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, bool restart);
 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              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1000             (!(idev->dev->flags & IFF_LOOPBACK) &&
1001              addr_type & IPV6_ADDR_LOOPBACK))
1002                 return ERR_PTR(-EADDRNOTAVAIL);
1003
1004         if (idev->dead) {
1005                 err = -ENODEV;                  /*XXX*/
1006                 goto out;
1007         }
1008
1009         if (idev->cnf.disable_ipv6) {
1010                 err = -EACCES;
1011                 goto out;
1012         }
1013
1014         /* validator notifier needs to be blocking;
1015          * do not call in atomic context
1016          */
1017         if (can_block) {
1018                 struct in6_validator_info i6vi = {
1019                         .i6vi_addr = *cfg->pfx,
1020                         .i6vi_dev = idev,
1021                         .extack = extack,
1022                 };
1023
1024                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1025                 err = notifier_to_errno(err);
1026                 if (err < 0)
1027                         goto out;
1028         }
1029
1030         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1031         if (!ifa) {
1032                 err = -ENOBUFS;
1033                 goto out;
1034         }
1035
1036         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1037         if (IS_ERR(f6i)) {
1038                 err = PTR_ERR(f6i);
1039                 f6i = NULL;
1040                 goto out;
1041         }
1042
1043         if (net->ipv6.devconf_all->disable_policy ||
1044             idev->cnf.disable_policy)
1045                 f6i->dst_nopolicy = true;
1046
1047         neigh_parms_data_state_setall(idev->nd_parms);
1048
1049         ifa->addr = *cfg->pfx;
1050         if (cfg->peer_pfx)
1051                 ifa->peer_addr = *cfg->peer_pfx;
1052
1053         spin_lock_init(&ifa->lock);
1054         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1055         INIT_HLIST_NODE(&ifa->addr_lst);
1056         ifa->scope = cfg->scope;
1057         ifa->prefix_len = cfg->plen;
1058         ifa->rt_priority = cfg->rt_priority;
1059         ifa->flags = cfg->ifa_flags;
1060         /* No need to add the TENTATIVE flag for addresses with NODAD */
1061         if (!(cfg->ifa_flags & IFA_F_NODAD))
1062                 ifa->flags |= IFA_F_TENTATIVE;
1063         ifa->valid_lft = cfg->valid_lft;
1064         ifa->prefered_lft = cfg->preferred_lft;
1065         ifa->cstamp = ifa->tstamp = jiffies;
1066         ifa->tokenized = false;
1067
1068         ifa->rt = f6i;
1069
1070         ifa->idev = idev;
1071         in6_dev_hold(idev);
1072
1073         /* For caller */
1074         refcount_set(&ifa->refcnt, 1);
1075
1076         rcu_read_lock_bh();
1077
1078         err = ipv6_add_addr_hash(idev->dev, ifa);
1079         if (err < 0) {
1080                 rcu_read_unlock_bh();
1081                 goto out;
1082         }
1083
1084         write_lock(&idev->lock);
1085
1086         /* Add to inet6_dev unicast addr list. */
1087         ipv6_link_dev_addr(idev, ifa);
1088
1089         if (ifa->flags&IFA_F_TEMPORARY) {
1090                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1091                 in6_ifa_hold(ifa);
1092         }
1093
1094         in6_ifa_hold(ifa);
1095         write_unlock(&idev->lock);
1096
1097         rcu_read_unlock_bh();
1098
1099         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1100 out:
1101         if (unlikely(err < 0)) {
1102                 fib6_info_release(f6i);
1103
1104                 if (ifa) {
1105                         if (ifa->idev)
1106                                 in6_dev_put(ifa->idev);
1107                         kfree(ifa);
1108                 }
1109                 ifa = ERR_PTR(err);
1110         }
1111
1112         return ifa;
1113 }
1114
1115 enum cleanup_prefix_rt_t {
1116         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1117         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1118         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1119 };
1120
1121 /*
1122  * Check, whether the prefix for ifp would still need a prefix route
1123  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1124  * constants.
1125  *
1126  * 1) we don't purge prefix if address was not permanent.
1127  *    prefix is managed by its own lifetime.
1128  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1129  * 3) if there are no addresses, delete prefix.
1130  * 4) if there are still other permanent address(es),
1131  *    corresponding prefix is still permanent.
1132  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1133  *    don't purge the prefix, assume user space is managing it.
1134  * 6) otherwise, update prefix lifetime to the
1135  *    longest valid lifetime among the corresponding
1136  *    addresses on the device.
1137  *    Note: subsequent RA will update lifetime.
1138  **/
1139 static enum cleanup_prefix_rt_t
1140 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1141 {
1142         struct inet6_ifaddr *ifa;
1143         struct inet6_dev *idev = ifp->idev;
1144         unsigned long lifetime;
1145         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1146
1147         *expires = jiffies;
1148
1149         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1150                 if (ifa == ifp)
1151                         continue;
1152                 if (ifa->prefix_len != ifp->prefix_len ||
1153                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1154                                        ifp->prefix_len))
1155                         continue;
1156                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1157                         return CLEANUP_PREFIX_RT_NOP;
1158
1159                 action = CLEANUP_PREFIX_RT_EXPIRE;
1160
1161                 spin_lock(&ifa->lock);
1162
1163                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1164                 /*
1165                  * Note: Because this address is
1166                  * not permanent, lifetime <
1167                  * LONG_MAX / HZ here.
1168                  */
1169                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1170                         *expires = ifa->tstamp + lifetime * HZ;
1171                 spin_unlock(&ifa->lock);
1172         }
1173
1174         return action;
1175 }
1176
1177 static void
1178 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1179                      bool del_rt, bool del_peer)
1180 {
1181         struct fib6_info *f6i;
1182
1183         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1184                                        ifp->prefix_len,
1185                                        ifp->idev->dev,
1186                                        0, RTF_GATEWAY | RTF_DEFAULT);
1187         if (f6i) {
1188                 if (del_rt)
1189                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1190                 else {
1191                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1192                                 fib6_set_expires(f6i, expires);
1193                         fib6_info_release(f6i);
1194                 }
1195         }
1196 }
1197
1198
1199 /* This function wants to get referenced ifp and releases it before return */
1200
1201 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1202 {
1203         int state;
1204         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1205         unsigned long expires;
1206
1207         ASSERT_RTNL();
1208
1209         spin_lock_bh(&ifp->lock);
1210         state = ifp->state;
1211         ifp->state = INET6_IFADDR_STATE_DEAD;
1212         spin_unlock_bh(&ifp->lock);
1213
1214         if (state == INET6_IFADDR_STATE_DEAD)
1215                 goto out;
1216
1217         spin_lock_bh(&addrconf_hash_lock);
1218         hlist_del_init_rcu(&ifp->addr_lst);
1219         spin_unlock_bh(&addrconf_hash_lock);
1220
1221         write_lock_bh(&ifp->idev->lock);
1222
1223         if (ifp->flags&IFA_F_TEMPORARY) {
1224                 list_del(&ifp->tmp_list);
1225                 if (ifp->ifpub) {
1226                         in6_ifa_put(ifp->ifpub);
1227                         ifp->ifpub = NULL;
1228                 }
1229                 __in6_ifa_put(ifp);
1230         }
1231
1232         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1233                 action = check_cleanup_prefix_route(ifp, &expires);
1234
1235         list_del_rcu(&ifp->if_list);
1236         __in6_ifa_put(ifp);
1237
1238         write_unlock_bh(&ifp->idev->lock);
1239
1240         addrconf_del_dad_work(ifp);
1241
1242         ipv6_ifa_notify(RTM_DELADDR, ifp);
1243
1244         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1245
1246         if (action != CLEANUP_PREFIX_RT_NOP) {
1247                 cleanup_prefix_route(ifp, expires,
1248                         action == CLEANUP_PREFIX_RT_DEL, false);
1249         }
1250
1251         /* clean up prefsrc entries */
1252         rt6_remove_prefsrc(ifp);
1253 out:
1254         in6_ifa_put(ifp);
1255 }
1256
1257 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1258                                 struct inet6_ifaddr *ift,
1259                                 bool block)
1260 {
1261         struct inet6_dev *idev = ifp->idev;
1262         struct in6_addr addr, *tmpaddr;
1263         unsigned long tmp_tstamp, age;
1264         unsigned long regen_advance;
1265         struct ifa6_config cfg;
1266         int ret = 0;
1267         unsigned long now = jiffies;
1268         long max_desync_factor;
1269         s32 cnf_temp_preferred_lft;
1270
1271         write_lock_bh(&idev->lock);
1272         if (ift) {
1273                 spin_lock_bh(&ift->lock);
1274                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1275                 spin_unlock_bh(&ift->lock);
1276                 tmpaddr = &addr;
1277         } else {
1278                 tmpaddr = NULL;
1279         }
1280 retry:
1281         in6_dev_hold(idev);
1282         if (idev->cnf.use_tempaddr <= 0) {
1283                 write_unlock_bh(&idev->lock);
1284                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1285                 in6_dev_put(idev);
1286                 ret = -1;
1287                 goto out;
1288         }
1289         spin_lock_bh(&ifp->lock);
1290         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1291                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1292                 spin_unlock_bh(&ifp->lock);
1293                 write_unlock_bh(&idev->lock);
1294                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1295                         __func__);
1296                 in6_dev_put(idev);
1297                 ret = -1;
1298                 goto out;
1299         }
1300         in6_ifa_hold(ifp);
1301         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1302         ipv6_try_regen_rndid(idev, tmpaddr);
1303         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1304         age = (now - ifp->tstamp) / HZ;
1305
1306         regen_advance = idev->cnf.regen_max_retry *
1307                         idev->cnf.dad_transmits *
1308                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1309
1310         /* recalculate max_desync_factor each time and update
1311          * idev->desync_factor if it's larger
1312          */
1313         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1314         max_desync_factor = min_t(__u32,
1315                                   idev->cnf.max_desync_factor,
1316                                   cnf_temp_preferred_lft - regen_advance);
1317
1318         if (unlikely(idev->desync_factor > max_desync_factor)) {
1319                 if (max_desync_factor > 0) {
1320                         get_random_bytes(&idev->desync_factor,
1321                                          sizeof(idev->desync_factor));
1322                         idev->desync_factor %= max_desync_factor;
1323                 } else {
1324                         idev->desync_factor = 0;
1325                 }
1326         }
1327
1328         memset(&cfg, 0, sizeof(cfg));
1329         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1330                               idev->cnf.temp_valid_lft + age);
1331         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1332         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1333
1334         cfg.plen = ifp->prefix_len;
1335         tmp_tstamp = ifp->tstamp;
1336         spin_unlock_bh(&ifp->lock);
1337
1338         write_unlock_bh(&idev->lock);
1339
1340         /* A temporary address is created only if this calculated Preferred
1341          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1342          * an implementation must not create a temporary address with a zero
1343          * Preferred Lifetime.
1344          * Use age calculation as in addrconf_verify to avoid unnecessary
1345          * temporary addresses being generated.
1346          */
1347         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1348         if (cfg.preferred_lft <= regen_advance + age) {
1349                 in6_ifa_put(ifp);
1350                 in6_dev_put(idev);
1351                 ret = -1;
1352                 goto out;
1353         }
1354
1355         cfg.ifa_flags = IFA_F_TEMPORARY;
1356         /* set in addrconf_prefix_rcv() */
1357         if (ifp->flags & IFA_F_OPTIMISTIC)
1358                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1359
1360         cfg.pfx = &addr;
1361         cfg.scope = ipv6_addr_scope(cfg.pfx);
1362
1363         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1364         if (IS_ERR(ift)) {
1365                 in6_ifa_put(ifp);
1366                 in6_dev_put(idev);
1367                 pr_info("%s: retry temporary address regeneration\n", __func__);
1368                 tmpaddr = &addr;
1369                 write_lock_bh(&idev->lock);
1370                 goto retry;
1371         }
1372
1373         spin_lock_bh(&ift->lock);
1374         ift->ifpub = ifp;
1375         ift->cstamp = now;
1376         ift->tstamp = tmp_tstamp;
1377         spin_unlock_bh(&ift->lock);
1378
1379         addrconf_dad_start(ift);
1380         in6_ifa_put(ift);
1381         in6_dev_put(idev);
1382 out:
1383         return ret;
1384 }
1385
1386 /*
1387  *      Choose an appropriate source address (RFC3484)
1388  */
1389 enum {
1390         IPV6_SADDR_RULE_INIT = 0,
1391         IPV6_SADDR_RULE_LOCAL,
1392         IPV6_SADDR_RULE_SCOPE,
1393         IPV6_SADDR_RULE_PREFERRED,
1394 #ifdef CONFIG_IPV6_MIP6
1395         IPV6_SADDR_RULE_HOA,
1396 #endif
1397         IPV6_SADDR_RULE_OIF,
1398         IPV6_SADDR_RULE_LABEL,
1399         IPV6_SADDR_RULE_PRIVACY,
1400         IPV6_SADDR_RULE_ORCHID,
1401         IPV6_SADDR_RULE_PREFIX,
1402 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1403         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1404 #endif
1405         IPV6_SADDR_RULE_MAX
1406 };
1407
1408 struct ipv6_saddr_score {
1409         int                     rule;
1410         int                     addr_type;
1411         struct inet6_ifaddr     *ifa;
1412         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1413         int                     scopedist;
1414         int                     matchlen;
1415 };
1416
1417 struct ipv6_saddr_dst {
1418         const struct in6_addr *addr;
1419         int ifindex;
1420         int scope;
1421         int label;
1422         unsigned int prefs;
1423 };
1424
1425 static inline int ipv6_saddr_preferred(int type)
1426 {
1427         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1428                 return 1;
1429         return 0;
1430 }
1431
1432 static bool ipv6_use_optimistic_addr(struct net *net,
1433                                      struct inet6_dev *idev)
1434 {
1435 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1436         if (!idev)
1437                 return false;
1438         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1439                 return false;
1440         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1441                 return false;
1442
1443         return true;
1444 #else
1445         return false;
1446 #endif
1447 }
1448
1449 static bool ipv6_allow_optimistic_dad(struct net *net,
1450                                       struct inet6_dev *idev)
1451 {
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453         if (!idev)
1454                 return false;
1455         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1456                 return false;
1457
1458         return true;
1459 #else
1460         return false;
1461 #endif
1462 }
1463
1464 static int ipv6_get_saddr_eval(struct net *net,
1465                                struct ipv6_saddr_score *score,
1466                                struct ipv6_saddr_dst *dst,
1467                                int i)
1468 {
1469         int ret;
1470
1471         if (i <= score->rule) {
1472                 switch (i) {
1473                 case IPV6_SADDR_RULE_SCOPE:
1474                         ret = score->scopedist;
1475                         break;
1476                 case IPV6_SADDR_RULE_PREFIX:
1477                         ret = score->matchlen;
1478                         break;
1479                 default:
1480                         ret = !!test_bit(i, score->scorebits);
1481                 }
1482                 goto out;
1483         }
1484
1485         switch (i) {
1486         case IPV6_SADDR_RULE_INIT:
1487                 /* Rule 0: remember if hiscore is not ready yet */
1488                 ret = !!score->ifa;
1489                 break;
1490         case IPV6_SADDR_RULE_LOCAL:
1491                 /* Rule 1: Prefer same address */
1492                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1493                 break;
1494         case IPV6_SADDR_RULE_SCOPE:
1495                 /* Rule 2: Prefer appropriate scope
1496                  *
1497                  *      ret
1498                  *       ^
1499                  *    -1 |  d 15
1500                  *    ---+--+-+---> scope
1501                  *       |
1502                  *       |             d is scope of the destination.
1503                  *  B-d  |  \
1504                  *       |   \      <- smaller scope is better if
1505                  *  B-15 |    \        if scope is enough for destination.
1506                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1507                  * d-C-1 | /
1508                  *       |/         <- greater is better
1509                  *   -C  /             if scope is not enough for destination.
1510                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1511                  *
1512                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1513                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1514                  * Assume B = 0 and we get C > 29.
1515                  */
1516                 ret = __ipv6_addr_src_scope(score->addr_type);
1517                 if (ret >= dst->scope)
1518                         ret = -ret;
1519                 else
1520                         ret -= 128;     /* 30 is enough */
1521                 score->scopedist = ret;
1522                 break;
1523         case IPV6_SADDR_RULE_PREFERRED:
1524             {
1525                 /* Rule 3: Avoid deprecated and optimistic addresses */
1526                 u8 avoid = IFA_F_DEPRECATED;
1527
1528                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1529                         avoid |= IFA_F_OPTIMISTIC;
1530                 ret = ipv6_saddr_preferred(score->addr_type) ||
1531                       !(score->ifa->flags & avoid);
1532                 break;
1533             }
1534 #ifdef CONFIG_IPV6_MIP6
1535         case IPV6_SADDR_RULE_HOA:
1536             {
1537                 /* Rule 4: Prefer home address */
1538                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1539                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1540                 break;
1541             }
1542 #endif
1543         case IPV6_SADDR_RULE_OIF:
1544                 /* Rule 5: Prefer outgoing interface */
1545                 ret = (!dst->ifindex ||
1546                        dst->ifindex == score->ifa->idev->dev->ifindex);
1547                 break;
1548         case IPV6_SADDR_RULE_LABEL:
1549                 /* Rule 6: Prefer matching label */
1550                 ret = ipv6_addr_label(net,
1551                                       &score->ifa->addr, score->addr_type,
1552                                       score->ifa->idev->dev->ifindex) == dst->label;
1553                 break;
1554         case IPV6_SADDR_RULE_PRIVACY:
1555             {
1556                 /* Rule 7: Prefer public address
1557                  * Note: prefer temporary address if use_tempaddr >= 2
1558                  */
1559                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1560                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1561                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1562                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1563                 break;
1564             }
1565         case IPV6_SADDR_RULE_ORCHID:
1566                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1567                  *          non-ORCHID vs non-ORCHID
1568                  */
1569                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1570                         ipv6_addr_orchid(dst->addr));
1571                 break;
1572         case IPV6_SADDR_RULE_PREFIX:
1573                 /* Rule 8: Use longest matching prefix */
1574                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1575                 if (ret > score->ifa->prefix_len)
1576                         ret = score->ifa->prefix_len;
1577                 score->matchlen = ret;
1578                 break;
1579 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1580         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1581                 /* Optimistic addresses still have lower precedence than other
1582                  * preferred addresses.
1583                  */
1584                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1585                 break;
1586 #endif
1587         default:
1588                 ret = 0;
1589         }
1590
1591         if (ret)
1592                 __set_bit(i, score->scorebits);
1593         score->rule = i;
1594 out:
1595         return ret;
1596 }
1597
1598 static int __ipv6_dev_get_saddr(struct net *net,
1599                                 struct ipv6_saddr_dst *dst,
1600                                 struct inet6_dev *idev,
1601                                 struct ipv6_saddr_score *scores,
1602                                 int hiscore_idx)
1603 {
1604         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1605
1606         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1607                 int i;
1608
1609                 /*
1610                  * - Tentative Address (RFC2462 section 5.4)
1611                  *  - A tentative address is not considered
1612                  *    "assigned to an interface" in the traditional
1613                  *    sense, unless it is also flagged as optimistic.
1614                  * - Candidate Source Address (section 4)
1615                  *  - In any case, anycast addresses, multicast
1616                  *    addresses, and the unspecified address MUST
1617                  *    NOT be included in a candidate set.
1618                  */
1619                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1620                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1621                         continue;
1622
1623                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1624
1625                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1626                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1627                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1628                                             idev->dev->name);
1629                         continue;
1630                 }
1631
1632                 score->rule = -1;
1633                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1634
1635                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1636                         int minihiscore, miniscore;
1637
1638                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1639                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1640
1641                         if (minihiscore > miniscore) {
1642                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1643                                     score->scopedist > 0) {
1644                                         /*
1645                                          * special case:
1646                                          * each remaining entry
1647                                          * has too small (not enough)
1648                                          * scope, because ifa entries
1649                                          * are sorted by their scope
1650                                          * values.
1651                                          */
1652                                         goto out;
1653                                 }
1654                                 break;
1655                         } else if (minihiscore < miniscore) {
1656                                 swap(hiscore, score);
1657                                 hiscore_idx = 1 - hiscore_idx;
1658
1659                                 /* restore our iterator */
1660                                 score->ifa = hiscore->ifa;
1661
1662                                 break;
1663                         }
1664                 }
1665         }
1666 out:
1667         return hiscore_idx;
1668 }
1669
1670 static int ipv6_get_saddr_master(struct net *net,
1671                                  const struct net_device *dst_dev,
1672                                  const struct net_device *master,
1673                                  struct ipv6_saddr_dst *dst,
1674                                  struct ipv6_saddr_score *scores,
1675                                  int hiscore_idx)
1676 {
1677         struct inet6_dev *idev;
1678
1679         idev = __in6_dev_get(dst_dev);
1680         if (idev)
1681                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1682                                                    scores, hiscore_idx);
1683
1684         idev = __in6_dev_get(master);
1685         if (idev)
1686                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1687                                                    scores, hiscore_idx);
1688
1689         return hiscore_idx;
1690 }
1691
1692 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1693                        const struct in6_addr *daddr, unsigned int prefs,
1694                        struct in6_addr *saddr)
1695 {
1696         struct ipv6_saddr_score scores[2], *hiscore;
1697         struct ipv6_saddr_dst dst;
1698         struct inet6_dev *idev;
1699         struct net_device *dev;
1700         int dst_type;
1701         bool use_oif_addr = false;
1702         int hiscore_idx = 0;
1703         int ret = 0;
1704
1705         dst_type = __ipv6_addr_type(daddr);
1706         dst.addr = daddr;
1707         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1708         dst.scope = __ipv6_addr_src_scope(dst_type);
1709         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1710         dst.prefs = prefs;
1711
1712         scores[hiscore_idx].rule = -1;
1713         scores[hiscore_idx].ifa = NULL;
1714
1715         rcu_read_lock();
1716
1717         /* Candidate Source Address (section 4)
1718          *  - multicast and link-local destination address,
1719          *    the set of candidate source address MUST only
1720          *    include addresses assigned to interfaces
1721          *    belonging to the same link as the outgoing
1722          *    interface.
1723          * (- For site-local destination addresses, the
1724          *    set of candidate source addresses MUST only
1725          *    include addresses assigned to interfaces
1726          *    belonging to the same site as the outgoing
1727          *    interface.)
1728          *  - "It is RECOMMENDED that the candidate source addresses
1729          *    be the set of unicast addresses assigned to the
1730          *    interface that will be used to send to the destination
1731          *    (the 'outgoing' interface)." (RFC 6724)
1732          */
1733         if (dst_dev) {
1734                 idev = __in6_dev_get(dst_dev);
1735                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1736                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1737                     (idev && idev->cnf.use_oif_addrs_only)) {
1738                         use_oif_addr = true;
1739                 }
1740         }
1741
1742         if (use_oif_addr) {
1743                 if (idev)
1744                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1745         } else {
1746                 const struct net_device *master;
1747                 int master_idx = 0;
1748
1749                 /* if dst_dev exists and is enslaved to an L3 device, then
1750                  * prefer addresses from dst_dev and then the master over
1751                  * any other enslaved devices in the L3 domain.
1752                  */
1753                 master = l3mdev_master_dev_rcu(dst_dev);
1754                 if (master) {
1755                         master_idx = master->ifindex;
1756
1757                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1758                                                             master, &dst,
1759                                                             scores, hiscore_idx);
1760
1761                         if (scores[hiscore_idx].ifa)
1762                                 goto out;
1763                 }
1764
1765                 for_each_netdev_rcu(net, dev) {
1766                         /* only consider addresses on devices in the
1767                          * same L3 domain
1768                          */
1769                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1770                                 continue;
1771                         idev = __in6_dev_get(dev);
1772                         if (!idev)
1773                                 continue;
1774                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1775                 }
1776         }
1777
1778 out:
1779         hiscore = &scores[hiscore_idx];
1780         if (!hiscore->ifa)
1781                 ret = -EADDRNOTAVAIL;
1782         else
1783                 *saddr = hiscore->ifa->addr;
1784
1785         rcu_read_unlock();
1786         return ret;
1787 }
1788 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1789
1790 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1791                       u32 banned_flags)
1792 {
1793         struct inet6_ifaddr *ifp;
1794         int err = -EADDRNOTAVAIL;
1795
1796         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1797                 if (ifp->scope > IFA_LINK)
1798                         break;
1799                 if (ifp->scope == IFA_LINK &&
1800                     !(ifp->flags & banned_flags)) {
1801                         *addr = ifp->addr;
1802                         err = 0;
1803                         break;
1804                 }
1805         }
1806         return err;
1807 }
1808
1809 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1810                     u32 banned_flags)
1811 {
1812         struct inet6_dev *idev;
1813         int err = -EADDRNOTAVAIL;
1814
1815         rcu_read_lock();
1816         idev = __in6_dev_get(dev);
1817         if (idev) {
1818                 read_lock_bh(&idev->lock);
1819                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1820                 read_unlock_bh(&idev->lock);
1821         }
1822         rcu_read_unlock();
1823         return err;
1824 }
1825
1826 static int ipv6_count_addresses(const struct inet6_dev *idev)
1827 {
1828         const struct inet6_ifaddr *ifp;
1829         int cnt = 0;
1830
1831         rcu_read_lock();
1832         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1833                 cnt++;
1834         rcu_read_unlock();
1835         return cnt;
1836 }
1837
1838 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1839                   const struct net_device *dev, int strict)
1840 {
1841         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1842                                        strict, IFA_F_TENTATIVE);
1843 }
1844 EXPORT_SYMBOL(ipv6_chk_addr);
1845
1846 /* device argument is used to find the L3 domain of interest. If
1847  * skip_dev_check is set, then the ifp device is not checked against
1848  * the passed in dev argument. So the 2 cases for addresses checks are:
1849  *   1. does the address exist in the L3 domain that dev is part of
1850  *      (skip_dev_check = true), or
1851  *
1852  *   2. does the address exist on the specific device
1853  *      (skip_dev_check = false)
1854  */
1855 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1856                             const struct net_device *dev, bool skip_dev_check,
1857                             int strict, u32 banned_flags)
1858 {
1859         unsigned int hash = inet6_addr_hash(net, addr);
1860         const struct net_device *l3mdev;
1861         struct inet6_ifaddr *ifp;
1862         u32 ifp_flags;
1863
1864         rcu_read_lock();
1865
1866         l3mdev = l3mdev_master_dev_rcu(dev);
1867         if (skip_dev_check)
1868                 dev = NULL;
1869
1870         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1871                 if (!net_eq(dev_net(ifp->idev->dev), net))
1872                         continue;
1873
1874                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1875                         continue;
1876
1877                 /* Decouple optimistic from tentative for evaluation here.
1878                  * Ban optimistic addresses explicitly, when required.
1879                  */
1880                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1881                             ? (ifp->flags&~IFA_F_TENTATIVE)
1882                             : ifp->flags;
1883                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1884                     !(ifp_flags&banned_flags) &&
1885                     (!dev || ifp->idev->dev == dev ||
1886                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1887                         rcu_read_unlock();
1888                         return 1;
1889                 }
1890         }
1891
1892         rcu_read_unlock();
1893         return 0;
1894 }
1895 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1896
1897
1898 /* Compares an address/prefix_len with addresses on device @dev.
1899  * If one is found it returns true.
1900  */
1901 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1902         const unsigned int prefix_len, struct net_device *dev)
1903 {
1904         const struct inet6_ifaddr *ifa;
1905         const struct inet6_dev *idev;
1906         bool ret = false;
1907
1908         rcu_read_lock();
1909         idev = __in6_dev_get(dev);
1910         if (idev) {
1911                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1912                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1913                         if (ret)
1914                                 break;
1915                 }
1916         }
1917         rcu_read_unlock();
1918
1919         return ret;
1920 }
1921 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1922
1923 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1924 {
1925         const struct inet6_ifaddr *ifa;
1926         const struct inet6_dev *idev;
1927         int     onlink;
1928
1929         onlink = 0;
1930         rcu_read_lock();
1931         idev = __in6_dev_get(dev);
1932         if (idev) {
1933                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1934                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1935                                                    ifa->prefix_len);
1936                         if (onlink)
1937                                 break;
1938                 }
1939         }
1940         rcu_read_unlock();
1941         return onlink;
1942 }
1943 EXPORT_SYMBOL(ipv6_chk_prefix);
1944
1945 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1946                                      struct net_device *dev, int strict)
1947 {
1948         unsigned int hash = inet6_addr_hash(net, addr);
1949         struct inet6_ifaddr *ifp, *result = NULL;
1950
1951         rcu_read_lock();
1952         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1953                 if (!net_eq(dev_net(ifp->idev->dev), net))
1954                         continue;
1955                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1956                         if (!dev || ifp->idev->dev == dev ||
1957                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1958                                 result = ifp;
1959                                 in6_ifa_hold(ifp);
1960                                 break;
1961                         }
1962                 }
1963         }
1964         rcu_read_unlock();
1965
1966         return result;
1967 }
1968
1969 /* Gets referenced address, destroys ifaddr */
1970
1971 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1972 {
1973         if (dad_failed)
1974                 ifp->flags |= IFA_F_DADFAILED;
1975
1976         if (ifp->flags&IFA_F_TEMPORARY) {
1977                 struct inet6_ifaddr *ifpub;
1978                 spin_lock_bh(&ifp->lock);
1979                 ifpub = ifp->ifpub;
1980                 if (ifpub) {
1981                         in6_ifa_hold(ifpub);
1982                         spin_unlock_bh(&ifp->lock);
1983                         ipv6_create_tempaddr(ifpub, ifp, true);
1984                         in6_ifa_put(ifpub);
1985                 } else {
1986                         spin_unlock_bh(&ifp->lock);
1987                 }
1988                 ipv6_del_addr(ifp);
1989         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1990                 spin_lock_bh(&ifp->lock);
1991                 addrconf_del_dad_work(ifp);
1992                 ifp->flags |= IFA_F_TENTATIVE;
1993                 if (dad_failed)
1994                         ifp->flags &= ~IFA_F_OPTIMISTIC;
1995                 spin_unlock_bh(&ifp->lock);
1996                 if (dad_failed)
1997                         ipv6_ifa_notify(0, ifp);
1998                 in6_ifa_put(ifp);
1999         } else {
2000                 ipv6_del_addr(ifp);
2001         }
2002 }
2003
2004 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2005 {
2006         int err = -ENOENT;
2007
2008         spin_lock_bh(&ifp->lock);
2009         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2010                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2011                 err = 0;
2012         }
2013         spin_unlock_bh(&ifp->lock);
2014
2015         return err;
2016 }
2017
2018 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2019 {
2020         struct inet6_dev *idev = ifp->idev;
2021         struct net *net = dev_net(ifp->idev->dev);
2022
2023         if (addrconf_dad_end(ifp)) {
2024                 in6_ifa_put(ifp);
2025                 return;
2026         }
2027
2028         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2029                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2030
2031         spin_lock_bh(&ifp->lock);
2032
2033         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2034                 struct in6_addr new_addr;
2035                 struct inet6_ifaddr *ifp2;
2036                 int retries = ifp->stable_privacy_retry + 1;
2037                 struct ifa6_config cfg = {
2038                         .pfx = &new_addr,
2039                         .plen = ifp->prefix_len,
2040                         .ifa_flags = ifp->flags,
2041                         .valid_lft = ifp->valid_lft,
2042                         .preferred_lft = ifp->prefered_lft,
2043                         .scope = ifp->scope,
2044                 };
2045
2046                 if (retries > net->ipv6.sysctl.idgen_retries) {
2047                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2048                                              ifp->idev->dev->name);
2049                         goto errdad;
2050                 }
2051
2052                 new_addr = ifp->addr;
2053                 if (ipv6_generate_stable_address(&new_addr, retries,
2054                                                  idev))
2055                         goto errdad;
2056
2057                 spin_unlock_bh(&ifp->lock);
2058
2059                 if (idev->cnf.max_addresses &&
2060                     ipv6_count_addresses(idev) >=
2061                     idev->cnf.max_addresses)
2062                         goto lock_errdad;
2063
2064                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2065                                      ifp->idev->dev->name);
2066
2067                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2068                 if (IS_ERR(ifp2))
2069                         goto lock_errdad;
2070
2071                 spin_lock_bh(&ifp2->lock);
2072                 ifp2->stable_privacy_retry = retries;
2073                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2074                 spin_unlock_bh(&ifp2->lock);
2075
2076                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2077                 in6_ifa_put(ifp2);
2078 lock_errdad:
2079                 spin_lock_bh(&ifp->lock);
2080         }
2081
2082 errdad:
2083         /* transition from _POSTDAD to _ERRDAD */
2084         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2085         spin_unlock_bh(&ifp->lock);
2086
2087         addrconf_mod_dad_work(ifp, 0);
2088         in6_ifa_put(ifp);
2089 }
2090
2091 /* Join to solicited addr multicast group.
2092  * caller must hold RTNL */
2093 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2094 {
2095         struct in6_addr maddr;
2096
2097         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2098                 return;
2099
2100         addrconf_addr_solict_mult(addr, &maddr);
2101         ipv6_dev_mc_inc(dev, &maddr);
2102 }
2103
2104 /* caller must hold RTNL */
2105 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2106 {
2107         struct in6_addr maddr;
2108
2109         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2110                 return;
2111
2112         addrconf_addr_solict_mult(addr, &maddr);
2113         __ipv6_dev_mc_dec(idev, &maddr);
2114 }
2115
2116 /* caller must hold RTNL */
2117 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2118 {
2119         struct in6_addr addr;
2120
2121         if (ifp->prefix_len >= 127) /* RFC 6164 */
2122                 return;
2123         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2124         if (ipv6_addr_any(&addr))
2125                 return;
2126         __ipv6_dev_ac_inc(ifp->idev, &addr);
2127 }
2128
2129 /* caller must hold RTNL */
2130 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2131 {
2132         struct in6_addr addr;
2133
2134         if (ifp->prefix_len >= 127) /* RFC 6164 */
2135                 return;
2136         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2137         if (ipv6_addr_any(&addr))
2138                 return;
2139         __ipv6_dev_ac_dec(ifp->idev, &addr);
2140 }
2141
2142 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2143 {
2144         switch (dev->addr_len) {
2145         case ETH_ALEN:
2146                 memcpy(eui, dev->dev_addr, 3);
2147                 eui[3] = 0xFF;
2148                 eui[4] = 0xFE;
2149                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2150                 break;
2151         case EUI64_ADDR_LEN:
2152                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2153                 eui[0] ^= 2;
2154                 break;
2155         default:
2156                 return -1;
2157         }
2158
2159         return 0;
2160 }
2161
2162 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2163 {
2164         union fwnet_hwaddr *ha;
2165
2166         if (dev->addr_len != FWNET_ALEN)
2167                 return -1;
2168
2169         ha = (union fwnet_hwaddr *)dev->dev_addr;
2170
2171         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2172         eui[0] ^= 2;
2173         return 0;
2174 }
2175
2176 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2177 {
2178         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2179         if (dev->addr_len != ARCNET_ALEN)
2180                 return -1;
2181         memset(eui, 0, 7);
2182         eui[7] = *(u8 *)dev->dev_addr;
2183         return 0;
2184 }
2185
2186 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2187 {
2188         if (dev->addr_len != INFINIBAND_ALEN)
2189                 return -1;
2190         memcpy(eui, dev->dev_addr + 12, 8);
2191         eui[0] |= 2;
2192         return 0;
2193 }
2194
2195 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2196 {
2197         if (addr == 0)
2198                 return -1;
2199         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2200                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2201                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2202                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2203                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2204                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2205         eui[1] = 0;
2206         eui[2] = 0x5E;
2207         eui[3] = 0xFE;
2208         memcpy(eui + 4, &addr, 4);
2209         return 0;
2210 }
2211
2212 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2213 {
2214         if (dev->priv_flags & IFF_ISATAP)
2215                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2216         return -1;
2217 }
2218
2219 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2220 {
2221         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2222 }
2223
2224 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2225 {
2226         memcpy(eui, dev->perm_addr, 3);
2227         memcpy(eui + 5, dev->perm_addr + 3, 3);
2228         eui[3] = 0xFF;
2229         eui[4] = 0xFE;
2230         eui[0] ^= 2;
2231         return 0;
2232 }
2233
2234 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2235 {
2236         switch (dev->type) {
2237         case ARPHRD_ETHER:
2238         case ARPHRD_FDDI:
2239                 return addrconf_ifid_eui48(eui, dev);
2240         case ARPHRD_ARCNET:
2241                 return addrconf_ifid_arcnet(eui, dev);
2242         case ARPHRD_INFINIBAND:
2243                 return addrconf_ifid_infiniband(eui, dev);
2244         case ARPHRD_SIT:
2245                 return addrconf_ifid_sit(eui, dev);
2246         case ARPHRD_IPGRE:
2247         case ARPHRD_TUNNEL:
2248                 return addrconf_ifid_gre(eui, dev);
2249         case ARPHRD_6LOWPAN:
2250                 return addrconf_ifid_6lowpan(eui, dev);
2251         case ARPHRD_IEEE1394:
2252                 return addrconf_ifid_ieee1394(eui, dev);
2253         case ARPHRD_TUNNEL6:
2254         case ARPHRD_IP6GRE:
2255         case ARPHRD_RAWIP:
2256                 return addrconf_ifid_ip6tnl(eui, dev);
2257         }
2258         return -1;
2259 }
2260
2261 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2262 {
2263         int err = -1;
2264         struct inet6_ifaddr *ifp;
2265
2266         read_lock_bh(&idev->lock);
2267         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2268                 if (ifp->scope > IFA_LINK)
2269                         break;
2270                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2271                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2272                         err = 0;
2273                         break;
2274                 }
2275         }
2276         read_unlock_bh(&idev->lock);
2277         return err;
2278 }
2279
2280 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2281 static void ipv6_regen_rndid(struct inet6_dev *idev)
2282 {
2283 regen:
2284         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2285         idev->rndid[0] &= ~0x02;
2286
2287         /*
2288          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2289          * check if generated address is not inappropriate
2290          *
2291          *  - Reserved subnet anycast (RFC 2526)
2292          *      11111101 11....11 1xxxxxxx
2293          *  - ISATAP (RFC4214) 6.1
2294          *      00-00-5E-FE-xx-xx-xx-xx
2295          *  - value 0
2296          *  - XXX: already assigned to an address on the device
2297          */
2298         if (idev->rndid[0] == 0xfd &&
2299             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2300             (idev->rndid[7]&0x80))
2301                 goto regen;
2302         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2303                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2304                         goto regen;
2305                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2306                         goto regen;
2307         }
2308 }
2309
2310 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2311 {
2312         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2313                 ipv6_regen_rndid(idev);
2314 }
2315
2316 /*
2317  *      Add prefix route.
2318  */
2319
2320 static void
2321 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2322                       struct net_device *dev, unsigned long expires,
2323                       u32 flags, gfp_t gfp_flags)
2324 {
2325         struct fib6_config cfg = {
2326                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2327                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2328                 .fc_ifindex = dev->ifindex,
2329                 .fc_expires = expires,
2330                 .fc_dst_len = plen,
2331                 .fc_flags = RTF_UP | flags,
2332                 .fc_nlinfo.nl_net = dev_net(dev),
2333                 .fc_protocol = RTPROT_KERNEL,
2334                 .fc_type = RTN_UNICAST,
2335         };
2336
2337         cfg.fc_dst = *pfx;
2338
2339         /* Prevent useless cloning on PtP SIT.
2340            This thing is done here expecting that the whole
2341            class of non-broadcast devices need not cloning.
2342          */
2343 #if IS_ENABLED(CONFIG_IPV6_SIT)
2344         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2345                 cfg.fc_flags |= RTF_NONEXTHOP;
2346 #endif
2347
2348         ip6_route_add(&cfg, gfp_flags, NULL);
2349 }
2350
2351
2352 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2353                                                   int plen,
2354                                                   const struct net_device *dev,
2355                                                   u32 flags, u32 noflags)
2356 {
2357         struct fib6_node *fn;
2358         struct fib6_info *rt = NULL;
2359         struct fib6_table *table;
2360         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2361
2362         table = fib6_get_table(dev_net(dev), tb_id);
2363         if (!table)
2364                 return NULL;
2365
2366         rcu_read_lock();
2367         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2368         if (!fn)
2369                 goto out;
2370
2371         for_each_fib6_node_rt_rcu(fn) {
2372                 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2373                         continue;
2374                 if ((rt->fib6_flags & flags) != flags)
2375                         continue;
2376                 if ((rt->fib6_flags & noflags) != 0)
2377                         continue;
2378                 if (!fib6_info_hold_safe(rt))
2379                         continue;
2380                 break;
2381         }
2382 out:
2383         rcu_read_unlock();
2384         return rt;
2385 }
2386
2387
2388 /* Create "default" multicast route to the interface */
2389
2390 static void addrconf_add_mroute(struct net_device *dev)
2391 {
2392         struct fib6_config cfg = {
2393                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2394                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2395                 .fc_ifindex = dev->ifindex,
2396                 .fc_dst_len = 8,
2397                 .fc_flags = RTF_UP,
2398                 .fc_type = RTN_MULTICAST,
2399                 .fc_nlinfo.nl_net = dev_net(dev),
2400                 .fc_protocol = RTPROT_KERNEL,
2401         };
2402
2403         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2404
2405         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2406 }
2407
2408 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2409 {
2410         struct inet6_dev *idev;
2411
2412         ASSERT_RTNL();
2413
2414         idev = ipv6_find_idev(dev);
2415         if (!idev)
2416                 return ERR_PTR(-ENOBUFS);
2417
2418         if (idev->cnf.disable_ipv6)
2419                 return ERR_PTR(-EACCES);
2420
2421         /* Add default multicast route */
2422         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2423                 addrconf_add_mroute(dev);
2424
2425         return idev;
2426 }
2427
2428 static void manage_tempaddrs(struct inet6_dev *idev,
2429                              struct inet6_ifaddr *ifp,
2430                              __u32 valid_lft, __u32 prefered_lft,
2431                              bool create, unsigned long now)
2432 {
2433         u32 flags;
2434         struct inet6_ifaddr *ift;
2435
2436         read_lock_bh(&idev->lock);
2437         /* update all temporary addresses in the list */
2438         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2439                 int age, max_valid, max_prefered;
2440
2441                 if (ifp != ift->ifpub)
2442                         continue;
2443
2444                 /* RFC 4941 section 3.3:
2445                  * If a received option will extend the lifetime of a public
2446                  * address, the lifetimes of temporary addresses should
2447                  * be extended, subject to the overall constraint that no
2448                  * temporary addresses should ever remain "valid" or "preferred"
2449                  * for a time longer than (TEMP_VALID_LIFETIME) or
2450                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2451                  */
2452                 age = (now - ift->cstamp) / HZ;
2453                 max_valid = idev->cnf.temp_valid_lft - age;
2454                 if (max_valid < 0)
2455                         max_valid = 0;
2456
2457                 max_prefered = idev->cnf.temp_prefered_lft -
2458                                idev->desync_factor - age;
2459                 if (max_prefered < 0)
2460                         max_prefered = 0;
2461
2462                 if (valid_lft > max_valid)
2463                         valid_lft = max_valid;
2464
2465                 if (prefered_lft > max_prefered)
2466                         prefered_lft = max_prefered;
2467
2468                 spin_lock(&ift->lock);
2469                 flags = ift->flags;
2470                 ift->valid_lft = valid_lft;
2471                 ift->prefered_lft = prefered_lft;
2472                 ift->tstamp = now;
2473                 if (prefered_lft > 0)
2474                         ift->flags &= ~IFA_F_DEPRECATED;
2475
2476                 spin_unlock(&ift->lock);
2477                 if (!(flags&IFA_F_TENTATIVE))
2478                         ipv6_ifa_notify(0, ift);
2479         }
2480
2481         if ((create || list_empty(&idev->tempaddr_list)) &&
2482             idev->cnf.use_tempaddr > 0) {
2483                 /* When a new public address is created as described
2484                  * in [ADDRCONF], also create a new temporary address.
2485                  * Also create a temporary address if it's enabled but
2486                  * no temporary address currently exists.
2487                  */
2488                 read_unlock_bh(&idev->lock);
2489                 ipv6_create_tempaddr(ifp, NULL, false);
2490         } else {
2491                 read_unlock_bh(&idev->lock);
2492         }
2493 }
2494
2495 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2496 {
2497         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2498                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2499 }
2500
2501 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2502                                  const struct prefix_info *pinfo,
2503                                  struct inet6_dev *in6_dev,
2504                                  const struct in6_addr *addr, int addr_type,
2505                                  u32 addr_flags, bool sllao, bool tokenized,
2506                                  __u32 valid_lft, u32 prefered_lft)
2507 {
2508         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2509         int create = 0, update_lft = 0;
2510
2511         if (!ifp && valid_lft) {
2512                 int max_addresses = in6_dev->cnf.max_addresses;
2513                 struct ifa6_config cfg = {
2514                         .pfx = addr,
2515                         .plen = pinfo->prefix_len,
2516                         .ifa_flags = addr_flags,
2517                         .valid_lft = valid_lft,
2518                         .preferred_lft = prefered_lft,
2519                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2520                 };
2521
2522 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2523                 if ((net->ipv6.devconf_all->optimistic_dad ||
2524                      in6_dev->cnf.optimistic_dad) &&
2525                     !net->ipv6.devconf_all->forwarding && sllao)
2526                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2527 #endif
2528
2529                 /* Do not allow to create too much of autoconfigured
2530                  * addresses; this would be too easy way to crash kernel.
2531                  */
2532                 if (!max_addresses ||
2533                     ipv6_count_addresses(in6_dev) < max_addresses)
2534                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2535
2536                 if (IS_ERR_OR_NULL(ifp))
2537                         return -1;
2538
2539                 create = 1;
2540                 spin_lock_bh(&ifp->lock);
2541                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2542                 ifp->cstamp = jiffies;
2543                 ifp->tokenized = tokenized;
2544                 spin_unlock_bh(&ifp->lock);
2545                 addrconf_dad_start(ifp);
2546         }
2547
2548         if (ifp) {
2549                 u32 flags;
2550                 unsigned long now;
2551                 u32 stored_lft;
2552
2553                 /* update lifetime (RFC2462 5.5.3 e) */
2554                 spin_lock_bh(&ifp->lock);
2555                 now = jiffies;
2556                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2557                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2558                 else
2559                         stored_lft = 0;
2560                 if (!create && stored_lft) {
2561                         const u32 minimum_lft = min_t(u32,
2562                                 stored_lft, MIN_VALID_LIFETIME);
2563                         valid_lft = max(valid_lft, minimum_lft);
2564
2565                         /* RFC4862 Section 5.5.3e:
2566                          * "Note that the preferred lifetime of the
2567                          *  corresponding address is always reset to
2568                          *  the Preferred Lifetime in the received
2569                          *  Prefix Information option, regardless of
2570                          *  whether the valid lifetime is also reset or
2571                          *  ignored."
2572                          *
2573                          * So we should always update prefered_lft here.
2574                          */
2575                         update_lft = 1;
2576                 }
2577
2578                 if (update_lft) {
2579                         ifp->valid_lft = valid_lft;
2580                         ifp->prefered_lft = prefered_lft;
2581                         ifp->tstamp = now;
2582                         flags = ifp->flags;
2583                         ifp->flags &= ~IFA_F_DEPRECATED;
2584                         spin_unlock_bh(&ifp->lock);
2585
2586                         if (!(flags&IFA_F_TENTATIVE))
2587                                 ipv6_ifa_notify(0, ifp);
2588                 } else
2589                         spin_unlock_bh(&ifp->lock);
2590
2591                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2592                                  create, now);
2593
2594                 in6_ifa_put(ifp);
2595                 addrconf_verify();
2596         }
2597
2598         return 0;
2599 }
2600 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2601
2602 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2603 {
2604         struct prefix_info *pinfo;
2605         __u32 valid_lft;
2606         __u32 prefered_lft;
2607         int addr_type, err;
2608         u32 addr_flags = 0;
2609         struct inet6_dev *in6_dev;
2610         struct net *net = dev_net(dev);
2611
2612         pinfo = (struct prefix_info *) opt;
2613
2614         if (len < sizeof(struct prefix_info)) {
2615                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2616                 return;
2617         }
2618
2619         /*
2620          *      Validation checks ([ADDRCONF], page 19)
2621          */
2622
2623         addr_type = ipv6_addr_type(&pinfo->prefix);
2624
2625         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2626                 return;
2627
2628         valid_lft = ntohl(pinfo->valid);
2629         prefered_lft = ntohl(pinfo->prefered);
2630
2631         if (prefered_lft > valid_lft) {
2632                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2633                 return;
2634         }
2635
2636         in6_dev = in6_dev_get(dev);
2637
2638         if (!in6_dev) {
2639                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2640                                     dev->name);
2641                 return;
2642         }
2643
2644         /*
2645          *      Two things going on here:
2646          *      1) Add routes for on-link prefixes
2647          *      2) Configure prefixes with the auto flag set
2648          */
2649
2650         if (pinfo->onlink) {
2651                 struct fib6_info *rt;
2652                 unsigned long rt_expires;
2653
2654                 /* Avoid arithmetic overflow. Really, we could
2655                  * save rt_expires in seconds, likely valid_lft,
2656                  * but it would require division in fib gc, that it
2657                  * not good.
2658                  */
2659                 if (HZ > USER_HZ)
2660                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2661                 else
2662                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2663
2664                 if (addrconf_finite_timeout(rt_expires))
2665                         rt_expires *= HZ;
2666
2667                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2668                                                pinfo->prefix_len,
2669                                                dev,
2670                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2671                                                RTF_GATEWAY | RTF_DEFAULT);
2672
2673                 if (rt) {
2674                         /* Autoconf prefix route */
2675                         if (valid_lft == 0) {
2676                                 ip6_del_rt(net, rt);
2677                                 rt = NULL;
2678                         } else if (addrconf_finite_timeout(rt_expires)) {
2679                                 /* not infinity */
2680                                 fib6_set_expires(rt, jiffies + rt_expires);
2681                         } else {
2682                                 fib6_clean_expires(rt);
2683                         }
2684                 } else if (valid_lft) {
2685                         clock_t expires = 0;
2686                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2687                         if (addrconf_finite_timeout(rt_expires)) {
2688                                 /* not infinity */
2689                                 flags |= RTF_EXPIRES;
2690                                 expires = jiffies_to_clock_t(rt_expires);
2691                         }
2692                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2693                                               0, dev, expires, flags,
2694                                               GFP_ATOMIC);
2695                 }
2696                 fib6_info_release(rt);
2697         }
2698
2699         /* Try to figure out our local address for this prefix */
2700
2701         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2702                 struct in6_addr addr;
2703                 bool tokenized = false, dev_addr_generated = false;
2704
2705                 if (pinfo->prefix_len == 64) {
2706                         memcpy(&addr, &pinfo->prefix, 8);
2707
2708                         if (!ipv6_addr_any(&in6_dev->token)) {
2709                                 read_lock_bh(&in6_dev->lock);
2710                                 memcpy(addr.s6_addr + 8,
2711                                        in6_dev->token.s6_addr + 8, 8);
2712                                 read_unlock_bh(&in6_dev->lock);
2713                                 tokenized = true;
2714                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2715                                    !ipv6_generate_stable_address(&addr, 0,
2716                                                                  in6_dev)) {
2717                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2718                                 goto ok;
2719                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2720                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2721                                 goto put;
2722                         } else {
2723                                 dev_addr_generated = true;
2724                         }
2725                         goto ok;
2726                 }
2727                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2728                                     pinfo->prefix_len);
2729                 goto put;
2730
2731 ok:
2732                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2733                                                    &addr, addr_type,
2734                                                    addr_flags, sllao,
2735                                                    tokenized, valid_lft,
2736                                                    prefered_lft);
2737                 if (err)
2738                         goto put;
2739
2740                 /* Ignore error case here because previous prefix add addr was
2741                  * successful which will be notified.
2742                  */
2743                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2744                                               addr_type, addr_flags, sllao,
2745                                               tokenized, valid_lft,
2746                                               prefered_lft,
2747                                               dev_addr_generated);
2748         }
2749         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2750 put:
2751         in6_dev_put(in6_dev);
2752 }
2753
2754 /*
2755  *      Set destination address.
2756  *      Special case for SIT interfaces where we create a new "virtual"
2757  *      device.
2758  */
2759 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2760 {
2761         struct in6_ifreq ireq;
2762         struct net_device *dev;
2763         int err = -EINVAL;
2764
2765         rtnl_lock();
2766
2767         err = -EFAULT;
2768         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2769                 goto err_exit;
2770
2771         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2772
2773         err = -ENODEV;
2774         if (!dev)
2775                 goto err_exit;
2776
2777 #if IS_ENABLED(CONFIG_IPV6_SIT)
2778         if (dev->type == ARPHRD_SIT) {
2779                 const struct net_device_ops *ops = dev->netdev_ops;
2780                 struct ifreq ifr;
2781                 struct ip_tunnel_parm p;
2782
2783                 err = -EADDRNOTAVAIL;
2784                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2785                         goto err_exit;
2786
2787                 memset(&p, 0, sizeof(p));
2788                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2789                 p.iph.saddr = 0;
2790                 p.iph.version = 4;
2791                 p.iph.ihl = 5;
2792                 p.iph.protocol = IPPROTO_IPV6;
2793                 p.iph.ttl = 64;
2794                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2795
2796                 if (ops->ndo_do_ioctl) {
2797                         mm_segment_t oldfs = get_fs();
2798
2799                         set_fs(KERNEL_DS);
2800                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2801                         set_fs(oldfs);
2802                 } else
2803                         err = -EOPNOTSUPP;
2804
2805                 if (err == 0) {
2806                         err = -ENOBUFS;
2807                         dev = __dev_get_by_name(net, p.name);
2808                         if (!dev)
2809                                 goto err_exit;
2810                         err = dev_open(dev);
2811                 }
2812         }
2813 #endif
2814
2815 err_exit:
2816         rtnl_unlock();
2817         return err;
2818 }
2819
2820 static int ipv6_mc_config(struct sock *sk, bool join,
2821                           const struct in6_addr *addr, int ifindex)
2822 {
2823         int ret;
2824
2825         ASSERT_RTNL();
2826
2827         lock_sock(sk);
2828         if (join)
2829                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2830         else
2831                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2832         release_sock(sk);
2833
2834         return ret;
2835 }
2836
2837 /*
2838  *      Manual configuration of address on an interface
2839  */
2840 static int inet6_addr_add(struct net *net, int ifindex,
2841                           struct ifa6_config *cfg,
2842                           struct netlink_ext_ack *extack)
2843 {
2844         struct inet6_ifaddr *ifp;
2845         struct inet6_dev *idev;
2846         struct net_device *dev;
2847         unsigned long timeout;
2848         clock_t expires;
2849         u32 flags;
2850
2851         ASSERT_RTNL();
2852
2853         if (cfg->plen > 128)
2854                 return -EINVAL;
2855
2856         /* check the lifetime */
2857         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2858                 return -EINVAL;
2859
2860         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2861                 return -EINVAL;
2862
2863         dev = __dev_get_by_index(net, ifindex);
2864         if (!dev)
2865                 return -ENODEV;
2866
2867         idev = addrconf_add_dev(dev);
2868         if (IS_ERR(idev))
2869                 return PTR_ERR(idev);
2870
2871         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2872                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2873                                          true, cfg->pfx, ifindex);
2874
2875                 if (ret < 0)
2876                         return ret;
2877         }
2878
2879         cfg->scope = ipv6_addr_scope(cfg->pfx);
2880
2881         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2882         if (addrconf_finite_timeout(timeout)) {
2883                 expires = jiffies_to_clock_t(timeout * HZ);
2884                 cfg->valid_lft = timeout;
2885                 flags = RTF_EXPIRES;
2886         } else {
2887                 expires = 0;
2888                 flags = 0;
2889                 cfg->ifa_flags |= IFA_F_PERMANENT;
2890         }
2891
2892         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2893         if (addrconf_finite_timeout(timeout)) {
2894                 if (timeout == 0)
2895                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2896                 cfg->preferred_lft = timeout;
2897         }
2898
2899         ifp = ipv6_add_addr(idev, cfg, true, extack);
2900         if (!IS_ERR(ifp)) {
2901                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2902                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2903                                               ifp->rt_priority, dev, expires,
2904                                               flags, GFP_KERNEL);
2905                 }
2906
2907                 /* Send a netlink notification if DAD is enabled and
2908                  * optimistic flag is not set
2909                  */
2910                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2911                         ipv6_ifa_notify(0, ifp);
2912                 /*
2913                  * Note that section 3.1 of RFC 4429 indicates
2914                  * that the Optimistic flag should not be set for
2915                  * manually configured addresses
2916                  */
2917                 addrconf_dad_start(ifp);
2918                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2919                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2920                                          cfg->preferred_lft, true, jiffies);
2921                 in6_ifa_put(ifp);
2922                 addrconf_verify_rtnl();
2923                 return 0;
2924         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2925                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2926                                cfg->pfx, ifindex);
2927         }
2928
2929         return PTR_ERR(ifp);
2930 }
2931
2932 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2933                           const struct in6_addr *pfx, unsigned int plen)
2934 {
2935         struct inet6_ifaddr *ifp;
2936         struct inet6_dev *idev;
2937         struct net_device *dev;
2938
2939         if (plen > 128)
2940                 return -EINVAL;
2941
2942         dev = __dev_get_by_index(net, ifindex);
2943         if (!dev)
2944                 return -ENODEV;
2945
2946         idev = __in6_dev_get(dev);
2947         if (!idev)
2948                 return -ENXIO;
2949
2950         read_lock_bh(&idev->lock);
2951         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2952                 if (ifp->prefix_len == plen &&
2953                     ipv6_addr_equal(pfx, &ifp->addr)) {
2954                         in6_ifa_hold(ifp);
2955                         read_unlock_bh(&idev->lock);
2956
2957                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2958                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2959                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2960                                                  jiffies);
2961                         ipv6_del_addr(ifp);
2962                         addrconf_verify_rtnl();
2963                         if (ipv6_addr_is_multicast(pfx)) {
2964                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2965                                                false, pfx, dev->ifindex);
2966                         }
2967                         return 0;
2968                 }
2969         }
2970         read_unlock_bh(&idev->lock);
2971         return -EADDRNOTAVAIL;
2972 }
2973
2974
2975 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2976 {
2977         struct ifa6_config cfg = {
2978                 .ifa_flags = IFA_F_PERMANENT,
2979                 .preferred_lft = INFINITY_LIFE_TIME,
2980                 .valid_lft = INFINITY_LIFE_TIME,
2981         };
2982         struct in6_ifreq ireq;
2983         int err;
2984
2985         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2986                 return -EPERM;
2987
2988         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2989                 return -EFAULT;
2990
2991         cfg.pfx = &ireq.ifr6_addr;
2992         cfg.plen = ireq.ifr6_prefixlen;
2993
2994         rtnl_lock();
2995         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
2996         rtnl_unlock();
2997         return err;
2998 }
2999
3000 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3001 {
3002         struct in6_ifreq ireq;
3003         int err;
3004
3005         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3006                 return -EPERM;
3007
3008         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3009                 return -EFAULT;
3010
3011         rtnl_lock();
3012         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3013                              ireq.ifr6_prefixlen);
3014         rtnl_unlock();
3015         return err;
3016 }
3017
3018 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3019                      int plen, int scope)
3020 {
3021         struct inet6_ifaddr *ifp;
3022         struct ifa6_config cfg = {
3023                 .pfx = addr,
3024                 .plen = plen,
3025                 .ifa_flags = IFA_F_PERMANENT,
3026                 .valid_lft = INFINITY_LIFE_TIME,
3027                 .preferred_lft = INFINITY_LIFE_TIME,
3028                 .scope = scope
3029         };
3030
3031         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3032         if (!IS_ERR(ifp)) {
3033                 spin_lock_bh(&ifp->lock);
3034                 ifp->flags &= ~IFA_F_TENTATIVE;
3035                 spin_unlock_bh(&ifp->lock);
3036                 rt_genid_bump_ipv6(dev_net(idev->dev));
3037                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3038                 in6_ifa_put(ifp);
3039         }
3040 }
3041
3042 #if IS_ENABLED(CONFIG_IPV6_SIT)
3043 static void sit_add_v4_addrs(struct inet6_dev *idev)
3044 {
3045         struct in6_addr addr;
3046         struct net_device *dev;
3047         struct net *net = dev_net(idev->dev);
3048         int scope, plen;
3049         u32 pflags = 0;
3050
3051         ASSERT_RTNL();
3052
3053         memset(&addr, 0, sizeof(struct in6_addr));
3054         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3055
3056         if (idev->dev->flags&IFF_POINTOPOINT) {
3057                 addr.s6_addr32[0] = htonl(0xfe800000);
3058                 scope = IFA_LINK;
3059                 plen = 64;
3060         } else {
3061                 scope = IPV6_ADDR_COMPATv4;
3062                 plen = 96;
3063                 pflags |= RTF_NONEXTHOP;
3064         }
3065
3066         if (addr.s6_addr32[3]) {
3067                 add_addr(idev, &addr, plen, scope);
3068                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3069                                       GFP_KERNEL);
3070                 return;
3071         }
3072
3073         for_each_netdev(net, dev) {
3074                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3075                 if (in_dev && (dev->flags & IFF_UP)) {
3076                         struct in_ifaddr *ifa;
3077
3078                         int flag = scope;
3079
3080                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3081
3082                                 addr.s6_addr32[3] = ifa->ifa_local;
3083
3084                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3085                                         continue;
3086                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3087                                         if (idev->dev->flags&IFF_POINTOPOINT)
3088                                                 continue;
3089                                         flag |= IFA_HOST;
3090                                 }
3091
3092                                 add_addr(idev, &addr, plen, flag);
3093                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3094                                                       0, pflags, GFP_KERNEL);
3095                         }
3096                 }
3097         }
3098 }
3099 #endif
3100
3101 static void init_loopback(struct net_device *dev)
3102 {
3103         struct inet6_dev  *idev;
3104
3105         /* ::1 */
3106
3107         ASSERT_RTNL();
3108
3109         idev = ipv6_find_idev(dev);
3110         if (!idev) {
3111                 pr_debug("%s: add_dev failed\n", __func__);
3112                 return;
3113         }
3114
3115         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3116 }
3117
3118 void addrconf_add_linklocal(struct inet6_dev *idev,
3119                             const struct in6_addr *addr, u32 flags)
3120 {
3121         struct ifa6_config cfg = {
3122                 .pfx = addr,
3123                 .plen = 64,
3124                 .ifa_flags = flags | IFA_F_PERMANENT,
3125                 .valid_lft = INFINITY_LIFE_TIME,
3126                 .preferred_lft = INFINITY_LIFE_TIME,
3127                 .scope = IFA_LINK
3128         };
3129         struct inet6_ifaddr *ifp;
3130
3131 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3132         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3133              idev->cnf.optimistic_dad) &&
3134             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3135                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3136 #endif
3137
3138         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3139         if (!IS_ERR(ifp)) {
3140                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3141                                       0, 0, GFP_ATOMIC);
3142                 addrconf_dad_start(ifp);
3143                 in6_ifa_put(ifp);
3144         }
3145 }
3146 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3147
3148 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3149 {
3150         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3151                 return true;
3152
3153         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3154             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3155                 return true;
3156
3157         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3158             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3159                 return true;
3160
3161         return false;
3162 }
3163
3164 static int ipv6_generate_stable_address(struct in6_addr *address,
3165                                         u8 dad_count,
3166                                         const struct inet6_dev *idev)
3167 {
3168         static DEFINE_SPINLOCK(lock);
3169         static __u32 digest[SHA_DIGEST_WORDS];
3170         static __u32 workspace[SHA_WORKSPACE_WORDS];
3171
3172         static union {
3173                 char __data[SHA_MESSAGE_BYTES];
3174                 struct {
3175                         struct in6_addr secret;
3176                         __be32 prefix[2];
3177                         unsigned char hwaddr[MAX_ADDR_LEN];
3178                         u8 dad_count;
3179                 } __packed;
3180         } data;
3181
3182         struct in6_addr secret;
3183         struct in6_addr temp;
3184         struct net *net = dev_net(idev->dev);
3185
3186         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3187
3188         if (idev->cnf.stable_secret.initialized)
3189                 secret = idev->cnf.stable_secret.secret;
3190         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3191                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3192         else
3193                 return -1;
3194
3195 retry:
3196         spin_lock_bh(&lock);
3197
3198         sha_init(digest);
3199         memset(&data, 0, sizeof(data));
3200         memset(workspace, 0, sizeof(workspace));
3201         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3202         data.prefix[0] = address->s6_addr32[0];
3203         data.prefix[1] = address->s6_addr32[1];
3204         data.secret = secret;
3205         data.dad_count = dad_count;
3206
3207         sha_transform(digest, data.__data, workspace);
3208
3209         temp = *address;
3210         temp.s6_addr32[2] = (__force __be32)digest[0];
3211         temp.s6_addr32[3] = (__force __be32)digest[1];
3212
3213         spin_unlock_bh(&lock);
3214
3215         if (ipv6_reserved_interfaceid(temp)) {
3216                 dad_count++;
3217                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3218                         return -1;
3219                 goto retry;
3220         }
3221
3222         *address = temp;
3223         return 0;
3224 }
3225
3226 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3227 {
3228         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3229
3230         if (s->initialized)
3231                 return;
3232         s = &idev->cnf.stable_secret;
3233         get_random_bytes(&s->secret, sizeof(s->secret));
3234         s->initialized = true;
3235 }
3236
3237 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3238 {
3239         struct in6_addr addr;
3240
3241         /* no link local addresses on L3 master devices */
3242         if (netif_is_l3_master(idev->dev))
3243                 return;
3244
3245         /* no link local addresses on devices flagged as slaves */
3246         if (idev->dev->flags & IFF_SLAVE)
3247                 return;
3248
3249         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3250
3251         switch (idev->cnf.addr_gen_mode) {
3252         case IN6_ADDR_GEN_MODE_RANDOM:
3253                 ipv6_gen_mode_random_init(idev);
3254                 /* fallthrough */
3255         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3256                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3257                         addrconf_add_linklocal(idev, &addr,
3258                                                IFA_F_STABLE_PRIVACY);
3259                 else if (prefix_route)
3260                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3261                                               0, 0, GFP_KERNEL);
3262                 break;
3263         case IN6_ADDR_GEN_MODE_EUI64:
3264                 /* addrconf_add_linklocal also adds a prefix_route and we
3265                  * only need to care about prefix routes if ipv6_generate_eui64
3266                  * couldn't generate one.
3267                  */
3268                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3269                         addrconf_add_linklocal(idev, &addr, 0);
3270                 else if (prefix_route)
3271                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3272                                               0, 0, GFP_KERNEL);
3273                 break;
3274         case IN6_ADDR_GEN_MODE_NONE:
3275         default:
3276                 /* will not add any link local address */
3277                 break;
3278         }
3279 }
3280
3281 static void addrconf_dev_config(struct net_device *dev)
3282 {
3283         struct inet6_dev *idev;
3284
3285         ASSERT_RTNL();
3286
3287         if ((dev->type != ARPHRD_ETHER) &&
3288             (dev->type != ARPHRD_FDDI) &&
3289             (dev->type != ARPHRD_ARCNET) &&
3290             (dev->type != ARPHRD_INFINIBAND) &&
3291             (dev->type != ARPHRD_IEEE1394) &&
3292             (dev->type != ARPHRD_TUNNEL6) &&
3293             (dev->type != ARPHRD_6LOWPAN) &&
3294             (dev->type != ARPHRD_IP6GRE) &&
3295             (dev->type != ARPHRD_IPGRE) &&
3296             (dev->type != ARPHRD_TUNNEL) &&
3297             (dev->type != ARPHRD_NONE) &&
3298             (dev->type != ARPHRD_RAWIP)) {
3299                 /* Alas, we support only Ethernet autoconfiguration. */
3300                 idev = __in6_dev_get(dev);
3301                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3302                     dev->flags & IFF_MULTICAST)
3303                         ipv6_mc_up(idev);
3304                 return;
3305         }
3306
3307         idev = addrconf_add_dev(dev);
3308         if (IS_ERR(idev))
3309                 return;
3310
3311         /* this device type has no EUI support */
3312         if (dev->type == ARPHRD_NONE &&
3313             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3314                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3315
3316         addrconf_addr_gen(idev, false);
3317 }
3318
3319 #if IS_ENABLED(CONFIG_IPV6_SIT)
3320 static void addrconf_sit_config(struct net_device *dev)
3321 {
3322         struct inet6_dev *idev;
3323
3324         ASSERT_RTNL();
3325
3326         /*
3327          * Configure the tunnel with one of our IPv4
3328          * addresses... we should configure all of
3329          * our v4 addrs in the tunnel
3330          */
3331
3332         idev = ipv6_find_idev(dev);
3333         if (!idev) {
3334                 pr_debug("%s: add_dev failed\n", __func__);
3335                 return;
3336         }
3337
3338         if (dev->priv_flags & IFF_ISATAP) {
3339                 addrconf_addr_gen(idev, false);
3340                 return;
3341         }
3342
3343         sit_add_v4_addrs(idev);
3344
3345         if (dev->flags&IFF_POINTOPOINT)
3346                 addrconf_add_mroute(dev);
3347 }
3348 #endif
3349
3350 #if IS_ENABLED(CONFIG_NET_IPGRE)
3351 static void addrconf_gre_config(struct net_device *dev)
3352 {
3353         struct inet6_dev *idev;
3354
3355         ASSERT_RTNL();
3356
3357         idev = ipv6_find_idev(dev);
3358         if (!idev) {
3359                 pr_debug("%s: add_dev failed\n", __func__);
3360                 return;
3361         }
3362
3363         addrconf_addr_gen(idev, true);
3364         if (dev->flags & IFF_POINTOPOINT)
3365                 addrconf_add_mroute(dev);
3366 }
3367 #endif
3368
3369 static int fixup_permanent_addr(struct net *net,
3370                                 struct inet6_dev *idev,
3371                                 struct inet6_ifaddr *ifp)
3372 {
3373         /* !fib6_node means the host route was removed from the
3374          * FIB, for example, if 'lo' device is taken down. In that
3375          * case regenerate the host route.
3376          */
3377         if (!ifp->rt || !ifp->rt->fib6_node) {
3378                 struct fib6_info *f6i, *prev;
3379
3380                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3381                                          GFP_ATOMIC);
3382                 if (IS_ERR(f6i))
3383                         return PTR_ERR(f6i);
3384
3385                 /* ifp->rt can be accessed outside of rtnl */
3386                 spin_lock(&ifp->lock);
3387                 prev = ifp->rt;
3388                 ifp->rt = f6i;
3389                 spin_unlock(&ifp->lock);
3390
3391                 fib6_info_release(prev);
3392         }
3393
3394         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3395                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3396                                       ifp->rt_priority, idev->dev, 0, 0,
3397                                       GFP_ATOMIC);
3398         }
3399
3400         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3401                 addrconf_dad_start(ifp);
3402
3403         return 0;
3404 }
3405
3406 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3407 {
3408         struct inet6_ifaddr *ifp, *tmp;
3409         struct inet6_dev *idev;
3410
3411         idev = __in6_dev_get(dev);
3412         if (!idev)
3413                 return;
3414
3415         write_lock_bh(&idev->lock);
3416
3417         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3418                 if ((ifp->flags & IFA_F_PERMANENT) &&
3419                     fixup_permanent_addr(net, idev, ifp) < 0) {
3420                         write_unlock_bh(&idev->lock);
3421                         in6_ifa_hold(ifp);
3422                         ipv6_del_addr(ifp);
3423                         write_lock_bh(&idev->lock);
3424
3425                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3426                                              idev->dev->name, &ifp->addr);
3427                 }
3428         }
3429
3430         write_unlock_bh(&idev->lock);
3431 }
3432
3433 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3434                            void *ptr)
3435 {
3436         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3437         struct netdev_notifier_change_info *change_info;
3438         struct netdev_notifier_changeupper_info *info;
3439         struct inet6_dev *idev = __in6_dev_get(dev);
3440         struct net *net = dev_net(dev);
3441         int run_pending = 0;
3442         int err;
3443
3444         switch (event) {
3445         case NETDEV_REGISTER:
3446                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3447                         idev = ipv6_add_dev(dev);
3448                         if (IS_ERR(idev))
3449                                 return notifier_from_errno(PTR_ERR(idev));
3450                 }
3451                 break;
3452
3453         case NETDEV_CHANGEMTU:
3454                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3455                 if (dev->mtu < IPV6_MIN_MTU) {
3456                         addrconf_ifdown(dev, dev != net->loopback_dev);
3457                         break;
3458                 }
3459
3460                 if (idev) {
3461                         rt6_mtu_change(dev, dev->mtu);
3462                         idev->cnf.mtu6 = dev->mtu;
3463                         break;
3464                 }
3465
3466                 /* allocate new idev */
3467                 idev = ipv6_add_dev(dev);
3468                 if (IS_ERR(idev))
3469                         break;
3470
3471                 /* device is still not ready */
3472                 if (!(idev->if_flags & IF_READY))
3473                         break;
3474
3475                 run_pending = 1;
3476
3477                 /* fall through */
3478
3479         case NETDEV_UP:
3480         case NETDEV_CHANGE:
3481                 if (dev->flags & IFF_SLAVE)
3482                         break;
3483
3484                 if (idev && idev->cnf.disable_ipv6)
3485                         break;
3486
3487                 if (event == NETDEV_UP) {
3488                         /* restore routes for permanent addresses */
3489                         addrconf_permanent_addr(net, dev);
3490
3491                         if (!addrconf_link_ready(dev)) {
3492                                 /* device is not ready yet. */
3493                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3494                                         dev->name);
3495                                 break;
3496                         }
3497
3498                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3499                                 idev = ipv6_add_dev(dev);
3500
3501                         if (!IS_ERR_OR_NULL(idev)) {
3502                                 idev->if_flags |= IF_READY;
3503                                 run_pending = 1;
3504                         }
3505                 } else if (event == NETDEV_CHANGE) {
3506                         if (!addrconf_link_ready(dev)) {
3507                                 /* device is still not ready. */
3508                                 rt6_sync_down_dev(dev, event);
3509                                 break;
3510                         }
3511
3512                         if (!IS_ERR_OR_NULL(idev)) {
3513                                 if (idev->if_flags & IF_READY) {
3514                                         /* device is already configured -
3515                                          * but resend MLD reports, we might
3516                                          * have roamed and need to update
3517                                          * multicast snooping switches
3518                                          */
3519                                         ipv6_mc_up(idev);
3520                                         change_info = ptr;
3521                                         if (change_info->flags_changed & IFF_NOARP)
3522                                                 addrconf_dad_run(idev, true);
3523                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3524                                         break;
3525                                 }
3526                                 idev->if_flags |= IF_READY;
3527                         }
3528
3529                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3530                                 dev->name);
3531
3532                         run_pending = 1;
3533                 }
3534
3535                 switch (dev->type) {
3536 #if IS_ENABLED(CONFIG_IPV6_SIT)
3537                 case ARPHRD_SIT:
3538                         addrconf_sit_config(dev);
3539                         break;
3540 #endif
3541 #if IS_ENABLED(CONFIG_NET_IPGRE)
3542                 case ARPHRD_IPGRE:
3543                         addrconf_gre_config(dev);
3544                         break;
3545 #endif
3546                 case ARPHRD_LOOPBACK:
3547                         init_loopback(dev);
3548                         break;
3549
3550                 default:
3551                         addrconf_dev_config(dev);
3552                         break;
3553                 }
3554
3555                 if (!IS_ERR_OR_NULL(idev)) {
3556                         if (run_pending)
3557                                 addrconf_dad_run(idev, false);
3558
3559                         /* Device has an address by now */
3560                         rt6_sync_up(dev, RTNH_F_DEAD);
3561
3562                         /*
3563                          * If the MTU changed during the interface down,
3564                          * when the interface up, the changed MTU must be
3565                          * reflected in the idev as well as routers.
3566                          */
3567                         if (idev->cnf.mtu6 != dev->mtu &&
3568                             dev->mtu >= IPV6_MIN_MTU) {
3569                                 rt6_mtu_change(dev, dev->mtu);
3570                                 idev->cnf.mtu6 = dev->mtu;
3571                         }
3572                         idev->tstamp = jiffies;
3573                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3574
3575                         /*
3576                          * If the changed mtu during down is lower than
3577                          * IPV6_MIN_MTU stop IPv6 on this interface.
3578                          */
3579                         if (dev->mtu < IPV6_MIN_MTU)
3580                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3581                 }
3582                 break;
3583
3584         case NETDEV_DOWN:
3585         case NETDEV_UNREGISTER:
3586                 /*
3587                  *      Remove all addresses from this interface.
3588                  */
3589                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3590                 break;
3591
3592         case NETDEV_CHANGENAME:
3593                 if (idev) {
3594                         snmp6_unregister_dev(idev);
3595                         addrconf_sysctl_unregister(idev);
3596                         err = addrconf_sysctl_register(idev);
3597                         if (err)
3598                                 return notifier_from_errno(err);
3599                         err = snmp6_register_dev(idev);
3600                         if (err) {
3601                                 addrconf_sysctl_unregister(idev);
3602                                 return notifier_from_errno(err);
3603                         }
3604                 }
3605                 break;
3606
3607         case NETDEV_PRE_TYPE_CHANGE:
3608         case NETDEV_POST_TYPE_CHANGE:
3609                 if (idev)
3610                         addrconf_type_change(dev, event);
3611                 break;
3612
3613         case NETDEV_CHANGEUPPER:
3614                 info = ptr;
3615
3616                 /* flush all routes if dev is linked to or unlinked from
3617                  * an L3 master device (e.g., VRF)
3618                  */
3619                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3620                         addrconf_ifdown(dev, 0);
3621         }
3622
3623         return NOTIFY_OK;
3624 }
3625
3626 /*
3627  *      addrconf module should be notified of a device going up
3628  */
3629 static struct notifier_block ipv6_dev_notf = {
3630         .notifier_call = addrconf_notify,
3631         .priority = ADDRCONF_NOTIFY_PRIORITY,
3632 };
3633
3634 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3635 {
3636         struct inet6_dev *idev;
3637         ASSERT_RTNL();
3638
3639         idev = __in6_dev_get(dev);
3640
3641         if (event == NETDEV_POST_TYPE_CHANGE)
3642                 ipv6_mc_remap(idev);
3643         else if (event == NETDEV_PRE_TYPE_CHANGE)
3644                 ipv6_mc_unmap(idev);
3645 }
3646
3647 static bool addr_is_local(const struct in6_addr *addr)
3648 {
3649         return ipv6_addr_type(addr) &
3650                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3651 }
3652
3653 static int addrconf_ifdown(struct net_device *dev, int how)
3654 {
3655         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3656         struct net *net = dev_net(dev);
3657         struct inet6_dev *idev;
3658         struct inet6_ifaddr *ifa, *tmp;
3659         bool keep_addr = false;
3660         int state, i;
3661
3662         ASSERT_RTNL();
3663
3664         rt6_disable_ip(dev, event);
3665
3666         idev = __in6_dev_get(dev);
3667         if (!idev)
3668                 return -ENODEV;
3669
3670         /*
3671          * Step 1: remove reference to ipv6 device from parent device.
3672          *         Do not dev_put!
3673          */
3674         if (how) {
3675                 idev->dead = 1;
3676
3677                 /* protected by rtnl_lock */
3678                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3679
3680                 /* Step 1.5: remove snmp6 entry */
3681                 snmp6_unregister_dev(idev);
3682
3683         }
3684
3685         /* combine the user config with event to determine if permanent
3686          * addresses are to be removed from address hash table
3687          */
3688         if (!how && !idev->cnf.disable_ipv6) {
3689                 /* aggregate the system setting and interface setting */
3690                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3691
3692                 if (!_keep_addr)
3693                         _keep_addr = idev->cnf.keep_addr_on_down;
3694
3695                 keep_addr = (_keep_addr > 0);
3696         }
3697
3698         /* Step 2: clear hash table */
3699         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3700                 struct hlist_head *h = &inet6_addr_lst[i];
3701
3702                 spin_lock_bh(&addrconf_hash_lock);
3703 restart:
3704                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3705                         if (ifa->idev == idev) {
3706                                 addrconf_del_dad_work(ifa);
3707                                 /* combined flag + permanent flag decide if
3708                                  * address is retained on a down event
3709                                  */
3710                                 if (!keep_addr ||
3711                                     !(ifa->flags & IFA_F_PERMANENT) ||
3712                                     addr_is_local(&ifa->addr)) {
3713                                         hlist_del_init_rcu(&ifa->addr_lst);
3714                                         goto restart;
3715                                 }
3716                         }
3717                 }
3718                 spin_unlock_bh(&addrconf_hash_lock);
3719         }
3720
3721         write_lock_bh(&idev->lock);
3722
3723         addrconf_del_rs_timer(idev);
3724
3725         /* Step 2: clear flags for stateless addrconf */
3726         if (!how)
3727                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3728
3729         /* Step 3: clear tempaddr list */
3730         while (!list_empty(&idev->tempaddr_list)) {
3731                 ifa = list_first_entry(&idev->tempaddr_list,
3732                                        struct inet6_ifaddr, tmp_list);
3733                 list_del(&ifa->tmp_list);
3734                 write_unlock_bh(&idev->lock);
3735                 spin_lock_bh(&ifa->lock);
3736
3737                 if (ifa->ifpub) {
3738                         in6_ifa_put(ifa->ifpub);
3739                         ifa->ifpub = NULL;
3740                 }
3741                 spin_unlock_bh(&ifa->lock);
3742                 in6_ifa_put(ifa);
3743                 write_lock_bh(&idev->lock);
3744         }
3745
3746         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3747                 struct fib6_info *rt = NULL;
3748                 bool keep;
3749
3750                 addrconf_del_dad_work(ifa);
3751
3752                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3753                         !addr_is_local(&ifa->addr);
3754
3755                 write_unlock_bh(&idev->lock);
3756                 spin_lock_bh(&ifa->lock);
3757
3758                 if (keep) {
3759                         /* set state to skip the notifier below */
3760                         state = INET6_IFADDR_STATE_DEAD;
3761                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3762                         if (!(ifa->flags & IFA_F_NODAD))
3763                                 ifa->flags |= IFA_F_TENTATIVE;
3764
3765                         rt = ifa->rt;
3766                         ifa->rt = NULL;
3767                 } else {
3768                         state = ifa->state;
3769                         ifa->state = INET6_IFADDR_STATE_DEAD;
3770                 }
3771
3772                 spin_unlock_bh(&ifa->lock);
3773
3774                 if (rt)
3775                         ip6_del_rt(net, rt);
3776
3777                 if (state != INET6_IFADDR_STATE_DEAD) {
3778                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3779                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3780                 } else {
3781                         if (idev->cnf.forwarding)
3782                                 addrconf_leave_anycast(ifa);
3783                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3784                 }
3785
3786                 write_lock_bh(&idev->lock);
3787                 if (!keep) {
3788                         list_del_rcu(&ifa->if_list);
3789                         in6_ifa_put(ifa);
3790                 }
3791         }
3792
3793         write_unlock_bh(&idev->lock);
3794
3795         /* Step 5: Discard anycast and multicast list */
3796         if (how) {
3797                 ipv6_ac_destroy_dev(idev);
3798                 ipv6_mc_destroy_dev(idev);
3799         } else {
3800                 ipv6_mc_down(idev);
3801         }
3802
3803         idev->tstamp = jiffies;
3804
3805         /* Last: Shot the device (if unregistered) */
3806         if (how) {
3807                 addrconf_sysctl_unregister(idev);
3808                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3809                 neigh_ifdown(&nd_tbl, dev);
3810                 in6_dev_put(idev);
3811         }
3812         return 0;
3813 }
3814
3815 static void addrconf_rs_timer(struct timer_list *t)
3816 {
3817         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3818         struct net_device *dev = idev->dev;
3819         struct in6_addr lladdr;
3820
3821         write_lock(&idev->lock);
3822         if (idev->dead || !(idev->if_flags & IF_READY))
3823                 goto out;
3824
3825         if (!ipv6_accept_ra(idev))
3826                 goto out;
3827
3828         /* Announcement received after solicitation was sent */
3829         if (idev->if_flags & IF_RA_RCVD)
3830                 goto out;
3831
3832         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3833                 write_unlock(&idev->lock);
3834                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3835                         ndisc_send_rs(dev, &lladdr,
3836                                       &in6addr_linklocal_allrouters);
3837                 else
3838                         goto put;
3839
3840                 write_lock(&idev->lock);
3841                 idev->rs_interval = rfc3315_s14_backoff_update(
3842                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3843                 /* The wait after the last probe can be shorter */
3844                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3845                                              idev->cnf.rtr_solicits) ?
3846                                       idev->cnf.rtr_solicit_delay :
3847                                       idev->rs_interval);
3848         } else {
3849                 /*
3850                  * Note: we do not support deprecated "all on-link"
3851                  * assumption any longer.
3852                  */
3853                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3854         }
3855
3856 out:
3857         write_unlock(&idev->lock);
3858 put:
3859         in6_dev_put(idev);
3860 }
3861
3862 /*
3863  *      Duplicate Address Detection
3864  */
3865 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3866 {
3867         unsigned long rand_num;
3868         struct inet6_dev *idev = ifp->idev;
3869         u64 nonce;
3870
3871         if (ifp->flags & IFA_F_OPTIMISTIC)
3872                 rand_num = 0;
3873         else
3874                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3875
3876         nonce = 0;
3877         if (idev->cnf.enhanced_dad ||
3878             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3879                 do
3880                         get_random_bytes(&nonce, 6);
3881                 while (nonce == 0);
3882         }
3883         ifp->dad_nonce = nonce;
3884         ifp->dad_probes = idev->cnf.dad_transmits;
3885         addrconf_mod_dad_work(ifp, rand_num);
3886 }
3887
3888 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3889 {
3890         struct inet6_dev *idev = ifp->idev;
3891         struct net_device *dev = idev->dev;
3892         bool bump_id, notify = false;
3893         struct net *net;
3894
3895         addrconf_join_solict(dev, &ifp->addr);
3896
3897         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3898
3899         read_lock_bh(&idev->lock);
3900         spin_lock(&ifp->lock);
3901         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3902                 goto out;
3903
3904         net = dev_net(dev);
3905         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3906             (net->ipv6.devconf_all->accept_dad < 1 &&
3907              idev->cnf.accept_dad < 1) ||
3908             !(ifp->flags&IFA_F_TENTATIVE) ||
3909             ifp->flags & IFA_F_NODAD) {
3910                 bool send_na = false;
3911
3912                 if (ifp->flags & IFA_F_TENTATIVE &&
3913                     !(ifp->flags & IFA_F_OPTIMISTIC))
3914                         send_na = true;
3915                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3916                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3917                 spin_unlock(&ifp->lock);
3918                 read_unlock_bh(&idev->lock);
3919
3920                 addrconf_dad_completed(ifp, bump_id, send_na);
3921                 return;
3922         }
3923
3924         if (!(idev->if_flags & IF_READY)) {
3925                 spin_unlock(&ifp->lock);
3926                 read_unlock_bh(&idev->lock);
3927                 /*
3928                  * If the device is not ready:
3929                  * - keep it tentative if it is a permanent address.
3930                  * - otherwise, kill it.
3931                  */
3932                 in6_ifa_hold(ifp);
3933                 addrconf_dad_stop(ifp, 0);
3934                 return;
3935         }
3936
3937         /*
3938          * Optimistic nodes can start receiving
3939          * Frames right away
3940          */
3941         if (ifp->flags & IFA_F_OPTIMISTIC) {
3942                 ip6_ins_rt(net, ifp->rt);
3943                 if (ipv6_use_optimistic_addr(net, idev)) {
3944                         /* Because optimistic nodes can use this address,
3945                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3946                          */
3947                         notify = true;
3948                 }
3949         }
3950
3951         addrconf_dad_kick(ifp);
3952 out:
3953         spin_unlock(&ifp->lock);
3954         read_unlock_bh(&idev->lock);
3955         if (notify)
3956                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3957 }
3958
3959 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3960 {
3961         bool begin_dad = false;
3962
3963         spin_lock_bh(&ifp->lock);
3964         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3965                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3966                 begin_dad = true;
3967         }
3968         spin_unlock_bh(&ifp->lock);
3969
3970         if (begin_dad)
3971                 addrconf_mod_dad_work(ifp, 0);
3972 }
3973
3974 static void addrconf_dad_work(struct work_struct *w)
3975 {
3976         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3977                                                 struct inet6_ifaddr,
3978                                                 dad_work);
3979         struct inet6_dev *idev = ifp->idev;
3980         bool bump_id, disable_ipv6 = false;
3981         struct in6_addr mcaddr;
3982
3983         enum {
3984                 DAD_PROCESS,
3985                 DAD_BEGIN,
3986                 DAD_ABORT,
3987         } action = DAD_PROCESS;
3988
3989         rtnl_lock();
3990
3991         spin_lock_bh(&ifp->lock);
3992         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3993                 action = DAD_BEGIN;
3994                 ifp->state = INET6_IFADDR_STATE_DAD;
3995         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3996                 action = DAD_ABORT;
3997                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3998
3999                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4000                      idev->cnf.accept_dad > 1) &&
4001                     !idev->cnf.disable_ipv6 &&
4002                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4003                         struct in6_addr addr;
4004
4005                         addr.s6_addr32[0] = htonl(0xfe800000);
4006                         addr.s6_addr32[1] = 0;
4007
4008                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4009                             ipv6_addr_equal(&ifp->addr, &addr)) {
4010                                 /* DAD failed for link-local based on MAC */
4011                                 idev->cnf.disable_ipv6 = 1;
4012
4013                                 pr_info("%s: IPv6 being disabled!\n",
4014                                         ifp->idev->dev->name);
4015                                 disable_ipv6 = true;
4016                         }
4017                 }
4018         }
4019         spin_unlock_bh(&ifp->lock);
4020
4021         if (action == DAD_BEGIN) {
4022                 addrconf_dad_begin(ifp);
4023                 goto out;
4024         } else if (action == DAD_ABORT) {
4025                 in6_ifa_hold(ifp);
4026                 addrconf_dad_stop(ifp, 1);
4027                 if (disable_ipv6)
4028                         addrconf_ifdown(idev->dev, 0);
4029                 goto out;
4030         }
4031
4032         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4033                 goto out;
4034
4035         write_lock_bh(&idev->lock);
4036         if (idev->dead || !(idev->if_flags & IF_READY)) {
4037                 write_unlock_bh(&idev->lock);
4038                 goto out;
4039         }
4040
4041         spin_lock(&ifp->lock);
4042         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4043                 spin_unlock(&ifp->lock);
4044                 write_unlock_bh(&idev->lock);
4045                 goto out;
4046         }
4047
4048         if (ifp->dad_probes == 0) {
4049                 bool send_na = false;
4050
4051                 /*
4052                  * DAD was successful
4053                  */
4054
4055                 if (ifp->flags & IFA_F_TENTATIVE &&
4056                     !(ifp->flags & IFA_F_OPTIMISTIC))
4057                         send_na = true;
4058                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4059                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4060                 spin_unlock(&ifp->lock);
4061                 write_unlock_bh(&idev->lock);
4062
4063                 addrconf_dad_completed(ifp, bump_id, send_na);
4064
4065                 goto out;
4066         }
4067
4068         ifp->dad_probes--;
4069         addrconf_mod_dad_work(ifp,
4070                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4071         spin_unlock(&ifp->lock);
4072         write_unlock_bh(&idev->lock);
4073
4074         /* send a neighbour solicitation for our addr */
4075         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4076         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4077                       ifp->dad_nonce);
4078 out:
4079         in6_ifa_put(ifp);
4080         rtnl_unlock();
4081 }
4082
4083 /* ifp->idev must be at least read locked */
4084 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4085 {
4086         struct inet6_ifaddr *ifpiter;
4087         struct inet6_dev *idev = ifp->idev;
4088
4089         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4090                 if (ifpiter->scope > IFA_LINK)
4091                         break;
4092                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4093                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4094                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4095                     IFA_F_PERMANENT)
4096                         return false;
4097         }
4098         return true;
4099 }
4100
4101 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4102                                    bool send_na)
4103 {
4104         struct net_device *dev = ifp->idev->dev;
4105         struct in6_addr lladdr;
4106         bool send_rs, send_mld;
4107
4108         addrconf_del_dad_work(ifp);
4109
4110         /*
4111          *      Configure the address for reception. Now it is valid.
4112          */
4113
4114         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4115
4116         /* If added prefix is link local and we are prepared to process
4117            router advertisements, start sending router solicitations.
4118          */
4119
4120         read_lock_bh(&ifp->idev->lock);
4121         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4122         send_rs = send_mld &&
4123                   ipv6_accept_ra(ifp->idev) &&
4124                   ifp->idev->cnf.rtr_solicits != 0 &&
4125                   (dev->flags&IFF_LOOPBACK) == 0;
4126         read_unlock_bh(&ifp->idev->lock);
4127
4128         /* While dad is in progress mld report's source address is in6_addrany.
4129          * Resend with proper ll now.
4130          */
4131         if (send_mld)
4132                 ipv6_mc_dad_complete(ifp->idev);
4133
4134         /* send unsolicited NA if enabled */
4135         if (send_na &&
4136             (ifp->idev->cnf.ndisc_notify ||
4137              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4138                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4139                               /*router=*/ !!ifp->idev->cnf.forwarding,
4140                               /*solicited=*/ false, /*override=*/ true,
4141                               /*inc_opt=*/ true);
4142         }
4143
4144         if (send_rs) {
4145                 /*
4146                  *      If a host as already performed a random delay
4147                  *      [...] as part of DAD [...] there is no need
4148                  *      to delay again before sending the first RS
4149                  */
4150                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4151                         return;
4152                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4153
4154                 write_lock_bh(&ifp->idev->lock);
4155                 spin_lock(&ifp->lock);
4156                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4157                         ifp->idev->cnf.rtr_solicit_interval);
4158                 ifp->idev->rs_probes = 1;
4159                 ifp->idev->if_flags |= IF_RS_SENT;
4160                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4161                 spin_unlock(&ifp->lock);
4162                 write_unlock_bh(&ifp->idev->lock);
4163         }
4164
4165         if (bump_id)
4166                 rt_genid_bump_ipv6(dev_net(dev));
4167
4168         /* Make sure that a new temporary address will be created
4169          * before this temporary address becomes deprecated.
4170          */
4171         if (ifp->flags & IFA_F_TEMPORARY)
4172                 addrconf_verify_rtnl();
4173 }
4174
4175 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4176 {
4177         struct inet6_ifaddr *ifp;
4178
4179         read_lock_bh(&idev->lock);
4180         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4181                 spin_lock(&ifp->lock);
4182                 if ((ifp->flags & IFA_F_TENTATIVE &&
4183                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4184                         if (restart)
4185                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4186                         addrconf_dad_kick(ifp);
4187                 }
4188                 spin_unlock(&ifp->lock);
4189         }
4190         read_unlock_bh(&idev->lock);
4191 }
4192
4193 #ifdef CONFIG_PROC_FS
4194 struct if6_iter_state {
4195         struct seq_net_private p;
4196         int bucket;
4197         int offset;
4198 };
4199
4200 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4201 {
4202         struct if6_iter_state *state = seq->private;
4203         struct net *net = seq_file_net(seq);
4204         struct inet6_ifaddr *ifa = NULL;
4205         int p = 0;
4206
4207         /* initial bucket if pos is 0 */
4208         if (pos == 0) {
4209                 state->bucket = 0;
4210                 state->offset = 0;
4211         }
4212
4213         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4214                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4215                                          addr_lst) {
4216                         if (!net_eq(dev_net(ifa->idev->dev), net))
4217                                 continue;
4218                         /* sync with offset */
4219                         if (p < state->offset) {
4220                                 p++;
4221                                 continue;
4222                         }
4223                         return ifa;
4224                 }
4225
4226                 /* prepare for next bucket */
4227                 state->offset = 0;
4228                 p = 0;
4229         }
4230         return NULL;
4231 }
4232
4233 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4234                                          struct inet6_ifaddr *ifa)
4235 {
4236         struct if6_iter_state *state = seq->private;
4237         struct net *net = seq_file_net(seq);
4238
4239         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4240                 if (!net_eq(dev_net(ifa->idev->dev), net))
4241                         continue;
4242                 state->offset++;
4243                 return ifa;
4244         }
4245
4246         state->offset = 0;
4247         while (++state->bucket < IN6_ADDR_HSIZE) {
4248                 hlist_for_each_entry_rcu(ifa,
4249                                      &inet6_addr_lst[state->bucket], addr_lst) {
4250                         if (!net_eq(dev_net(ifa->idev->dev), net))
4251                                 continue;
4252                         return ifa;
4253                 }
4254         }
4255
4256         return NULL;
4257 }
4258
4259 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4260         __acquires(rcu)
4261 {
4262         rcu_read_lock();
4263         return if6_get_first(seq, *pos);
4264 }
4265
4266 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4267 {
4268         struct inet6_ifaddr *ifa;
4269
4270         ifa = if6_get_next(seq, v);
4271         ++*pos;
4272         return ifa;
4273 }
4274
4275 static void if6_seq_stop(struct seq_file *seq, void *v)
4276         __releases(rcu)
4277 {
4278         rcu_read_unlock();
4279 }
4280
4281 static int if6_seq_show(struct seq_file *seq, void *v)
4282 {
4283         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4284         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4285                    &ifp->addr,
4286                    ifp->idev->dev->ifindex,
4287                    ifp->prefix_len,
4288                    ifp->scope,
4289                    (u8) ifp->flags,
4290                    ifp->idev->dev->name);
4291         return 0;
4292 }
4293
4294 static const struct seq_operations if6_seq_ops = {
4295         .start  = if6_seq_start,
4296         .next   = if6_seq_next,
4297         .show   = if6_seq_show,
4298         .stop   = if6_seq_stop,
4299 };
4300
4301 static int __net_init if6_proc_net_init(struct net *net)
4302 {
4303         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4304                         sizeof(struct if6_iter_state)))
4305                 return -ENOMEM;
4306         return 0;
4307 }
4308
4309 static void __net_exit if6_proc_net_exit(struct net *net)
4310 {
4311         remove_proc_entry("if_inet6", net->proc_net);
4312 }
4313
4314 static struct pernet_operations if6_proc_net_ops = {
4315         .init = if6_proc_net_init,
4316         .exit = if6_proc_net_exit,
4317 };
4318
4319 int __init if6_proc_init(void)
4320 {
4321         return register_pernet_subsys(&if6_proc_net_ops);
4322 }
4323
4324 void if6_proc_exit(void)
4325 {
4326         unregister_pernet_subsys(&if6_proc_net_ops);
4327 }
4328 #endif  /* CONFIG_PROC_FS */
4329
4330 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4331 /* Check if address is a home address configured on any interface. */
4332 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4333 {
4334         unsigned int hash = inet6_addr_hash(net, addr);
4335         struct inet6_ifaddr *ifp = NULL;
4336         int ret = 0;
4337
4338         rcu_read_lock();
4339         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4340                 if (!net_eq(dev_net(ifp->idev->dev), net))
4341                         continue;
4342                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4343                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4344                         ret = 1;
4345                         break;
4346                 }
4347         }
4348         rcu_read_unlock();
4349         return ret;
4350 }
4351 #endif
4352
4353 /*
4354  *      Periodic address status verification
4355  */
4356
4357 static void addrconf_verify_rtnl(void)
4358 {
4359         unsigned long now, next, next_sec, next_sched;
4360         struct inet6_ifaddr *ifp;
4361         int i;
4362
4363         ASSERT_RTNL();
4364
4365         rcu_read_lock_bh();
4366         now = jiffies;
4367         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4368
4369         cancel_delayed_work(&addr_chk_work);
4370
4371         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4372 restart:
4373                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4374                         unsigned long age;
4375
4376                         /* When setting preferred_lft to a value not zero or
4377                          * infinity, while valid_lft is infinity
4378                          * IFA_F_PERMANENT has a non-infinity life time.
4379                          */
4380                         if ((ifp->flags & IFA_F_PERMANENT) &&
4381                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4382                                 continue;
4383
4384                         spin_lock(&ifp->lock);
4385                         /* We try to batch several events at once. */
4386                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4387
4388                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4389                             age >= ifp->valid_lft) {
4390                                 spin_unlock(&ifp->lock);
4391                                 in6_ifa_hold(ifp);
4392                                 ipv6_del_addr(ifp);
4393                                 goto restart;
4394                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4395                                 spin_unlock(&ifp->lock);
4396                                 continue;
4397                         } else if (age >= ifp->prefered_lft) {
4398                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4399                                 int deprecate = 0;
4400
4401                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4402                                         deprecate = 1;
4403                                         ifp->flags |= IFA_F_DEPRECATED;
4404                                 }
4405
4406                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4407                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4408                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4409
4410                                 spin_unlock(&ifp->lock);
4411
4412                                 if (deprecate) {
4413                                         in6_ifa_hold(ifp);
4414
4415                                         ipv6_ifa_notify(0, ifp);
4416                                         in6_ifa_put(ifp);
4417                                         goto restart;
4418                                 }
4419                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4420                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4421                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4422                                         ifp->idev->cnf.dad_transmits *
4423                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4424
4425                                 if (age >= ifp->prefered_lft - regen_advance) {
4426                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4427                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4428                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4429                                         if (!ifp->regen_count && ifpub) {
4430                                                 ifp->regen_count++;
4431                                                 in6_ifa_hold(ifp);
4432                                                 in6_ifa_hold(ifpub);
4433                                                 spin_unlock(&ifp->lock);
4434
4435                                                 spin_lock(&ifpub->lock);
4436                                                 ifpub->regen_count = 0;
4437                                                 spin_unlock(&ifpub->lock);
4438                                                 rcu_read_unlock_bh();
4439                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4440                                                 in6_ifa_put(ifpub);
4441                                                 in6_ifa_put(ifp);
4442                                                 rcu_read_lock_bh();
4443                                                 goto restart;
4444                                         }
4445                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4446                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4447                                 spin_unlock(&ifp->lock);
4448                         } else {
4449                                 /* ifp->prefered_lft <= ifp->valid_lft */
4450                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4451                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4452                                 spin_unlock(&ifp->lock);
4453                         }
4454                 }
4455         }
4456
4457         next_sec = round_jiffies_up(next);
4458         next_sched = next;
4459
4460         /* If rounded timeout is accurate enough, accept it. */
4461         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4462                 next_sched = next_sec;
4463
4464         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4465         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4466                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4467
4468         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4469                  now, next, next_sec, next_sched);
4470         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4471         rcu_read_unlock_bh();
4472 }
4473
4474 static void addrconf_verify_work(struct work_struct *w)
4475 {
4476         rtnl_lock();
4477         addrconf_verify_rtnl();
4478         rtnl_unlock();
4479 }
4480
4481 static void addrconf_verify(void)
4482 {
4483         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4484 }
4485
4486 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4487                                      struct in6_addr **peer_pfx)
4488 {
4489         struct in6_addr *pfx = NULL;
4490
4491         *peer_pfx = NULL;
4492
4493         if (addr)
4494                 pfx = nla_data(addr);
4495
4496         if (local) {
4497                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4498                         *peer_pfx = pfx;
4499                 pfx = nla_data(local);
4500         }
4501
4502         return pfx;
4503 }
4504
4505 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4506         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4507         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4508         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4509         [IFA_FLAGS]             = { .len = sizeof(u32) },
4510         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4511 };
4512
4513 static int
4514 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4515                   struct netlink_ext_ack *extack)
4516 {
4517         struct net *net = sock_net(skb->sk);
4518         struct ifaddrmsg *ifm;
4519         struct nlattr *tb[IFA_MAX+1];
4520         struct in6_addr *pfx, *peer_pfx;
4521         u32 ifa_flags;
4522         int err;
4523
4524         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4525                           extack);
4526         if (err < 0)
4527                 return err;
4528
4529         ifm = nlmsg_data(nlh);
4530         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4531         if (!pfx)
4532                 return -EINVAL;
4533
4534         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4535
4536         /* We ignore other flags so far. */
4537         ifa_flags &= IFA_F_MANAGETEMPADDR;
4538
4539         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4540                               ifm->ifa_prefixlen);
4541 }
4542
4543 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4544                                unsigned long expires, u32 flags,
4545                                bool modify_peer)
4546 {
4547         struct fib6_info *f6i;
4548         u32 prio;
4549
4550         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4551                                         ifp->prefix_len,
4552                                         ifp->idev->dev,
4553                                         0, RTF_GATEWAY | RTF_DEFAULT);
4554         if (!f6i)
4555                 return -ENOENT;
4556
4557         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4558         if (f6i->fib6_metric != prio) {
4559                 /* delete old one */
4560                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4561
4562                 /* add new one */
4563                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4564                                       ifp->prefix_len,
4565                                       ifp->rt_priority, ifp->idev->dev,
4566                                       expires, flags, GFP_KERNEL);
4567         } else {
4568                 if (!expires)
4569                         fib6_clean_expires(f6i);
4570                 else
4571                         fib6_set_expires(f6i, expires);
4572
4573                 fib6_info_release(f6i);
4574         }
4575
4576         return 0;
4577 }
4578
4579 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4580 {
4581         u32 flags;
4582         clock_t expires;
4583         unsigned long timeout;
4584         bool was_managetempaddr;
4585         bool had_prefixroute;
4586         bool new_peer = false;
4587
4588         ASSERT_RTNL();
4589
4590         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4591                 return -EINVAL;
4592
4593         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4594             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4595                 return -EINVAL;
4596
4597         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4598                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4599
4600         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4601         if (addrconf_finite_timeout(timeout)) {
4602                 expires = jiffies_to_clock_t(timeout * HZ);
4603                 cfg->valid_lft = timeout;
4604                 flags = RTF_EXPIRES;
4605         } else {
4606                 expires = 0;
4607                 flags = 0;
4608                 cfg->ifa_flags |= IFA_F_PERMANENT;
4609         }
4610
4611         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4612         if (addrconf_finite_timeout(timeout)) {
4613                 if (timeout == 0)
4614                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4615                 cfg->preferred_lft = timeout;
4616         }
4617
4618         if (cfg->peer_pfx &&
4619             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4620                 if (!ipv6_addr_any(&ifp->peer_addr))
4621                         cleanup_prefix_route(ifp, expires, true, true);
4622                 new_peer = true;
4623         }
4624
4625         spin_lock_bh(&ifp->lock);
4626         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4627         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4628                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4629         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4630                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4631                         IFA_F_NOPREFIXROUTE);
4632         ifp->flags |= cfg->ifa_flags;
4633         ifp->tstamp = jiffies;
4634         ifp->valid_lft = cfg->valid_lft;
4635         ifp->prefered_lft = cfg->preferred_lft;
4636
4637         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4638                 ifp->rt_priority = cfg->rt_priority;
4639
4640         if (new_peer)
4641                 ifp->peer_addr = *cfg->peer_pfx;
4642
4643         spin_unlock_bh(&ifp->lock);
4644         if (!(ifp->flags&IFA_F_TENTATIVE))
4645                 ipv6_ifa_notify(0, ifp);
4646
4647         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4648                 int rc = -ENOENT;
4649
4650                 if (had_prefixroute)
4651                         rc = modify_prefix_route(ifp, expires, flags, false);
4652
4653                 /* prefix route could have been deleted; if so restore it */
4654                 if (rc == -ENOENT) {
4655                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4656                                               ifp->rt_priority, ifp->idev->dev,
4657                                               expires, flags, GFP_KERNEL);
4658                 }
4659
4660                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4661                         rc = modify_prefix_route(ifp, expires, flags, true);
4662
4663                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4664                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4665                                               ifp->rt_priority, ifp->idev->dev,
4666                                               expires, flags, GFP_KERNEL);
4667                 }
4668         } else if (had_prefixroute) {
4669                 enum cleanup_prefix_rt_t action;
4670                 unsigned long rt_expires;
4671
4672                 write_lock_bh(&ifp->idev->lock);
4673                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4674                 write_unlock_bh(&ifp->idev->lock);
4675
4676                 if (action != CLEANUP_PREFIX_RT_NOP) {
4677                         cleanup_prefix_route(ifp, rt_expires,
4678                                 action == CLEANUP_PREFIX_RT_DEL, false);
4679                 }
4680         }
4681
4682         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4683                 if (was_managetempaddr &&
4684                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4685                         cfg->valid_lft = 0;
4686                         cfg->preferred_lft = 0;
4687                 }
4688                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4689                                  cfg->preferred_lft, !was_managetempaddr,
4690                                  jiffies);
4691         }
4692
4693         addrconf_verify_rtnl();
4694
4695         return 0;
4696 }
4697
4698 static int
4699 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4700                   struct netlink_ext_ack *extack)
4701 {
4702         struct net *net = sock_net(skb->sk);
4703         struct ifaddrmsg *ifm;
4704         struct nlattr *tb[IFA_MAX+1];
4705         struct in6_addr *peer_pfx;
4706         struct inet6_ifaddr *ifa;
4707         struct net_device *dev;
4708         struct inet6_dev *idev;
4709         struct ifa6_config cfg;
4710         int err;
4711
4712         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4713                           extack);
4714         if (err < 0)
4715                 return err;
4716
4717         memset(&cfg, 0, sizeof(cfg));
4718
4719         ifm = nlmsg_data(nlh);
4720         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4721         if (!cfg.pfx)
4722                 return -EINVAL;
4723
4724         cfg.peer_pfx = peer_pfx;
4725         cfg.plen = ifm->ifa_prefixlen;
4726         if (tb[IFA_RT_PRIORITY])
4727                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4728
4729         cfg.valid_lft = INFINITY_LIFE_TIME;
4730         cfg.preferred_lft = INFINITY_LIFE_TIME;
4731
4732         if (tb[IFA_CACHEINFO]) {
4733                 struct ifa_cacheinfo *ci;
4734
4735                 ci = nla_data(tb[IFA_CACHEINFO]);
4736                 cfg.valid_lft = ci->ifa_valid;
4737                 cfg.preferred_lft = ci->ifa_prefered;
4738         }
4739
4740         dev =  __dev_get_by_index(net, ifm->ifa_index);
4741         if (!dev)
4742                 return -ENODEV;
4743
4744         if (tb[IFA_FLAGS])
4745                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4746         else
4747                 cfg.ifa_flags = ifm->ifa_flags;
4748
4749         /* We ignore other flags so far. */
4750         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4751                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4752                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4753
4754         idev = ipv6_find_idev(dev);
4755         if (!idev)
4756                 return -ENOBUFS;
4757
4758         if (!ipv6_allow_optimistic_dad(net, idev))
4759                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4760
4761         if (cfg.ifa_flags & IFA_F_NODAD &&
4762             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4763                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4764                 return -EINVAL;
4765         }
4766
4767         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4768         if (!ifa) {
4769                 /*
4770                  * It would be best to check for !NLM_F_CREATE here but
4771                  * userspace already relies on not having to provide this.
4772                  */
4773                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4774         }
4775
4776         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4777             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4778                 err = -EEXIST;
4779         else
4780                 err = inet6_addr_modify(ifa, &cfg);
4781
4782         in6_ifa_put(ifa);
4783
4784         return err;
4785 }
4786
4787 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4788                           u8 scope, int ifindex)
4789 {
4790         struct ifaddrmsg *ifm;
4791
4792         ifm = nlmsg_data(nlh);
4793         ifm->ifa_family = AF_INET6;
4794         ifm->ifa_prefixlen = prefixlen;
4795         ifm->ifa_flags = flags;
4796         ifm->ifa_scope = scope;
4797         ifm->ifa_index = ifindex;
4798 }
4799
4800 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4801                          unsigned long tstamp, u32 preferred, u32 valid)
4802 {
4803         struct ifa_cacheinfo ci;
4804
4805         ci.cstamp = cstamp_delta(cstamp);
4806         ci.tstamp = cstamp_delta(tstamp);
4807         ci.ifa_prefered = preferred;
4808         ci.ifa_valid = valid;
4809
4810         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4811 }
4812
4813 static inline int rt_scope(int ifa_scope)
4814 {
4815         if (ifa_scope & IFA_HOST)
4816                 return RT_SCOPE_HOST;
4817         else if (ifa_scope & IFA_LINK)
4818                 return RT_SCOPE_LINK;
4819         else if (ifa_scope & IFA_SITE)
4820                 return RT_SCOPE_SITE;
4821         else
4822                 return RT_SCOPE_UNIVERSE;
4823 }
4824
4825 static inline int inet6_ifaddr_msgsize(void)
4826 {
4827         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4828                + nla_total_size(16) /* IFA_LOCAL */
4829                + nla_total_size(16) /* IFA_ADDRESS */
4830                + nla_total_size(sizeof(struct ifa_cacheinfo))
4831                + nla_total_size(4)  /* IFA_FLAGS */
4832                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4833 }
4834
4835 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4836                              u32 portid, u32 seq, int event, unsigned int flags)
4837 {
4838         struct nlmsghdr  *nlh;
4839         u32 preferred, valid;
4840
4841         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4842         if (!nlh)
4843                 return -EMSGSIZE;
4844
4845         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4846                       ifa->idev->dev->ifindex);
4847
4848         if (!((ifa->flags&IFA_F_PERMANENT) &&
4849               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4850                 preferred = ifa->prefered_lft;
4851                 valid = ifa->valid_lft;
4852                 if (preferred != INFINITY_LIFE_TIME) {
4853                         long tval = (jiffies - ifa->tstamp)/HZ;
4854                         if (preferred > tval)
4855                                 preferred -= tval;
4856                         else
4857                                 preferred = 0;
4858                         if (valid != INFINITY_LIFE_TIME) {
4859                                 if (valid > tval)
4860                                         valid -= tval;
4861                                 else
4862                                         valid = 0;
4863                         }
4864                 }
4865         } else {
4866                 preferred = INFINITY_LIFE_TIME;
4867                 valid = INFINITY_LIFE_TIME;
4868         }
4869
4870         if (!ipv6_addr_any(&ifa->peer_addr)) {
4871                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4872                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4873                         goto error;
4874         } else
4875                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4876                         goto error;
4877
4878         if (ifa->rt_priority &&
4879             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4880                 goto error;
4881
4882         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4883                 goto error;
4884
4885         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4886                 goto error;
4887
4888         nlmsg_end(skb, nlh);
4889         return 0;
4890
4891 error:
4892         nlmsg_cancel(skb, nlh);
4893         return -EMSGSIZE;
4894 }
4895
4896 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4897                                 u32 portid, u32 seq, int event, u16 flags)
4898 {
4899         struct nlmsghdr  *nlh;
4900         u8 scope = RT_SCOPE_UNIVERSE;
4901         int ifindex = ifmca->idev->dev->ifindex;
4902
4903         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4904                 scope = RT_SCOPE_SITE;
4905
4906         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4907         if (!nlh)
4908                 return -EMSGSIZE;
4909
4910         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4911         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4912             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4913                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4914                 nlmsg_cancel(skb, nlh);
4915                 return -EMSGSIZE;
4916         }
4917
4918         nlmsg_end(skb, nlh);
4919         return 0;
4920 }
4921
4922 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4923                                 u32 portid, u32 seq, int event, unsigned int flags)
4924 {
4925         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4926         int ifindex = dev ? dev->ifindex : 1;
4927         struct nlmsghdr  *nlh;
4928         u8 scope = RT_SCOPE_UNIVERSE;
4929
4930         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4931                 scope = RT_SCOPE_SITE;
4932
4933         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4934         if (!nlh)
4935                 return -EMSGSIZE;
4936
4937         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4938         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4939             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4940                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4941                 nlmsg_cancel(skb, nlh);
4942                 return -EMSGSIZE;
4943         }
4944
4945         nlmsg_end(skb, nlh);
4946         return 0;
4947 }
4948
4949 enum addr_type_t {
4950         UNICAST_ADDR,
4951         MULTICAST_ADDR,
4952         ANYCAST_ADDR,
4953 };
4954
4955 /* called with rcu_read_lock() */
4956 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4957                           struct netlink_callback *cb, enum addr_type_t type,
4958                           int s_ip_idx, int *p_ip_idx)
4959 {
4960         struct ifmcaddr6 *ifmca;
4961         struct ifacaddr6 *ifaca;
4962         int err = 1;
4963         int ip_idx = *p_ip_idx;
4964
4965         read_lock_bh(&idev->lock);
4966         switch (type) {
4967         case UNICAST_ADDR: {
4968                 struct inet6_ifaddr *ifa;
4969
4970                 /* unicast address incl. temp addr */
4971                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4972                         if (ip_idx < s_ip_idx)
4973                                 goto next;
4974                         err = inet6_fill_ifaddr(skb, ifa,
4975                                                 NETLINK_CB(cb->skb).portid,
4976                                                 cb->nlh->nlmsg_seq,
4977                                                 RTM_NEWADDR,
4978                                                 NLM_F_MULTI);
4979                         if (err < 0)
4980                                 break;
4981                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4982 next:
4983                         ip_idx++;
4984                 }
4985                 break;
4986         }
4987         case MULTICAST_ADDR:
4988                 /* multicast address */
4989                 for (ifmca = idev->mc_list; ifmca;
4990                      ifmca = ifmca->next, ip_idx++) {
4991                         if (ip_idx < s_ip_idx)
4992                                 continue;
4993                         err = inet6_fill_ifmcaddr(skb, ifmca,
4994                                                   NETLINK_CB(cb->skb).portid,
4995                                                   cb->nlh->nlmsg_seq,
4996                                                   RTM_GETMULTICAST,
4997                                                   NLM_F_MULTI);
4998                         if (err < 0)
4999                                 break;
5000                 }
5001                 break;
5002         case ANYCAST_ADDR:
5003                 /* anycast address */
5004                 for (ifaca = idev->ac_list; ifaca;
5005                      ifaca = ifaca->aca_next, ip_idx++) {
5006                         if (ip_idx < s_ip_idx)
5007                                 continue;
5008                         err = inet6_fill_ifacaddr(skb, ifaca,
5009                                                   NETLINK_CB(cb->skb).portid,
5010                                                   cb->nlh->nlmsg_seq,
5011                                                   RTM_GETANYCAST,
5012                                                   NLM_F_MULTI);
5013                         if (err < 0)
5014                                 break;
5015                 }
5016                 break;
5017         default:
5018                 break;
5019         }
5020         read_unlock_bh(&idev->lock);
5021         *p_ip_idx = ip_idx;
5022         return err;
5023 }
5024
5025 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5026                            enum addr_type_t type)
5027 {
5028         struct net *net = sock_net(skb->sk);
5029         int h, s_h;
5030         int idx, ip_idx;
5031         int s_idx, s_ip_idx;
5032         struct net_device *dev;
5033         struct inet6_dev *idev;
5034         struct hlist_head *head;
5035
5036         s_h = cb->args[0];
5037         s_idx = idx = cb->args[1];
5038         s_ip_idx = ip_idx = cb->args[2];
5039
5040         rcu_read_lock();
5041         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
5042         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5043                 idx = 0;
5044                 head = &net->dev_index_head[h];
5045                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5046                         if (idx < s_idx)
5047                                 goto cont;
5048                         if (h > s_h || idx > s_idx)
5049                                 s_ip_idx = 0;
5050                         ip_idx = 0;
5051                         idev = __in6_dev_get(dev);
5052                         if (!idev)
5053                                 goto cont;
5054
5055                         if (in6_dump_addrs(idev, skb, cb, type,
5056                                            s_ip_idx, &ip_idx) < 0)
5057                                 goto done;
5058 cont:
5059                         idx++;
5060                 }
5061         }
5062 done:
5063         rcu_read_unlock();
5064         cb->args[0] = h;
5065         cb->args[1] = idx;
5066         cb->args[2] = ip_idx;
5067
5068         return skb->len;
5069 }
5070
5071 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5072 {
5073         enum addr_type_t type = UNICAST_ADDR;
5074
5075         return inet6_dump_addr(skb, cb, type);
5076 }
5077
5078 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5079 {
5080         enum addr_type_t type = MULTICAST_ADDR;
5081
5082         return inet6_dump_addr(skb, cb, type);
5083 }
5084
5085
5086 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5087 {
5088         enum addr_type_t type = ANYCAST_ADDR;
5089
5090         return inet6_dump_addr(skb, cb, type);
5091 }
5092
5093 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5094                              struct netlink_ext_ack *extack)
5095 {
5096         struct net *net = sock_net(in_skb->sk);
5097         struct ifaddrmsg *ifm;
5098         struct nlattr *tb[IFA_MAX+1];
5099         struct in6_addr *addr = NULL, *peer;
5100         struct net_device *dev = NULL;
5101         struct inet6_ifaddr *ifa;
5102         struct sk_buff *skb;
5103         int err;
5104
5105         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5106                           extack);
5107         if (err < 0)
5108                 return err;
5109
5110         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5111         if (!addr)
5112                 return -EINVAL;
5113
5114         ifm = nlmsg_data(nlh);
5115         if (ifm->ifa_index)
5116                 dev = dev_get_by_index(net, ifm->ifa_index);
5117
5118         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5119         if (!ifa) {
5120                 err = -EADDRNOTAVAIL;
5121                 goto errout;
5122         }
5123
5124         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5125         if (!skb) {
5126                 err = -ENOBUFS;
5127                 goto errout_ifa;
5128         }
5129
5130         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5131                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5132         if (err < 0) {
5133                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5134                 WARN_ON(err == -EMSGSIZE);
5135                 kfree_skb(skb);
5136                 goto errout_ifa;
5137         }
5138         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5139 errout_ifa:
5140         in6_ifa_put(ifa);
5141 errout:
5142         if (dev)
5143                 dev_put(dev);
5144         return err;
5145 }
5146
5147 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5148 {
5149         struct sk_buff *skb;
5150         struct net *net = dev_net(ifa->idev->dev);
5151         int err = -ENOBUFS;
5152
5153         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5154         if (!skb)
5155                 goto errout;
5156
5157         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5158         if (err < 0) {
5159                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5160                 WARN_ON(err == -EMSGSIZE);
5161                 kfree_skb(skb);
5162                 goto errout;
5163         }
5164         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5165         return;
5166 errout:
5167         if (err < 0)
5168                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5169 }
5170
5171 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5172                                 __s32 *array, int bytes)
5173 {
5174         BUG_ON(bytes < (DEVCONF_MAX * 4));
5175
5176         memset(array, 0, bytes);
5177         array[DEVCONF_FORWARDING] = cnf->forwarding;
5178         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5179         array[DEVCONF_MTU6] = cnf->mtu6;
5180         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5181         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5182         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5183         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5184         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5185         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5186                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5187         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5188                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5189         array[DEVCONF_RTR_SOLICIT_DELAY] =
5190                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5191         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5192         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5193                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5194         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5195                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5196         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5197         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5198         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5199         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5200         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5201         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5202         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5203         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5204         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5205 #ifdef CONFIG_IPV6_ROUTER_PREF
5206         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5207         array[DEVCONF_RTR_PROBE_INTERVAL] =
5208                 jiffies_to_msecs(cnf->rtr_probe_interval);
5209 #ifdef CONFIG_IPV6_ROUTE_INFO
5210         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5211         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5212 #endif
5213 #endif
5214         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5215         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5216 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5217         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5218         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5219 #endif
5220 #ifdef CONFIG_IPV6_MROUTE
5221         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5222 #endif
5223         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5224         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5225         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5226         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5227         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5228         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5229         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5230         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5231         /* we omit DEVCONF_STABLE_SECRET for now */
5232         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5233         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5234         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5235         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5236         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5237 #ifdef CONFIG_IPV6_SEG6_HMAC
5238         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5239 #endif
5240         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5241         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5242         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5243         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5244 }
5245
5246 static inline size_t inet6_ifla6_size(void)
5247 {
5248         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5249              + nla_total_size(sizeof(struct ifla_cacheinfo))
5250              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5251              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5252              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5253              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5254              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5255              + 0;
5256 }
5257
5258 static inline size_t inet6_if_nlmsg_size(void)
5259 {
5260         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5261                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5262                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5263                + nla_total_size(4) /* IFLA_MTU */
5264                + nla_total_size(4) /* IFLA_LINK */
5265                + nla_total_size(1) /* IFLA_OPERSTATE */
5266                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5267 }
5268
5269 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5270                                         int bytes)
5271 {
5272         int i;
5273         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5274         BUG_ON(pad < 0);
5275
5276         /* Use put_unaligned() because stats may not be aligned for u64. */
5277         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5278         for (i = 1; i < ICMP6_MIB_MAX; i++)
5279                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5280
5281         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5282 }
5283
5284 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5285                                         int bytes, size_t syncpoff)
5286 {
5287         int i, c;
5288         u64 buff[IPSTATS_MIB_MAX];
5289         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5290
5291         BUG_ON(pad < 0);
5292
5293         memset(buff, 0, sizeof(buff));
5294         buff[0] = IPSTATS_MIB_MAX;
5295
5296         for_each_possible_cpu(c) {
5297                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5298                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5299         }
5300
5301         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5302         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5303 }
5304
5305 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5306                              int bytes)
5307 {
5308         switch (attrtype) {
5309         case IFLA_INET6_STATS:
5310                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5311                                      offsetof(struct ipstats_mib, syncp));
5312                 break;
5313         case IFLA_INET6_ICMP6STATS:
5314                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5315                 break;
5316         }
5317 }
5318
5319 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5320                                   u32 ext_filter_mask)
5321 {
5322         struct nlattr *nla;
5323         struct ifla_cacheinfo ci;
5324
5325         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5326                 goto nla_put_failure;
5327         ci.max_reasm_len = IPV6_MAXPLEN;
5328         ci.tstamp = cstamp_delta(idev->tstamp);
5329         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5330         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5331         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5332                 goto nla_put_failure;
5333         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5334         if (!nla)
5335                 goto nla_put_failure;
5336         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5337
5338         /* XXX - MC not implemented */
5339
5340         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5341                 return 0;
5342
5343         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5344         if (!nla)
5345                 goto nla_put_failure;
5346         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5347
5348         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5349         if (!nla)
5350                 goto nla_put_failure;
5351         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5352
5353         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5354         if (!nla)
5355                 goto nla_put_failure;
5356
5357         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5358                 goto nla_put_failure;
5359
5360         read_lock_bh(&idev->lock);
5361         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5362         read_unlock_bh(&idev->lock);
5363
5364         return 0;
5365
5366 nla_put_failure:
5367         return -EMSGSIZE;
5368 }
5369
5370 static size_t inet6_get_link_af_size(const struct net_device *dev,
5371                                      u32 ext_filter_mask)
5372 {
5373         if (!__in6_dev_get(dev))
5374                 return 0;
5375
5376         return inet6_ifla6_size();
5377 }
5378
5379 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5380                               u32 ext_filter_mask)
5381 {
5382         struct inet6_dev *idev = __in6_dev_get(dev);
5383
5384         if (!idev)
5385                 return -ENODATA;
5386
5387         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5388                 return -EMSGSIZE;
5389
5390         return 0;
5391 }
5392
5393 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5394 {
5395         struct inet6_ifaddr *ifp;
5396         struct net_device *dev = idev->dev;
5397         bool clear_token, update_rs = false;
5398         struct in6_addr ll_addr;
5399
5400         ASSERT_RTNL();
5401
5402         if (!token)
5403                 return -EINVAL;
5404         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5405                 return -EINVAL;
5406         if (!ipv6_accept_ra(idev))
5407                 return -EINVAL;
5408         if (idev->cnf.rtr_solicits == 0)
5409                 return -EINVAL;
5410
5411         write_lock_bh(&idev->lock);
5412
5413         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5414         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5415
5416         write_unlock_bh(&idev->lock);
5417
5418         clear_token = ipv6_addr_any(token);
5419         if (clear_token)
5420                 goto update_lft;
5421
5422         if (!idev->dead && (idev->if_flags & IF_READY) &&
5423             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5424                              IFA_F_OPTIMISTIC)) {
5425                 /* If we're not ready, then normal ifup will take care
5426                  * of this. Otherwise, we need to request our rs here.
5427                  */
5428                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5429                 update_rs = true;
5430         }
5431
5432 update_lft:
5433         write_lock_bh(&idev->lock);
5434
5435         if (update_rs) {
5436                 idev->if_flags |= IF_RS_SENT;
5437                 idev->rs_interval = rfc3315_s14_backoff_init(
5438                         idev->cnf.rtr_solicit_interval);
5439                 idev->rs_probes = 1;
5440                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5441         }
5442
5443         /* Well, that's kinda nasty ... */
5444         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5445                 spin_lock(&ifp->lock);
5446                 if (ifp->tokenized) {
5447                         ifp->valid_lft = 0;
5448                         ifp->prefered_lft = 0;
5449                 }
5450                 spin_unlock(&ifp->lock);
5451         }
5452
5453         write_unlock_bh(&idev->lock);
5454         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5455         addrconf_verify_rtnl();
5456         return 0;
5457 }
5458
5459 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5460         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5461         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5462 };
5463
5464 static int inet6_validate_link_af(const struct net_device *dev,
5465                                   const struct nlattr *nla)
5466 {
5467         struct nlattr *tb[IFLA_INET6_MAX + 1];
5468
5469         if (dev && !__in6_dev_get(dev))
5470                 return -EAFNOSUPPORT;
5471
5472         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5473                                 NULL);
5474 }
5475
5476 static int check_addr_gen_mode(int mode)
5477 {
5478         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5479             mode != IN6_ADDR_GEN_MODE_NONE &&
5480             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5481             mode != IN6_ADDR_GEN_MODE_RANDOM)
5482                 return -EINVAL;
5483         return 1;
5484 }
5485
5486 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5487                                 int mode)
5488 {
5489         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5490             !idev->cnf.stable_secret.initialized &&
5491             !net->ipv6.devconf_dflt->stable_secret.initialized)
5492                 return -EINVAL;
5493         return 1;
5494 }
5495
5496 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5497 {
5498         int err = -EINVAL;
5499         struct inet6_dev *idev = __in6_dev_get(dev);
5500         struct nlattr *tb[IFLA_INET6_MAX + 1];
5501
5502         if (!idev)
5503                 return -EAFNOSUPPORT;
5504
5505         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5506                 BUG();
5507
5508         if (tb[IFLA_INET6_TOKEN]) {
5509                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5510                 if (err)
5511                         return err;
5512         }
5513
5514         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5515                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5516
5517                 if (check_addr_gen_mode(mode) < 0 ||
5518                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5519                         return -EINVAL;
5520
5521                 idev->cnf.addr_gen_mode = mode;
5522                 err = 0;
5523         }
5524
5525         return err;
5526 }
5527
5528 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5529                              u32 portid, u32 seq, int event, unsigned int flags)
5530 {
5531         struct net_device *dev = idev->dev;
5532         struct ifinfomsg *hdr;
5533         struct nlmsghdr *nlh;
5534         void *protoinfo;
5535
5536         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5537         if (!nlh)
5538                 return -EMSGSIZE;
5539
5540         hdr = nlmsg_data(nlh);
5541         hdr->ifi_family = AF_INET6;
5542         hdr->__ifi_pad = 0;
5543         hdr->ifi_type = dev->type;
5544         hdr->ifi_index = dev->ifindex;
5545         hdr->ifi_flags = dev_get_flags(dev);
5546         hdr->ifi_change = 0;
5547
5548         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5549             (dev->addr_len &&
5550              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5551             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5552             (dev->ifindex != dev_get_iflink(dev) &&
5553              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5554             nla_put_u8(skb, IFLA_OPERSTATE,
5555                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5556                 goto nla_put_failure;
5557         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5558         if (!protoinfo)
5559                 goto nla_put_failure;
5560
5561         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5562                 goto nla_put_failure;
5563
5564         nla_nest_end(skb, protoinfo);
5565         nlmsg_end(skb, nlh);
5566         return 0;
5567
5568 nla_put_failure:
5569         nlmsg_cancel(skb, nlh);
5570         return -EMSGSIZE;
5571 }
5572
5573 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5574 {
5575         struct net *net = sock_net(skb->sk);
5576         int h, s_h;
5577         int idx = 0, s_idx;
5578         struct net_device *dev;
5579         struct inet6_dev *idev;
5580         struct hlist_head *head;
5581
5582         s_h = cb->args[0];
5583         s_idx = cb->args[1];
5584
5585         rcu_read_lock();
5586         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5587                 idx = 0;
5588                 head = &net->dev_index_head[h];
5589                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5590                         if (idx < s_idx)
5591                                 goto cont;
5592                         idev = __in6_dev_get(dev);
5593                         if (!idev)
5594                                 goto cont;
5595                         if (inet6_fill_ifinfo(skb, idev,
5596                                               NETLINK_CB(cb->skb).portid,
5597                                               cb->nlh->nlmsg_seq,
5598                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5599                                 goto out;
5600 cont:
5601                         idx++;
5602                 }
5603         }
5604 out:
5605         rcu_read_unlock();
5606         cb->args[1] = idx;
5607         cb->args[0] = h;
5608
5609         return skb->len;
5610 }
5611
5612 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5613 {
5614         struct sk_buff *skb;
5615         struct net *net = dev_net(idev->dev);
5616         int err = -ENOBUFS;
5617
5618         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5619         if (!skb)
5620                 goto errout;
5621
5622         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5623         if (err < 0) {
5624                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5625                 WARN_ON(err == -EMSGSIZE);
5626                 kfree_skb(skb);
5627                 goto errout;
5628         }
5629         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5630         return;
5631 errout:
5632         if (err < 0)
5633                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5634 }
5635
5636 static inline size_t inet6_prefix_nlmsg_size(void)
5637 {
5638         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5639                + nla_total_size(sizeof(struct in6_addr))
5640                + nla_total_size(sizeof(struct prefix_cacheinfo));
5641 }
5642
5643 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5644                              struct prefix_info *pinfo, u32 portid, u32 seq,
5645                              int event, unsigned int flags)
5646 {
5647         struct prefixmsg *pmsg;
5648         struct nlmsghdr *nlh;
5649         struct prefix_cacheinfo ci;
5650
5651         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5652         if (!nlh)
5653                 return -EMSGSIZE;
5654
5655         pmsg = nlmsg_data(nlh);
5656         pmsg->prefix_family = AF_INET6;
5657         pmsg->prefix_pad1 = 0;
5658         pmsg->prefix_pad2 = 0;
5659         pmsg->prefix_ifindex = idev->dev->ifindex;
5660         pmsg->prefix_len = pinfo->prefix_len;
5661         pmsg->prefix_type = pinfo->type;
5662         pmsg->prefix_pad3 = 0;
5663         pmsg->prefix_flags = 0;
5664         if (pinfo->onlink)
5665                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5666         if (pinfo->autoconf)
5667                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5668
5669         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5670                 goto nla_put_failure;
5671         ci.preferred_time = ntohl(pinfo->prefered);
5672         ci.valid_time = ntohl(pinfo->valid);
5673         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5674                 goto nla_put_failure;
5675         nlmsg_end(skb, nlh);
5676         return 0;
5677
5678 nla_put_failure:
5679         nlmsg_cancel(skb, nlh);
5680         return -EMSGSIZE;
5681 }
5682
5683 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5684                          struct prefix_info *pinfo)
5685 {
5686         struct sk_buff *skb;
5687         struct net *net = dev_net(idev->dev);
5688         int err = -ENOBUFS;
5689
5690         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5691         if (!skb)
5692                 goto errout;
5693
5694         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5695         if (err < 0) {
5696                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5697                 WARN_ON(err == -EMSGSIZE);
5698                 kfree_skb(skb);
5699                 goto errout;
5700         }
5701         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5702         return;
5703 errout:
5704         if (err < 0)
5705                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5706 }
5707
5708 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5709 {
5710         struct net *net = dev_net(ifp->idev->dev);
5711
5712         if (event)
5713                 ASSERT_RTNL();
5714
5715         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5716
5717         switch (event) {
5718         case RTM_NEWADDR:
5719                 /*
5720                  * If the address was optimistic we inserted the route at the
5721                  * start of our DAD process, so we don't need to do it again.
5722                  * If the device was taken down in the middle of the DAD
5723                  * cycle there is a race where we could get here without a
5724                  * host route, so nothing to insert. That will be fixed when
5725                  * the device is brought up.
5726                  */
5727                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
5728                         ip6_ins_rt(net, ifp->rt);
5729                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
5730                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
5731                                 &ifp->addr, ifp->idev->dev->name);
5732                 }
5733
5734                 if (ifp->idev->cnf.forwarding)
5735                         addrconf_join_anycast(ifp);
5736                 if (!ipv6_addr_any(&ifp->peer_addr))
5737                         addrconf_prefix_route(&ifp->peer_addr, 128,
5738                                               ifp->rt_priority, ifp->idev->dev,
5739                                               0, 0, GFP_ATOMIC);
5740                 break;
5741         case RTM_DELADDR:
5742                 if (ifp->idev->cnf.forwarding)
5743                         addrconf_leave_anycast(ifp);
5744                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5745                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5746                         struct fib6_info *rt;
5747
5748                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5749                                                        ifp->idev->dev, 0, 0);
5750                         if (rt)
5751                                 ip6_del_rt(net, rt);
5752                 }
5753                 if (ifp->rt) {
5754                         ip6_del_rt(net, ifp->rt);
5755                         ifp->rt = NULL;
5756                 }
5757                 rt_genid_bump_ipv6(net);
5758                 break;
5759         }
5760         atomic_inc(&net->ipv6.dev_addr_genid);
5761 }
5762
5763 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5764 {
5765         rcu_read_lock_bh();
5766         if (likely(ifp->idev->dead == 0))
5767                 __ipv6_ifa_notify(event, ifp);
5768         rcu_read_unlock_bh();
5769 }
5770
5771 #ifdef CONFIG_SYSCTL
5772
5773 static
5774 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5775                            void __user *buffer, size_t *lenp, loff_t *ppos)
5776 {
5777         int *valp = ctl->data;
5778         int val = *valp;
5779         loff_t pos = *ppos;
5780         struct ctl_table lctl;
5781         int ret;
5782
5783         /*
5784          * ctl->data points to idev->cnf.forwarding, we should
5785          * not modify it until we get the rtnl lock.
5786          */
5787         lctl = *ctl;
5788         lctl.data = &val;
5789
5790         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5791
5792         if (write)
5793                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5794         if (ret)
5795                 *ppos = pos;
5796         return ret;
5797 }
5798
5799 static
5800 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5801                         void __user *buffer, size_t *lenp, loff_t *ppos)
5802 {
5803         struct inet6_dev *idev = ctl->extra1;
5804         int min_mtu = IPV6_MIN_MTU;
5805         struct ctl_table lctl;
5806
5807         lctl = *ctl;
5808         lctl.extra1 = &min_mtu;
5809         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5810
5811         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5812 }
5813
5814 static void dev_disable_change(struct inet6_dev *idev)
5815 {
5816         struct netdev_notifier_info info;
5817
5818         if (!idev || !idev->dev)
5819                 return;
5820
5821         netdev_notifier_info_init(&info, idev->dev);
5822         if (idev->cnf.disable_ipv6)
5823                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5824         else
5825                 addrconf_notify(NULL, NETDEV_UP, &info);
5826 }
5827
5828 static void addrconf_disable_change(struct net *net, __s32 newf)
5829 {
5830         struct net_device *dev;
5831         struct inet6_dev *idev;
5832
5833         for_each_netdev(net, dev) {
5834                 idev = __in6_dev_get(dev);
5835                 if (idev) {
5836                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5837                         idev->cnf.disable_ipv6 = newf;
5838                         if (changed)
5839                                 dev_disable_change(idev);
5840                 }
5841         }
5842 }
5843
5844 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5845 {
5846         struct net *net;
5847         int old;
5848
5849         if (!rtnl_trylock())
5850                 return restart_syscall();
5851
5852         net = (struct net *)table->extra2;
5853         old = *p;
5854         *p = newf;
5855
5856         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5857                 rtnl_unlock();
5858                 return 0;
5859         }
5860
5861         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5862                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5863                 addrconf_disable_change(net, newf);
5864         } else if ((!newf) ^ (!old))
5865                 dev_disable_change((struct inet6_dev *)table->extra1);
5866
5867         rtnl_unlock();
5868         return 0;
5869 }
5870
5871 static
5872 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5873                             void __user *buffer, size_t *lenp, loff_t *ppos)
5874 {
5875         int *valp = ctl->data;
5876         int val = *valp;
5877         loff_t pos = *ppos;
5878         struct ctl_table lctl;
5879         int ret;
5880
5881         /*
5882          * ctl->data points to idev->cnf.disable_ipv6, we should
5883          * not modify it until we get the rtnl lock.
5884          */
5885         lctl = *ctl;
5886         lctl.data = &val;
5887
5888         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5889
5890         if (write)
5891                 ret = addrconf_disable_ipv6(ctl, valp, val);
5892         if (ret)
5893                 *ppos = pos;
5894         return ret;
5895 }
5896
5897 static
5898 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5899                               void __user *buffer, size_t *lenp, loff_t *ppos)
5900 {
5901         int *valp = ctl->data;
5902         int ret;
5903         int old, new;
5904
5905         old = *valp;
5906         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5907         new = *valp;
5908
5909         if (write && old != new) {
5910                 struct net *net = ctl->extra2;
5911
5912                 if (!rtnl_trylock())
5913                         return restart_syscall();
5914
5915                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5916                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5917                                                      NETCONFA_PROXY_NEIGH,
5918                                                      NETCONFA_IFINDEX_DEFAULT,
5919                                                      net->ipv6.devconf_dflt);
5920                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5921                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5922                                                      NETCONFA_PROXY_NEIGH,
5923                                                      NETCONFA_IFINDEX_ALL,
5924                                                      net->ipv6.devconf_all);
5925                 else {
5926                         struct inet6_dev *idev = ctl->extra1;
5927
5928                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5929                                                      NETCONFA_PROXY_NEIGH,
5930                                                      idev->dev->ifindex,
5931                                                      &idev->cnf);
5932                 }
5933                 rtnl_unlock();
5934         }
5935
5936         return ret;
5937 }
5938
5939 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5940                                          void __user *buffer, size_t *lenp,
5941                                          loff_t *ppos)
5942 {
5943         int ret = 0;
5944         u32 new_val;
5945         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5946         struct net *net = (struct net *)ctl->extra2;
5947         struct ctl_table tmp = {
5948                 .data = &new_val,
5949                 .maxlen = sizeof(new_val),
5950                 .mode = ctl->mode,
5951         };
5952
5953         if (!rtnl_trylock())
5954                 return restart_syscall();
5955
5956         new_val = *((u32 *)ctl->data);
5957
5958         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
5959         if (ret != 0)
5960                 goto out;
5961
5962         if (write) {
5963                 if (check_addr_gen_mode(new_val) < 0) {
5964                         ret = -EINVAL;
5965                         goto out;
5966                 }
5967
5968                 if (idev) {
5969                         if (check_stable_privacy(idev, net, new_val) < 0) {
5970                                 ret = -EINVAL;
5971                                 goto out;
5972                         }
5973
5974                         if (idev->cnf.addr_gen_mode != new_val) {
5975                                 idev->cnf.addr_gen_mode = new_val;
5976                                 addrconf_dev_config(idev->dev);
5977                         }
5978                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
5979                         struct net_device *dev;
5980
5981                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
5982                         for_each_netdev(net, dev) {
5983                                 idev = __in6_dev_get(dev);
5984                                 if (idev &&
5985                                     idev->cnf.addr_gen_mode != new_val) {
5986                                         idev->cnf.addr_gen_mode = new_val;
5987                                         addrconf_dev_config(idev->dev);
5988                                 }
5989                         }
5990                 }
5991
5992                 *((u32 *)ctl->data) = new_val;
5993         }
5994
5995 out:
5996         rtnl_unlock();
5997
5998         return ret;
5999 }
6000
6001 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6002                                          void __user *buffer, size_t *lenp,
6003                                          loff_t *ppos)
6004 {
6005         int err;
6006         struct in6_addr addr;
6007         char str[IPV6_MAX_STRLEN];
6008         struct ctl_table lctl = *ctl;
6009         struct net *net = ctl->extra2;
6010         struct ipv6_stable_secret *secret = ctl->data;
6011
6012         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6013                 return -EIO;
6014
6015         lctl.maxlen = IPV6_MAX_STRLEN;
6016         lctl.data = str;
6017
6018         if (!rtnl_trylock())
6019                 return restart_syscall();
6020
6021         if (!write && !secret->initialized) {
6022                 err = -EIO;
6023                 goto out;
6024         }
6025
6026         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6027         if (err >= sizeof(str)) {
6028                 err = -EIO;
6029                 goto out;
6030         }
6031
6032         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6033         if (err || !write)
6034                 goto out;
6035
6036         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6037                 err = -EIO;
6038                 goto out;
6039         }
6040
6041         secret->initialized = true;
6042         secret->secret = addr;
6043
6044         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6045                 struct net_device *dev;
6046
6047                 for_each_netdev(net, dev) {
6048                         struct inet6_dev *idev = __in6_dev_get(dev);
6049
6050                         if (idev) {
6051                                 idev->cnf.addr_gen_mode =
6052                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6053                         }
6054                 }
6055         } else {
6056                 struct inet6_dev *idev = ctl->extra1;
6057
6058                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6059         }
6060
6061 out:
6062         rtnl_unlock();
6063
6064         return err;
6065 }
6066
6067 static
6068 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6069                                                 int write,
6070                                                 void __user *buffer,
6071                                                 size_t *lenp,
6072                                                 loff_t *ppos)
6073 {
6074         int *valp = ctl->data;
6075         int val = *valp;
6076         loff_t pos = *ppos;
6077         struct ctl_table lctl;
6078         int ret;
6079
6080         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6081          * we should not modify it until we get the rtnl lock.
6082          */
6083         lctl = *ctl;
6084         lctl.data = &val;
6085
6086         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6087
6088         if (write)
6089                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6090         if (ret)
6091                 *ppos = pos;
6092         return ret;
6093 }
6094
6095 static
6096 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6097 {
6098         if (rt) {
6099                 if (action)
6100                         rt->dst.flags |= DST_NOPOLICY;
6101                 else
6102                         rt->dst.flags &= ~DST_NOPOLICY;
6103         }
6104 }
6105
6106 static
6107 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6108 {
6109         struct inet6_ifaddr *ifa;
6110
6111         read_lock_bh(&idev->lock);
6112         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6113                 spin_lock(&ifa->lock);
6114                 if (ifa->rt) {
6115                         struct fib6_info *rt = ifa->rt;
6116                         int cpu;
6117
6118                         rcu_read_lock();
6119                         ifa->rt->dst_nopolicy = val ? true : false;
6120                         if (rt->rt6i_pcpu) {
6121                                 for_each_possible_cpu(cpu) {
6122                                         struct rt6_info **rtp;
6123
6124                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6125                                         addrconf_set_nopolicy(*rtp, val);
6126                                 }
6127                         }
6128                         rcu_read_unlock();
6129                 }
6130                 spin_unlock(&ifa->lock);
6131         }
6132         read_unlock_bh(&idev->lock);
6133 }
6134
6135 static
6136 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6137 {
6138         struct inet6_dev *idev;
6139         struct net *net;
6140
6141         if (!rtnl_trylock())
6142                 return restart_syscall();
6143
6144         *valp = val;
6145
6146         net = (struct net *)ctl->extra2;
6147         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6148                 rtnl_unlock();
6149                 return 0;
6150         }
6151
6152         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6153                 struct net_device *dev;
6154
6155                 for_each_netdev(net, dev) {
6156                         idev = __in6_dev_get(dev);
6157                         if (idev)
6158                                 addrconf_disable_policy_idev(idev, val);
6159                 }
6160         } else {
6161                 idev = (struct inet6_dev *)ctl->extra1;
6162                 addrconf_disable_policy_idev(idev, val);
6163         }
6164
6165         rtnl_unlock();
6166         return 0;
6167 }
6168
6169 static
6170 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6171                                    void __user *buffer, size_t *lenp,
6172                                    loff_t *ppos)
6173 {
6174         int *valp = ctl->data;
6175         int val = *valp;
6176         loff_t pos = *ppos;
6177         struct ctl_table lctl;
6178         int ret;
6179
6180         lctl = *ctl;
6181         lctl.data = &val;
6182         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6183
6184         if (write && (*valp != val))
6185                 ret = addrconf_disable_policy(ctl, valp, val);
6186
6187         if (ret)
6188                 *ppos = pos;
6189
6190         return ret;
6191 }
6192
6193 static int minus_one = -1;
6194 static const int zero = 0;
6195 static const int one = 1;
6196 static const int two_five_five = 255;
6197
6198 static const struct ctl_table addrconf_sysctl[] = {
6199         {
6200                 .procname       = "forwarding",
6201                 .data           = &ipv6_devconf.forwarding,
6202                 .maxlen         = sizeof(int),
6203                 .mode           = 0644,
6204                 .proc_handler   = addrconf_sysctl_forward,
6205         },
6206         {
6207                 .procname       = "hop_limit",
6208                 .data           = &ipv6_devconf.hop_limit,
6209                 .maxlen         = sizeof(int),
6210                 .mode           = 0644,
6211                 .proc_handler   = proc_dointvec_minmax,
6212                 .extra1         = (void *)&one,
6213                 .extra2         = (void *)&two_five_five,
6214         },
6215         {
6216                 .procname       = "mtu",
6217                 .data           = &ipv6_devconf.mtu6,
6218                 .maxlen         = sizeof(int),
6219                 .mode           = 0644,
6220                 .proc_handler   = addrconf_sysctl_mtu,
6221         },
6222         {
6223                 .procname       = "accept_ra",
6224                 .data           = &ipv6_devconf.accept_ra,
6225                 .maxlen         = sizeof(int),
6226                 .mode           = 0644,
6227                 .proc_handler   = proc_dointvec,
6228         },
6229         {
6230                 .procname       = "accept_redirects",
6231                 .data           = &ipv6_devconf.accept_redirects,
6232                 .maxlen         = sizeof(int),
6233                 .mode           = 0644,
6234                 .proc_handler   = proc_dointvec,
6235         },
6236         {
6237                 .procname       = "autoconf",
6238                 .data           = &ipv6_devconf.autoconf,
6239                 .maxlen         = sizeof(int),
6240                 .mode           = 0644,
6241                 .proc_handler   = proc_dointvec,
6242         },
6243         {
6244                 .procname       = "dad_transmits",
6245                 .data           = &ipv6_devconf.dad_transmits,
6246                 .maxlen         = sizeof(int),
6247                 .mode           = 0644,
6248                 .proc_handler   = proc_dointvec,
6249         },
6250         {
6251                 .procname       = "router_solicitations",
6252                 .data           = &ipv6_devconf.rtr_solicits,
6253                 .maxlen         = sizeof(int),
6254                 .mode           = 0644,
6255                 .proc_handler   = proc_dointvec_minmax,
6256                 .extra1         = &minus_one,
6257         },
6258         {
6259                 .procname       = "router_solicitation_interval",
6260                 .data           = &ipv6_devconf.rtr_solicit_interval,
6261                 .maxlen         = sizeof(int),
6262                 .mode           = 0644,
6263                 .proc_handler   = proc_dointvec_jiffies,
6264         },
6265         {
6266                 .procname       = "router_solicitation_max_interval",
6267                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6268                 .maxlen         = sizeof(int),
6269                 .mode           = 0644,
6270                 .proc_handler   = proc_dointvec_jiffies,
6271         },
6272         {
6273                 .procname       = "router_solicitation_delay",
6274                 .data           = &ipv6_devconf.rtr_solicit_delay,
6275                 .maxlen         = sizeof(int),
6276                 .mode           = 0644,
6277                 .proc_handler   = proc_dointvec_jiffies,
6278         },
6279         {
6280                 .procname       = "force_mld_version",
6281                 .data           = &ipv6_devconf.force_mld_version,
6282                 .maxlen         = sizeof(int),
6283                 .mode           = 0644,
6284                 .proc_handler   = proc_dointvec,
6285         },
6286         {
6287                 .procname       = "mldv1_unsolicited_report_interval",
6288                 .data           =
6289                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6290                 .maxlen         = sizeof(int),
6291                 .mode           = 0644,
6292                 .proc_handler   = proc_dointvec_ms_jiffies,
6293         },
6294         {
6295                 .procname       = "mldv2_unsolicited_report_interval",
6296                 .data           =
6297                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6298                 .maxlen         = sizeof(int),
6299                 .mode           = 0644,
6300                 .proc_handler   = proc_dointvec_ms_jiffies,
6301         },
6302         {
6303                 .procname       = "use_tempaddr",
6304                 .data           = &ipv6_devconf.use_tempaddr,
6305                 .maxlen         = sizeof(int),
6306                 .mode           = 0644,
6307                 .proc_handler   = proc_dointvec,
6308         },
6309         {
6310                 .procname       = "temp_valid_lft",
6311                 .data           = &ipv6_devconf.temp_valid_lft,
6312                 .maxlen         = sizeof(int),
6313                 .mode           = 0644,
6314                 .proc_handler   = proc_dointvec,
6315         },
6316         {
6317                 .procname       = "temp_prefered_lft",
6318                 .data           = &ipv6_devconf.temp_prefered_lft,
6319                 .maxlen         = sizeof(int),
6320                 .mode           = 0644,
6321                 .proc_handler   = proc_dointvec,
6322         },
6323         {
6324                 .procname       = "regen_max_retry",
6325                 .data           = &ipv6_devconf.regen_max_retry,
6326                 .maxlen         = sizeof(int),
6327                 .mode           = 0644,
6328                 .proc_handler   = proc_dointvec,
6329         },
6330         {
6331                 .procname       = "max_desync_factor",
6332                 .data           = &ipv6_devconf.max_desync_factor,
6333                 .maxlen         = sizeof(int),
6334                 .mode           = 0644,
6335                 .proc_handler   = proc_dointvec,
6336         },
6337         {
6338                 .procname       = "max_addresses",
6339                 .data           = &ipv6_devconf.max_addresses,
6340                 .maxlen         = sizeof(int),
6341                 .mode           = 0644,
6342                 .proc_handler   = proc_dointvec,
6343         },
6344         {
6345                 .procname       = "accept_ra_defrtr",
6346                 .data           = &ipv6_devconf.accept_ra_defrtr,
6347                 .maxlen         = sizeof(int),
6348                 .mode           = 0644,
6349                 .proc_handler   = proc_dointvec,
6350         },
6351         {
6352                 .procname       = "accept_ra_min_hop_limit",
6353                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6354                 .maxlen         = sizeof(int),
6355                 .mode           = 0644,
6356                 .proc_handler   = proc_dointvec,
6357         },
6358         {
6359                 .procname       = "accept_ra_pinfo",
6360                 .data           = &ipv6_devconf.accept_ra_pinfo,
6361                 .maxlen         = sizeof(int),
6362                 .mode           = 0644,
6363                 .proc_handler   = proc_dointvec,
6364         },
6365 #ifdef CONFIG_IPV6_ROUTER_PREF
6366         {
6367                 .procname       = "accept_ra_rtr_pref",
6368                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6369                 .maxlen         = sizeof(int),
6370                 .mode           = 0644,
6371                 .proc_handler   = proc_dointvec,
6372         },
6373         {
6374                 .procname       = "router_probe_interval",
6375                 .data           = &ipv6_devconf.rtr_probe_interval,
6376                 .maxlen         = sizeof(int),
6377                 .mode           = 0644,
6378                 .proc_handler   = proc_dointvec_jiffies,
6379         },
6380 #ifdef CONFIG_IPV6_ROUTE_INFO
6381         {
6382                 .procname       = "accept_ra_rt_info_min_plen",
6383                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6384                 .maxlen         = sizeof(int),
6385                 .mode           = 0644,
6386                 .proc_handler   = proc_dointvec,
6387         },
6388         {
6389                 .procname       = "accept_ra_rt_info_max_plen",
6390                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6391                 .maxlen         = sizeof(int),
6392                 .mode           = 0644,
6393                 .proc_handler   = proc_dointvec,
6394         },
6395 #endif
6396 #endif
6397         {
6398                 .procname       = "proxy_ndp",
6399                 .data           = &ipv6_devconf.proxy_ndp,
6400                 .maxlen         = sizeof(int),
6401                 .mode           = 0644,
6402                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6403         },
6404         {
6405                 .procname       = "accept_source_route",
6406                 .data           = &ipv6_devconf.accept_source_route,
6407                 .maxlen         = sizeof(int),
6408                 .mode           = 0644,
6409                 .proc_handler   = proc_dointvec,
6410         },
6411 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6412         {
6413                 .procname       = "optimistic_dad",
6414                 .data           = &ipv6_devconf.optimistic_dad,
6415                 .maxlen         = sizeof(int),
6416                 .mode           = 0644,
6417                 .proc_handler   = proc_dointvec,
6418         },
6419         {
6420                 .procname       = "use_optimistic",
6421                 .data           = &ipv6_devconf.use_optimistic,
6422                 .maxlen         = sizeof(int),
6423                 .mode           = 0644,
6424                 .proc_handler   = proc_dointvec,
6425         },
6426 #endif
6427 #ifdef CONFIG_IPV6_MROUTE
6428         {
6429                 .procname       = "mc_forwarding",
6430                 .data           = &ipv6_devconf.mc_forwarding,
6431                 .maxlen         = sizeof(int),
6432                 .mode           = 0444,
6433                 .proc_handler   = proc_dointvec,
6434         },
6435 #endif
6436         {
6437                 .procname       = "disable_ipv6",
6438                 .data           = &ipv6_devconf.disable_ipv6,
6439                 .maxlen         = sizeof(int),
6440                 .mode           = 0644,
6441                 .proc_handler   = addrconf_sysctl_disable,
6442         },
6443         {
6444                 .procname       = "accept_dad",
6445                 .data           = &ipv6_devconf.accept_dad,
6446                 .maxlen         = sizeof(int),
6447                 .mode           = 0644,
6448                 .proc_handler   = proc_dointvec,
6449         },
6450         {
6451                 .procname       = "force_tllao",
6452                 .data           = &ipv6_devconf.force_tllao,
6453                 .maxlen         = sizeof(int),
6454                 .mode           = 0644,
6455                 .proc_handler   = proc_dointvec
6456         },
6457         {
6458                 .procname       = "ndisc_notify",
6459                 .data           = &ipv6_devconf.ndisc_notify,
6460                 .maxlen         = sizeof(int),
6461                 .mode           = 0644,
6462                 .proc_handler   = proc_dointvec
6463         },
6464         {
6465                 .procname       = "suppress_frag_ndisc",
6466                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6467                 .maxlen         = sizeof(int),
6468                 .mode           = 0644,
6469                 .proc_handler   = proc_dointvec
6470         },
6471         {
6472                 .procname       = "accept_ra_from_local",
6473                 .data           = &ipv6_devconf.accept_ra_from_local,
6474                 .maxlen         = sizeof(int),
6475                 .mode           = 0644,
6476                 .proc_handler   = proc_dointvec,
6477         },
6478         {
6479                 .procname       = "accept_ra_mtu",
6480                 .data           = &ipv6_devconf.accept_ra_mtu,
6481                 .maxlen         = sizeof(int),
6482                 .mode           = 0644,
6483                 .proc_handler   = proc_dointvec,
6484         },
6485         {
6486                 .procname       = "stable_secret",
6487                 .data           = &ipv6_devconf.stable_secret,
6488                 .maxlen         = IPV6_MAX_STRLEN,
6489                 .mode           = 0600,
6490                 .proc_handler   = addrconf_sysctl_stable_secret,
6491         },
6492         {
6493                 .procname       = "use_oif_addrs_only",
6494                 .data           = &ipv6_devconf.use_oif_addrs_only,
6495                 .maxlen         = sizeof(int),
6496                 .mode           = 0644,
6497                 .proc_handler   = proc_dointvec,
6498         },
6499         {
6500                 .procname       = "ignore_routes_with_linkdown",
6501                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6502                 .maxlen         = sizeof(int),
6503                 .mode           = 0644,
6504                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6505         },
6506         {
6507                 .procname       = "drop_unicast_in_l2_multicast",
6508                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6509                 .maxlen         = sizeof(int),
6510                 .mode           = 0644,
6511                 .proc_handler   = proc_dointvec,
6512         },
6513         {
6514                 .procname       = "drop_unsolicited_na",
6515                 .data           = &ipv6_devconf.drop_unsolicited_na,
6516                 .maxlen         = sizeof(int),
6517                 .mode           = 0644,
6518                 .proc_handler   = proc_dointvec,
6519         },
6520         {
6521                 .procname       = "keep_addr_on_down",
6522                 .data           = &ipv6_devconf.keep_addr_on_down,
6523                 .maxlen         = sizeof(int),
6524                 .mode           = 0644,
6525                 .proc_handler   = proc_dointvec,
6526
6527         },
6528         {
6529                 .procname       = "seg6_enabled",
6530                 .data           = &ipv6_devconf.seg6_enabled,
6531                 .maxlen         = sizeof(int),
6532                 .mode           = 0644,
6533                 .proc_handler   = proc_dointvec,
6534         },
6535 #ifdef CONFIG_IPV6_SEG6_HMAC
6536         {
6537                 .procname       = "seg6_require_hmac",
6538                 .data           = &ipv6_devconf.seg6_require_hmac,
6539                 .maxlen         = sizeof(int),
6540                 .mode           = 0644,
6541                 .proc_handler   = proc_dointvec,
6542         },
6543 #endif
6544         {
6545                 .procname       = "enhanced_dad",
6546                 .data           = &ipv6_devconf.enhanced_dad,
6547                 .maxlen         = sizeof(int),
6548                 .mode           = 0644,
6549                 .proc_handler   = proc_dointvec,
6550         },
6551         {
6552                 .procname               = "addr_gen_mode",
6553                 .data                   = &ipv6_devconf.addr_gen_mode,
6554                 .maxlen                 = sizeof(int),
6555                 .mode                   = 0644,
6556                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6557         },
6558         {
6559                 .procname       = "disable_policy",
6560                 .data           = &ipv6_devconf.disable_policy,
6561                 .maxlen         = sizeof(int),
6562                 .mode           = 0644,
6563                 .proc_handler   = addrconf_sysctl_disable_policy,
6564         },
6565         {
6566                 .procname       = "ndisc_tclass",
6567                 .data           = &ipv6_devconf.ndisc_tclass,
6568                 .maxlen         = sizeof(int),
6569                 .mode           = 0644,
6570                 .proc_handler   = proc_dointvec_minmax,
6571                 .extra1         = (void *)&zero,
6572                 .extra2         = (void *)&two_five_five,
6573         },
6574         {
6575                 /* sentinel */
6576         }
6577 };
6578
6579 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6580                 struct inet6_dev *idev, struct ipv6_devconf *p)
6581 {
6582         int i, ifindex;
6583         struct ctl_table *table;
6584         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6585
6586         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6587         if (!table)
6588                 goto out;
6589
6590         for (i = 0; table[i].data; i++) {
6591                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6592                 /* If one of these is already set, then it is not safe to
6593                  * overwrite either of them: this makes proc_dointvec_minmax
6594                  * usable.
6595                  */
6596                 if (!table[i].extra1 && !table[i].extra2) {
6597                         table[i].extra1 = idev; /* embedded; no ref */
6598                         table[i].extra2 = net;
6599                 }
6600         }
6601
6602         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6603
6604         p->sysctl_header = register_net_sysctl(net, path, table);
6605         if (!p->sysctl_header)
6606                 goto free;
6607
6608         if (!strcmp(dev_name, "all"))
6609                 ifindex = NETCONFA_IFINDEX_ALL;
6610         else if (!strcmp(dev_name, "default"))
6611                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6612         else
6613                 ifindex = idev->dev->ifindex;
6614         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6615                                      ifindex, p);
6616         return 0;
6617
6618 free:
6619         kfree(table);
6620 out:
6621         return -ENOBUFS;
6622 }
6623
6624 static void __addrconf_sysctl_unregister(struct net *net,
6625                                          struct ipv6_devconf *p, int ifindex)
6626 {
6627         struct ctl_table *table;
6628
6629         if (!p->sysctl_header)
6630                 return;
6631
6632         table = p->sysctl_header->ctl_table_arg;
6633         unregister_net_sysctl_table(p->sysctl_header);
6634         p->sysctl_header = NULL;
6635         kfree(table);
6636
6637         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6638 }
6639
6640 static int addrconf_sysctl_register(struct inet6_dev *idev)
6641 {
6642         int err;
6643
6644         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6645                 return -EINVAL;
6646
6647         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6648                                     &ndisc_ifinfo_sysctl_change);
6649         if (err)
6650                 return err;
6651         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6652                                          idev, &idev->cnf);
6653         if (err)
6654                 neigh_sysctl_unregister(idev->nd_parms);
6655
6656         return err;
6657 }
6658
6659 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6660 {
6661         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6662                                      idev->dev->ifindex);
6663         neigh_sysctl_unregister(idev->nd_parms);
6664 }
6665
6666
6667 #endif
6668
6669 static int __net_init addrconf_init_net(struct net *net)
6670 {
6671         int err = -ENOMEM;
6672         struct ipv6_devconf *all, *dflt;
6673
6674         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6675         if (!all)
6676                 goto err_alloc_all;
6677
6678         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6679         if (!dflt)
6680                 goto err_alloc_dflt;
6681
6682         /* these will be inherited by all namespaces */
6683         dflt->autoconf = ipv6_defaults.autoconf;
6684         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6685
6686         dflt->stable_secret.initialized = false;
6687         all->stable_secret.initialized = false;
6688
6689         net->ipv6.devconf_all = all;
6690         net->ipv6.devconf_dflt = dflt;
6691
6692 #ifdef CONFIG_SYSCTL
6693         err = __addrconf_sysctl_register(net, "all", NULL, all);
6694         if (err < 0)
6695                 goto err_reg_all;
6696
6697         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6698         if (err < 0)
6699                 goto err_reg_dflt;
6700 #endif
6701         return 0;
6702
6703 #ifdef CONFIG_SYSCTL
6704 err_reg_dflt:
6705         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6706 err_reg_all:
6707         kfree(dflt);
6708 #endif
6709 err_alloc_dflt:
6710         kfree(all);
6711 err_alloc_all:
6712         return err;
6713 }
6714
6715 static void __net_exit addrconf_exit_net(struct net *net)
6716 {
6717 #ifdef CONFIG_SYSCTL
6718         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6719                                      NETCONFA_IFINDEX_DEFAULT);
6720         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6721                                      NETCONFA_IFINDEX_ALL);
6722 #endif
6723         kfree(net->ipv6.devconf_dflt);
6724         kfree(net->ipv6.devconf_all);
6725 }
6726
6727 static struct pernet_operations addrconf_ops = {
6728         .init = addrconf_init_net,
6729         .exit = addrconf_exit_net,
6730 };
6731
6732 static struct rtnl_af_ops inet6_ops __read_mostly = {
6733         .family           = AF_INET6,
6734         .fill_link_af     = inet6_fill_link_af,
6735         .get_link_af_size = inet6_get_link_af_size,
6736         .validate_link_af = inet6_validate_link_af,
6737         .set_link_af      = inet6_set_link_af,
6738 };
6739
6740 /*
6741  *      Init / cleanup code
6742  */
6743
6744 int __init addrconf_init(void)
6745 {
6746         struct inet6_dev *idev;
6747         int i, err;
6748
6749         err = ipv6_addr_label_init();
6750         if (err < 0) {
6751                 pr_crit("%s: cannot initialize default policy table: %d\n",
6752                         __func__, err);
6753                 goto out;
6754         }
6755
6756         err = register_pernet_subsys(&addrconf_ops);
6757         if (err < 0)
6758                 goto out_addrlabel;
6759
6760         addrconf_wq = create_workqueue("ipv6_addrconf");
6761         if (!addrconf_wq) {
6762                 err = -ENOMEM;
6763                 goto out_nowq;
6764         }
6765
6766         /* The addrconf netdev notifier requires that loopback_dev
6767          * has it's ipv6 private information allocated and setup
6768          * before it can bring up and give link-local addresses
6769          * to other devices which are up.
6770          *
6771          * Unfortunately, loopback_dev is not necessarily the first
6772          * entry in the global dev_base list of net devices.  In fact,
6773          * it is likely to be the very last entry on that list.
6774          * So this causes the notifier registry below to try and
6775          * give link-local addresses to all devices besides loopback_dev
6776          * first, then loopback_dev, which cases all the non-loopback_dev
6777          * devices to fail to get a link-local address.
6778          *
6779          * So, as a temporary fix, allocate the ipv6 structure for
6780          * loopback_dev first by hand.
6781          * Longer term, all of the dependencies ipv6 has upon the loopback
6782          * device and it being up should be removed.
6783          */
6784         rtnl_lock();
6785         idev = ipv6_add_dev(init_net.loopback_dev);
6786         rtnl_unlock();
6787         if (IS_ERR(idev)) {
6788                 err = PTR_ERR(idev);
6789                 goto errlo;
6790         }
6791
6792         ip6_route_init_special_entries();
6793
6794         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6795                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6796
6797         register_netdevice_notifier(&ipv6_dev_notf);
6798
6799         addrconf_verify();
6800
6801         rtnl_af_register(&inet6_ops);
6802
6803         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6804                                    NULL, inet6_dump_ifinfo, 0);
6805         if (err < 0)
6806                 goto errout;
6807
6808         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6809                                    inet6_rtm_newaddr, NULL, 0);
6810         if (err < 0)
6811                 goto errout;
6812         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6813                                    inet6_rtm_deladdr, NULL, 0);
6814         if (err < 0)
6815                 goto errout;
6816         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6817                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
6818                                    RTNL_FLAG_DOIT_UNLOCKED);
6819         if (err < 0)
6820                 goto errout;
6821         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6822                                    NULL, inet6_dump_ifmcaddr, 0);
6823         if (err < 0)
6824                 goto errout;
6825         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6826                                    NULL, inet6_dump_ifacaddr, 0);
6827         if (err < 0)
6828                 goto errout;
6829         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6830                                    inet6_netconf_get_devconf,
6831                                    inet6_netconf_dump_devconf,
6832                                    RTNL_FLAG_DOIT_UNLOCKED);
6833         if (err < 0)
6834                 goto errout;
6835         err = ipv6_addr_label_rtnl_register();
6836         if (err < 0)
6837                 goto errout;
6838
6839         return 0;
6840 errout:
6841         rtnl_unregister_all(PF_INET6);
6842         rtnl_af_unregister(&inet6_ops);
6843         unregister_netdevice_notifier(&ipv6_dev_notf);
6844 errlo:
6845         destroy_workqueue(addrconf_wq);
6846 out_nowq:
6847         unregister_pernet_subsys(&addrconf_ops);
6848 out_addrlabel:
6849         ipv6_addr_label_cleanup();
6850 out:
6851         return err;
6852 }
6853
6854 void addrconf_cleanup(void)
6855 {
6856         struct net_device *dev;
6857         int i;
6858
6859         unregister_netdevice_notifier(&ipv6_dev_notf);
6860         unregister_pernet_subsys(&addrconf_ops);
6861         ipv6_addr_label_cleanup();
6862
6863         rtnl_af_unregister(&inet6_ops);
6864
6865         rtnl_lock();
6866
6867         /* clean dev list */
6868         for_each_netdev(&init_net, dev) {
6869                 if (__in6_dev_get(dev) == NULL)
6870                         continue;
6871                 addrconf_ifdown(dev, 1);
6872         }
6873         addrconf_ifdown(init_net.loopback_dev, 2);
6874
6875         /*
6876          *      Check hash table.
6877          */
6878         spin_lock_bh(&addrconf_hash_lock);
6879         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6880                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6881         spin_unlock_bh(&addrconf_hash_lock);
6882         cancel_delayed_work(&addr_chk_work);
6883         rtnl_unlock();
6884
6885         destroy_workqueue(addrconf_wq);
6886 }