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