         return 0;
1891 }
1892 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1893
1894
1895 /* Compares an address/prefix_len with addresses on device @dev.
1896  * If one is found it returns true.
1897  */
1898 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1899         const unsigned int prefix_len, struct net_device *dev)
1900 {
1901         const struct inet6_ifaddr *ifa;
1902         const struct inet6_dev *idev;
1903         bool ret = false;
1904
1905         rcu_read_lock();
1906         idev = __in6_dev_get(dev);
1907         if (idev) {
1908                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1909                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1910                         if (ret)
1911                                 break;
1912                 }
1913         }
1914         rcu_read_unlock();
1915
1916         return ret;
1917 }
1918 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1919
1920 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1921 {
1922         const struct inet6_ifaddr *ifa;
1923         const struct inet6_dev *idev;
1924         int     onlink;
1925
1926         onlink = 0;
1927         rcu_read_lock();
1928         idev = __in6_dev_get(dev);
1929         if (idev) {
1930                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1931                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1932                                                    ifa->prefix_len);
1933                         if (onlink)
1934                                 break;
1935                 }
1936         }
1937         rcu_read_unlock();
1938         return onlink;
1939 }
1940 EXPORT_SYMBOL(ipv6_chk_prefix);
1941
1942 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1943                                      struct net_device *dev, int strict)
1944 {
1945         unsigned int hash = inet6_addr_hash(net, addr);
1946         struct inet6_ifaddr *ifp, *result = NULL;
1947
1948         rcu_read_lock();
1949         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1950                 if (!net_eq(dev_net(ifp->idev->dev), net))
1951                         continue;
1952                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1953                         if (!dev || ifp->idev->dev == dev ||
1954                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1955                                 result = ifp;
1956                                 in6_ifa_hold(ifp);
1957                                 break;
1958                         }
1959                 }
1960         }
1961         rcu_read_unlock();
1962
1963         return result;
1964 }
1965
1966 /* Gets referenced address, destroys ifaddr */
1967
1968 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1969 {
1970         if (dad_failed)
1971                 ifp->flags |= IFA_F_DADFAILED;
1972
1973         if (ifp->flags&IFA_F_TEMPORARY) {
1974                 struct inet6_ifaddr *ifpub;
1975                 spin_lock_bh(&ifp->lock);
1976                 ifpub = ifp->ifpub;
1977                 if (ifpub) {
1978                         in6_ifa_hold(ifpub);
1979                         spin_unlock_bh(&ifp->lock);
1980                         ipv6_create_tempaddr(ifpub, ifp, true);
1981                         in6_ifa_put(ifpub);
1982                 } else {
1983                         spin_unlock_bh(&ifp->lock);
1984                 }
1985                 ipv6_del_addr(ifp);
1986         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1987                 spin_lock_bh(&ifp->lock);
1988                 addrconf_del_dad_work(ifp);
1989                 ifp->flags |= IFA_F_TENTATIVE;
1990                 if (dad_failed)
1991                         ifp->flags &= ~IFA_F_OPTIMISTIC;
1992                 spin_unlock_bh(&ifp->lock);
1993                 if (dad_failed)
1994                         ipv6_ifa_notify(0, ifp);
1995                 in6_ifa_put(ifp);
1996         } else {
1997                 ipv6_del_addr(ifp);
1998         }
1999 }
2000
2001 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2002 {
2003         int err = -ENOENT;
2004
2005         spin_lock_bh(&ifp->lock);
2006         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2007                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2008                 err = 0;
2009         }
2010         spin_unlock_bh(&ifp->lock);
2011
2012         return err;
2013 }
2014
2015 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2016 {
2017         struct inet6_dev *idev = ifp->idev;
2018         struct net *net = dev_net(ifp->idev->dev);
2019
2020         if (addrconf_dad_end(ifp)) {
2021                 in6_ifa_put(ifp);
2022                 return;
2023         }
2024
2025         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2026                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2027
2028         spin_lock_bh(&ifp->lock);
2029
2030         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2031                 struct in6_addr new_addr;
2032                 struct inet6_ifaddr *ifp2;
2033                 int retries = ifp->stable_privacy_retry + 1;
2034                 struct ifa6_config cfg = {
2035                         .pfx = &new_addr,
2036                         .plen = ifp->prefix_len,
2037                         .ifa_flags = ifp->flags,
2038                         .valid_lft = ifp->valid_lft,
2039                         .preferred_lft = ifp->prefered_lft,
2040                         .scope = ifp->scope,
2041                 };
2042
2043                 if (retries > net->ipv6.sysctl.idgen_retries) {
2044                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2045                                              ifp->idev->dev->name);
2046                         goto errdad;
2047                 }
2048
2049                 new_addr = ifp->addr;
2050                 if (ipv6_generate_stable_address(&new_addr, retries,
2051                                                  idev))
2052                         goto errdad;
2053
2054                 spin_unlock_bh(&ifp->lock);
2055
2056                 if (idev->cnf.max_addresses &&
2057                     ipv6_count_addresses(idev) >=
2058                     idev->cnf.max_addresses)
2059                         goto lock_errdad;
2060
2061                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2062                                      ifp->idev->dev->name);
2063
2064                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2065                 if (IS_ERR(ifp2))
2066                         goto lock_errdad;
2067
2068                 spin_lock_bh(&ifp2->lock);
2069                 ifp2->stable_privacy_retry = retries;
2070                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2071                 spin_unlock_bh(&ifp2->lock);
2072
2073                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2074                 in6_ifa_put(ifp2);
2075 lock_errdad:
2076                 spin_lock_bh(&ifp->lock);
2077         }
2078
2079 errdad:
2080         /* transition from _POSTDAD to _ERRDAD */
2081         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2082         spin_unlock_bh(&ifp->lock);
2083
2084         addrconf_mod_dad_work(ifp, 0);
2085         in6_ifa_put(ifp);
2086 }
2087
2088 /* Join to solicited addr multicast group.
2089  * caller must hold RTNL */
2090 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2091 {
2092         struct in6_addr maddr;
2093
2094         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2095                 return;
2096
2097         addrconf_addr_solict_mult(addr, &maddr);
2098         ipv6_dev_mc_inc(dev, &maddr);
2099 }
2100
2101 /* caller must hold RTNL */
2102 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2103 {
2104         struct in6_addr maddr;
2105
2106         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2107                 return;
2108
2109         addrconf_addr_solict_mult(addr, &maddr);
2110         __ipv6_dev_mc_dec(idev, &maddr);
2111 }
2112
2113 /* caller must hold RTNL */
2114 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2115 {
2116         struct in6_addr addr;
2117
2118         if (ifp->prefix_len >= 127) /* RFC 6164 */
2119                 return;
2120         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2121         if (ipv6_addr_any(&addr))
2122                 return;
2123         __ipv6_dev_ac_inc(ifp->idev, &addr);
2124 }
2125
2126 /* caller must hold RTNL */
2127 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2128 {
2129         struct in6_addr addr;
2130
2131         if (ifp->prefix_len >= 127) /* RFC 6164 */
2132                 return;
2133         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2134         if (ipv6_addr_any(&addr))
2135                 return;
2136         __ipv6_dev_ac_dec(ifp->idev, &addr);
2137 }
2138
2139 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2140 {
2141         switch (dev->addr_len) {
2142         case ETH_ALEN:
2143                 memcpy(eui, dev->dev_addr, 3);
2144                 eui[3] = 0xFF;
2145                 eui[4] = 0xFE;
2146                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2147                 break;
2148         case EUI64_ADDR_LEN:
2149                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2150                 eui[0] ^= 2;
2151                 break;
2152         default:
2153                 return -1;
2154         }
2155
2156         return 0;
2157 }
2158
2159 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2160 {
2161         union fwnet_hwaddr *ha;
2162
2163         if (dev->addr_len != FWNET_ALEN)
2164                 return -1;
2165
2166         ha = (union fwnet_hwaddr *)dev->dev_addr;
2167
2168         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2169         eui[0] ^= 2;
2170         return 0;
2171 }
2172
2173 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2174 {
2175         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2176         if (dev->addr_len != ARCNET_ALEN)
2177                 return -1;
2178         memset(eui, 0, 7);
2179         eui[7] = *(u8 *)dev->dev_addr;
2180         return 0;
2181 }
2182
2183 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2184 {
2185         if (dev->addr_len != INFINIBAND_ALEN)
2186                 return -1;
2187         memcpy(eui, dev->dev_addr + 12, 8);
2188         eui[0] |= 2;
2189         return 0;
2190 }
2191
2192 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2193 {
2194         if (addr == 0)
2195                 return -1;
2196         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2197                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2198                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2199                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2200                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2201                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2202         eui[1] = 0;
2203         eui[2] = 0x5E;
2204         eui[3] = 0xFE;
2205         memcpy(eui + 4, &addr, 4);
2206         return 0;
2207 }
2208
2209 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2210 {
2211         if (dev->priv_flags & IFF_ISATAP)
2212                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2213         return -1;
2214 }
2215
2216 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2217 {
2218         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2219 }
2220
2221 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2222 {
2223         memcpy(eui, dev->perm_addr, 3);
2224         memcpy(eui + 5, dev->perm_addr + 3, 3);
2225         eui[3] = 0xFF;
2226         eui[4] = 0xFE;
2227         eui[0] ^= 2;
2228         return 0;
2229 }
2230
2231 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2232 {
2233         switch (dev->type) {
2234         case ARPHRD_ETHER:
2235         case ARPHRD_FDDI:
2236                 return addrconf_ifid_eui48(eui, dev);
2237         case ARPHRD_ARCNET:
2238                 return addrconf_ifid_arcnet(eui, dev);
2239         case ARPHRD_INFINIBAND:
2240                 return addrconf_ifid_infiniband(eui, dev);
2241         case ARPHRD_SIT:
2242                 return addrconf_ifid_sit(eui, dev);
2243         case ARPHRD_IPGRE:
2244         case ARPHRD_TUNNEL:
2245                 return addrconf_ifid_gre(eui, dev);
2246         case ARPHRD_6LOWPAN:
2247                 return addrconf_ifid_6lowpan(eui, dev);
2248         case ARPHRD_IEEE1394:
2249                 return addrconf_ifid_ieee1394(eui, dev);
2250         case ARPHRD_TUNNEL6:
2251         case ARPHRD_IP6GRE:
2252         case ARPHRD_RAWIP:
2253                 return addrconf_ifid_ip6tnl(eui, dev);
2254         }
2255         return -1;
2256 }
2257
2258 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2259 {
2260         int err = -1;
2261         struct inet6_ifaddr *ifp;
2262
2263         read_lock_bh(&idev->lock);
2264         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2265                 if (ifp->scope > IFA_LINK)
2266                         break;
2267                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2268                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2269                         err = 0;
2270                         break;
2271                 }
2272         }
2273         read_unlock_bh(&idev->lock);
2274         return err;
2275 }
2276
2277 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2278 static void ipv6_regen_rndid(struct inet6_dev *idev)
2279 {
2280 regen:
2281         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2282         idev->rndid[0] &= ~0x02;
2283
2284         /*
2285          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2286          * check if generated address is not inappropriate
2287          *
2288          *  - Reserved subnet anycast (RFC 2526)
2289          *      11111101 11....11 1xxxxxxx
2290          *  - ISATAP (RFC4214) 6.1
2291          *      00-00-5E-FE-xx-xx-xx-xx
2292          *  - value 0
2293          *  - XXX: already assigned to an address on the device
2294          */
2295         if (idev->rndid[0] == 0xfd &&
2296             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2297             (idev->rndid[7]&0x80))
2298                 goto regen;
2299         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2300                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2301                         goto regen;
2302                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2303                         goto regen;
2304         }
2305 }
2306
2307 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2308 {
2309         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2310                 ipv6_regen_rndid(idev);
2311 }
2312
2313 /*
2314  *      Add prefix route.
2315  */
2316
2317 static void
2318 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2319                       struct net_device *dev, unsigned long expires,
2320                       u32 flags, gfp_t gfp_flags)
2321 {
2322         struct fib6_config cfg = {
2323                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2324                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2325                 .fc_ifindex = dev->ifindex,
2326                 .fc_expires = expires,
2327                 .fc_dst_len = plen,
2328                 .fc_flags = RTF_UP | flags,
2329                 .fc_nlinfo.nl_net = dev_net(dev),
2330                 .fc_protocol = RTPROT_KERNEL,
2331                 .fc_type = RTN_UNICAST,
2332         };
2333
2334         cfg.fc_dst = *pfx;
2335
2336         /* Prevent useless cloning on PtP SIT.
2337            This thing is done here expecting that the whole
2338            class of non-broadcast devices need not cloning.
2339          */
2340 #if IS_ENABLED(CONFIG_IPV6_SIT)
2341         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2342                 cfg.fc_flags |= RTF_NONEXTHOP;
2343 #endif
2344
2345         ip6_route_add(&cfg, gfp_flags, NULL);
2346 }
2347
2348
2349 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2350                                                   int plen,
2351                                                   const struct net_device *dev,
2352                                                   u32 flags, u32 noflags)
2353 {
2354         struct fib6_node *fn;
2355         struct fib6_info *rt = NULL;
2356         struct fib6_table *table;
2357         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2358
2359         table = fib6_get_table(dev_net(dev), tb_id);
2360         if (!table)
2361                 return NULL;
2362
2363         rcu_read_lock();
2364         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2365         if (!fn)
2366                 goto out;
2367
2368         for_each_fib6_node_rt_rcu(fn) {
2369                 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2370                         continue;
2371                 if ((rt->fib6_flags & flags) != flags)
2372                         continue;
2373                 if ((rt->fib6_flags & noflags) != 0)
2374                         continue;
2375                 if (!fib6_info_hold_safe(rt))
2376                         continue;
2377                 break;
2378         }
2379 out:
2380         rcu_read_unlock();
2381         return rt;
2382 }
2383
2384
2385 /* Create "default" multicast route to the interface */
2386
2387 static void addrconf_add_mroute(struct net_device *dev)
2388 {
2389         struct fib6_config cfg = {
2390                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2391                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2392                 .fc_ifindex = dev->ifindex,
2393                 .fc_dst_len = 8,
2394                 .fc_flags = RTF_UP,
2395                 .fc_type = RTN_UNICAST,
2396                 .fc_nlinfo.nl_net = dev_net(dev),
2397         };
2398
2399         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2400
2401         ip6_route_add(&cfg, GFP_ATOMIC, NULL);
2402 }
2403
2404 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2405 {
2406         struct inet6_dev *idev;
2407
2408         ASSERT_RTNL();
2409
2410         idev = ipv6_find_idev(dev);
2411         if (!idev)
2412                 return ERR_PTR(-ENOBUFS);
2413
2414         if (idev->cnf.disable_ipv6)
2415                 return ERR_PTR(-EACCES);
2416
2417         /* Add default multicast route */
2418         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2419                 addrconf_add_mroute(dev);
2420
2421         return idev;
2422 }
2423
2424 static void manage_tempaddrs(struct inet6_dev *idev,
2425                              struct inet6_ifaddr *ifp,
2426                              __u32 valid_lft, __u32 prefered_lft,
2427                              bool create, unsigned long now)
2428 {
2429         u32 flags;
2430         struct inet6_ifaddr *ift;
2431
2432         read_lock_bh(&idev->lock);
2433         /* update all temporary addresses in the list */
2434         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2435                 int age, max_valid, max_prefered;
2436
2437                 if (ifp != ift->ifpub)
2438                         continue;
2439
2440                 /* RFC 4941 section 3.3:
2441                  * If a received option will extend the lifetime of a public
2442                  * address, the lifetimes of temporary addresses should
2443                  * be extended, subject to the overall constraint that no
2444                  * temporary addresses should ever remain "valid" or "preferred"
2445                  * for a time longer than (TEMP_VALID_LIFETIME) or
2446                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2447                  */
2448                 age = (now - ift->cstamp) / HZ;
2449                 max_valid = idev->cnf.temp_valid_lft - age;
2450                 if (max_valid < 0)
2451                         max_valid = 0;
2452
2453                 max_prefered = idev->cnf.temp_prefered_lft -
2454                                idev->desync_factor - age;
2455                 if (max_prefered < 0)
2456                         max_prefered = 0;
2457
2458                 if (valid_lft > max_valid)
2459                         valid_lft = max_valid;
2460
2461                 if (prefered_lft > max_prefered)
2462                         prefered_lft = max_prefered;
2463
2464                 spin_lock(&ift->lock);
2465                 flags = ift->flags;
2466                 ift->valid_lft = valid_lft;
2467                 ift->prefered_lft = prefered_lft;
2468                 ift->tstamp = now;
2469                 if (prefered_lft > 0)
2470                         ift->flags &= ~IFA_F_DEPRECATED;
2471
2472                 spin_unlock(&ift->lock);
2473                 if (!(flags&IFA_F_TENTATIVE))
2474                         ipv6_ifa_notify(0, ift);
2475         }
2476
2477         if ((create || list_empty(&idev->tempaddr_list)) &&
2478             idev->cnf.use_tempaddr > 0) {
2479                 /* When a new public address is created as described
2480                  * in [ADDRCONF], also create a new temporary address.
2481                  * Also create a temporary address if it's enabled but
2482                  * no temporary address currently exists.
2483                  */
2484                 read_unlock_bh(&idev->lock);
2485                 ipv6_create_tempaddr(ifp, NULL, false);
2486         } else {
2487                 read_unlock_bh(&idev->lock);
2488         }
2489 }
2490
2491 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2492 {
2493         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2494                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2495 }
2496
2497 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2498                                  const struct prefix_info *pinfo,
2499                                  struct inet6_dev *in6_dev,
2500                                  const struct in6_addr *addr, int addr_type,
2501                                  u32 addr_flags, bool sllao, bool tokenized,
2502                                  __u32 valid_lft, u32 prefered_lft)
2503 {
2504         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2505         int create = 0, update_lft = 0;
2506
2507         if (!ifp && valid_lft) {
2508                 int max_addresses = in6_dev->cnf.max_addresses;
2509                 struct ifa6_config cfg = {
2510                         .pfx = addr,
2511                         .plen = pinfo->prefix_len,
2512                         .ifa_flags = addr_flags,
2513                         .valid_lft = valid_lft,
2514                         .preferred_lft = prefered_lft,
2515                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2516                 };
2517
2518 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2519                 if ((net->ipv6.devconf_all->optimistic_dad ||
2520                      in6_dev->cnf.optimistic_dad) &&
2521                     !net->ipv6.devconf_all->forwarding && sllao)
2522                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2523 #endif
2524
2525                 /* Do not allow to create too much of autoconfigured
2526                  * addresses; this would be too easy way to crash kernel.
2527                  */
2528                 if (!max_addresses ||
2529                     ipv6_count_addresses(in6_dev) < max_addresses)
2530                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2531
2532                 if (IS_ERR_OR_NULL(ifp))
2533                         return -1;
2534
2535                 create = 1;
2536                 spin_lock_bh(&ifp->lock);
2537                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2538                 ifp->cstamp = jiffies;
2539                 ifp->tokenized = tokenized;
2540                 spin_unlock_bh(&ifp->lock);
2541                 addrconf_dad_start(ifp);
2542         }
2543
2544         if (ifp) {
2545                 u32 flags;
2546                 unsigned long now;
2547                 u32 stored_lft;
2548
2549                 /* update lifetime (RFC2462 5.5.3 e) */
2550                 spin_lock_bh(&ifp->lock);
2551                 now = jiffies;
2552                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2553                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2554                 else
2555                         stored_lft = 0;
2556                 if (!create && stored_lft) {
2557                         const u32 minimum_lft = min_t(u32,
2558                                 stored_lft, MIN_VALID_LIFETIME);
2559                         valid_lft = max(valid_lft, minimum_lft);
2560
2561                         /* RFC4862 Section 5.5.3e:
2562                          * "Note that the preferred lifetime of the
2563                          *  corresponding address is always reset to
2564                          *  the Preferred Lifetime in the received
2565                          *  Prefix Information option, regardless of
2566                          *  whether the valid lifetime is also reset or
2567                          *  ignored."
2568                          *
2569                          * So we should always update prefered_lft here.
2570                          */
2571                         update_lft = 1;
2572                 }
2573
2574                 if (update_lft) {
2575                         ifp->valid_lft = valid_lft;
2576                         ifp->prefered_lft = prefered_lft;
2577                         ifp->tstamp = now;
2578                         flags = ifp->flags;
2579                         ifp->flags &= ~IFA_F_DEPRECATED;
2580                         spin_unlock_bh(&ifp->lock);
2581
2582                         if (!(flags&IFA_F_TENTATIVE))
2583                                 ipv6_ifa_notify(0, ifp);
2584                 } else
2585                         spin_unlock_bh(&ifp->lock);
2586
2587                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2588                                  create, now);
2589
2590                 in6_ifa_put(ifp);
2591                 addrconf_verify();
2592         }
2593
2594         return 0;
2595 }
2596 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2597
2598 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2599 {
2600         struct prefix_info *pinfo;
2601         __u32 valid_lft;
2602         __u32 prefered_lft;
2603         int addr_type, err;
2604         u32 addr_flags = 0;
2605         struct inet6_dev *in6_dev;
2606         struct net *net = dev_net(dev);
2607
2608         pinfo = (struct prefix_info *) opt;
2609
2610         if (len < sizeof(struct prefix_info)) {
2611                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2612                 return;
2613         }
2614
2615         /*
2616          *      Validation checks ([ADDRCONF], page 19)
2617          */
2618
2619         addr_type = ipv6_addr_type(&pinfo->prefix);
2620
2621         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2622                 return;
2623
2624         valid_lft = ntohl(pinfo->valid);
2625         prefered_lft = ntohl(pinfo->prefered);
2626
2627         if (prefered_lft > valid_lft) {
2628                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2629                 return;
2630         }
2631
2632         in6_dev = in6_dev_get(dev);
2633
2634         if (!in6_dev) {
2635                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2636                                     dev->name);
2637                 return;
2638         }
2639
2640         /*
2641          *      Two things going on here:
2642          *      1) Add routes for on-link prefixes
2643          *      2) Configure prefixes with the auto flag set
2644          */
2645
2646         if (pinfo->onlink) {
2647                 struct fib6_info *rt;
2648                 unsigned long rt_expires;
2649
2650                 /* Avoid arithmetic overflow. Really, we could
2651                  * save rt_expires in seconds, likely valid_lft,
2652                  * but it would require division in fib gc, that it
2653                  * not good.
2654                  */
2655                 if (HZ > USER_HZ)
2656                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2657                 else
2658                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2659
2660                 if (addrconf_finite_timeout(rt_expires))
2661                         rt_expires *= HZ;
2662
2663                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2664                                                pinfo->prefix_len,
2665                                                dev,
2666                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2667                                                RTF_GATEWAY | RTF_DEFAULT);
2668
2669                 if (rt) {
2670                         /* Autoconf prefix route */
2671                         if (valid_lft == 0) {
2672                                 ip6_del_rt(net, rt);
2673                                 rt = NULL;
2674                         } else if (addrconf_finite_timeout(rt_expires)) {
2675                                 /* not infinity */
2676                                 fib6_set_expires(rt, jiffies + rt_expires);
2677                         } else {
2678                                 fib6_clean_expires(rt);
2679                         }
2680                 } else if (valid_lft) {
2681                         clock_t expires = 0;
2682                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2683                         if (addrconf_finite_timeout(rt_expires)) {
2684                                 /* not infinity */
2685                                 flags |= RTF_EXPIRES;
2686                                 expires = jiffies_to_clock_t(rt_expires);
2687                         }
2688                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2689                                               0, dev, expires, flags,
2690                                               GFP_ATOMIC);
2691                 }
2692                 fib6_info_release(rt);
2693         }
2694
2695         /* Try to figure out our local address for this prefix */
2696
2697         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2698                 struct in6_addr addr;
2699                 bool tokenized = false, dev_addr_generated = false;
2700
2701                 if (pinfo->prefix_len == 64) {
2702                         memcpy(&addr, &pinfo->prefix, 8);
2703
2704                         if (!ipv6_addr_any(&in6_dev->token)) {
2705                                 read_lock_bh(&in6_dev->lock);
2706                                 memcpy(addr.s6_addr + 8,
2707                                        in6_dev->token.s6_addr + 8, 8);
2708                                 read_unlock_bh(&in6_dev->lock);
2709                                 tokenized = true;
2710                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2711                                    !ipv6_generate_stable_address(&addr, 0,
2712                                                                  in6_dev)) {
2713                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2714                                 goto ok;
2715                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2716                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2717                                 goto put;
2718                         } else {
2719                                 dev_addr_generated = true;
2720                         }
2721                         goto ok;
2722                 }
2723                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2724                                     pinfo->prefix_len);
2725                 goto put;
2726
2727 ok:
2728                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2729                                                    &addr, addr_type,
2730                                                    addr_flags, sllao,
2731                                                    tokenized, valid_lft,
2732                                                    prefered_lft);
2733                 if (err)
2734                         goto put;
2735
2736                 /* Ignore error case here because previous prefix add addr was
2737                  * successful which will be notified.
2738                  */
2739                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2740                                               addr_type, addr_flags, sllao,
2741                                               tokenized, valid_lft,
2742                                               prefered_lft,
2743                                               dev_addr_generated);
2744         }
2745         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2746 put:
2747         in6_dev_put(in6_dev);
2748 }
2749
2750 /*
2751  *      Set destination address.
2752  *      Special case for SIT interfaces where we create a new "virtual"
2753  *      device.
2754  */
2755 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2756 {
2757         struct in6_ifreq ireq;
2758         struct net_device *dev;
2759         int err = -EINVAL;
2760
2761         rtnl_lock();
2762
2763         err = -EFAULT;
2764         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2765                 goto err_exit;
2766
2767         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2768
2769         err = -ENODEV;
2770         if (!dev)
2771                 goto err_exit;
2772
2773 #if IS_ENABLED(CONFIG_IPV6_SIT)
2774         if (dev->type == ARPHRD_SIT) {
2775                 const struct net_device_ops *ops = dev->netdev_ops;
2776                 struct ifreq ifr;
2777                 struct ip_tunnel_parm p;
2778
2779                 err = -EADDRNOTAVAIL;
2780                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2781                         goto err_exit;
2782
2783                 memset(&p, 0, sizeof(p));
2784                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2785                 p.iph.saddr = 0;
2786                 p.iph.version = 4;
2787                 p.iph.ihl = 5;
2788                 p.iph.protocol = IPPROTO_IPV6;
2789                 p.iph.ttl = 64;
2790                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2791
2792                 if (ops->ndo_do_ioctl) {
2793                         mm_segment_t oldfs = get_fs();
2794
2795                         set_fs(KERNEL_DS);
2796                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2797                         set_fs(oldfs);
2798                 } else
2799                         err = -EOPNOTSUPP;
2800
2801                 if (err == 0) {
2802                         err = -ENOBUFS;
2803                         dev = __dev_get_by_name(net, p.name);
2804                         if (!dev)
2805                                 goto err_exit;
2806                         err = dev_open(dev);
2807                 }
2808         }
2809 #endif
2810
2811 err_exit:
2812         rtnl_unlock();
2813         return err;
2814 }
2815
2816 static int ipv6_mc_config(struct sock *sk, bool join,
2817                           const struct in6_addr *addr, int ifindex)
2818 {
2819         int ret;
2820
2821         ASSERT_RTNL();
2822
2823         lock_sock(sk);
2824         if (join)
2825                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2826         else
2827                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2828         release_sock(sk);
2829
2830         return ret;
2831 }
2832
2833 /*
2834  *      Manual configuration of address on an interface
2835  */
2836 static int inet6_addr_add(struct net *net, int ifindex,
2837                           struct ifa6_config *cfg,
2838                           struct netlink_ext_ack *extack)
2839 {
2840         struct inet6_ifaddr *ifp;
2841         struct inet6_dev *idev;
2842         struct net_device *dev;
2843         unsigned long timeout;
2844         clock_t expires;
2845         u32 flags;
2846
2847         ASSERT_RTNL();
2848
2849         if (cfg->plen > 128)
2850                 return -EINVAL;
2851
2852         /* check the lifetime */
2853         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2854                 return -EINVAL;
2855
2856         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2857                 return -EINVAL;
2858
2859         dev = __dev_get_by_index(net, ifindex);
2860         if (!dev)
2861                 return -ENODEV;
2862
2863         idev = addrconf_add_dev(dev);
2864         if (IS_ERR(idev))
2865                 return PTR_ERR(idev);
2866
2867         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2868                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2869                                          true, cfg->pfx, ifindex);
2870
2871                 if (ret < 0)
2872                         return ret;
2873         }
2874
2875         cfg->scope = ipv6_addr_scope(cfg->pfx);
2876
2877         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2878         if (addrconf_finite_timeout(timeout)) {
2879                 expires = jiffies_to_clock_t(timeout * HZ);
2880                 cfg->valid_lft = timeout;
2881                 flags = RTF_EXPIRES;
2882         } else {
2883                 expires = 0;
2884                 flags = 0;
2885                 cfg->ifa_flags |= IFA_F_PERMANENT;
2886         }
2887
2888         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2889         if (addrconf_finite_timeout(timeout)) {
2890                 if (timeout == 0)
2891                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2892                 cfg->preferred_lft = timeout;
2893         }
2894
2895         ifp = ipv6_add_addr(idev, cfg, true, extack);
2896         if (!IS_ERR(ifp)) {
2897                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2898                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2899                                               ifp->rt_priority, dev, expires,
2900                                               flags, GFP_KERNEL);
2901                 }
2902
2903                 /* Send a netlink notification if DAD is enabled and
2904                  * optimistic flag is not set
2905                  */
2906                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2907                         ipv6_ifa_notify(0, ifp);
2908                 /*
2909                  * Note that section 3.1 of RFC 4429 indicates
2910                  * that the Optimistic flag should not be set for
2911                  * manually configured addresses
2912                  */
2913                 addrconf_dad_start(ifp);
2914                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2915                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2916                                          cfg->preferred_lft, true, jiffies);
2917                 in6_ifa_put(ifp);
2918                 addrconf_verify_rtnl();
2919                 return 0;
2920         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2921                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2922                                cfg->pfx, ifindex);
2923         }
2924
2925         return PTR_ERR(ifp);
2926 }
2927
2928 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2929                           const struct in6_addr *pfx, unsigned int plen)
2930 {
2931         struct inet6_ifaddr *ifp;
2932         struct inet6_dev *idev;
2933         struct net_device *dev;
2934
2935         if (plen > 128)
2936                 return -EINVAL;
2937
2938         dev = __dev_get_by_index(net, ifindex);
2939         if (!dev)
2940                 return -ENODEV;
2941
2942         idev = __in6_dev_get(dev);
2943         if (!idev)
2944                 return -ENXIO;
2945
2946         read_lock_bh(&idev->lock);
2947         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2948                 if (ifp->prefix_len == plen &&
2949                     ipv6_addr_equal(pfx, &ifp->addr)) {
2950                         in6_ifa_hold(ifp);
2951                         read_unlock_bh(&idev->lock);
2952
2953                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2954                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2955                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2956                                                  jiffies);
2957                         ipv6_del_addr(ifp);
2958                         addrconf_verify_rtnl();
2959                         if (ipv6_addr_is_multicast(pfx)) {
2960                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2961                                                false, pfx, dev->ifindex);
2962                         }
2963                         return 0;
2964                 }
2965         }
2966         read_unlock_bh(&idev->lock);
2967         return -EADDRNOTAVAIL;
2968 }
2969
2970
2971 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2972 {
2973         struct ifa6_config cfg = {
2974                 .ifa_flags = IFA_F_PERMANENT,
2975                 .preferred_lft = INFINITY_LIFE_TIME,
2976                 .valid_lft = INFINITY_LIFE_TIME,
2977         };
2978         struct in6_ifreq ireq;
2979         int err;
2980
2981         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2982                 return -EPERM;
2983
2984         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2985                 return -EFAULT;
2986
2987         cfg.pfx = &ireq.ifr6_addr;
2988         cfg.plen = ireq.ifr6_prefixlen;
2989
2990         rtnl_lock();
2991         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
2992         rtnl_unlock();
2993         return err;
2994 }
2995
2996 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2997 {
2998         struct in6_ifreq ireq;
2999         int err;
3000
3001         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3002                 return -EPERM;
3003
3004         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3005                 return -EFAULT;
3006
3007         rtnl_lock();
3008         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3009                              ireq.ifr6_prefixlen);
3010         rtnl_unlock();
3011         return err;
3012 }
3013
3014 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3015                      int plen, int scope)
3016 {
3017         struct inet6_ifaddr *ifp;
3018         struct ifa6_config cfg = {
3019                 .pfx = addr,
3020                 .plen = plen,
3021                 .ifa_flags = IFA_F_PERMANENT,
3022                 .valid_lft = INFINITY_LIFE_TIME,
3023                 .preferred_lft = INFINITY_LIFE_TIME,
3024                 .scope = scope
3025         };
3026
3027         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3028         if (!IS_ERR(ifp)) {
3029                 spin_lock_bh(&ifp->lock);
3030                 ifp->flags &= ~IFA_F_TENTATIVE;
3031                 spin_unlock_bh(&ifp->lock);
3032                 rt_genid_bump_ipv6(dev_net(idev->dev));
3033                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3034                 in6_ifa_put(ifp);
3035         }
3036 }
3037
3038 #if IS_ENABLED(CONFIG_IPV6_SIT)
3039 static void sit_add_v4_addrs(struct inet6_dev *idev)
3040 {
3041         struct in6_addr addr;
3042         struct net_device *dev;
3043         struct net *net = dev_net(idev->dev);
3044         int scope, plen;
3045         u32 pflags = 0;
3046
3047         ASSERT_RTNL();
3048
3049         memset(&addr, 0, sizeof(struct in6_addr));
3050         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3051
3052         if (idev->dev->flags&IFF_POINTOPOINT) {
3053                 addr.s6_addr32[0] = htonl(0xfe800000);
3054                 scope = IFA_LINK;
3055                 plen = 64;
3056         } else {
3057                 scope = IPV6_ADDR_COMPATv4;
3058                 plen = 96;
3059                 pflags |= RTF_NONEXTHOP;
3060         }
3061
3062         if (addr.s6_addr32[3]) {
3063                 add_addr(idev, &addr, plen, scope);
3064                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3065                                       GFP_ATOMIC);
3066                 return;
3067         }
3068
3069         for_each_netdev(net, dev) {
3070                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3071                 if (in_dev && (dev->flags & IFF_UP)) {
3072                         struct in_ifaddr *ifa;
3073
3074                         int flag = scope;
3075
3076                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3077
3078                                 addr.s6_addr32[3] = ifa->ifa_local;
3079
3080                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3081                                         continue;
3082                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3083                                         if (idev->dev->flags&IFF_POINTOPOINT)
3084                                                 continue;
3085                                         flag |= IFA_HOST;
3086                                 }
3087
3088                                 add_addr(idev, &addr, plen, flag);
3089                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3090                                                       0, pflags, GFP_ATOMIC);
3091                         }
3092                 }
3093         }
3094 }
3095 #endif
3096
3097 static void init_loopback(struct net_device *dev)
3098 {
3099         struct inet6_dev  *idev;
3100
3101         /* ::1 */
3102
3103         ASSERT_RTNL();
3104
3105         idev = ipv6_find_idev(dev);
3106         if (!idev) {
3107                 pr_debug("%s: add_dev failed\n", __func__);
3108                 return;
3109         }
3110
3111         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3112 }
3113
3114 void addrconf_add_linklocal(struct inet6_dev *idev,
3115                             const struct in6_addr *addr, u32 flags)
3116 {
3117         struct ifa6_config cfg = {
3118                 .pfx = addr,
3119                 .plen = 64,
3120                 .ifa_flags = flags | IFA_F_PERMANENT,
3121                 .valid_lft = INFINITY_LIFE_TIME,
3122                 .preferred_lft = INFINITY_LIFE_TIME,
3123                 .scope = IFA_LINK
3124         };
3125         struct inet6_ifaddr *ifp;
3126
3127 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3128         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3129              idev->cnf.optimistic_dad) &&
3130             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3131                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3132 #endif
3133
3134         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3135         if (!IS_ERR(ifp)) {
3136                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3137                                       0, 0, GFP_ATOMIC);
3138                 addrconf_dad_start(ifp);
3139                 in6_ifa_put(ifp);
3140         }
3141 }
3142 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3143
3144 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3145 {
3146         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3147                 return true;
3148
3149         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3150             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3151                 return true;
3152
3153         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3154             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3155                 return true;
3156
3157         return false;
3158 }
3159
3160 static int ipv6_generate_stable_address(struct in6_addr *address,
3161                                         u8 dad_count,
3162                                         const struct inet6_dev *idev)
3163 {
3164         static DEFINE_SPINLOCK(lock);
3165         static __u32 digest[SHA_DIGEST_WORDS];
3166         static __u32 workspace[SHA_WORKSPACE_WORDS];
3167
3168         static union {
3169                 char __data[SHA_MESSAGE_BYTES];
3170                 struct {
3171                         struct in6_addr secret;
3172                         __be32 prefix[2];
3173                         unsigned char hwaddr[MAX_ADDR_LEN];
3174                         u8 dad_count;
3175                 } __packed;
3176         } data;
3177
3178         struct in6_addr secret;
3179         struct in6_addr temp;
3180         struct net *net = dev_net(idev->dev);
3181
3182         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3183
3184         if (idev->cnf.stable_secret.initialized)
3185                 secret = idev->cnf.stable_secret.secret;
3186         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3187                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3188         else
3189                 return -1;
3190
3191 retry:
3192         spin_lock_bh(&lock);
3193
3194         sha_init(digest);
3195         memset(&data, 0, sizeof(data));
3196         memset(workspace, 0, sizeof(workspace));
3197         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3198         data.prefix[0] = address->s6_addr32[0];
3199         data.prefix[1] = address->s6_addr32[1];
3200         data.secret = secret;
3201         data.dad_count = dad_count;
3202
3203         sha_transform(digest, data.__data, workspace);
3204
3205         temp = *address;
3206         temp.s6_addr32[2] = (__force __be32)digest[0];
3207         temp.s6_addr32[3] = (__force __be32)digest[1];
3208
3209         spin_unlock_bh(&lock);
3210
3211         if (ipv6_reserved_interfaceid(temp)) {
3212                 dad_count++;
3213                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3214                         return -1;
3215                 goto retry;
3216         }
3217
3218         *address = temp;
3219         return 0;
3220 }
3221
3222 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3223 {
3224         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3225
3226         if (s->initialized)
3227                 return;
3228         s = &idev->cnf.stable_secret;
3229         get_random_bytes(&s->secret, sizeof(s->secret));
3230         s->initialized = true;
3231 }
3232
3233 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3234 {
3235         struct in6_addr addr;
3236
3237         /* no link local addresses on L3 master devices */
3238         if (netif_is_l3_master(idev->dev))
3239                 return;
3240
3241         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3242
3243         switch (idev->cnf.addr_gen_mode) {
3244         case IN6_ADDR_GEN_MODE_RANDOM:
3245                 ipv6_gen_mode_random_init(idev);
3246                 /* fallthrough */
3247         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3248                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3249                         addrconf_add_linklocal(idev, &addr,
3250                                                IFA_F_STABLE_PRIVACY);
3251                 else if (prefix_route)
3252                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3253                                               0, 0, GFP_KERNEL);
3254                 break;
3255         case IN6_ADDR_GEN_MODE_EUI64:
3256                 /* addrconf_add_linklocal also adds a prefix_route and we
3257                  * only need to care about prefix routes if ipv6_generate_eui64
3258                  * couldn't generate one.
3259                  */
3260                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3261                         addrconf_add_linklocal(idev, &addr, 0);
3262                 else if (prefix_route)
3263                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3264                                               0, 0, GFP_KERNEL);
3265                 break;
3266         case IN6_ADDR_GEN_MODE_NONE:
3267         default:
3268                 /* will not add any link local address */
3269                 break;
3270         }
3271 }
3272
3273 static void addrconf_dev_config(struct net_device *dev)
3274 {
3275         struct inet6_dev *idev;
3276
3277         ASSERT_RTNL();
3278
3279         if ((dev->type != ARPHRD_ETHER) &&
3280             (dev->type != ARPHRD_FDDI) &&
3281             (dev->type != ARPHRD_ARCNET) &&
3282             (dev->type != ARPHRD_INFINIBAND) &&
3283             (dev->type != ARPHRD_IEEE1394) &&
3284             (dev->type != ARPHRD_TUNNEL6) &&
3285             (dev->type != ARPHRD_6LOWPAN) &&
3286             (dev->type != ARPHRD_IP6GRE) &&
3287             (dev->type != ARPHRD_IPGRE) &&
3288             (dev->type != ARPHRD_TUNNEL) &&
3289             (dev->type != ARPHRD_NONE) &&
3290             (dev->type != ARPHRD_RAWIP)) {
3291                 /* Alas, we support only Ethernet autoconfiguration. */
3292                 return;
3293         }
3294
3295         idev = addrconf_add_dev(dev);
3296         if (IS_ERR(idev))
3297                 return;
3298
3299         /* this device type has no EUI support */
3300         if (dev->type == ARPHRD_NONE &&
3301             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3302                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3303
3304         addrconf_addr_gen(idev, false);
3305 }
3306
3307 #if IS_ENABLED(CONFIG_IPV6_SIT)
3308 static void addrconf_sit_config(struct net_device *dev)
3309 {
3310         struct inet6_dev *idev;
3311
3312         ASSERT_RTNL();
3313
3314         /*
3315          * Configure the tunnel with one of our IPv4
3316          * addresses... we should configure all of
3317          * our v4 addrs in the tunnel
3318          */
3319
3320         idev = ipv6_find_idev(dev);
3321         if (!idev) {
3322                 pr_debug("%s: add_dev failed\n", __func__);
3323                 return;
3324         }
3325
3326         if (dev->priv_flags & IFF_ISATAP) {
3327                 addrconf_addr_gen(idev, false);
3328                 return;
3329         }
3330
3331         sit_add_v4_addrs(idev);
3332
3333         if (dev->flags&IFF_POINTOPOINT)
3334                 addrconf_add_mroute(dev);
3335 }
3336 #endif
3337
3338 #if IS_ENABLED(CONFIG_NET_IPGRE)
3339 static void addrconf_gre_config(struct net_device *dev)
3340 {
3341         struct inet6_dev *idev;
3342
3343         ASSERT_RTNL();
3344
3345         idev = ipv6_find_idev(dev);
3346         if (!idev) {
3347                 pr_debug("%s: add_dev failed\n", __func__);
3348                 return;
3349         }
3350
3351         addrconf_addr_gen(idev, true);
3352         if (dev->flags & IFF_POINTOPOINT)
3353                 addrconf_add_mroute(dev);
3354 }
3355 #endif
3356
3357 static int fixup_permanent_addr(struct net *net,
3358                                 struct inet6_dev *idev,
3359                                 struct inet6_ifaddr *ifp)
3360 {
3361         /* !fib6_node means the host route was removed from the
3362          * FIB, for example, if 'lo' device is taken down. In that
3363          * case regenerate the host route.
3364          */
3365         if (!ifp->rt || !ifp->rt->fib6_node) {
3366                 struct fib6_info *f6i, *prev;
3367
3368                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3369                                          GFP_ATOMIC);
3370                 if (IS_ERR(f6i))
3371                         return PTR_ERR(f6i);
3372
3373                 /* ifp->rt can be accessed outside of rtnl */
3374                 spin_lock(&ifp->lock);
3375                 prev = ifp->rt;
3376                 ifp->rt = f6i;
3377                 spin_unlock(&ifp->lock);
3378
3379                 fib6_info_release(prev);
3380         }
3381
3382         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3383                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3384                                       ifp->rt_priority, idev->dev, 0, 0,
3385                                       GFP_ATOMIC);
3386         }
3387
3388         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3389                 addrconf_dad_start(ifp);
3390
3391         return 0;
3392 }
3393
3394 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3395 {
3396         struct inet6_ifaddr *ifp, *tmp;
3397         struct inet6_dev *idev;
3398
3399         idev = __in6_dev_get(dev);
3400         if (!idev)
3401                 return;
3402
3403         write_lock_bh(&idev->lock);
3404
3405         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3406                 if ((ifp->flags & IFA_F_PERMANENT) &&
3407                     fixup_permanent_addr(net, idev, ifp) < 0) {
3408                         write_unlock_bh(&idev->lock);
3409                         in6_ifa_hold(ifp);
3410                         ipv6_del_addr(ifp);
3411                         write_lock_bh(&idev->lock);
3412
3413                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3414                                              idev->dev->name, &ifp->addr);
3415                 }
3416         }
3417
3418         write_unlock_bh(&idev->lock);
3419 }
3420
3421 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3422                            void *ptr)
3423 {
3424         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3425         struct netdev_notifier_changeupper_info *info;
3426         struct inet6_dev *idev = __in6_dev_get(dev);
3427         struct net *net = dev_net(dev);
3428         int run_pending = 0;
3429         int err;
3430
3431         switch (event) {
3432         case NETDEV_REGISTER:
3433                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3434                         idev = ipv6_add_dev(dev);
3435                         if (IS_ERR(idev))
3436                                 return notifier_from_errno(PTR_ERR(idev));
3437                 }
3438                 break;
3439
3440         case NETDEV_CHANGEMTU:
3441                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3442                 if (dev->mtu < IPV6_MIN_MTU) {
3443                         addrconf_ifdown(dev, dev != net->loopback_dev);
3444                         break;
3445                 }
3446
3447                 if (idev) {
3448                         rt6_mtu_change(dev, dev->mtu);
3449                         idev->cnf.mtu6 = dev->mtu;
3450                         break;
3451                 }
3452
3453                 /* allocate new idev */
3454                 idev = ipv6_add_dev(dev);
3455                 if (IS_ERR(idev))
3456                         break;
3457
3458                 /* device is still not ready */
3459                 if (!(idev->if_flags & IF_READY))
3460                         break;
3461
3462                 run_pending = 1;
3463
3464                 /* fall through */
3465
3466         case NETDEV_UP:
3467         case NETDEV_CHANGE:
3468                 if (dev->flags & IFF_SLAVE)
3469                         break;
3470
3471                 if (idev && idev->cnf.disable_ipv6)
3472                         break;
3473
3474                 if (event == NETDEV_UP) {
3475                         /* restore routes for permanent addresses */
3476                         addrconf_permanent_addr(net, dev);
3477
3478                         if (!addrconf_link_ready(dev)) {
3479                                 /* device is not ready yet. */
3480                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3481                                         dev->name);
3482                                 break;
3483                         }
3484
3485                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3486                                 idev = ipv6_add_dev(dev);
3487
3488                         if (!IS_ERR_OR_NULL(idev)) {
3489                                 idev->if_flags |= IF_READY;
3490                                 run_pending = 1;
3491                         }
3492                 } else if (event == NETDEV_CHANGE) {
3493                         if (!addrconf_link_ready(dev)) {
3494                                 /* device is still not ready. */
3495                                 rt6_sync_down_dev(dev, event);
3496                                 break;
3497                         }
3498
3499                         if (idev) {
3500                                 if (idev->if_flags & IF_READY) {
3501                                         /* device is already configured -
3502                                          * but resend MLD reports, we might
3503                                          * have roamed and need to update
3504                                          * multicast snooping switches
3505                                          */
3506                                         ipv6_mc_up(idev);
3507                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3508                                         break;
3509                                 }
3510                                 idev->if_flags |= IF_READY;
3511                         }
3512
3513                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3514                                 dev->name);
3515
3516                         run_pending = 1;
3517                 }
3518
3519                 switch (dev->type) {
3520 #if IS_ENABLED(CONFIG_IPV6_SIT)
3521                 case ARPHRD_SIT:
3522                         addrconf_sit_config(dev);
3523                         break;
3524 #endif
3525 #if IS_ENABLED(CONFIG_NET_IPGRE)
3526                 case ARPHRD_IPGRE:
3527                         addrconf_gre_config(dev);
3528                         break;
3529 #endif
3530                 case ARPHRD_LOOPBACK:
3531                         init_loopback(dev);
3532                         break;
3533
3534                 default:
3535                         addrconf_dev_config(dev);
3536                         break;
3537                 }
3538
3539                 if (!IS_ERR_OR_NULL(idev)) {
3540                         if (run_pending)
3541                                 addrconf_dad_run(idev);
3542
3543                         /* Device has an address by now */
3544                         rt6_sync_up(dev, RTNH_F_DEAD);
3545
3546                         /*
3547                          * If the MTU changed during the interface down,
3548                          * when the interface up, the changed MTU must be
3549                          * reflected in the idev as well as routers.
3550                          */
3551                         if (idev->cnf.mtu6 != dev->mtu &&
3552                             dev->mtu >= IPV6_MIN_MTU) {
3553                                 rt6_mtu_change(dev, dev->mtu);
3554                                 idev->cnf.mtu6 = dev->mtu;
3555                         }
3556                         idev->tstamp = jiffies;
3557                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3558
3559                         /*
3560                          * If the changed mtu during down is lower than
3561                          * IPV6_MIN_MTU stop IPv6 on this interface.
3562                          */
3563                         if (dev->mtu < IPV6_MIN_MTU)
3564                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3565                 }
3566                 break;
3567
3568         case NETDEV_DOWN:
3569         case NETDEV_UNREGISTER:
3570                 /*
3571                  *      Remove all addresses from this interface.
3572                  */
3573                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3574                 break;
3575
3576         case NETDEV_CHANGENAME:
3577                 if (idev) {
3578                         snmp6_unregister_dev(idev);
3579                         addrconf_sysctl_unregister(idev);
3580                         err = addrconf_sysctl_register(idev);
3581                         if (err)
3582                                 return notifier_from_errno(err);
3583                         err = snmp6_register_dev(idev);
3584                         if (err) {
3585                                 addrconf_sysctl_unregister(idev);
3586                                 return notifier_from_errno(err);
3587                         }
3588                 }
3589                 break;
3590
3591         case NETDEV_PRE_TYPE_CHANGE:
3592         case NETDEV_POST_TYPE_CHANGE:
3593                 if (idev)
3594                         addrconf_type_change(dev, event);
3595                 break;
3596
3597         case NETDEV_CHANGEUPPER:
3598                 info = ptr;
3599
3600                 /* flush all routes if dev is linked to or unlinked from
3601                  * an L3 master device (e.g., VRF)
3602                  */
3603                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3604                         addrconf_ifdown(dev, 0);
3605         }
3606
3607         return NOTIFY_OK;
3608 }
3609
3610 /*
3611  *      addrconf module should be notified of a device going up
3612  */
3613 static struct notifier_block ipv6_dev_notf = {
3614         .notifier_call = addrconf_notify,
3615         .priority = ADDRCONF_NOTIFY_PRIORITY,
3616 };
3617
3618 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3619 {
3620         struct inet6_dev *idev;
3621         ASSERT_RTNL();
3622
3623         idev = __in6_dev_get(dev);
3624
3625         if (event == NETDEV_POST_TYPE_CHANGE)
3626                 ipv6_mc_remap(idev);
3627         else if (event == NETDEV_PRE_TYPE_CHANGE)
3628                 ipv6_mc_unmap(idev);
3629 }
3630
3631 static bool addr_is_local(const struct in6_addr *addr)
3632 {
3633         return ipv6_addr_type(addr) &
3634                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3635 }
3636
3637 static int addrconf_ifdown(struct net_device *dev, int how)
3638 {
3639         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3640         struct net *net = dev_net(dev);
3641         struct inet6_dev *idev;
3642         struct inet6_ifaddr *ifa, *tmp;
3643         bool keep_addr = false;
3644         int state, i;
3645
3646         ASSERT_RTNL();
3647
3648         rt6_disable_ip(dev, event);
3649
3650         idev = __in6_dev_get(dev);
3651         if (!idev)
3652                 return -ENODEV;
3653
3654         /*
3655          * Step 1: remove reference to ipv6 device from parent device.
3656          *         Do not dev_put!
3657          */
3658         if (how) {
3659                 idev->dead = 1;
3660
3661                 /* protected by rtnl_lock */
3662                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3663
3664                 /* Step 1.5: remove snmp6 entry */
3665                 snmp6_unregister_dev(idev);
3666
3667         }
3668
3669         /* combine the user config with event to determine if permanent
3670          * addresses are to be removed from address hash table
3671          */
3672         if (!how && !idev->cnf.disable_ipv6) {
3673                 /* aggregate the system setting and interface setting */
3674                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3675
3676                 if (!_keep_addr)
3677                         _keep_addr = idev->cnf.keep_addr_on_down;
3678
3679                 keep_addr = (_keep_addr > 0);
3680         }
3681
3682         /* Step 2: clear hash table */
3683         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3684                 struct hlist_head *h = &inet6_addr_lst[i];
3685
3686                 spin_lock_bh(&addrconf_hash_lock);
3687 restart:
3688                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3689                         if (ifa->idev == idev) {
3690                                 addrconf_del_dad_work(ifa);
3691                                 /* combined flag + permanent flag decide if
3692                                  * address is retained on a down event
3693                                  */
3694                                 if (!keep_addr ||
3695                                     !(ifa->flags & IFA_F_PERMANENT) ||
3696                                     addr_is_local(&ifa->addr)) {
3697                                         hlist_del_init_rcu(&ifa->addr_lst);
3698                                         goto restart;
3699                                 }
3700                         }
3701                 }
3702                 spin_unlock_bh(&addrconf_hash_lock);
3703         }
3704
3705         write_lock_bh(&idev->lock);
3706
3707         addrconf_del_rs_timer(idev);
3708
3709         /* Step 2: clear flags for stateless addrconf */
3710         if (!how)
3711                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3712
3713         /* Step 3: clear tempaddr list */
3714         while (!list_empty(&idev->tempaddr_list)) {
3715                 ifa = list_first_entry(&idev->tempaddr_list,
3716                                        struct inet6_ifaddr, tmp_list);
3717                 list_del(&ifa->tmp_list);
3718                 write_unlock_bh(&idev->lock);
3719                 spin_lock_bh(&ifa->lock);
3720
3721                 if (ifa->ifpub) {
3722                         in6_ifa_put(ifa->ifpub);
3723                         ifa->ifpub = NULL;
3724                 }
3725                 spin_unlock_bh(&ifa->lock);
3726                 in6_ifa_put(ifa);
3727                 write_lock_bh(&idev->lock);
3728         }
3729
3730         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3731                 struct fib6_info *rt = NULL;
3732                 bool keep;
3733
3734                 addrconf_del_dad_work(ifa);
3735
3736                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3737                         !addr_is_local(&ifa->addr);
3738
3739                 write_unlock_bh(&idev->lock);
3740                 spin_lock_bh(&ifa->lock);
3741
3742                 if (keep) {
3743                         /* set state to skip the notifier below */
3744                         state = INET6_IFADDR_STATE_DEAD;
3745                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3746                         if (!(ifa->flags & IFA_F_NODAD))
3747                                 ifa->flags |= IFA_F_TENTATIVE;
3748
3749                         rt = ifa->rt;
3750                         ifa->rt = NULL;
3751                 } else {
3752                         state = ifa->state;
3753                         ifa->state = INET6_IFADDR_STATE_DEAD;
3754                 }
3755
3756                 spin_unlock_bh(&ifa->lock);
3757
3758                 if (rt)
3759                         ip6_del_rt(net, rt);
3760
3761                 if (state != INET6_IFADDR_STATE_DEAD) {
3762                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3763                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3764                 } else {
3765                         if (idev->cnf.forwarding)
3766                                 addrconf_leave_anycast(ifa);
3767                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3768                 }
3769
3770                 write_lock_bh(&idev->lock);
3771                 if (!keep) {
3772                         list_del_rcu(&ifa->if_list);
3773                         in6_ifa_put(ifa);
3774                 }
3775         }
3776
3777         write_unlock_bh(&idev->lock);
3778
3779         /* Step 5: Discard anycast and multicast list */
3780         if (how) {
3781                 ipv6_ac_destroy_dev(idev);
3782                 ipv6_mc_destroy_dev(idev);
3783         } else {
3784                 ipv6_mc_down(idev);
3785         }
3786
3787         idev->tstamp = jiffies;
3788
3789         /* Last: Shot the device (if unregistered) */
3790         if (how) {
3791                 addrconf_sysctl_unregister(idev);
3792                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3793                 neigh_ifdown(&nd_tbl, dev);
3794                 in6_dev_put(idev);
3795         }
3796         return 0;
3797 }
3798
3799 static void addrconf_rs_timer(struct timer_list *t)
3800 {
3801         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3802         struct net_device *dev = idev->dev;
3803         struct in6_addr lladdr;
3804
3805         write_lock(&idev->lock);
3806         if (idev->dead || !(idev->if_flags & IF_READY))
3807                 goto out;
3808
3809         if (!ipv6_accept_ra(idev))
3810                 goto out;
3811
3812         /* Announcement received after solicitation was sent */
3813         if (idev->if_flags & IF_RA_RCVD)
3814                 goto out;
3815
3816         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3817                 write_unlock(&idev->lock);
3818                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3819                         ndisc_send_rs(dev, &lladdr,
3820                                       &in6addr_linklocal_allrouters);
3821                 else
3822                         goto put;
3823
3824                 write_lock(&idev->lock);
3825                 idev->rs_interval = rfc3315_s14_backoff_update(
3826                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3827                 /* The wait after the last probe can be shorter */
3828                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3829                                              idev->cnf.rtr_solicits) ?
3830                                       idev->cnf.rtr_solicit_delay :
3831                                       idev->rs_interval);
3832         } else {
3833                 /*
3834                  * Note: we do not support deprecated "all on-link"
3835                  * assumption any longer.
3836                  */
3837                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3838         }
3839
3840 out:
3841         write_unlock(&idev->lock);
3842 put:
3843         in6_dev_put(idev);
3844 }
3845
3846 /*
3847  *      Duplicate Address Detection
3848  */
3849 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3850 {
3851         unsigned long rand_num;
3852         struct inet6_dev *idev = ifp->idev;
3853         u64 nonce;
3854
3855         if (ifp->flags & IFA_F_OPTIMISTIC)
3856                 rand_num = 0;
3857         else
3858                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3859
3860         nonce = 0;
3861         if (idev->cnf.enhanced_dad ||
3862             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3863                 do
3864                         get_random_bytes(&nonce, 6);
3865                 while (nonce == 0);
3866         }
3867         ifp->dad_nonce = nonce;
3868         ifp->dad_probes = idev->cnf.dad_transmits;
3869         addrconf_mod_dad_work(ifp, rand_num);
3870 }
3871
3872 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3873 {
3874         struct inet6_dev *idev = ifp->idev;
3875         struct net_device *dev = idev->dev;
3876         bool bump_id, notify = false;
3877         struct net *net;
3878
3879         addrconf_join_solict(dev, &ifp->addr);
3880
3881         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3882
3883         read_lock_bh(&idev->lock);
3884         spin_lock(&ifp->lock);
3885         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3886                 goto out;
3887
3888         net = dev_net(dev);
3889         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3890             (net->ipv6.devconf_all->accept_dad < 1 &&
3891              idev->cnf.accept_dad < 1) ||
3892             !(ifp->flags&IFA_F_TENTATIVE) ||
3893             ifp->flags & IFA_F_NODAD) {
3894                 bool send_na = false;
3895
3896                 if (ifp->flags & IFA_F_TENTATIVE &&
3897                     !(ifp->flags & IFA_F_OPTIMISTIC))
3898                         send_na = true;
3899                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3900                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3901                 spin_unlock(&ifp->lock);
3902                 read_unlock_bh(&idev->lock);
3903
3904                 addrconf_dad_completed(ifp, bump_id, send_na);
3905                 return;
3906         }
3907
3908         if (!(idev->if_flags & IF_READY)) {
3909                 spin_unlock(&ifp->lock);
3910                 read_unlock_bh(&idev->lock);
3911                 /*
3912                  * If the device is not ready:
3913                  * - keep it tentative if it is a permanent address.
3914                  * - otherwise, kill it.
3915                  */
3916                 in6_ifa_hold(ifp);
3917                 addrconf_dad_stop(ifp, 0);
3918                 return;
3919         }
3920
3921         /*
3922          * Optimistic nodes can start receiving
3923          * Frames right away
3924          */
3925         if (ifp->flags & IFA_F_OPTIMISTIC) {
3926                 ip6_ins_rt(net, ifp->rt);
3927                 if (ipv6_use_optimistic_addr(net, idev)) {
3928                         /* Because optimistic nodes can use this address,
3929                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3930                          */
3931                         notify = true;
3932                 }
3933         }
3934
3935         addrconf_dad_kick(ifp);
3936 out:
3937         spin_unlock(&ifp->lock);
3938         read_unlock_bh(&idev->lock);
3939         if (notify)
3940                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3941 }
3942
3943 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3944 {
3945         bool begin_dad = false;
3946
3947         spin_lock_bh(&ifp->lock);
3948         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3949                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3950                 begin_dad = true;
3951         }
3952         spin_unlock_bh(&ifp->lock);
3953
3954         if (begin_dad)
3955                 addrconf_mod_dad_work(ifp, 0);
3956 }
3957
3958 static void addrconf_dad_work(struct work_struct *w)
3959 {
3960         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3961                                                 struct inet6_ifaddr,
3962                                                 dad_work);
3963         struct inet6_dev *idev = ifp->idev;
3964         bool bump_id, disable_ipv6 = false;
3965         struct in6_addr mcaddr;
3966
3967         enum {
3968                 DAD_PROCESS,
3969                 DAD_BEGIN,
3970                 DAD_ABORT,
3971         } action = DAD_PROCESS;
3972
3973         rtnl_lock();
3974
3975         spin_lock_bh(&ifp->lock);
3976         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3977                 action = DAD_BEGIN;
3978                 ifp->state = INET6_IFADDR_STATE_DAD;
3979         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3980                 action = DAD_ABORT;
3981                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3982
3983                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3984                      idev->cnf.accept_dad > 1) &&
3985                     !idev->cnf.disable_ipv6 &&
3986                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3987                         struct in6_addr addr;
3988
3989                         addr.s6_addr32[0] = htonl(0xfe800000);
3990                         addr.s6_addr32[1] = 0;
3991
3992                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3993                             ipv6_addr_equal(&ifp->addr, &addr)) {
3994                                 /* DAD failed for link-local based on MAC */
3995                                 idev->cnf.disable_ipv6 = 1;
3996
3997                                 pr_info("%s: IPv6 being disabled!\n",
3998                                         ifp->idev->dev->name);
3999                                 disable_ipv6 = true;
4000                         }
4001                 }
4002         }
4003         spin_unlock_bh(&ifp->lock);
4004
4005         if (action == DAD_BEGIN) {
4006                 addrconf_dad_begin(ifp);
4007                 goto out;
4008         } else if (action == DAD_ABORT) {
4009                 in6_ifa_hold(ifp);
4010                 addrconf_dad_stop(ifp, 1);
4011                 if (disable_ipv6)
4012                         addrconf_ifdown(idev->dev, 0);
4013                 goto out;
4014         }
4015
4016         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4017                 goto out;
4018
4019         write_lock_bh(&idev->lock);
4020         if (idev->dead || !(idev->if_flags & IF_READY)) {
4021                 write_unlock_bh(&idev->lock);
4022                 goto out;
4023         }
4024
4025         spin_lock(&ifp->lock);
4026         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4027                 spin_unlock(&ifp->lock);
4028                 write_unlock_bh(&idev->lock);
4029                 goto out;
4030         }
4031
4032         if (ifp->dad_probes == 0) {
4033                 bool send_na = false;
4034
4035                 /*
4036                  * DAD was successful
4037                  */
4038
4039                 if (ifp->flags & IFA_F_TENTATIVE &&
4040                     !(ifp->flags & IFA_F_OPTIMISTIC))
4041                         send_na = true;
4042                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4043                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4044                 spin_unlock(&ifp->lock);
4045                 write_unlock_bh(&idev->lock);
4046
4047                 addrconf_dad_completed(ifp, bump_id, send_na);
4048
4049                 goto out;
4050         }
4051
4052         ifp->dad_probes--;
4053         addrconf_mod_dad_work(ifp,
4054                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4055         spin_unlock(&ifp->lock);
4056         write_unlock_bh(&idev->lock);
4057
4058         /* send a neighbour solicitation for our addr */
4059         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4060         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4061                       ifp->dad_nonce);
4062 out:
4063         in6_ifa_put(ifp);
4064         rtnl_unlock();
4065 }
4066
4067 /* ifp->idev must be at least read locked */
4068 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4069 {
4070         struct inet6_ifaddr *ifpiter;
4071         struct inet6_dev *idev = ifp->idev;
4072
4073         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4074                 if (ifpiter->scope > IFA_LINK)
4075                         break;
4076                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4077                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4078                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4079                     IFA_F_PERMANENT)
4080                         return false;
4081         }
4082         return true;
4083 }
4084
4085 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4086                                    bool send_na)
4087 {
4088         struct net_device *dev = ifp->idev->dev;
4089         struct in6_addr lladdr;
4090         bool send_rs, send_mld;
4091
4092         addrconf_del_dad_work(ifp);
4093
4094         /*
4095          *      Configure the address for reception. Now it is valid.
4096          */
4097
4098         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4099
4100         /* If added prefix is link local and we are prepared to process
4101            router advertisements, start sending router solicitations.
4102          */
4103
4104         read_lock_bh(&ifp->idev->lock);
4105         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4106         send_rs = send_mld &&
4107                   ipv6_accept_ra(ifp->idev) &&
4108                   ifp->idev->cnf.rtr_solicits != 0 &&
4109                   (dev->flags&IFF_LOOPBACK) == 0;
4110         read_unlock_bh(&ifp->idev->lock);
4111
4112         /* While dad is in progress mld report's source address is in6_addrany.
4113          * Resend with proper ll now.
4114          */
4115         if (send_mld)
4116                 ipv6_mc_dad_complete(ifp->idev);
4117
4118         /* send unsolicited NA if enabled */
4119         if (send_na &&
4120             (ifp->idev->cnf.ndisc_notify ||
4121              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4122                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4123                               /*router=*/ !!ifp->idev->cnf.forwarding,
4124                               /*solicited=*/ false, /*override=*/ true,
4125                               /*inc_opt=*/ true);
4126         }
4127
4128         if (send_rs) {
4129                 /*
4130                  *      If a host as already performed a random delay
4131                  *      [...] as part of DAD [...] there is no need
4132                  *      to delay again before sending the first RS
4133                  */
4134                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4135                         return;
4136                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4137
4138                 write_lock_bh(&ifp->idev->lock);
4139                 spin_lock(&ifp->lock);
4140                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4141                         ifp->idev->cnf.rtr_solicit_interval);
4142                 ifp->idev->rs_probes = 1;
4143                 ifp->idev->if_flags |= IF_RS_SENT;
4144                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4145                 spin_unlock(&ifp->lock);
4146                 write_unlock_bh(&ifp->idev->lock);
4147         }
4148
4149         if (bump_id)
4150                 rt_genid_bump_ipv6(dev_net(dev));
4151
4152         /* Make sure that a new temporary address will be created
4153          * before this temporary address becomes deprecated.
4154          */
4155         if (ifp->flags & IFA_F_TEMPORARY)
4156                 addrconf_verify_rtnl();
4157 }
4158
4159 static void addrconf_dad_run(struct inet6_dev *idev)
4160 {
4161         struct inet6_ifaddr *ifp;
4162
4163         read_lock_bh(&idev->lock);
4164         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4165                 spin_lock(&ifp->lock);
4166                 if (ifp->flags & IFA_F_TENTATIVE &&
4167                     ifp->state == INET6_IFADDR_STATE_DAD)
4168                         addrconf_dad_kick(ifp);
4169                 spin_unlock(&ifp->lock);
4170         }
4171         read_unlock_bh(&idev->lock);
4172 }
4173
4174 #ifdef CONFIG_PROC_FS
4175 struct if6_iter_state {
4176         struct seq_net_private p;
4177         int bucket;
4178         int offset;
4179 };
4180
4181 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4182 {
4183         struct if6_iter_state *state = seq->private;
4184         struct net *net = seq_file_net(seq);
4185         struct inet6_ifaddr *ifa = NULL;
4186         int p = 0;
4187
4188         /* initial bucket if pos is 0 */
4189         if (pos == 0) {
4190                 state->bucket = 0;
4191                 state->offset = 0;
4192         }
4193
4194         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4195                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4196                                          addr_lst) {
4197                         if (!net_eq(dev_net(ifa->idev->dev), net))
4198                                 continue;
4199                         /* sync with offset */
4200                         if (p < state->offset) {
4201                                 p++;
4202                                 continue;
4203                         }
4204                         state->offset++;
4205                         return ifa;
4206                 }
4207
4208                 /* prepare for next bucket */
4209                 state->offset = 0;
4210                 p = 0;
4211         }
4212         return NULL;
4213 }
4214
4215 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4216                                          struct inet6_ifaddr *ifa)
4217 {
4218         struct if6_iter_state *state = seq->private;
4219         struct net *net = seq_file_net(seq);
4220
4221         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4222                 if (!net_eq(dev_net(ifa->idev->dev), net))
4223                         continue;
4224                 state->offset++;
4225                 return ifa;
4226         }
4227
4228         while (++state->bucket < IN6_ADDR_HSIZE) {
4229                 state->offset = 0;
4230                 hlist_for_each_entry_rcu(ifa,
4231                                      &inet6_addr_lst[state->bucket], addr_lst) {
4232                         if (!net_eq(dev_net(ifa->idev->dev), net))
4233                                 continue;
4234                         state->offset++;
4235                         return ifa;
4236                 }
4237         }
4238
4239         return NULL;
4240 }
4241
4242 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4243         __acquires(rcu)
4244 {
4245         rcu_read_lock();
4246         return if6_get_first(seq, *pos);
4247 }
4248
4249 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4250 {
4251         struct inet6_ifaddr *ifa;
4252
4253         ifa = if6_get_next(seq, v);
4254         ++*pos;
4255         return ifa;
4256 }
4257
4258 static void if6_seq_stop(struct seq_file *seq, void *v)
4259         __releases(rcu)
4260 {
4261         rcu_read_unlock();
4262 }
4263
4264 static int if6_seq_show(struct seq_file *seq, void *v)
4265 {
4266         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4267         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4268                    &ifp->addr,
4269                    ifp->idev->dev->ifindex,
4270                    ifp->prefix_len,
4271                    ifp->scope,
4272                    (u8) ifp->flags,
4273                    ifp->idev->dev->name);
4274         return 0;
4275 }
4276
4277 static const struct seq_operations if6_seq_ops = {
4278         .start  = if6_seq_start,
4279         .next   = if6_seq_next,
4280         .show   = if6_seq_show,
4281         .stop   = if6_seq_stop,
4282 };
4283
4284 static int __net_init if6_proc_net_init(struct net *net)
4285 {
4286         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4287                         sizeof(struct if6_iter_state)))
4288                 return -ENOMEM;
4289         return 0;
4290 }
4291
4292 static void __net_exit if6_proc_net_exit(struct net *net)
4293 {
4294         remove_proc_entry("if_inet6", net->proc_net);
4295 }
4296
4297 static struct pernet_operations if6_proc_net_ops = {
4298         .init = if6_proc_net_init,
4299         .exit = if6_proc_net_exit,
4300 };
4301
4302 int __init if6_proc_init(void)
4303 {
4304         return register_pernet_subsys(&if6_proc_net_ops);
4305 }
4306
4307 void if6_proc_exit(void)
4308 {
4309         unregister_pernet_subsys(&if6_proc_net_ops);
4310 }
4311 #endif  /* CONFIG_PROC_FS */
4312
4313 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4314 /* Check if address is a home address configured on any interface. */
4315 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4316 {
4317         unsigned int hash = inet6_addr_hash(net, addr);
4318         struct inet6_ifaddr *ifp = NULL;
4319         int ret = 0;
4320
4321         rcu_read_lock();
4322         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4323                 if (!net_eq(dev_net(ifp->idev->dev), net))
4324                         continue;
4325                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4326                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4327                         ret = 1;
4328                         break;
4329                 }
4330         }
4331         rcu_read_unlock();
4332         return ret;
4333 }
4334 #endif
4335
4336 /*
4337  *      Periodic address status verification
4338  */
4339
4340 static void addrconf_verify_rtnl(void)
4341 {
4342         unsigned long now, next, next_sec, next_sched;
4343         struct inet6_ifaddr *ifp;
4344         int i;
4345
4346         ASSERT_RTNL();
4347
4348         rcu_read_lock_bh();
4349         now = jiffies;
4350         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4351
4352         cancel_delayed_work(&addr_chk_work);
4353
4354         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4355 restart:
4356                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4357                         unsigned long age;
4358
4359                         /* When setting preferred_lft to a value not zero or
4360                          * infinity, while valid_lft is infinity
4361                          * IFA_F_PERMANENT has a non-infinity life time.
4362                          */
4363                         if ((ifp->flags & IFA_F_PERMANENT) &&
4364                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4365                                 continue;
4366
4367                         spin_lock(&ifp->lock);
4368                         /* We try to batch several events at once. */
4369                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4370
4371                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4372                             age >= ifp->valid_lft) {
4373                                 spin_unlock(&ifp->lock);
4374                                 in6_ifa_hold(ifp);
4375                                 ipv6_del_addr(ifp);
4376                                 goto restart;
4377                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4378                                 spin_unlock(&ifp->lock);
4379                                 continue;
4380                         } else if (age >= ifp->prefered_lft) {
4381                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4382                                 int deprecate = 0;
4383
4384                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4385                                         deprecate = 1;
4386                                         ifp->flags |= IFA_F_DEPRECATED;
4387                                 }
4388
4389                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4390                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4391                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4392
4393                                 spin_unlock(&ifp->lock);
4394
4395                                 if (deprecate) {
4396                                         in6_ifa_hold(ifp);
4397
4398                                         ipv6_ifa_notify(0, ifp);
4399                                         in6_ifa_put(ifp);
4400                                         goto restart;
4401                                 }
4402                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4403                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4404                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4405                                         ifp->idev->cnf.dad_transmits *
4406                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4407
4408                                 if (age >= ifp->prefered_lft - regen_advance) {
4409                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4410                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4411                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4412                                         if (!ifp->regen_count && ifpub) {
4413                                                 ifp->regen_count++;
4414                                                 in6_ifa_hold(ifp);
4415                                                 in6_ifa_hold(ifpub);
4416                                                 spin_unlock(&ifp->lock);
4417
4418                                                 spin_lock(&ifpub->lock);
4419                                                 ifpub->regen_count = 0;
4420                                                 spin_unlock(&ifpub->lock);
4421                                                 rcu_read_unlock_bh();
4422                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4423                                                 in6_ifa_put(ifpub);
4424                                                 in6_ifa_put(ifp);
4425                                                 rcu_read_lock_bh();
4426                                                 goto restart;
4427                                         }
4428                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4429                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4430                                 spin_unlock(&ifp->lock);
4431                         } else {
4432                                 /* ifp->prefered_lft <= ifp->valid_lft */
4433                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4434                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4435                                 spin_unlock(&ifp->lock);
4436                         }
4437                 }
4438         }
4439
4440         next_sec = round_jiffies_up(next);
4441         next_sched = next;
4442
4443         /* If rounded timeout is accurate enough, accept it. */
4444         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4445                 next_sched = next_sec;
4446
4447         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4448         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4449                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4450
4451         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4452                  now, next, next_sec, next_sched);
4453         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4454         rcu_read_unlock_bh();
4455 }
4456
4457 static void addrconf_verify_work(struct work_struct *w)
4458 {
4459         rtnl_lock();
4460         addrconf_verify_rtnl();
4461         rtnl_unlock();
4462 }
4463
4464 static void addrconf_verify(void)
4465 {
4466         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4467 }
4468
4469 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4470                                      struct in6_addr **peer_pfx)
4471 {
4472         struct in6_addr *pfx = NULL;
4473
4474         *peer_pfx = NULL;
4475
4476         if (addr)
4477                 pfx = nla_data(addr);
4478
4479         if (local) {
4480                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4481                         *peer_pfx = pfx;
4482                 pfx = nla_data(local);
4483         }
4484
4485         return pfx;
4486 }
4487
4488 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4489         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4490         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4491         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4492         [IFA_FLAGS]             = { .len = sizeof(u32) },
4493         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4494 };
4495
4496 static int
4497 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4498                   struct netlink_ext_ack *extack)
4499 {
4500         struct net *net = sock_net(skb->sk);
4501         struct ifaddrmsg *ifm;
4502         struct nlattr *tb[IFA_MAX+1];
4503         struct in6_addr *pfx, *peer_pfx;
4504         u32 ifa_flags;
4505         int err;
4506
4507         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4508                           extack);
4509         if (err < 0)
4510                 return err;
4511
4512         ifm = nlmsg_data(nlh);
4513         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4514         if (!pfx)
4515                 return -EINVAL;
4516
4517         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4518
4519         /* We ignore other flags so far. */
4520         ifa_flags &= IFA_F_MANAGETEMPADDR;
4521
4522         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4523                               ifm->ifa_prefixlen);
4524 }
4525
4526 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4527                                unsigned long expires, u32 flags)
4528 {
4529         struct fib6_info *f6i;
4530         u32 prio;
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         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4540         if (f6i->fib6_metric != prio) {
4541                 /* delete old one */
4542                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4543
4544                 /* add new one */
4545                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4546                                       ifp->rt_priority, ifp->idev->dev,
4547                                       expires, flags, GFP_KERNEL);
4548         } else {
4549                 if (!expires)
4550                         fib6_clean_expires(f6i);
4551                 else
4552                         fib6_set_expires(f6i, expires);
4553
4554                 fib6_info_release(f6i);
4555         }
4556
4557         return 0;
4558 }
4559
4560 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4561 {
4562         u32 flags;
4563         clock_t expires;
4564         unsigned long timeout;
4565         bool was_managetempaddr;
4566         bool had_prefixroute;
4567
4568         ASSERT_RTNL();
4569
4570         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4571                 return -EINVAL;
4572
4573         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4574             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4575                 return -EINVAL;
4576
4577         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4578                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4579
4580         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4581         if (addrconf_finite_timeout(timeout)) {
4582                 expires = jiffies_to_clock_t(timeout * HZ);
4583                 cfg->valid_lft = timeout;
4584                 flags = RTF_EXPIRES;
4585         } else {
4586                 expires = 0;
4587                 flags = 0;
4588                 cfg->ifa_flags |= IFA_F_PERMANENT;
4589         }
4590
4591         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4592         if (addrconf_finite_timeout(timeout)) {
4593                 if (timeout == 0)
4594                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4595                 cfg->preferred_lft = timeout;
4596         }
4597
4598         spin_lock_bh(&ifp->lock);
4599         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4600         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4601                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4602         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4603                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4604                         IFA_F_NOPREFIXROUTE);
4605         ifp->flags |= cfg->ifa_flags;
4606         ifp->tstamp = jiffies;
4607         ifp->valid_lft = cfg->valid_lft;
4608         ifp->prefered_lft = cfg->preferred_lft;
4609
4610         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4611                 ifp->rt_priority = cfg->rt_priority;
4612
4613         spin_unlock_bh(&ifp->lock);
4614         if (!(ifp->flags&IFA_F_TENTATIVE))
4615                 ipv6_ifa_notify(0, ifp);
4616
4617         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4618                 int rc = -ENOENT;
4619
4620                 if (had_prefixroute)
4621                         rc = modify_prefix_route(ifp, expires, flags);
4622
4623                 /* prefix route could have been deleted; if so restore it */
4624                 if (rc == -ENOENT) {
4625                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4626                                               ifp->rt_priority, ifp->idev->dev,
4627                                               expires, flags, GFP_KERNEL);
4628                 }
4629         } else if (had_prefixroute) {
4630                 enum cleanup_prefix_rt_t action;
4631                 unsigned long rt_expires;
4632
4633                 write_lock_bh(&ifp->idev->lock);
4634                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4635                 write_unlock_bh(&ifp->idev->lock);
4636
4637                 if (action != CLEANUP_PREFIX_RT_NOP) {
4638                         cleanup_prefix_route(ifp, rt_expires,
4639                                 action == CLEANUP_PREFIX_RT_DEL);
4640                 }
4641         }
4642
4643         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4644                 if (was_managetempaddr &&
4645                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4646                         cfg->valid_lft = 0;
4647                         cfg->preferred_lft = 0;
4648                 }
4649                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4650                                  cfg->preferred_lft, !was_managetempaddr,
4651                                  jiffies);
4652         }
4653
4654         addrconf_verify_rtnl();
4655
4656         return 0;
4657 }
4658
4659 static int
4660 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4661                   struct netlink_ext_ack *extack)
4662 {
4663         struct net *net = sock_net(skb->sk);
4664         struct ifaddrmsg *ifm;
4665         struct nlattr *tb[IFA_MAX+1];
4666         struct in6_addr *peer_pfx;
4667         struct inet6_ifaddr *ifa;
4668         struct net_device *dev;
4669         struct inet6_dev *idev;
4670         struct ifa6_config cfg;
4671         int err;
4672
4673         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4674                           extack);
4675         if (err < 0)
4676                 return err;
4677
4678         memset(&cfg, 0, sizeof(cfg));
4679
4680         ifm = nlmsg_data(nlh);
4681         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4682         if (!cfg.pfx)
4683                 return -EINVAL;
4684
4685         cfg.peer_pfx = peer_pfx;
4686         cfg.plen = ifm->ifa_prefixlen;
4687         if (tb[IFA_RT_PRIORITY])
4688                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4689
4690         cfg.valid_lft = INFINITY_LIFE_TIME;
4691         cfg.preferred_lft = INFINITY_LIFE_TIME;
4692
4693         if (tb[IFA_CACHEINFO]) {
4694                 struct ifa_cacheinfo *ci;
4695
4696                 ci = nla_data(tb[IFA_CACHEINFO]);
4697                 cfg.valid_lft = ci->ifa_valid;
4698                 cfg.preferred_lft = ci->ifa_prefered;
4699         }
4700
4701         dev =  __dev_get_by_index(net, ifm->ifa_index);
4702         if (!dev)
4703                 return -ENODEV;
4704
4705         if (tb[IFA_FLAGS])
4706                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4707         else
4708                 cfg.ifa_flags = ifm->ifa_flags;
4709
4710         /* We ignore other flags so far. */
4711         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4712                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4713                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4714
4715         idev = ipv6_find_idev(dev);
4716         if (IS_ERR(idev))
4717                 return PTR_ERR(idev);
4718
4719         if (!ipv6_allow_optimistic_dad(net, idev))
4720                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4721
4722         if (cfg.ifa_flags & IFA_F_NODAD &&
4723             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4724                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4725                 return -EINVAL;
4726         }
4727
4728         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4729         if (!ifa) {
4730                 /*
4731                  * It would be best to check for !NLM_F_CREATE here but
4732                  * userspace already relies on not having to provide this.
4733                  */
4734                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4735         }
4736
4737         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4738             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4739                 err = -EEXIST;
4740         else
4741                 err = inet6_addr_modify(ifa, &cfg);
4742
4743         in6_ifa_put(ifa);
4744
4745         return err;
4746 }
4747
4748 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4749                           u8 scope, int ifindex)
4750 {
4751         struct ifaddrmsg *ifm;
4752
4753         ifm = nlmsg_data(nlh);
4754         ifm->ifa_family = AF_INET6;
4755         ifm->ifa_prefixlen = prefixlen;
4756         ifm->ifa_flags = flags;
4757         ifm->ifa_scope = scope;
4758         ifm->ifa_index = ifindex;
4759 }
4760
4761 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4762                          unsigned long tstamp, u32 preferred, u32 valid)
4763 {
4764         struct ifa_cacheinfo ci;
4765
4766         ci.cstamp = cstamp_delta(cstamp);
4767         ci.tstamp = cstamp_delta(tstamp);
4768         ci.ifa_prefered = preferred;
4769         ci.ifa_valid = valid;
4770
4771         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4772 }
4773
4774 static inline int rt_scope(int ifa_scope)
4775 {
4776         if (ifa_scope & IFA_HOST)
4777                 return RT_SCOPE_HOST;
4778         else if (ifa_scope & IFA_LINK)
4779                 return RT_SCOPE_LINK;
4780         else if (ifa_scope & IFA_SITE)
4781                 return RT_SCOPE_SITE;
4782         else
4783                 return RT_SCOPE_UNIVERSE;
4784 }
4785
4786 static inline int inet6_ifaddr_msgsize(void)
4787 {
4788         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4789                + nla_total_size(16) /* IFA_LOCAL */
4790                + nla_total_size(16) /* IFA_ADDRESS */
4791                + nla_total_size(sizeof(struct ifa_cacheinfo))
4792                + nla_total_size(4)  /* IFA_FLAGS */
4793                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4794 }
4795
4796 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4797                              u32 portid, u32 seq, int event, unsigned int flags)
4798 {
4799         struct nlmsghdr  *nlh;
4800         u32 preferred, valid;
4801
4802         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4803         if (!nlh)
4804                 return -EMSGSIZE;
4805
4806         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4807                       ifa->idev->dev->ifindex);
4808
4809         if (!((ifa->flags&IFA_F_PERMANENT) &&
4810               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4811                 preferred = ifa->prefered_lft;
4812                 valid = ifa->valid_lft;
4813                 if (preferred != INFINITY_LIFE_TIME) {
4814                         long tval = (jiffies - ifa->tstamp)/HZ;
4815                         if (preferred > tval)
4816                                 preferred -= tval;
4817                         else
4818                                 preferred = 0;
4819                         if (valid != INFINITY_LIFE_TIME) {
4820                                 if (valid > tval)
4821                                         valid -= tval;
4822                                 else
4823                                         valid = 0;
4824                         }
4825                 }
4826         } else {
4827                 preferred = INFINITY_LIFE_TIME;
4828                 valid = INFINITY_LIFE_TIME;
4829         }
4830
4831         if (!ipv6_addr_any(&ifa->peer_addr)) {
4832                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4833                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4834                         goto error;
4835         } else
4836                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4837                         goto error;
4838
4839         if (ifa->rt_priority &&
4840             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4841                 goto error;
4842
4843         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4844                 goto error;
4845
4846         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4847                 goto error;
4848
4849         nlmsg_end(skb, nlh);
4850         return 0;
4851
4852 error:
4853         nlmsg_cancel(skb, nlh);
4854         return -EMSGSIZE;
4855 }
4856
4857 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4858                                 u32 portid, u32 seq, int event, u16 flags)
4859 {
4860         struct nlmsghdr  *nlh;
4861         u8 scope = RT_SCOPE_UNIVERSE;
4862         int ifindex = ifmca->idev->dev->ifindex;
4863
4864         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4865                 scope = RT_SCOPE_SITE;
4866
4867         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4868         if (!nlh)
4869                 return -EMSGSIZE;
4870
4871         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4872         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4873             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4874                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4875                 nlmsg_cancel(skb, nlh);
4876                 return -EMSGSIZE;
4877         }
4878
4879         nlmsg_end(skb, nlh);
4880         return 0;
4881 }
4882
4883 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4884                                 u32 portid, u32 seq, int event, unsigned int flags)
4885 {
4886         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4887         int ifindex = dev ? dev->ifindex : 1;
4888         struct nlmsghdr  *nlh;
4889         u8 scope = RT_SCOPE_UNIVERSE;
4890
4891         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4892                 scope = RT_SCOPE_SITE;
4893
4894         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4895         if (!nlh)
4896                 return -EMSGSIZE;
4897
4898         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4899         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4900             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4901                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4902                 nlmsg_cancel(skb, nlh);
4903                 return -EMSGSIZE;
4904         }
4905
4906         nlmsg_end(skb, nlh);
4907         return 0;
4908 }
4909
4910 enum addr_type_t {
4911         UNICAST_ADDR,
4912         MULTICAST_ADDR,
4913         ANYCAST_ADDR,
4914 };
4915
4916 /* called with rcu_read_lock() */
4917 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4918                           struct netlink_callback *cb, enum addr_type_t type,
4919                           int s_ip_idx, int *p_ip_idx)
4920 {
4921         struct ifmcaddr6 *ifmca;
4922         struct ifacaddr6 *ifaca;
4923         int err = 1;
4924         int ip_idx = *p_ip_idx;
4925
4926         read_lock_bh(&idev->lock);
4927         switch (type) {
4928         case UNICAST_ADDR: {
4929                 struct inet6_ifaddr *ifa;
4930
4931                 /* unicast address incl. temp addr */
4932                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4933                         if (++ip_idx < s_ip_idx)
4934                                 continue;
4935                         err = inet6_fill_ifaddr(skb, ifa,
4936                                                 NETLINK_CB(cb->skb).portid,
4937                                                 cb->nlh->nlmsg_seq,
4938                                                 RTM_NEWADDR,
4939                                                 NLM_F_MULTI);
4940                         if (err < 0)
4941                                 break;
4942                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4943                 }
4944                 break;
4945         }
4946         case MULTICAST_ADDR:
4947                 /* multicast address */
4948                 for (ifmca = idev->mc_list; ifmca;
4949                      ifmca = ifmca->next, ip_idx++) {
4950                         if (ip_idx < s_ip_idx)
4951                                 continue;
4952                         err = inet6_fill_ifmcaddr(skb, ifmca,
4953                                                   NETLINK_CB(cb->skb).portid,
4954                                                   cb->nlh->nlmsg_seq,
4955                                                   RTM_GETMULTICAST,
4956                                                   NLM_F_MULTI);
4957                         if (err < 0)
4958                                 break;
4959                 }
4960                 break;
4961         case ANYCAST_ADDR:
4962                 /* anycast address */
4963                 for (ifaca = idev->ac_list; ifaca;
4964                      ifaca = ifaca->aca_next, ip_idx++) {
4965                         if (ip_idx < s_ip_idx)
4966                                 continue;
4967                         err = inet6_fill_ifacaddr(skb, ifaca,
4968                                                   NETLINK_CB(cb->skb).portid,
4969                                                   cb->nlh->nlmsg_seq,
4970                                                   RTM_GETANYCAST,
4971                                                   NLM_F_MULTI);
4972                         if (err < 0)
4973                                 break;
4974                 }
4975                 break;
4976         default:
4977                 break;
4978         }
4979         read_unlock_bh(&idev->lock);
4980         *p_ip_idx = ip_idx;
4981         return err;
4982 }
4983
4984 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4985                            enum addr_type_t type)
4986 {
4987         struct net *net = sock_net(skb->sk);
4988         int h, s_h;
4989         int idx, ip_idx;
4990         int s_idx, s_ip_idx;
4991         struct net_device *dev;
4992         struct inet6_dev *idev;
4993         struct hlist_head *head;
4994
4995         s_h = cb->args[0];
4996         s_idx = idx = cb->args[1];
4997         s_ip_idx = ip_idx = cb->args[2];
4998
4999         rcu_read_lock();
5000         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
5001         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5002                 idx = 0;
5003                 head = &net->dev_index_head[h];
5004                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5005                         if (idx < s_idx)
5006                                 goto cont;
5007                         if (h > s_h || idx > s_idx)
5008                                 s_ip_idx = 0;
5009                         ip_idx = 0;
5010                         idev = __in6_dev_get(dev);
5011                         if (!idev)
5012                                 goto cont;
5013
5014                         if (in6_dump_addrs(idev, skb, cb, type,
5015                                            s_ip_idx, &ip_idx) < 0)
5016                                 goto done;
5017 cont:
5018                         idx++;
5019                 }
5020         }
5021 done:
5022         rcu_read_unlock();
5023         cb->args[0] = h;
5024         cb->args[1] = idx;
5025         cb->args[2] = ip_idx;
5026
5027         return skb->len;
5028 }
5029
5030 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5031 {
5032         enum addr_type_t type = UNICAST_ADDR;
5033
5034         return inet6_dump_addr(skb, cb, type);
5035 }
5036
5037 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5038 {
5039         enum addr_type_t type = MULTICAST_ADDR;
5040
5041         return inet6_dump_addr(skb, cb, type);
5042 }
5043
5044
5045 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5046 {
5047         enum addr_type_t type = ANYCAST_ADDR;
5048
5049         return inet6_dump_addr(skb, cb, type);
5050 }
5051
5052 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5053                              struct netlink_ext_ack *extack)
5054 {
5055         struct net *net = sock_net(in_skb->sk);
5056         struct ifaddrmsg *ifm;
5057         struct nlattr *tb[IFA_MAX+1];
5058         struct in6_addr *addr = NULL, *peer;
5059         struct net_device *dev = NULL;
5060         struct inet6_ifaddr *ifa;
5061         struct sk_buff *skb;
5062         int err;
5063
5064         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5065                           extack);
5066         if (err < 0)
5067                 return err;
5068
5069         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5070         if (!addr)
5071                 return -EINVAL;
5072
5073         ifm = nlmsg_data(nlh);
5074         if (ifm->ifa_index)
5075                 dev = dev_get_by_index(net, ifm->ifa_index);
5076
5077         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5078         if (!ifa) {
5079                 err = -EADDRNOTAVAIL;
5080                 goto errout;
5081         }
5082
5083         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5084         if (!skb) {
5085                 err = -ENOBUFS;
5086                 goto errout_ifa;
5087         }
5088
5089         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5090                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5091         if (err < 0) {
5092                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5093                 WARN_ON(err == -EMSGSIZE);
5094                 kfree_skb(skb);
5095                 goto errout_ifa;
5096         }
5097         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5098 errout_ifa:
5099         in6_ifa_put(ifa);
5100 errout:
5101         if (dev)
5102                 dev_put(dev);
5103         return err;
5104 }
5105
5106 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5107 {
5108         struct sk_buff *skb;
5109         struct net *net = dev_net(ifa->idev->dev);
5110         int err = -ENOBUFS;
5111
5112         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5113         if (!skb)
5114                 goto errout;
5115
5116         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5117         if (err < 0) {
5118                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5119                 WARN_ON(err == -EMSGSIZE);
5120                 kfree_skb(skb);
5121                 goto errout;
5122         }
5123         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5124         return;
5125 errout:
5126         if (err < 0)
5127                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5128 }
5129
5130 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5131                                 __s32 *array, int bytes)
5132 {
5133         BUG_ON(bytes < (DEVCONF_MAX * 4));
5134
5135         memset(array, 0, bytes);
5136         array[DEVCONF_FORWARDING] = cnf->forwarding;
5137         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5138         array[DEVCONF_MTU6] = cnf->mtu6;
5139         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5140         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5141         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5142         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5143         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5144         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5145                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5146         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5147                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5148         array[DEVCONF_RTR_SOLICIT_DELAY] =
5149                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5150         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5151         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5152                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5153         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5154                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5155         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5156         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5157         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5158         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5159         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5160         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5161         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5162         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5163         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5164 #ifdef CONFIG_IPV6_ROUTER_PREF
5165         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5166         array[DEVCONF_RTR_PROBE_INTERVAL] =
5167                 jiffies_to_msecs(cnf->rtr_probe_interval);
5168 #ifdef CONFIG_IPV6_ROUTE_INFO
5169         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5170         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5171 #endif
5172 #endif
5173         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5174         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5175 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5176         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5177         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5178 #endif
5179 #ifdef CONFIG_IPV6_MROUTE
5180         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5181 #endif
5182         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5183         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5184         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5185         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5186         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5187         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5188         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5189         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5190         /* we omit DEVCONF_STABLE_SECRET for now */
5191         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5192         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5193         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5194         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5195         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5196 #ifdef CONFIG_IPV6_SEG6_HMAC
5197         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5198 #endif
5199         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5200         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5201         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5202         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5203 }
5204
5205 static inline size_t inet6_ifla6_size(void)
5206 {
5207         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5208              + nla_total_size(sizeof(struct ifla_cacheinfo))
5209              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5210              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5211              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5212              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5213              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5214              + 0;
5215 }
5216
5217 static inline size_t inet6_if_nlmsg_size(void)
5218 {
5219         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5220                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5221                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5222                + nla_total_size(4) /* IFLA_MTU */
5223                + nla_total_size(4) /* IFLA_LINK */
5224                + nla_total_size(1) /* IFLA_OPERSTATE */
5225                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5226 }
5227
5228 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5229                                         int bytes)
5230 {
5231         int i;
5232         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5233         BUG_ON(pad < 0);
5234
5235         /* Use put_unaligned() because stats may not be aligned for u64. */
5236         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5237         for (i = 1; i < ICMP6_MIB_MAX; i++)
5238                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5239
5240         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5241 }
5242
5243 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5244                                         int bytes, size_t syncpoff)
5245 {
5246         int i, c;
5247         u64 buff[IPSTATS_MIB_MAX];
5248         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5249
5250         BUG_ON(pad < 0);
5251
5252         memset(buff, 0, sizeof(buff));
5253         buff[0] = IPSTATS_MIB_MAX;
5254
5255         for_each_possible_cpu(c) {
5256                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5257                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5258         }
5259
5260         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5261         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5262 }
5263
5264 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5265                              int bytes)
5266 {
5267         switch (attrtype) {
5268         case IFLA_INET6_STATS:
5269                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5270                                      offsetof(struct ipstats_mib, syncp));
5271                 break;
5272         case IFLA_INET6_ICMP6STATS:
5273                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5274                 break;
5275         }
5276 }
5277
5278 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5279                                   u32 ext_filter_mask)
5280 {
5281         struct nlattr *nla;
5282         struct ifla_cacheinfo ci;
5283
5284         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5285                 goto nla_put_failure;
5286         ci.max_reasm_len = IPV6_MAXPLEN;
5287         ci.tstamp = cstamp_delta(idev->tstamp);
5288         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5289         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5290         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5291                 goto nla_put_failure;
5292         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5293         if (!nla)
5294                 goto nla_put_failure;
5295         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5296
5297         /* XXX - MC not implemented */
5298
5299         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5300                 return 0;
5301
5302         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5303         if (!nla)
5304                 goto nla_put_failure;
5305         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5306
5307         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5308         if (!nla)
5309                 goto nla_put_failure;
5310         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5311
5312         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5313         if (!nla)
5314                 goto nla_put_failure;
5315
5316         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5317                 goto nla_put_failure;
5318
5319         read_lock_bh(&idev->lock);
5320         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5321         read_unlock_bh(&idev->lock);
5322
5323         return 0;
5324
5325 nla_put_failure:
5326         return -EMSGSIZE;
5327 }
5328
5329 static size_t inet6_get_link_af_size(const struct net_device *dev,
5330                                      u32 ext_filter_mask)
5331 {
5332         if (!__in6_dev_get(dev))
5333                 return 0;
5334
5335         return inet6_ifla6_size();
5336 }
5337
5338 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5339                               u32 ext_filter_mask)
5340 {
5341         struct inet6_dev *idev = __in6_dev_get(dev);
5342
5343         if (!idev)
5344                 return -ENODATA;
5345
5346         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5347                 return -EMSGSIZE;
5348
5349         return 0;
5350 }
5351
5352 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5353 {
5354         struct inet6_ifaddr *ifp;
5355         struct net_device *dev = idev->dev;
5356         bool clear_token, update_rs = false;
5357         struct in6_addr ll_addr;
5358
5359         ASSERT_RTNL();
5360
5361         if (!token)
5362                 return -EINVAL;
5363         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5364                 return -EINVAL;
5365         if (!ipv6_accept_ra(idev))
5366                 return -EINVAL;
5367         if (idev->cnf.rtr_solicits == 0)
5368                 return -EINVAL;
5369
5370         write_lock_bh(&idev->lock);
5371
5372         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5373         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5374
5375         write_unlock_bh(&idev->lock);
5376
5377         clear_token = ipv6_addr_any(token);
5378         if (clear_token)
5379                 goto update_lft;
5380
5381         if (!idev->dead && (idev->if_flags & IF_READY) &&
5382             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5383                              IFA_F_OPTIMISTIC)) {
5384                 /* If we're not ready, then normal ifup will take care
5385                  * of this. Otherwise, we need to request our rs here.
5386                  */
5387                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5388                 update_rs = true;
5389         }
5390
5391 update_lft:
5392         write_lock_bh(&idev->lock);
5393
5394         if (update_rs) {
5395                 idev->if_flags |= IF_RS_SENT;
5396                 idev->rs_interval = rfc3315_s14_backoff_init(
5397                         idev->cnf.rtr_solicit_interval);
5398                 idev->rs_probes = 1;
5399                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5400         }
5401
5402         /* Well, that's kinda nasty ... */
5403         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5404                 spin_lock(&ifp->lock);
5405                 if (ifp->tokenized) {
5406                         ifp->valid_lft = 0;
5407                         ifp->prefered_lft = 0;
5408                 }
5409                 spin_unlock(&ifp->lock);
5410         }
5411
5412         write_unlock_bh(&idev->lock);
5413         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5414         addrconf_verify_rtnl();
5415         return 0;
5416 }
5417
5418 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5419         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5420         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5421 };
5422
5423 static int inet6_validate_link_af(const struct net_device *dev,
5424                                   const struct nlattr *nla)
5425 {
5426         struct nlattr *tb[IFLA_INET6_MAX + 1];
5427
5428         if (dev && !__in6_dev_get(dev))
5429                 return -EAFNOSUPPORT;
5430
5431         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5432                                 NULL);
5433 }
5434
5435 static int check_addr_gen_mode(int mode)
5436 {
5437         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5438             mode != IN6_ADDR_GEN_MODE_NONE &&
5439             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5440             mode != IN6_ADDR_GEN_MODE_RANDOM)
5441                 return -EINVAL;
5442         return 1;
5443 }
5444
5445 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5446                                 int mode)
5447 {
5448         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5449             !idev->cnf.stable_secret.initialized &&
5450             !net->ipv6.devconf_dflt->stable_secret.initialized)
5451                 return -EINVAL;
5452         return 1;
5453 }
5454
5455 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5456 {
5457         int err = -EINVAL;
5458         struct inet6_dev *idev = __in6_dev_get(dev);
5459         struct nlattr *tb[IFLA_INET6_MAX + 1];
5460
5461         if (!idev)
5462                 return -EAFNOSUPPORT;
5463
5464         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5465                 BUG();
5466
5467         if (tb[IFLA_INET6_TOKEN]) {
5468                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5469                 if (err)
5470                         return err;
5471         }
5472
5473         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5474                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5475
5476                 if (check_addr_gen_mode(mode) < 0 ||
5477                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5478                         return -EINVAL;
5479
5480                 idev->cnf.addr_gen_mode = mode;
5481                 err = 0;
5482         }
5483
5484         return err;
5485 }
5486
5487 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5488                              u32 portid, u32 seq, int event, unsigned int flags)
5489 {
5490         struct net_device *dev = idev->dev;
5491         struct ifinfomsg *hdr;
5492         struct nlmsghdr *nlh;
5493         void *protoinfo;
5494
5495         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5496         if (!nlh)
5497                 return -EMSGSIZE;
5498
5499         hdr = nlmsg_data(nlh);
5500         hdr->ifi_family = AF_INET6;
5501         hdr->__ifi_pad = 0;
5502         hdr->ifi_type = dev->type;
5503         hdr->ifi_index = dev->ifindex;
5504         hdr->ifi_flags = dev_get_flags(dev);
5505         hdr->ifi_change = 0;
5506
5507         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5508             (dev->addr_len &&
5509              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5510             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5511             (dev->ifindex != dev_get_iflink(dev) &&
5512              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5513             nla_put_u8(skb, IFLA_OPERSTATE,
5514                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5515                 goto nla_put_failure;
5516         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5517         if (!protoinfo)
5518                 goto nla_put_failure;
5519
5520         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5521                 goto nla_put_failure;
5522
5523         nla_nest_end(skb, protoinfo);
5524         nlmsg_end(skb, nlh);
5525         return 0;
5526
5527 nla_put_failure:
5528         nlmsg_cancel(skb, nlh);
5529         return -EMSGSIZE;
5530 }
5531
5532 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5533 {
5534         struct net *net = sock_net(skb->sk);
5535         int h, s_h;
5536         int idx = 0, s_idx;
5537         struct net_device *dev;
5538         struct inet6_dev *idev;
5539         struct hlist_head *head;
5540
5541         s_h = cb->args[0];
5542         s_idx = cb->args[1];
5543
5544         rcu_read_lock();
5545         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5546                 idx = 0;
5547                 head = &net->dev_index_head[h];
5548                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5549                         if (idx < s_idx)
5550                                 goto cont;
5551                         idev = __in6_dev_get(dev);
5552                         if (!idev)
5553                                 goto cont;
5554                         if (inet6_fill_ifinfo(skb, idev,
5555                                               NETLINK_CB(cb->skb).portid,
5556                                               cb->nlh->nlmsg_seq,
5557                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5558                                 goto out;
5559 cont:
5560                         idx++;
5561                 }
5562         }
5563 out:
5564         rcu_read_unlock();
5565         cb->args[1] = idx;
5566         cb->args[0] = h;
5567
5568         return skb->len;
5569 }
5570
5571 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5572 {
5573         struct sk_buff *skb;
5574         struct net *net = dev_net(idev->dev);
5575         int err = -ENOBUFS;
5576
5577         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5578         if (!skb)
5579                 goto errout;
5580
5581         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5582         if (err < 0) {
5583                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5584                 WARN_ON(err == -EMSGSIZE);
5585                 kfree_skb(skb);
5586                 goto errout;
5587         }
5588         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5589         return;
5590 errout:
5591         if (err < 0)
5592                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5593 }
5594
5595 static inline size_t inet6_prefix_nlmsg_size(void)
5596 {
5597         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5598                + nla_total_size(sizeof(struct in6_addr))
5599                + nla_total_size(sizeof(struct prefix_cacheinfo));
5600 }
5601
5602 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5603                              struct prefix_info *pinfo, u32 portid, u32 seq,
5604                              int event, unsigned int flags)
5605 {
5606         struct prefixmsg *pmsg;
5607         struct nlmsghdr *nlh;
5608         struct prefix_cacheinfo ci;
5609
5610         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5611         if (!nlh)
5612                 return -EMSGSIZE;
5613
5614         pmsg = nlmsg_data(nlh);
5615         pmsg->prefix_family = AF_INET6;
5616         pmsg->prefix_pad1 = 0;
5617         pmsg->prefix_pad2 = 0;
5618         pmsg->prefix_ifindex = idev->dev->ifindex;
5619         pmsg->prefix_len = pinfo->prefix_len;
5620         pmsg->prefix_type = pinfo->type;
5621         pmsg->prefix_pad3 = 0;
5622         pmsg->prefix_flags = 0;
5623         if (pinfo->onlink)
5624                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5625         if (pinfo->autoconf)
5626                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5627
5628         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5629                 goto nla_put_failure;
5630         ci.preferred_time = ntohl(pinfo->prefered);
5631         ci.valid_time = ntohl(pinfo->valid);
5632         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5633                 goto nla_put_failure;
5634         nlmsg_end(skb, nlh);
5635         return 0;
5636
5637 nla_put_failure:
5638         nlmsg_cancel(skb, nlh);
5639         return -EMSGSIZE;
5640 }
5641
5642 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5643                          struct prefix_info *pinfo)
5644 {
5645         struct sk_buff *skb;
5646         struct net *net = dev_net(idev->dev);
5647         int err = -ENOBUFS;
5648
5649         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5650         if (!skb)
5651                 goto errout;
5652
5653         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5654         if (err < 0) {
5655                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5656                 WARN_ON(err == -EMSGSIZE);
5657                 kfree_skb(skb);
5658                 goto errout;
5659         }
5660         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5661         return;
5662 errout:
5663         if (err < 0)
5664                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5665 }
5666
5667 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5668 {
5669         struct net *net = dev_net(ifp->idev->dev);
5670
5671         if (event)
5672                 ASSERT_RTNL();
5673
5674         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5675
5676         switch (event) {
5677         case RTM_NEWADDR:
5678                 /*
5679                  * If the address was optimistic
5680                  * we inserted the route at the start of
5681                  * our DAD process, so we don't need
5682                  * to do it again
5683                  */
5684                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5685                         ip6_ins_rt(net, ifp->rt);
5686                 if (ifp->idev->cnf.forwarding)
5687                         addrconf_join_anycast(ifp);
5688                 if (!ipv6_addr_any(&ifp->peer_addr))
5689                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5690                                               ifp->idev->dev, 0, 0,
5691                                               GFP_ATOMIC);
5692                 break;
5693         case RTM_DELADDR:
5694                 if (ifp->idev->cnf.forwarding)
5695                         addrconf_leave_anycast(ifp);
5696                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5697                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5698                         struct fib6_info *rt;
5699
5700                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5701                                                        ifp->idev->dev, 0, 0);
5702                         if (rt)
5703                                 ip6_del_rt(net, rt);
5704                 }
5705                 if (ifp->rt) {
5706                         ip6_del_rt(net, ifp->rt);
5707                         ifp->rt = NULL;
5708                 }
5709                 rt_genid_bump_ipv6(net);
5710                 break;
5711         }
5712         atomic_inc(&net->ipv6.dev_addr_genid);
5713 }
5714
5715 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5716 {
5717         rcu_read_lock_bh();
5718         if (likely(ifp->idev->dead == 0))
5719                 __ipv6_ifa_notify(event, ifp);
5720         rcu_read_unlock_bh();
5721 }
5722
5723 #ifdef CONFIG_SYSCTL
5724
5725 static
5726 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5727                            void __user *buffer, size_t *lenp, loff_t *ppos)
5728 {
5729         int *valp = ctl->data;
5730         int val = *valp;
5731         loff_t pos = *ppos;
5732         struct ctl_table lctl;
5733         int ret;
5734
5735         /*
5736          * ctl->data points to idev->cnf.forwarding, we should
5737          * not modify it until we get the rtnl lock.
5738          */
5739         lctl = *ctl;
5740         lctl.data = &val;
5741
5742         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5743
5744         if (write)
5745                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5746         if (ret)
5747                 *ppos = pos;
5748         return ret;
5749 }
5750
5751 static
5752 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5753                         void __user *buffer, size_t *lenp, loff_t *ppos)
5754 {
5755         struct inet6_dev *idev = ctl->extra1;
5756         int min_mtu = IPV6_MIN_MTU;
5757         struct ctl_table lctl;
5758
5759         lctl = *ctl;
5760         lctl.extra1 = &min_mtu;
5761         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5762
5763         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5764 }
5765
5766 static void dev_disable_change(struct inet6_dev *idev)
5767 {
5768         struct netdev_notifier_info info;
5769
5770         if (!idev || !idev->dev)
5771                 return;
5772
5773         netdev_notifier_info_init(&info, idev->dev);
5774         if (idev->cnf.disable_ipv6)
5775                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5776         else
5777                 addrconf_notify(NULL, NETDEV_UP, &info);
5778 }
5779
5780 static void addrconf_disable_change(struct net *net, __s32 newf)
5781 {
5782         struct net_device *dev;
5783         struct inet6_dev *idev;
5784
5785         for_each_netdev(net, dev) {
5786                 idev = __in6_dev_get(dev);
5787                 if (idev) {
5788                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5789                         idev->cnf.disable_ipv6 = newf;
5790                         if (changed)
5791                                 dev_disable_change(idev);
5792                 }
5793         }
5794 }
5795
5796 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5797 {
5798         struct net *net;
5799         int old;
5800
5801         if (!rtnl_trylock())
5802                 return restart_syscall();
5803
5804         net = (struct net *)table->extra2;
5805         old = *p;
5806         *p = newf;
5807
5808         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5809                 rtnl_unlock();
5810                 return 0;
5811         }
5812
5813         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5814                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5815                 addrconf_disable_change(net, newf);
5816         } else if ((!newf) ^ (!old))
5817                 dev_disable_change((struct inet6_dev *)table->extra1);
5818
5819         rtnl_unlock();
5820         return 0;
5821 }
5822
5823 static
5824 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5825                             void __user *buffer, size_t *lenp, loff_t *ppos)
5826 {
5827         int *valp = ctl->data;
5828         int val = *valp;
5829         loff_t pos = *ppos;
5830         struct ctl_table lctl;
5831         int ret;
5832
5833         /*
5834          * ctl->data points to idev->cnf.disable_ipv6, we should
5835          * not modify it until we get the rtnl lock.
5836          */
5837         lctl = *ctl;
5838         lctl.data = &val;
5839
5840         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5841
5842         if (write)
5843                 ret = addrconf_disable_ipv6(ctl, valp, val);
5844         if (ret)
5845                 *ppos = pos;
5846         return ret;
5847 }
5848
5849 static
5850 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5851                               void __user *buffer, size_t *lenp, loff_t *ppos)
5852 {
5853         int *valp = ctl->data;
5854         int ret;
5855         int old, new;
5856
5857         old = *valp;
5858         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5859         new = *valp;
5860
5861         if (write && old != new) {
5862                 struct net *net = ctl->extra2;
5863
5864                 if (!rtnl_trylock())
5865                         return restart_syscall();
5866
5867                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5868                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5869                                                      NETCONFA_PROXY_NEIGH,
5870                                                      NETCONFA_IFINDEX_DEFAULT,
5871                                                      net->ipv6.devconf_dflt);
5872                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5873                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5874                                                      NETCONFA_PROXY_NEIGH,
5875                                                      NETCONFA_IFINDEX_ALL,
5876                                                      net->ipv6.devconf_all);
5877                 else {
5878                         struct inet6_dev *idev = ctl->extra1;
5879
5880                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5881                                                      NETCONFA_PROXY_NEIGH,
5882                                                      idev->dev->ifindex,
5883                                                      &idev->cnf);
5884                 }
5885                 rtnl_unlock();
5886         }
5887
5888         return ret;
5889 }
5890
5891 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5892                                          void __user *buffer, size_t *lenp,
5893                                          loff_t *ppos)
5894 {
5895         int ret = 0;
5896         u32 new_val;
5897         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5898         struct net *net = (struct net *)ctl->extra2;
5899         struct ctl_table tmp = {
5900                 .data = &new_val,
5901                 .maxlen = sizeof(new_val),
5902                 .mode = ctl->mode,
5903         };
5904
5905         if (!rtnl_trylock())
5906                 return restart_syscall();
5907
5908         new_val = *((u32 *)ctl->data);
5909
5910         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
5911         if (ret != 0)
5912                 goto out;
5913
5914         if (write) {
5915                 if (check_addr_gen_mode(new_val) < 0) {
5916                         ret = -EINVAL;
5917                         goto out;
5918                 }
5919
5920                 if (idev) {
5921                         if (check_stable_privacy(idev, net, new_val) < 0) {
5922                                 ret = -EINVAL;
5923                                 goto out;
5924                         }
5925
5926                         if (idev->cnf.addr_gen_mode != new_val) {
5927                                 idev->cnf.addr_gen_mode = new_val;
5928                                 addrconf_dev_config(idev->dev);
5929                         }
5930                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
5931                         struct net_device *dev;
5932
5933                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
5934                         for_each_netdev(net, dev) {
5935                                 idev = __in6_dev_get(dev);
5936                                 if (idev &&
5937                                     idev->cnf.addr_gen_mode != new_val) {
5938                                         idev->cnf.addr_gen_mode = new_val;
5939                                         addrconf_dev_config(idev->dev);
5940                                 }
5941                         }
5942                 }
5943
5944                 *((u32 *)ctl->data) = new_val;
5945         }
5946
5947 out:
5948         rtnl_unlock();
5949
5950         return ret;
5951 }
5952
5953 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5954                                          void __user *buffer, size_t *lenp,
5955                                          loff_t *ppos)
5956 {
5957         int err;
5958         struct in6_addr addr;
5959         char str[IPV6_MAX_STRLEN];
5960         struct ctl_table lctl = *ctl;
5961         struct net *net = ctl->extra2;
5962         struct ipv6_stable_secret *secret = ctl->data;
5963
5964         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5965                 return -EIO;
5966
5967         lctl.maxlen = IPV6_MAX_STRLEN;
5968         lctl.data = str;
5969
5970         if (!rtnl_trylock())
5971                 return restart_syscall();
5972
5973         if (!write && !secret->initialized) {
5974                 err = -EIO;
5975                 goto out;
5976         }
5977
5978         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5979         if (err >= sizeof(str)) {
5980                 err = -EIO;
5981                 goto out;
5982         }
5983
5984         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5985         if (err || !write)
5986                 goto out;
5987
5988         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5989                 err = -EIO;
5990                 goto out;
5991         }
5992
5993         secret->initialized = true;
5994         secret->secret = addr;
5995
5996         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5997                 struct net_device *dev;
5998
5999                 for_each_netdev(net, dev) {
6000                         struct inet6_dev *idev = __in6_dev_get(dev);
6001
6002                         if (idev) {
6003                                 idev->cnf.addr_gen_mode =
6004                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6005                         }
6006                 }
6007         } else {
6008                 struct inet6_dev *idev = ctl->extra1;
6009
6010                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6011         }
6012
6013 out:
6014         rtnl_unlock();
6015
6016         return err;
6017 }
6018
6019 static
6020 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6021                                                 int write,
6022                                                 void __user *buffer,
6023                                                 size_t *lenp,
6024                                                 loff_t *ppos)
6025 {
6026         int *valp = ctl->data;
6027         int val = *valp;
6028         loff_t pos = *ppos;
6029         struct ctl_table lctl;
6030         int ret;
6031
6032         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6033          * we should not modify it until we get the rtnl lock.
6034          */
6035         lctl = *ctl;
6036         lctl.data = &val;
6037
6038         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6039
6040         if (write)
6041                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6042         if (ret)
6043                 *ppos = pos;
6044         return ret;
6045 }
6046
6047 static
6048 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6049 {
6050         if (rt) {
6051                 if (action)
6052                         rt->dst.flags |= DST_NOPOLICY;
6053                 else
6054                         rt->dst.flags &= ~DST_NOPOLICY;
6055         }
6056 }
6057
6058 static
6059 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6060 {
6061         struct inet6_ifaddr *ifa;
6062
6063         read_lock_bh(&idev->lock);
6064         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6065                 spin_lock(&ifa->lock);
6066                 if (ifa->rt) {
6067                         struct fib6_info *rt = ifa->rt;
6068                         int cpu;
6069
6070                         rcu_read_lock();
6071                         ifa->rt->dst_nopolicy = val ? true : false;
6072                         if (rt->rt6i_pcpu) {
6073                                 for_each_possible_cpu(cpu) {
6074                                         struct rt6_info **rtp;
6075
6076                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6077                                         addrconf_set_nopolicy(*rtp, val);
6078                                 }
6079                         }
6080                         rcu_read_unlock();
6081                 }
6082                 spin_unlock(&ifa->lock);
6083         }
6084         read_unlock_bh(&idev->lock);
6085 }
6086
6087 static
6088 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6089 {
6090         struct inet6_dev *idev;
6091         struct net *net;
6092
6093         if (!rtnl_trylock())
6094                 return restart_syscall();
6095
6096         *valp = val;
6097
6098         net = (struct net *)ctl->extra2;
6099         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6100                 rtnl_unlock();
6101                 return 0;
6102         }
6103
6104         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6105                 struct net_device *dev;
6106
6107                 for_each_netdev(net, dev) {
6108                         idev = __in6_dev_get(dev);
6109                         if (idev)
6110                                 addrconf_disable_policy_idev(idev, val);
6111                 }
6112         } else {
6113                 idev = (struct inet6_dev *)ctl->extra1;
6114                 addrconf_disable_policy_idev(idev, val);
6115         }
6116
6117         rtnl_unlock();
6118         return 0;
6119 }
6120
6121 static
6122 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6123                                    void __user *buffer, size_t *lenp,
6124                                    loff_t *ppos)
6125 {
6126         int *valp = ctl->data;
6127         int val = *valp;
6128         loff_t pos = *ppos;
6129         struct ctl_table lctl;
6130         int ret;
6131
6132         lctl = *ctl;
6133         lctl.data = &val;
6134         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6135
6136         if (write && (*valp != val))
6137                 ret = addrconf_disable_policy(ctl, valp, val);
6138
6139         if (ret)
6140                 *ppos = pos;
6141
6142         return ret;
6143 }
6144
6145 static int minus_one = -1;
6146 static const int zero = 0;
6147 static const int one = 1;
6148 static const int two_five_five = 255;
6149
6150 static const struct ctl_table addrconf_sysctl[] = {
6151         {
6152                 .procname       = "forwarding",
6153                 .data           = &ipv6_devconf.forwarding,
6154                 .maxlen         = sizeof(int),
6155                 .mode           = 0644,
6156                 .proc_handler   = addrconf_sysctl_forward,
6157         },
6158         {
6159                 .procname       = "hop_limit",
6160                 .data           = &ipv6_devconf.hop_limit,
6161                 .maxlen         = sizeof(int),
6162                 .mode           = 0644,
6163                 .proc_handler   = proc_dointvec_minmax,
6164                 .extra1         = (void *)&one,
6165                 .extra2         = (void *)&two_five_five,
6166         },
6167         {
6168                 .procname       = "mtu",
6169                 .data           = &ipv6_devconf.mtu6,
6170                 .maxlen         = sizeof(int),
6171                 .mode           = 0644,
6172                 .proc_handler   = addrconf_sysctl_mtu,
6173         },
6174         {
6175                 .procname       = "accept_ra",
6176                 .data           = &ipv6_devconf.accept_ra,
6177                 .maxlen         = sizeof(int),
6178                 .mode           = 0644,
6179                 .proc_handler   = proc_dointvec,
6180         },
6181         {
6182                 .procname       = "accept_redirects",
6183                 .data           = &ipv6_devconf.accept_redirects,
6184                 .maxlen         = sizeof(int),
6185                 .mode           = 0644,
6186                 .proc_handler   = proc_dointvec,
6187         },
6188         {
6189                 .procname       = "autoconf",
6190                 .data           = &ipv6_devconf.autoconf,
6191                 .maxlen         = sizeof(int),
6192                 .mode           = 0644,
6193                 .proc_handler   = proc_dointvec,
6194         },
6195         {
6196                 .procname       = "dad_transmits",
6197                 .data           = &ipv6_devconf.dad_transmits,
6198                 .maxlen         = sizeof(int),
6199                 .mode           = 0644,
6200                 .proc_handler   = proc_dointvec,
6201         },
6202         {
6203                 .procname       = "router_solicitations",
6204                 .data           = &ipv6_devconf.rtr_solicits,
6205                 .maxlen         = sizeof(int),
6206                 .mode           = 0644,
6207                 .proc_handler   = proc_dointvec_minmax,
6208                 .extra1         = &minus_one,
6209         },
6210         {
6211                 .procname       = "router_solicitation_interval",
6212                 .data           = &ipv6_devconf.rtr_solicit_interval,
6213                 .maxlen         = sizeof(int),
6214                 .mode           = 0644,
6215                 .proc_handler   = proc_dointvec_jiffies,
6216         },
6217         {
6218                 .procname       = "router_solicitation_max_interval",
6219                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6220                 .maxlen         = sizeof(int),
6221                 .mode           = 0644,
6222                 .proc_handler   = proc_dointvec_jiffies,
6223         },
6224         {
6225                 .procname       = "router_solicitation_delay",
6226                 .data           = &ipv6_devconf.rtr_solicit_delay,
6227                 .maxlen         = sizeof(int),
6228                 .mode           = 0644,
6229                 .proc_handler   = proc_dointvec_jiffies,
6230         },
6231         {
6232                 .procname       = "force_mld_version",
6233                 .data           = &ipv6_devconf.force_mld_version,
6234                 .maxlen         = sizeof(int),
6235                 .mode           = 0644,
6236                 .proc_handler   = proc_dointvec,
6237         },
6238         {
6239                 .procname       = "mldv1_unsolicited_report_interval",
6240                 .data           =
6241                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6242                 .maxlen         = sizeof(int),
6243                 .mode           = 0644,
6244                 .proc_handler   = proc_dointvec_ms_jiffies,
6245         },
6246         {
6247                 .procname       = "mldv2_unsolicited_report_interval",
6248                 .data           =
6249                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6250                 .maxlen         = sizeof(int),
6251                 .mode           = 0644,
6252                 .proc_handler   = proc_dointvec_ms_jiffies,
6253         },
6254         {
6255                 .procname       = "use_tempaddr",
6256                 .data           = &ipv6_devconf.use_tempaddr,
6257                 .maxlen         = sizeof(int),
6258                 .mode           = 0644,
6259                 .proc_handler   = proc_dointvec,
6260         },
6261         {
6262                 .procname       = "temp_valid_lft",
6263                 .data           = &ipv6_devconf.temp_valid_lft,
6264                 .maxlen         = sizeof(int),
6265                 .mode           = 0644,
6266                 .proc_handler   = proc_dointvec,
6267         },
6268         {
6269                 .procname       = "temp_prefered_lft",
6270                 .data           = &ipv6_devconf.temp_prefered_lft,
6271                 .maxlen         = sizeof(int),
6272                 .mode           = 0644,
6273                 .proc_handler   = proc_dointvec,
6274         },
6275         {
6276                 .procname       = "regen_max_retry",
6277                 .data           = &ipv6_devconf.regen_max_retry,
6278                 .maxlen         = sizeof(int),
6279                 .mode           = 0644,
6280                 .proc_handler   = proc_dointvec,
6281         },
6282         {
6283                 .procname       = "max_desync_factor",
6284                 .data           = &ipv6_devconf.max_desync_factor,
6285                 .maxlen         = sizeof(int),
6286                 .mode           = 0644,
6287                 .proc_handler   = proc_dointvec,
6288         },
6289         {
6290                 .procname       = "max_addresses",
6291                 .data           = &ipv6_devconf.max_addresses,
6292                 .maxlen         = sizeof(int),
6293                 .mode           = 0644,
6294                 .proc_handler   = proc_dointvec,
6295         },
6296         {
6297                 .procname       = "accept_ra_defrtr",
6298                 .data           = &ipv6_devconf.accept_ra_defrtr,
6299                 .maxlen         = sizeof(int),
6300                 .mode           = 0644,
6301                 .proc_handler   = proc_dointvec,
6302         },
6303         {
6304                 .procname       = "accept_ra_min_hop_limit",
6305                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6306                 .maxlen         = sizeof(int),
6307                 .mode           = 0644,
6308                 .proc_handler   = proc_dointvec,
6309         },
6310         {
6311                 .procname       = "accept_ra_pinfo",
6312                 .data           = &ipv6_devconf.accept_ra_pinfo,
6313                 .maxlen         = sizeof(int),
6314                 .mode           = 0644,
6315                 .proc_handler   = proc_dointvec,
6316         },
6317 #ifdef CONFIG_IPV6_ROUTER_PREF
6318         {
6319                 .procname       = "accept_ra_rtr_pref",
6320                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6321                 .maxlen         = sizeof(int),
6322                 .mode           = 0644,
6323                 .proc_handler   = proc_dointvec,
6324         },
6325         {
6326                 .procname       = "router_probe_interval",
6327                 .data           = &ipv6_devconf.rtr_probe_interval,
6328                 .maxlen         = sizeof(int),
6329                 .mode           = 0644,
6330                 .proc_handler   = proc_dointvec_jiffies,
6331         },
6332 #ifdef CONFIG_IPV6_ROUTE_INFO
6333         {
6334                 .procname       = "accept_ra_rt_info_min_plen",
6335                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6336                 .maxlen         = sizeof(int),
6337                 .mode           = 0644,
6338                 .proc_handler   = proc_dointvec,
6339         },
6340         {
6341                 .procname       = "accept_ra_rt_info_max_plen",
6342                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6343                 .maxlen         = sizeof(int),
6344                 .mode           = 0644,
6345                 .proc_handler   = proc_dointvec,
6346         },
6347 #endif
6348 #endif
6349         {
6350                 .procname       = "proxy_ndp",
6351                 .data           = &ipv6_devconf.proxy_ndp,
6352                 .maxlen         = sizeof(int),
6353                 .mode           = 0644,
6354                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6355         },
6356         {
6357                 .procname       = "accept_source_route",
6358                 .data           = &ipv6_devconf.accept_source_route,
6359                 .maxlen         = sizeof(int),
6360                 .mode           = 0644,
6361                 .proc_handler   = proc_dointvec,
6362         },
6363 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6364         {
6365                 .procname       = "optimistic_dad",
6366                 .data           = &ipv6_devconf.optimistic_dad,
6367                 .maxlen         = sizeof(int),
6368                 .mode           = 0644,
6369                 .proc_handler   = proc_dointvec,
6370         },
6371         {
6372                 .procname       = "use_optimistic",
6373                 .data           = &ipv6_devconf.use_optimistic,
6374                 .maxlen         = sizeof(int),
6375                 .mode           = 0644,
6376                 .proc_handler   = proc_dointvec,
6377         },
6378 #endif
6379 #ifdef CONFIG_IPV6_MROUTE
6380         {
6381                 .procname       = "mc_forwarding",
6382                 .data           = &ipv6_devconf.mc_forwarding,
6383                 .maxlen         = sizeof(int),
6384                 .mode           = 0444,
6385                 .proc_handler   = proc_dointvec,
6386         },
6387 #endif
6388         {
6389                 .procname       = "disable_ipv6",
6390                 .data           = &ipv6_devconf.disable_ipv6,
6391                 .maxlen         = sizeof(int),
6392                 .mode           = 0644,
6393                 .proc_handler   = addrconf_sysctl_disable,
6394         },
6395         {
6396                 .procname       = "accept_dad",
6397                 .data           = &ipv6_devconf.accept_dad,
6398                 .maxlen         = sizeof(int),
6399                 .mode           = 0644,
6400                 .proc_handler   = proc_dointvec,
6401         },
6402         {
6403                 .procname       = "force_tllao",
6404                 .data           = &ipv6_devconf.force_tllao,
6405                 .maxlen         = sizeof(int),
6406                 .mode           = 0644,
6407                 .proc_handler   = proc_dointvec
6408         },
6409         {
6410                 .procname       = "ndisc_notify",
6411                 .data           = &ipv6_devconf.ndisc_notify,
6412                 .maxlen         = sizeof(int),
6413                 .mode           = 0644,
6414                 .proc_handler   = proc_dointvec
6415         },
6416         {
6417                 .procname       = "suppress_frag_ndisc",
6418                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6419                 .maxlen         = sizeof(int),
6420                 .mode           = 0644,
6421                 .proc_handler   = proc_dointvec
6422         },
6423         {
6424                 .procname       = "accept_ra_from_local",
6425                 .data           = &ipv6_devconf.accept_ra_from_local,
6426                 .maxlen         = sizeof(int),
6427                 .mode           = 0644,
6428                 .proc_handler   = proc_dointvec,
6429         },
6430         {
6431                 .procname       = "accept_ra_mtu",
6432                 .data           = &ipv6_devconf.accept_ra_mtu,
6433                 .maxlen         = sizeof(int),
6434                 .mode           = 0644,
6435                 .proc_handler   = proc_dointvec,
6436         },
6437         {
6438                 .procname       = "stable_secret",
6439                 .data           = &ipv6_devconf.stable_secret,
6440                 .maxlen         = IPV6_MAX_STRLEN,
6441                 .mode           = 0600,
6442                 .proc_handler   = addrconf_sysctl_stable_secret,
6443         },
6444         {
6445                 .procname       = "use_oif_addrs_only",
6446                 .data           = &ipv6_devconf.use_oif_addrs_only,
6447                 .maxlen         = sizeof(int),
6448                 .mode           = 0644,
6449                 .proc_handler   = proc_dointvec,
6450         },
6451         {
6452                 .procname       = "ignore_routes_with_linkdown",
6453                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6454                 .maxlen         = sizeof(int),
6455                 .mode           = 0644,
6456                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6457         },
6458         {
6459                 .procname       = "drop_unicast_in_l2_multicast",
6460                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6461                 .maxlen         = sizeof(int),
6462                 .mode           = 0644,
6463                 .proc_handler   = proc_dointvec,
6464         },
6465         {
6466                 .procname       = "drop_unsolicited_na",
6467                 .data           = &ipv6_devconf.drop_unsolicited_na,
6468                 .maxlen         = sizeof(int),
6469                 .mode           = 0644,
6470                 .proc_handler   = proc_dointvec,
6471         },
6472         {
6473                 .procname       = "keep_addr_on_down",
6474                 .data           = &ipv6_devconf.keep_addr_on_down,
6475                 .maxlen         = sizeof(int),
6476                 .mode           = 0644,
6477                 .proc_handler   = proc_dointvec,
6478
6479         },
6480         {
6481                 .procname       = "seg6_enabled",
6482                 .data           = &ipv6_devconf.seg6_enabled,
6483                 .maxlen         = sizeof(int),
6484                 .mode           = 0644,
6485                 .proc_handler   = proc_dointvec,
6486         },
6487 #ifdef CONFIG_IPV6_SEG6_HMAC
6488         {
6489                 .procname       = "seg6_require_hmac",
6490                 .data           = &ipv6_devconf.seg6_require_hmac,
6491                 .maxlen         = sizeof(int),
6492                 .mode           = 0644,
6493                 .proc_handler   = proc_dointvec,
6494         },
6495 #endif
6496         {
6497                 .procname       = "enhanced_dad",
6498                 .data           = &ipv6_devconf.enhanced_dad,
6499                 .maxlen         = sizeof(int),
6500                 .mode           = 0644,
6501                 .proc_handler   = proc_dointvec,
6502         },
6503         {
6504                 .procname               = "addr_gen_mode",
6505                 .data                   = &ipv6_devconf.addr_gen_mode,
6506                 .maxlen                 = sizeof(int),
6507                 .mode                   = 0644,
6508                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6509         },
6510         {
6511                 .procname       = "disable_policy",
6512                 .data           = &ipv6_devconf.disable_policy,
6513                 .maxlen         = sizeof(int),
6514                 .mode           = 0644,
6515                 .proc_handler   = addrconf_sysctl_disable_policy,
6516         },
6517         {
6518                 .procname       = "ndisc_tclass",
6519                 .data           = &ipv6_devconf.ndisc_tclass,
6520                 .maxlen         = sizeof(int),
6521                 .mode           = 0644,
6522                 .proc_handler   = proc_dointvec_minmax,
6523                 .extra1         = (void *)&zero,
6524                 .extra2         = (void *)&two_five_five,
6525         },
6526         {
6527                 /* sentinel */
6528         }
6529 };
6530
6531 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6532                 struct inet6_dev *idev, struct ipv6_devconf *p)
6533 {
6534         int i, ifindex;
6535         struct ctl_table *table;
6536         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6537
6538         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6539         if (!table)
6540                 goto out;
6541
6542         for (i = 0; table[i].data; i++) {
6543                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6544                 /* If one of these is already set, then it is not safe to
6545                  * overwrite either of them: this makes proc_dointvec_minmax
6546                  * usable.
6547                  */
6548                 if (!table[i].extra1 && !table[i].extra2) {
6549                         table[i].extra1 = idev; /* embedded; no ref */
6550                         table[i].extra2 = net;
6551                 }
6552         }
6553
6554         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6555
6556         p->sysctl_header = register_net_sysctl(net, path, table);
6557         if (!p->sysctl_header)
6558                 goto free;
6559
6560         if (!strcmp(dev_name, "all"))
6561                 ifindex = NETCONFA_IFINDEX_ALL;
6562         else if (!strcmp(dev_name, "default"))
6563                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6564         else
6565                 ifindex = idev->dev->ifindex;
6566         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6567                                      ifindex, p);
6568         return 0;
6569
6570 free:
6571         kfree(table);
6572 out:
6573         return -ENOBUFS;
6574 }
6575
6576 static void __addrconf_sysctl_unregister(struct net *net,
6577                                          struct ipv6_devconf *p, int ifindex)
6578 {
6579         struct ctl_table *table;
6580
6581         if (!p->sysctl_header)
6582                 return;
6583
6584         table = p->sysctl_header->ctl_table_arg;
6585         unregister_net_sysctl_table(p->sysctl_header);
6586         p->sysctl_header = NULL;
6587         kfree(table);
6588
6589         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6590 }
6591
6592 static int addrconf_sysctl_register(struct inet6_dev *idev)
6593 {
6594         int err;
6595
6596         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6597                 return -EINVAL;
6598
6599         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6600                                     &ndisc_ifinfo_sysctl_change);
6601         if (err)
6602                 return err;
6603         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6604                                          idev, &idev->cnf);
6605         if (err)
6606                 neigh_sysctl_unregister(idev->nd_parms);
6607
6608         return err;
6609 }
6610
6611 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6612 {
6613         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6614                                      idev->dev->ifindex);
6615         neigh_sysctl_unregister(idev->nd_parms);
6616 }
6617
6618
6619 #endif
6620
6621 static int __net_init addrconf_init_net(struct net *net)
6622 {
6623         int err = -ENOMEM;
6624         struct ipv6_devconf *all, *dflt;
6625
6626         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6627         if (!all)
6628                 goto err_alloc_all;
6629
6630         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6631         if (!dflt)
6632                 goto err_alloc_dflt;
6633
6634         /* these will be inherited by all namespaces */
6635         dflt->autoconf = ipv6_defaults.autoconf;
6636         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6637
6638         dflt->stable_secret.initialized = false;
6639         all->stable_secret.initialized = false;
6640
6641         net->ipv6.devconf_all = all;
6642         net->ipv6.devconf_dflt = dflt;
6643
6644 #ifdef CONFIG_SYSCTL
6645         err = __addrconf_sysctl_register(net, "all", NULL, all);
6646         if (err < 0)
6647                 goto err_reg_all;
6648
6649         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6650         if (err < 0)
6651                 goto err_reg_dflt;
6652 #endif
6653         return 0;
6654
6655 #ifdef CONFIG_SYSCTL
6656 err_reg_dflt:
6657         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6658 err_reg_all:
6659         kfree(dflt);
6660 #endif
6661 err_alloc_dflt:
6662         kfree(all);
6663 err_alloc_all:
6664         return err;
6665 }
6666
6667 static void __net_exit addrconf_exit_net(struct net *net)
6668 {
6669 #ifdef CONFIG_SYSCTL
6670         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6671                                      NETCONFA_IFINDEX_DEFAULT);
6672         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6673                                      NETCONFA_IFINDEX_ALL);
6674 #endif
6675         kfree(net->ipv6.devconf_dflt);
6676         kfree(net->ipv6.devconf_all);
6677 }
6678
6679 static struct pernet_operations addrconf_ops = {
6680         .init = addrconf_init_net,
6681         .exit = addrconf_exit_net,
6682 };
6683
6684 static struct rtnl_af_ops inet6_ops __read_mostly = {
6685         .family           = AF_INET6,
6686         .fill_link_af     = inet6_fill_link_af,
6687         .get_link_af_size = inet6_get_link_af_size,
6688         .validate_link_af = inet6_validate_link_af,
6689         .set_link_af      = inet6_set_link_af,
6690 };
6691
6692 /*
6693  *      Init / cleanup code
6694  */
6695
6696 int __init addrconf_init(void)
6697 {
6698         struct inet6_dev *idev;
6699         int i, err;
6700
6701         err = ipv6_addr_label_init();
6702         if (err < 0) {
6703                 pr_crit("%s: cannot initialize default policy table: %d\n",
6704                         __func__, err);
6705                 goto out;
6706         }
6707
6708         err = register_pernet_subsys(&addrconf_ops);
6709         if (err < 0)
6710                 goto out_addrlabel;
6711
6712         addrconf_wq = create_workqueue("ipv6_addrconf");
6713         if (!addrconf_wq) {
6714                 err = -ENOMEM;
6715                 goto out_nowq;
6716         }
6717
6718         /* The addrconf netdev notifier requires that loopback_dev
6719          * has it's ipv6 private information allocated and setup
6720          * before it can bring up and give link-local addresses
6721          * to other devices which are up.
6722          *
6723          * Unfortunately, loopback_dev is not necessarily the first
6724          * entry in the global dev_base list of net devices.  In fact,
6725          * it is likely to be the very last entry on that list.
6726          * So this causes the notifier registry below to try and
6727          * give link-local addresses to all devices besides loopback_dev
6728          * first, then loopback_dev, which cases all the non-loopback_dev
6729          * devices to fail to get a link-local address.
6730          *
6731          * So, as a temporary fix, allocate the ipv6 structure for
6732          * loopback_dev first by hand.
6733          * Longer term, all of the dependencies ipv6 has upon the loopback
6734          * device and it being up should be removed.
6735          */
6736         rtnl_lock();
6737         idev = ipv6_add_dev(init_net.loopback_dev);
6738         rtnl_unlock();
6739         if (IS_ERR(idev)) {
6740                 err = PTR_ERR(idev);
6741                 goto errlo;
6742         }
6743
6744         ip6_route_init_special_entries();
6745
6746         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6747                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6748
6749         register_netdevice_notifier(&ipv6_dev_notf);
6750
6751         addrconf_verify();
6752
6753         rtnl_af_register(&inet6_ops);
6754
6755         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6756                                    NULL, inet6_dump_ifinfo, 0);
6757         if (err < 0)
6758                 goto errout;
6759
6760         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6761                                    inet6_rtm_newaddr, NULL, 0);
6762         if (err < 0)
6763                 goto errout;
6764         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6765                                    inet6_rtm_deladdr, NULL, 0);
6766         if (err < 0)
6767                 goto errout;
6768         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6769                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
6770                                    RTNL_FLAG_DOIT_UNLOCKED);
6771         if (err < 0)
6772                 goto errout;
6773         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6774                                    NULL, inet6_dump_ifmcaddr, 0);
6775         if (err < 0)
6776                 goto errout;
6777         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6778                                    NULL, inet6_dump_ifacaddr, 0);
6779         if (err < 0)
6780                 goto errout;
6781         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6782                                    inet6_netconf_get_devconf,
6783                                    inet6_netconf_dump_devconf,
6784                                    RTNL_FLAG_DOIT_UNLOCKED);
6785         if (err < 0)
6786                 goto errout;
6787         err = ipv6_addr_label_rtnl_register();
6788         if (err < 0)
6789                 goto errout;
6790
6791         return 0;
6792 errout:
6793         rtnl_unregister_all(PF_INET6);
6794         rtnl_af_unregister(&inet6_ops);
6795         unregister_netdevice_notifier(&ipv6_dev_notf);
6796 errlo:
6797         destroy_workqueue(addrconf_wq);
6798 out_nowq:
6799         unregister_pernet_subsys(&addrconf_ops);
6800 out_addrlabel:
6801         ipv6_addr_label_cleanup();
6802 out:
6803         return err;
6804 }
6805
6806 void addrconf_cleanup(void)
6807 {
6808         struct net_device *dev;
6809         int i;
6810
6811         unregister_netdevice_notifier(&ipv6_dev_notf);
6812         unregister_pernet_subsys(&addrconf_ops);
6813         ipv6_addr_label_cleanup();
6814
6815         rtnl_af_unregister(&inet6_ops);
6816
6817         rtnl_lock();
6818
6819         /* clean dev list */
6820         for_each_netdev(&init_net, dev) {
6821                 if (__in6_dev_get(dev) == NULL)
6822                         continue;
6823                 addrconf_ifdown(dev, 1);
6824         }
6825         addrconf_ifdown(init_net.loopback_dev, 2);
6826
6827         /*
6828          *      Check hash table.
6829          */
6830         spin_lock_bh(&addrconf_hash_lock);
6831         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6832                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6833         spin_unlock_bh(&addrconf_hash_lock);
6834         cancel_delayed_work(&addr_chk_work);
6835         rtnl_unlock();
6836
6837         destroy_workqueue(addrconf_wq);
6838 }