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 /* Set to 3 to get tracing... */
  98 #define ACONF_DEBUG 2
  99
 100 #if ACONF_DEBUG >= 3
 101 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 102 #else
 103 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 104 #endif
 105
 106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 107
 108 #define IPV6_MAX_STRLEN \
 109         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 110
 111 static inline u32 cstamp_delta(unsigned long cstamp)
 112 {
 113         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 114 }
 115
 116 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 117 {
 118         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 119         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 120         do_div(tmp, 1000000);
 121         return (s32)tmp;
 122 }
 123
 124 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 125 {
 126         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 127         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 128         do_div(tmp, 1000000);
 129         if ((s32)tmp > mrt) {
 130                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 131                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 132                 do_div(tmp, 1000000);
 133         }
 134         return (s32)tmp;
 135 }
 136
 137 #ifdef CONFIG_SYSCTL
 138 static int addrconf_sysctl_register(struct inet6_dev *idev);
 139 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 140 #else
 141 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 142 {
 143         return 0;
 144 }
 145
 146 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 147 {
 148 }
 149 #endif
 150
 151 static void ipv6_regen_rndid(struct inet6_dev *idev);
 152 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 153
 154 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 155 static int ipv6_count_addresses(struct inet6_dev *idev);
 156 static int ipv6_generate_stable_address(struct in6_addr *addr,
 157                                         u8 dad_count,
 158                                         const struct inet6_dev *idev);
 159
 160 /*
 161  *      Configured unicast address hash table
 162  */
 163 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 164 static DEFINE_SPINLOCK(addrconf_hash_lock);
 165
 166 static void addrconf_verify(void);
 167 static void addrconf_verify_rtnl(void);
 168 static void addrconf_verify_work(struct work_struct *);
 169
 170 static struct workqueue_struct *addrconf_wq;
 171 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 172
 173 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 174 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 175
 176 static void addrconf_type_change(struct net_device *dev,
 177                                  unsigned long event);
 178 static int addrconf_ifdown(struct net_device *dev, int how);
 179
 180 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 181                                                   int plen,
 182                                                   const struct net_device *dev,
 183                                                   u32 flags, u32 noflags);
 184
 185 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 186 static void addrconf_dad_work(struct work_struct *w);
 187 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id);
 188 static void addrconf_dad_run(struct inet6_dev *idev);
 189 static void addrconf_rs_timer(unsigned long data);
 190 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 191 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 192
 193 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 194                                 struct prefix_info *pinfo);
 195 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 196                                struct net_device *dev);
 197
 198 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 199         .forwarding             = 0,
 200         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 201         .mtu6                   = IPV6_MIN_MTU,
 202         .accept_ra              = 1,
 203         .accept_redirects       = 1,
 204         .autoconf               = 1,
 205         .force_mld_version      = 0,
 206         .mldv1_unsolicited_report_interval = 10 * HZ,
 207         .mldv2_unsolicited_report_interval = HZ,
 208         .dad_transmits          = 1,
 209         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 210         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 211         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 212         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 213         .use_tempaddr           = 0,
 214         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 215         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 216         .regen_max_retry        = REGEN_MAX_RETRY,
 217         .max_desync_factor      = MAX_DESYNC_FACTOR,
 218         .max_addresses          = IPV6_MAX_ADDRESSES,
 219         .accept_ra_defrtr       = 1,
 220         .accept_ra_from_local   = 0,
 221         .accept_ra_min_hop_limit= 1,
 222         .accept_ra_pinfo        = 1,
 223 #ifdef CONFIG_IPV6_ROUTER_PREF
 224         .accept_ra_rtr_pref     = 1,
 225         .rtr_probe_interval     = 60 * HZ,
 226 #ifdef CONFIG_IPV6_ROUTE_INFO
 227         .accept_ra_rt_info_min_plen = 0,
 228         .accept_ra_rt_info_max_plen = 0,
 229 #endif
 230 #endif
 231         .proxy_ndp              = 0,
 232         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 233         .disable_ipv6           = 0,
 234         .accept_dad             = 1,
 235         .suppress_frag_ndisc    = 1,
 236         .accept_ra_mtu          = 1,
 237         .stable_secret          = {
 238                 .initialized = false,
 239         },
 240         .use_oif_addrs_only     = 0,
 241         .ignore_routes_with_linkdown = 0,
 242         .keep_addr_on_down      = 0,
 243         .seg6_enabled           = 0,
 244 #ifdef CONFIG_IPV6_SEG6_HMAC
 245         .seg6_require_hmac      = 0,
 246 #endif
 247         .enhanced_dad           = 1,
 248         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 249         .disable_policy         = 0,
 250 };
 251
 252 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 253         .forwarding             = 0,
 254         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 255         .mtu6                   = IPV6_MIN_MTU,
 256         .accept_ra              = 1,
 257         .accept_redirects       = 1,
 258         .autoconf               = 1,
 259         .force_mld_version      = 0,
 260         .mldv1_unsolicited_report_interval = 10 * HZ,
 261         .mldv2_unsolicited_report_interval = HZ,
 262         .dad_transmits          = 1,
 263         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 264         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 265         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 266         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 267         .use_tempaddr           = 0,
 268         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 269         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 270         .regen_max_retry        = REGEN_MAX_RETRY,
 271         .max_desync_factor      = MAX_DESYNC_FACTOR,
 272         .max_addresses          = IPV6_MAX_ADDRESSES,
 273         .accept_ra_defrtr       = 1,
 274         .accept_ra_from_local   = 0,
 275         .accept_ra_min_hop_limit= 1,
 276         .accept_ra_pinfo        = 1,
 277 #ifdef CONFIG_IPV6_ROUTER_PREF
 278         .accept_ra_rtr_pref     = 1,
 279         .rtr_probe_interval     = 60 * HZ,
 280 #ifdef CONFIG_IPV6_ROUTE_INFO
 281         .accept_ra_rt_info_min_plen = 0,
 282         .accept_ra_rt_info_max_plen = 0,
 283 #endif
 284 #endif
 285         .proxy_ndp              = 0,
 286         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 287         .disable_ipv6           = 0,
 288         .accept_dad             = 1,
 289         .suppress_frag_ndisc    = 1,
 290         .accept_ra_mtu          = 1,
 291         .stable_secret          = {
 292                 .initialized = false,
 293         },
 294         .use_oif_addrs_only     = 0,
 295         .ignore_routes_with_linkdown = 0,
 296         .keep_addr_on_down      = 0,
 297         .seg6_enabled           = 0,
 298 #ifdef CONFIG_IPV6_SEG6_HMAC
 299         .seg6_require_hmac      = 0,
 300 #endif
 301         .enhanced_dad           = 1,
 302         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 303         .disable_policy         = 0,
 304 };
 305
 306 /* Check if a valid qdisc is available */
 307 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 308 {
 309         return !qdisc_tx_is_noop(dev);
 310 }
 311
 312 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 313 {
 314         if (del_timer(&idev->rs_timer))
 315                 __in6_dev_put(idev);
 316 }
 317
 318 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 319 {
 320         if (cancel_delayed_work(&ifp->dad_work))
 321                 __in6_ifa_put(ifp);
 322 }
 323
 324 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 325                                   unsigned long when)
 326 {
 327         if (!timer_pending(&idev->rs_timer))
 328                 in6_dev_hold(idev);
 329         mod_timer(&idev->rs_timer, jiffies + when);
 330 }
 331
 332 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 333                                    unsigned long delay)
 334 {
 335         in6_ifa_hold(ifp);
 336         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 337                 in6_ifa_put(ifp);
 338 }
 339
 340 static int snmp6_alloc_dev(struct inet6_dev *idev)
 341 {
 342         int i;
 343
 344         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 345         if (!idev->stats.ipv6)
 346                 goto err_ip;
 347
 348         for_each_possible_cpu(i) {
 349                 struct ipstats_mib *addrconf_stats;
 350                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 351                 u64_stats_init(&addrconf_stats->syncp);
 352         }
 353
 354
 355         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 356                                         GFP_KERNEL);
 357         if (!idev->stats.icmpv6dev)
 358                 goto err_icmp;
 359         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 360                                            GFP_KERNEL);
 361         if (!idev->stats.icmpv6msgdev)
 362                 goto err_icmpmsg;
 363
 364         return 0;
 365
 366 err_icmpmsg:
 367         kfree(idev->stats.icmpv6dev);
 368 err_icmp:
 369         free_percpu(idev->stats.ipv6);
 370 err_ip:
 371         return -ENOMEM;
 372 }
 373
 374 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 375 {
 376         struct inet6_dev *ndev;
 377         int err = -ENOMEM;
 378
 379         ASSERT_RTNL();
 380
 381         if (dev->mtu < IPV6_MIN_MTU)
 382                 return ERR_PTR(-EINVAL);
 383
 384         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 385         if (!ndev)
 386                 return ERR_PTR(err);
 387
 388         rwlock_init(&ndev->lock);
 389         ndev->dev = dev;
 390         INIT_LIST_HEAD(&ndev->addr_list);
 391         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 392                     (unsigned long)ndev);
 393         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 394
 395         if (ndev->cnf.stable_secret.initialized)
 396                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 397         else
 398                 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
 399
 400         ndev->cnf.mtu6 = dev->mtu;
 401         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 402         if (!ndev->nd_parms) {
 403                 kfree(ndev);
 404                 return ERR_PTR(err);
 405         }
 406         if (ndev->cnf.forwarding)
 407                 dev_disable_lro(dev);
 408         /* We refer to the device */
 409         dev_hold(dev);
 410
 411         if (snmp6_alloc_dev(ndev) < 0) {
 412                 ADBG(KERN_WARNING
 413                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 414                         __func__, dev->name);
 415                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 416                 dev_put(dev);
 417                 kfree(ndev);
 418                 return ERR_PTR(err);
 419         }
 420
 421         if (snmp6_register_dev(ndev) < 0) {
 422                 ADBG(KERN_WARNING
 423                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 424                         __func__, dev->name);
 425                 goto err_release;
 426         }
 427
 428         /* One reference from device. */
 429         refcount_set(&ndev->refcnt, 1);
 430
 431         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 432                 ndev->cnf.accept_dad = -1;
 433
 434 #if IS_ENABLED(CONFIG_IPV6_SIT)
 435         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 436                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 437                 ndev->cnf.rtr_solicits = 0;
 438         }
 439 #endif
 440
 441         INIT_LIST_HEAD(&ndev->tempaddr_list);
 442         ndev->desync_factor = U32_MAX;
 443         if ((dev->flags&IFF_LOOPBACK) ||
 444             dev->type == ARPHRD_TUNNEL ||
 445             dev->type == ARPHRD_TUNNEL6 ||
 446             dev->type == ARPHRD_SIT ||
 447             dev->type == ARPHRD_NONE) {
 448                 ndev->cnf.use_tempaddr = -1;
 449         } else
 450                 ipv6_regen_rndid(ndev);
 451
 452         ndev->token = in6addr_any;
 453
 454         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 455                 ndev->if_flags |= IF_READY;
 456
 457         ipv6_mc_init_dev(ndev);
 458         ndev->tstamp = jiffies;
 459         err = addrconf_sysctl_register(ndev);
 460         if (err) {
 461                 ipv6_mc_destroy_dev(ndev);
 462                 snmp6_unregister_dev(ndev);
 463                 goto err_release;
 464         }
 465         /* protected by rtnl_lock */
 466         rcu_assign_pointer(dev->ip6_ptr, ndev);
 467
 468         /* Join interface-local all-node multicast group */
 469         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 470
 471         /* Join all-node multicast group */
 472         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 473
 474         /* Join all-router multicast group if forwarding is set */
 475         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 476                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 477
 478         return ndev;
 479
 480 err_release:
 481         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 482         ndev->dead = 1;
 483         in6_dev_finish_destroy(ndev);
 484         return ERR_PTR(err);
 485 }
 486
 487 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 488 {
 489         struct inet6_dev *idev;
 490
 491         ASSERT_RTNL();
 492
 493         idev = __in6_dev_get(dev);
 494         if (!idev) {
 495                 idev = ipv6_add_dev(dev);
 496                 if (IS_ERR(idev))
 497                         return NULL;
 498         }
 499
 500         if (dev->flags&IFF_UP)
 501                 ipv6_mc_up(idev);
 502         return idev;
 503 }
 504
 505 static int inet6_netconf_msgsize_devconf(int type)
 506 {
 507         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 508                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 509         bool all = false;
 510
 511         if (type == NETCONFA_ALL)
 512                 all = true;
 513
 514         if (all || type == NETCONFA_FORWARDING)
 515                 size += nla_total_size(4);
 516 #ifdef CONFIG_IPV6_MROUTE
 517         if (all || type == NETCONFA_MC_FORWARDING)
 518                 size += nla_total_size(4);
 519 #endif
 520         if (all || type == NETCONFA_PROXY_NEIGH)
 521                 size += nla_total_size(4);
 522
 523         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 524                 size += nla_total_size(4);
 525
 526         return size;
 527 }
 528
 529 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 530                                       struct ipv6_devconf *devconf, u32 portid,
 531                                       u32 seq, int event, unsigned int flags,
 532                                       int type)
 533 {
 534         struct nlmsghdr  *nlh;
 535         struct netconfmsg *ncm;
 536         bool all = false;
 537
 538         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 539                         flags);
 540         if (!nlh)
 541                 return -EMSGSIZE;
 542
 543         if (type == NETCONFA_ALL)
 544                 all = true;
 545
 546         ncm = nlmsg_data(nlh);
 547         ncm->ncm_family = AF_INET6;
 548
 549         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 550                 goto nla_put_failure;
 551
 552         if (!devconf)
 553                 goto out;
 554
 555         if ((all || type == NETCONFA_FORWARDING) &&
 556             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 557                 goto nla_put_failure;
 558 #ifdef CONFIG_IPV6_MROUTE
 559         if ((all || type == NETCONFA_MC_FORWARDING) &&
 560             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 561                         devconf->mc_forwarding) < 0)
 562                 goto nla_put_failure;
 563 #endif
 564         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 565             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 566                 goto nla_put_failure;
 567
 568         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 569             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 570                         devconf->ignore_routes_with_linkdown) < 0)
 571                 goto nla_put_failure;
 572
 573 out:
 574         nlmsg_end(skb, nlh);
 575         return 0;
 576
 577 nla_put_failure:
 578         nlmsg_cancel(skb, nlh);
 579         return -EMSGSIZE;
 580 }
 581
 582 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 583                                   int ifindex, struct ipv6_devconf *devconf)
 584 {
 585         struct sk_buff *skb;
 586         int err = -ENOBUFS;
 587
 588         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 589         if (!skb)
 590                 goto errout;
 591
 592         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 593                                          event, 0, type);
 594         if (err < 0) {
 595                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 596                 WARN_ON(err == -EMSGSIZE);
 597                 kfree_skb(skb);
 598                 goto errout;
 599         }
 600         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 601         return;
 602 errout:
 603         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 604 }
 605
 606 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 607         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 608         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 609         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 610         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 611 };
 612
 613 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 614                                      struct nlmsghdr *nlh,
 615                                      struct netlink_ext_ack *extack)
 616 {
 617         struct net *net = sock_net(in_skb->sk);
 618         struct nlattr *tb[NETCONFA_MAX+1];
 619         struct netconfmsg *ncm;
 620         struct sk_buff *skb;
 621         struct ipv6_devconf *devconf;
 622         struct inet6_dev *in6_dev;
 623         struct net_device *dev;
 624         int ifindex;
 625         int err;
 626
 627         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 628                           devconf_ipv6_policy, extack);
 629         if (err < 0)
 630                 goto errout;
 631
 632         err = -EINVAL;
 633         if (!tb[NETCONFA_IFINDEX])
 634                 goto errout;
 635
 636         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 637         switch (ifindex) {
 638         case NETCONFA_IFINDEX_ALL:
 639                 devconf = net->ipv6.devconf_all;
 640                 break;
 641         case NETCONFA_IFINDEX_DEFAULT:
 642                 devconf = net->ipv6.devconf_dflt;
 643                 break;
 644         default:
 645                 dev = __dev_get_by_index(net, ifindex);
 646                 if (!dev)
 647                         goto errout;
 648                 in6_dev = __in6_dev_get(dev);
 649                 if (!in6_dev)
 650                         goto errout;
 651                 devconf = &in6_dev->cnf;
 652                 break;
 653         }
 654
 655         err = -ENOBUFS;
 656         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
 657         if (!skb)
 658                 goto errout;
 659
 660         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 661                                          NETLINK_CB(in_skb).portid,
 662                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 663                                          NETCONFA_ALL);
 664         if (err < 0) {
 665                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 666                 WARN_ON(err == -EMSGSIZE);
 667                 kfree_skb(skb);
 668                 goto errout;
 669         }
 670         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 671 errout:
 672         return err;
 673 }
 674
 675 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 676                                       struct netlink_callback *cb)
 677 {
 678         struct net *net = sock_net(skb->sk);
 679         int h, s_h;
 680         int idx, s_idx;
 681         struct net_device *dev;
 682         struct inet6_dev *idev;
 683         struct hlist_head *head;
 684
 685         s_h = cb->args[0];
 686         s_idx = idx = cb->args[1];
 687
 688         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 689                 idx = 0;
 690                 head = &net->dev_index_head[h];
 691                 rcu_read_lock();
 692                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 693                           net->dev_base_seq;
 694                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 695                         if (idx < s_idx)
 696                                 goto cont;
 697                         idev = __in6_dev_get(dev);
 698                         if (!idev)
 699                                 goto cont;
 700
 701                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 702                                                        &idev->cnf,
 703                                                        NETLINK_CB(cb->skb).portid,
 704                                                        cb->nlh->nlmsg_seq,
 705                                                        RTM_NEWNETCONF,
 706                                                        NLM_F_MULTI,
 707                                                        NETCONFA_ALL) < 0) {
 708                                 rcu_read_unlock();
 709                                 goto done;
 710                         }
 711                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 712 cont:
 713                         idx++;
 714                 }
 715                 rcu_read_unlock();
 716         }
 717         if (h == NETDEV_HASHENTRIES) {
 718                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 719                                                net->ipv6.devconf_all,
 720                                                NETLINK_CB(cb->skb).portid,
 721                                                cb->nlh->nlmsg_seq,
 722                                                RTM_NEWNETCONF, NLM_F_MULTI,
 723                                                NETCONFA_ALL) < 0)
 724                         goto done;
 725                 else
 726                         h++;
 727         }
 728         if (h == NETDEV_HASHENTRIES + 1) {
 729                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 730                                                net->ipv6.devconf_dflt,
 731                                                NETLINK_CB(cb->skb).portid,
 732                                                cb->nlh->nlmsg_seq,
 733                                                RTM_NEWNETCONF, NLM_F_MULTI,
 734                                                NETCONFA_ALL) < 0)
 735                         goto done;
 736                 else
 737                         h++;
 738         }
 739 done:
 740         cb->args[0] = h;
 741         cb->args[1] = idx;
 742
 743         return skb->len;
 744 }
 745
 746 #ifdef CONFIG_SYSCTL
 747 static void dev_forward_change(struct inet6_dev *idev)
 748 {
 749         struct net_device *dev;
 750         struct inet6_ifaddr *ifa;
 751
 752         if (!idev)
 753                 return;
 754         dev = idev->dev;
 755         if (idev->cnf.forwarding)
 756                 dev_disable_lro(dev);
 757         if (dev->flags & IFF_MULTICAST) {
 758                 if (idev->cnf.forwarding) {
 759                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 760                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 761                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 762                 } else {
 763                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 764                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 765                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 766                 }
 767         }
 768
 769         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 770                 if (ifa->flags&IFA_F_TENTATIVE)
 771                         continue;
 772                 if (idev->cnf.forwarding)
 773                         addrconf_join_anycast(ifa);
 774                 else
 775                         addrconf_leave_anycast(ifa);
 776         }
 777         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 778                                      NETCONFA_FORWARDING,
 779                                      dev->ifindex, &idev->cnf);
 780 }
 781
 782
 783 static void addrconf_forward_change(struct net *net, __s32 newf)
 784 {
 785         struct net_device *dev;
 786         struct inet6_dev *idev;
 787
 788         for_each_netdev(net, dev) {
 789                 idev = __in6_dev_get(dev);
 790                 if (idev) {
 791                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 792                         idev->cnf.forwarding = newf;
 793                         if (changed)
 794                                 dev_forward_change(idev);
 795                 }
 796         }
 797 }
 798
 799 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 800 {
 801         struct net *net;
 802         int old;
 803
 804         if (!rtnl_trylock())
 805                 return restart_syscall();
 806
 807         net = (struct net *)table->extra2;
 808         old = *p;
 809         *p = newf;
 810
 811         if (p == &net->ipv6.devconf_dflt->forwarding) {
 812                 if ((!newf) ^ (!old))
 813                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 814                                                      NETCONFA_FORWARDING,
 815                                                      NETCONFA_IFINDEX_DEFAULT,
 816                                                      net->ipv6.devconf_dflt);
 817                 rtnl_unlock();
 818                 return 0;
 819         }
 820
 821         if (p == &net->ipv6.devconf_all->forwarding) {
 822                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 823
 824                 net->ipv6.devconf_dflt->forwarding = newf;
 825                 if ((!newf) ^ (!old_dflt))
 826                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 827                                                      NETCONFA_FORWARDING,
 828                                                      NETCONFA_IFINDEX_DEFAULT,
 829                                                      net->ipv6.devconf_dflt);
 830
 831                 addrconf_forward_change(net, newf);
 832                 if ((!newf) ^ (!old))
 833                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 834                                                      NETCONFA_FORWARDING,
 835                                                      NETCONFA_IFINDEX_ALL,
 836                                                      net->ipv6.devconf_all);
 837         } else if ((!newf) ^ (!old))
 838                 dev_forward_change((struct inet6_dev *)table->extra1);
 839         rtnl_unlock();
 840
 841         if (newf)
 842                 rt6_purge_dflt_routers(net);
 843         return 1;
 844 }
 845
 846 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 847 {
 848         struct net_device *dev;
 849         struct inet6_dev *idev;
 850
 851         for_each_netdev(net, dev) {
 852                 idev = __in6_dev_get(dev);
 853                 if (idev) {
 854                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 855
 856                         idev->cnf.ignore_routes_with_linkdown = newf;
 857                         if (changed)
 858                                 inet6_netconf_notify_devconf(dev_net(dev),
 859                                                              RTM_NEWNETCONF,
 860                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 861                                                              dev->ifindex,
 862                                                              &idev->cnf);
 863                 }
 864         }
 865 }
 866
 867 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 868 {
 869         struct net *net;
 870         int old;
 871
 872         if (!rtnl_trylock())
 873                 return restart_syscall();
 874
 875         net = (struct net *)table->extra2;
 876         old = *p;
 877         *p = newf;
 878
 879         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 880                 if ((!newf) ^ (!old))
 881                         inet6_netconf_notify_devconf(net,
 882                                                      RTM_NEWNETCONF,
 883                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 884                                                      NETCONFA_IFINDEX_DEFAULT,
 885                                                      net->ipv6.devconf_dflt);
 886                 rtnl_unlock();
 887                 return 0;
 888         }
 889
 890         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 891                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 892                 addrconf_linkdown_change(net, newf);
 893                 if ((!newf) ^ (!old))
 894                         inet6_netconf_notify_devconf(net,
 895                                                      RTM_NEWNETCONF,
 896                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 897                                                      NETCONFA_IFINDEX_ALL,
 898                                                      net->ipv6.devconf_all);
 899         }
 900         rtnl_unlock();
 901
 902         return 1;
 903 }
 904
 905 #endif
 906
 907 /* Nobody refers to this ifaddr, destroy it */
 908 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 909 {
 910         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 911
 912 #ifdef NET_REFCNT_DEBUG
 913         pr_debug("%s\n", __func__);
 914 #endif
 915
 916         in6_dev_put(ifp->idev);
 917
 918         if (cancel_delayed_work(&ifp->dad_work))
 919                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 920                           ifp);
 921
 922         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 923                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 924                 return;
 925         }
 926         ip6_rt_put(ifp->rt);
 927
 928         kfree_rcu(ifp, rcu);
 929 }
 930
 931 static void
 932 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 933 {
 934         struct list_head *p;
 935         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 936
 937         /*
 938          * Each device address list is sorted in order of scope -
 939          * global before linklocal.
 940          */
 941         list_for_each(p, &idev->addr_list) {
 942                 struct inet6_ifaddr *ifa
 943                         = list_entry(p, struct inet6_ifaddr, if_list);
 944                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 945                         break;
 946         }
 947
 948         list_add_tail(&ifp->if_list, p);
 949 }
 950
 951 static u32 inet6_addr_hash(const struct in6_addr *addr)
 952 {
 953         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 954 }
 955
 956 /* On success it returns ifp with increased reference count */
 957
 958 static struct inet6_ifaddr *
 959 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 960               const struct in6_addr *peer_addr, int pfxlen,
 961               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 962 {
 963         struct net *net = dev_net(idev->dev);
 964         struct inet6_ifaddr *ifa = NULL;
 965         struct rt6_info *rt;
 966         struct in6_validator_info i6vi;
 967         unsigned int hash;
 968         int err = 0;
 969         int addr_type = ipv6_addr_type(addr);
 970
 971         if (addr_type == IPV6_ADDR_ANY ||
 972             addr_type & IPV6_ADDR_MULTICAST ||
 973             (!(idev->dev->flags & IFF_LOOPBACK) &&
 974              addr_type & IPV6_ADDR_LOOPBACK))
 975                 return ERR_PTR(-EADDRNOTAVAIL);
 976
 977         rcu_read_lock_bh();
 978
 979         in6_dev_hold(idev);
 980
 981         if (idev->dead) {
 982                 err = -ENODEV;                  /*XXX*/
 983                 goto out2;
 984         }
 985
 986         if (idev->cnf.disable_ipv6) {
 987                 err = -EACCES;
 988                 goto out2;
 989         }
 990
 991         i6vi.i6vi_addr = *addr;
 992         i6vi.i6vi_dev = idev;
 993         rcu_read_unlock_bh();
 994
 995         err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
 996
 997         rcu_read_lock_bh();
 998         err = notifier_to_errno(err);
 999         if (err)
1000                 goto out2;
1001
1002         spin_lock(&addrconf_hash_lock);
1003
1004         /* Ignore adding duplicate addresses on an interface */
1005         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
1006                 ADBG("ipv6_add_addr: already assigned\n");
1007                 err = -EEXIST;
1008                 goto out;
1009         }
1010
1011         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
1012
1013         if (!ifa) {
1014                 ADBG("ipv6_add_addr: malloc failed\n");
1015                 err = -ENOBUFS;
1016                 goto out;
1017         }
1018
1019         rt = addrconf_dst_alloc(idev, addr, false);
1020         if (IS_ERR(rt)) {
1021                 err = PTR_ERR(rt);
1022                 goto out;
1023         }
1024
1025         if (net->ipv6.devconf_all->disable_policy ||
1026             idev->cnf.disable_policy)
1027                 rt->dst.flags |= DST_NOPOLICY;
1028
1029         neigh_parms_data_state_setall(idev->nd_parms);
1030
1031         ifa->addr = *addr;
1032         if (peer_addr)
1033                 ifa->peer_addr = *peer_addr;
1034
1035         spin_lock_init(&ifa->lock);
1036         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1037         INIT_HLIST_NODE(&ifa->addr_lst);
1038         ifa->scope = scope;
1039         ifa->prefix_len = pfxlen;
1040         ifa->flags = flags;
1041         /* No need to add the TENTATIVE flag for addresses with NODAD */
1042         if (!(flags & IFA_F_NODAD))
1043                 ifa->flags |= IFA_F_TENTATIVE;
1044         ifa->valid_lft = valid_lft;
1045         ifa->prefered_lft = prefered_lft;
1046         ifa->cstamp = ifa->tstamp = jiffies;
1047         ifa->tokenized = false;
1048
1049         ifa->rt = rt;
1050
1051         ifa->idev = idev;
1052         /* For caller */
1053         refcount_set(&ifa->refcnt, 1);
1054
1055         /* Add to big hash table */
1056         hash = inet6_addr_hash(addr);
1057
1058         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1059         spin_unlock(&addrconf_hash_lock);
1060
1061         write_lock(&idev->lock);
1062         /* Add to inet6_dev unicast addr list. */
1063         ipv6_link_dev_addr(idev, ifa);
1064
1065         if (ifa->flags&IFA_F_TEMPORARY) {
1066                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1067                 in6_ifa_hold(ifa);
1068         }
1069
1070         in6_ifa_hold(ifa);
1071         write_unlock(&idev->lock);
1072 out2:
1073         rcu_read_unlock_bh();
1074
1075         if (likely(err == 0))
1076                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1077         else {
1078                 kfree(ifa);
1079                 in6_dev_put(idev);
1080                 ifa = ERR_PTR(err);
1081         }
1082
1083         return ifa;
1084 out:
1085         spin_unlock(&addrconf_hash_lock);
1086         goto out2;
1087 }
1088
1089 enum cleanup_prefix_rt_t {
1090         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1091         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1092         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1093 };
1094
1095 /*
1096  * Check, whether the prefix for ifp would still need a prefix route
1097  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1098  * constants.
1099  *
1100  * 1) we don't purge prefix if address was not permanent.
1101  *    prefix is managed by its own lifetime.
1102  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1103  * 3) if there are no addresses, delete prefix.
1104  * 4) if there are still other permanent address(es),
1105  *    corresponding prefix is still permanent.
1106  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1107  *    don't purge the prefix, assume user space is managing it.
1108  * 6) otherwise, update prefix lifetime to the
1109  *    longest valid lifetime among the corresponding
1110  *    addresses on the device.
1111  *    Note: subsequent RA will update lifetime.
1112  **/
1113 static enum cleanup_prefix_rt_t
1114 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1115 {
1116         struct inet6_ifaddr *ifa;
1117         struct inet6_dev *idev = ifp->idev;
1118         unsigned long lifetime;
1119         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1120
1121         *expires = jiffies;
1122
1123         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1124                 if (ifa == ifp)
1125                         continue;
1126                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1127                                        ifp->prefix_len))
1128                         continue;
1129                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1130                         return CLEANUP_PREFIX_RT_NOP;
1131
1132                 action = CLEANUP_PREFIX_RT_EXPIRE;
1133
1134                 spin_lock(&ifa->lock);
1135
1136                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1137                 /*
1138                  * Note: Because this address is
1139                  * not permanent, lifetime <
1140                  * LONG_MAX / HZ here.
1141                  */
1142                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1143                         *expires = ifa->tstamp + lifetime * HZ;
1144                 spin_unlock(&ifa->lock);
1145         }
1146
1147         return action;
1148 }
1149
1150 static void
1151 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1152 {
1153         struct rt6_info *rt;
1154
1155         rt = addrconf_get_prefix_route(&ifp->addr,
1156                                        ifp->prefix_len,
1157                                        ifp->idev->dev,
1158                                        0, RTF_GATEWAY | RTF_DEFAULT);
1159         if (rt) {
1160                 if (del_rt)
1161                         ip6_del_rt(rt);
1162                 else {
1163                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1164                                 rt6_set_expires(rt, expires);
1165                         ip6_rt_put(rt);
1166                 }
1167         }
1168 }
1169
1170
1171 /* This function wants to get referenced ifp and releases it before return */
1172
1173 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1174 {
1175         int state;
1176         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1177         unsigned long expires;
1178
1179         ASSERT_RTNL();
1180
1181         spin_lock_bh(&ifp->lock);
1182         state = ifp->state;
1183         ifp->state = INET6_IFADDR_STATE_DEAD;
1184         spin_unlock_bh(&ifp->lock);
1185
1186         if (state == INET6_IFADDR_STATE_DEAD)
1187                 goto out;
1188
1189         spin_lock_bh(&addrconf_hash_lock);
1190         hlist_del_init_rcu(&ifp->addr_lst);
1191         spin_unlock_bh(&addrconf_hash_lock);
1192
1193         write_lock_bh(&ifp->idev->lock);
1194
1195         if (ifp->flags&IFA_F_TEMPORARY) {
1196                 list_del(&ifp->tmp_list);
1197                 if (ifp->ifpub) {
1198                         in6_ifa_put(ifp->ifpub);
1199                         ifp->ifpub = NULL;
1200                 }
1201                 __in6_ifa_put(ifp);
1202         }
1203
1204         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1205                 action = check_cleanup_prefix_route(ifp, &expires);
1206
1207         list_del_init(&ifp->if_list);
1208         __in6_ifa_put(ifp);
1209
1210         write_unlock_bh(&ifp->idev->lock);
1211
1212         addrconf_del_dad_work(ifp);
1213
1214         ipv6_ifa_notify(RTM_DELADDR, ifp);
1215
1216         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1217
1218         if (action != CLEANUP_PREFIX_RT_NOP) {
1219                 cleanup_prefix_route(ifp, expires,
1220                         action == CLEANUP_PREFIX_RT_DEL);
1221         }
1222
1223         /* clean up prefsrc entries */
1224         rt6_remove_prefsrc(ifp);
1225 out:
1226         in6_ifa_put(ifp);
1227 }
1228
1229 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1230 {
1231         struct inet6_dev *idev = ifp->idev;
1232         struct in6_addr addr, *tmpaddr;
1233         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1234         unsigned long regen_advance;
1235         int tmp_plen;
1236         int ret = 0;
1237         u32 addr_flags;
1238         unsigned long now = jiffies;
1239         long max_desync_factor;
1240         s32 cnf_temp_preferred_lft;
1241
1242         write_lock_bh(&idev->lock);
1243         if (ift) {
1244                 spin_lock_bh(&ift->lock);
1245                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1246                 spin_unlock_bh(&ift->lock);
1247                 tmpaddr = &addr;
1248         } else {
1249                 tmpaddr = NULL;
1250         }
1251 retry:
1252         in6_dev_hold(idev);
1253         if (idev->cnf.use_tempaddr <= 0) {
1254                 write_unlock_bh(&idev->lock);
1255                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1256                 in6_dev_put(idev);
1257                 ret = -1;
1258                 goto out;
1259         }
1260         spin_lock_bh(&ifp->lock);
1261         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1262                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1263                 spin_unlock_bh(&ifp->lock);
1264                 write_unlock_bh(&idev->lock);
1265                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1266                         __func__);
1267                 in6_dev_put(idev);
1268                 ret = -1;
1269                 goto out;
1270         }
1271         in6_ifa_hold(ifp);
1272         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1273         ipv6_try_regen_rndid(idev, tmpaddr);
1274         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1275         age = (now - ifp->tstamp) / HZ;
1276
1277         regen_advance = idev->cnf.regen_max_retry *
1278                         idev->cnf.dad_transmits *
1279                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1280
1281         /* recalculate max_desync_factor each time and update
1282          * idev->desync_factor if it's larger
1283          */
1284         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1285         max_desync_factor = min_t(__u32,
1286                                   idev->cnf.max_desync_factor,
1287                                   cnf_temp_preferred_lft - regen_advance);
1288
1289         if (unlikely(idev->desync_factor > max_desync_factor)) {
1290                 if (max_desync_factor > 0) {
1291                         get_random_bytes(&idev->desync_factor,
1292                                          sizeof(idev->desync_factor));
1293                         idev->desync_factor %= max_desync_factor;
1294                 } else {
1295                         idev->desync_factor = 0;
1296                 }
1297         }
1298
1299         tmp_valid_lft = min_t(__u32,
1300                               ifp->valid_lft,
1301                               idev->cnf.temp_valid_lft + age);
1302         tmp_prefered_lft = cnf_temp_preferred_lft + age -
1303                             idev->desync_factor;
1304         tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1305         tmp_plen = ifp->prefix_len;
1306         tmp_tstamp = ifp->tstamp;
1307         spin_unlock_bh(&ifp->lock);
1308
1309         write_unlock_bh(&idev->lock);
1310
1311         /* A temporary address is created only if this calculated Preferred
1312          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1313          * an implementation must not create a temporary address with a zero
1314          * Preferred Lifetime.
1315          * Use age calculation as in addrconf_verify to avoid unnecessary
1316          * temporary addresses being generated.
1317          */
1318         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1319         if (tmp_prefered_lft <= regen_advance + age) {
1320                 in6_ifa_put(ifp);
1321                 in6_dev_put(idev);
1322                 ret = -1;
1323                 goto out;
1324         }
1325
1326         addr_flags = IFA_F_TEMPORARY;
1327         /* set in addrconf_prefix_rcv() */
1328         if (ifp->flags & IFA_F_OPTIMISTIC)
1329                 addr_flags |= IFA_F_OPTIMISTIC;
1330
1331         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1332                             ipv6_addr_scope(&addr), addr_flags,
1333                             tmp_valid_lft, tmp_prefered_lft);
1334         if (IS_ERR(ift)) {
1335                 in6_ifa_put(ifp);
1336                 in6_dev_put(idev);
1337                 pr_info("%s: retry temporary address regeneration\n", __func__);
1338                 tmpaddr = &addr;
1339                 write_lock_bh(&idev->lock);
1340                 goto retry;
1341         }
1342
1343         spin_lock_bh(&ift->lock);
1344         ift->ifpub = ifp;
1345         ift->cstamp = now;
1346         ift->tstamp = tmp_tstamp;
1347         spin_unlock_bh(&ift->lock);
1348
1349         addrconf_dad_start(ift);
1350         in6_ifa_put(ift);
1351         in6_dev_put(idev);
1352 out:
1353         return ret;
1354 }
1355
1356 /*
1357  *      Choose an appropriate source address (RFC3484)
1358  */
1359 enum {
1360         IPV6_SADDR_RULE_INIT = 0,
1361         IPV6_SADDR_RULE_LOCAL,
1362         IPV6_SADDR_RULE_SCOPE,
1363         IPV6_SADDR_RULE_PREFERRED,
1364 #ifdef CONFIG_IPV6_MIP6
1365         IPV6_SADDR_RULE_HOA,
1366 #endif
1367         IPV6_SADDR_RULE_OIF,
1368         IPV6_SADDR_RULE_LABEL,
1369         IPV6_SADDR_RULE_PRIVACY,
1370         IPV6_SADDR_RULE_ORCHID,
1371         IPV6_SADDR_RULE_PREFIX,
1372 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1373         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1374 #endif
1375         IPV6_SADDR_RULE_MAX
1376 };
1377
1378 struct ipv6_saddr_score {
1379         int                     rule;
1380         int                     addr_type;
1381         struct inet6_ifaddr     *ifa;
1382         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1383         int                     scopedist;
1384         int                     matchlen;
1385 };
1386
1387 struct ipv6_saddr_dst {
1388         const struct in6_addr *addr;
1389         int ifindex;
1390         int scope;
1391         int label;
1392         unsigned int prefs;
1393 };
1394
1395 static inline int ipv6_saddr_preferred(int type)
1396 {
1397         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1398                 return 1;
1399         return 0;
1400 }
1401
1402 static bool ipv6_use_optimistic_addr(struct net *net,
1403                                      struct inet6_dev *idev)
1404 {
1405 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1406         if (!idev)
1407                 return false;
1408         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1409                 return false;
1410         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1411                 return false;
1412
1413         return true;
1414 #else
1415         return false;
1416 #endif
1417 }
1418
1419 static int ipv6_get_saddr_eval(struct net *net,
1420                                struct ipv6_saddr_score *score,
1421                                struct ipv6_saddr_dst *dst,
1422                                int i)
1423 {
1424         int ret;
1425
1426         if (i <= score->rule) {
1427                 switch (i) {
1428                 case IPV6_SADDR_RULE_SCOPE:
1429                         ret = score->scopedist;
1430                         break;
1431                 case IPV6_SADDR_RULE_PREFIX:
1432                         ret = score->matchlen;
1433                         break;
1434                 default:
1435                         ret = !!test_bit(i, score->scorebits);
1436                 }
1437                 goto out;
1438         }
1439
1440         switch (i) {
1441         case IPV6_SADDR_RULE_INIT:
1442                 /* Rule 0: remember if hiscore is not ready yet */
1443                 ret = !!score->ifa;
1444                 break;
1445         case IPV6_SADDR_RULE_LOCAL:
1446                 /* Rule 1: Prefer same address */
1447                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1448                 break;
1449         case IPV6_SADDR_RULE_SCOPE:
1450                 /* Rule 2: Prefer appropriate scope
1451                  *
1452                  *      ret
1453                  *       ^
1454                  *    -1 |  d 15
1455                  *    ---+--+-+---> scope
1456                  *       |
1457                  *       |             d is scope of the destination.
1458                  *  B-d  |  \
1459                  *       |   \      <- smaller scope is better if
1460                  *  B-15 |    \        if scope is enough for destination.
1461                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1462                  * d-C-1 | /
1463                  *       |/         <- greater is better
1464                  *   -C  /             if scope is not enough for destination.
1465                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1466                  *
1467                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1468                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1469                  * Assume B = 0 and we get C > 29.
1470                  */
1471                 ret = __ipv6_addr_src_scope(score->addr_type);
1472                 if (ret >= dst->scope)
1473                         ret = -ret;
1474                 else
1475                         ret -= 128;     /* 30 is enough */
1476                 score->scopedist = ret;
1477                 break;
1478         case IPV6_SADDR_RULE_PREFERRED:
1479             {
1480                 /* Rule 3: Avoid deprecated and optimistic addresses */
1481                 u8 avoid = IFA_F_DEPRECATED;
1482
1483                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1484                         avoid |= IFA_F_OPTIMISTIC;
1485                 ret = ipv6_saddr_preferred(score->addr_type) ||
1486                       !(score->ifa->flags & avoid);
1487                 break;
1488             }
1489 #ifdef CONFIG_IPV6_MIP6
1490         case IPV6_SADDR_RULE_HOA:
1491             {
1492                 /* Rule 4: Prefer home address */
1493                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1494                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1495                 break;
1496             }
1497 #endif
1498         case IPV6_SADDR_RULE_OIF:
1499                 /* Rule 5: Prefer outgoing interface */
1500                 ret = (!dst->ifindex ||
1501                        dst->ifindex == score->ifa->idev->dev->ifindex);
1502                 break;
1503         case IPV6_SADDR_RULE_LABEL:
1504                 /* Rule 6: Prefer matching label */
1505                 ret = ipv6_addr_label(net,
1506                                       &score->ifa->addr, score->addr_type,
1507                                       score->ifa->idev->dev->ifindex) == dst->label;
1508                 break;
1509         case IPV6_SADDR_RULE_PRIVACY:
1510             {
1511                 /* Rule 7: Prefer public address
1512                  * Note: prefer temporary address if use_tempaddr >= 2
1513                  */
1514                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1515                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1516                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1517                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1518                 break;
1519             }
1520         case IPV6_SADDR_RULE_ORCHID:
1521                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1522                  *          non-ORCHID vs non-ORCHID
1523                  */
1524                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1525                         ipv6_addr_orchid(dst->addr));
1526                 break;
1527         case IPV6_SADDR_RULE_PREFIX:
1528                 /* Rule 8: Use longest matching prefix */
1529                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1530                 if (ret > score->ifa->prefix_len)
1531                         ret = score->ifa->prefix_len;
1532                 score->matchlen = ret;
1533                 break;
1534 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1535         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1536                 /* Optimistic addresses still have lower precedence than other
1537                  * preferred addresses.
1538                  */
1539                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1540                 break;
1541 #endif
1542         default:
1543                 ret = 0;
1544         }
1545
1546         if (ret)
1547                 __set_bit(i, score->scorebits);
1548         score->rule = i;
1549 out:
1550         return ret;
1551 }
1552
1553 static int __ipv6_dev_get_saddr(struct net *net,
1554                                 struct ipv6_saddr_dst *dst,
1555                                 struct inet6_dev *idev,
1556                                 struct ipv6_saddr_score *scores,
1557                                 int hiscore_idx)
1558 {
1559         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1560
1561         read_lock_bh(&idev->lock);
1562         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1563                 int i;
1564
1565                 /*
1566                  * - Tentative Address (RFC2462 section 5.4)
1567                  *  - A tentative address is not considered
1568                  *    "assigned to an interface" in the traditional
1569                  *    sense, unless it is also flagged as optimistic.
1570                  * - Candidate Source Address (section 4)
1571                  *  - In any case, anycast addresses, multicast
1572                  *    addresses, and the unspecified address MUST
1573                  *    NOT be included in a candidate set.
1574                  */
1575                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1576                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1577                         continue;
1578
1579                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1580
1581                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1582                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1583                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1584                                             idev->dev->name);
1585                         continue;
1586                 }
1587
1588                 score->rule = -1;
1589                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1590
1591                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1592                         int minihiscore, miniscore;
1593
1594                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1595                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1596
1597                         if (minihiscore > miniscore) {
1598                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1599                                     score->scopedist > 0) {
1600                                         /*
1601                                          * special case:
1602                                          * each remaining entry
1603                                          * has too small (not enough)
1604                                          * scope, because ifa entries
1605                                          * are sorted by their scope
1606                                          * values.
1607                                          */
1608                                         goto out;
1609                                 }
1610                                 break;
1611                         } else if (minihiscore < miniscore) {
1612                                 if (hiscore->ifa)
1613                                         in6_ifa_put(hiscore->ifa);
1614
1615                                 in6_ifa_hold(score->ifa);
1616
1617                                 swap(hiscore, score);
1618                                 hiscore_idx = 1 - hiscore_idx;
1619
1620                                 /* restore our iterator */
1621                                 score->ifa = hiscore->ifa;
1622
1623                                 break;
1624                         }
1625                 }
1626         }
1627 out:
1628         read_unlock_bh(&idev->lock);
1629         return hiscore_idx;
1630 }
1631
1632 static int ipv6_get_saddr_master(struct net *net,
1633                                  const struct net_device *dst_dev,
1634                                  const struct net_device *master,
1635                                  struct ipv6_saddr_dst *dst,
1636                                  struct ipv6_saddr_score *scores,
1637                                  int hiscore_idx)
1638 {
1639         struct inet6_dev *idev;
1640
1641         idev = __in6_dev_get(dst_dev);
1642         if (idev)
1643                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1644                                                    scores, hiscore_idx);
1645
1646         idev = __in6_dev_get(master);
1647         if (idev)
1648                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1649                                                    scores, hiscore_idx);
1650
1651         return hiscore_idx;
1652 }
1653
1654 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1655                        const struct in6_addr *daddr, unsigned int prefs,
1656                        struct in6_addr *saddr)
1657 {
1658         struct ipv6_saddr_score scores[2], *hiscore;
1659         struct ipv6_saddr_dst dst;
1660         struct inet6_dev *idev;
1661         struct net_device *dev;
1662         int dst_type;
1663         bool use_oif_addr = false;
1664         int hiscore_idx = 0;
1665
1666         dst_type = __ipv6_addr_type(daddr);
1667         dst.addr = daddr;
1668         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1669         dst.scope = __ipv6_addr_src_scope(dst_type);
1670         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1671         dst.prefs = prefs;
1672
1673         scores[hiscore_idx].rule = -1;
1674         scores[hiscore_idx].ifa = NULL;
1675
1676         rcu_read_lock();
1677
1678         /* Candidate Source Address (section 4)
1679          *  - multicast and link-local destination address,
1680          *    the set of candidate source address MUST only
1681          *    include addresses assigned to interfaces
1682          *    belonging to the same link as the outgoing
1683          *    interface.
1684          * (- For site-local destination addresses, the
1685          *    set of candidate source addresses MUST only
1686          *    include addresses assigned to interfaces
1687          *    belonging to the same site as the outgoing
1688          *    interface.)
1689          *  - "It is RECOMMENDED that the candidate source addresses
1690          *    be the set of unicast addresses assigned to the
1691          *    interface that will be used to send to the destination
1692          *    (the 'outgoing' interface)." (RFC 6724)
1693          */
1694         if (dst_dev) {
1695                 idev = __in6_dev_get(dst_dev);
1696                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1697                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1698                     (idev && idev->cnf.use_oif_addrs_only)) {
1699                         use_oif_addr = true;
1700                 }
1701         }
1702
1703         if (use_oif_addr) {
1704                 if (idev)
1705                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1706         } else {
1707                 const struct net_device *master;
1708                 int master_idx = 0;
1709
1710                 /* if dst_dev exists and is enslaved to an L3 device, then
1711                  * prefer addresses from dst_dev and then the master over
1712                  * any other enslaved devices in the L3 domain.
1713                  */
1714                 master = l3mdev_master_dev_rcu(dst_dev);
1715                 if (master) {
1716                         master_idx = master->ifindex;
1717
1718                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1719                                                             master, &dst,
1720                                                             scores, hiscore_idx);
1721
1722                         if (scores[hiscore_idx].ifa)
1723                                 goto out;
1724                 }
1725
1726                 for_each_netdev_rcu(net, dev) {
1727                         /* only consider addresses on devices in the
1728                          * same L3 domain
1729                          */
1730                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1731                                 continue;
1732                         idev = __in6_dev_get(dev);
1733                         if (!idev)
1734                                 continue;
1735                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1736                 }
1737         }
1738
1739 out:
1740         rcu_read_unlock();
1741
1742         hiscore = &scores[hiscore_idx];
1743         if (!hiscore->ifa)
1744                 return -EADDRNOTAVAIL;
1745
1746         *saddr = hiscore->ifa->addr;
1747         in6_ifa_put(hiscore->ifa);
1748         return 0;
1749 }
1750 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1751
1752 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1753                       u32 banned_flags)
1754 {
1755         struct inet6_ifaddr *ifp;
1756         int err = -EADDRNOTAVAIL;
1757
1758         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1759                 if (ifp->scope > IFA_LINK)
1760                         break;
1761                 if (ifp->scope == IFA_LINK &&
1762                     !(ifp->flags & banned_flags)) {
1763                         *addr = ifp->addr;
1764                         err = 0;
1765                         break;
1766                 }
1767         }
1768         return err;
1769 }
1770
1771 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1772                     u32 banned_flags)
1773 {
1774         struct inet6_dev *idev;
1775         int err = -EADDRNOTAVAIL;
1776
1777         rcu_read_lock();
1778         idev = __in6_dev_get(dev);
1779         if (idev) {
1780                 read_lock_bh(&idev->lock);
1781                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1782                 read_unlock_bh(&idev->lock);
1783         }
1784         rcu_read_unlock();
1785         return err;
1786 }
1787
1788 static int ipv6_count_addresses(struct inet6_dev *idev)
1789 {
1790         int cnt = 0;
1791         struct inet6_ifaddr *ifp;
1792
1793         read_lock_bh(&idev->lock);
1794         list_for_each_entry(ifp, &idev->addr_list, if_list)
1795                 cnt++;
1796         read_unlock_bh(&idev->lock);
1797         return cnt;
1798 }
1799
1800 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1801                   const struct net_device *dev, int strict)
1802 {
1803         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1804 }
1805 EXPORT_SYMBOL(ipv6_chk_addr);
1806
1807 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1808                             const struct net_device *dev, int strict,
1809                             u32 banned_flags)
1810 {
1811         struct inet6_ifaddr *ifp;
1812         unsigned int hash = inet6_addr_hash(addr);
1813         u32 ifp_flags;
1814
1815         rcu_read_lock_bh();
1816         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1817                 if (!net_eq(dev_net(ifp->idev->dev), net))
1818                         continue;
1819                 /* Decouple optimistic from tentative for evaluation here.
1820                  * Ban optimistic addresses explicitly, when required.
1821                  */
1822                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1823                             ? (ifp->flags&~IFA_F_TENTATIVE)
1824                             : ifp->flags;
1825                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1826                     !(ifp_flags&banned_flags) &&
1827                     (!dev || ifp->idev->dev == dev ||
1828                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1829                         rcu_read_unlock_bh();
1830                         return 1;
1831                 }
1832         }
1833
1834         rcu_read_unlock_bh();
1835         return 0;
1836 }
1837 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1838
1839 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1840                                struct net_device *dev)
1841 {
1842         unsigned int hash = inet6_addr_hash(addr);
1843         struct inet6_ifaddr *ifp;
1844
1845         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1846                 if (!net_eq(dev_net(ifp->idev->dev), net))
1847                         continue;
1848                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1849                         if (!dev || ifp->idev->dev == dev)
1850                                 return true;
1851                 }
1852         }
1853         return false;
1854 }
1855
1856 /* Compares an address/prefix_len with addresses on device @dev.
1857  * If one is found it returns true.
1858  */
1859 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1860         const unsigned int prefix_len, struct net_device *dev)
1861 {
1862         struct inet6_dev *idev;
1863         struct inet6_ifaddr *ifa;
1864         bool ret = false;
1865
1866         rcu_read_lock();
1867         idev = __in6_dev_get(dev);
1868         if (idev) {
1869                 read_lock_bh(&idev->lock);
1870                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1871                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1872                         if (ret)
1873                                 break;
1874                 }
1875                 read_unlock_bh(&idev->lock);
1876         }
1877         rcu_read_unlock();
1878
1879         return ret;
1880 }
1881 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1882
1883 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1884 {
1885         struct inet6_dev *idev;
1886         struct inet6_ifaddr *ifa;
1887         int     onlink;
1888
1889         onlink = 0;
1890         rcu_read_lock();
1891         idev = __in6_dev_get(dev);
1892         if (idev) {
1893                 read_lock_bh(&idev->lock);
1894                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1895                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1896                                                    ifa->prefix_len);
1897                         if (onlink)
1898                                 break;
1899                 }
1900                 read_unlock_bh(&idev->lock);
1901         }
1902         rcu_read_unlock();
1903         return onlink;
1904 }
1905 EXPORT_SYMBOL(ipv6_chk_prefix);
1906
1907 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1908                                      struct net_device *dev, int strict)
1909 {
1910         struct inet6_ifaddr *ifp, *result = NULL;
1911         unsigned int hash = inet6_addr_hash(addr);
1912
1913         rcu_read_lock_bh();
1914         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1915                 if (!net_eq(dev_net(ifp->idev->dev), net))
1916                         continue;
1917                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1918                         if (!dev || ifp->idev->dev == dev ||
1919                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1920                                 result = ifp;
1921                                 in6_ifa_hold(ifp);
1922                                 break;
1923                         }
1924                 }
1925         }
1926         rcu_read_unlock_bh();
1927
1928         return result;
1929 }
1930
1931 /* Gets referenced address, destroys ifaddr */
1932
1933 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1934 {
1935         if (dad_failed)
1936                 ifp->flags |= IFA_F_DADFAILED;
1937
1938         if (ifp->flags&IFA_F_TEMPORARY) {
1939                 struct inet6_ifaddr *ifpub;
1940                 spin_lock_bh(&ifp->lock);
1941                 ifpub = ifp->ifpub;
1942                 if (ifpub) {
1943                         in6_ifa_hold(ifpub);
1944                         spin_unlock_bh(&ifp->lock);
1945                         ipv6_create_tempaddr(ifpub, ifp);
1946                         in6_ifa_put(ifpub);
1947                 } else {
1948                         spin_unlock_bh(&ifp->lock);
1949                 }
1950                 ipv6_del_addr(ifp);
1951         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1952                 spin_lock_bh(&ifp->lock);
1953                 addrconf_del_dad_work(ifp);
1954                 ifp->flags |= IFA_F_TENTATIVE;
1955                 spin_unlock_bh(&ifp->lock);
1956                 if (dad_failed)
1957                         ipv6_ifa_notify(0, ifp);
1958                 in6_ifa_put(ifp);
1959         } else {
1960                 ipv6_del_addr(ifp);
1961         }
1962 }
1963
1964 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1965 {
1966         int err = -ENOENT;
1967
1968         spin_lock_bh(&ifp->lock);
1969         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1970                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1971                 err = 0;
1972         }
1973         spin_unlock_bh(&ifp->lock);
1974
1975         return err;
1976 }
1977
1978 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1979 {
1980         struct inet6_dev *idev = ifp->idev;
1981         struct net *net = dev_net(ifp->idev->dev);
1982
1983         if (addrconf_dad_end(ifp)) {
1984                 in6_ifa_put(ifp);
1985                 return;
1986         }
1987
1988         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1989                              ifp->idev->dev->name, &ifp->addr);
1990
1991         spin_lock_bh(&ifp->lock);
1992
1993         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1994                 int scope = ifp->scope;
1995                 u32 flags = ifp->flags;
1996                 struct in6_addr new_addr;
1997                 struct inet6_ifaddr *ifp2;
1998                 u32 valid_lft, preferred_lft;
1999                 int pfxlen = ifp->prefix_len;
2000                 int retries = ifp->stable_privacy_retry + 1;
2001
2002                 if (retries > net->ipv6.sysctl.idgen_retries) {
2003                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2004                                              ifp->idev->dev->name);
2005                         goto errdad;
2006                 }
2007
2008                 new_addr = ifp->addr;
2009                 if (ipv6_generate_stable_address(&new_addr, retries,
2010                                                  idev))
2011                         goto errdad;
2012
2013                 valid_lft = ifp->valid_lft;
2014                 preferred_lft = ifp->prefered_lft;
2015
2016                 spin_unlock_bh(&ifp->lock);
2017
2018                 if (idev->cnf.max_addresses &&
2019                     ipv6_count_addresses(idev) >=
2020                     idev->cnf.max_addresses)
2021                         goto lock_errdad;
2022
2023                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2024                                      ifp->idev->dev->name);
2025
2026                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2027                                      scope, flags, valid_lft,
2028                                      preferred_lft);
2029                 if (IS_ERR(ifp2))
2030                         goto lock_errdad;
2031
2032                 spin_lock_bh(&ifp2->lock);
2033                 ifp2->stable_privacy_retry = retries;
2034                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2035                 spin_unlock_bh(&ifp2->lock);
2036
2037                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2038                 in6_ifa_put(ifp2);
2039 lock_errdad:
2040                 spin_lock_bh(&ifp->lock);
2041         }
2042
2043 errdad:
2044         /* transition from _POSTDAD to _ERRDAD */
2045         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2046         spin_unlock_bh(&ifp->lock);
2047
2048         addrconf_mod_dad_work(ifp, 0);
2049         in6_ifa_put(ifp);
2050 }
2051
2052 /* Join to solicited addr multicast group.
2053  * caller must hold RTNL */
2054 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2055 {
2056         struct in6_addr maddr;
2057
2058         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2059                 return;
2060
2061         addrconf_addr_solict_mult(addr, &maddr);
2062         ipv6_dev_mc_inc(dev, &maddr);
2063 }
2064
2065 /* caller must hold RTNL */
2066 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2067 {
2068         struct in6_addr maddr;
2069
2070         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2071                 return;
2072
2073         addrconf_addr_solict_mult(addr, &maddr);
2074         __ipv6_dev_mc_dec(idev, &maddr);
2075 }
2076
2077 /* caller must hold RTNL */
2078 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2079 {
2080         struct in6_addr addr;
2081
2082         if (ifp->prefix_len >= 127) /* RFC 6164 */
2083                 return;
2084         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2085         if (ipv6_addr_any(&addr))
2086                 return;
2087         __ipv6_dev_ac_inc(ifp->idev, &addr);
2088 }
2089
2090 /* caller must hold RTNL */
2091 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2092 {
2093         struct in6_addr addr;
2094
2095         if (ifp->prefix_len >= 127) /* RFC 6164 */
2096                 return;
2097         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2098         if (ipv6_addr_any(&addr))
2099                 return;
2100         __ipv6_dev_ac_dec(ifp->idev, &addr);
2101 }
2102
2103 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2104 {
2105         switch (dev->addr_len) {
2106         case ETH_ALEN:
2107                 memcpy(eui, dev->dev_addr, 3);
2108                 eui[3] = 0xFF;
2109                 eui[4] = 0xFE;
2110                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2111                 break;
2112         case EUI64_ADDR_LEN:
2113                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2114                 eui[0] ^= 2;
2115                 break;
2116         default:
2117                 return -1;
2118         }
2119
2120         return 0;
2121 }
2122
2123 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2124 {
2125         union fwnet_hwaddr *ha;
2126
2127         if (dev->addr_len != FWNET_ALEN)
2128                 return -1;
2129
2130         ha = (union fwnet_hwaddr *)dev->dev_addr;
2131
2132         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2133         eui[0] ^= 2;
2134         return 0;
2135 }
2136
2137 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2138 {
2139         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2140         if (dev->addr_len != ARCNET_ALEN)
2141                 return -1;
2142         memset(eui, 0, 7);
2143         eui[7] = *(u8 *)dev->dev_addr;
2144         return 0;
2145 }
2146
2147 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2148 {
2149         if (dev->addr_len != INFINIBAND_ALEN)
2150                 return -1;
2151         memcpy(eui, dev->dev_addr + 12, 8);
2152         eui[0] |= 2;
2153         return 0;
2154 }
2155
2156 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2157 {
2158         if (addr == 0)
2159                 return -1;
2160         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2161                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2162                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2163                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2164                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2165                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2166         eui[1] = 0;
2167         eui[2] = 0x5E;
2168         eui[3] = 0xFE;
2169         memcpy(eui + 4, &addr, 4);
2170         return 0;
2171 }
2172
2173 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2174 {
2175         if (dev->priv_flags & IFF_ISATAP)
2176                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2177         return -1;
2178 }
2179
2180 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2181 {
2182         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2183 }
2184
2185 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2186 {
2187         memcpy(eui, dev->perm_addr, 3);
2188         memcpy(eui + 5, dev->perm_addr + 3, 3);
2189         eui[3] = 0xFF;
2190         eui[4] = 0xFE;
2191         eui[0] ^= 2;
2192         return 0;
2193 }
2194
2195 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2196 {
2197         switch (dev->type) {
2198         case ARPHRD_ETHER:
2199         case ARPHRD_FDDI:
2200                 return addrconf_ifid_eui48(eui, dev);
2201         case ARPHRD_ARCNET:
2202                 return addrconf_ifid_arcnet(eui, dev);
2203         case ARPHRD_INFINIBAND:
2204                 return addrconf_ifid_infiniband(eui, dev);
2205         case ARPHRD_SIT:
2206                 return addrconf_ifid_sit(eui, dev);
2207         case ARPHRD_IPGRE:
2208         case ARPHRD_TUNNEL:
2209                 return addrconf_ifid_gre(eui, dev);
2210         case ARPHRD_6LOWPAN:
2211                 return addrconf_ifid_6lowpan(eui, dev);
2212         case ARPHRD_IEEE1394:
2213                 return addrconf_ifid_ieee1394(eui, dev);
2214         case ARPHRD_TUNNEL6:
2215         case ARPHRD_IP6GRE:
2216                 return addrconf_ifid_ip6tnl(eui, dev);
2217         }
2218         return -1;
2219 }
2220
2221 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2222 {
2223         int err = -1;
2224         struct inet6_ifaddr *ifp;
2225
2226         read_lock_bh(&idev->lock);
2227         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2228                 if (ifp->scope > IFA_LINK)
2229                         break;
2230                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2231                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2232                         err = 0;
2233                         break;
2234                 }
2235         }
2236         read_unlock_bh(&idev->lock);
2237         return err;
2238 }
2239
2240 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2241 static void ipv6_regen_rndid(struct inet6_dev *idev)
2242 {
2243 regen:
2244         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2245         idev->rndid[0] &= ~0x02;
2246
2247         /*
2248          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2249          * check if generated address is not inappropriate
2250          *
2251          *  - Reserved subnet anycast (RFC 2526)
2252          *      11111101 11....11 1xxxxxxx
2253          *  - ISATAP (RFC4214) 6.1
2254          *      00-00-5E-FE-xx-xx-xx-xx
2255          *  - value 0
2256          *  - XXX: already assigned to an address on the device
2257          */
2258         if (idev->rndid[0] == 0xfd &&
2259             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2260             (idev->rndid[7]&0x80))
2261                 goto regen;
2262         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2263                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2264                         goto regen;
2265                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2266                         goto regen;
2267         }
2268 }
2269
2270 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2271 {
2272         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2273                 ipv6_regen_rndid(idev);
2274 }
2275
2276 /*
2277  *      Add prefix route.
2278  */
2279
2280 static void
2281 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2282                       unsigned long expires, u32 flags)
2283 {
2284         struct fib6_config cfg = {
2285                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2286                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2287                 .fc_ifindex = dev->ifindex,
2288                 .fc_expires = expires,
2289                 .fc_dst_len = plen,
2290                 .fc_flags = RTF_UP | flags,
2291                 .fc_nlinfo.nl_net = dev_net(dev),
2292                 .fc_protocol = RTPROT_KERNEL,
2293         };
2294
2295         cfg.fc_dst = *pfx;
2296
2297         /* Prevent useless cloning on PtP SIT.
2298            This thing is done here expecting that the whole
2299            class of non-broadcast devices need not cloning.
2300          */
2301 #if IS_ENABLED(CONFIG_IPV6_SIT)
2302         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2303                 cfg.fc_flags |= RTF_NONEXTHOP;
2304 #endif
2305
2306         ip6_route_add(&cfg, NULL);
2307 }
2308
2309
2310 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2311                                                   int plen,
2312                                                   const struct net_device *dev,
2313                                                   u32 flags, u32 noflags)
2314 {
2315         struct fib6_node *fn;
2316         struct rt6_info *rt = NULL;
2317         struct fib6_table *table;
2318         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2319
2320         table = fib6_get_table(dev_net(dev), tb_id);
2321         if (!table)
2322                 return NULL;
2323
2324         read_lock_bh(&table->tb6_lock);
2325         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2326         if (!fn)
2327                 goto out;
2328
2329         noflags |= RTF_CACHE;
2330         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2331                 if (rt->dst.dev->ifindex != dev->ifindex)
2332                         continue;
2333                 if ((rt->rt6i_flags & flags) != flags)
2334                         continue;
2335                 if ((rt->rt6i_flags & noflags) != 0)
2336                         continue;
2337                 dst_hold(&rt->dst);
2338                 break;
2339         }
2340 out:
2341         read_unlock_bh(&table->tb6_lock);
2342         return rt;
2343 }
2344
2345
2346 /* Create "default" multicast route to the interface */
2347
2348 static void addrconf_add_mroute(struct net_device *dev)
2349 {
2350         struct fib6_config cfg = {
2351                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2352                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2353                 .fc_ifindex = dev->ifindex,
2354                 .fc_dst_len = 8,
2355                 .fc_flags = RTF_UP,
2356                 .fc_nlinfo.nl_net = dev_net(dev),
2357         };
2358
2359         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2360
2361         ip6_route_add(&cfg, NULL);
2362 }
2363
2364 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2365 {
2366         struct inet6_dev *idev;
2367
2368         ASSERT_RTNL();
2369
2370         idev = ipv6_find_idev(dev);
2371         if (!idev)
2372                 return ERR_PTR(-ENOBUFS);
2373
2374         if (idev->cnf.disable_ipv6)
2375                 return ERR_PTR(-EACCES);
2376
2377         /* Add default multicast route */
2378         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2379                 addrconf_add_mroute(dev);
2380
2381         return idev;
2382 }
2383
2384 static void manage_tempaddrs(struct inet6_dev *idev,
2385                              struct inet6_ifaddr *ifp,
2386                              __u32 valid_lft, __u32 prefered_lft,
2387                              bool create, unsigned long now)
2388 {
2389         u32 flags;
2390         struct inet6_ifaddr *ift;
2391
2392         read_lock_bh(&idev->lock);
2393         /* update all temporary addresses in the list */
2394         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2395                 int age, max_valid, max_prefered;
2396
2397                 if (ifp != ift->ifpub)
2398                         continue;
2399
2400                 /* RFC 4941 section 3.3:
2401                  * If a received option will extend the lifetime of a public
2402                  * address, the lifetimes of temporary addresses should
2403                  * be extended, subject to the overall constraint that no
2404                  * temporary addresses should ever remain "valid" or "preferred"
2405                  * for a time longer than (TEMP_VALID_LIFETIME) or
2406                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2407                  */
2408                 age = (now - ift->cstamp) / HZ;
2409                 max_valid = idev->cnf.temp_valid_lft - age;
2410                 if (max_valid < 0)
2411                         max_valid = 0;
2412
2413                 max_prefered = idev->cnf.temp_prefered_lft -
2414                                idev->desync_factor - age;
2415                 if (max_prefered < 0)
2416                         max_prefered = 0;
2417
2418                 if (valid_lft > max_valid)
2419                         valid_lft = max_valid;
2420
2421                 if (prefered_lft > max_prefered)
2422                         prefered_lft = max_prefered;
2423
2424                 spin_lock(&ift->lock);
2425                 flags = ift->flags;
2426                 ift->valid_lft = valid_lft;
2427                 ift->prefered_lft = prefered_lft;
2428                 ift->tstamp = now;
2429                 if (prefered_lft > 0)
2430                         ift->flags &= ~IFA_F_DEPRECATED;
2431
2432                 spin_unlock(&ift->lock);
2433                 if (!(flags&IFA_F_TENTATIVE))
2434                         ipv6_ifa_notify(0, ift);
2435         }
2436
2437         if ((create || list_empty(&idev->tempaddr_list)) &&
2438             idev->cnf.use_tempaddr > 0) {
2439                 /* When a new public address is created as described
2440                  * in [ADDRCONF], also create a new temporary address.
2441                  * Also create a temporary address if it's enabled but
2442                  * no temporary address currently exists.
2443                  */
2444                 read_unlock_bh(&idev->lock);
2445                 ipv6_create_tempaddr(ifp, NULL);
2446         } else {
2447                 read_unlock_bh(&idev->lock);
2448         }
2449 }
2450
2451 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2452 {
2453         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2454                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2455 }
2456
2457 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2458                                  const struct prefix_info *pinfo,
2459                                  struct inet6_dev *in6_dev,
2460                                  const struct in6_addr *addr, int addr_type,
2461                                  u32 addr_flags, bool sllao, bool tokenized,
2462                                  __u32 valid_lft, u32 prefered_lft)
2463 {
2464         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2465         int create = 0, update_lft = 0;
2466
2467         if (!ifp && valid_lft) {
2468                 int max_addresses = in6_dev->cnf.max_addresses;
2469
2470 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2471                 if ((net->ipv6.devconf_all->optimistic_dad ||
2472                      in6_dev->cnf.optimistic_dad) &&
2473                     !net->ipv6.devconf_all->forwarding && sllao)
2474                         addr_flags |= IFA_F_OPTIMISTIC;
2475 #endif
2476
2477                 /* Do not allow to create too much of autoconfigured
2478                  * addresses; this would be too easy way to crash kernel.
2479                  */
2480                 if (!max_addresses ||
2481                     ipv6_count_addresses(in6_dev) < max_addresses)
2482                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2483                                             pinfo->prefix_len,
2484                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2485                                             addr_flags, valid_lft,
2486                                             prefered_lft);
2487
2488                 if (IS_ERR_OR_NULL(ifp))
2489                         return -1;
2490
2491                 update_lft = 0;
2492                 create = 1;
2493                 spin_lock_bh(&ifp->lock);
2494                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2495                 ifp->cstamp = jiffies;
2496                 ifp->tokenized = tokenized;
2497                 spin_unlock_bh(&ifp->lock);
2498                 addrconf_dad_start(ifp);
2499         }
2500
2501         if (ifp) {
2502                 u32 flags;
2503                 unsigned long now;
2504                 u32 stored_lft;
2505
2506                 /* update lifetime (RFC2462 5.5.3 e) */
2507                 spin_lock_bh(&ifp->lock);
2508                 now = jiffies;
2509                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2510                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2511                 else
2512                         stored_lft = 0;
2513                 if (!update_lft && !create && stored_lft) {
2514                         const u32 minimum_lft = min_t(u32,
2515                                 stored_lft, MIN_VALID_LIFETIME);
2516                         valid_lft = max(valid_lft, minimum_lft);
2517
2518                         /* RFC4862 Section 5.5.3e:
2519                          * "Note that the preferred lifetime of the
2520                          *  corresponding address is always reset to
2521                          *  the Preferred Lifetime in the received
2522                          *  Prefix Information option, regardless of
2523                          *  whether the valid lifetime is also reset or
2524                          *  ignored."
2525                          *
2526                          * So we should always update prefered_lft here.
2527                          */
2528                         update_lft = 1;
2529                 }
2530
2531                 if (update_lft) {
2532                         ifp->valid_lft = valid_lft;
2533                         ifp->prefered_lft = prefered_lft;
2534                         ifp->tstamp = now;
2535                         flags = ifp->flags;
2536                         ifp->flags &= ~IFA_F_DEPRECATED;
2537                         spin_unlock_bh(&ifp->lock);
2538
2539                         if (!(flags&IFA_F_TENTATIVE))
2540                                 ipv6_ifa_notify(0, ifp);
2541                 } else
2542                         spin_unlock_bh(&ifp->lock);
2543
2544                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2545                                  create, now);
2546
2547                 in6_ifa_put(ifp);
2548                 addrconf_verify();
2549         }
2550
2551         return 0;
2552 }
2553 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2554
2555 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2556 {
2557         struct prefix_info *pinfo;
2558         __u32 valid_lft;
2559         __u32 prefered_lft;
2560         int addr_type, err;
2561         u32 addr_flags = 0;
2562         struct inet6_dev *in6_dev;
2563         struct net *net = dev_net(dev);
2564
2565         pinfo = (struct prefix_info *) opt;
2566
2567         if (len < sizeof(struct prefix_info)) {
2568                 ADBG("addrconf: prefix option too short\n");
2569                 return;
2570         }
2571
2572         /*
2573          *      Validation checks ([ADDRCONF], page 19)
2574          */
2575
2576         addr_type = ipv6_addr_type(&pinfo->prefix);
2577
2578         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2579                 return;
2580
2581         valid_lft = ntohl(pinfo->valid);
2582         prefered_lft = ntohl(pinfo->prefered);
2583
2584         if (prefered_lft > valid_lft) {
2585                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2586                 return;
2587         }
2588
2589         in6_dev = in6_dev_get(dev);
2590
2591         if (!in6_dev) {
2592                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2593                                     dev->name);
2594                 return;
2595         }
2596
2597         /*
2598          *      Two things going on here:
2599          *      1) Add routes for on-link prefixes
2600          *      2) Configure prefixes with the auto flag set
2601          */
2602
2603         if (pinfo->onlink) {
2604                 struct rt6_info *rt;
2605                 unsigned long rt_expires;
2606
2607                 /* Avoid arithmetic overflow. Really, we could
2608                  * save rt_expires in seconds, likely valid_lft,
2609                  * but it would require division in fib gc, that it
2610                  * not good.
2611                  */
2612                 if (HZ > USER_HZ)
2613                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2614                 else
2615                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2616
2617                 if (addrconf_finite_timeout(rt_expires))
2618                         rt_expires *= HZ;
2619
2620                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2621                                                pinfo->prefix_len,
2622                                                dev,
2623                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2624                                                RTF_GATEWAY | RTF_DEFAULT);
2625
2626                 if (rt) {
2627                         /* Autoconf prefix route */
2628                         if (valid_lft == 0) {
2629                                 ip6_del_rt(rt);
2630                                 rt = NULL;
2631                         } else if (addrconf_finite_timeout(rt_expires)) {
2632                                 /* not infinity */
2633                                 rt6_set_expires(rt, jiffies + rt_expires);
2634                         } else {
2635                                 rt6_clean_expires(rt);
2636                         }
2637                 } else if (valid_lft) {
2638                         clock_t expires = 0;
2639                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2640                         if (addrconf_finite_timeout(rt_expires)) {
2641                                 /* not infinity */
2642                                 flags |= RTF_EXPIRES;
2643                                 expires = jiffies_to_clock_t(rt_expires);
2644                         }
2645                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2646                                               dev, expires, flags);
2647                 }
2648                 ip6_rt_put(rt);
2649         }
2650
2651         /* Try to figure out our local address for this prefix */
2652
2653         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2654                 struct in6_addr addr;
2655                 bool tokenized = false, dev_addr_generated = false;
2656
2657                 if (pinfo->prefix_len == 64) {
2658                         memcpy(&addr, &pinfo->prefix, 8);
2659
2660                         if (!ipv6_addr_any(&in6_dev->token)) {
2661                                 read_lock_bh(&in6_dev->lock);
2662                                 memcpy(addr.s6_addr + 8,
2663                                        in6_dev->token.s6_addr + 8, 8);
2664                                 read_unlock_bh(&in6_dev->lock);
2665                                 tokenized = true;
2666                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2667                                    !ipv6_generate_stable_address(&addr, 0,
2668                                                                  in6_dev)) {
2669                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2670                                 goto ok;
2671                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2672                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2673                                 goto put;
2674                         } else {
2675                                 dev_addr_generated = true;
2676                         }
2677                         goto ok;
2678                 }
2679                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2680                                     pinfo->prefix_len);
2681                 goto put;
2682
2683 ok:
2684                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2685                                                    &addr, addr_type,
2686                                                    addr_flags, sllao,
2687                                                    tokenized, valid_lft,
2688                                                    prefered_lft);
2689                 if (err)
2690                         goto put;
2691
2692                 /* Ignore error case here because previous prefix add addr was
2693                  * successful which will be notified.
2694                  */
2695                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2696                                               addr_type, addr_flags, sllao,
2697                                               tokenized, valid_lft,
2698                                               prefered_lft,
2699                                               dev_addr_generated);
2700         }
2701         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2702 put:
2703         in6_dev_put(in6_dev);
2704 }
2705
2706 /*
2707  *      Set destination address.
2708  *      Special case for SIT interfaces where we create a new "virtual"
2709  *      device.
2710  */
2711 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2712 {
2713         struct in6_ifreq ireq;
2714         struct net_device *dev;
2715         int err = -EINVAL;
2716
2717         rtnl_lock();
2718
2719         err = -EFAULT;
2720         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2721                 goto err_exit;
2722
2723         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2724
2725         err = -ENODEV;
2726         if (!dev)
2727                 goto err_exit;
2728
2729 #if IS_ENABLED(CONFIG_IPV6_SIT)
2730         if (dev->type == ARPHRD_SIT) {
2731                 const struct net_device_ops *ops = dev->netdev_ops;
2732                 struct ifreq ifr;
2733                 struct ip_tunnel_parm p;
2734
2735                 err = -EADDRNOTAVAIL;
2736                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2737                         goto err_exit;
2738
2739                 memset(&p, 0, sizeof(p));
2740                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2741                 p.iph.saddr = 0;
2742                 p.iph.version = 4;
2743                 p.iph.ihl = 5;
2744                 p.iph.protocol = IPPROTO_IPV6;
2745                 p.iph.ttl = 64;
2746                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2747
2748                 if (ops->ndo_do_ioctl) {
2749                         mm_segment_t oldfs = get_fs();
2750
2751                         set_fs(KERNEL_DS);
2752                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2753                         set_fs(oldfs);
2754                 } else
2755                         err = -EOPNOTSUPP;
2756
2757                 if (err == 0) {
2758                         err = -ENOBUFS;
2759                         dev = __dev_get_by_name(net, p.name);
2760                         if (!dev)
2761                                 goto err_exit;
2762                         err = dev_open(dev);
2763                 }
2764         }
2765 #endif
2766
2767 err_exit:
2768         rtnl_unlock();
2769         return err;
2770 }
2771
2772 static int ipv6_mc_config(struct sock *sk, bool join,
2773                           const struct in6_addr *addr, int ifindex)
2774 {
2775         int ret;
2776
2777         ASSERT_RTNL();
2778
2779         lock_sock(sk);
2780         if (join)
2781                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2782         else
2783                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2784         release_sock(sk);
2785
2786         return ret;
2787 }
2788
2789 /*
2790  *      Manual configuration of address on an interface
2791  */
2792 static int inet6_addr_add(struct net *net, int ifindex,
2793                           const struct in6_addr *pfx,
2794                           const struct in6_addr *peer_pfx,
2795                           unsigned int plen, __u32 ifa_flags,
2796                           __u32 prefered_lft, __u32 valid_lft)
2797 {
2798         struct inet6_ifaddr *ifp;
2799         struct inet6_dev *idev;
2800         struct net_device *dev;
2801         unsigned long timeout;
2802         clock_t expires;
2803         int scope;
2804         u32 flags;
2805
2806         ASSERT_RTNL();
2807
2808         if (plen > 128)
2809                 return -EINVAL;
2810
2811         /* check the lifetime */
2812         if (!valid_lft || prefered_lft > valid_lft)
2813                 return -EINVAL;
2814
2815         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2816                 return -EINVAL;
2817
2818         dev = __dev_get_by_index(net, ifindex);
2819         if (!dev)
2820                 return -ENODEV;
2821
2822         idev = addrconf_add_dev(dev);
2823         if (IS_ERR(idev))
2824                 return PTR_ERR(idev);
2825
2826         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2827                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2828                                          true, pfx, ifindex);
2829
2830                 if (ret < 0)
2831                         return ret;
2832         }
2833
2834         scope = ipv6_addr_scope(pfx);
2835
2836         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2837         if (addrconf_finite_timeout(timeout)) {
2838                 expires = jiffies_to_clock_t(timeout * HZ);
2839                 valid_lft = timeout;
2840                 flags = RTF_EXPIRES;
2841         } else {
2842                 expires = 0;
2843                 flags = 0;
2844                 ifa_flags |= IFA_F_PERMANENT;
2845         }
2846
2847         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2848         if (addrconf_finite_timeout(timeout)) {
2849                 if (timeout == 0)
2850                         ifa_flags |= IFA_F_DEPRECATED;
2851                 prefered_lft = timeout;
2852         }
2853
2854         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2855                             valid_lft, prefered_lft);
2856
2857         if (!IS_ERR(ifp)) {
2858                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2859                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2860                                               expires, flags);
2861                 }
2862
2863                 /*
2864                  * Note that section 3.1 of RFC 4429 indicates
2865                  * that the Optimistic flag should not be set for
2866                  * manually configured addresses
2867                  */
2868                 addrconf_dad_start(ifp);
2869                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2870                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2871                                          true, jiffies);
2872                 in6_ifa_put(ifp);
2873                 addrconf_verify_rtnl();
2874                 return 0;
2875         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2876                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2877                                false, pfx, ifindex);
2878         }
2879
2880         return PTR_ERR(ifp);
2881 }
2882
2883 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2884                           const struct in6_addr *pfx, unsigned int plen)
2885 {
2886         struct inet6_ifaddr *ifp;
2887         struct inet6_dev *idev;
2888         struct net_device *dev;
2889
2890         if (plen > 128)
2891                 return -EINVAL;
2892
2893         dev = __dev_get_by_index(net, ifindex);
2894         if (!dev)
2895                 return -ENODEV;
2896
2897         idev = __in6_dev_get(dev);
2898         if (!idev)
2899                 return -ENXIO;
2900
2901         read_lock_bh(&idev->lock);
2902         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2903                 if (ifp->prefix_len == plen &&
2904                     ipv6_addr_equal(pfx, &ifp->addr)) {
2905                         in6_ifa_hold(ifp);
2906                         read_unlock_bh(&idev->lock);
2907
2908                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2909                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2910                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2911                                                  jiffies);
2912                         ipv6_del_addr(ifp);
2913                         addrconf_verify_rtnl();
2914                         if (ipv6_addr_is_multicast(pfx)) {
2915                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2916                                                false, pfx, dev->ifindex);
2917                         }
2918                         return 0;
2919                 }
2920         }
2921         read_unlock_bh(&idev->lock);
2922         return -EADDRNOTAVAIL;
2923 }
2924
2925
2926 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2927 {
2928         struct in6_ifreq ireq;
2929         int err;
2930
2931         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2932                 return -EPERM;
2933
2934         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2935                 return -EFAULT;
2936
2937         rtnl_lock();
2938         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2939                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2940                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2941         rtnl_unlock();
2942         return err;
2943 }
2944
2945 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2946 {
2947         struct in6_ifreq ireq;
2948         int err;
2949
2950         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2951                 return -EPERM;
2952
2953         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2954                 return -EFAULT;
2955
2956         rtnl_lock();
2957         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2958                              ireq.ifr6_prefixlen);
2959         rtnl_unlock();
2960         return err;
2961 }
2962
2963 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2964                      int plen, int scope)
2965 {
2966         struct inet6_ifaddr *ifp;
2967
2968         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2969                             scope, IFA_F_PERMANENT,
2970                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2971         if (!IS_ERR(ifp)) {
2972                 spin_lock_bh(&ifp->lock);
2973                 ifp->flags &= ~IFA_F_TENTATIVE;
2974                 spin_unlock_bh(&ifp->lock);
2975                 rt_genid_bump_ipv6(dev_net(idev->dev));
2976                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2977                 in6_ifa_put(ifp);
2978         }
2979 }
2980
2981 #if IS_ENABLED(CONFIG_IPV6_SIT)
2982 static void sit_add_v4_addrs(struct inet6_dev *idev)
2983 {
2984         struct in6_addr addr;
2985         struct net_device *dev;
2986         struct net *net = dev_net(idev->dev);
2987         int scope, plen;
2988         u32 pflags = 0;
2989
2990         ASSERT_RTNL();
2991
2992         memset(&addr, 0, sizeof(struct in6_addr));
2993         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2994
2995         if (idev->dev->flags&IFF_POINTOPOINT) {
2996                 addr.s6_addr32[0] = htonl(0xfe800000);
2997                 scope = IFA_LINK;
2998                 plen = 64;
2999         } else {
3000                 scope = IPV6_ADDR_COMPATv4;
3001                 plen = 96;
3002                 pflags |= RTF_NONEXTHOP;
3003         }
3004
3005         if (addr.s6_addr32[3]) {
3006                 add_addr(idev, &addr, plen, scope);
3007                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3008                 return;
3009         }
3010
3011         for_each_netdev(net, dev) {
3012                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3013                 if (in_dev && (dev->flags & IFF_UP)) {
3014                         struct in_ifaddr *ifa;
3015
3016                         int flag = scope;
3017
3018                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3019
3020                                 addr.s6_addr32[3] = ifa->ifa_local;
3021
3022                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3023                                         continue;
3024                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3025                                         if (idev->dev->flags&IFF_POINTOPOINT)
3026                                                 continue;
3027                                         flag |= IFA_HOST;
3028                                 }
3029
3030                                 add_addr(idev, &addr, plen, flag);
3031                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3032                                                       pflags);
3033                         }
3034                 }
3035         }
3036 }
3037 #endif
3038
3039 static void init_loopback(struct net_device *dev)
3040 {
3041         struct inet6_dev  *idev;
3042
3043         /* ::1 */
3044
3045         ASSERT_RTNL();
3046
3047         idev = ipv6_find_idev(dev);
3048         if (!idev) {
3049                 pr_debug("%s: add_dev failed\n", __func__);
3050                 return;
3051         }
3052
3053         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3054 }
3055
3056 void addrconf_add_linklocal(struct inet6_dev *idev,
3057                             const struct in6_addr *addr, u32 flags)
3058 {
3059         struct inet6_ifaddr *ifp;
3060         u32 addr_flags = flags | IFA_F_PERMANENT;
3061
3062 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3063         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3064              idev->cnf.optimistic_dad) &&
3065             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3066                 addr_flags |= IFA_F_OPTIMISTIC;
3067 #endif
3068
3069         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3070                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3071         if (!IS_ERR(ifp)) {
3072                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3073                 addrconf_dad_start(ifp);
3074                 in6_ifa_put(ifp);
3075         }
3076 }
3077 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3078
3079 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3080 {
3081         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3082                 return true;
3083
3084         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3085             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3086                 return true;
3087
3088         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3089             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3090                 return true;
3091
3092         return false;
3093 }
3094
3095 static int ipv6_generate_stable_address(struct in6_addr *address,
3096                                         u8 dad_count,
3097                                         const struct inet6_dev *idev)
3098 {
3099         static DEFINE_SPINLOCK(lock);
3100         static __u32 digest[SHA_DIGEST_WORDS];
3101         static __u32 workspace[SHA_WORKSPACE_WORDS];
3102
3103         static union {
3104                 char __data[SHA_MESSAGE_BYTES];
3105                 struct {
3106                         struct in6_addr secret;
3107                         __be32 prefix[2];
3108                         unsigned char hwaddr[MAX_ADDR_LEN];
3109                         u8 dad_count;
3110                 } __packed;
3111         } data;
3112
3113         struct in6_addr secret;
3114         struct in6_addr temp;
3115         struct net *net = dev_net(idev->dev);
3116
3117         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3118
3119         if (idev->cnf.stable_secret.initialized)
3120                 secret = idev->cnf.stable_secret.secret;
3121         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3122                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3123         else
3124                 return -1;
3125
3126 retry:
3127         spin_lock_bh(&lock);
3128
3129         sha_init(digest);
3130         memset(&data, 0, sizeof(data));
3131         memset(workspace, 0, sizeof(workspace));
3132         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3133         data.prefix[0] = address->s6_addr32[0];
3134         data.prefix[1] = address->s6_addr32[1];
3135         data.secret = secret;
3136         data.dad_count = dad_count;
3137
3138         sha_transform(digest, data.__data, workspace);
3139
3140         temp = *address;
3141         temp.s6_addr32[2] = (__force __be32)digest[0];
3142         temp.s6_addr32[3] = (__force __be32)digest[1];
3143
3144         spin_unlock_bh(&lock);
3145
3146         if (ipv6_reserved_interfaceid(temp)) {
3147                 dad_count++;
3148                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3149                         return -1;
3150                 goto retry;
3151         }
3152
3153         *address = temp;
3154         return 0;
3155 }
3156
3157 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3158 {
3159         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3160
3161         if (s->initialized)
3162                 return;
3163         s = &idev->cnf.stable_secret;
3164         get_random_bytes(&s->secret, sizeof(s->secret));
3165         s->initialized = true;
3166 }
3167
3168 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3169 {
3170         struct in6_addr addr;
3171
3172         /* no link local addresses on L3 master devices */
3173         if (netif_is_l3_master(idev->dev))
3174                 return;
3175
3176         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3177
3178         switch (idev->cnf.addr_gen_mode) {
3179         case IN6_ADDR_GEN_MODE_RANDOM:
3180                 ipv6_gen_mode_random_init(idev);
3181                 /* fallthrough */
3182         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3183                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3184                         addrconf_add_linklocal(idev, &addr,
3185                                                IFA_F_STABLE_PRIVACY);
3186                 else if (prefix_route)
3187                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3188                 break;
3189         case IN6_ADDR_GEN_MODE_EUI64:
3190                 /* addrconf_add_linklocal also adds a prefix_route and we
3191                  * only need to care about prefix routes if ipv6_generate_eui64
3192                  * couldn't generate one.
3193                  */
3194                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3195                         addrconf_add_linklocal(idev, &addr, 0);
3196                 else if (prefix_route)
3197                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3198                 break;
3199         case IN6_ADDR_GEN_MODE_NONE:
3200         default:
3201                 /* will not add any link local address */
3202                 break;
3203         }
3204 }
3205
3206 static void addrconf_dev_config(struct net_device *dev)
3207 {
3208         struct inet6_dev *idev;
3209
3210         ASSERT_RTNL();
3211
3212         if ((dev->type != ARPHRD_ETHER) &&
3213             (dev->type != ARPHRD_FDDI) &&
3214             (dev->type != ARPHRD_ARCNET) &&
3215             (dev->type != ARPHRD_INFINIBAND) &&
3216             (dev->type != ARPHRD_IEEE1394) &&
3217             (dev->type != ARPHRD_TUNNEL6) &&
3218             (dev->type != ARPHRD_6LOWPAN) &&
3219             (dev->type != ARPHRD_IP6GRE) &&
3220             (dev->type != ARPHRD_IPGRE) &&
3221             (dev->type != ARPHRD_TUNNEL) &&
3222             (dev->type != ARPHRD_NONE)) {
3223                 /* Alas, we support only Ethernet autoconfiguration. */
3224                 return;
3225         }
3226
3227         idev = addrconf_add_dev(dev);
3228         if (IS_ERR(idev))
3229                 return;
3230
3231         /* this device type has no EUI support */
3232         if (dev->type == ARPHRD_NONE &&
3233             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3234                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3235
3236         addrconf_addr_gen(idev, false);
3237 }
3238
3239 #if IS_ENABLED(CONFIG_IPV6_SIT)
3240 static void addrconf_sit_config(struct net_device *dev)
3241 {
3242         struct inet6_dev *idev;
3243
3244         ASSERT_RTNL();
3245
3246         /*
3247          * Configure the tunnel with one of our IPv4
3248          * addresses... we should configure all of
3249          * our v4 addrs in the tunnel
3250          */
3251
3252         idev = ipv6_find_idev(dev);
3253         if (!idev) {
3254                 pr_debug("%s: add_dev failed\n", __func__);
3255                 return;
3256         }
3257
3258         if (dev->priv_flags & IFF_ISATAP) {
3259                 addrconf_addr_gen(idev, false);
3260                 return;
3261         }
3262
3263         sit_add_v4_addrs(idev);
3264
3265         if (dev->flags&IFF_POINTOPOINT)
3266                 addrconf_add_mroute(dev);
3267 }
3268 #endif
3269
3270 #if IS_ENABLED(CONFIG_NET_IPGRE)
3271 static void addrconf_gre_config(struct net_device *dev)
3272 {
3273         struct inet6_dev *idev;
3274
3275         ASSERT_RTNL();
3276
3277         idev = ipv6_find_idev(dev);
3278         if (!idev) {
3279                 pr_debug("%s: add_dev failed\n", __func__);
3280                 return;
3281         }
3282
3283         addrconf_addr_gen(idev, true);
3284         if (dev->flags & IFF_POINTOPOINT)
3285                 addrconf_add_mroute(dev);
3286 }
3287 #endif
3288
3289 static int fixup_permanent_addr(struct inet6_dev *idev,
3290                                 struct inet6_ifaddr *ifp)
3291 {
3292         /* !rt6i_node means the host route was removed from the
3293          * FIB, for example, if 'lo' device is taken down. In that
3294          * case regenerate the host route.
3295          */
3296         if (!ifp->rt || !ifp->rt->rt6i_node) {
3297                 struct rt6_info *rt, *prev;
3298
3299                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3300                 if (unlikely(IS_ERR(rt)))
3301                         return PTR_ERR(rt);
3302
3303                 /* ifp->rt can be accessed outside of rtnl */
3304                 spin_lock(&ifp->lock);
3305                 prev = ifp->rt;
3306                 ifp->rt = rt;
3307                 spin_unlock(&ifp->lock);
3308
3309                 ip6_rt_put(prev);
3310         }
3311
3312         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3313                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3314                                       idev->dev, 0, 0);
3315         }
3316
3317         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3318                 addrconf_dad_start(ifp);
3319
3320         return 0;
3321 }
3322
3323 static void addrconf_permanent_addr(struct net_device *dev)
3324 {
3325         struct inet6_ifaddr *ifp, *tmp;
3326         struct inet6_dev *idev;
3327
3328         idev = __in6_dev_get(dev);
3329         if (!idev)
3330                 return;
3331
3332         write_lock_bh(&idev->lock);
3333
3334         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3335                 if ((ifp->flags & IFA_F_PERMANENT) &&
3336                     fixup_permanent_addr(idev, ifp) < 0) {
3337                         write_unlock_bh(&idev->lock);
3338                         ipv6_del_addr(ifp);
3339                         write_lock_bh(&idev->lock);
3340
3341                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3342                                              idev->dev->name, &ifp->addr);
3343                 }
3344         }
3345
3346         write_unlock_bh(&idev->lock);
3347 }
3348
3349 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3350                            void *ptr)
3351 {
3352         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3353         struct netdev_notifier_changeupper_info *info;
3354         struct inet6_dev *idev = __in6_dev_get(dev);
3355         struct net *net = dev_net(dev);
3356         int run_pending = 0;
3357         int err;
3358
3359         switch (event) {
3360         case NETDEV_REGISTER:
3361                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3362                         idev = ipv6_add_dev(dev);
3363                         if (IS_ERR(idev))
3364                                 return notifier_from_errno(PTR_ERR(idev));
3365                 }
3366                 break;
3367
3368         case NETDEV_CHANGEMTU:
3369                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3370                 if (dev->mtu < IPV6_MIN_MTU) {
3371                         addrconf_ifdown(dev, dev != net->loopback_dev);
3372                         break;
3373                 }
3374
3375                 if (idev) {
3376                         rt6_mtu_change(dev, dev->mtu);
3377                         idev->cnf.mtu6 = dev->mtu;
3378                         break;
3379                 }
3380
3381                 /* allocate new idev */
3382                 idev = ipv6_add_dev(dev);
3383                 if (IS_ERR(idev))
3384                         break;
3385
3386                 /* device is still not ready */
3387                 if (!(idev->if_flags & IF_READY))
3388                         break;
3389
3390                 run_pending = 1;
3391
3392                 /* fall through */
3393
3394         case NETDEV_UP:
3395         case NETDEV_CHANGE:
3396                 if (dev->flags & IFF_SLAVE)
3397                         break;
3398
3399                 if (idev && idev->cnf.disable_ipv6)
3400                         break;
3401
3402                 if (event == NETDEV_UP) {
3403                         /* restore routes for permanent addresses */
3404                         addrconf_permanent_addr(dev);
3405
3406                         if (!addrconf_qdisc_ok(dev)) {
3407                                 /* device is not ready yet. */
3408                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3409                                         dev->name);
3410                                 break;
3411                         }
3412
3413                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3414                                 idev = ipv6_add_dev(dev);
3415
3416                         if (!IS_ERR_OR_NULL(idev)) {
3417                                 idev->if_flags |= IF_READY;
3418                                 run_pending = 1;
3419                         }
3420                 } else if (event == NETDEV_CHANGE) {
3421                         if (!addrconf_qdisc_ok(dev)) {
3422                                 /* device is still not ready. */
3423                                 break;
3424                         }
3425
3426                         if (idev) {
3427                                 if (idev->if_flags & IF_READY) {
3428                                         /* device is already configured -
3429                                          * but resend MLD reports, we might
3430                                          * have roamed and need to update
3431                                          * multicast snooping switches
3432                                          */
3433                                         ipv6_mc_up(idev);
3434                                         break;
3435                                 }
3436                                 idev->if_flags |= IF_READY;
3437                         }
3438
3439                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3440                                 dev->name);
3441
3442                         run_pending = 1;
3443                 }
3444
3445                 switch (dev->type) {
3446 #if IS_ENABLED(CONFIG_IPV6_SIT)
3447                 case ARPHRD_SIT:
3448                         addrconf_sit_config(dev);
3449                         break;
3450 #endif
3451 #if IS_ENABLED(CONFIG_NET_IPGRE)
3452                 case ARPHRD_IPGRE:
3453                         addrconf_gre_config(dev);
3454                         break;
3455 #endif
3456                 case ARPHRD_LOOPBACK:
3457                         init_loopback(dev);
3458                         break;
3459
3460                 default:
3461                         addrconf_dev_config(dev);
3462                         break;
3463                 }
3464
3465                 if (!IS_ERR_OR_NULL(idev)) {
3466                         if (run_pending)
3467                                 addrconf_dad_run(idev);
3468
3469                         /*
3470                          * If the MTU changed during the interface down,
3471                          * when the interface up, the changed MTU must be
3472                          * reflected in the idev as well as routers.
3473                          */
3474                         if (idev->cnf.mtu6 != dev->mtu &&
3475                             dev->mtu >= IPV6_MIN_MTU) {
3476                                 rt6_mtu_change(dev, dev->mtu);
3477                                 idev->cnf.mtu6 = dev->mtu;
3478                         }
3479                         idev->tstamp = jiffies;
3480                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3481
3482                         /*
3483                          * If the changed mtu during down is lower than
3484                          * IPV6_MIN_MTU stop IPv6 on this interface.
3485                          */
3486                         if (dev->mtu < IPV6_MIN_MTU)
3487                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3488                 }
3489                 break;
3490
3491         case NETDEV_DOWN:
3492         case NETDEV_UNREGISTER:
3493                 /*
3494                  *      Remove all addresses from this interface.
3495                  */
3496                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3497                 break;
3498
3499         case NETDEV_CHANGENAME:
3500                 if (idev) {
3501                         snmp6_unregister_dev(idev);
3502                         addrconf_sysctl_unregister(idev);
3503                         err = addrconf_sysctl_register(idev);
3504                         if (err)
3505                                 return notifier_from_errno(err);
3506                         err = snmp6_register_dev(idev);
3507                         if (err) {
3508                                 addrconf_sysctl_unregister(idev);
3509                                 return notifier_from_errno(err);
3510                         }
3511                 }
3512                 break;
3513
3514         case NETDEV_PRE_TYPE_CHANGE:
3515         case NETDEV_POST_TYPE_CHANGE:
3516                 if (idev)
3517                         addrconf_type_change(dev, event);
3518                 break;
3519
3520         case NETDEV_CHANGEUPPER:
3521                 info = ptr;
3522
3523                 /* flush all routes if dev is linked to or unlinked from
3524                  * an L3 master device (e.g., VRF)
3525                  */
3526                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3527                         addrconf_ifdown(dev, 0);
3528         }
3529
3530         return NOTIFY_OK;
3531 }
3532
3533 /*
3534  *      addrconf module should be notified of a device going up
3535  */
3536 static struct notifier_block ipv6_dev_notf = {
3537         .notifier_call = addrconf_notify,
3538         .priority = ADDRCONF_NOTIFY_PRIORITY,
3539 };
3540
3541 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3542 {
3543         struct inet6_dev *idev;
3544         ASSERT_RTNL();
3545
3546         idev = __in6_dev_get(dev);
3547
3548         if (event == NETDEV_POST_TYPE_CHANGE)
3549                 ipv6_mc_remap(idev);
3550         else if (event == NETDEV_PRE_TYPE_CHANGE)
3551                 ipv6_mc_unmap(idev);
3552 }
3553
3554 static bool addr_is_local(const struct in6_addr *addr)
3555 {
3556         return ipv6_addr_type(addr) &
3557                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3558 }
3559
3560 static int addrconf_ifdown(struct net_device *dev, int how)
3561 {
3562         struct net *net = dev_net(dev);
3563         struct inet6_dev *idev;
3564         struct inet6_ifaddr *ifa, *tmp;
3565         struct list_head del_list;
3566         int _keep_addr;
3567         bool keep_addr;
3568         int state, i;
3569
3570         ASSERT_RTNL();
3571
3572         rt6_ifdown(net, dev);
3573         neigh_ifdown(&nd_tbl, dev);
3574
3575         idev = __in6_dev_get(dev);
3576         if (!idev)
3577                 return -ENODEV;
3578
3579         /*
3580          * Step 1: remove reference to ipv6 device from parent device.
3581          *         Do not dev_put!
3582          */
3583         if (how) {
3584                 idev->dead = 1;
3585
3586                 /* protected by rtnl_lock */
3587                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3588
3589                 /* Step 1.5: remove snmp6 entry */
3590                 snmp6_unregister_dev(idev);
3591
3592         }
3593
3594         /* aggregate the system setting and interface setting */
3595         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3596         if (!_keep_addr)
3597                 _keep_addr = idev->cnf.keep_addr_on_down;
3598
3599         /* combine the user config with event to determine if permanent
3600          * addresses are to be removed from address hash table
3601          */
3602         keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3603
3604         /* Step 2: clear hash table */
3605         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3606                 struct hlist_head *h = &inet6_addr_lst[i];
3607
3608                 spin_lock_bh(&addrconf_hash_lock);
3609 restart:
3610                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3611                         if (ifa->idev == idev) {
3612                                 addrconf_del_dad_work(ifa);
3613                                 /* combined flag + permanent flag decide if
3614                                  * address is retained on a down event
3615                                  */
3616                                 if (!keep_addr ||
3617                                     !(ifa->flags & IFA_F_PERMANENT) ||
3618                                     addr_is_local(&ifa->addr)) {
3619                                         hlist_del_init_rcu(&ifa->addr_lst);
3620                                         goto restart;
3621                                 }
3622                         }
3623                 }
3624                 spin_unlock_bh(&addrconf_hash_lock);
3625         }
3626
3627         write_lock_bh(&idev->lock);
3628
3629         addrconf_del_rs_timer(idev);
3630
3631         /* Step 2: clear flags for stateless addrconf */
3632         if (!how)
3633                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3634
3635         /* Step 3: clear tempaddr list */
3636         while (!list_empty(&idev->tempaddr_list)) {
3637                 ifa = list_first_entry(&idev->tempaddr_list,
3638                                        struct inet6_ifaddr, tmp_list);
3639                 list_del(&ifa->tmp_list);
3640                 write_unlock_bh(&idev->lock);
3641                 spin_lock_bh(&ifa->lock);
3642
3643                 if (ifa->ifpub) {
3644                         in6_ifa_put(ifa->ifpub);
3645                         ifa->ifpub = NULL;
3646                 }
3647                 spin_unlock_bh(&ifa->lock);
3648                 in6_ifa_put(ifa);
3649                 write_lock_bh(&idev->lock);
3650         }
3651
3652         /* re-combine the user config with event to determine if permanent
3653          * addresses are to be removed from the interface list
3654          */
3655         keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3656
3657         INIT_LIST_HEAD(&del_list);
3658         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3659                 struct rt6_info *rt = NULL;
3660                 bool keep;
3661
3662                 addrconf_del_dad_work(ifa);
3663
3664                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3665                         !addr_is_local(&ifa->addr);
3666                 if (!keep)
3667                         list_move(&ifa->if_list, &del_list);
3668
3669                 write_unlock_bh(&idev->lock);
3670                 spin_lock_bh(&ifa->lock);
3671
3672                 if (keep) {
3673                         /* set state to skip the notifier below */
3674                         state = INET6_IFADDR_STATE_DEAD;
3675                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3676                         if (!(ifa->flags & IFA_F_NODAD))
3677                                 ifa->flags |= IFA_F_TENTATIVE;
3678
3679                         rt = ifa->rt;
3680                         ifa->rt = NULL;
3681                 } else {
3682                         state = ifa->state;
3683                         ifa->state = INET6_IFADDR_STATE_DEAD;
3684                 }
3685
3686                 spin_unlock_bh(&ifa->lock);
3687
3688                 if (rt)
3689                         ip6_del_rt(rt);
3690
3691                 if (state != INET6_IFADDR_STATE_DEAD) {
3692                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3693                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3694                 } else {
3695                         if (idev->cnf.forwarding)
3696                                 addrconf_leave_anycast(ifa);
3697                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3698                 }
3699
3700                 write_lock_bh(&idev->lock);
3701         }
3702
3703         write_unlock_bh(&idev->lock);
3704
3705         /* now clean up addresses to be removed */
3706         while (!list_empty(&del_list)) {
3707                 ifa = list_first_entry(&del_list,
3708                                        struct inet6_ifaddr, if_list);
3709                 list_del(&ifa->if_list);
3710
3711                 in6_ifa_put(ifa);
3712         }
3713
3714         /* Step 5: Discard anycast and multicast list */
3715         if (how) {
3716                 ipv6_ac_destroy_dev(idev);
3717                 ipv6_mc_destroy_dev(idev);
3718         } else {
3719                 ipv6_mc_down(idev);
3720         }
3721
3722         idev->tstamp = jiffies;
3723
3724         /* Last: Shot the device (if unregistered) */
3725         if (how) {
3726                 addrconf_sysctl_unregister(idev);
3727                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3728                 neigh_ifdown(&nd_tbl, dev);
3729                 in6_dev_put(idev);
3730         }
3731         return 0;
3732 }
3733
3734 static void addrconf_rs_timer(unsigned long data)
3735 {
3736         struct inet6_dev *idev = (struct inet6_dev *)data;
3737         struct net_device *dev = idev->dev;
3738         struct in6_addr lladdr;
3739
3740         write_lock(&idev->lock);
3741         if (idev->dead || !(idev->if_flags & IF_READY))
3742                 goto out;
3743
3744         if (!ipv6_accept_ra(idev))
3745                 goto out;
3746
3747         /* Announcement received after solicitation was sent */
3748         if (idev->if_flags & IF_RA_RCVD)
3749                 goto out;
3750
3751         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3752                 write_unlock(&idev->lock);
3753                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3754                         ndisc_send_rs(dev, &lladdr,
3755                                       &in6addr_linklocal_allrouters);
3756                 else
3757                         goto put;
3758
3759                 write_lock(&idev->lock);
3760                 idev->rs_interval = rfc3315_s14_backoff_update(
3761                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3762                 /* The wait after the last probe can be shorter */
3763                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3764                                              idev->cnf.rtr_solicits) ?
3765                                       idev->cnf.rtr_solicit_delay :
3766                                       idev->rs_interval);
3767         } else {
3768                 /*
3769                  * Note: we do not support deprecated "all on-link"
3770                  * assumption any longer.
3771                  */
3772                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3773         }
3774
3775 out:
3776         write_unlock(&idev->lock);
3777 put:
3778         in6_dev_put(idev);
3779 }
3780
3781 /*
3782  *      Duplicate Address Detection
3783  */
3784 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3785 {
3786         unsigned long rand_num;
3787         struct inet6_dev *idev = ifp->idev;
3788         u64 nonce;
3789
3790         if (ifp->flags & IFA_F_OPTIMISTIC)
3791                 rand_num = 0;
3792         else
3793                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3794
3795         nonce = 0;
3796         if (idev->cnf.enhanced_dad ||
3797             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3798                 do
3799                         get_random_bytes(&nonce, 6);
3800                 while (nonce == 0);
3801         }
3802         ifp->dad_nonce = nonce;
3803         ifp->dad_probes = idev->cnf.dad_transmits;
3804         addrconf_mod_dad_work(ifp, rand_num);
3805 }
3806
3807 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3808 {
3809         struct inet6_dev *idev = ifp->idev;
3810         struct net_device *dev = idev->dev;
3811         bool bump_id, notify = false;
3812
3813         addrconf_join_solict(dev, &ifp->addr);
3814
3815         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3816
3817         read_lock_bh(&idev->lock);
3818         spin_lock(&ifp->lock);
3819         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3820                 goto out;
3821
3822         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3823             dev_net(dev)->ipv6.devconf_all->accept_dad < 1 ||
3824             idev->cnf.accept_dad < 1 ||
3825             !(ifp->flags&IFA_F_TENTATIVE) ||
3826             ifp->flags & IFA_F_NODAD) {
3827                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3828                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3829                 spin_unlock(&ifp->lock);
3830                 read_unlock_bh(&idev->lock);
3831
3832                 addrconf_dad_completed(ifp, bump_id);
3833                 return;
3834         }
3835
3836         if (!(idev->if_flags & IF_READY)) {
3837                 spin_unlock(&ifp->lock);
3838                 read_unlock_bh(&idev->lock);
3839                 /*
3840                  * If the device is not ready:
3841                  * - keep it tentative if it is a permanent address.
3842                  * - otherwise, kill it.
3843                  */
3844                 in6_ifa_hold(ifp);
3845                 addrconf_dad_stop(ifp, 0);
3846                 return;
3847         }
3848
3849         /*
3850          * Optimistic nodes can start receiving
3851          * Frames right away
3852          */
3853         if (ifp->flags & IFA_F_OPTIMISTIC) {
3854                 ip6_ins_rt(ifp->rt);
3855                 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3856                         /* Because optimistic nodes can use this address,
3857                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3858                          */
3859                         notify = true;
3860                 }
3861         }
3862
3863         addrconf_dad_kick(ifp);
3864 out:
3865         spin_unlock(&ifp->lock);
3866         read_unlock_bh(&idev->lock);
3867         if (notify)
3868                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3869 }
3870
3871 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3872 {
3873         bool begin_dad = false;
3874
3875         spin_lock_bh(&ifp->lock);
3876         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3877                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3878                 begin_dad = true;
3879         }
3880         spin_unlock_bh(&ifp->lock);
3881
3882         if (begin_dad)
3883                 addrconf_mod_dad_work(ifp, 0);
3884 }
3885
3886 static void addrconf_dad_work(struct work_struct *w)
3887 {
3888         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3889                                                 struct inet6_ifaddr,
3890                                                 dad_work);
3891         struct inet6_dev *idev = ifp->idev;
3892         bool bump_id, disable_ipv6 = false;
3893         struct in6_addr mcaddr;
3894
3895         enum {
3896                 DAD_PROCESS,
3897                 DAD_BEGIN,
3898                 DAD_ABORT,
3899         } action = DAD_PROCESS;
3900
3901         rtnl_lock();
3902
3903         spin_lock_bh(&ifp->lock);
3904         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3905                 action = DAD_BEGIN;
3906                 ifp->state = INET6_IFADDR_STATE_DAD;
3907         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3908                 action = DAD_ABORT;
3909                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3910
3911                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3912                      idev->cnf.accept_dad > 1) &&
3913                     !idev->cnf.disable_ipv6 &&
3914                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3915                         struct in6_addr addr;
3916
3917                         addr.s6_addr32[0] = htonl(0xfe800000);
3918                         addr.s6_addr32[1] = 0;
3919
3920                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3921                             ipv6_addr_equal(&ifp->addr, &addr)) {
3922                                 /* DAD failed for link-local based on MAC */
3923                                 idev->cnf.disable_ipv6 = 1;
3924
3925                                 pr_info("%s: IPv6 being disabled!\n",
3926                                         ifp->idev->dev->name);
3927                                 disable_ipv6 = true;
3928                         }
3929                 }
3930         }
3931         spin_unlock_bh(&ifp->lock);
3932
3933         if (action == DAD_BEGIN) {
3934                 addrconf_dad_begin(ifp);
3935                 goto out;
3936         } else if (action == DAD_ABORT) {
3937                 in6_ifa_hold(ifp);
3938                 addrconf_dad_stop(ifp, 1);
3939                 if (disable_ipv6)
3940                         addrconf_ifdown(idev->dev, 0);
3941                 goto out;
3942         }
3943
3944         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3945                 goto out;
3946
3947         write_lock_bh(&idev->lock);
3948         if (idev->dead || !(idev->if_flags & IF_READY)) {
3949                 write_unlock_bh(&idev->lock);
3950                 goto out;
3951         }
3952
3953         spin_lock(&ifp->lock);
3954         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3955                 spin_unlock(&ifp->lock);
3956                 write_unlock_bh(&idev->lock);
3957                 goto out;
3958         }
3959
3960         if (ifp->dad_probes == 0) {
3961                 /*
3962                  * DAD was successful
3963                  */
3964
3965                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3966                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3967                 spin_unlock(&ifp->lock);
3968                 write_unlock_bh(&idev->lock);
3969
3970                 addrconf_dad_completed(ifp, bump_id);
3971
3972                 goto out;
3973         }
3974
3975         ifp->dad_probes--;
3976         addrconf_mod_dad_work(ifp,
3977                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3978         spin_unlock(&ifp->lock);
3979         write_unlock_bh(&idev->lock);
3980
3981         /* send a neighbour solicitation for our addr */
3982         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3983         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
3984                       ifp->dad_nonce);
3985 out:
3986         in6_ifa_put(ifp);
3987         rtnl_unlock();
3988 }
3989
3990 /* ifp->idev must be at least read locked */
3991 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3992 {
3993         struct inet6_ifaddr *ifpiter;
3994         struct inet6_dev *idev = ifp->idev;
3995
3996         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3997                 if (ifpiter->scope > IFA_LINK)
3998                         break;
3999                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4000                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4001                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4002                     IFA_F_PERMANENT)
4003                         return false;
4004         }
4005         return true;
4006 }
4007
4008 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
4009 {
4010         struct net_device *dev = ifp->idev->dev;
4011         struct in6_addr lladdr;
4012         bool send_rs, send_mld;
4013
4014         addrconf_del_dad_work(ifp);
4015
4016         /*
4017          *      Configure the address for reception. Now it is valid.
4018          */
4019
4020         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4021
4022         /* If added prefix is link local and we are prepared to process
4023            router advertisements, start sending router solicitations.
4024          */
4025
4026         read_lock_bh(&ifp->idev->lock);
4027         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4028         send_rs = send_mld &&
4029                   ipv6_accept_ra(ifp->idev) &&
4030                   ifp->idev->cnf.rtr_solicits != 0 &&
4031                   (dev->flags&IFF_LOOPBACK) == 0;
4032         read_unlock_bh(&ifp->idev->lock);
4033
4034         /* While dad is in progress mld report's source address is in6_addrany.
4035          * Resend with proper ll now.
4036          */
4037         if (send_mld)
4038                 ipv6_mc_dad_complete(ifp->idev);
4039
4040         if (send_rs) {
4041                 /*
4042                  *      If a host as already performed a random delay
4043                  *      [...] as part of DAD [...] there is no need
4044                  *      to delay again before sending the first RS
4045                  */
4046                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4047                         return;
4048                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4049
4050                 write_lock_bh(&ifp->idev->lock);
4051                 spin_lock(&ifp->lock);
4052                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4053                         ifp->idev->cnf.rtr_solicit_interval);
4054                 ifp->idev->rs_probes = 1;
4055                 ifp->idev->if_flags |= IF_RS_SENT;
4056                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4057                 spin_unlock(&ifp->lock);
4058                 write_unlock_bh(&ifp->idev->lock);
4059         }
4060
4061         if (bump_id)
4062                 rt_genid_bump_ipv6(dev_net(dev));
4063
4064         /* Make sure that a new temporary address will be created
4065          * before this temporary address becomes deprecated.
4066          */
4067         if (ifp->flags & IFA_F_TEMPORARY)
4068                 addrconf_verify_rtnl();
4069 }
4070
4071 static void addrconf_dad_run(struct inet6_dev *idev)
4072 {
4073         struct inet6_ifaddr *ifp;
4074
4075         read_lock_bh(&idev->lock);
4076         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4077                 spin_lock(&ifp->lock);
4078                 if (ifp->flags & IFA_F_TENTATIVE &&
4079                     ifp->state == INET6_IFADDR_STATE_DAD)
4080                         addrconf_dad_kick(ifp);
4081                 spin_unlock(&ifp->lock);
4082         }
4083         read_unlock_bh(&idev->lock);
4084 }
4085
4086 #ifdef CONFIG_PROC_FS
4087 struct if6_iter_state {
4088         struct seq_net_private p;
4089         int bucket;
4090         int offset;
4091 };
4092
4093 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4094 {
4095         struct inet6_ifaddr *ifa = NULL;
4096         struct if6_iter_state *state = seq->private;
4097         struct net *net = seq_file_net(seq);
4098         int p = 0;
4099
4100         /* initial bucket if pos is 0 */
4101         if (pos == 0) {
4102                 state->bucket = 0;
4103                 state->offset = 0;
4104         }
4105
4106         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4107                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4108                                          addr_lst) {
4109                         if (!net_eq(dev_net(ifa->idev->dev), net))
4110                                 continue;
4111                         /* sync with offset */
4112                         if (p < state->offset) {
4113                                 p++;
4114                                 continue;
4115                         }
4116                         state->offset++;
4117                         return ifa;
4118                 }
4119
4120                 /* prepare for next bucket */
4121                 state->offset = 0;
4122                 p = 0;
4123         }
4124         return NULL;
4125 }
4126
4127 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4128                                          struct inet6_ifaddr *ifa)
4129 {
4130         struct if6_iter_state *state = seq->private;
4131         struct net *net = seq_file_net(seq);
4132
4133         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4134                 if (!net_eq(dev_net(ifa->idev->dev), net))
4135                         continue;
4136                 state->offset++;
4137                 return ifa;
4138         }
4139
4140         while (++state->bucket < IN6_ADDR_HSIZE) {
4141                 state->offset = 0;
4142                 hlist_for_each_entry_rcu_bh(ifa,
4143                                      &inet6_addr_lst[state->bucket], addr_lst) {
4144                         if (!net_eq(dev_net(ifa->idev->dev), net))
4145                                 continue;
4146                         state->offset++;
4147                         return ifa;
4148                 }
4149         }
4150
4151         return NULL;
4152 }
4153
4154 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4155         __acquires(rcu_bh)
4156 {
4157         rcu_read_lock_bh();
4158         return if6_get_first(seq, *pos);
4159 }
4160
4161 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4162 {
4163         struct inet6_ifaddr *ifa;
4164
4165         ifa = if6_get_next(seq, v);
4166         ++*pos;
4167         return ifa;
4168 }
4169
4170 static void if6_seq_stop(struct seq_file *seq, void *v)
4171         __releases(rcu_bh)
4172 {
4173         rcu_read_unlock_bh();
4174 }
4175
4176 static int if6_seq_show(struct seq_file *seq, void *v)
4177 {
4178         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4179         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4180                    &ifp->addr,
4181                    ifp->idev->dev->ifindex,
4182                    ifp->prefix_len,
4183                    ifp->scope,
4184                    (u8) ifp->flags,
4185                    ifp->idev->dev->name);
4186         return 0;
4187 }
4188
4189 static const struct seq_operations if6_seq_ops = {
4190         .start  = if6_seq_start,
4191         .next   = if6_seq_next,
4192         .show   = if6_seq_show,
4193         .stop   = if6_seq_stop,
4194 };
4195
4196 static int if6_seq_open(struct inode *inode, struct file *file)
4197 {
4198         return seq_open_net(inode, file, &if6_seq_ops,
4199                             sizeof(struct if6_iter_state));
4200 }
4201
4202 static const struct file_operations if6_fops = {
4203         .owner          = THIS_MODULE,
4204         .open           = if6_seq_open,
4205         .read           = seq_read,
4206         .llseek         = seq_lseek,
4207         .release        = seq_release_net,
4208 };
4209
4210 static int __net_init if6_proc_net_init(struct net *net)
4211 {
4212         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4213                 return -ENOMEM;
4214         return 0;
4215 }
4216
4217 static void __net_exit if6_proc_net_exit(struct net *net)
4218 {
4219         remove_proc_entry("if_inet6", net->proc_net);
4220 }
4221
4222 static struct pernet_operations if6_proc_net_ops = {
4223         .init = if6_proc_net_init,
4224         .exit = if6_proc_net_exit,
4225 };
4226
4227 int __init if6_proc_init(void)
4228 {
4229         return register_pernet_subsys(&if6_proc_net_ops);
4230 }
4231
4232 void if6_proc_exit(void)
4233 {
4234         unregister_pernet_subsys(&if6_proc_net_ops);
4235 }
4236 #endif  /* CONFIG_PROC_FS */
4237
4238 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4239 /* Check if address is a home address configured on any interface. */
4240 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4241 {
4242         int ret = 0;
4243         struct inet6_ifaddr *ifp = NULL;
4244         unsigned int hash = inet6_addr_hash(addr);
4245
4246         rcu_read_lock_bh();
4247         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4248                 if (!net_eq(dev_net(ifp->idev->dev), net))
4249                         continue;
4250                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4251                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4252                         ret = 1;
4253                         break;
4254                 }
4255         }
4256         rcu_read_unlock_bh();
4257         return ret;
4258 }
4259 #endif
4260
4261 /*
4262  *      Periodic address status verification
4263  */
4264
4265 static void addrconf_verify_rtnl(void)
4266 {
4267         unsigned long now, next, next_sec, next_sched;
4268         struct inet6_ifaddr *ifp;
4269         int i;
4270
4271         ASSERT_RTNL();
4272
4273         rcu_read_lock_bh();
4274         now = jiffies;
4275         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4276
4277         cancel_delayed_work(&addr_chk_work);
4278
4279         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4280 restart:
4281                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4282                         unsigned long age;
4283
4284                         /* When setting preferred_lft to a value not zero or
4285                          * infinity, while valid_lft is infinity
4286                          * IFA_F_PERMANENT has a non-infinity life time.
4287                          */
4288                         if ((ifp->flags & IFA_F_PERMANENT) &&
4289                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4290                                 continue;
4291
4292                         spin_lock(&ifp->lock);
4293                         /* We try to batch several events at once. */
4294                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4295
4296                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4297                             age >= ifp->valid_lft) {
4298                                 spin_unlock(&ifp->lock);
4299                                 in6_ifa_hold(ifp);
4300                                 ipv6_del_addr(ifp);
4301                                 goto restart;
4302                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4303                                 spin_unlock(&ifp->lock);
4304                                 continue;
4305                         } else if (age >= ifp->prefered_lft) {
4306                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4307                                 int deprecate = 0;
4308
4309                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4310                                         deprecate = 1;
4311                                         ifp->flags |= IFA_F_DEPRECATED;
4312                                 }
4313
4314                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4315                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4316                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4317
4318                                 spin_unlock(&ifp->lock);
4319
4320                                 if (deprecate) {
4321                                         in6_ifa_hold(ifp);
4322
4323                                         ipv6_ifa_notify(0, ifp);
4324                                         in6_ifa_put(ifp);
4325                                         goto restart;
4326                                 }
4327                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4328                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4329                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4330                                         ifp->idev->cnf.dad_transmits *
4331                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4332
4333                                 if (age >= ifp->prefered_lft - regen_advance) {
4334                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4335                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4336                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4337                                         if (!ifp->regen_count && ifpub) {
4338                                                 ifp->regen_count++;
4339                                                 in6_ifa_hold(ifp);
4340                                                 in6_ifa_hold(ifpub);
4341                                                 spin_unlock(&ifp->lock);
4342
4343                                                 spin_lock(&ifpub->lock);
4344                                                 ifpub->regen_count = 0;
4345                                                 spin_unlock(&ifpub->lock);
4346                                                 ipv6_create_tempaddr(ifpub, ifp);
4347                                                 in6_ifa_put(ifpub);
4348                                                 in6_ifa_put(ifp);
4349                                                 goto restart;
4350                                         }
4351                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4352                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4353                                 spin_unlock(&ifp->lock);
4354                         } else {
4355                                 /* ifp->prefered_lft <= ifp->valid_lft */
4356                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4357                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4358                                 spin_unlock(&ifp->lock);
4359                         }
4360                 }
4361         }
4362
4363         next_sec = round_jiffies_up(next);
4364         next_sched = next;
4365
4366         /* If rounded timeout is accurate enough, accept it. */
4367         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4368                 next_sched = next_sec;
4369
4370         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4371         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4372                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4373
4374         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4375               now, next, next_sec, next_sched);
4376         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4377         rcu_read_unlock_bh();
4378 }
4379
4380 static void addrconf_verify_work(struct work_struct *w)
4381 {
4382         rtnl_lock();
4383         addrconf_verify_rtnl();
4384         rtnl_unlock();
4385 }
4386
4387 static void addrconf_verify(void)
4388 {
4389         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4390 }
4391
4392 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4393                                      struct in6_addr **peer_pfx)
4394 {
4395         struct in6_addr *pfx = NULL;
4396
4397         *peer_pfx = NULL;
4398
4399         if (addr)
4400                 pfx = nla_data(addr);
4401
4402         if (local) {
4403                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4404                         *peer_pfx = pfx;
4405                 pfx = nla_data(local);
4406         }
4407
4408         return pfx;
4409 }
4410
4411 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4412         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4413         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4414         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4415         [IFA_FLAGS]             = { .len = sizeof(u32) },
4416 };
4417
4418 static int
4419 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4420                   struct netlink_ext_ack *extack)
4421 {
4422         struct net *net = sock_net(skb->sk);
4423         struct ifaddrmsg *ifm;
4424         struct nlattr *tb[IFA_MAX+1];
4425         struct in6_addr *pfx, *peer_pfx;
4426         u32 ifa_flags;
4427         int err;
4428
4429         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4430                           extack);
4431         if (err < 0)
4432                 return err;
4433
4434         ifm = nlmsg_data(nlh);
4435         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4436         if (!pfx)
4437                 return -EINVAL;
4438
4439         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4440
4441         /* We ignore other flags so far. */
4442         ifa_flags &= IFA_F_MANAGETEMPADDR;
4443
4444         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4445                               ifm->ifa_prefixlen);
4446 }
4447
4448 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4449                              u32 prefered_lft, u32 valid_lft)
4450 {
4451         u32 flags;
4452         clock_t expires;
4453         unsigned long timeout;
4454         bool was_managetempaddr;
4455         bool had_prefixroute;
4456
4457         ASSERT_RTNL();
4458
4459         if (!valid_lft || (prefered_lft > valid_lft))
4460                 return -EINVAL;
4461
4462         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4463             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4464                 return -EINVAL;
4465
4466         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4467         if (addrconf_finite_timeout(timeout)) {
4468                 expires = jiffies_to_clock_t(timeout * HZ);
4469                 valid_lft = timeout;
4470                 flags = RTF_EXPIRES;
4471         } else {
4472                 expires = 0;
4473                 flags = 0;
4474                 ifa_flags |= IFA_F_PERMANENT;
4475         }
4476
4477         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4478         if (addrconf_finite_timeout(timeout)) {
4479                 if (timeout == 0)
4480                         ifa_flags |= IFA_F_DEPRECATED;
4481                 prefered_lft = timeout;
4482         }
4483
4484         spin_lock_bh(&ifp->lock);
4485         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4486         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4487                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4488         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4489                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4490                         IFA_F_NOPREFIXROUTE);
4491         ifp->flags |= ifa_flags;
4492         ifp->tstamp = jiffies;
4493         ifp->valid_lft = valid_lft;
4494         ifp->prefered_lft = prefered_lft;
4495
4496         spin_unlock_bh(&ifp->lock);
4497         if (!(ifp->flags&IFA_F_TENTATIVE))
4498                 ipv6_ifa_notify(0, ifp);
4499
4500         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4501                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4502                                       expires, flags);
4503         } else if (had_prefixroute) {
4504                 enum cleanup_prefix_rt_t action;
4505                 unsigned long rt_expires;
4506
4507                 write_lock_bh(&ifp->idev->lock);
4508                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4509                 write_unlock_bh(&ifp->idev->lock);
4510
4511                 if (action != CLEANUP_PREFIX_RT_NOP) {
4512                         cleanup_prefix_route(ifp, rt_expires,
4513                                 action == CLEANUP_PREFIX_RT_DEL);
4514                 }
4515         }
4516
4517         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4518                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4519                         valid_lft = prefered_lft = 0;
4520                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4521                                  !was_managetempaddr, jiffies);
4522         }
4523
4524         addrconf_verify_rtnl();
4525
4526         return 0;
4527 }
4528
4529 static int
4530 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4531                   struct netlink_ext_ack *extack)
4532 {
4533         struct net *net = sock_net(skb->sk);
4534         struct ifaddrmsg *ifm;
4535         struct nlattr *tb[IFA_MAX+1];
4536         struct in6_addr *pfx, *peer_pfx;
4537         struct inet6_ifaddr *ifa;
4538         struct net_device *dev;
4539         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4540         u32 ifa_flags;
4541         int err;
4542
4543         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4544                           extack);
4545         if (err < 0)
4546                 return err;
4547
4548         ifm = nlmsg_data(nlh);
4549         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4550         if (!pfx)
4551                 return -EINVAL;
4552
4553         if (tb[IFA_CACHEINFO]) {
4554                 struct ifa_cacheinfo *ci;
4555
4556                 ci = nla_data(tb[IFA_CACHEINFO]);
4557                 valid_lft = ci->ifa_valid;
4558                 preferred_lft = ci->ifa_prefered;
4559         } else {
4560                 preferred_lft = INFINITY_LIFE_TIME;
4561                 valid_lft = INFINITY_LIFE_TIME;
4562         }
4563
4564         dev =  __dev_get_by_index(net, ifm->ifa_index);
4565         if (!dev)
4566                 return -ENODEV;
4567
4568         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4569
4570         /* We ignore other flags so far. */
4571         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4572                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4573
4574         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4575         if (!ifa) {
4576                 /*
4577                  * It would be best to check for !NLM_F_CREATE here but
4578                  * userspace already relies on not having to provide this.
4579                  */
4580                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4581                                       ifm->ifa_prefixlen, ifa_flags,
4582                                       preferred_lft, valid_lft);
4583         }
4584
4585         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4586             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4587                 err = -EEXIST;
4588         else
4589                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4590
4591         in6_ifa_put(ifa);
4592
4593         return err;
4594 }
4595
4596 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4597                           u8 scope, int ifindex)
4598 {
4599         struct ifaddrmsg *ifm;
4600
4601         ifm = nlmsg_data(nlh);
4602         ifm->ifa_family = AF_INET6;
4603         ifm->ifa_prefixlen = prefixlen;
4604         ifm->ifa_flags = flags;
4605         ifm->ifa_scope = scope;
4606         ifm->ifa_index = ifindex;
4607 }
4608
4609 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4610                          unsigned long tstamp, u32 preferred, u32 valid)
4611 {
4612         struct ifa_cacheinfo ci;
4613
4614         ci.cstamp = cstamp_delta(cstamp);
4615         ci.tstamp = cstamp_delta(tstamp);
4616         ci.ifa_prefered = preferred;
4617         ci.ifa_valid = valid;
4618
4619         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4620 }
4621
4622 static inline int rt_scope(int ifa_scope)
4623 {
4624         if (ifa_scope & IFA_HOST)
4625                 return RT_SCOPE_HOST;
4626         else if (ifa_scope & IFA_LINK)
4627                 return RT_SCOPE_LINK;
4628         else if (ifa_scope & IFA_SITE)
4629                 return RT_SCOPE_SITE;
4630         else
4631                 return RT_SCOPE_UNIVERSE;
4632 }
4633
4634 static inline int inet6_ifaddr_msgsize(void)
4635 {
4636         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4637                + nla_total_size(16) /* IFA_LOCAL */
4638                + nla_total_size(16) /* IFA_ADDRESS */
4639                + nla_total_size(sizeof(struct ifa_cacheinfo))
4640                + nla_total_size(4)  /* IFA_FLAGS */;
4641 }
4642
4643 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4644                              u32 portid, u32 seq, int event, unsigned int flags)
4645 {
4646         struct nlmsghdr  *nlh;
4647         u32 preferred, valid;
4648
4649         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4650         if (!nlh)
4651                 return -EMSGSIZE;
4652
4653         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4654                       ifa->idev->dev->ifindex);
4655
4656         if (!((ifa->flags&IFA_F_PERMANENT) &&
4657               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4658                 preferred = ifa->prefered_lft;
4659                 valid = ifa->valid_lft;
4660                 if (preferred != INFINITY_LIFE_TIME) {
4661                         long tval = (jiffies - ifa->tstamp)/HZ;
4662                         if (preferred > tval)
4663                                 preferred -= tval;
4664                         else
4665                                 preferred = 0;
4666                         if (valid != INFINITY_LIFE_TIME) {
4667                                 if (valid > tval)
4668                                         valid -= tval;
4669                                 else
4670                                         valid = 0;
4671                         }
4672                 }
4673         } else {
4674                 preferred = INFINITY_LIFE_TIME;
4675                 valid = INFINITY_LIFE_TIME;
4676         }
4677
4678         if (!ipv6_addr_any(&ifa->peer_addr)) {
4679                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4680                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4681                         goto error;
4682         } else
4683                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4684                         goto error;
4685
4686         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4687                 goto error;
4688
4689         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4690                 goto error;
4691
4692         nlmsg_end(skb, nlh);
4693         return 0;
4694
4695 error:
4696         nlmsg_cancel(skb, nlh);
4697         return -EMSGSIZE;
4698 }
4699
4700 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4701                                 u32 portid, u32 seq, int event, u16 flags)
4702 {
4703         struct nlmsghdr  *nlh;
4704         u8 scope = RT_SCOPE_UNIVERSE;
4705         int ifindex = ifmca->idev->dev->ifindex;
4706
4707         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4708                 scope = RT_SCOPE_SITE;
4709
4710         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4711         if (!nlh)
4712                 return -EMSGSIZE;
4713
4714         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4715         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4716             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4717                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4718                 nlmsg_cancel(skb, nlh);
4719                 return -EMSGSIZE;
4720         }
4721
4722         nlmsg_end(skb, nlh);
4723         return 0;
4724 }
4725
4726 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4727                                 u32 portid, u32 seq, int event, unsigned int flags)
4728 {
4729         struct nlmsghdr  *nlh;
4730         u8 scope = RT_SCOPE_UNIVERSE;
4731         int ifindex = ifaca->aca_idev->dev->ifindex;
4732
4733         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4734                 scope = RT_SCOPE_SITE;
4735
4736         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4737         if (!nlh)
4738                 return -EMSGSIZE;
4739
4740         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4741         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4742             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4743                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4744                 nlmsg_cancel(skb, nlh);
4745                 return -EMSGSIZE;
4746         }
4747
4748         nlmsg_end(skb, nlh);
4749         return 0;
4750 }
4751
4752 enum addr_type_t {
4753         UNICAST_ADDR,
4754         MULTICAST_ADDR,
4755         ANYCAST_ADDR,
4756 };
4757
4758 /* called with rcu_read_lock() */
4759 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4760                           struct netlink_callback *cb, enum addr_type_t type,
4761                           int s_ip_idx, int *p_ip_idx)
4762 {
4763         struct ifmcaddr6 *ifmca;
4764         struct ifacaddr6 *ifaca;
4765         int err = 1;
4766         int ip_idx = *p_ip_idx;
4767
4768         read_lock_bh(&idev->lock);
4769         switch (type) {
4770         case UNICAST_ADDR: {
4771                 struct inet6_ifaddr *ifa;
4772
4773                 /* unicast address incl. temp addr */
4774                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4775                         if (++ip_idx < s_ip_idx)
4776                                 continue;
4777                         err = inet6_fill_ifaddr(skb, ifa,
4778                                                 NETLINK_CB(cb->skb).portid,
4779                                                 cb->nlh->nlmsg_seq,
4780                                                 RTM_NEWADDR,
4781                                                 NLM_F_MULTI);
4782                         if (err < 0)
4783                                 break;
4784                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4785                 }
4786                 break;
4787         }
4788         case MULTICAST_ADDR:
4789                 /* multicast address */
4790                 for (ifmca = idev->mc_list; ifmca;
4791                      ifmca = ifmca->next, ip_idx++) {
4792                         if (ip_idx < s_ip_idx)
4793                                 continue;
4794                         err = inet6_fill_ifmcaddr(skb, ifmca,
4795                                                   NETLINK_CB(cb->skb).portid,
4796                                                   cb->nlh->nlmsg_seq,
4797                                                   RTM_GETMULTICAST,
4798                                                   NLM_F_MULTI);
4799                         if (err < 0)
4800                                 break;
4801                 }
4802                 break;
4803         case ANYCAST_ADDR:
4804                 /* anycast address */
4805                 for (ifaca = idev->ac_list; ifaca;
4806                      ifaca = ifaca->aca_next, ip_idx++) {
4807                         if (ip_idx < s_ip_idx)
4808                                 continue;
4809                         err = inet6_fill_ifacaddr(skb, ifaca,
4810                                                   NETLINK_CB(cb->skb).portid,
4811                                                   cb->nlh->nlmsg_seq,
4812                                                   RTM_GETANYCAST,
4813                                                   NLM_F_MULTI);
4814                         if (err < 0)
4815                                 break;
4816                 }
4817                 break;
4818         default:
4819                 break;
4820         }
4821         read_unlock_bh(&idev->lock);
4822         *p_ip_idx = ip_idx;
4823         return err;
4824 }
4825
4826 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4827                            enum addr_type_t type)
4828 {
4829         struct net *net = sock_net(skb->sk);
4830         int h, s_h;
4831         int idx, ip_idx;
4832         int s_idx, s_ip_idx;
4833         struct net_device *dev;
4834         struct inet6_dev *idev;
4835         struct hlist_head *head;
4836
4837         s_h = cb->args[0];
4838         s_idx = idx = cb->args[1];
4839         s_ip_idx = ip_idx = cb->args[2];
4840
4841         rcu_read_lock();
4842         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4843         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4844                 idx = 0;
4845                 head = &net->dev_index_head[h];
4846                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4847                         if (idx < s_idx)
4848                                 goto cont;
4849                         if (h > s_h || idx > s_idx)
4850                                 s_ip_idx = 0;
4851                         ip_idx = 0;
4852                         idev = __in6_dev_get(dev);
4853                         if (!idev)
4854                                 goto cont;
4855
4856                         if (in6_dump_addrs(idev, skb, cb, type,
4857                                            s_ip_idx, &ip_idx) < 0)
4858                                 goto done;
4859 cont:
4860                         idx++;
4861                 }
4862         }
4863 done:
4864         rcu_read_unlock();
4865         cb->args[0] = h;
4866         cb->args[1] = idx;
4867         cb->args[2] = ip_idx;
4868
4869         return skb->len;
4870 }
4871
4872 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4873 {
4874         enum addr_type_t type = UNICAST_ADDR;
4875
4876         return inet6_dump_addr(skb, cb, type);
4877 }
4878
4879 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4880 {
4881         enum addr_type_t type = MULTICAST_ADDR;
4882
4883         return inet6_dump_addr(skb, cb, type);
4884 }
4885
4886
4887 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4888 {
4889         enum addr_type_t type = ANYCAST_ADDR;
4890
4891         return inet6_dump_addr(skb, cb, type);
4892 }
4893
4894 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4895                              struct netlink_ext_ack *extack)
4896 {
4897         struct net *net = sock_net(in_skb->sk);
4898         struct ifaddrmsg *ifm;
4899         struct nlattr *tb[IFA_MAX+1];
4900         struct in6_addr *addr = NULL, *peer;
4901         struct net_device *dev = NULL;
4902         struct inet6_ifaddr *ifa;
4903         struct sk_buff *skb;
4904         int err;
4905
4906         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4907                           extack);
4908         if (err < 0)
4909                 goto errout;
4910
4911         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4912         if (!addr) {
4913                 err = -EINVAL;
4914                 goto errout;
4915         }
4916
4917         ifm = nlmsg_data(nlh);
4918         if (ifm->ifa_index)
4919                 dev = __dev_get_by_index(net, ifm->ifa_index);
4920
4921         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4922         if (!ifa) {
4923                 err = -EADDRNOTAVAIL;
4924                 goto errout;
4925         }
4926
4927         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4928         if (!skb) {
4929                 err = -ENOBUFS;
4930                 goto errout_ifa;
4931         }
4932
4933         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4934                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4935         if (err < 0) {
4936                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4937                 WARN_ON(err == -EMSGSIZE);
4938                 kfree_skb(skb);
4939                 goto errout_ifa;
4940         }
4941         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4942 errout_ifa:
4943         in6_ifa_put(ifa);
4944 errout:
4945         return err;
4946 }
4947
4948 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4949 {
4950         struct sk_buff *skb;
4951         struct net *net = dev_net(ifa->idev->dev);
4952         int err = -ENOBUFS;
4953
4954         /* Don't send DELADDR notification for TENTATIVE address,
4955          * since NEWADDR notification is sent only after removing
4956          * TENTATIVE flag, if DAD has not failed.
4957          */
4958         if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
4959             event == RTM_DELADDR)
4960                 return;
4961
4962         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4963         if (!skb)
4964                 goto errout;
4965
4966         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4967         if (err < 0) {
4968                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4969                 WARN_ON(err == -EMSGSIZE);
4970                 kfree_skb(skb);
4971                 goto errout;
4972         }
4973         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4974         return;
4975 errout:
4976         if (err < 0)
4977                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4978 }
4979
4980 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4981                                 __s32 *array, int bytes)
4982 {
4983         BUG_ON(bytes < (DEVCONF_MAX * 4));
4984
4985         memset(array, 0, bytes);
4986         array[DEVCONF_FORWARDING] = cnf->forwarding;
4987         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4988         array[DEVCONF_MTU6] = cnf->mtu6;
4989         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4990         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4991         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4992         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4993         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4994         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4995                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4996         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4997                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4998         array[DEVCONF_RTR_SOLICIT_DELAY] =
4999                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5000         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5001         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5002                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5003         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5004                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5005         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5006         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5007         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5008         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5009         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5010         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5011         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5012         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5013         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5014 #ifdef CONFIG_IPV6_ROUTER_PREF
5015         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5016         array[DEVCONF_RTR_PROBE_INTERVAL] =
5017                 jiffies_to_msecs(cnf->rtr_probe_interval);
5018 #ifdef CONFIG_IPV6_ROUTE_INFO
5019         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5020         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5021 #endif
5022 #endif
5023         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5024         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5025 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5026         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5027         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5028 #endif
5029 #ifdef CONFIG_IPV6_MROUTE
5030         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5031 #endif
5032         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5033         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5034         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5035         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5036         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5037         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5038         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5039         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5040         /* we omit DEVCONF_STABLE_SECRET for now */
5041         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5042         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5043         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5044         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5045         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5046 #ifdef CONFIG_IPV6_SEG6_HMAC
5047         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5048 #endif
5049         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5050         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5051         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5052 }
5053
5054 static inline size_t inet6_ifla6_size(void)
5055 {
5056         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5057              + nla_total_size(sizeof(struct ifla_cacheinfo))
5058              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5059              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5060              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5061              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5062 }
5063
5064 static inline size_t inet6_if_nlmsg_size(void)
5065 {
5066         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5067                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5068                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5069                + nla_total_size(4) /* IFLA_MTU */
5070                + nla_total_size(4) /* IFLA_LINK */
5071                + nla_total_size(1) /* IFLA_OPERSTATE */
5072                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5073 }
5074
5075 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5076                                         int bytes)
5077 {
5078         int i;
5079         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5080         BUG_ON(pad < 0);
5081
5082         /* Use put_unaligned() because stats may not be aligned for u64. */
5083         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5084         for (i = 1; i < ICMP6_MIB_MAX; i++)
5085                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5086
5087         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5088 }
5089
5090 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5091                                         int bytes, size_t syncpoff)
5092 {
5093         int i, c;
5094         u64 buff[IPSTATS_MIB_MAX];
5095         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5096
5097         BUG_ON(pad < 0);
5098
5099         memset(buff, 0, sizeof(buff));
5100         buff[0] = IPSTATS_MIB_MAX;
5101
5102         for_each_possible_cpu(c) {
5103                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5104                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5105         }
5106
5107         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5108         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5109 }
5110
5111 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5112                              int bytes)
5113 {
5114         switch (attrtype) {
5115         case IFLA_INET6_STATS:
5116                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5117                                      offsetof(struct ipstats_mib, syncp));
5118                 break;
5119         case IFLA_INET6_ICMP6STATS:
5120                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5121                 break;
5122         }
5123 }
5124
5125 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5126                                   u32 ext_filter_mask)
5127 {
5128         struct nlattr *nla;
5129         struct ifla_cacheinfo ci;
5130
5131         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5132                 goto nla_put_failure;
5133         ci.max_reasm_len = IPV6_MAXPLEN;
5134         ci.tstamp = cstamp_delta(idev->tstamp);
5135         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5136         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5137         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5138                 goto nla_put_failure;
5139         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5140         if (!nla)
5141                 goto nla_put_failure;
5142         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5143
5144         /* XXX - MC not implemented */
5145
5146         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5147                 return 0;
5148
5149         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5150         if (!nla)
5151                 goto nla_put_failure;
5152         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5153
5154         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5155         if (!nla)
5156                 goto nla_put_failure;
5157         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5158
5159         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5160         if (!nla)
5161                 goto nla_put_failure;
5162
5163         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5164                 goto nla_put_failure;
5165
5166         read_lock_bh(&idev->lock);
5167         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5168         read_unlock_bh(&idev->lock);
5169
5170         return 0;
5171
5172 nla_put_failure:
5173         return -EMSGSIZE;
5174 }
5175
5176 static size_t inet6_get_link_af_size(const struct net_device *dev,
5177                                      u32 ext_filter_mask)
5178 {
5179         if (!__in6_dev_get(dev))
5180                 return 0;
5181
5182         return inet6_ifla6_size();
5183 }
5184
5185 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5186                               u32 ext_filter_mask)
5187 {
5188         struct inet6_dev *idev = __in6_dev_get(dev);
5189
5190         if (!idev)
5191                 return -ENODATA;
5192
5193         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5194                 return -EMSGSIZE;
5195
5196         return 0;
5197 }
5198
5199 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5200 {
5201         struct inet6_ifaddr *ifp;
5202         struct net_device *dev = idev->dev;
5203         bool clear_token, update_rs = false;
5204         struct in6_addr ll_addr;
5205
5206         ASSERT_RTNL();
5207
5208         if (!token)
5209                 return -EINVAL;
5210         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5211                 return -EINVAL;
5212         if (!ipv6_accept_ra(idev))
5213                 return -EINVAL;
5214         if (idev->cnf.rtr_solicits == 0)
5215                 return -EINVAL;
5216
5217         write_lock_bh(&idev->lock);
5218
5219         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5220         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5221
5222         write_unlock_bh(&idev->lock);
5223
5224         clear_token = ipv6_addr_any(token);
5225         if (clear_token)
5226                 goto update_lft;
5227
5228         if (!idev->dead && (idev->if_flags & IF_READY) &&
5229             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5230                              IFA_F_OPTIMISTIC)) {
5231                 /* If we're not ready, then normal ifup will take care
5232                  * of this. Otherwise, we need to request our rs here.
5233                  */
5234                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5235                 update_rs = true;
5236         }
5237
5238 update_lft:
5239         write_lock_bh(&idev->lock);
5240
5241         if (update_rs) {
5242                 idev->if_flags |= IF_RS_SENT;
5243                 idev->rs_interval = rfc3315_s14_backoff_init(
5244                         idev->cnf.rtr_solicit_interval);
5245                 idev->rs_probes = 1;
5246                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5247         }
5248
5249         /* Well, that's kinda nasty ... */
5250         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5251                 spin_lock(&ifp->lock);
5252                 if (ifp->tokenized) {
5253                         ifp->valid_lft = 0;
5254                         ifp->prefered_lft = 0;
5255                 }
5256                 spin_unlock(&ifp->lock);
5257         }
5258
5259         write_unlock_bh(&idev->lock);
5260         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5261         addrconf_verify_rtnl();
5262         return 0;
5263 }
5264
5265 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5266         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5267         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5268 };
5269
5270 static int inet6_validate_link_af(const struct net_device *dev,
5271                                   const struct nlattr *nla)
5272 {
5273         struct nlattr *tb[IFLA_INET6_MAX + 1];
5274
5275         if (dev && !__in6_dev_get(dev))
5276                 return -EAFNOSUPPORT;
5277
5278         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5279                                 NULL);
5280 }
5281
5282 static int check_addr_gen_mode(int mode)
5283 {
5284         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5285             mode != IN6_ADDR_GEN_MODE_NONE &&
5286             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5287             mode != IN6_ADDR_GEN_MODE_RANDOM)
5288                 return -EINVAL;
5289         return 1;
5290 }
5291
5292 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5293                                 int mode)
5294 {
5295         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5296             !idev->cnf.stable_secret.initialized &&
5297             !net->ipv6.devconf_dflt->stable_secret.initialized)
5298                 return -EINVAL;
5299         return 1;
5300 }
5301
5302 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5303 {
5304         int err = -EINVAL;
5305         struct inet6_dev *idev = __in6_dev_get(dev);
5306         struct nlattr *tb[IFLA_INET6_MAX + 1];
5307
5308         if (!idev)
5309                 return -EAFNOSUPPORT;
5310
5311         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5312                 BUG();
5313
5314         if (tb[IFLA_INET6_TOKEN]) {
5315                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5316                 if (err)
5317                         return err;
5318         }
5319
5320         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5321                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5322
5323                 if (check_addr_gen_mode(mode) < 0 ||
5324                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5325                         return -EINVAL;
5326
5327                 idev->cnf.addr_gen_mode = mode;
5328                 err = 0;
5329         }
5330
5331         return err;
5332 }
5333
5334 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5335                              u32 portid, u32 seq, int event, unsigned int flags)
5336 {
5337         struct net_device *dev = idev->dev;
5338         struct ifinfomsg *hdr;
5339         struct nlmsghdr *nlh;
5340         void *protoinfo;
5341
5342         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5343         if (!nlh)
5344                 return -EMSGSIZE;
5345
5346         hdr = nlmsg_data(nlh);
5347         hdr->ifi_family = AF_INET6;
5348         hdr->__ifi_pad = 0;
5349         hdr->ifi_type = dev->type;
5350         hdr->ifi_index = dev->ifindex;
5351         hdr->ifi_flags = dev_get_flags(dev);
5352         hdr->ifi_change = 0;
5353
5354         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5355             (dev->addr_len &&
5356              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5357             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5358             (dev->ifindex != dev_get_iflink(dev) &&
5359              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5360             nla_put_u8(skb, IFLA_OPERSTATE,
5361                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5362                 goto nla_put_failure;
5363         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5364         if (!protoinfo)
5365                 goto nla_put_failure;
5366
5367         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5368                 goto nla_put_failure;
5369
5370         nla_nest_end(skb, protoinfo);
5371         nlmsg_end(skb, nlh);
5372         return 0;
5373
5374 nla_put_failure:
5375         nlmsg_cancel(skb, nlh);
5376         return -EMSGSIZE;
5377 }
5378
5379 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5380 {
5381         struct net *net = sock_net(skb->sk);
5382         int h, s_h;
5383         int idx = 0, s_idx;
5384         struct net_device *dev;
5385         struct inet6_dev *idev;
5386         struct hlist_head *head;
5387
5388         s_h = cb->args[0];
5389         s_idx = cb->args[1];
5390
5391         rcu_read_lock();
5392         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5393                 idx = 0;
5394                 head = &net->dev_index_head[h];
5395                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5396                         if (idx < s_idx)
5397                                 goto cont;
5398                         idev = __in6_dev_get(dev);
5399                         if (!idev)
5400                                 goto cont;
5401                         if (inet6_fill_ifinfo(skb, idev,
5402                                               NETLINK_CB(cb->skb).portid,
5403                                               cb->nlh->nlmsg_seq,
5404                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5405                                 goto out;
5406 cont:
5407                         idx++;
5408                 }
5409         }
5410 out:
5411         rcu_read_unlock();
5412         cb->args[1] = idx;
5413         cb->args[0] = h;
5414
5415         return skb->len;
5416 }
5417
5418 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5419 {
5420         struct sk_buff *skb;
5421         struct net *net = dev_net(idev->dev);
5422         int err = -ENOBUFS;
5423
5424         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5425         if (!skb)
5426                 goto errout;
5427
5428         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5429         if (err < 0) {
5430                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5431                 WARN_ON(err == -EMSGSIZE);
5432                 kfree_skb(skb);
5433                 goto errout;
5434         }
5435         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5436         return;
5437 errout:
5438         if (err < 0)
5439                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5440 }
5441
5442 static inline size_t inet6_prefix_nlmsg_size(void)
5443 {
5444         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5445                + nla_total_size(sizeof(struct in6_addr))
5446                + nla_total_size(sizeof(struct prefix_cacheinfo));
5447 }
5448
5449 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5450                              struct prefix_info *pinfo, u32 portid, u32 seq,
5451                              int event, unsigned int flags)
5452 {
5453         struct prefixmsg *pmsg;
5454         struct nlmsghdr *nlh;
5455         struct prefix_cacheinfo ci;
5456
5457         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5458         if (!nlh)
5459                 return -EMSGSIZE;
5460
5461         pmsg = nlmsg_data(nlh);
5462         pmsg->prefix_family = AF_INET6;
5463         pmsg->prefix_pad1 = 0;
5464         pmsg->prefix_pad2 = 0;
5465         pmsg->prefix_ifindex = idev->dev->ifindex;
5466         pmsg->prefix_len = pinfo->prefix_len;
5467         pmsg->prefix_type = pinfo->type;
5468         pmsg->prefix_pad3 = 0;
5469         pmsg->prefix_flags = 0;
5470         if (pinfo->onlink)
5471                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5472         if (pinfo->autoconf)
5473                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5474
5475         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5476                 goto nla_put_failure;
5477         ci.preferred_time = ntohl(pinfo->prefered);
5478         ci.valid_time = ntohl(pinfo->valid);
5479         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5480                 goto nla_put_failure;
5481         nlmsg_end(skb, nlh);
5482         return 0;
5483
5484 nla_put_failure:
5485         nlmsg_cancel(skb, nlh);
5486         return -EMSGSIZE;
5487 }
5488
5489 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5490                          struct prefix_info *pinfo)
5491 {
5492         struct sk_buff *skb;
5493         struct net *net = dev_net(idev->dev);
5494         int err = -ENOBUFS;
5495
5496         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5497         if (!skb)
5498                 goto errout;
5499
5500         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5501         if (err < 0) {
5502                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5503                 WARN_ON(err == -EMSGSIZE);
5504                 kfree_skb(skb);
5505                 goto errout;
5506         }
5507         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5508         return;
5509 errout:
5510         if (err < 0)
5511                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5512 }
5513
5514 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5515 {
5516         struct net *net = dev_net(ifp->idev->dev);
5517
5518         if (event)
5519                 ASSERT_RTNL();
5520
5521         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5522
5523         switch (event) {
5524         case RTM_NEWADDR:
5525                 /*
5526                  * If the address was optimistic
5527                  * we inserted the route at the start of
5528                  * our DAD process, so we don't need
5529                  * to do it again
5530                  */
5531                 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5532                         ip6_ins_rt(ifp->rt);
5533                 if (ifp->idev->cnf.forwarding)
5534                         addrconf_join_anycast(ifp);
5535                 if (!ipv6_addr_any(&ifp->peer_addr))
5536                         addrconf_prefix_route(&ifp->peer_addr, 128,
5537                                               ifp->idev->dev, 0, 0);
5538                 break;
5539         case RTM_DELADDR:
5540                 if (ifp->idev->cnf.forwarding)
5541                         addrconf_leave_anycast(ifp);
5542                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5543                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5544                         struct rt6_info *rt;
5545
5546                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5547                                                        ifp->idev->dev, 0, 0);
5548                         if (rt)
5549                                 ip6_del_rt(rt);
5550                 }
5551                 if (ifp->rt) {
5552                         if (dst_hold_safe(&ifp->rt->dst))
5553                                 ip6_del_rt(ifp->rt);
5554                 }
5555                 rt_genid_bump_ipv6(net);
5556                 break;
5557         }
5558         atomic_inc(&net->ipv6.dev_addr_genid);
5559 }
5560
5561 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5562 {
5563         rcu_read_lock_bh();
5564         if (likely(ifp->idev->dead == 0))
5565                 __ipv6_ifa_notify(event, ifp);
5566         rcu_read_unlock_bh();
5567 }
5568
5569 #ifdef CONFIG_SYSCTL
5570
5571 static
5572 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5573                            void __user *buffer, size_t *lenp, loff_t *ppos)
5574 {
5575         int *valp = ctl->data;
5576         int val = *valp;
5577         loff_t pos = *ppos;
5578         struct ctl_table lctl;
5579         int ret;
5580
5581         /*
5582          * ctl->data points to idev->cnf.forwarding, we should
5583          * not modify it until we get the rtnl lock.
5584          */
5585         lctl = *ctl;
5586         lctl.data = &val;
5587
5588         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5589
5590         if (write)
5591                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5592         if (ret)
5593                 *ppos = pos;
5594         return ret;
5595 }
5596
5597 static
5598 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5599                         void __user *buffer, size_t *lenp, loff_t *ppos)
5600 {
5601         struct inet6_dev *idev = ctl->extra1;
5602         int min_mtu = IPV6_MIN_MTU;
5603         struct ctl_table lctl;
5604
5605         lctl = *ctl;
5606         lctl.extra1 = &min_mtu;
5607         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5608
5609         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5610 }
5611
5612 static void dev_disable_change(struct inet6_dev *idev)
5613 {
5614         struct netdev_notifier_info info;
5615
5616         if (!idev || !idev->dev)
5617                 return;
5618
5619         netdev_notifier_info_init(&info, idev->dev);
5620         if (idev->cnf.disable_ipv6)
5621                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5622         else
5623                 addrconf_notify(NULL, NETDEV_UP, &info);
5624 }
5625
5626 static void addrconf_disable_change(struct net *net, __s32 newf)
5627 {
5628         struct net_device *dev;
5629         struct inet6_dev *idev;
5630
5631         for_each_netdev(net, dev) {
5632                 idev = __in6_dev_get(dev);
5633                 if (idev) {
5634                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5635                         idev->cnf.disable_ipv6 = newf;
5636                         if (changed)
5637                                 dev_disable_change(idev);
5638                 }
5639         }
5640 }
5641
5642 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5643 {
5644         struct net *net;
5645         int old;
5646
5647         if (!rtnl_trylock())
5648                 return restart_syscall();
5649
5650         net = (struct net *)table->extra2;
5651         old = *p;
5652         *p = newf;
5653
5654         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5655                 rtnl_unlock();
5656                 return 0;
5657         }
5658
5659         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5660                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5661                 addrconf_disable_change(net, newf);
5662         } else if ((!newf) ^ (!old))
5663                 dev_disable_change((struct inet6_dev *)table->extra1);
5664
5665         rtnl_unlock();
5666         return 0;
5667 }
5668
5669 static
5670 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5671                             void __user *buffer, size_t *lenp, loff_t *ppos)
5672 {
5673         int *valp = ctl->data;
5674         int val = *valp;
5675         loff_t pos = *ppos;
5676         struct ctl_table lctl;
5677         int ret;
5678
5679         /*
5680          * ctl->data points to idev->cnf.disable_ipv6, we should
5681          * not modify it until we get the rtnl lock.
5682          */
5683         lctl = *ctl;
5684         lctl.data = &val;
5685
5686         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5687
5688         if (write)
5689                 ret = addrconf_disable_ipv6(ctl, valp, val);
5690         if (ret)
5691                 *ppos = pos;
5692         return ret;
5693 }
5694
5695 static
5696 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5697                               void __user *buffer, size_t *lenp, loff_t *ppos)
5698 {
5699         int *valp = ctl->data;
5700         int ret;
5701         int old, new;
5702
5703         old = *valp;
5704         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5705         new = *valp;
5706
5707         if (write && old != new) {
5708                 struct net *net = ctl->extra2;
5709
5710                 if (!rtnl_trylock())
5711                         return restart_syscall();
5712
5713                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5714                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5715                                                      NETCONFA_PROXY_NEIGH,
5716                                                      NETCONFA_IFINDEX_DEFAULT,
5717                                                      net->ipv6.devconf_dflt);
5718                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5719                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5720                                                      NETCONFA_PROXY_NEIGH,
5721                                                      NETCONFA_IFINDEX_ALL,
5722                                                      net->ipv6.devconf_all);
5723                 else {
5724                         struct inet6_dev *idev = ctl->extra1;
5725
5726                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5727                                                      NETCONFA_PROXY_NEIGH,
5728                                                      idev->dev->ifindex,
5729                                                      &idev->cnf);
5730                 }
5731                 rtnl_unlock();
5732         }
5733
5734         return ret;
5735 }
5736
5737 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5738                                          void __user *buffer, size_t *lenp,
5739                                          loff_t *ppos)
5740 {
5741         int ret = 0;
5742         int new_val;
5743         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5744         struct net *net = (struct net *)ctl->extra2;
5745
5746         if (!rtnl_trylock())
5747                 return restart_syscall();
5748
5749         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5750
5751         if (write) {
5752                 new_val = *((int *)ctl->data);
5753
5754                 if (check_addr_gen_mode(new_val) < 0) {
5755                         ret = -EINVAL;
5756                         goto out;
5757                 }
5758
5759                 /* request for default */
5760                 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5761                         ipv6_devconf_dflt.addr_gen_mode = new_val;
5762
5763                 /* request for individual net device */
5764                 } else {
5765                         if (!idev)
5766                                 goto out;
5767
5768                         if (check_stable_privacy(idev, net, new_val) < 0) {
5769                                 ret = -EINVAL;
5770                                 goto out;
5771                         }
5772
5773                         if (idev->cnf.addr_gen_mode != new_val) {
5774                                 idev->cnf.addr_gen_mode = new_val;
5775                                 addrconf_dev_config(idev->dev);
5776                         }
5777                 }
5778         }
5779
5780 out:
5781         rtnl_unlock();
5782
5783         return ret;
5784 }
5785
5786 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5787                                          void __user *buffer, size_t *lenp,
5788                                          loff_t *ppos)
5789 {
5790         int err;
5791         struct in6_addr addr;
5792         char str[IPV6_MAX_STRLEN];
5793         struct ctl_table lctl = *ctl;
5794         struct net *net = ctl->extra2;
5795         struct ipv6_stable_secret *secret = ctl->data;
5796
5797         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5798                 return -EIO;
5799
5800         lctl.maxlen = IPV6_MAX_STRLEN;
5801         lctl.data = str;
5802
5803         if (!rtnl_trylock())
5804                 return restart_syscall();
5805
5806         if (!write && !secret->initialized) {
5807                 err = -EIO;
5808                 goto out;
5809         }
5810
5811         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5812         if (err >= sizeof(str)) {
5813                 err = -EIO;
5814                 goto out;
5815         }
5816
5817         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5818         if (err || !write)
5819                 goto out;
5820
5821         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5822                 err = -EIO;
5823                 goto out;
5824         }
5825
5826         secret->initialized = true;
5827         secret->secret = addr;
5828
5829         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5830                 struct net_device *dev;
5831
5832                 for_each_netdev(net, dev) {
5833                         struct inet6_dev *idev = __in6_dev_get(dev);
5834
5835                         if (idev) {
5836                                 idev->cnf.addr_gen_mode =
5837                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5838                         }
5839                 }
5840         } else {
5841                 struct inet6_dev *idev = ctl->extra1;
5842
5843                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5844         }
5845
5846 out:
5847         rtnl_unlock();
5848
5849         return err;
5850 }
5851
5852 static
5853 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5854                                                 int write,
5855                                                 void __user *buffer,
5856                                                 size_t *lenp,
5857                                                 loff_t *ppos)
5858 {
5859         int *valp = ctl->data;
5860         int val = *valp;
5861         loff_t pos = *ppos;
5862         struct ctl_table lctl;
5863         int ret;
5864
5865         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5866          * we should not modify it until we get the rtnl lock.
5867          */
5868         lctl = *ctl;
5869         lctl.data = &val;
5870
5871         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5872
5873         if (write)
5874                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5875         if (ret)
5876                 *ppos = pos;
5877         return ret;
5878 }
5879
5880 static
5881 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5882 {
5883         if (rt) {
5884                 if (action)
5885                         rt->dst.flags |= DST_NOPOLICY;
5886                 else
5887                         rt->dst.flags &= ~DST_NOPOLICY;
5888         }
5889 }
5890
5891 static
5892 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5893 {
5894         struct inet6_ifaddr *ifa;
5895
5896         read_lock_bh(&idev->lock);
5897         list_for_each_entry(ifa, &idev->addr_list, if_list) {
5898                 spin_lock(&ifa->lock);
5899                 if (ifa->rt) {
5900                         struct rt6_info *rt = ifa->rt;
5901                         struct fib6_table *table = rt->rt6i_table;
5902                         int cpu;
5903
5904                         read_lock(&table->tb6_lock);
5905                         addrconf_set_nopolicy(ifa->rt, val);
5906                         if (rt->rt6i_pcpu) {
5907                                 for_each_possible_cpu(cpu) {
5908                                         struct rt6_info **rtp;
5909
5910                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5911                                         addrconf_set_nopolicy(*rtp, val);
5912                                 }
5913                         }
5914                         read_unlock(&table->tb6_lock);
5915                 }
5916                 spin_unlock(&ifa->lock);
5917         }
5918         read_unlock_bh(&idev->lock);
5919 }
5920
5921 static
5922 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5923 {
5924         struct inet6_dev *idev;
5925         struct net *net;
5926
5927         if (!rtnl_trylock())
5928                 return restart_syscall();
5929
5930         *valp = val;
5931
5932         net = (struct net *)ctl->extra2;
5933         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5934                 rtnl_unlock();
5935                 return 0;
5936         }
5937
5938         if (valp == &net->ipv6.devconf_all->disable_policy)  {
5939                 struct net_device *dev;
5940
5941                 for_each_netdev(net, dev) {
5942                         idev = __in6_dev_get(dev);
5943                         if (idev)
5944                                 addrconf_disable_policy_idev(idev, val);
5945                 }
5946         } else {
5947                 idev = (struct inet6_dev *)ctl->extra1;
5948                 addrconf_disable_policy_idev(idev, val);
5949         }
5950
5951         rtnl_unlock();
5952         return 0;
5953 }
5954
5955 static
5956 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
5957                                    void __user *buffer, size_t *lenp,
5958                                    loff_t *ppos)
5959 {
5960         int *valp = ctl->data;
5961         int val = *valp;
5962         loff_t pos = *ppos;
5963         struct ctl_table lctl;
5964         int ret;
5965
5966         lctl = *ctl;
5967         lctl.data = &val;
5968         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5969
5970         if (write && (*valp != val))
5971                 ret = addrconf_disable_policy(ctl, valp, val);
5972
5973         if (ret)
5974                 *ppos = pos;
5975
5976         return ret;
5977 }
5978
5979 static int minus_one = -1;
5980 static const int one = 1;
5981 static const int two_five_five = 255;
5982
5983 static const struct ctl_table addrconf_sysctl[] = {
5984         {
5985                 .procname       = "forwarding",
5986                 .data           = &ipv6_devconf.forwarding,
5987                 .maxlen         = sizeof(int),
5988                 .mode           = 0644,
5989                 .proc_handler   = addrconf_sysctl_forward,
5990         },
5991         {
5992                 .procname       = "hop_limit",
5993                 .data           = &ipv6_devconf.hop_limit,
5994                 .maxlen         = sizeof(int),
5995                 .mode           = 0644,
5996                 .proc_handler   = proc_dointvec_minmax,
5997                 .extra1         = (void *)&one,
5998                 .extra2         = (void *)&two_five_five,
5999         },
6000         {
6001                 .procname       = "mtu",
6002                 .data           = &ipv6_devconf.mtu6,
6003                 .maxlen         = sizeof(int),
6004                 .mode           = 0644,
6005                 .proc_handler   = addrconf_sysctl_mtu,
6006         },
6007         {
6008                 .procname       = "accept_ra",
6009                 .data           = &ipv6_devconf.accept_ra,
6010                 .maxlen         = sizeof(int),
6011                 .mode           = 0644,
6012                 .proc_handler   = proc_dointvec,
6013         },
6014         {
6015                 .procname       = "accept_redirects",
6016                 .data           = &ipv6_devconf.accept_redirects,
6017                 .maxlen         = sizeof(int),
6018                 .mode           = 0644,
6019                 .proc_handler   = proc_dointvec,
6020         },
6021         {
6022                 .procname       = "autoconf",
6023                 .data           = &ipv6_devconf.autoconf,
6024                 .maxlen         = sizeof(int),
6025                 .mode           = 0644,
6026                 .proc_handler   = proc_dointvec,
6027         },
6028         {
6029                 .procname       = "dad_transmits",
6030                 .data           = &ipv6_devconf.dad_transmits,
6031                 .maxlen         = sizeof(int),
6032                 .mode           = 0644,
6033                 .proc_handler   = proc_dointvec,
6034         },
6035         {
6036                 .procname       = "router_solicitations",
6037                 .data           = &ipv6_devconf.rtr_solicits,
6038                 .maxlen         = sizeof(int),
6039                 .mode           = 0644,
6040                 .proc_handler   = proc_dointvec_minmax,
6041                 .extra1         = &minus_one,
6042         },
6043         {
6044                 .procname       = "router_solicitation_interval",
6045                 .data           = &ipv6_devconf.rtr_solicit_interval,
6046                 .maxlen         = sizeof(int),
6047                 .mode           = 0644,
6048                 .proc_handler   = proc_dointvec_jiffies,
6049         },
6050         {
6051                 .procname       = "router_solicitation_max_interval",
6052                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6053                 .maxlen         = sizeof(int),
6054                 .mode           = 0644,
6055                 .proc_handler   = proc_dointvec_jiffies,
6056         },
6057         {
6058                 .procname       = "router_solicitation_delay",
6059                 .data           = &ipv6_devconf.rtr_solicit_delay,
6060                 .maxlen         = sizeof(int),
6061                 .mode           = 0644,
6062                 .proc_handler   = proc_dointvec_jiffies,
6063         },
6064         {
6065                 .procname       = "force_mld_version",
6066                 .data           = &ipv6_devconf.force_mld_version,
6067                 .maxlen         = sizeof(int),
6068                 .mode           = 0644,
6069                 .proc_handler   = proc_dointvec,
6070         },
6071         {
6072                 .procname       = "mldv1_unsolicited_report_interval",
6073                 .data           =
6074                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6075                 .maxlen         = sizeof(int),
6076                 .mode           = 0644,
6077                 .proc_handler   = proc_dointvec_ms_jiffies,
6078         },
6079         {
6080                 .procname       = "mldv2_unsolicited_report_interval",
6081                 .data           =
6082                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6083                 .maxlen         = sizeof(int),
6084                 .mode           = 0644,
6085                 .proc_handler   = proc_dointvec_ms_jiffies,
6086         },
6087         {
6088                 .procname       = "use_tempaddr",
6089                 .data           = &ipv6_devconf.use_tempaddr,
6090                 .maxlen         = sizeof(int),
6091                 .mode           = 0644,
6092                 .proc_handler   = proc_dointvec,
6093         },
6094         {
6095                 .procname       = "temp_valid_lft",
6096                 .data           = &ipv6_devconf.temp_valid_lft,
6097                 .maxlen         = sizeof(int),
6098                 .mode           = 0644,
6099                 .proc_handler   = proc_dointvec,
6100         },
6101         {
6102                 .procname       = "temp_prefered_lft",
6103                 .data           = &ipv6_devconf.temp_prefered_lft,
6104                 .maxlen         = sizeof(int),
6105                 .mode           = 0644,
6106                 .proc_handler   = proc_dointvec,
6107         },
6108         {
6109                 .procname       = "regen_max_retry",
6110                 .data           = &ipv6_devconf.regen_max_retry,
6111                 .maxlen         = sizeof(int),
6112                 .mode           = 0644,
6113                 .proc_handler   = proc_dointvec,
6114         },
6115         {
6116                 .procname       = "max_desync_factor",
6117                 .data           = &ipv6_devconf.max_desync_factor,
6118                 .maxlen         = sizeof(int),
6119                 .mode           = 0644,
6120                 .proc_handler   = proc_dointvec,
6121         },
6122         {
6123                 .procname       = "max_addresses",
6124                 .data           = &ipv6_devconf.max_addresses,
6125                 .maxlen         = sizeof(int),
6126                 .mode           = 0644,
6127                 .proc_handler   = proc_dointvec,
6128         },
6129         {
6130                 .procname       = "accept_ra_defrtr",
6131                 .data           = &ipv6_devconf.accept_ra_defrtr,
6132                 .maxlen         = sizeof(int),
6133                 .mode           = 0644,
6134                 .proc_handler   = proc_dointvec,
6135         },
6136         {
6137                 .procname       = "accept_ra_min_hop_limit",
6138                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6139                 .maxlen         = sizeof(int),
6140                 .mode           = 0644,
6141                 .proc_handler   = proc_dointvec,
6142         },
6143         {
6144                 .procname       = "accept_ra_pinfo",
6145                 .data           = &ipv6_devconf.accept_ra_pinfo,
6146                 .maxlen         = sizeof(int),
6147                 .mode           = 0644,
6148                 .proc_handler   = proc_dointvec,
6149         },
6150 #ifdef CONFIG_IPV6_ROUTER_PREF
6151         {
6152                 .procname       = "accept_ra_rtr_pref",
6153                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6154                 .maxlen         = sizeof(int),
6155                 .mode           = 0644,
6156                 .proc_handler   = proc_dointvec,
6157         },
6158         {
6159                 .procname       = "router_probe_interval",
6160                 .data           = &ipv6_devconf.rtr_probe_interval,
6161                 .maxlen         = sizeof(int),
6162                 .mode           = 0644,
6163                 .proc_handler   = proc_dointvec_jiffies,
6164         },
6165 #ifdef CONFIG_IPV6_ROUTE_INFO
6166         {
6167                 .procname       = "accept_ra_rt_info_min_plen",
6168                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6169                 .maxlen         = sizeof(int),
6170                 .mode           = 0644,
6171                 .proc_handler   = proc_dointvec,
6172         },
6173         {
6174                 .procname       = "accept_ra_rt_info_max_plen",
6175                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6176                 .maxlen         = sizeof(int),
6177                 .mode           = 0644,
6178                 .proc_handler   = proc_dointvec,
6179         },
6180 #endif
6181 #endif
6182         {
6183                 .procname       = "proxy_ndp",
6184                 .data           = &ipv6_devconf.proxy_ndp,
6185                 .maxlen         = sizeof(int),
6186                 .mode           = 0644,
6187                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6188         },
6189         {
6190                 .procname       = "accept_source_route",
6191                 .data           = &ipv6_devconf.accept_source_route,
6192                 .maxlen         = sizeof(int),
6193                 .mode           = 0644,
6194                 .proc_handler   = proc_dointvec,
6195         },
6196 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6197         {
6198                 .procname       = "optimistic_dad",
6199                 .data           = &ipv6_devconf.optimistic_dad,
6200                 .maxlen         = sizeof(int),
6201                 .mode           = 0644,
6202                 .proc_handler   = proc_dointvec,
6203         },
6204         {
6205                 .procname       = "use_optimistic",
6206                 .data           = &ipv6_devconf.use_optimistic,
6207                 .maxlen         = sizeof(int),
6208                 .mode           = 0644,
6209                 .proc_handler   = proc_dointvec,
6210         },
6211 #endif
6212 #ifdef CONFIG_IPV6_MROUTE
6213         {
6214                 .procname       = "mc_forwarding",
6215                 .data           = &ipv6_devconf.mc_forwarding,
6216                 .maxlen         = sizeof(int),
6217                 .mode           = 0444,
6218                 .proc_handler   = proc_dointvec,
6219         },
6220 #endif
6221         {
6222                 .procname       = "disable_ipv6",
6223                 .data           = &ipv6_devconf.disable_ipv6,
6224                 .maxlen         = sizeof(int),
6225                 .mode           = 0644,
6226                 .proc_handler   = addrconf_sysctl_disable,
6227         },
6228         {
6229                 .procname       = "accept_dad",
6230                 .data           = &ipv6_devconf.accept_dad,
6231                 .maxlen         = sizeof(int),
6232                 .mode           = 0644,
6233                 .proc_handler   = proc_dointvec,
6234         },
6235         {
6236                 .procname       = "force_tllao",
6237                 .data           = &ipv6_devconf.force_tllao,
6238                 .maxlen         = sizeof(int),
6239                 .mode           = 0644,
6240                 .proc_handler   = proc_dointvec
6241         },
6242         {
6243                 .procname       = "ndisc_notify",
6244                 .data           = &ipv6_devconf.ndisc_notify,
6245                 .maxlen         = sizeof(int),
6246                 .mode           = 0644,
6247                 .proc_handler   = proc_dointvec
6248         },
6249         {
6250                 .procname       = "suppress_frag_ndisc",
6251                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6252                 .maxlen         = sizeof(int),
6253                 .mode           = 0644,
6254                 .proc_handler   = proc_dointvec
6255         },
6256         {
6257                 .procname       = "accept_ra_from_local",
6258                 .data           = &ipv6_devconf.accept_ra_from_local,
6259                 .maxlen         = sizeof(int),
6260                 .mode           = 0644,
6261                 .proc_handler   = proc_dointvec,
6262         },
6263         {
6264                 .procname       = "accept_ra_mtu",
6265                 .data           = &ipv6_devconf.accept_ra_mtu,
6266                 .maxlen         = sizeof(int),
6267                 .mode           = 0644,
6268                 .proc_handler   = proc_dointvec,
6269         },
6270         {
6271                 .procname       = "stable_secret",
6272                 .data           = &ipv6_devconf.stable_secret,
6273                 .maxlen         = IPV6_MAX_STRLEN,
6274                 .mode           = 0600,
6275                 .proc_handler   = addrconf_sysctl_stable_secret,
6276         },
6277         {
6278                 .procname       = "use_oif_addrs_only",
6279                 .data           = &ipv6_devconf.use_oif_addrs_only,
6280                 .maxlen         = sizeof(int),
6281                 .mode           = 0644,
6282                 .proc_handler   = proc_dointvec,
6283         },
6284         {
6285                 .procname       = "ignore_routes_with_linkdown",
6286                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6287                 .maxlen         = sizeof(int),
6288                 .mode           = 0644,
6289                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6290         },
6291         {
6292                 .procname       = "drop_unicast_in_l2_multicast",
6293                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6294                 .maxlen         = sizeof(int),
6295                 .mode           = 0644,
6296                 .proc_handler   = proc_dointvec,
6297         },
6298         {
6299                 .procname       = "drop_unsolicited_na",
6300                 .data           = &ipv6_devconf.drop_unsolicited_na,
6301                 .maxlen         = sizeof(int),
6302                 .mode           = 0644,
6303                 .proc_handler   = proc_dointvec,
6304         },
6305         {
6306                 .procname       = "keep_addr_on_down",
6307                 .data           = &ipv6_devconf.keep_addr_on_down,
6308                 .maxlen         = sizeof(int),
6309                 .mode           = 0644,
6310                 .proc_handler   = proc_dointvec,
6311
6312         },
6313         {
6314                 .procname       = "seg6_enabled",
6315                 .data           = &ipv6_devconf.seg6_enabled,
6316                 .maxlen         = sizeof(int),
6317                 .mode           = 0644,
6318                 .proc_handler   = proc_dointvec,
6319         },
6320 #ifdef CONFIG_IPV6_SEG6_HMAC
6321         {
6322                 .procname       = "seg6_require_hmac",
6323                 .data           = &ipv6_devconf.seg6_require_hmac,
6324                 .maxlen         = sizeof(int),
6325                 .mode           = 0644,
6326                 .proc_handler   = proc_dointvec,
6327         },
6328 #endif
6329         {
6330                 .procname       = "enhanced_dad",
6331                 .data           = &ipv6_devconf.enhanced_dad,
6332                 .maxlen         = sizeof(int),
6333                 .mode           = 0644,
6334                 .proc_handler   = proc_dointvec,
6335         },
6336         {
6337                 .procname               = "addr_gen_mode",
6338                 .data                   = &ipv6_devconf.addr_gen_mode,
6339                 .maxlen                 = sizeof(int),
6340                 .mode                   = 0644,
6341                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6342         },
6343         {
6344                 .procname       = "disable_policy",
6345                 .data           = &ipv6_devconf.disable_policy,
6346                 .maxlen         = sizeof(int),
6347                 .mode           = 0644,
6348                 .proc_handler   = addrconf_sysctl_disable_policy,
6349         },
6350         {
6351                 /* sentinel */
6352         }
6353 };
6354
6355 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6356                 struct inet6_dev *idev, struct ipv6_devconf *p)
6357 {
6358         int i, ifindex;
6359         struct ctl_table *table;
6360         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6361
6362         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6363         if (!table)
6364                 goto out;
6365
6366         for (i = 0; table[i].data; i++) {
6367                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6368                 /* If one of these is already set, then it is not safe to
6369                  * overwrite either of them: this makes proc_dointvec_minmax
6370                  * usable.
6371                  */
6372                 if (!table[i].extra1 && !table[i].extra2) {
6373                         table[i].extra1 = idev; /* embedded; no ref */
6374                         table[i].extra2 = net;
6375                 }
6376         }
6377
6378         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6379
6380         p->sysctl_header = register_net_sysctl(net, path, table);
6381         if (!p->sysctl_header)
6382                 goto free;
6383
6384         if (!strcmp(dev_name, "all"))
6385                 ifindex = NETCONFA_IFINDEX_ALL;
6386         else if (!strcmp(dev_name, "default"))
6387                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6388         else
6389                 ifindex = idev->dev->ifindex;
6390         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6391                                      ifindex, p);
6392         return 0;
6393
6394 free:
6395         kfree(table);
6396 out:
6397         return -ENOBUFS;
6398 }
6399
6400 static void __addrconf_sysctl_unregister(struct net *net,
6401                                          struct ipv6_devconf *p, int ifindex)
6402 {
6403         struct ctl_table *table;
6404
6405         if (!p->sysctl_header)
6406                 return;
6407
6408         table = p->sysctl_header->ctl_table_arg;
6409         unregister_net_sysctl_table(p->sysctl_header);
6410         p->sysctl_header = NULL;
6411         kfree(table);
6412
6413         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6414 }
6415
6416 static int addrconf_sysctl_register(struct inet6_dev *idev)
6417 {
6418         int err;
6419
6420         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6421                 return -EINVAL;
6422
6423         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6424                                     &ndisc_ifinfo_sysctl_change);
6425         if (err)
6426                 return err;
6427         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6428                                          idev, &idev->cnf);
6429         if (err)
6430                 neigh_sysctl_unregister(idev->nd_parms);
6431
6432         return err;
6433 }
6434
6435 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6436 {
6437         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6438                                      idev->dev->ifindex);
6439         neigh_sysctl_unregister(idev->nd_parms);
6440 }
6441
6442
6443 #endif
6444
6445 static int __net_init addrconf_init_net(struct net *net)
6446 {
6447         int err = -ENOMEM;
6448         struct ipv6_devconf *all, *dflt;
6449
6450         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6451         if (!all)
6452                 goto err_alloc_all;
6453
6454         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6455         if (!dflt)
6456                 goto err_alloc_dflt;
6457
6458         /* these will be inherited by all namespaces */
6459         dflt->autoconf = ipv6_defaults.autoconf;
6460         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6461
6462         dflt->stable_secret.initialized = false;
6463         all->stable_secret.initialized = false;
6464
6465         net->ipv6.devconf_all = all;
6466         net->ipv6.devconf_dflt = dflt;
6467
6468 #ifdef CONFIG_SYSCTL
6469         err = __addrconf_sysctl_register(net, "all", NULL, all);
6470         if (err < 0)
6471                 goto err_reg_all;
6472
6473         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6474         if (err < 0)
6475                 goto err_reg_dflt;
6476 #endif
6477         return 0;
6478
6479 #ifdef CONFIG_SYSCTL
6480 err_reg_dflt:
6481         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6482 err_reg_all:
6483         kfree(dflt);
6484 #endif
6485 err_alloc_dflt:
6486         kfree(all);
6487 err_alloc_all:
6488         return err;
6489 }
6490
6491 static void __net_exit addrconf_exit_net(struct net *net)
6492 {
6493 #ifdef CONFIG_SYSCTL
6494         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6495                                      NETCONFA_IFINDEX_DEFAULT);
6496         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6497                                      NETCONFA_IFINDEX_ALL);
6498 #endif
6499         kfree(net->ipv6.devconf_dflt);
6500         kfree(net->ipv6.devconf_all);
6501 }
6502
6503 static struct pernet_operations addrconf_ops = {
6504         .init = addrconf_init_net,
6505         .exit = addrconf_exit_net,
6506 };
6507
6508 static struct rtnl_af_ops inet6_ops __read_mostly = {
6509         .family           = AF_INET6,
6510         .fill_link_af     = inet6_fill_link_af,
6511         .get_link_af_size = inet6_get_link_af_size,
6512         .validate_link_af = inet6_validate_link_af,
6513         .set_link_af      = inet6_set_link_af,
6514 };
6515
6516 /*
6517  *      Init / cleanup code
6518  */
6519
6520 int __init addrconf_init(void)
6521 {
6522         struct inet6_dev *idev;
6523         int i, err;
6524
6525         err = ipv6_addr_label_init();
6526         if (err < 0) {
6527                 pr_crit("%s: cannot initialize default policy table: %d\n",
6528                         __func__, err);
6529                 goto out;
6530         }
6531
6532         err = register_pernet_subsys(&addrconf_ops);
6533         if (err < 0)
6534                 goto out_addrlabel;
6535
6536         addrconf_wq = create_workqueue("ipv6_addrconf");
6537         if (!addrconf_wq) {
6538                 err = -ENOMEM;
6539                 goto out_nowq;
6540         }
6541
6542         /* The addrconf netdev notifier requires that loopback_dev
6543          * has it's ipv6 private information allocated and setup
6544          * before it can bring up and give link-local addresses
6545          * to other devices which are up.
6546          *
6547          * Unfortunately, loopback_dev is not necessarily the first
6548          * entry in the global dev_base list of net devices.  In fact,
6549          * it is likely to be the very last entry on that list.
6550          * So this causes the notifier registry below to try and
6551          * give link-local addresses to all devices besides loopback_dev
6552          * first, then loopback_dev, which cases all the non-loopback_dev
6553          * devices to fail to get a link-local address.
6554          *
6555          * So, as a temporary fix, allocate the ipv6 structure for
6556          * loopback_dev first by hand.
6557          * Longer term, all of the dependencies ipv6 has upon the loopback
6558          * device and it being up should be removed.
6559          */
6560         rtnl_lock();
6561         idev = ipv6_add_dev(init_net.loopback_dev);
6562         rtnl_unlock();
6563         if (IS_ERR(idev)) {
6564                 err = PTR_ERR(idev);
6565                 goto errlo;
6566         }
6567
6568         ip6_route_init_special_entries();
6569
6570         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6571                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6572
6573         register_netdevice_notifier(&ipv6_dev_notf);
6574
6575         addrconf_verify();
6576
6577         rtnl_af_register(&inet6_ops);
6578
6579         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6580                               0);
6581         if (err < 0)
6582                 goto errout;
6583
6584         /* Only the first call to __rtnl_register can fail */
6585         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, 0);
6586         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, 0);
6587         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6588                         inet6_dump_ifaddr, 0);
6589         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6590                         inet6_dump_ifmcaddr, 0);
6591         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6592                         inet6_dump_ifacaddr, 0);
6593         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6594                         inet6_netconf_dump_devconf, 0);
6595
6596         ipv6_addr_label_rtnl_register();
6597
6598         return 0;
6599 errout:
6600         rtnl_af_unregister(&inet6_ops);
6601         unregister_netdevice_notifier(&ipv6_dev_notf);
6602 errlo:
6603         destroy_workqueue(addrconf_wq);
6604 out_nowq:
6605         unregister_pernet_subsys(&addrconf_ops);
6606 out_addrlabel:
6607         ipv6_addr_label_cleanup();
6608 out:
6609         return err;
6610 }
6611
6612 void addrconf_cleanup(void)
6613 {
6614         struct net_device *dev;
6615         int i;
6616
6617         unregister_netdevice_notifier(&ipv6_dev_notf);
6618         unregister_pernet_subsys(&addrconf_ops);
6619         ipv6_addr_label_cleanup();
6620
6621         rtnl_lock();
6622
6623         __rtnl_af_unregister(&inet6_ops);
6624
6625         /* clean dev list */
6626         for_each_netdev(&init_net, dev) {
6627                 if (__in6_dev_get(dev) == NULL)
6628                         continue;
6629                 addrconf_ifdown(dev, 1);
6630         }
6631         addrconf_ifdown(init_net.loopback_dev, 2);
6632
6633         /*
6634          *      Check hash table.
6635          */
6636         spin_lock_bh(&addrconf_hash_lock);
6637         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6638                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6639         spin_unlock_bh(&addrconf_hash_lock);
6640         cancel_delayed_work(&addr_chk_work);
6641         rtnl_unlock();
6642
6643         destroy_workqueue(addrconf_wq);
6644 }