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_max_plen = 0,
 228 #endif
 229 #endif
 230         .proxy_ndp              = 0,
 231         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 232         .disable_ipv6           = 0,
 233         .accept_dad             = 1,
 234         .suppress_frag_ndisc    = 1,
 235         .accept_ra_mtu          = 1,
 236         .stable_secret          = {
 237                 .initialized = false,
 238         },
 239         .use_oif_addrs_only     = 0,
 240         .ignore_routes_with_linkdown = 0,
 241         .keep_addr_on_down      = 0,
 242         .seg6_enabled           = 0,
 243 #ifdef CONFIG_IPV6_SEG6_HMAC
 244         .seg6_require_hmac      = 0,
 245 #endif
 246         .enhanced_dad           = 1,
 247         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 248 };
 249
 250 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 251         .forwarding             = 0,
 252         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 253         .mtu6                   = IPV6_MIN_MTU,
 254         .accept_ra              = 1,
 255         .accept_redirects       = 1,
 256         .autoconf               = 1,
 257         .force_mld_version      = 0,
 258         .mldv1_unsolicited_report_interval = 10 * HZ,
 259         .mldv2_unsolicited_report_interval = HZ,
 260         .dad_transmits          = 1,
 261         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 262         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 263         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 264         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 265         .use_tempaddr           = 0,
 266         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 267         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 268         .regen_max_retry        = REGEN_MAX_RETRY,
 269         .max_desync_factor      = MAX_DESYNC_FACTOR,
 270         .max_addresses          = IPV6_MAX_ADDRESSES,
 271         .accept_ra_defrtr       = 1,
 272         .accept_ra_from_local   = 0,
 273         .accept_ra_min_hop_limit= 1,
 274         .accept_ra_pinfo        = 1,
 275 #ifdef CONFIG_IPV6_ROUTER_PREF
 276         .accept_ra_rtr_pref     = 1,
 277         .rtr_probe_interval     = 60 * HZ,
 278 #ifdef CONFIG_IPV6_ROUTE_INFO
 279         .accept_ra_rt_info_max_plen = 0,
 280 #endif
 281 #endif
 282         .proxy_ndp              = 0,
 283         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 284         .disable_ipv6           = 0,
 285         .accept_dad             = 1,
 286         .suppress_frag_ndisc    = 1,
 287         .accept_ra_mtu          = 1,
 288         .stable_secret          = {
 289                 .initialized = false,
 290         },
 291         .use_oif_addrs_only     = 0,
 292         .ignore_routes_with_linkdown = 0,
 293         .keep_addr_on_down      = 0,
 294         .seg6_enabled           = 0,
 295 #ifdef CONFIG_IPV6_SEG6_HMAC
 296         .seg6_require_hmac      = 0,
 297 #endif
 298         .enhanced_dad           = 1,
 299         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 300 };
 301
 302 /* Check if a valid qdisc is available */
 303 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 304 {
 305         return !qdisc_tx_is_noop(dev);
 306 }
 307
 308 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 309 {
 310         if (del_timer(&idev->rs_timer))
 311                 __in6_dev_put(idev);
 312 }
 313
 314 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 315 {
 316         if (cancel_delayed_work(&ifp->dad_work))
 317                 __in6_ifa_put(ifp);
 318 }
 319
 320 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 321                                   unsigned long when)
 322 {
 323         if (!timer_pending(&idev->rs_timer))
 324                 in6_dev_hold(idev);
 325         mod_timer(&idev->rs_timer, jiffies + when);
 326 }
 327
 328 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 329                                    unsigned long delay)
 330 {
 331         if (!delayed_work_pending(&ifp->dad_work))
 332                 in6_ifa_hold(ifp);
 333         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 334 }
 335
 336 static int snmp6_alloc_dev(struct inet6_dev *idev)
 337 {
 338         int i;
 339
 340         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 341         if (!idev->stats.ipv6)
 342                 goto err_ip;
 343
 344         for_each_possible_cpu(i) {
 345                 struct ipstats_mib *addrconf_stats;
 346                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 347                 u64_stats_init(&addrconf_stats->syncp);
 348         }
 349
 350
 351         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 352                                         GFP_KERNEL);
 353         if (!idev->stats.icmpv6dev)
 354                 goto err_icmp;
 355         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 356                                            GFP_KERNEL);
 357         if (!idev->stats.icmpv6msgdev)
 358                 goto err_icmpmsg;
 359
 360         return 0;
 361
 362 err_icmpmsg:
 363         kfree(idev->stats.icmpv6dev);
 364 err_icmp:
 365         free_percpu(idev->stats.ipv6);
 366 err_ip:
 367         return -ENOMEM;
 368 }
 369
 370 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 371 {
 372         struct inet6_dev *ndev;
 373         int err = -ENOMEM;
 374
 375         ASSERT_RTNL();
 376
 377         if (dev->mtu < IPV6_MIN_MTU)
 378                 return ERR_PTR(-EINVAL);
 379
 380         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 381         if (!ndev)
 382                 return ERR_PTR(err);
 383
 384         rwlock_init(&ndev->lock);
 385         ndev->dev = dev;
 386         INIT_LIST_HEAD(&ndev->addr_list);
 387         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 388                     (unsigned long)ndev);
 389         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 390
 391         if (ndev->cnf.stable_secret.initialized)
 392                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 393         else
 394                 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
 395
 396         ndev->cnf.mtu6 = dev->mtu;
 397         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 398         if (!ndev->nd_parms) {
 399                 kfree(ndev);
 400                 return ERR_PTR(err);
 401         }
 402         if (ndev->cnf.forwarding)
 403                 dev_disable_lro(dev);
 404         /* We refer to the device */
 405         dev_hold(dev);
 406
 407         if (snmp6_alloc_dev(ndev) < 0) {
 408                 ADBG(KERN_WARNING
 409                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 410                         __func__, dev->name);
 411                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 412                 dev_put(dev);
 413                 kfree(ndev);
 414                 return ERR_PTR(err);
 415         }
 416
 417         if (snmp6_register_dev(ndev) < 0) {
 418                 ADBG(KERN_WARNING
 419                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 420                         __func__, dev->name);
 421                 goto err_release;
 422         }
 423
 424         /* One reference from device. */
 425         in6_dev_hold(ndev);
 426
 427         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 428                 ndev->cnf.accept_dad = -1;
 429
 430 #if IS_ENABLED(CONFIG_IPV6_SIT)
 431         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 432                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 433                 ndev->cnf.rtr_solicits = 0;
 434         }
 435 #endif
 436
 437         INIT_LIST_HEAD(&ndev->tempaddr_list);
 438         ndev->desync_factor = U32_MAX;
 439         if ((dev->flags&IFF_LOOPBACK) ||
 440             dev->type == ARPHRD_TUNNEL ||
 441             dev->type == ARPHRD_TUNNEL6 ||
 442             dev->type == ARPHRD_SIT ||
 443             dev->type == ARPHRD_NONE) {
 444                 ndev->cnf.use_tempaddr = -1;
 445         } else
 446                 ipv6_regen_rndid(ndev);
 447
 448         ndev->token = in6addr_any;
 449
 450         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 451                 ndev->if_flags |= IF_READY;
 452
 453         ipv6_mc_init_dev(ndev);
 454         ndev->tstamp = jiffies;
 455         err = addrconf_sysctl_register(ndev);
 456         if (err) {
 457                 ipv6_mc_destroy_dev(ndev);
 458                 snmp6_unregister_dev(ndev);
 459                 goto err_release;
 460         }
 461         /* protected by rtnl_lock */
 462         rcu_assign_pointer(dev->ip6_ptr, ndev);
 463
 464         /* Join interface-local all-node multicast group */
 465         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 466
 467         /* Join all-node multicast group */
 468         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 469
 470         /* Join all-router multicast group if forwarding is set */
 471         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 472                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 473
 474         return ndev;
 475
 476 err_release:
 477         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 478         ndev->dead = 1;
 479         in6_dev_finish_destroy(ndev);
 480         return ERR_PTR(err);
 481 }
 482
 483 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 484 {
 485         struct inet6_dev *idev;
 486
 487         ASSERT_RTNL();
 488
 489         idev = __in6_dev_get(dev);
 490         if (!idev) {
 491                 idev = ipv6_add_dev(dev);
 492                 if (IS_ERR(idev))
 493                         return NULL;
 494         }
 495
 496         if (dev->flags&IFF_UP)
 497                 ipv6_mc_up(idev);
 498         return idev;
 499 }
 500
 501 static int inet6_netconf_msgsize_devconf(int type)
 502 {
 503         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 504                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 505         bool all = false;
 506
 507         if (type == NETCONFA_ALL)
 508                 all = true;
 509
 510         if (all || type == NETCONFA_FORWARDING)
 511                 size += nla_total_size(4);
 512 #ifdef CONFIG_IPV6_MROUTE
 513         if (all || type == NETCONFA_MC_FORWARDING)
 514                 size += nla_total_size(4);
 515 #endif
 516         if (all || type == NETCONFA_PROXY_NEIGH)
 517                 size += nla_total_size(4);
 518
 519         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 520                 size += nla_total_size(4);
 521
 522         return size;
 523 }
 524
 525 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 526                                       struct ipv6_devconf *devconf, u32 portid,
 527                                       u32 seq, int event, unsigned int flags,
 528                                       int type)
 529 {
 530         struct nlmsghdr  *nlh;
 531         struct netconfmsg *ncm;
 532         bool all = false;
 533
 534         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 535                         flags);
 536         if (!nlh)
 537                 return -EMSGSIZE;
 538
 539         if (type == NETCONFA_ALL)
 540                 all = true;
 541
 542         ncm = nlmsg_data(nlh);
 543         ncm->ncm_family = AF_INET6;
 544
 545         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 546                 goto nla_put_failure;
 547
 548         if ((all || type == NETCONFA_FORWARDING) &&
 549             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 550                 goto nla_put_failure;
 551 #ifdef CONFIG_IPV6_MROUTE
 552         if ((all || type == NETCONFA_MC_FORWARDING) &&
 553             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 554                         devconf->mc_forwarding) < 0)
 555                 goto nla_put_failure;
 556 #endif
 557         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 558             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 559                 goto nla_put_failure;
 560
 561         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 562             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 563                         devconf->ignore_routes_with_linkdown) < 0)
 564                 goto nla_put_failure;
 565
 566         nlmsg_end(skb, nlh);
 567         return 0;
 568
 569 nla_put_failure:
 570         nlmsg_cancel(skb, nlh);
 571         return -EMSGSIZE;
 572 }
 573
 574 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 575                                   struct ipv6_devconf *devconf)
 576 {
 577         struct sk_buff *skb;
 578         int err = -ENOBUFS;
 579
 580         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 581         if (!skb)
 582                 goto errout;
 583
 584         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 585                                          RTM_NEWNETCONF, 0, type);
 586         if (err < 0) {
 587                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 588                 WARN_ON(err == -EMSGSIZE);
 589                 kfree_skb(skb);
 590                 goto errout;
 591         }
 592         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 593         return;
 594 errout:
 595         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 596 }
 597
 598 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 599         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 600         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 601         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 602         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 603 };
 604
 605 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 606                                      struct nlmsghdr *nlh)
 607 {
 608         struct net *net = sock_net(in_skb->sk);
 609         struct nlattr *tb[NETCONFA_MAX+1];
 610         struct netconfmsg *ncm;
 611         struct sk_buff *skb;
 612         struct ipv6_devconf *devconf;
 613         struct inet6_dev *in6_dev;
 614         struct net_device *dev;
 615         int ifindex;
 616         int err;
 617
 618         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 619                           devconf_ipv6_policy);
 620         if (err < 0)
 621                 goto errout;
 622
 623         err = -EINVAL;
 624         if (!tb[NETCONFA_IFINDEX])
 625                 goto errout;
 626
 627         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 628         switch (ifindex) {
 629         case NETCONFA_IFINDEX_ALL:
 630                 devconf = net->ipv6.devconf_all;
 631                 break;
 632         case NETCONFA_IFINDEX_DEFAULT:
 633                 devconf = net->ipv6.devconf_dflt;
 634                 break;
 635         default:
 636                 dev = __dev_get_by_index(net, ifindex);
 637                 if (!dev)
 638                         goto errout;
 639                 in6_dev = __in6_dev_get(dev);
 640                 if (!in6_dev)
 641                         goto errout;
 642                 devconf = &in6_dev->cnf;
 643                 break;
 644         }
 645
 646         err = -ENOBUFS;
 647         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
 648         if (!skb)
 649                 goto errout;
 650
 651         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 652                                          NETLINK_CB(in_skb).portid,
 653                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 654                                          NETCONFA_ALL);
 655         if (err < 0) {
 656                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 657                 WARN_ON(err == -EMSGSIZE);
 658                 kfree_skb(skb);
 659                 goto errout;
 660         }
 661         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 662 errout:
 663         return err;
 664 }
 665
 666 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 667                                       struct netlink_callback *cb)
 668 {
 669         struct net *net = sock_net(skb->sk);
 670         int h, s_h;
 671         int idx, s_idx;
 672         struct net_device *dev;
 673         struct inet6_dev *idev;
 674         struct hlist_head *head;
 675
 676         s_h = cb->args[0];
 677         s_idx = idx = cb->args[1];
 678
 679         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 680                 idx = 0;
 681                 head = &net->dev_index_head[h];
 682                 rcu_read_lock();
 683                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 684                           net->dev_base_seq;
 685                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 686                         if (idx < s_idx)
 687                                 goto cont;
 688                         idev = __in6_dev_get(dev);
 689                         if (!idev)
 690                                 goto cont;
 691
 692                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 693                                                        &idev->cnf,
 694                                                        NETLINK_CB(cb->skb).portid,
 695                                                        cb->nlh->nlmsg_seq,
 696                                                        RTM_NEWNETCONF,
 697                                                        NLM_F_MULTI,
 698                                                        NETCONFA_ALL) < 0) {
 699                                 rcu_read_unlock();
 700                                 goto done;
 701                         }
 702                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 703 cont:
 704                         idx++;
 705                 }
 706                 rcu_read_unlock();
 707         }
 708         if (h == NETDEV_HASHENTRIES) {
 709                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 710                                                net->ipv6.devconf_all,
 711                                                NETLINK_CB(cb->skb).portid,
 712                                                cb->nlh->nlmsg_seq,
 713                                                RTM_NEWNETCONF, NLM_F_MULTI,
 714                                                NETCONFA_ALL) < 0)
 715                         goto done;
 716                 else
 717                         h++;
 718         }
 719         if (h == NETDEV_HASHENTRIES + 1) {
 720                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 721                                                net->ipv6.devconf_dflt,
 722                                                NETLINK_CB(cb->skb).portid,
 723                                                cb->nlh->nlmsg_seq,
 724                                                RTM_NEWNETCONF, NLM_F_MULTI,
 725                                                NETCONFA_ALL) < 0)
 726                         goto done;
 727                 else
 728                         h++;
 729         }
 730 done:
 731         cb->args[0] = h;
 732         cb->args[1] = idx;
 733
 734         return skb->len;
 735 }
 736
 737 #ifdef CONFIG_SYSCTL
 738 static void dev_forward_change(struct inet6_dev *idev)
 739 {
 740         struct net_device *dev;
 741         struct inet6_ifaddr *ifa;
 742
 743         if (!idev)
 744                 return;
 745         dev = idev->dev;
 746         if (idev->cnf.forwarding)
 747                 dev_disable_lro(dev);
 748         if (dev->flags & IFF_MULTICAST) {
 749                 if (idev->cnf.forwarding) {
 750                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 751                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 752                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 753                 } else {
 754                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 755                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 756                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 757                 }
 758         }
 759
 760         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 761                 if (ifa->flags&IFA_F_TENTATIVE)
 762                         continue;
 763                 if (idev->cnf.forwarding)
 764                         addrconf_join_anycast(ifa);
 765                 else
 766                         addrconf_leave_anycast(ifa);
 767         }
 768         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 769                                      dev->ifindex, &idev->cnf);
 770 }
 771
 772
 773 static void addrconf_forward_change(struct net *net, __s32 newf)
 774 {
 775         struct net_device *dev;
 776         struct inet6_dev *idev;
 777
 778         for_each_netdev(net, dev) {
 779                 idev = __in6_dev_get(dev);
 780                 if (idev) {
 781                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 782                         idev->cnf.forwarding = newf;
 783                         if (changed)
 784                                 dev_forward_change(idev);
 785                 }
 786         }
 787 }
 788
 789 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 790 {
 791         struct net *net;
 792         int old;
 793
 794         if (!rtnl_trylock())
 795                 return restart_syscall();
 796
 797         net = (struct net *)table->extra2;
 798         old = *p;
 799         *p = newf;
 800
 801         if (p == &net->ipv6.devconf_dflt->forwarding) {
 802                 if ((!newf) ^ (!old))
 803                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 804                                                      NETCONFA_IFINDEX_DEFAULT,
 805                                                      net->ipv6.devconf_dflt);
 806                 rtnl_unlock();
 807                 return 0;
 808         }
 809
 810         if (p == &net->ipv6.devconf_all->forwarding) {
 811                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 812
 813                 net->ipv6.devconf_dflt->forwarding = newf;
 814                 if ((!newf) ^ (!old_dflt))
 815                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 816                                                      NETCONFA_IFINDEX_DEFAULT,
 817                                                      net->ipv6.devconf_dflt);
 818
 819                 addrconf_forward_change(net, newf);
 820                 if ((!newf) ^ (!old))
 821                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 822                                                      NETCONFA_IFINDEX_ALL,
 823                                                      net->ipv6.devconf_all);
 824         } else if ((!newf) ^ (!old))
 825                 dev_forward_change((struct inet6_dev *)table->extra1);
 826         rtnl_unlock();
 827
 828         if (newf)
 829                 rt6_purge_dflt_routers(net);
 830         return 1;
 831 }
 832
 833 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 834 {
 835         struct net_device *dev;
 836         struct inet6_dev *idev;
 837
 838         for_each_netdev(net, dev) {
 839                 idev = __in6_dev_get(dev);
 840                 if (idev) {
 841                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 842
 843                         idev->cnf.ignore_routes_with_linkdown = newf;
 844                         if (changed)
 845                                 inet6_netconf_notify_devconf(dev_net(dev),
 846                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 847                                                              dev->ifindex,
 848                                                              &idev->cnf);
 849                 }
 850         }
 851 }
 852
 853 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 854 {
 855         struct net *net;
 856         int old;
 857
 858         if (!rtnl_trylock())
 859                 return restart_syscall();
 860
 861         net = (struct net *)table->extra2;
 862         old = *p;
 863         *p = newf;
 864
 865         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 866                 if ((!newf) ^ (!old))
 867                         inet6_netconf_notify_devconf(net,
 868                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 869                                                      NETCONFA_IFINDEX_DEFAULT,
 870                                                      net->ipv6.devconf_dflt);
 871                 rtnl_unlock();
 872                 return 0;
 873         }
 874
 875         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 876                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 877                 addrconf_linkdown_change(net, newf);
 878                 if ((!newf) ^ (!old))
 879                         inet6_netconf_notify_devconf(net,
 880                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 881                                                      NETCONFA_IFINDEX_ALL,
 882                                                      net->ipv6.devconf_all);
 883         }
 884         rtnl_unlock();
 885
 886         return 1;
 887 }
 888
 889 #endif
 890
 891 /* Nobody refers to this ifaddr, destroy it */
 892 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 893 {
 894         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 895
 896 #ifdef NET_REFCNT_DEBUG
 897         pr_debug("%s\n", __func__);
 898 #endif
 899
 900         in6_dev_put(ifp->idev);
 901
 902         if (cancel_delayed_work(&ifp->dad_work))
 903                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 904                           ifp);
 905
 906         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 907                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 908                 return;
 909         }
 910         ip6_rt_put(ifp->rt);
 911
 912         kfree_rcu(ifp, rcu);
 913 }
 914
 915 static void
 916 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 917 {
 918         struct list_head *p;
 919         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 920
 921         /*
 922          * Each device address list is sorted in order of scope -
 923          * global before linklocal.
 924          */
 925         list_for_each(p, &idev->addr_list) {
 926                 struct inet6_ifaddr *ifa
 927                         = list_entry(p, struct inet6_ifaddr, if_list);
 928                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 929                         break;
 930         }
 931
 932         list_add_tail(&ifp->if_list, p);
 933 }
 934
 935 static u32 inet6_addr_hash(const struct in6_addr *addr)
 936 {
 937         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 938 }
 939
 940 /* On success it returns ifp with increased reference count */
 941
 942 static struct inet6_ifaddr *
 943 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 944               const struct in6_addr *peer_addr, int pfxlen,
 945               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 946 {
 947         struct inet6_ifaddr *ifa = NULL;
 948         struct rt6_info *rt;
 949         unsigned int hash;
 950         int err = 0;
 951         int addr_type = ipv6_addr_type(addr);
 952
 953         if (addr_type == IPV6_ADDR_ANY ||
 954             addr_type & IPV6_ADDR_MULTICAST ||
 955             (!(idev->dev->flags & IFF_LOOPBACK) &&
 956              addr_type & IPV6_ADDR_LOOPBACK))
 957                 return ERR_PTR(-EADDRNOTAVAIL);
 958
 959         rcu_read_lock_bh();
 960         if (idev->dead) {
 961                 err = -ENODEV;                  /*XXX*/
 962                 goto out2;
 963         }
 964
 965         if (idev->cnf.disable_ipv6) {
 966                 err = -EACCES;
 967                 goto out2;
 968         }
 969
 970         spin_lock(&addrconf_hash_lock);
 971
 972         /* Ignore adding duplicate addresses on an interface */
 973         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 974                 ADBG("ipv6_add_addr: already assigned\n");
 975                 err = -EEXIST;
 976                 goto out;
 977         }
 978
 979         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 980
 981         if (!ifa) {
 982                 ADBG("ipv6_add_addr: malloc failed\n");
 983                 err = -ENOBUFS;
 984                 goto out;
 985         }
 986
 987         rt = addrconf_dst_alloc(idev, addr, false);
 988         if (IS_ERR(rt)) {
 989                 err = PTR_ERR(rt);
 990                 goto out;
 991         }
 992
 993         neigh_parms_data_state_setall(idev->nd_parms);
 994
 995         ifa->addr = *addr;
 996         if (peer_addr)
 997                 ifa->peer_addr = *peer_addr;
 998
 999         spin_lock_init(&ifa->lock);
1000         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1001         INIT_HLIST_NODE(&ifa->addr_lst);
1002         ifa->scope = scope;
1003         ifa->prefix_len = pfxlen;
1004         ifa->flags = flags | IFA_F_TENTATIVE;
1005         ifa->valid_lft = valid_lft;
1006         ifa->prefered_lft = prefered_lft;
1007         ifa->cstamp = ifa->tstamp = jiffies;
1008         ifa->tokenized = false;
1009
1010         ifa->rt = rt;
1011
1012         ifa->idev = idev;
1013         in6_dev_hold(idev);
1014         /* For caller */
1015         in6_ifa_hold(ifa);
1016
1017         /* Add to big hash table */
1018         hash = inet6_addr_hash(addr);
1019
1020         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1021         spin_unlock(&addrconf_hash_lock);
1022
1023         write_lock(&idev->lock);
1024         /* Add to inet6_dev unicast addr list. */
1025         ipv6_link_dev_addr(idev, ifa);
1026
1027         if (ifa->flags&IFA_F_TEMPORARY) {
1028                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1029                 in6_ifa_hold(ifa);
1030         }
1031
1032         in6_ifa_hold(ifa);
1033         write_unlock(&idev->lock);
1034 out2:
1035         rcu_read_unlock_bh();
1036
1037         if (likely(err == 0))
1038                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1039         else {
1040                 kfree(ifa);
1041                 ifa = ERR_PTR(err);
1042         }
1043
1044         return ifa;
1045 out:
1046         spin_unlock(&addrconf_hash_lock);
1047         goto out2;
1048 }
1049
1050 enum cleanup_prefix_rt_t {
1051         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1052         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1053         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1054 };
1055
1056 /*
1057  * Check, whether the prefix for ifp would still need a prefix route
1058  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1059  * constants.
1060  *
1061  * 1) we don't purge prefix if address was not permanent.
1062  *    prefix is managed by its own lifetime.
1063  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1064  * 3) if there are no addresses, delete prefix.
1065  * 4) if there are still other permanent address(es),
1066  *    corresponding prefix is still permanent.
1067  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1068  *    don't purge the prefix, assume user space is managing it.
1069  * 6) otherwise, update prefix lifetime to the
1070  *    longest valid lifetime among the corresponding
1071  *    addresses on the device.
1072  *    Note: subsequent RA will update lifetime.
1073  **/
1074 static enum cleanup_prefix_rt_t
1075 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1076 {
1077         struct inet6_ifaddr *ifa;
1078         struct inet6_dev *idev = ifp->idev;
1079         unsigned long lifetime;
1080         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1081
1082         *expires = jiffies;
1083
1084         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1085                 if (ifa == ifp)
1086                         continue;
1087                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1088                                        ifp->prefix_len))
1089                         continue;
1090                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1091                         return CLEANUP_PREFIX_RT_NOP;
1092
1093                 action = CLEANUP_PREFIX_RT_EXPIRE;
1094
1095                 spin_lock(&ifa->lock);
1096
1097                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1098                 /*
1099                  * Note: Because this address is
1100                  * not permanent, lifetime <
1101                  * LONG_MAX / HZ here.
1102                  */
1103                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1104                         *expires = ifa->tstamp + lifetime * HZ;
1105                 spin_unlock(&ifa->lock);
1106         }
1107
1108         return action;
1109 }
1110
1111 static void
1112 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1113 {
1114         struct rt6_info *rt;
1115
1116         rt = addrconf_get_prefix_route(&ifp->addr,
1117                                        ifp->prefix_len,
1118                                        ifp->idev->dev,
1119                                        0, RTF_GATEWAY | RTF_DEFAULT);
1120         if (rt) {
1121                 if (del_rt)
1122                         ip6_del_rt(rt);
1123                 else {
1124                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1125                                 rt6_set_expires(rt, expires);
1126                         ip6_rt_put(rt);
1127                 }
1128         }
1129 }
1130
1131
1132 /* This function wants to get referenced ifp and releases it before return */
1133
1134 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1135 {
1136         int state;
1137         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1138         unsigned long expires;
1139
1140         ASSERT_RTNL();
1141
1142         spin_lock_bh(&ifp->lock);
1143         state = ifp->state;
1144         ifp->state = INET6_IFADDR_STATE_DEAD;
1145         spin_unlock_bh(&ifp->lock);
1146
1147         if (state == INET6_IFADDR_STATE_DEAD)
1148                 goto out;
1149
1150         spin_lock_bh(&addrconf_hash_lock);
1151         hlist_del_init_rcu(&ifp->addr_lst);
1152         spin_unlock_bh(&addrconf_hash_lock);
1153
1154         write_lock_bh(&ifp->idev->lock);
1155
1156         if (ifp->flags&IFA_F_TEMPORARY) {
1157                 list_del(&ifp->tmp_list);
1158                 if (ifp->ifpub) {
1159                         in6_ifa_put(ifp->ifpub);
1160                         ifp->ifpub = NULL;
1161                 }
1162                 __in6_ifa_put(ifp);
1163         }
1164
1165         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1166                 action = check_cleanup_prefix_route(ifp, &expires);
1167
1168         list_del_init(&ifp->if_list);
1169         __in6_ifa_put(ifp);
1170
1171         write_unlock_bh(&ifp->idev->lock);
1172
1173         addrconf_del_dad_work(ifp);
1174
1175         ipv6_ifa_notify(RTM_DELADDR, ifp);
1176
1177         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1178
1179         if (action != CLEANUP_PREFIX_RT_NOP) {
1180                 cleanup_prefix_route(ifp, expires,
1181                         action == CLEANUP_PREFIX_RT_DEL);
1182         }
1183
1184         /* clean up prefsrc entries */
1185         rt6_remove_prefsrc(ifp);
1186 out:
1187         in6_ifa_put(ifp);
1188 }
1189
1190 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1191 {
1192         struct inet6_dev *idev = ifp->idev;
1193         struct in6_addr addr, *tmpaddr;
1194         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1195         unsigned long regen_advance;
1196         int tmp_plen;
1197         int ret = 0;
1198         u32 addr_flags;
1199         unsigned long now = jiffies;
1200         long max_desync_factor;
1201         s32 cnf_temp_preferred_lft;
1202
1203         write_lock_bh(&idev->lock);
1204         if (ift) {
1205                 spin_lock_bh(&ift->lock);
1206                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1207                 spin_unlock_bh(&ift->lock);
1208                 tmpaddr = &addr;
1209         } else {
1210                 tmpaddr = NULL;
1211         }
1212 retry:
1213         in6_dev_hold(idev);
1214         if (idev->cnf.use_tempaddr <= 0) {
1215                 write_unlock_bh(&idev->lock);
1216                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1217                 in6_dev_put(idev);
1218                 ret = -1;
1219                 goto out;
1220         }
1221         spin_lock_bh(&ifp->lock);
1222         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1223                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1224                 spin_unlock_bh(&ifp->lock);
1225                 write_unlock_bh(&idev->lock);
1226                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1227                         __func__);
1228                 in6_dev_put(idev);
1229                 ret = -1;
1230                 goto out;
1231         }
1232         in6_ifa_hold(ifp);
1233         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1234         ipv6_try_regen_rndid(idev, tmpaddr);
1235         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1236         age = (now - ifp->tstamp) / HZ;
1237
1238         regen_advance = idev->cnf.regen_max_retry *
1239                         idev->cnf.dad_transmits *
1240                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1241
1242         /* recalculate max_desync_factor each time and update
1243          * idev->desync_factor if it's larger
1244          */
1245         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1246         max_desync_factor = min_t(__u32,
1247                                   idev->cnf.max_desync_factor,
1248                                   cnf_temp_preferred_lft - regen_advance);
1249
1250         if (unlikely(idev->desync_factor > max_desync_factor)) {
1251                 if (max_desync_factor > 0) {
1252                         get_random_bytes(&idev->desync_factor,
1253                                          sizeof(idev->desync_factor));
1254                         idev->desync_factor %= max_desync_factor;
1255                 } else {
1256                         idev->desync_factor = 0;
1257                 }
1258         }
1259
1260         tmp_valid_lft = min_t(__u32,
1261                               ifp->valid_lft,
1262                               idev->cnf.temp_valid_lft + age);
1263         tmp_prefered_lft = cnf_temp_preferred_lft + age -
1264                             idev->desync_factor;
1265         tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1266         tmp_plen = ifp->prefix_len;
1267         tmp_tstamp = ifp->tstamp;
1268         spin_unlock_bh(&ifp->lock);
1269
1270         write_unlock_bh(&idev->lock);
1271
1272         /* A temporary address is created only if this calculated Preferred
1273          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1274          * an implementation must not create a temporary address with a zero
1275          * Preferred Lifetime.
1276          * Use age calculation as in addrconf_verify to avoid unnecessary
1277          * temporary addresses being generated.
1278          */
1279         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1280         if (tmp_prefered_lft <= regen_advance + age) {
1281                 in6_ifa_put(ifp);
1282                 in6_dev_put(idev);
1283                 ret = -1;
1284                 goto out;
1285         }
1286
1287         addr_flags = IFA_F_TEMPORARY;
1288         /* set in addrconf_prefix_rcv() */
1289         if (ifp->flags & IFA_F_OPTIMISTIC)
1290                 addr_flags |= IFA_F_OPTIMISTIC;
1291
1292         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1293                             ipv6_addr_scope(&addr), addr_flags,
1294                             tmp_valid_lft, tmp_prefered_lft);
1295         if (IS_ERR(ift)) {
1296                 in6_ifa_put(ifp);
1297                 in6_dev_put(idev);
1298                 pr_info("%s: retry temporary address regeneration\n", __func__);
1299                 tmpaddr = &addr;
1300                 write_lock_bh(&idev->lock);
1301                 goto retry;
1302         }
1303
1304         spin_lock_bh(&ift->lock);
1305         ift->ifpub = ifp;
1306         ift->cstamp = now;
1307         ift->tstamp = tmp_tstamp;
1308         spin_unlock_bh(&ift->lock);
1309
1310         addrconf_dad_start(ift);
1311         in6_ifa_put(ift);
1312         in6_dev_put(idev);
1313 out:
1314         return ret;
1315 }
1316
1317 /*
1318  *      Choose an appropriate source address (RFC3484)
1319  */
1320 enum {
1321         IPV6_SADDR_RULE_INIT = 0,
1322         IPV6_SADDR_RULE_LOCAL,
1323         IPV6_SADDR_RULE_SCOPE,
1324         IPV6_SADDR_RULE_PREFERRED,
1325 #ifdef CONFIG_IPV6_MIP6
1326         IPV6_SADDR_RULE_HOA,
1327 #endif
1328         IPV6_SADDR_RULE_OIF,
1329         IPV6_SADDR_RULE_LABEL,
1330         IPV6_SADDR_RULE_PRIVACY,
1331         IPV6_SADDR_RULE_ORCHID,
1332         IPV6_SADDR_RULE_PREFIX,
1333 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1334         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1335 #endif
1336         IPV6_SADDR_RULE_MAX
1337 };
1338
1339 struct ipv6_saddr_score {
1340         int                     rule;
1341         int                     addr_type;
1342         struct inet6_ifaddr     *ifa;
1343         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1344         int                     scopedist;
1345         int                     matchlen;
1346 };
1347
1348 struct ipv6_saddr_dst {
1349         const struct in6_addr *addr;
1350         int ifindex;
1351         int scope;
1352         int label;
1353         unsigned int prefs;
1354 };
1355
1356 static inline int ipv6_saddr_preferred(int type)
1357 {
1358         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1359                 return 1;
1360         return 0;
1361 }
1362
1363 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1364 {
1365 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1366         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1367 #else
1368         return false;
1369 #endif
1370 }
1371
1372 static int ipv6_get_saddr_eval(struct net *net,
1373                                struct ipv6_saddr_score *score,
1374                                struct ipv6_saddr_dst *dst,
1375                                int i)
1376 {
1377         int ret;
1378
1379         if (i <= score->rule) {
1380                 switch (i) {
1381                 case IPV6_SADDR_RULE_SCOPE:
1382                         ret = score->scopedist;
1383                         break;
1384                 case IPV6_SADDR_RULE_PREFIX:
1385                         ret = score->matchlen;
1386                         break;
1387                 default:
1388                         ret = !!test_bit(i, score->scorebits);
1389                 }
1390                 goto out;
1391         }
1392
1393         switch (i) {
1394         case IPV6_SADDR_RULE_INIT:
1395                 /* Rule 0: remember if hiscore is not ready yet */
1396                 ret = !!score->ifa;
1397                 break;
1398         case IPV6_SADDR_RULE_LOCAL:
1399                 /* Rule 1: Prefer same address */
1400                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1401                 break;
1402         case IPV6_SADDR_RULE_SCOPE:
1403                 /* Rule 2: Prefer appropriate scope
1404                  *
1405                  *      ret
1406                  *       ^
1407                  *    -1 |  d 15
1408                  *    ---+--+-+---> scope
1409                  *       |
1410                  *       |             d is scope of the destination.
1411                  *  B-d  |  \
1412                  *       |   \      <- smaller scope is better if
1413                  *  B-15 |    \        if scope is enough for destination.
1414                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1415                  * d-C-1 | /
1416                  *       |/         <- greater is better
1417                  *   -C  /             if scope is not enough for destination.
1418                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1419                  *
1420                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1421                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1422                  * Assume B = 0 and we get C > 29.
1423                  */
1424                 ret = __ipv6_addr_src_scope(score->addr_type);
1425                 if (ret >= dst->scope)
1426                         ret = -ret;
1427                 else
1428                         ret -= 128;     /* 30 is enough */
1429                 score->scopedist = ret;
1430                 break;
1431         case IPV6_SADDR_RULE_PREFERRED:
1432             {
1433                 /* Rule 3: Avoid deprecated and optimistic addresses */
1434                 u8 avoid = IFA_F_DEPRECATED;
1435
1436                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1437                         avoid |= IFA_F_OPTIMISTIC;
1438                 ret = ipv6_saddr_preferred(score->addr_type) ||
1439                       !(score->ifa->flags & avoid);
1440                 break;
1441             }
1442 #ifdef CONFIG_IPV6_MIP6
1443         case IPV6_SADDR_RULE_HOA:
1444             {
1445                 /* Rule 4: Prefer home address */
1446                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1447                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1448                 break;
1449             }
1450 #endif
1451         case IPV6_SADDR_RULE_OIF:
1452                 /* Rule 5: Prefer outgoing interface */
1453                 ret = (!dst->ifindex ||
1454                        dst->ifindex == score->ifa->idev->dev->ifindex);
1455                 break;
1456         case IPV6_SADDR_RULE_LABEL:
1457                 /* Rule 6: Prefer matching label */
1458                 ret = ipv6_addr_label(net,
1459                                       &score->ifa->addr, score->addr_type,
1460                                       score->ifa->idev->dev->ifindex) == dst->label;
1461                 break;
1462         case IPV6_SADDR_RULE_PRIVACY:
1463             {
1464                 /* Rule 7: Prefer public address
1465                  * Note: prefer temporary address if use_tempaddr >= 2
1466                  */
1467                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1468                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1469                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1470                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1471                 break;
1472             }
1473         case IPV6_SADDR_RULE_ORCHID:
1474                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1475                  *          non-ORCHID vs non-ORCHID
1476                  */
1477                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1478                         ipv6_addr_orchid(dst->addr));
1479                 break;
1480         case IPV6_SADDR_RULE_PREFIX:
1481                 /* Rule 8: Use longest matching prefix */
1482                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1483                 if (ret > score->ifa->prefix_len)
1484                         ret = score->ifa->prefix_len;
1485                 score->matchlen = ret;
1486                 break;
1487 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1488         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1489                 /* Optimistic addresses still have lower precedence than other
1490                  * preferred addresses.
1491                  */
1492                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1493                 break;
1494 #endif
1495         default:
1496                 ret = 0;
1497         }
1498
1499         if (ret)
1500                 __set_bit(i, score->scorebits);
1501         score->rule = i;
1502 out:
1503         return ret;
1504 }
1505
1506 static int __ipv6_dev_get_saddr(struct net *net,
1507                                 struct ipv6_saddr_dst *dst,
1508                                 struct inet6_dev *idev,
1509                                 struct ipv6_saddr_score *scores,
1510                                 int hiscore_idx)
1511 {
1512         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1513
1514         read_lock_bh(&idev->lock);
1515         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1516                 int i;
1517
1518                 /*
1519                  * - Tentative Address (RFC2462 section 5.4)
1520                  *  - A tentative address is not considered
1521                  *    "assigned to an interface" in the traditional
1522                  *    sense, unless it is also flagged as optimistic.
1523                  * - Candidate Source Address (section 4)
1524                  *  - In any case, anycast addresses, multicast
1525                  *    addresses, and the unspecified address MUST
1526                  *    NOT be included in a candidate set.
1527                  */
1528                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1529                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1530                         continue;
1531
1532                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1533
1534                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1535                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1536                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1537                                             idev->dev->name);
1538                         continue;
1539                 }
1540
1541                 score->rule = -1;
1542                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1543
1544                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1545                         int minihiscore, miniscore;
1546
1547                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1548                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1549
1550                         if (minihiscore > miniscore) {
1551                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1552                                     score->scopedist > 0) {
1553                                         /*
1554                                          * special case:
1555                                          * each remaining entry
1556                                          * has too small (not enough)
1557                                          * scope, because ifa entries
1558                                          * are sorted by their scope
1559                                          * values.
1560                                          */
1561                                         goto out;
1562                                 }
1563                                 break;
1564                         } else if (minihiscore < miniscore) {
1565                                 if (hiscore->ifa)
1566                                         in6_ifa_put(hiscore->ifa);
1567
1568                                 in6_ifa_hold(score->ifa);
1569
1570                                 swap(hiscore, score);
1571                                 hiscore_idx = 1 - hiscore_idx;
1572
1573                                 /* restore our iterator */
1574                                 score->ifa = hiscore->ifa;
1575
1576                                 break;
1577                         }
1578                 }
1579         }
1580 out:
1581         read_unlock_bh(&idev->lock);
1582         return hiscore_idx;
1583 }
1584
1585 static int ipv6_get_saddr_master(struct net *net,
1586                                  const struct net_device *dst_dev,
1587                                  const struct net_device *master,
1588                                  struct ipv6_saddr_dst *dst,
1589                                  struct ipv6_saddr_score *scores,
1590                                  int hiscore_idx)
1591 {
1592         struct inet6_dev *idev;
1593
1594         idev = __in6_dev_get(dst_dev);
1595         if (idev)
1596                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1597                                                    scores, hiscore_idx);
1598
1599         idev = __in6_dev_get(master);
1600         if (idev)
1601                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1602                                                    scores, hiscore_idx);
1603
1604         return hiscore_idx;
1605 }
1606
1607 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1608                        const struct in6_addr *daddr, unsigned int prefs,
1609                        struct in6_addr *saddr)
1610 {
1611         struct ipv6_saddr_score scores[2], *hiscore;
1612         struct ipv6_saddr_dst dst;
1613         struct inet6_dev *idev;
1614         struct net_device *dev;
1615         int dst_type;
1616         bool use_oif_addr = false;
1617         int hiscore_idx = 0;
1618
1619         dst_type = __ipv6_addr_type(daddr);
1620         dst.addr = daddr;
1621         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1622         dst.scope = __ipv6_addr_src_scope(dst_type);
1623         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1624         dst.prefs = prefs;
1625
1626         scores[hiscore_idx].rule = -1;
1627         scores[hiscore_idx].ifa = NULL;
1628
1629         rcu_read_lock();
1630
1631         /* Candidate Source Address (section 4)
1632          *  - multicast and link-local destination address,
1633          *    the set of candidate source address MUST only
1634          *    include addresses assigned to interfaces
1635          *    belonging to the same link as the outgoing
1636          *    interface.
1637          * (- For site-local destination addresses, the
1638          *    set of candidate source addresses MUST only
1639          *    include addresses assigned to interfaces
1640          *    belonging to the same site as the outgoing
1641          *    interface.)
1642          *  - "It is RECOMMENDED that the candidate source addresses
1643          *    be the set of unicast addresses assigned to the
1644          *    interface that will be used to send to the destination
1645          *    (the 'outgoing' interface)." (RFC 6724)
1646          */
1647         if (dst_dev) {
1648                 idev = __in6_dev_get(dst_dev);
1649                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1650                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1651                     (idev && idev->cnf.use_oif_addrs_only)) {
1652                         use_oif_addr = true;
1653                 }
1654         }
1655
1656         if (use_oif_addr) {
1657                 if (idev)
1658                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1659         } else {
1660                 const struct net_device *master;
1661                 int master_idx = 0;
1662
1663                 /* if dst_dev exists and is enslaved to an L3 device, then
1664                  * prefer addresses from dst_dev and then the master over
1665                  * any other enslaved devices in the L3 domain.
1666                  */
1667                 master = l3mdev_master_dev_rcu(dst_dev);
1668                 if (master) {
1669                         master_idx = master->ifindex;
1670
1671                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1672                                                             master, &dst,
1673                                                             scores, hiscore_idx);
1674
1675                         if (scores[hiscore_idx].ifa)
1676                                 goto out;
1677                 }
1678
1679                 for_each_netdev_rcu(net, dev) {
1680                         /* only consider addresses on devices in the
1681                          * same L3 domain
1682                          */
1683                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1684                                 continue;
1685                         idev = __in6_dev_get(dev);
1686                         if (!idev)
1687                                 continue;
1688                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1689                 }
1690         }
1691
1692 out:
1693         rcu_read_unlock();
1694
1695         hiscore = &scores[hiscore_idx];
1696         if (!hiscore->ifa)
1697                 return -EADDRNOTAVAIL;
1698
1699         *saddr = hiscore->ifa->addr;
1700         in6_ifa_put(hiscore->ifa);
1701         return 0;
1702 }
1703 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1704
1705 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1706                       u32 banned_flags)
1707 {
1708         struct inet6_ifaddr *ifp;
1709         int err = -EADDRNOTAVAIL;
1710
1711         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1712                 if (ifp->scope > IFA_LINK)
1713                         break;
1714                 if (ifp->scope == IFA_LINK &&
1715                     !(ifp->flags & banned_flags)) {
1716                         *addr = ifp->addr;
1717                         err = 0;
1718                         break;
1719                 }
1720         }
1721         return err;
1722 }
1723
1724 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1725                     u32 banned_flags)
1726 {
1727         struct inet6_dev *idev;
1728         int err = -EADDRNOTAVAIL;
1729
1730         rcu_read_lock();
1731         idev = __in6_dev_get(dev);
1732         if (idev) {
1733                 read_lock_bh(&idev->lock);
1734                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1735                 read_unlock_bh(&idev->lock);
1736         }
1737         rcu_read_unlock();
1738         return err;
1739 }
1740
1741 static int ipv6_count_addresses(struct inet6_dev *idev)
1742 {
1743         int cnt = 0;
1744         struct inet6_ifaddr *ifp;
1745
1746         read_lock_bh(&idev->lock);
1747         list_for_each_entry(ifp, &idev->addr_list, if_list)
1748                 cnt++;
1749         read_unlock_bh(&idev->lock);
1750         return cnt;
1751 }
1752
1753 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1754                   const struct net_device *dev, int strict)
1755 {
1756         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1757 }
1758 EXPORT_SYMBOL(ipv6_chk_addr);
1759
1760 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1761                             const struct net_device *dev, int strict,
1762                             u32 banned_flags)
1763 {
1764         struct inet6_ifaddr *ifp;
1765         unsigned int hash = inet6_addr_hash(addr);
1766         u32 ifp_flags;
1767
1768         rcu_read_lock_bh();
1769         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1770                 if (!net_eq(dev_net(ifp->idev->dev), net))
1771                         continue;
1772                 /* Decouple optimistic from tentative for evaluation here.
1773                  * Ban optimistic addresses explicitly, when required.
1774                  */
1775                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1776                             ? (ifp->flags&~IFA_F_TENTATIVE)
1777                             : ifp->flags;
1778                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1779                     !(ifp_flags&banned_flags) &&
1780                     (!dev || ifp->idev->dev == dev ||
1781                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1782                         rcu_read_unlock_bh();
1783                         return 1;
1784                 }
1785         }
1786
1787         rcu_read_unlock_bh();
1788         return 0;
1789 }
1790 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1791
1792 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1793                                struct net_device *dev)
1794 {
1795         unsigned int hash = inet6_addr_hash(addr);
1796         struct inet6_ifaddr *ifp;
1797
1798         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1799                 if (!net_eq(dev_net(ifp->idev->dev), net))
1800                         continue;
1801                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1802                         if (!dev || ifp->idev->dev == dev)
1803                                 return true;
1804                 }
1805         }
1806         return false;
1807 }
1808
1809 /* Compares an address/prefix_len with addresses on device @dev.
1810  * If one is found it returns true.
1811  */
1812 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1813         const unsigned int prefix_len, struct net_device *dev)
1814 {
1815         struct inet6_dev *idev;
1816         struct inet6_ifaddr *ifa;
1817         bool ret = false;
1818
1819         rcu_read_lock();
1820         idev = __in6_dev_get(dev);
1821         if (idev) {
1822                 read_lock_bh(&idev->lock);
1823                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1824                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1825                         if (ret)
1826                                 break;
1827                 }
1828                 read_unlock_bh(&idev->lock);
1829         }
1830         rcu_read_unlock();
1831
1832         return ret;
1833 }
1834 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1835
1836 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1837 {
1838         struct inet6_dev *idev;
1839         struct inet6_ifaddr *ifa;
1840         int     onlink;
1841
1842         onlink = 0;
1843         rcu_read_lock();
1844         idev = __in6_dev_get(dev);
1845         if (idev) {
1846                 read_lock_bh(&idev->lock);
1847                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1848                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1849                                                    ifa->prefix_len);
1850                         if (onlink)
1851                                 break;
1852                 }
1853                 read_unlock_bh(&idev->lock);
1854         }
1855         rcu_read_unlock();
1856         return onlink;
1857 }
1858 EXPORT_SYMBOL(ipv6_chk_prefix);
1859
1860 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1861                                      struct net_device *dev, int strict)
1862 {
1863         struct inet6_ifaddr *ifp, *result = NULL;
1864         unsigned int hash = inet6_addr_hash(addr);
1865
1866         rcu_read_lock_bh();
1867         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1868                 if (!net_eq(dev_net(ifp->idev->dev), net))
1869                         continue;
1870                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1871                         if (!dev || ifp->idev->dev == dev ||
1872                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1873                                 result = ifp;
1874                                 in6_ifa_hold(ifp);
1875                                 break;
1876                         }
1877                 }
1878         }
1879         rcu_read_unlock_bh();
1880
1881         return result;
1882 }
1883
1884 /* Gets referenced address, destroys ifaddr */
1885
1886 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1887 {
1888         if (dad_failed)
1889                 ifp->flags |= IFA_F_DADFAILED;
1890
1891         if (ifp->flags&IFA_F_PERMANENT) {
1892                 spin_lock_bh(&ifp->lock);
1893                 addrconf_del_dad_work(ifp);
1894                 ifp->flags |= IFA_F_TENTATIVE;
1895                 spin_unlock_bh(&ifp->lock);
1896                 if (dad_failed)
1897                         ipv6_ifa_notify(0, ifp);
1898                 in6_ifa_put(ifp);
1899         } else if (ifp->flags&IFA_F_TEMPORARY) {
1900                 struct inet6_ifaddr *ifpub;
1901                 spin_lock_bh(&ifp->lock);
1902                 ifpub = ifp->ifpub;
1903                 if (ifpub) {
1904                         in6_ifa_hold(ifpub);
1905                         spin_unlock_bh(&ifp->lock);
1906                         ipv6_create_tempaddr(ifpub, ifp);
1907                         in6_ifa_put(ifpub);
1908                 } else {
1909                         spin_unlock_bh(&ifp->lock);
1910                 }
1911                 ipv6_del_addr(ifp);
1912         } else {
1913                 ipv6_del_addr(ifp);
1914         }
1915 }
1916
1917 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1918 {
1919         int err = -ENOENT;
1920
1921         spin_lock_bh(&ifp->lock);
1922         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1923                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1924                 err = 0;
1925         }
1926         spin_unlock_bh(&ifp->lock);
1927
1928         return err;
1929 }
1930
1931 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1932 {
1933         struct inet6_dev *idev = ifp->idev;
1934         struct net *net = dev_net(ifp->idev->dev);
1935
1936         if (addrconf_dad_end(ifp)) {
1937                 in6_ifa_put(ifp);
1938                 return;
1939         }
1940
1941         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1942                              ifp->idev->dev->name, &ifp->addr);
1943
1944         spin_lock_bh(&ifp->lock);
1945
1946         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1947                 int scope = ifp->scope;
1948                 u32 flags = ifp->flags;
1949                 struct in6_addr new_addr;
1950                 struct inet6_ifaddr *ifp2;
1951                 u32 valid_lft, preferred_lft;
1952                 int pfxlen = ifp->prefix_len;
1953                 int retries = ifp->stable_privacy_retry + 1;
1954
1955                 if (retries > net->ipv6.sysctl.idgen_retries) {
1956                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1957                                              ifp->idev->dev->name);
1958                         goto errdad;
1959                 }
1960
1961                 new_addr = ifp->addr;
1962                 if (ipv6_generate_stable_address(&new_addr, retries,
1963                                                  idev))
1964                         goto errdad;
1965
1966                 valid_lft = ifp->valid_lft;
1967                 preferred_lft = ifp->prefered_lft;
1968
1969                 spin_unlock_bh(&ifp->lock);
1970
1971                 if (idev->cnf.max_addresses &&
1972                     ipv6_count_addresses(idev) >=
1973                     idev->cnf.max_addresses)
1974                         goto lock_errdad;
1975
1976                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1977                                      ifp->idev->dev->name);
1978
1979                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1980                                      scope, flags, valid_lft,
1981                                      preferred_lft);
1982                 if (IS_ERR(ifp2))
1983                         goto lock_errdad;
1984
1985                 spin_lock_bh(&ifp2->lock);
1986                 ifp2->stable_privacy_retry = retries;
1987                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1988                 spin_unlock_bh(&ifp2->lock);
1989
1990                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1991                 in6_ifa_put(ifp2);
1992 lock_errdad:
1993                 spin_lock_bh(&ifp->lock);
1994         }
1995
1996 errdad:
1997         /* transition from _POSTDAD to _ERRDAD */
1998         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1999         spin_unlock_bh(&ifp->lock);
2000
2001         addrconf_mod_dad_work(ifp, 0);
2002         in6_ifa_put(ifp);
2003 }
2004
2005 /* Join to solicited addr multicast group.
2006  * caller must hold RTNL */
2007 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2008 {
2009         struct in6_addr maddr;
2010
2011         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2012                 return;
2013
2014         addrconf_addr_solict_mult(addr, &maddr);
2015         ipv6_dev_mc_inc(dev, &maddr);
2016 }
2017
2018 /* caller must hold RTNL */
2019 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2020 {
2021         struct in6_addr maddr;
2022
2023         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2024                 return;
2025
2026         addrconf_addr_solict_mult(addr, &maddr);
2027         __ipv6_dev_mc_dec(idev, &maddr);
2028 }
2029
2030 /* caller must hold RTNL */
2031 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2032 {
2033         struct in6_addr addr;
2034
2035         if (ifp->prefix_len >= 127) /* RFC 6164 */
2036                 return;
2037         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2038         if (ipv6_addr_any(&addr))
2039                 return;
2040         __ipv6_dev_ac_inc(ifp->idev, &addr);
2041 }
2042
2043 /* caller must hold RTNL */
2044 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2045 {
2046         struct in6_addr addr;
2047
2048         if (ifp->prefix_len >= 127) /* RFC 6164 */
2049                 return;
2050         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2051         if (ipv6_addr_any(&addr))
2052                 return;
2053         __ipv6_dev_ac_dec(ifp->idev, &addr);
2054 }
2055
2056 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2057 {
2058         if (dev->addr_len != EUI64_ADDR_LEN)
2059                 return -1;
2060         memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2061         eui[0] ^= 2;
2062         return 0;
2063 }
2064
2065 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2066 {
2067         union fwnet_hwaddr *ha;
2068
2069         if (dev->addr_len != FWNET_ALEN)
2070                 return -1;
2071
2072         ha = (union fwnet_hwaddr *)dev->dev_addr;
2073
2074         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2075         eui[0] ^= 2;
2076         return 0;
2077 }
2078
2079 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2080 {
2081         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2082         if (dev->addr_len != ARCNET_ALEN)
2083                 return -1;
2084         memset(eui, 0, 7);
2085         eui[7] = *(u8 *)dev->dev_addr;
2086         return 0;
2087 }
2088
2089 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2090 {
2091         if (dev->addr_len != INFINIBAND_ALEN)
2092                 return -1;
2093         memcpy(eui, dev->dev_addr + 12, 8);
2094         eui[0] |= 2;
2095         return 0;
2096 }
2097
2098 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2099 {
2100         if (addr == 0)
2101                 return -1;
2102         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2103                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2104                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2105                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2106                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2107                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2108         eui[1] = 0;
2109         eui[2] = 0x5E;
2110         eui[3] = 0xFE;
2111         memcpy(eui + 4, &addr, 4);
2112         return 0;
2113 }
2114
2115 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2116 {
2117         if (dev->priv_flags & IFF_ISATAP)
2118                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2119         return -1;
2120 }
2121
2122 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2123 {
2124         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2125 }
2126
2127 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2128 {
2129         memcpy(eui, dev->perm_addr, 3);
2130         memcpy(eui + 5, dev->perm_addr + 3, 3);
2131         eui[3] = 0xFF;
2132         eui[4] = 0xFE;
2133         eui[0] ^= 2;
2134         return 0;
2135 }
2136
2137 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2138 {
2139         switch (dev->type) {
2140         case ARPHRD_ETHER:
2141         case ARPHRD_FDDI:
2142                 return addrconf_ifid_eui48(eui, dev);
2143         case ARPHRD_ARCNET:
2144                 return addrconf_ifid_arcnet(eui, dev);
2145         case ARPHRD_INFINIBAND:
2146                 return addrconf_ifid_infiniband(eui, dev);
2147         case ARPHRD_SIT:
2148                 return addrconf_ifid_sit(eui, dev);
2149         case ARPHRD_IPGRE:
2150         case ARPHRD_TUNNEL:
2151                 return addrconf_ifid_gre(eui, dev);
2152         case ARPHRD_6LOWPAN:
2153                 return addrconf_ifid_eui64(eui, dev);
2154         case ARPHRD_IEEE1394:
2155                 return addrconf_ifid_ieee1394(eui, dev);
2156         case ARPHRD_TUNNEL6:
2157         case ARPHRD_IP6GRE:
2158                 return addrconf_ifid_ip6tnl(eui, dev);
2159         }
2160         return -1;
2161 }
2162
2163 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2164 {
2165         int err = -1;
2166         struct inet6_ifaddr *ifp;
2167
2168         read_lock_bh(&idev->lock);
2169         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2170                 if (ifp->scope > IFA_LINK)
2171                         break;
2172                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2173                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2174                         err = 0;
2175                         break;
2176                 }
2177         }
2178         read_unlock_bh(&idev->lock);
2179         return err;
2180 }
2181
2182 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2183 static void ipv6_regen_rndid(struct inet6_dev *idev)
2184 {
2185 regen:
2186         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2187         idev->rndid[0] &= ~0x02;
2188
2189         /*
2190          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2191          * check if generated address is not inappropriate
2192          *
2193          *  - Reserved subnet anycast (RFC 2526)
2194          *      11111101 11....11 1xxxxxxx
2195          *  - ISATAP (RFC4214) 6.1
2196          *      00-00-5E-FE-xx-xx-xx-xx
2197          *  - value 0
2198          *  - XXX: already assigned to an address on the device
2199          */
2200         if (idev->rndid[0] == 0xfd &&
2201             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2202             (idev->rndid[7]&0x80))
2203                 goto regen;
2204         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2205                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2206                         goto regen;
2207                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2208                         goto regen;
2209         }
2210 }
2211
2212 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2213 {
2214         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2215                 ipv6_regen_rndid(idev);
2216 }
2217
2218 /*
2219  *      Add prefix route.
2220  */
2221
2222 static void
2223 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2224                       unsigned long expires, u32 flags)
2225 {
2226         struct fib6_config cfg = {
2227                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2228                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2229                 .fc_ifindex = dev->ifindex,
2230                 .fc_expires = expires,
2231                 .fc_dst_len = plen,
2232                 .fc_flags = RTF_UP | flags,
2233                 .fc_nlinfo.nl_net = dev_net(dev),
2234                 .fc_protocol = RTPROT_KERNEL,
2235         };
2236
2237         cfg.fc_dst = *pfx;
2238
2239         /* Prevent useless cloning on PtP SIT.
2240            This thing is done here expecting that the whole
2241            class of non-broadcast devices need not cloning.
2242          */
2243 #if IS_ENABLED(CONFIG_IPV6_SIT)
2244         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2245                 cfg.fc_flags |= RTF_NONEXTHOP;
2246 #endif
2247
2248         ip6_route_add(&cfg);
2249 }
2250
2251
2252 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2253                                                   int plen,
2254                                                   const struct net_device *dev,
2255                                                   u32 flags, u32 noflags)
2256 {
2257         struct fib6_node *fn;
2258         struct rt6_info *rt = NULL;
2259         struct fib6_table *table;
2260         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2261
2262         table = fib6_get_table(dev_net(dev), tb_id);
2263         if (!table)
2264                 return NULL;
2265
2266         read_lock_bh(&table->tb6_lock);
2267         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2268         if (!fn)
2269                 goto out;
2270
2271         noflags |= RTF_CACHE;
2272         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2273                 if (rt->dst.dev->ifindex != dev->ifindex)
2274                         continue;
2275                 if ((rt->rt6i_flags & flags) != flags)
2276                         continue;
2277                 if ((rt->rt6i_flags & noflags) != 0)
2278                         continue;
2279                 dst_hold(&rt->dst);
2280                 break;
2281         }
2282 out:
2283         read_unlock_bh(&table->tb6_lock);
2284         return rt;
2285 }
2286
2287
2288 /* Create "default" multicast route to the interface */
2289
2290 static void addrconf_add_mroute(struct net_device *dev)
2291 {
2292         struct fib6_config cfg = {
2293                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2294                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2295                 .fc_ifindex = dev->ifindex,
2296                 .fc_dst_len = 8,
2297                 .fc_flags = RTF_UP,
2298                 .fc_nlinfo.nl_net = dev_net(dev),
2299         };
2300
2301         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2302
2303         ip6_route_add(&cfg);
2304 }
2305
2306 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2307 {
2308         struct inet6_dev *idev;
2309
2310         ASSERT_RTNL();
2311
2312         idev = ipv6_find_idev(dev);
2313         if (!idev)
2314                 return ERR_PTR(-ENOBUFS);
2315
2316         if (idev->cnf.disable_ipv6)
2317                 return ERR_PTR(-EACCES);
2318
2319         /* Add default multicast route */
2320         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2321                 addrconf_add_mroute(dev);
2322
2323         return idev;
2324 }
2325
2326 static void manage_tempaddrs(struct inet6_dev *idev,
2327                              struct inet6_ifaddr *ifp,
2328                              __u32 valid_lft, __u32 prefered_lft,
2329                              bool create, unsigned long now)
2330 {
2331         u32 flags;
2332         struct inet6_ifaddr *ift;
2333
2334         read_lock_bh(&idev->lock);
2335         /* update all temporary addresses in the list */
2336         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2337                 int age, max_valid, max_prefered;
2338
2339                 if (ifp != ift->ifpub)
2340                         continue;
2341
2342                 /* RFC 4941 section 3.3:
2343                  * If a received option will extend the lifetime of a public
2344                  * address, the lifetimes of temporary addresses should
2345                  * be extended, subject to the overall constraint that no
2346                  * temporary addresses should ever remain "valid" or "preferred"
2347                  * for a time longer than (TEMP_VALID_LIFETIME) or
2348                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2349                  */
2350                 age = (now - ift->cstamp) / HZ;
2351                 max_valid = idev->cnf.temp_valid_lft - age;
2352                 if (max_valid < 0)
2353                         max_valid = 0;
2354
2355                 max_prefered = idev->cnf.temp_prefered_lft -
2356                                idev->desync_factor - age;
2357                 if (max_prefered < 0)
2358                         max_prefered = 0;
2359
2360                 if (valid_lft > max_valid)
2361                         valid_lft = max_valid;
2362
2363                 if (prefered_lft > max_prefered)
2364                         prefered_lft = max_prefered;
2365
2366                 spin_lock(&ift->lock);
2367                 flags = ift->flags;
2368                 ift->valid_lft = valid_lft;
2369                 ift->prefered_lft = prefered_lft;
2370                 ift->tstamp = now;
2371                 if (prefered_lft > 0)
2372                         ift->flags &= ~IFA_F_DEPRECATED;
2373
2374                 spin_unlock(&ift->lock);
2375                 if (!(flags&IFA_F_TENTATIVE))
2376                         ipv6_ifa_notify(0, ift);
2377         }
2378
2379         if ((create || list_empty(&idev->tempaddr_list)) &&
2380             idev->cnf.use_tempaddr > 0) {
2381                 /* When a new public address is created as described
2382                  * in [ADDRCONF], also create a new temporary address.
2383                  * Also create a temporary address if it's enabled but
2384                  * no temporary address currently exists.
2385                  */
2386                 read_unlock_bh(&idev->lock);
2387                 ipv6_create_tempaddr(ifp, NULL);
2388         } else {
2389                 read_unlock_bh(&idev->lock);
2390         }
2391 }
2392
2393 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2394 {
2395         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2396                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2397 }
2398
2399 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2400                                  const struct prefix_info *pinfo,
2401                                  struct inet6_dev *in6_dev,
2402                                  const struct in6_addr *addr, int addr_type,
2403                                  u32 addr_flags, bool sllao, bool tokenized,
2404                                  __u32 valid_lft, u32 prefered_lft)
2405 {
2406         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2407         int create = 0, update_lft = 0;
2408
2409         if (!ifp && valid_lft) {
2410                 int max_addresses = in6_dev->cnf.max_addresses;
2411
2412 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2413                 if (in6_dev->cnf.optimistic_dad &&
2414                     !net->ipv6.devconf_all->forwarding && sllao)
2415                         addr_flags |= IFA_F_OPTIMISTIC;
2416 #endif
2417
2418                 /* Do not allow to create too much of autoconfigured
2419                  * addresses; this would be too easy way to crash kernel.
2420                  */
2421                 if (!max_addresses ||
2422                     ipv6_count_addresses(in6_dev) < max_addresses)
2423                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2424                                             pinfo->prefix_len,
2425                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2426                                             addr_flags, valid_lft,
2427                                             prefered_lft);
2428
2429                 if (IS_ERR_OR_NULL(ifp))
2430                         return -1;
2431
2432                 update_lft = 0;
2433                 create = 1;
2434                 spin_lock_bh(&ifp->lock);
2435                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2436                 ifp->cstamp = jiffies;
2437                 ifp->tokenized = tokenized;
2438                 spin_unlock_bh(&ifp->lock);
2439                 addrconf_dad_start(ifp);
2440         }
2441
2442         if (ifp) {
2443                 u32 flags;
2444                 unsigned long now;
2445                 u32 stored_lft;
2446
2447                 /* update lifetime (RFC2462 5.5.3 e) */
2448                 spin_lock_bh(&ifp->lock);
2449                 now = jiffies;
2450                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2451                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2452                 else
2453                         stored_lft = 0;
2454                 if (!update_lft && !create && stored_lft) {
2455                         const u32 minimum_lft = min_t(u32,
2456                                 stored_lft, MIN_VALID_LIFETIME);
2457                         valid_lft = max(valid_lft, minimum_lft);
2458
2459                         /* RFC4862 Section 5.5.3e:
2460                          * "Note that the preferred lifetime of the
2461                          *  corresponding address is always reset to
2462                          *  the Preferred Lifetime in the received
2463                          *  Prefix Information option, regardless of
2464                          *  whether the valid lifetime is also reset or
2465                          *  ignored."
2466                          *
2467                          * So we should always update prefered_lft here.
2468                          */
2469                         update_lft = 1;
2470                 }
2471
2472                 if (update_lft) {
2473                         ifp->valid_lft = valid_lft;
2474                         ifp->prefered_lft = prefered_lft;
2475                         ifp->tstamp = now;
2476                         flags = ifp->flags;
2477                         ifp->flags &= ~IFA_F_DEPRECATED;
2478                         spin_unlock_bh(&ifp->lock);
2479
2480                         if (!(flags&IFA_F_TENTATIVE))
2481                                 ipv6_ifa_notify(0, ifp);
2482                 } else
2483                         spin_unlock_bh(&ifp->lock);
2484
2485                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2486                                  create, now);
2487
2488                 in6_ifa_put(ifp);
2489                 addrconf_verify();
2490         }
2491
2492         return 0;
2493 }
2494 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2495
2496 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2497 {
2498         struct prefix_info *pinfo;
2499         __u32 valid_lft;
2500         __u32 prefered_lft;
2501         int addr_type, err;
2502         u32 addr_flags = 0;
2503         struct inet6_dev *in6_dev;
2504         struct net *net = dev_net(dev);
2505
2506         pinfo = (struct prefix_info *) opt;
2507
2508         if (len < sizeof(struct prefix_info)) {
2509                 ADBG("addrconf: prefix option too short\n");
2510                 return;
2511         }
2512
2513         /*
2514          *      Validation checks ([ADDRCONF], page 19)
2515          */
2516
2517         addr_type = ipv6_addr_type(&pinfo->prefix);
2518
2519         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2520                 return;
2521
2522         valid_lft = ntohl(pinfo->valid);
2523         prefered_lft = ntohl(pinfo->prefered);
2524
2525         if (prefered_lft > valid_lft) {
2526                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2527                 return;
2528         }
2529
2530         in6_dev = in6_dev_get(dev);
2531
2532         if (!in6_dev) {
2533                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2534                                     dev->name);
2535                 return;
2536         }
2537
2538         /*
2539          *      Two things going on here:
2540          *      1) Add routes for on-link prefixes
2541          *      2) Configure prefixes with the auto flag set
2542          */
2543
2544         if (pinfo->onlink) {
2545                 struct rt6_info *rt;
2546                 unsigned long rt_expires;
2547
2548                 /* Avoid arithmetic overflow. Really, we could
2549                  * save rt_expires in seconds, likely valid_lft,
2550                  * but it would require division in fib gc, that it
2551                  * not good.
2552                  */
2553                 if (HZ > USER_HZ)
2554                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2555                 else
2556                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2557
2558                 if (addrconf_finite_timeout(rt_expires))
2559                         rt_expires *= HZ;
2560
2561                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2562                                                pinfo->prefix_len,
2563                                                dev,
2564                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2565                                                RTF_GATEWAY | RTF_DEFAULT);
2566
2567                 if (rt) {
2568                         /* Autoconf prefix route */
2569                         if (valid_lft == 0) {
2570                                 ip6_del_rt(rt);
2571                                 rt = NULL;
2572                         } else if (addrconf_finite_timeout(rt_expires)) {
2573                                 /* not infinity */
2574                                 rt6_set_expires(rt, jiffies + rt_expires);
2575                         } else {
2576                                 rt6_clean_expires(rt);
2577                         }
2578                 } else if (valid_lft) {
2579                         clock_t expires = 0;
2580                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2581                         if (addrconf_finite_timeout(rt_expires)) {
2582                                 /* not infinity */
2583                                 flags |= RTF_EXPIRES;
2584                                 expires = jiffies_to_clock_t(rt_expires);
2585                         }
2586                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2587                                               dev, expires, flags);
2588                 }
2589                 ip6_rt_put(rt);
2590         }
2591
2592         /* Try to figure out our local address for this prefix */
2593
2594         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2595                 struct in6_addr addr;
2596                 bool tokenized = false, dev_addr_generated = false;
2597
2598                 if (pinfo->prefix_len == 64) {
2599                         memcpy(&addr, &pinfo->prefix, 8);
2600
2601                         if (!ipv6_addr_any(&in6_dev->token)) {
2602                                 read_lock_bh(&in6_dev->lock);
2603                                 memcpy(addr.s6_addr + 8,
2604                                        in6_dev->token.s6_addr + 8, 8);
2605                                 read_unlock_bh(&in6_dev->lock);
2606                                 tokenized = true;
2607                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2608                                    !ipv6_generate_stable_address(&addr, 0,
2609                                                                  in6_dev)) {
2610                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2611                                 goto ok;
2612                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2613                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2614                                 goto put;
2615                         } else {
2616                                 dev_addr_generated = true;
2617                         }
2618                         goto ok;
2619                 }
2620                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2621                                     pinfo->prefix_len);
2622                 goto put;
2623
2624 ok:
2625                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2626                                                    &addr, addr_type,
2627                                                    addr_flags, sllao,
2628                                                    tokenized, valid_lft,
2629                                                    prefered_lft);
2630                 if (err)
2631                         goto put;
2632
2633                 /* Ignore error case here because previous prefix add addr was
2634                  * successful which will be notified.
2635                  */
2636                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2637                                               addr_type, addr_flags, sllao,
2638                                               tokenized, valid_lft,
2639                                               prefered_lft,
2640                                               dev_addr_generated);
2641         }
2642         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2643 put:
2644         in6_dev_put(in6_dev);
2645 }
2646
2647 /*
2648  *      Set destination address.
2649  *      Special case for SIT interfaces where we create a new "virtual"
2650  *      device.
2651  */
2652 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2653 {
2654         struct in6_ifreq ireq;
2655         struct net_device *dev;
2656         int err = -EINVAL;
2657
2658         rtnl_lock();
2659
2660         err = -EFAULT;
2661         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2662                 goto err_exit;
2663
2664         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2665
2666         err = -ENODEV;
2667         if (!dev)
2668                 goto err_exit;
2669
2670 #if IS_ENABLED(CONFIG_IPV6_SIT)
2671         if (dev->type == ARPHRD_SIT) {
2672                 const struct net_device_ops *ops = dev->netdev_ops;
2673                 struct ifreq ifr;
2674                 struct ip_tunnel_parm p;
2675
2676                 err = -EADDRNOTAVAIL;
2677                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2678                         goto err_exit;
2679
2680                 memset(&p, 0, sizeof(p));
2681                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2682                 p.iph.saddr = 0;
2683                 p.iph.version = 4;
2684                 p.iph.ihl = 5;
2685                 p.iph.protocol = IPPROTO_IPV6;
2686                 p.iph.ttl = 64;
2687                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2688
2689                 if (ops->ndo_do_ioctl) {
2690                         mm_segment_t oldfs = get_fs();
2691
2692                         set_fs(KERNEL_DS);
2693                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2694                         set_fs(oldfs);
2695                 } else
2696                         err = -EOPNOTSUPP;
2697
2698                 if (err == 0) {
2699                         err = -ENOBUFS;
2700                         dev = __dev_get_by_name(net, p.name);
2701                         if (!dev)
2702                                 goto err_exit;
2703                         err = dev_open(dev);
2704                 }
2705         }
2706 #endif
2707
2708 err_exit:
2709         rtnl_unlock();
2710         return err;
2711 }
2712
2713 static int ipv6_mc_config(struct sock *sk, bool join,
2714                           const struct in6_addr *addr, int ifindex)
2715 {
2716         int ret;
2717
2718         ASSERT_RTNL();
2719
2720         lock_sock(sk);
2721         if (join)
2722                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2723         else
2724                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2725         release_sock(sk);
2726
2727         return ret;
2728 }
2729
2730 /*
2731  *      Manual configuration of address on an interface
2732  */
2733 static int inet6_addr_add(struct net *net, int ifindex,
2734                           const struct in6_addr *pfx,
2735                           const struct in6_addr *peer_pfx,
2736                           unsigned int plen, __u32 ifa_flags,
2737                           __u32 prefered_lft, __u32 valid_lft)
2738 {
2739         struct inet6_ifaddr *ifp;
2740         struct inet6_dev *idev;
2741         struct net_device *dev;
2742         unsigned long timeout;
2743         clock_t expires;
2744         int scope;
2745         u32 flags;
2746
2747         ASSERT_RTNL();
2748
2749         if (plen > 128)
2750                 return -EINVAL;
2751
2752         /* check the lifetime */
2753         if (!valid_lft || prefered_lft > valid_lft)
2754                 return -EINVAL;
2755
2756         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2757                 return -EINVAL;
2758
2759         dev = __dev_get_by_index(net, ifindex);
2760         if (!dev)
2761                 return -ENODEV;
2762
2763         idev = addrconf_add_dev(dev);
2764         if (IS_ERR(idev))
2765                 return PTR_ERR(idev);
2766
2767         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2768                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2769                                          true, pfx, ifindex);
2770
2771                 if (ret < 0)
2772                         return ret;
2773         }
2774
2775         scope = ipv6_addr_scope(pfx);
2776
2777         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2778         if (addrconf_finite_timeout(timeout)) {
2779                 expires = jiffies_to_clock_t(timeout * HZ);
2780                 valid_lft = timeout;
2781                 flags = RTF_EXPIRES;
2782         } else {
2783                 expires = 0;
2784                 flags = 0;
2785                 ifa_flags |= IFA_F_PERMANENT;
2786         }
2787
2788         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2789         if (addrconf_finite_timeout(timeout)) {
2790                 if (timeout == 0)
2791                         ifa_flags |= IFA_F_DEPRECATED;
2792                 prefered_lft = timeout;
2793         }
2794
2795         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2796                             valid_lft, prefered_lft);
2797
2798         if (!IS_ERR(ifp)) {
2799                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2800                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2801                                               expires, flags);
2802                 }
2803
2804                 /*
2805                  * Note that section 3.1 of RFC 4429 indicates
2806                  * that the Optimistic flag should not be set for
2807                  * manually configured addresses
2808                  */
2809                 addrconf_dad_start(ifp);
2810                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2811                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2812                                          true, jiffies);
2813                 in6_ifa_put(ifp);
2814                 addrconf_verify_rtnl();
2815                 return 0;
2816         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2817                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2818                                false, pfx, ifindex);
2819         }
2820
2821         return PTR_ERR(ifp);
2822 }
2823
2824 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2825                           const struct in6_addr *pfx, unsigned int plen)
2826 {
2827         struct inet6_ifaddr *ifp;
2828         struct inet6_dev *idev;
2829         struct net_device *dev;
2830
2831         if (plen > 128)
2832                 return -EINVAL;
2833
2834         dev = __dev_get_by_index(net, ifindex);
2835         if (!dev)
2836                 return -ENODEV;
2837
2838         idev = __in6_dev_get(dev);
2839         if (!idev)
2840                 return -ENXIO;
2841
2842         read_lock_bh(&idev->lock);
2843         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2844                 if (ifp->prefix_len == plen &&
2845                     ipv6_addr_equal(pfx, &ifp->addr)) {
2846                         in6_ifa_hold(ifp);
2847                         read_unlock_bh(&idev->lock);
2848
2849                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2850                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2851                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2852                                                  jiffies);
2853                         ipv6_del_addr(ifp);
2854                         addrconf_verify_rtnl();
2855                         if (ipv6_addr_is_multicast(pfx)) {
2856                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2857                                                false, pfx, dev->ifindex);
2858                         }
2859                         return 0;
2860                 }
2861         }
2862         read_unlock_bh(&idev->lock);
2863         return -EADDRNOTAVAIL;
2864 }
2865
2866
2867 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2868 {
2869         struct in6_ifreq ireq;
2870         int err;
2871
2872         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2873                 return -EPERM;
2874
2875         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2876                 return -EFAULT;
2877
2878         rtnl_lock();
2879         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2880                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2881                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2882         rtnl_unlock();
2883         return err;
2884 }
2885
2886 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2887 {
2888         struct in6_ifreq ireq;
2889         int err;
2890
2891         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2892                 return -EPERM;
2893
2894         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2895                 return -EFAULT;
2896
2897         rtnl_lock();
2898         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2899                              ireq.ifr6_prefixlen);
2900         rtnl_unlock();
2901         return err;
2902 }
2903
2904 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2905                      int plen, int scope)
2906 {
2907         struct inet6_ifaddr *ifp;
2908
2909         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2910                             scope, IFA_F_PERMANENT,
2911                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2912         if (!IS_ERR(ifp)) {
2913                 spin_lock_bh(&ifp->lock);
2914                 ifp->flags &= ~IFA_F_TENTATIVE;
2915                 spin_unlock_bh(&ifp->lock);
2916                 rt_genid_bump_ipv6(dev_net(idev->dev));
2917                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2918                 in6_ifa_put(ifp);
2919         }
2920 }
2921
2922 #if IS_ENABLED(CONFIG_IPV6_SIT)
2923 static void sit_add_v4_addrs(struct inet6_dev *idev)
2924 {
2925         struct in6_addr addr;
2926         struct net_device *dev;
2927         struct net *net = dev_net(idev->dev);
2928         int scope, plen;
2929         u32 pflags = 0;
2930
2931         ASSERT_RTNL();
2932
2933         memset(&addr, 0, sizeof(struct in6_addr));
2934         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2935
2936         if (idev->dev->flags&IFF_POINTOPOINT) {
2937                 addr.s6_addr32[0] = htonl(0xfe800000);
2938                 scope = IFA_LINK;
2939                 plen = 64;
2940         } else {
2941                 scope = IPV6_ADDR_COMPATv4;
2942                 plen = 96;
2943                 pflags |= RTF_NONEXTHOP;
2944         }
2945
2946         if (addr.s6_addr32[3]) {
2947                 add_addr(idev, &addr, plen, scope);
2948                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2949                 return;
2950         }
2951
2952         for_each_netdev(net, dev) {
2953                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2954                 if (in_dev && (dev->flags & IFF_UP)) {
2955                         struct in_ifaddr *ifa;
2956
2957                         int flag = scope;
2958
2959                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2960
2961                                 addr.s6_addr32[3] = ifa->ifa_local;
2962
2963                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2964                                         continue;
2965                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2966                                         if (idev->dev->flags&IFF_POINTOPOINT)
2967                                                 continue;
2968                                         flag |= IFA_HOST;
2969                                 }
2970
2971                                 add_addr(idev, &addr, plen, flag);
2972                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2973                                                       pflags);
2974                         }
2975                 }
2976         }
2977 }
2978 #endif
2979
2980 static void init_loopback(struct net_device *dev)
2981 {
2982         struct inet6_dev  *idev;
2983         struct net_device *sp_dev;
2984         struct inet6_ifaddr *sp_ifa;
2985         struct rt6_info *sp_rt;
2986
2987         /* ::1 */
2988
2989         ASSERT_RTNL();
2990
2991         idev = ipv6_find_idev(dev);
2992         if (!idev) {
2993                 pr_debug("%s: add_dev failed\n", __func__);
2994                 return;
2995         }
2996
2997         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2998
2999         /* Add routes to other interface's IPv6 addresses */
3000         for_each_netdev(dev_net(dev), sp_dev) {
3001                 if (!strcmp(sp_dev->name, dev->name))
3002                         continue;
3003
3004                 idev = __in6_dev_get(sp_dev);
3005                 if (!idev)
3006                         continue;
3007
3008                 read_lock_bh(&idev->lock);
3009                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
3010
3011                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
3012                                 continue;
3013
3014                         if (sp_ifa->rt) {
3015                                 /* This dst has been added to garbage list when
3016                                  * lo device down, release this obsolete dst and
3017                                  * reallocate a new router for ifa.
3018                                  */
3019                                 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3020                                         ip6_rt_put(sp_ifa->rt);
3021                                         sp_ifa->rt = NULL;
3022                                 } else {
3023                                         continue;
3024                                 }
3025                         }
3026
3027                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3028
3029                         /* Failure cases are ignored */
3030                         if (!IS_ERR(sp_rt)) {
3031                                 sp_ifa->rt = sp_rt;
3032                                 ip6_ins_rt(sp_rt);
3033                         }
3034                 }
3035                 read_unlock_bh(&idev->lock);
3036         }
3037 }
3038
3039 void addrconf_add_linklocal(struct inet6_dev *idev,
3040                             const struct in6_addr *addr, u32 flags)
3041 {
3042         struct inet6_ifaddr *ifp;
3043         u32 addr_flags = flags | IFA_F_PERMANENT;
3044
3045 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3046         if (idev->cnf.optimistic_dad &&
3047             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3048                 addr_flags |= IFA_F_OPTIMISTIC;
3049 #endif
3050
3051         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3052                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3053         if (!IS_ERR(ifp)) {
3054                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3055                 addrconf_dad_start(ifp);
3056                 in6_ifa_put(ifp);
3057         }
3058 }
3059 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3060
3061 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3062 {
3063         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3064                 return true;
3065
3066         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3067             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3068                 return true;
3069
3070         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3071             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3072                 return true;
3073
3074         return false;
3075 }
3076
3077 static int ipv6_generate_stable_address(struct in6_addr *address,
3078                                         u8 dad_count,
3079                                         const struct inet6_dev *idev)
3080 {
3081         static DEFINE_SPINLOCK(lock);
3082         static __u32 digest[SHA_DIGEST_WORDS];
3083         static __u32 workspace[SHA_WORKSPACE_WORDS];
3084
3085         static union {
3086                 char __data[SHA_MESSAGE_BYTES];
3087                 struct {
3088                         struct in6_addr secret;
3089                         __be32 prefix[2];
3090                         unsigned char hwaddr[MAX_ADDR_LEN];
3091                         u8 dad_count;
3092                 } __packed;
3093         } data;
3094
3095         struct in6_addr secret;
3096         struct in6_addr temp;
3097         struct net *net = dev_net(idev->dev);
3098
3099         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3100
3101         if (idev->cnf.stable_secret.initialized)
3102                 secret = idev->cnf.stable_secret.secret;
3103         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3104                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3105         else
3106                 return -1;
3107
3108 retry:
3109         spin_lock_bh(&lock);
3110
3111         sha_init(digest);
3112         memset(&data, 0, sizeof(data));
3113         memset(workspace, 0, sizeof(workspace));
3114         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3115         data.prefix[0] = address->s6_addr32[0];
3116         data.prefix[1] = address->s6_addr32[1];
3117         data.secret = secret;
3118         data.dad_count = dad_count;
3119
3120         sha_transform(digest, data.__data, workspace);
3121
3122         temp = *address;
3123         temp.s6_addr32[2] = (__force __be32)digest[0];
3124         temp.s6_addr32[3] = (__force __be32)digest[1];
3125
3126         spin_unlock_bh(&lock);
3127
3128         if (ipv6_reserved_interfaceid(temp)) {
3129                 dad_count++;
3130                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3131                         return -1;
3132                 goto retry;
3133         }
3134
3135         *address = temp;
3136         return 0;
3137 }
3138
3139 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3140 {
3141         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3142
3143         if (s->initialized)
3144                 return;
3145         s = &idev->cnf.stable_secret;
3146         get_random_bytes(&s->secret, sizeof(s->secret));
3147         s->initialized = true;
3148 }
3149
3150 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3151 {
3152         struct in6_addr addr;
3153
3154         /* no link local addresses on L3 master devices */
3155         if (netif_is_l3_master(idev->dev))
3156                 return;
3157
3158         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3159
3160         switch (idev->cnf.addr_gen_mode) {
3161         case IN6_ADDR_GEN_MODE_RANDOM:
3162                 ipv6_gen_mode_random_init(idev);
3163                 /* fallthrough */
3164         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3165                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3166                         addrconf_add_linklocal(idev, &addr,
3167                                                IFA_F_STABLE_PRIVACY);
3168                 else if (prefix_route)
3169                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3170                 break;
3171         case IN6_ADDR_GEN_MODE_EUI64:
3172                 /* addrconf_add_linklocal also adds a prefix_route and we
3173                  * only need to care about prefix routes if ipv6_generate_eui64
3174                  * couldn't generate one.
3175                  */
3176                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3177                         addrconf_add_linklocal(idev, &addr, 0);
3178                 else if (prefix_route)
3179                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3180                 break;
3181         case IN6_ADDR_GEN_MODE_NONE:
3182         default:
3183                 /* will not add any link local address */
3184                 break;
3185         }
3186 }
3187
3188 static void addrconf_dev_config(struct net_device *dev)
3189 {
3190         struct inet6_dev *idev;
3191
3192         ASSERT_RTNL();
3193
3194         if ((dev->type != ARPHRD_ETHER) &&
3195             (dev->type != ARPHRD_FDDI) &&
3196             (dev->type != ARPHRD_ARCNET) &&
3197             (dev->type != ARPHRD_INFINIBAND) &&
3198             (dev->type != ARPHRD_IEEE1394) &&
3199             (dev->type != ARPHRD_TUNNEL6) &&
3200             (dev->type != ARPHRD_6LOWPAN) &&
3201             (dev->type != ARPHRD_IP6GRE) &&
3202             (dev->type != ARPHRD_IPGRE) &&
3203             (dev->type != ARPHRD_TUNNEL) &&
3204             (dev->type != ARPHRD_NONE)) {
3205                 /* Alas, we support only Ethernet autoconfiguration. */
3206                 return;
3207         }
3208
3209         idev = addrconf_add_dev(dev);
3210         if (IS_ERR(idev))
3211                 return;
3212
3213         /* this device type has no EUI support */
3214         if (dev->type == ARPHRD_NONE &&
3215             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3216                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3217
3218         addrconf_addr_gen(idev, false);
3219 }
3220
3221 #if IS_ENABLED(CONFIG_IPV6_SIT)
3222 static void addrconf_sit_config(struct net_device *dev)
3223 {
3224         struct inet6_dev *idev;
3225
3226         ASSERT_RTNL();
3227
3228         /*
3229          * Configure the tunnel with one of our IPv4
3230          * addresses... we should configure all of
3231          * our v4 addrs in the tunnel
3232          */
3233
3234         idev = ipv6_find_idev(dev);
3235         if (!idev) {
3236                 pr_debug("%s: add_dev failed\n", __func__);
3237                 return;
3238         }
3239
3240         if (dev->priv_flags & IFF_ISATAP) {
3241                 addrconf_addr_gen(idev, false);
3242                 return;
3243         }
3244
3245         sit_add_v4_addrs(idev);
3246
3247         if (dev->flags&IFF_POINTOPOINT)
3248                 addrconf_add_mroute(dev);
3249 }
3250 #endif
3251
3252 #if IS_ENABLED(CONFIG_NET_IPGRE)
3253 static void addrconf_gre_config(struct net_device *dev)
3254 {
3255         struct inet6_dev *idev;
3256
3257         ASSERT_RTNL();
3258
3259         idev = ipv6_find_idev(dev);
3260         if (!idev) {
3261                 pr_debug("%s: add_dev failed\n", __func__);
3262                 return;
3263         }
3264
3265         addrconf_addr_gen(idev, true);
3266         if (dev->flags & IFF_POINTOPOINT)
3267                 addrconf_add_mroute(dev);
3268 }
3269 #endif
3270
3271 static int fixup_permanent_addr(struct inet6_dev *idev,
3272                                 struct inet6_ifaddr *ifp)
3273 {
3274         if (!ifp->rt) {
3275                 struct rt6_info *rt;
3276
3277                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3278                 if (unlikely(IS_ERR(rt)))
3279                         return PTR_ERR(rt);
3280
3281                 ifp->rt = rt;
3282         }
3283
3284         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3285                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3286                                       idev->dev, 0, 0);
3287         }
3288
3289         addrconf_dad_start(ifp);
3290
3291         return 0;
3292 }
3293
3294 static void addrconf_permanent_addr(struct net_device *dev)
3295 {
3296         struct inet6_ifaddr *ifp, *tmp;
3297         struct inet6_dev *idev;
3298
3299         idev = __in6_dev_get(dev);
3300         if (!idev)
3301                 return;
3302
3303         write_lock_bh(&idev->lock);
3304
3305         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3306                 if ((ifp->flags & IFA_F_PERMANENT) &&
3307                     fixup_permanent_addr(idev, ifp) < 0) {
3308                         write_unlock_bh(&idev->lock);
3309                         ipv6_del_addr(ifp);
3310                         write_lock_bh(&idev->lock);
3311
3312                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3313                                              idev->dev->name, &ifp->addr);
3314                 }
3315         }
3316
3317         write_unlock_bh(&idev->lock);
3318 }
3319
3320 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3321                            void *ptr)
3322 {
3323         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3324         struct netdev_notifier_changeupper_info *info;
3325         struct inet6_dev *idev = __in6_dev_get(dev);
3326         int run_pending = 0;
3327         int err;
3328
3329         switch (event) {
3330         case NETDEV_REGISTER:
3331                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3332                         idev = ipv6_add_dev(dev);
3333                         if (IS_ERR(idev))
3334                                 return notifier_from_errno(PTR_ERR(idev));
3335                 }
3336                 break;
3337
3338         case NETDEV_CHANGEMTU:
3339                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3340                 if (dev->mtu < IPV6_MIN_MTU) {
3341                         addrconf_ifdown(dev, 1);
3342                         break;
3343                 }
3344
3345                 if (idev) {
3346                         rt6_mtu_change(dev, dev->mtu);
3347                         idev->cnf.mtu6 = dev->mtu;
3348                         break;
3349                 }
3350
3351                 /* allocate new idev */
3352                 idev = ipv6_add_dev(dev);
3353                 if (IS_ERR(idev))
3354                         break;
3355
3356                 /* device is still not ready */
3357                 if (!(idev->if_flags & IF_READY))
3358                         break;
3359
3360                 run_pending = 1;
3361
3362                 /* fall through */
3363
3364         case NETDEV_UP:
3365         case NETDEV_CHANGE:
3366                 if (dev->flags & IFF_SLAVE)
3367                         break;
3368
3369                 if (idev && idev->cnf.disable_ipv6)
3370                         break;
3371
3372                 if (event == NETDEV_UP) {
3373                         /* restore routes for permanent addresses */
3374                         addrconf_permanent_addr(dev);
3375
3376                         if (!addrconf_qdisc_ok(dev)) {
3377                                 /* device is not ready yet. */
3378                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3379                                         dev->name);
3380                                 break;
3381                         }
3382
3383                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3384                                 idev = ipv6_add_dev(dev);
3385
3386                         if (!IS_ERR_OR_NULL(idev)) {
3387                                 idev->if_flags |= IF_READY;
3388                                 run_pending = 1;
3389                         }
3390                 } else if (event == NETDEV_CHANGE) {
3391                         if (!addrconf_qdisc_ok(dev)) {
3392                                 /* device is still not ready. */
3393                                 break;
3394                         }
3395
3396                         if (idev) {
3397                                 if (idev->if_flags & IF_READY) {
3398                                         /* device is already configured -
3399                                          * but resend MLD reports, we might
3400                                          * have roamed and need to update
3401                                          * multicast snooping switches
3402                                          */
3403                                         ipv6_mc_up(idev);
3404                                         break;
3405                                 }
3406                                 idev->if_flags |= IF_READY;
3407                         }
3408
3409                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3410                                 dev->name);
3411
3412                         run_pending = 1;
3413                 }
3414
3415                 switch (dev->type) {
3416 #if IS_ENABLED(CONFIG_IPV6_SIT)
3417                 case ARPHRD_SIT:
3418                         addrconf_sit_config(dev);
3419                         break;
3420 #endif
3421 #if IS_ENABLED(CONFIG_NET_IPGRE)
3422                 case ARPHRD_IPGRE:
3423                         addrconf_gre_config(dev);
3424                         break;
3425 #endif
3426                 case ARPHRD_LOOPBACK:
3427                         init_loopback(dev);
3428                         break;
3429
3430                 default:
3431                         addrconf_dev_config(dev);
3432                         break;
3433                 }
3434
3435                 if (!IS_ERR_OR_NULL(idev)) {
3436                         if (run_pending)
3437                                 addrconf_dad_run(idev);
3438
3439                         /*
3440                          * If the MTU changed during the interface down,
3441                          * when the interface up, the changed MTU must be
3442                          * reflected in the idev as well as routers.
3443                          */
3444                         if (idev->cnf.mtu6 != dev->mtu &&
3445                             dev->mtu >= IPV6_MIN_MTU) {
3446                                 rt6_mtu_change(dev, dev->mtu);
3447                                 idev->cnf.mtu6 = dev->mtu;
3448                         }
3449                         idev->tstamp = jiffies;
3450                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3451
3452                         /*
3453                          * If the changed mtu during down is lower than
3454                          * IPV6_MIN_MTU stop IPv6 on this interface.
3455                          */
3456                         if (dev->mtu < IPV6_MIN_MTU)
3457                                 addrconf_ifdown(dev, 1);
3458                 }
3459                 break;
3460
3461         case NETDEV_DOWN:
3462         case NETDEV_UNREGISTER:
3463                 /*
3464                  *      Remove all addresses from this interface.
3465                  */
3466                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3467                 break;
3468
3469         case NETDEV_CHANGENAME:
3470                 if (idev) {
3471                         snmp6_unregister_dev(idev);
3472                         addrconf_sysctl_unregister(idev);
3473                         err = addrconf_sysctl_register(idev);
3474                         if (err)
3475                                 return notifier_from_errno(err);
3476                         err = snmp6_register_dev(idev);
3477                         if (err) {
3478                                 addrconf_sysctl_unregister(idev);
3479                                 return notifier_from_errno(err);
3480                         }
3481                 }
3482                 break;
3483
3484         case NETDEV_PRE_TYPE_CHANGE:
3485         case NETDEV_POST_TYPE_CHANGE:
3486                 if (idev)
3487                         addrconf_type_change(dev, event);
3488                 break;
3489
3490         case NETDEV_CHANGEUPPER:
3491                 info = ptr;
3492
3493                 /* flush all routes if dev is linked to or unlinked from
3494                  * an L3 master device (e.g., VRF)
3495                  */
3496                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3497                         addrconf_ifdown(dev, 0);
3498         }
3499
3500         return NOTIFY_OK;
3501 }
3502
3503 /*
3504  *      addrconf module should be notified of a device going up
3505  */
3506 static struct notifier_block ipv6_dev_notf = {
3507         .notifier_call = addrconf_notify,
3508 };
3509
3510 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3511 {
3512         struct inet6_dev *idev;
3513         ASSERT_RTNL();
3514
3515         idev = __in6_dev_get(dev);
3516
3517         if (event == NETDEV_POST_TYPE_CHANGE)
3518                 ipv6_mc_remap(idev);
3519         else if (event == NETDEV_PRE_TYPE_CHANGE)
3520                 ipv6_mc_unmap(idev);
3521 }
3522
3523 static bool addr_is_local(const struct in6_addr *addr)
3524 {
3525         return ipv6_addr_type(addr) &
3526                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3527 }
3528
3529 static int addrconf_ifdown(struct net_device *dev, int how)
3530 {
3531         struct net *net = dev_net(dev);
3532         struct inet6_dev *idev;
3533         struct inet6_ifaddr *ifa, *tmp;
3534         struct list_head del_list;
3535         int _keep_addr;
3536         bool keep_addr;
3537         int state, i;
3538
3539         ASSERT_RTNL();
3540
3541         rt6_ifdown(net, dev);
3542         neigh_ifdown(&nd_tbl, dev);
3543
3544         idev = __in6_dev_get(dev);
3545         if (!idev)
3546                 return -ENODEV;
3547
3548         /*
3549          * Step 1: remove reference to ipv6 device from parent device.
3550          *         Do not dev_put!
3551          */
3552         if (how) {
3553                 idev->dead = 1;
3554
3555                 /* protected by rtnl_lock */
3556                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3557
3558                 /* Step 1.5: remove snmp6 entry */
3559                 snmp6_unregister_dev(idev);
3560
3561         }
3562
3563         /* aggregate the system setting and interface setting */
3564         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3565         if (!_keep_addr)
3566                 _keep_addr = idev->cnf.keep_addr_on_down;
3567
3568         /* combine the user config with event to determine if permanent
3569          * addresses are to be removed from address hash table
3570          */
3571         keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3572
3573         /* Step 2: clear hash table */
3574         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3575                 struct hlist_head *h = &inet6_addr_lst[i];
3576
3577                 spin_lock_bh(&addrconf_hash_lock);
3578 restart:
3579                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3580                         if (ifa->idev == idev) {
3581                                 addrconf_del_dad_work(ifa);
3582                                 /* combined flag + permanent flag decide if
3583                                  * address is retained on a down event
3584                                  */
3585                                 if (!keep_addr ||
3586                                     !(ifa->flags & IFA_F_PERMANENT) ||
3587                                     addr_is_local(&ifa->addr)) {
3588                                         hlist_del_init_rcu(&ifa->addr_lst);
3589                                         goto restart;
3590                                 }
3591                         }
3592                 }
3593                 spin_unlock_bh(&addrconf_hash_lock);
3594         }
3595
3596         write_lock_bh(&idev->lock);
3597
3598         addrconf_del_rs_timer(idev);
3599
3600         /* Step 2: clear flags for stateless addrconf */
3601         if (!how)
3602                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3603
3604         /* Step 3: clear tempaddr list */
3605         while (!list_empty(&idev->tempaddr_list)) {
3606                 ifa = list_first_entry(&idev->tempaddr_list,
3607                                        struct inet6_ifaddr, tmp_list);
3608                 list_del(&ifa->tmp_list);
3609                 write_unlock_bh(&idev->lock);
3610                 spin_lock_bh(&ifa->lock);
3611
3612                 if (ifa->ifpub) {
3613                         in6_ifa_put(ifa->ifpub);
3614                         ifa->ifpub = NULL;
3615                 }
3616                 spin_unlock_bh(&ifa->lock);
3617                 in6_ifa_put(ifa);
3618                 write_lock_bh(&idev->lock);
3619         }
3620
3621         /* re-combine the user config with event to determine if permanent
3622          * addresses are to be removed from the interface list
3623          */
3624         keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3625
3626         INIT_LIST_HEAD(&del_list);
3627         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3628                 struct rt6_info *rt = NULL;
3629
3630                 addrconf_del_dad_work(ifa);
3631
3632                 write_unlock_bh(&idev->lock);
3633                 spin_lock_bh(&ifa->lock);
3634
3635                 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3636                     !addr_is_local(&ifa->addr)) {
3637                         /* set state to skip the notifier below */
3638                         state = INET6_IFADDR_STATE_DEAD;
3639                         ifa->state = 0;
3640                         if (!(ifa->flags & IFA_F_NODAD))
3641                                 ifa->flags |= IFA_F_TENTATIVE;
3642
3643                         rt = ifa->rt;
3644                         ifa->rt = NULL;
3645                 } else {
3646                         state = ifa->state;
3647                         ifa->state = INET6_IFADDR_STATE_DEAD;
3648
3649                         list_move(&ifa->if_list, &del_list);
3650                 }
3651
3652                 spin_unlock_bh(&ifa->lock);
3653
3654                 if (rt)
3655                         ip6_del_rt(rt);
3656
3657                 if (state != INET6_IFADDR_STATE_DEAD) {
3658                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3659                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3660                 } else {
3661                         if (idev->cnf.forwarding)
3662                                 addrconf_leave_anycast(ifa);
3663                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3664                 }
3665
3666                 write_lock_bh(&idev->lock);
3667         }
3668
3669         write_unlock_bh(&idev->lock);
3670
3671         /* now clean up addresses to be removed */
3672         while (!list_empty(&del_list)) {
3673                 ifa = list_first_entry(&del_list,
3674                                        struct inet6_ifaddr, if_list);
3675                 list_del(&ifa->if_list);
3676
3677                 in6_ifa_put(ifa);
3678         }
3679
3680         /* Step 5: Discard anycast and multicast list */
3681         if (how) {
3682                 ipv6_ac_destroy_dev(idev);
3683                 ipv6_mc_destroy_dev(idev);
3684         } else {
3685                 ipv6_mc_down(idev);
3686         }
3687
3688         idev->tstamp = jiffies;
3689
3690         /* Last: Shot the device (if unregistered) */
3691         if (how) {
3692                 addrconf_sysctl_unregister(idev);
3693                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3694                 neigh_ifdown(&nd_tbl, dev);
3695                 in6_dev_put(idev);
3696         }
3697         return 0;
3698 }
3699
3700 static void addrconf_rs_timer(unsigned long data)
3701 {
3702         struct inet6_dev *idev = (struct inet6_dev *)data;
3703         struct net_device *dev = idev->dev;
3704         struct in6_addr lladdr;
3705
3706         write_lock(&idev->lock);
3707         if (idev->dead || !(idev->if_flags & IF_READY))
3708                 goto out;
3709
3710         if (!ipv6_accept_ra(idev))
3711                 goto out;
3712
3713         /* Announcement received after solicitation was sent */
3714         if (idev->if_flags & IF_RA_RCVD)
3715                 goto out;
3716
3717         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3718                 write_unlock(&idev->lock);
3719                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3720                         ndisc_send_rs(dev, &lladdr,
3721                                       &in6addr_linklocal_allrouters);
3722                 else
3723                         goto put;
3724
3725                 write_lock(&idev->lock);
3726                 idev->rs_interval = rfc3315_s14_backoff_update(
3727                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3728                 /* The wait after the last probe can be shorter */
3729                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3730                                              idev->cnf.rtr_solicits) ?
3731                                       idev->cnf.rtr_solicit_delay :
3732                                       idev->rs_interval);
3733         } else {
3734                 /*
3735                  * Note: we do not support deprecated "all on-link"
3736                  * assumption any longer.
3737                  */
3738                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3739         }
3740
3741 out:
3742         write_unlock(&idev->lock);
3743 put:
3744         in6_dev_put(idev);
3745 }
3746
3747 /*
3748  *      Duplicate Address Detection
3749  */
3750 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3751 {
3752         unsigned long rand_num;
3753         struct inet6_dev *idev = ifp->idev;
3754         u64 nonce;
3755
3756         if (ifp->flags & IFA_F_OPTIMISTIC)
3757                 rand_num = 0;
3758         else
3759                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3760
3761         nonce = 0;
3762         if (idev->cnf.enhanced_dad ||
3763             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3764                 do
3765                         get_random_bytes(&nonce, 6);
3766                 while (nonce == 0);
3767         }
3768         ifp->dad_nonce = nonce;
3769         ifp->dad_probes = idev->cnf.dad_transmits;
3770         addrconf_mod_dad_work(ifp, rand_num);
3771 }
3772
3773 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3774 {
3775         struct inet6_dev *idev = ifp->idev;
3776         struct net_device *dev = idev->dev;
3777         bool bump_id, notify = false;
3778
3779         addrconf_join_solict(dev, &ifp->addr);
3780
3781         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3782
3783         read_lock_bh(&idev->lock);
3784         spin_lock(&ifp->lock);
3785         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3786                 goto out;
3787
3788         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3789             idev->cnf.accept_dad < 1 ||
3790             !(ifp->flags&IFA_F_TENTATIVE) ||
3791             ifp->flags & IFA_F_NODAD) {
3792                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3793                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3794                 spin_unlock(&ifp->lock);
3795                 read_unlock_bh(&idev->lock);
3796
3797                 addrconf_dad_completed(ifp, bump_id);
3798                 return;
3799         }
3800
3801         if (!(idev->if_flags & IF_READY)) {
3802                 spin_unlock(&ifp->lock);
3803                 read_unlock_bh(&idev->lock);
3804                 /*
3805                  * If the device is not ready:
3806                  * - keep it tentative if it is a permanent address.
3807                  * - otherwise, kill it.
3808                  */
3809                 in6_ifa_hold(ifp);
3810                 addrconf_dad_stop(ifp, 0);
3811                 return;
3812         }
3813
3814         /*
3815          * Optimistic nodes can start receiving
3816          * Frames right away
3817          */
3818         if (ifp->flags & IFA_F_OPTIMISTIC) {
3819                 ip6_ins_rt(ifp->rt);
3820                 if (ipv6_use_optimistic_addr(idev)) {
3821                         /* Because optimistic nodes can use this address,
3822                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3823                          */
3824                         notify = true;
3825                 }
3826         }
3827
3828         addrconf_dad_kick(ifp);
3829 out:
3830         spin_unlock(&ifp->lock);
3831         read_unlock_bh(&idev->lock);
3832         if (notify)
3833                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3834 }
3835
3836 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3837 {
3838         bool begin_dad = false;
3839
3840         spin_lock_bh(&ifp->lock);
3841         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3842                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3843                 begin_dad = true;
3844         }
3845         spin_unlock_bh(&ifp->lock);
3846
3847         if (begin_dad)
3848                 addrconf_mod_dad_work(ifp, 0);
3849 }
3850
3851 static void addrconf_dad_work(struct work_struct *w)
3852 {
3853         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3854                                                 struct inet6_ifaddr,
3855                                                 dad_work);
3856         struct inet6_dev *idev = ifp->idev;
3857         bool bump_id, disable_ipv6 = false;
3858         struct in6_addr mcaddr;
3859
3860         enum {
3861                 DAD_PROCESS,
3862                 DAD_BEGIN,
3863                 DAD_ABORT,
3864         } action = DAD_PROCESS;
3865
3866         rtnl_lock();
3867
3868         spin_lock_bh(&ifp->lock);
3869         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3870                 action = DAD_BEGIN;
3871                 ifp->state = INET6_IFADDR_STATE_DAD;
3872         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3873                 action = DAD_ABORT;
3874                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3875
3876                 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3877                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3878                         struct in6_addr addr;
3879
3880                         addr.s6_addr32[0] = htonl(0xfe800000);
3881                         addr.s6_addr32[1] = 0;
3882
3883                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3884                             ipv6_addr_equal(&ifp->addr, &addr)) {
3885                                 /* DAD failed for link-local based on MAC */
3886                                 idev->cnf.disable_ipv6 = 1;
3887
3888                                 pr_info("%s: IPv6 being disabled!\n",
3889                                         ifp->idev->dev->name);
3890                                 disable_ipv6 = true;
3891                         }
3892                 }
3893         }
3894         spin_unlock_bh(&ifp->lock);
3895
3896         if (action == DAD_BEGIN) {
3897                 addrconf_dad_begin(ifp);
3898                 goto out;
3899         } else if (action == DAD_ABORT) {
3900                 in6_ifa_hold(ifp);
3901                 addrconf_dad_stop(ifp, 1);
3902                 if (disable_ipv6)
3903                         addrconf_ifdown(idev->dev, 0);
3904                 goto out;
3905         }
3906
3907         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3908                 goto out;
3909
3910         write_lock_bh(&idev->lock);
3911         if (idev->dead || !(idev->if_flags & IF_READY)) {
3912                 write_unlock_bh(&idev->lock);
3913                 goto out;
3914         }
3915
3916         spin_lock(&ifp->lock);
3917         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3918                 spin_unlock(&ifp->lock);
3919                 write_unlock_bh(&idev->lock);
3920                 goto out;
3921         }
3922
3923         if (ifp->dad_probes == 0) {
3924                 /*
3925                  * DAD was successful
3926                  */
3927
3928                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3929                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3930                 spin_unlock(&ifp->lock);
3931                 write_unlock_bh(&idev->lock);
3932
3933                 addrconf_dad_completed(ifp, bump_id);
3934
3935                 goto out;
3936         }
3937
3938         ifp->dad_probes--;
3939         addrconf_mod_dad_work(ifp,
3940                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3941         spin_unlock(&ifp->lock);
3942         write_unlock_bh(&idev->lock);
3943
3944         /* send a neighbour solicitation for our addr */
3945         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3946         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
3947                       ifp->dad_nonce);
3948 out:
3949         in6_ifa_put(ifp);
3950         rtnl_unlock();
3951 }
3952
3953 /* ifp->idev must be at least read locked */
3954 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3955 {
3956         struct inet6_ifaddr *ifpiter;
3957         struct inet6_dev *idev = ifp->idev;
3958
3959         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3960                 if (ifpiter->scope > IFA_LINK)
3961                         break;
3962                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3963                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3964                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3965                     IFA_F_PERMANENT)
3966                         return false;
3967         }
3968         return true;
3969 }
3970
3971 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
3972 {
3973         struct net_device *dev = ifp->idev->dev;
3974         struct in6_addr lladdr;
3975         bool send_rs, send_mld;
3976
3977         addrconf_del_dad_work(ifp);
3978
3979         /*
3980          *      Configure the address for reception. Now it is valid.
3981          */
3982
3983         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3984
3985         /* If added prefix is link local and we are prepared to process
3986            router advertisements, start sending router solicitations.
3987          */
3988
3989         read_lock_bh(&ifp->idev->lock);
3990         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3991         send_rs = send_mld &&
3992                   ipv6_accept_ra(ifp->idev) &&
3993                   ifp->idev->cnf.rtr_solicits != 0 &&
3994                   (dev->flags&IFF_LOOPBACK) == 0;
3995         read_unlock_bh(&ifp->idev->lock);
3996
3997         /* While dad is in progress mld report's source address is in6_addrany.
3998          * Resend with proper ll now.
3999          */
4000         if (send_mld)
4001                 ipv6_mc_dad_complete(ifp->idev);
4002
4003         if (send_rs) {
4004                 /*
4005                  *      If a host as already performed a random delay
4006                  *      [...] as part of DAD [...] there is no need
4007                  *      to delay again before sending the first RS
4008                  */
4009                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4010                         return;
4011                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4012
4013                 write_lock_bh(&ifp->idev->lock);
4014                 spin_lock(&ifp->lock);
4015                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4016                         ifp->idev->cnf.rtr_solicit_interval);
4017                 ifp->idev->rs_probes = 1;
4018                 ifp->idev->if_flags |= IF_RS_SENT;
4019                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4020                 spin_unlock(&ifp->lock);
4021                 write_unlock_bh(&ifp->idev->lock);
4022         }
4023
4024         if (bump_id)
4025                 rt_genid_bump_ipv6(dev_net(dev));
4026
4027         /* Make sure that a new temporary address will be created
4028          * before this temporary address becomes deprecated.
4029          */
4030         if (ifp->flags & IFA_F_TEMPORARY)
4031                 addrconf_verify_rtnl();
4032 }
4033
4034 static void addrconf_dad_run(struct inet6_dev *idev)
4035 {
4036         struct inet6_ifaddr *ifp;
4037
4038         read_lock_bh(&idev->lock);
4039         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4040                 spin_lock(&ifp->lock);
4041                 if (ifp->flags & IFA_F_TENTATIVE &&
4042                     ifp->state == INET6_IFADDR_STATE_DAD)
4043                         addrconf_dad_kick(ifp);
4044                 spin_unlock(&ifp->lock);
4045         }
4046         read_unlock_bh(&idev->lock);
4047 }
4048
4049 #ifdef CONFIG_PROC_FS
4050 struct if6_iter_state {
4051         struct seq_net_private p;
4052         int bucket;
4053         int offset;
4054 };
4055
4056 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4057 {
4058         struct inet6_ifaddr *ifa = NULL;
4059         struct if6_iter_state *state = seq->private;
4060         struct net *net = seq_file_net(seq);
4061         int p = 0;
4062
4063         /* initial bucket if pos is 0 */
4064         if (pos == 0) {
4065                 state->bucket = 0;
4066                 state->offset = 0;
4067         }
4068
4069         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4070                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4071                                          addr_lst) {
4072                         if (!net_eq(dev_net(ifa->idev->dev), net))
4073                                 continue;
4074                         /* sync with offset */
4075                         if (p < state->offset) {
4076                                 p++;
4077                                 continue;
4078                         }
4079                         state->offset++;
4080                         return ifa;
4081                 }
4082
4083                 /* prepare for next bucket */
4084                 state->offset = 0;
4085                 p = 0;
4086         }
4087         return NULL;
4088 }
4089
4090 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4091                                          struct inet6_ifaddr *ifa)
4092 {
4093         struct if6_iter_state *state = seq->private;
4094         struct net *net = seq_file_net(seq);
4095
4096         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4097                 if (!net_eq(dev_net(ifa->idev->dev), net))
4098                         continue;
4099                 state->offset++;
4100                 return ifa;
4101         }
4102
4103         while (++state->bucket < IN6_ADDR_HSIZE) {
4104                 state->offset = 0;
4105                 hlist_for_each_entry_rcu_bh(ifa,
4106                                      &inet6_addr_lst[state->bucket], addr_lst) {
4107                         if (!net_eq(dev_net(ifa->idev->dev), net))
4108                                 continue;
4109                         state->offset++;
4110                         return ifa;
4111                 }
4112         }
4113
4114         return NULL;
4115 }
4116
4117 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4118         __acquires(rcu_bh)
4119 {
4120         rcu_read_lock_bh();
4121         return if6_get_first(seq, *pos);
4122 }
4123
4124 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4125 {
4126         struct inet6_ifaddr *ifa;
4127
4128         ifa = if6_get_next(seq, v);
4129         ++*pos;
4130         return ifa;
4131 }
4132
4133 static void if6_seq_stop(struct seq_file *seq, void *v)
4134         __releases(rcu_bh)
4135 {
4136         rcu_read_unlock_bh();
4137 }
4138
4139 static int if6_seq_show(struct seq_file *seq, void *v)
4140 {
4141         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4142         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4143                    &ifp->addr,
4144                    ifp->idev->dev->ifindex,
4145                    ifp->prefix_len,
4146                    ifp->scope,
4147                    (u8) ifp->flags,
4148                    ifp->idev->dev->name);
4149         return 0;
4150 }
4151
4152 static const struct seq_operations if6_seq_ops = {
4153         .start  = if6_seq_start,
4154         .next   = if6_seq_next,
4155         .show   = if6_seq_show,
4156         .stop   = if6_seq_stop,
4157 };
4158
4159 static int if6_seq_open(struct inode *inode, struct file *file)
4160 {
4161         return seq_open_net(inode, file, &if6_seq_ops,
4162                             sizeof(struct if6_iter_state));
4163 }
4164
4165 static const struct file_operations if6_fops = {
4166         .owner          = THIS_MODULE,
4167         .open           = if6_seq_open,
4168         .read           = seq_read,
4169         .llseek         = seq_lseek,
4170         .release        = seq_release_net,
4171 };
4172
4173 static int __net_init if6_proc_net_init(struct net *net)
4174 {
4175         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4176                 return -ENOMEM;
4177         return 0;
4178 }
4179
4180 static void __net_exit if6_proc_net_exit(struct net *net)
4181 {
4182         remove_proc_entry("if_inet6", net->proc_net);
4183 }
4184
4185 static struct pernet_operations if6_proc_net_ops = {
4186         .init = if6_proc_net_init,
4187         .exit = if6_proc_net_exit,
4188 };
4189
4190 int __init if6_proc_init(void)
4191 {
4192         return register_pernet_subsys(&if6_proc_net_ops);
4193 }
4194
4195 void if6_proc_exit(void)
4196 {
4197         unregister_pernet_subsys(&if6_proc_net_ops);
4198 }
4199 #endif  /* CONFIG_PROC_FS */
4200
4201 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4202 /* Check if address is a home address configured on any interface. */
4203 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4204 {
4205         int ret = 0;
4206         struct inet6_ifaddr *ifp = NULL;
4207         unsigned int hash = inet6_addr_hash(addr);
4208
4209         rcu_read_lock_bh();
4210         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4211                 if (!net_eq(dev_net(ifp->idev->dev), net))
4212                         continue;
4213                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4214                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4215                         ret = 1;
4216                         break;
4217                 }
4218         }
4219         rcu_read_unlock_bh();
4220         return ret;
4221 }
4222 #endif
4223
4224 /*
4225  *      Periodic address status verification
4226  */
4227
4228 static void addrconf_verify_rtnl(void)
4229 {
4230         unsigned long now, next, next_sec, next_sched;
4231         struct inet6_ifaddr *ifp;
4232         int i;
4233
4234         ASSERT_RTNL();
4235
4236         rcu_read_lock_bh();
4237         now = jiffies;
4238         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4239
4240         cancel_delayed_work(&addr_chk_work);
4241
4242         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4243 restart:
4244                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4245                         unsigned long age;
4246
4247                         /* When setting preferred_lft to a value not zero or
4248                          * infinity, while valid_lft is infinity
4249                          * IFA_F_PERMANENT has a non-infinity life time.
4250                          */
4251                         if ((ifp->flags & IFA_F_PERMANENT) &&
4252                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4253                                 continue;
4254
4255                         spin_lock(&ifp->lock);
4256                         /* We try to batch several events at once. */
4257                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4258
4259                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4260                             age >= ifp->valid_lft) {
4261                                 spin_unlock(&ifp->lock);
4262                                 in6_ifa_hold(ifp);
4263                                 ipv6_del_addr(ifp);
4264                                 goto restart;
4265                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4266                                 spin_unlock(&ifp->lock);
4267                                 continue;
4268                         } else if (age >= ifp->prefered_lft) {
4269                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4270                                 int deprecate = 0;
4271
4272                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4273                                         deprecate = 1;
4274                                         ifp->flags |= IFA_F_DEPRECATED;
4275                                 }
4276
4277                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4278                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4279                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4280
4281                                 spin_unlock(&ifp->lock);
4282
4283                                 if (deprecate) {
4284                                         in6_ifa_hold(ifp);
4285
4286                                         ipv6_ifa_notify(0, ifp);
4287                                         in6_ifa_put(ifp);
4288                                         goto restart;
4289                                 }
4290                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4291                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4292                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4293                                         ifp->idev->cnf.dad_transmits *
4294                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4295
4296                                 if (age >= ifp->prefered_lft - regen_advance) {
4297                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4298                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4299                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4300                                         if (!ifp->regen_count && ifpub) {
4301                                                 ifp->regen_count++;
4302                                                 in6_ifa_hold(ifp);
4303                                                 in6_ifa_hold(ifpub);
4304                                                 spin_unlock(&ifp->lock);
4305
4306                                                 spin_lock(&ifpub->lock);
4307                                                 ifpub->regen_count = 0;
4308                                                 spin_unlock(&ifpub->lock);
4309                                                 ipv6_create_tempaddr(ifpub, ifp);
4310                                                 in6_ifa_put(ifpub);
4311                                                 in6_ifa_put(ifp);
4312                                                 goto restart;
4313                                         }
4314                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4315                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4316                                 spin_unlock(&ifp->lock);
4317                         } else {
4318                                 /* ifp->prefered_lft <= ifp->valid_lft */
4319                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4320                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4321                                 spin_unlock(&ifp->lock);
4322                         }
4323                 }
4324         }
4325
4326         next_sec = round_jiffies_up(next);
4327         next_sched = next;
4328
4329         /* If rounded timeout is accurate enough, accept it. */
4330         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4331                 next_sched = next_sec;
4332
4333         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4334         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4335                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4336
4337         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4338               now, next, next_sec, next_sched);
4339         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4340         rcu_read_unlock_bh();
4341 }
4342
4343 static void addrconf_verify_work(struct work_struct *w)
4344 {
4345         rtnl_lock();
4346         addrconf_verify_rtnl();
4347         rtnl_unlock();
4348 }
4349
4350 static void addrconf_verify(void)
4351 {
4352         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4353 }
4354
4355 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4356                                      struct in6_addr **peer_pfx)
4357 {
4358         struct in6_addr *pfx = NULL;
4359
4360         *peer_pfx = NULL;
4361
4362         if (addr)
4363                 pfx = nla_data(addr);
4364
4365         if (local) {
4366                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4367                         *peer_pfx = pfx;
4368                 pfx = nla_data(local);
4369         }
4370
4371         return pfx;
4372 }
4373
4374 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4375         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4376         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4377         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4378         [IFA_FLAGS]             = { .len = sizeof(u32) },
4379 };
4380
4381 static int
4382 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4383 {
4384         struct net *net = sock_net(skb->sk);
4385         struct ifaddrmsg *ifm;
4386         struct nlattr *tb[IFA_MAX+1];
4387         struct in6_addr *pfx, *peer_pfx;
4388         u32 ifa_flags;
4389         int err;
4390
4391         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4392         if (err < 0)
4393                 return err;
4394
4395         ifm = nlmsg_data(nlh);
4396         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4397         if (!pfx)
4398                 return -EINVAL;
4399
4400         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4401
4402         /* We ignore other flags so far. */
4403         ifa_flags &= IFA_F_MANAGETEMPADDR;
4404
4405         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4406                               ifm->ifa_prefixlen);
4407 }
4408
4409 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4410                              u32 prefered_lft, u32 valid_lft)
4411 {
4412         u32 flags;
4413         clock_t expires;
4414         unsigned long timeout;
4415         bool was_managetempaddr;
4416         bool had_prefixroute;
4417
4418         ASSERT_RTNL();
4419
4420         if (!valid_lft || (prefered_lft > valid_lft))
4421                 return -EINVAL;
4422
4423         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4424             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4425                 return -EINVAL;
4426
4427         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4428         if (addrconf_finite_timeout(timeout)) {
4429                 expires = jiffies_to_clock_t(timeout * HZ);
4430                 valid_lft = timeout;
4431                 flags = RTF_EXPIRES;
4432         } else {
4433                 expires = 0;
4434                 flags = 0;
4435                 ifa_flags |= IFA_F_PERMANENT;
4436         }
4437
4438         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4439         if (addrconf_finite_timeout(timeout)) {
4440                 if (timeout == 0)
4441                         ifa_flags |= IFA_F_DEPRECATED;
4442                 prefered_lft = timeout;
4443         }
4444
4445         spin_lock_bh(&ifp->lock);
4446         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4447         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4448                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4449         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4450                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4451                         IFA_F_NOPREFIXROUTE);
4452         ifp->flags |= ifa_flags;
4453         ifp->tstamp = jiffies;
4454         ifp->valid_lft = valid_lft;
4455         ifp->prefered_lft = prefered_lft;
4456
4457         spin_unlock_bh(&ifp->lock);
4458         if (!(ifp->flags&IFA_F_TENTATIVE))
4459                 ipv6_ifa_notify(0, ifp);
4460
4461         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4462                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4463                                       expires, flags);
4464         } else if (had_prefixroute) {
4465                 enum cleanup_prefix_rt_t action;
4466                 unsigned long rt_expires;
4467
4468                 write_lock_bh(&ifp->idev->lock);
4469                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4470                 write_unlock_bh(&ifp->idev->lock);
4471
4472                 if (action != CLEANUP_PREFIX_RT_NOP) {
4473                         cleanup_prefix_route(ifp, rt_expires,
4474                                 action == CLEANUP_PREFIX_RT_DEL);
4475                 }
4476         }
4477
4478         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4479                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4480                         valid_lft = prefered_lft = 0;
4481                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4482                                  !was_managetempaddr, jiffies);
4483         }
4484
4485         addrconf_verify_rtnl();
4486
4487         return 0;
4488 }
4489
4490 static int
4491 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4492 {
4493         struct net *net = sock_net(skb->sk);
4494         struct ifaddrmsg *ifm;
4495         struct nlattr *tb[IFA_MAX+1];
4496         struct in6_addr *pfx, *peer_pfx;
4497         struct inet6_ifaddr *ifa;
4498         struct net_device *dev;
4499         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4500         u32 ifa_flags;
4501         int err;
4502
4503         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4504         if (err < 0)
4505                 return err;
4506
4507         ifm = nlmsg_data(nlh);
4508         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4509         if (!pfx)
4510                 return -EINVAL;
4511
4512         if (tb[IFA_CACHEINFO]) {
4513                 struct ifa_cacheinfo *ci;
4514
4515                 ci = nla_data(tb[IFA_CACHEINFO]);
4516                 valid_lft = ci->ifa_valid;
4517                 preferred_lft = ci->ifa_prefered;
4518         } else {
4519                 preferred_lft = INFINITY_LIFE_TIME;
4520                 valid_lft = INFINITY_LIFE_TIME;
4521         }
4522
4523         dev =  __dev_get_by_index(net, ifm->ifa_index);
4524         if (!dev)
4525                 return -ENODEV;
4526
4527         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4528
4529         /* We ignore other flags so far. */
4530         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4531                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4532
4533         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4534         if (!ifa) {
4535                 /*
4536                  * It would be best to check for !NLM_F_CREATE here but
4537                  * userspace already relies on not having to provide this.
4538                  */
4539                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4540                                       ifm->ifa_prefixlen, ifa_flags,
4541                                       preferred_lft, valid_lft);
4542         }
4543
4544         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4545             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4546                 err = -EEXIST;
4547         else
4548                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4549
4550         in6_ifa_put(ifa);
4551
4552         return err;
4553 }
4554
4555 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4556                           u8 scope, int ifindex)
4557 {
4558         struct ifaddrmsg *ifm;
4559
4560         ifm = nlmsg_data(nlh);
4561         ifm->ifa_family = AF_INET6;
4562         ifm->ifa_prefixlen = prefixlen;
4563         ifm->ifa_flags = flags;
4564         ifm->ifa_scope = scope;
4565         ifm->ifa_index = ifindex;
4566 }
4567
4568 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4569                          unsigned long tstamp, u32 preferred, u32 valid)
4570 {
4571         struct ifa_cacheinfo ci;
4572
4573         ci.cstamp = cstamp_delta(cstamp);
4574         ci.tstamp = cstamp_delta(tstamp);
4575         ci.ifa_prefered = preferred;
4576         ci.ifa_valid = valid;
4577
4578         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4579 }
4580
4581 static inline int rt_scope(int ifa_scope)
4582 {
4583         if (ifa_scope & IFA_HOST)
4584                 return RT_SCOPE_HOST;
4585         else if (ifa_scope & IFA_LINK)
4586                 return RT_SCOPE_LINK;
4587         else if (ifa_scope & IFA_SITE)
4588                 return RT_SCOPE_SITE;
4589         else
4590                 return RT_SCOPE_UNIVERSE;
4591 }
4592
4593 static inline int inet6_ifaddr_msgsize(void)
4594 {
4595         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4596                + nla_total_size(16) /* IFA_LOCAL */
4597                + nla_total_size(16) /* IFA_ADDRESS */
4598                + nla_total_size(sizeof(struct ifa_cacheinfo))
4599                + nla_total_size(4)  /* IFA_FLAGS */;
4600 }
4601
4602 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4603                              u32 portid, u32 seq, int event, unsigned int flags)
4604 {
4605         struct nlmsghdr  *nlh;
4606         u32 preferred, valid;
4607
4608         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4609         if (!nlh)
4610                 return -EMSGSIZE;
4611
4612         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4613                       ifa->idev->dev->ifindex);
4614
4615         if (!((ifa->flags&IFA_F_PERMANENT) &&
4616               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4617                 preferred = ifa->prefered_lft;
4618                 valid = ifa->valid_lft;
4619                 if (preferred != INFINITY_LIFE_TIME) {
4620                         long tval = (jiffies - ifa->tstamp)/HZ;
4621                         if (preferred > tval)
4622                                 preferred -= tval;
4623                         else
4624                                 preferred = 0;
4625                         if (valid != INFINITY_LIFE_TIME) {
4626                                 if (valid > tval)
4627                                         valid -= tval;
4628                                 else
4629                                         valid = 0;
4630                         }
4631                 }
4632         } else {
4633                 preferred = INFINITY_LIFE_TIME;
4634                 valid = INFINITY_LIFE_TIME;
4635         }
4636
4637         if (!ipv6_addr_any(&ifa->peer_addr)) {
4638                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4639                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4640                         goto error;
4641         } else
4642                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4643                         goto error;
4644
4645         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4646                 goto error;
4647
4648         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4649                 goto error;
4650
4651         nlmsg_end(skb, nlh);
4652         return 0;
4653
4654 error:
4655         nlmsg_cancel(skb, nlh);
4656         return -EMSGSIZE;
4657 }
4658
4659 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4660                                 u32 portid, u32 seq, int event, u16 flags)
4661 {
4662         struct nlmsghdr  *nlh;
4663         u8 scope = RT_SCOPE_UNIVERSE;
4664         int ifindex = ifmca->idev->dev->ifindex;
4665
4666         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4667                 scope = RT_SCOPE_SITE;
4668
4669         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4670         if (!nlh)
4671                 return -EMSGSIZE;
4672
4673         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4674         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4675             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4676                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4677                 nlmsg_cancel(skb, nlh);
4678                 return -EMSGSIZE;
4679         }
4680
4681         nlmsg_end(skb, nlh);
4682         return 0;
4683 }
4684
4685 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4686                                 u32 portid, u32 seq, int event, unsigned int flags)
4687 {
4688         struct nlmsghdr  *nlh;
4689         u8 scope = RT_SCOPE_UNIVERSE;
4690         int ifindex = ifaca->aca_idev->dev->ifindex;
4691
4692         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4693                 scope = RT_SCOPE_SITE;
4694
4695         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4696         if (!nlh)
4697                 return -EMSGSIZE;
4698
4699         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4700         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4701             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4702                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4703                 nlmsg_cancel(skb, nlh);
4704                 return -EMSGSIZE;
4705         }
4706
4707         nlmsg_end(skb, nlh);
4708         return 0;
4709 }
4710
4711 enum addr_type_t {
4712         UNICAST_ADDR,
4713         MULTICAST_ADDR,
4714         ANYCAST_ADDR,
4715 };
4716
4717 /* called with rcu_read_lock() */
4718 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4719                           struct netlink_callback *cb, enum addr_type_t type,
4720                           int s_ip_idx, int *p_ip_idx)
4721 {
4722         struct ifmcaddr6 *ifmca;
4723         struct ifacaddr6 *ifaca;
4724         int err = 1;
4725         int ip_idx = *p_ip_idx;
4726
4727         read_lock_bh(&idev->lock);
4728         switch (type) {
4729         case UNICAST_ADDR: {
4730                 struct inet6_ifaddr *ifa;
4731
4732                 /* unicast address incl. temp addr */
4733                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4734                         if (++ip_idx < s_ip_idx)
4735                                 continue;
4736                         err = inet6_fill_ifaddr(skb, ifa,
4737                                                 NETLINK_CB(cb->skb).portid,
4738                                                 cb->nlh->nlmsg_seq,
4739                                                 RTM_NEWADDR,
4740                                                 NLM_F_MULTI);
4741                         if (err < 0)
4742                                 break;
4743                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4744                 }
4745                 break;
4746         }
4747         case MULTICAST_ADDR:
4748                 /* multicast address */
4749                 for (ifmca = idev->mc_list; ifmca;
4750                      ifmca = ifmca->next, ip_idx++) {
4751                         if (ip_idx < s_ip_idx)
4752                                 continue;
4753                         err = inet6_fill_ifmcaddr(skb, ifmca,
4754                                                   NETLINK_CB(cb->skb).portid,
4755                                                   cb->nlh->nlmsg_seq,
4756                                                   RTM_GETMULTICAST,
4757                                                   NLM_F_MULTI);
4758                         if (err < 0)
4759                                 break;
4760                 }
4761                 break;
4762         case ANYCAST_ADDR:
4763                 /* anycast address */
4764                 for (ifaca = idev->ac_list; ifaca;
4765                      ifaca = ifaca->aca_next, ip_idx++) {
4766                         if (ip_idx < s_ip_idx)
4767                                 continue;
4768                         err = inet6_fill_ifacaddr(skb, ifaca,
4769                                                   NETLINK_CB(cb->skb).portid,
4770                                                   cb->nlh->nlmsg_seq,
4771                                                   RTM_GETANYCAST,
4772                                                   NLM_F_MULTI);
4773                         if (err < 0)
4774                                 break;
4775                 }
4776                 break;
4777         default:
4778                 break;
4779         }
4780         read_unlock_bh(&idev->lock);
4781         *p_ip_idx = ip_idx;
4782         return err;
4783 }
4784
4785 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4786                            enum addr_type_t type)
4787 {
4788         struct net *net = sock_net(skb->sk);
4789         int h, s_h;
4790         int idx, ip_idx;
4791         int s_idx, s_ip_idx;
4792         struct net_device *dev;
4793         struct inet6_dev *idev;
4794         struct hlist_head *head;
4795
4796         s_h = cb->args[0];
4797         s_idx = idx = cb->args[1];
4798         s_ip_idx = ip_idx = cb->args[2];
4799
4800         rcu_read_lock();
4801         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4802         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4803                 idx = 0;
4804                 head = &net->dev_index_head[h];
4805                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4806                         if (idx < s_idx)
4807                                 goto cont;
4808                         if (h > s_h || idx > s_idx)
4809                                 s_ip_idx = 0;
4810                         ip_idx = 0;
4811                         idev = __in6_dev_get(dev);
4812                         if (!idev)
4813                                 goto cont;
4814
4815                         if (in6_dump_addrs(idev, skb, cb, type,
4816                                            s_ip_idx, &ip_idx) < 0)
4817                                 goto done;
4818 cont:
4819                         idx++;
4820                 }
4821         }
4822 done:
4823         rcu_read_unlock();
4824         cb->args[0] = h;
4825         cb->args[1] = idx;
4826         cb->args[2] = ip_idx;
4827
4828         return skb->len;
4829 }
4830
4831 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4832 {
4833         enum addr_type_t type = UNICAST_ADDR;
4834
4835         return inet6_dump_addr(skb, cb, type);
4836 }
4837
4838 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4839 {
4840         enum addr_type_t type = MULTICAST_ADDR;
4841
4842         return inet6_dump_addr(skb, cb, type);
4843 }
4844
4845
4846 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4847 {
4848         enum addr_type_t type = ANYCAST_ADDR;
4849
4850         return inet6_dump_addr(skb, cb, type);
4851 }
4852
4853 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4854 {
4855         struct net *net = sock_net(in_skb->sk);
4856         struct ifaddrmsg *ifm;
4857         struct nlattr *tb[IFA_MAX+1];
4858         struct in6_addr *addr = NULL, *peer;
4859         struct net_device *dev = NULL;
4860         struct inet6_ifaddr *ifa;
4861         struct sk_buff *skb;
4862         int err;
4863
4864         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4865         if (err < 0)
4866                 goto errout;
4867
4868         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4869         if (!addr) {
4870                 err = -EINVAL;
4871                 goto errout;
4872         }
4873
4874         ifm = nlmsg_data(nlh);
4875         if (ifm->ifa_index)
4876                 dev = __dev_get_by_index(net, ifm->ifa_index);
4877
4878         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4879         if (!ifa) {
4880                 err = -EADDRNOTAVAIL;
4881                 goto errout;
4882         }
4883
4884         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4885         if (!skb) {
4886                 err = -ENOBUFS;
4887                 goto errout_ifa;
4888         }
4889
4890         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4891                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4892         if (err < 0) {
4893                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4894                 WARN_ON(err == -EMSGSIZE);
4895                 kfree_skb(skb);
4896                 goto errout_ifa;
4897         }
4898         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4899 errout_ifa:
4900         in6_ifa_put(ifa);
4901 errout:
4902         return err;
4903 }
4904
4905 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4906 {
4907         struct sk_buff *skb;
4908         struct net *net = dev_net(ifa->idev->dev);
4909         int err = -ENOBUFS;
4910
4911         /* Don't send DELADDR notification for TENTATIVE address,
4912          * since NEWADDR notification is sent only after removing
4913          * TENTATIVE flag.
4914          */
4915         if (ifa->flags & IFA_F_TENTATIVE && event == RTM_DELADDR)
4916                 return;
4917
4918         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4919         if (!skb)
4920                 goto errout;
4921
4922         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4923         if (err < 0) {
4924                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4925                 WARN_ON(err == -EMSGSIZE);
4926                 kfree_skb(skb);
4927                 goto errout;
4928         }
4929         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4930         return;
4931 errout:
4932         if (err < 0)
4933                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4934 }
4935
4936 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4937                                 __s32 *array, int bytes)
4938 {
4939         BUG_ON(bytes < (DEVCONF_MAX * 4));
4940
4941         memset(array, 0, bytes);
4942         array[DEVCONF_FORWARDING] = cnf->forwarding;
4943         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4944         array[DEVCONF_MTU6] = cnf->mtu6;
4945         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4946         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4947         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4948         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4949         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4950         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4951                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4952         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4953                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4954         array[DEVCONF_RTR_SOLICIT_DELAY] =
4955                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4956         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4957         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4958                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4959         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4960                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4961         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4962         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4963         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4964         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4965         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4966         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4967         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4968         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4969         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4970 #ifdef CONFIG_IPV6_ROUTER_PREF
4971         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4972         array[DEVCONF_RTR_PROBE_INTERVAL] =
4973                 jiffies_to_msecs(cnf->rtr_probe_interval);
4974 #ifdef CONFIG_IPV6_ROUTE_INFO
4975         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4976 #endif
4977 #endif
4978         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4979         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4980 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4981         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4982         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4983 #endif
4984 #ifdef CONFIG_IPV6_MROUTE
4985         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4986 #endif
4987         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4988         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4989         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4990         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4991         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4992         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4993         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4994         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4995         /* we omit DEVCONF_STABLE_SECRET for now */
4996         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4997         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4998         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4999         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5000         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5001 #ifdef CONFIG_IPV6_SEG6_HMAC
5002         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5003 #endif
5004         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5005         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5006 }
5007
5008 static inline size_t inet6_ifla6_size(void)
5009 {
5010         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5011              + nla_total_size(sizeof(struct ifla_cacheinfo))
5012              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5013              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5014              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5015              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5016 }
5017
5018 static inline size_t inet6_if_nlmsg_size(void)
5019 {
5020         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5021                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5022                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5023                + nla_total_size(4) /* IFLA_MTU */
5024                + nla_total_size(4) /* IFLA_LINK */
5025                + nla_total_size(1) /* IFLA_OPERSTATE */
5026                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5027 }
5028
5029 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5030                                         int bytes)
5031 {
5032         int i;
5033         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5034         BUG_ON(pad < 0);
5035
5036         /* Use put_unaligned() because stats may not be aligned for u64. */
5037         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5038         for (i = 1; i < ICMP6_MIB_MAX; i++)
5039                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5040
5041         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5042 }
5043
5044 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5045                                         int bytes, size_t syncpoff)
5046 {
5047         int i, c;
5048         u64 buff[IPSTATS_MIB_MAX];
5049         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5050
5051         BUG_ON(pad < 0);
5052
5053         memset(buff, 0, sizeof(buff));
5054         buff[0] = IPSTATS_MIB_MAX;
5055
5056         for_each_possible_cpu(c) {
5057                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5058                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5059         }
5060
5061         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5062         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5063 }
5064
5065 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5066                              int bytes)
5067 {
5068         switch (attrtype) {
5069         case IFLA_INET6_STATS:
5070                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5071                                      offsetof(struct ipstats_mib, syncp));
5072                 break;
5073         case IFLA_INET6_ICMP6STATS:
5074                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5075                 break;
5076         }
5077 }
5078
5079 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5080                                   u32 ext_filter_mask)
5081 {
5082         struct nlattr *nla;
5083         struct ifla_cacheinfo ci;
5084
5085         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5086                 goto nla_put_failure;
5087         ci.max_reasm_len = IPV6_MAXPLEN;
5088         ci.tstamp = cstamp_delta(idev->tstamp);
5089         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5090         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5091         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5092                 goto nla_put_failure;
5093         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5094         if (!nla)
5095                 goto nla_put_failure;
5096         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5097
5098         /* XXX - MC not implemented */
5099
5100         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5101                 return 0;
5102
5103         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5104         if (!nla)
5105                 goto nla_put_failure;
5106         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5107
5108         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5109         if (!nla)
5110                 goto nla_put_failure;
5111         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5112
5113         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5114         if (!nla)
5115                 goto nla_put_failure;
5116
5117         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5118                 goto nla_put_failure;
5119
5120         read_lock_bh(&idev->lock);
5121         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5122         read_unlock_bh(&idev->lock);
5123
5124         return 0;
5125
5126 nla_put_failure:
5127         return -EMSGSIZE;
5128 }
5129
5130 static size_t inet6_get_link_af_size(const struct net_device *dev,
5131                                      u32 ext_filter_mask)
5132 {
5133         if (!__in6_dev_get(dev))
5134                 return 0;
5135
5136         return inet6_ifla6_size();
5137 }
5138
5139 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5140                               u32 ext_filter_mask)
5141 {
5142         struct inet6_dev *idev = __in6_dev_get(dev);
5143
5144         if (!idev)
5145                 return -ENODATA;
5146
5147         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5148                 return -EMSGSIZE;
5149
5150         return 0;
5151 }
5152
5153 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5154 {
5155         struct inet6_ifaddr *ifp;
5156         struct net_device *dev = idev->dev;
5157         bool clear_token, update_rs = false;
5158         struct in6_addr ll_addr;
5159
5160         ASSERT_RTNL();
5161
5162         if (!token)
5163                 return -EINVAL;
5164         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5165                 return -EINVAL;
5166         if (!ipv6_accept_ra(idev))
5167                 return -EINVAL;
5168         if (idev->cnf.rtr_solicits == 0)
5169                 return -EINVAL;
5170
5171         write_lock_bh(&idev->lock);
5172
5173         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5174         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5175
5176         write_unlock_bh(&idev->lock);
5177
5178         clear_token = ipv6_addr_any(token);
5179         if (clear_token)
5180                 goto update_lft;
5181
5182         if (!idev->dead && (idev->if_flags & IF_READY) &&
5183             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5184                              IFA_F_OPTIMISTIC)) {
5185                 /* If we're not ready, then normal ifup will take care
5186                  * of this. Otherwise, we need to request our rs here.
5187                  */
5188                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5189                 update_rs = true;
5190         }
5191
5192 update_lft:
5193         write_lock_bh(&idev->lock);
5194
5195         if (update_rs) {
5196                 idev->if_flags |= IF_RS_SENT;
5197                 idev->rs_interval = rfc3315_s14_backoff_init(
5198                         idev->cnf.rtr_solicit_interval);
5199                 idev->rs_probes = 1;
5200                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5201         }
5202
5203         /* Well, that's kinda nasty ... */
5204         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5205                 spin_lock(&ifp->lock);
5206                 if (ifp->tokenized) {
5207                         ifp->valid_lft = 0;
5208                         ifp->prefered_lft = 0;
5209                 }
5210                 spin_unlock(&ifp->lock);
5211         }
5212
5213         write_unlock_bh(&idev->lock);
5214         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5215         addrconf_verify_rtnl();
5216         return 0;
5217 }
5218
5219 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5220         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5221         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5222 };
5223
5224 static int inet6_validate_link_af(const struct net_device *dev,
5225                                   const struct nlattr *nla)
5226 {
5227         struct nlattr *tb[IFLA_INET6_MAX + 1];
5228
5229         if (dev && !__in6_dev_get(dev))
5230                 return -EAFNOSUPPORT;
5231
5232         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5233 }
5234
5235 static int check_addr_gen_mode(int mode)
5236 {
5237         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5238             mode != IN6_ADDR_GEN_MODE_NONE &&
5239             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5240             mode != IN6_ADDR_GEN_MODE_RANDOM)
5241                 return -EINVAL;
5242         return 1;
5243 }
5244
5245 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5246                                 int mode)
5247 {
5248         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5249             !idev->cnf.stable_secret.initialized &&
5250             !net->ipv6.devconf_dflt->stable_secret.initialized)
5251                 return -EINVAL;
5252         return 1;
5253 }
5254
5255 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5256 {
5257         int err = -EINVAL;
5258         struct inet6_dev *idev = __in6_dev_get(dev);
5259         struct nlattr *tb[IFLA_INET6_MAX + 1];
5260
5261         if (!idev)
5262                 return -EAFNOSUPPORT;
5263
5264         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5265                 BUG();
5266
5267         if (tb[IFLA_INET6_TOKEN]) {
5268                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5269                 if (err)
5270                         return err;
5271         }
5272
5273         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5274                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5275
5276                 if (check_addr_gen_mode(mode) < 0 ||
5277                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5278                         return -EINVAL;
5279
5280                 idev->cnf.addr_gen_mode = mode;
5281                 err = 0;
5282         }
5283
5284         return err;
5285 }
5286
5287 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5288                              u32 portid, u32 seq, int event, unsigned int flags)
5289 {
5290         struct net_device *dev = idev->dev;
5291         struct ifinfomsg *hdr;
5292         struct nlmsghdr *nlh;
5293         void *protoinfo;
5294
5295         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5296         if (!nlh)
5297                 return -EMSGSIZE;
5298
5299         hdr = nlmsg_data(nlh);
5300         hdr->ifi_family = AF_INET6;
5301         hdr->__ifi_pad = 0;
5302         hdr->ifi_type = dev->type;
5303         hdr->ifi_index = dev->ifindex;
5304         hdr->ifi_flags = dev_get_flags(dev);
5305         hdr->ifi_change = 0;
5306
5307         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5308             (dev->addr_len &&
5309              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5310             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5311             (dev->ifindex != dev_get_iflink(dev) &&
5312              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5313             nla_put_u8(skb, IFLA_OPERSTATE,
5314                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5315                 goto nla_put_failure;
5316         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5317         if (!protoinfo)
5318                 goto nla_put_failure;
5319
5320         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5321                 goto nla_put_failure;
5322
5323         nla_nest_end(skb, protoinfo);
5324         nlmsg_end(skb, nlh);
5325         return 0;
5326
5327 nla_put_failure:
5328         nlmsg_cancel(skb, nlh);
5329         return -EMSGSIZE;
5330 }
5331
5332 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5333 {
5334         struct net *net = sock_net(skb->sk);
5335         int h, s_h;
5336         int idx = 0, s_idx;
5337         struct net_device *dev;
5338         struct inet6_dev *idev;
5339         struct hlist_head *head;
5340
5341         s_h = cb->args[0];
5342         s_idx = cb->args[1];
5343
5344         rcu_read_lock();
5345         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5346                 idx = 0;
5347                 head = &net->dev_index_head[h];
5348                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5349                         if (idx < s_idx)
5350                                 goto cont;
5351                         idev = __in6_dev_get(dev);
5352                         if (!idev)
5353                                 goto cont;
5354                         if (inet6_fill_ifinfo(skb, idev,
5355                                               NETLINK_CB(cb->skb).portid,
5356                                               cb->nlh->nlmsg_seq,
5357                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5358                                 goto out;
5359 cont:
5360                         idx++;
5361                 }
5362         }
5363 out:
5364         rcu_read_unlock();
5365         cb->args[1] = idx;
5366         cb->args[0] = h;
5367
5368         return skb->len;
5369 }
5370
5371 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5372 {
5373         struct sk_buff *skb;
5374         struct net *net = dev_net(idev->dev);
5375         int err = -ENOBUFS;
5376
5377         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5378         if (!skb)
5379                 goto errout;
5380
5381         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5382         if (err < 0) {
5383                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5384                 WARN_ON(err == -EMSGSIZE);
5385                 kfree_skb(skb);
5386                 goto errout;
5387         }
5388         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5389         return;
5390 errout:
5391         if (err < 0)
5392                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5393 }
5394
5395 static inline size_t inet6_prefix_nlmsg_size(void)
5396 {
5397         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5398                + nla_total_size(sizeof(struct in6_addr))
5399                + nla_total_size(sizeof(struct prefix_cacheinfo));
5400 }
5401
5402 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5403                              struct prefix_info *pinfo, u32 portid, u32 seq,
5404                              int event, unsigned int flags)
5405 {
5406         struct prefixmsg *pmsg;
5407         struct nlmsghdr *nlh;
5408         struct prefix_cacheinfo ci;
5409
5410         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5411         if (!nlh)
5412                 return -EMSGSIZE;
5413
5414         pmsg = nlmsg_data(nlh);
5415         pmsg->prefix_family = AF_INET6;
5416         pmsg->prefix_pad1 = 0;
5417         pmsg->prefix_pad2 = 0;
5418         pmsg->prefix_ifindex = idev->dev->ifindex;
5419         pmsg->prefix_len = pinfo->prefix_len;
5420         pmsg->prefix_type = pinfo->type;
5421         pmsg->prefix_pad3 = 0;
5422         pmsg->prefix_flags = 0;
5423         if (pinfo->onlink)
5424                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5425         if (pinfo->autoconf)
5426                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5427
5428         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5429                 goto nla_put_failure;
5430         ci.preferred_time = ntohl(pinfo->prefered);
5431         ci.valid_time = ntohl(pinfo->valid);
5432         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5433                 goto nla_put_failure;
5434         nlmsg_end(skb, nlh);
5435         return 0;
5436
5437 nla_put_failure:
5438         nlmsg_cancel(skb, nlh);
5439         return -EMSGSIZE;
5440 }
5441
5442 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5443                          struct prefix_info *pinfo)
5444 {
5445         struct sk_buff *skb;
5446         struct net *net = dev_net(idev->dev);
5447         int err = -ENOBUFS;
5448
5449         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5450         if (!skb)
5451                 goto errout;
5452
5453         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5454         if (err < 0) {
5455                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5456                 WARN_ON(err == -EMSGSIZE);
5457                 kfree_skb(skb);
5458                 goto errout;
5459         }
5460         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5461         return;
5462 errout:
5463         if (err < 0)
5464                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5465 }
5466
5467 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5468 {
5469         struct net *net = dev_net(ifp->idev->dev);
5470
5471         if (event)
5472                 ASSERT_RTNL();
5473
5474         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5475
5476         switch (event) {
5477         case RTM_NEWADDR:
5478                 /*
5479                  * If the address was optimistic
5480                  * we inserted the route at the start of
5481                  * our DAD process, so we don't need
5482                  * to do it again
5483                  */
5484                 if (!(ifp->rt->rt6i_node))
5485                         ip6_ins_rt(ifp->rt);
5486                 if (ifp->idev->cnf.forwarding)
5487                         addrconf_join_anycast(ifp);
5488                 if (!ipv6_addr_any(&ifp->peer_addr))
5489                         addrconf_prefix_route(&ifp->peer_addr, 128,
5490                                               ifp->idev->dev, 0, 0);
5491                 break;
5492         case RTM_DELADDR:
5493                 if (ifp->idev->cnf.forwarding)
5494                         addrconf_leave_anycast(ifp);
5495                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5496                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5497                         struct rt6_info *rt;
5498
5499                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5500                                                        ifp->idev->dev, 0, 0);
5501                         if (rt)
5502                                 ip6_del_rt(rt);
5503                 }
5504                 if (ifp->rt) {
5505                         dst_hold(&ifp->rt->dst);
5506                         ip6_del_rt(ifp->rt);
5507                 }
5508                 rt_genid_bump_ipv6(net);
5509                 break;
5510         }
5511         atomic_inc(&net->ipv6.dev_addr_genid);
5512 }
5513
5514 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5515 {
5516         rcu_read_lock_bh();
5517         if (likely(ifp->idev->dead == 0))
5518                 __ipv6_ifa_notify(event, ifp);
5519         rcu_read_unlock_bh();
5520 }
5521
5522 #ifdef CONFIG_SYSCTL
5523
5524 static
5525 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5526                            void __user *buffer, size_t *lenp, loff_t *ppos)
5527 {
5528         int *valp = ctl->data;
5529         int val = *valp;
5530         loff_t pos = *ppos;
5531         struct ctl_table lctl;
5532         int ret;
5533
5534         /*
5535          * ctl->data points to idev->cnf.forwarding, we should
5536          * not modify it until we get the rtnl lock.
5537          */
5538         lctl = *ctl;
5539         lctl.data = &val;
5540
5541         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5542
5543         if (write)
5544                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5545         if (ret)
5546                 *ppos = pos;
5547         return ret;
5548 }
5549
5550 static
5551 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5552                         void __user *buffer, size_t *lenp, loff_t *ppos)
5553 {
5554         struct inet6_dev *idev = ctl->extra1;
5555         int min_mtu = IPV6_MIN_MTU;
5556         struct ctl_table lctl;
5557
5558         lctl = *ctl;
5559         lctl.extra1 = &min_mtu;
5560         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5561
5562         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5563 }
5564
5565 static void dev_disable_change(struct inet6_dev *idev)
5566 {
5567         struct netdev_notifier_info info;
5568
5569         if (!idev || !idev->dev)
5570                 return;
5571
5572         netdev_notifier_info_init(&info, idev->dev);
5573         if (idev->cnf.disable_ipv6)
5574                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5575         else
5576                 addrconf_notify(NULL, NETDEV_UP, &info);
5577 }
5578
5579 static void addrconf_disable_change(struct net *net, __s32 newf)
5580 {
5581         struct net_device *dev;
5582         struct inet6_dev *idev;
5583
5584         for_each_netdev(net, dev) {
5585                 idev = __in6_dev_get(dev);
5586                 if (idev) {
5587                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5588                         idev->cnf.disable_ipv6 = newf;
5589                         if (changed)
5590                                 dev_disable_change(idev);
5591                 }
5592         }
5593 }
5594
5595 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5596 {
5597         struct net *net;
5598         int old;
5599
5600         if (!rtnl_trylock())
5601                 return restart_syscall();
5602
5603         net = (struct net *)table->extra2;
5604         old = *p;
5605         *p = newf;
5606
5607         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5608                 rtnl_unlock();
5609                 return 0;
5610         }
5611
5612         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5613                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5614                 addrconf_disable_change(net, newf);
5615         } else if ((!newf) ^ (!old))
5616                 dev_disable_change((struct inet6_dev *)table->extra1);
5617
5618         rtnl_unlock();
5619         return 0;
5620 }
5621
5622 static
5623 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5624                             void __user *buffer, size_t *lenp, loff_t *ppos)
5625 {
5626         int *valp = ctl->data;
5627         int val = *valp;
5628         loff_t pos = *ppos;
5629         struct ctl_table lctl;
5630         int ret;
5631
5632         /*
5633          * ctl->data points to idev->cnf.disable_ipv6, we should
5634          * not modify it until we get the rtnl lock.
5635          */
5636         lctl = *ctl;
5637         lctl.data = &val;
5638
5639         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5640
5641         if (write)
5642                 ret = addrconf_disable_ipv6(ctl, valp, val);
5643         if (ret)
5644                 *ppos = pos;
5645         return ret;
5646 }
5647
5648 static
5649 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5650                               void __user *buffer, size_t *lenp, loff_t *ppos)
5651 {
5652         int *valp = ctl->data;
5653         int ret;
5654         int old, new;
5655
5656         old = *valp;
5657         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5658         new = *valp;
5659
5660         if (write && old != new) {
5661                 struct net *net = ctl->extra2;
5662
5663                 if (!rtnl_trylock())
5664                         return restart_syscall();
5665
5666                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5667                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5668                                                      NETCONFA_IFINDEX_DEFAULT,
5669                                                      net->ipv6.devconf_dflt);
5670                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5671                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5672                                                      NETCONFA_IFINDEX_ALL,
5673                                                      net->ipv6.devconf_all);
5674                 else {
5675                         struct inet6_dev *idev = ctl->extra1;
5676
5677                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5678                                                      idev->dev->ifindex,
5679                                                      &idev->cnf);
5680                 }
5681                 rtnl_unlock();
5682         }
5683
5684         return ret;
5685 }
5686
5687 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5688                                          void __user *buffer, size_t *lenp,
5689                                          loff_t *ppos)
5690 {
5691         int ret = 0;
5692         int new_val;
5693         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5694         struct net *net = (struct net *)ctl->extra2;
5695
5696         if (!rtnl_trylock())
5697                 return restart_syscall();
5698
5699         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5700
5701         if (write) {
5702                 new_val = *((int *)ctl->data);
5703
5704                 if (check_addr_gen_mode(new_val) < 0) {
5705                         ret = -EINVAL;
5706                         goto out;
5707                 }
5708
5709                 /* request for default */
5710                 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5711                         ipv6_devconf_dflt.addr_gen_mode = new_val;
5712
5713                 /* request for individual net device */
5714                 } else {
5715                         if (!idev)
5716                                 goto out;
5717
5718                         if (check_stable_privacy(idev, net, new_val) < 0) {
5719                                 ret = -EINVAL;
5720                                 goto out;
5721                         }
5722
5723                         if (idev->cnf.addr_gen_mode != new_val) {
5724                                 idev->cnf.addr_gen_mode = new_val;
5725                                 addrconf_dev_config(idev->dev);
5726                         }
5727                 }
5728         }
5729
5730 out:
5731         rtnl_unlock();
5732
5733         return ret;
5734 }
5735
5736 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5737                                          void __user *buffer, size_t *lenp,
5738                                          loff_t *ppos)
5739 {
5740         int err;
5741         struct in6_addr addr;
5742         char str[IPV6_MAX_STRLEN];
5743         struct ctl_table lctl = *ctl;
5744         struct net *net = ctl->extra2;
5745         struct ipv6_stable_secret *secret = ctl->data;
5746
5747         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5748                 return -EIO;
5749
5750         lctl.maxlen = IPV6_MAX_STRLEN;
5751         lctl.data = str;
5752
5753         if (!rtnl_trylock())
5754                 return restart_syscall();
5755
5756         if (!write && !secret->initialized) {
5757                 err = -EIO;
5758                 goto out;
5759         }
5760
5761         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5762         if (err >= sizeof(str)) {
5763                 err = -EIO;
5764                 goto out;
5765         }
5766
5767         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5768         if (err || !write)
5769                 goto out;
5770
5771         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5772                 err = -EIO;
5773                 goto out;
5774         }
5775
5776         secret->initialized = true;
5777         secret->secret = addr;
5778
5779         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5780                 struct net_device *dev;
5781
5782                 for_each_netdev(net, dev) {
5783                         struct inet6_dev *idev = __in6_dev_get(dev);
5784
5785                         if (idev) {
5786                                 idev->cnf.addr_gen_mode =
5787                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5788                         }
5789                 }
5790         } else {
5791                 struct inet6_dev *idev = ctl->extra1;
5792
5793                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5794         }
5795
5796 out:
5797         rtnl_unlock();
5798
5799         return err;
5800 }
5801
5802 static
5803 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5804                                                 int write,
5805                                                 void __user *buffer,
5806                                                 size_t *lenp,
5807                                                 loff_t *ppos)
5808 {
5809         int *valp = ctl->data;
5810         int val = *valp;
5811         loff_t pos = *ppos;
5812         struct ctl_table lctl;
5813         int ret;
5814
5815         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5816          * we should not modify it until we get the rtnl lock.
5817          */
5818         lctl = *ctl;
5819         lctl.data = &val;
5820
5821         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5822
5823         if (write)
5824                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5825         if (ret)
5826                 *ppos = pos;
5827         return ret;
5828 }
5829
5830 static int minus_one = -1;
5831 static const int one = 1;
5832 static const int two_five_five = 255;
5833
5834 static const struct ctl_table addrconf_sysctl[] = {
5835         {
5836                 .procname       = "forwarding",
5837                 .data           = &ipv6_devconf.forwarding,
5838                 .maxlen         = sizeof(int),
5839                 .mode           = 0644,
5840                 .proc_handler   = addrconf_sysctl_forward,
5841         },
5842         {
5843                 .procname       = "hop_limit",
5844                 .data           = &ipv6_devconf.hop_limit,
5845                 .maxlen         = sizeof(int),
5846                 .mode           = 0644,
5847                 .proc_handler   = proc_dointvec_minmax,
5848                 .extra1         = (void *)&one,
5849                 .extra2         = (void *)&two_five_five,
5850         },
5851         {
5852                 .procname       = "mtu",
5853                 .data           = &ipv6_devconf.mtu6,
5854                 .maxlen         = sizeof(int),
5855                 .mode           = 0644,
5856                 .proc_handler   = addrconf_sysctl_mtu,
5857         },
5858         {
5859                 .procname       = "accept_ra",
5860                 .data           = &ipv6_devconf.accept_ra,
5861                 .maxlen         = sizeof(int),
5862                 .mode           = 0644,
5863                 .proc_handler   = proc_dointvec,
5864         },
5865         {
5866                 .procname       = "accept_redirects",
5867                 .data           = &ipv6_devconf.accept_redirects,
5868                 .maxlen         = sizeof(int),
5869                 .mode           = 0644,
5870                 .proc_handler   = proc_dointvec,
5871         },
5872         {
5873                 .procname       = "autoconf",
5874                 .data           = &ipv6_devconf.autoconf,
5875                 .maxlen         = sizeof(int),
5876                 .mode           = 0644,
5877                 .proc_handler   = proc_dointvec,
5878         },
5879         {
5880                 .procname       = "dad_transmits",
5881                 .data           = &ipv6_devconf.dad_transmits,
5882                 .maxlen         = sizeof(int),
5883                 .mode           = 0644,
5884                 .proc_handler   = proc_dointvec,
5885         },
5886         {
5887                 .procname       = "router_solicitations",
5888                 .data           = &ipv6_devconf.rtr_solicits,
5889                 .maxlen         = sizeof(int),
5890                 .mode           = 0644,
5891                 .proc_handler   = proc_dointvec_minmax,
5892                 .extra1         = &minus_one,
5893         },
5894         {
5895                 .procname       = "router_solicitation_interval",
5896                 .data           = &ipv6_devconf.rtr_solicit_interval,
5897                 .maxlen         = sizeof(int),
5898                 .mode           = 0644,
5899                 .proc_handler   = proc_dointvec_jiffies,
5900         },
5901         {
5902                 .procname       = "router_solicitation_max_interval",
5903                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
5904                 .maxlen         = sizeof(int),
5905                 .mode           = 0644,
5906                 .proc_handler   = proc_dointvec_jiffies,
5907         },
5908         {
5909                 .procname       = "router_solicitation_delay",
5910                 .data           = &ipv6_devconf.rtr_solicit_delay,
5911                 .maxlen         = sizeof(int),
5912                 .mode           = 0644,
5913                 .proc_handler   = proc_dointvec_jiffies,
5914         },
5915         {
5916                 .procname       = "force_mld_version",
5917                 .data           = &ipv6_devconf.force_mld_version,
5918                 .maxlen         = sizeof(int),
5919                 .mode           = 0644,
5920                 .proc_handler   = proc_dointvec,
5921         },
5922         {
5923                 .procname       = "mldv1_unsolicited_report_interval",
5924                 .data           =
5925                         &ipv6_devconf.mldv1_unsolicited_report_interval,
5926                 .maxlen         = sizeof(int),
5927                 .mode           = 0644,
5928                 .proc_handler   = proc_dointvec_ms_jiffies,
5929         },
5930         {
5931                 .procname       = "mldv2_unsolicited_report_interval",
5932                 .data           =
5933                         &ipv6_devconf.mldv2_unsolicited_report_interval,
5934                 .maxlen         = sizeof(int),
5935                 .mode           = 0644,
5936                 .proc_handler   = proc_dointvec_ms_jiffies,
5937         },
5938         {
5939                 .procname       = "use_tempaddr",
5940                 .data           = &ipv6_devconf.use_tempaddr,
5941                 .maxlen         = sizeof(int),
5942                 .mode           = 0644,
5943                 .proc_handler   = proc_dointvec,
5944         },
5945         {
5946                 .procname       = "temp_valid_lft",
5947                 .data           = &ipv6_devconf.temp_valid_lft,
5948                 .maxlen         = sizeof(int),
5949                 .mode           = 0644,
5950                 .proc_handler   = proc_dointvec,
5951         },
5952         {
5953                 .procname       = "temp_prefered_lft",
5954                 .data           = &ipv6_devconf.temp_prefered_lft,
5955                 .maxlen         = sizeof(int),
5956                 .mode           = 0644,
5957                 .proc_handler   = proc_dointvec,
5958         },
5959         {
5960                 .procname       = "regen_max_retry",
5961                 .data           = &ipv6_devconf.regen_max_retry,
5962                 .maxlen         = sizeof(int),
5963                 .mode           = 0644,
5964                 .proc_handler   = proc_dointvec,
5965         },
5966         {
5967                 .procname       = "max_desync_factor",
5968                 .data           = &ipv6_devconf.max_desync_factor,
5969                 .maxlen         = sizeof(int),
5970                 .mode           = 0644,
5971                 .proc_handler   = proc_dointvec,
5972         },
5973         {
5974                 .procname       = "max_addresses",
5975                 .data           = &ipv6_devconf.max_addresses,
5976                 .maxlen         = sizeof(int),
5977                 .mode           = 0644,
5978                 .proc_handler   = proc_dointvec,
5979         },
5980         {
5981                 .procname       = "accept_ra_defrtr",
5982                 .data           = &ipv6_devconf.accept_ra_defrtr,
5983                 .maxlen         = sizeof(int),
5984                 .mode           = 0644,
5985                 .proc_handler   = proc_dointvec,
5986         },
5987         {
5988                 .procname       = "accept_ra_min_hop_limit",
5989                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5990                 .maxlen         = sizeof(int),
5991                 .mode           = 0644,
5992                 .proc_handler   = proc_dointvec,
5993         },
5994         {
5995                 .procname       = "accept_ra_pinfo",
5996                 .data           = &ipv6_devconf.accept_ra_pinfo,
5997                 .maxlen         = sizeof(int),
5998                 .mode           = 0644,
5999                 .proc_handler   = proc_dointvec,
6000         },
6001 #ifdef CONFIG_IPV6_ROUTER_PREF
6002         {
6003                 .procname       = "accept_ra_rtr_pref",
6004                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6005                 .maxlen         = sizeof(int),
6006                 .mode           = 0644,
6007                 .proc_handler   = proc_dointvec,
6008         },
6009         {
6010                 .procname       = "router_probe_interval",
6011                 .data           = &ipv6_devconf.rtr_probe_interval,
6012                 .maxlen         = sizeof(int),
6013                 .mode           = 0644,
6014                 .proc_handler   = proc_dointvec_jiffies,
6015         },
6016 #ifdef CONFIG_IPV6_ROUTE_INFO
6017         {
6018                 .procname       = "accept_ra_rt_info_max_plen",
6019                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6020                 .maxlen         = sizeof(int),
6021                 .mode           = 0644,
6022                 .proc_handler   = proc_dointvec,
6023         },
6024 #endif
6025 #endif
6026         {
6027                 .procname       = "proxy_ndp",
6028                 .data           = &ipv6_devconf.proxy_ndp,
6029                 .maxlen         = sizeof(int),
6030                 .mode           = 0644,
6031                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6032         },
6033         {
6034                 .procname       = "accept_source_route",
6035                 .data           = &ipv6_devconf.accept_source_route,
6036                 .maxlen         = sizeof(int),
6037                 .mode           = 0644,
6038                 .proc_handler   = proc_dointvec,
6039         },
6040 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6041         {
6042                 .procname       = "optimistic_dad",
6043                 .data           = &ipv6_devconf.optimistic_dad,
6044                 .maxlen         = sizeof(int),
6045                 .mode           = 0644,
6046                 .proc_handler   = proc_dointvec,
6047         },
6048         {
6049                 .procname       = "use_optimistic",
6050                 .data           = &ipv6_devconf.use_optimistic,
6051                 .maxlen         = sizeof(int),
6052                 .mode           = 0644,
6053                 .proc_handler   = proc_dointvec,
6054         },
6055 #endif
6056 #ifdef CONFIG_IPV6_MROUTE
6057         {
6058                 .procname       = "mc_forwarding",
6059                 .data           = &ipv6_devconf.mc_forwarding,
6060                 .maxlen         = sizeof(int),
6061                 .mode           = 0444,
6062                 .proc_handler   = proc_dointvec,
6063         },
6064 #endif
6065         {
6066                 .procname       = "disable_ipv6",
6067                 .data           = &ipv6_devconf.disable_ipv6,
6068                 .maxlen         = sizeof(int),
6069                 .mode           = 0644,
6070                 .proc_handler   = addrconf_sysctl_disable,
6071         },
6072         {
6073                 .procname       = "accept_dad",
6074                 .data           = &ipv6_devconf.accept_dad,
6075                 .maxlen         = sizeof(int),
6076                 .mode           = 0644,
6077                 .proc_handler   = proc_dointvec,
6078         },
6079         {
6080                 .procname       = "force_tllao",
6081                 .data           = &ipv6_devconf.force_tllao,
6082                 .maxlen         = sizeof(int),
6083                 .mode           = 0644,
6084                 .proc_handler   = proc_dointvec
6085         },
6086         {
6087                 .procname       = "ndisc_notify",
6088                 .data           = &ipv6_devconf.ndisc_notify,
6089                 .maxlen         = sizeof(int),
6090                 .mode           = 0644,
6091                 .proc_handler   = proc_dointvec
6092         },
6093         {
6094                 .procname       = "suppress_frag_ndisc",
6095                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6096                 .maxlen         = sizeof(int),
6097                 .mode           = 0644,
6098                 .proc_handler   = proc_dointvec
6099         },
6100         {
6101                 .procname       = "accept_ra_from_local",
6102                 .data           = &ipv6_devconf.accept_ra_from_local,
6103                 .maxlen         = sizeof(int),
6104                 .mode           = 0644,
6105                 .proc_handler   = proc_dointvec,
6106         },
6107         {
6108                 .procname       = "accept_ra_mtu",
6109                 .data           = &ipv6_devconf.accept_ra_mtu,
6110                 .maxlen         = sizeof(int),
6111                 .mode           = 0644,
6112                 .proc_handler   = proc_dointvec,
6113         },
6114         {
6115                 .procname       = "stable_secret",
6116                 .data           = &ipv6_devconf.stable_secret,
6117                 .maxlen         = IPV6_MAX_STRLEN,
6118                 .mode           = 0600,
6119                 .proc_handler   = addrconf_sysctl_stable_secret,
6120         },
6121         {
6122                 .procname       = "use_oif_addrs_only",
6123                 .data           = &ipv6_devconf.use_oif_addrs_only,
6124                 .maxlen         = sizeof(int),
6125                 .mode           = 0644,
6126                 .proc_handler   = proc_dointvec,
6127         },
6128         {
6129                 .procname       = "ignore_routes_with_linkdown",
6130                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6131                 .maxlen         = sizeof(int),
6132                 .mode           = 0644,
6133                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6134         },
6135         {
6136                 .procname       = "drop_unicast_in_l2_multicast",
6137                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6138                 .maxlen         = sizeof(int),
6139                 .mode           = 0644,
6140                 .proc_handler   = proc_dointvec,
6141         },
6142         {
6143                 .procname       = "drop_unsolicited_na",
6144                 .data           = &ipv6_devconf.drop_unsolicited_na,
6145                 .maxlen         = sizeof(int),
6146                 .mode           = 0644,
6147                 .proc_handler   = proc_dointvec,
6148         },
6149         {
6150                 .procname       = "keep_addr_on_down",
6151                 .data           = &ipv6_devconf.keep_addr_on_down,
6152                 .maxlen         = sizeof(int),
6153                 .mode           = 0644,
6154                 .proc_handler   = proc_dointvec,
6155
6156         },
6157         {
6158                 .procname       = "seg6_enabled",
6159                 .data           = &ipv6_devconf.seg6_enabled,
6160                 .maxlen         = sizeof(int),
6161                 .mode           = 0644,
6162                 .proc_handler   = proc_dointvec,
6163         },
6164 #ifdef CONFIG_IPV6_SEG6_HMAC
6165         {
6166                 .procname       = "seg6_require_hmac",
6167                 .data           = &ipv6_devconf.seg6_require_hmac,
6168                 .maxlen         = sizeof(int),
6169                 .mode           = 0644,
6170                 .proc_handler   = proc_dointvec,
6171         },
6172 #endif
6173         {
6174                 .procname       = "enhanced_dad",
6175                 .data           = &ipv6_devconf.enhanced_dad,
6176                 .maxlen         = sizeof(int),
6177                 .mode           = 0644,
6178                 .proc_handler   = proc_dointvec,
6179         },
6180         {
6181                 .procname               = "addr_gen_mode",
6182                 .data                   = &ipv6_devconf.addr_gen_mode,
6183                 .maxlen                 = sizeof(int),
6184                 .mode                   = 0644,
6185                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6186         },
6187         {
6188                 /* sentinel */
6189         }
6190 };
6191
6192 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6193                 struct inet6_dev *idev, struct ipv6_devconf *p)
6194 {
6195         int i, ifindex;
6196         struct ctl_table *table;
6197         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6198
6199         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6200         if (!table)
6201                 goto out;
6202
6203         for (i = 0; table[i].data; i++) {
6204                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6205                 /* If one of these is already set, then it is not safe to
6206                  * overwrite either of them: this makes proc_dointvec_minmax
6207                  * usable.
6208                  */
6209                 if (!table[i].extra1 && !table[i].extra2) {
6210                         table[i].extra1 = idev; /* embedded; no ref */
6211                         table[i].extra2 = net;
6212                 }
6213         }
6214
6215         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6216
6217         p->sysctl_header = register_net_sysctl(net, path, table);
6218         if (!p->sysctl_header)
6219                 goto free;
6220
6221         if (!strcmp(dev_name, "all"))
6222                 ifindex = NETCONFA_IFINDEX_ALL;
6223         else if (!strcmp(dev_name, "default"))
6224                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6225         else
6226                 ifindex = idev->dev->ifindex;
6227         inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6228         return 0;
6229
6230 free:
6231         kfree(table);
6232 out:
6233         return -ENOBUFS;
6234 }
6235
6236 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6237 {
6238         struct ctl_table *table;
6239
6240         if (!p->sysctl_header)
6241                 return;
6242
6243         table = p->sysctl_header->ctl_table_arg;
6244         unregister_net_sysctl_table(p->sysctl_header);
6245         p->sysctl_header = NULL;
6246         kfree(table);
6247 }
6248
6249 static int addrconf_sysctl_register(struct inet6_dev *idev)
6250 {
6251         int err;
6252
6253         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6254                 return -EINVAL;
6255
6256         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6257                                     &ndisc_ifinfo_sysctl_change);
6258         if (err)
6259                 return err;
6260         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6261                                          idev, &idev->cnf);
6262         if (err)
6263                 neigh_sysctl_unregister(idev->nd_parms);
6264
6265         return err;
6266 }
6267
6268 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6269 {
6270         __addrconf_sysctl_unregister(&idev->cnf);
6271         neigh_sysctl_unregister(idev->nd_parms);
6272 }
6273
6274
6275 #endif
6276
6277 static int __net_init addrconf_init_net(struct net *net)
6278 {
6279         int err = -ENOMEM;
6280         struct ipv6_devconf *all, *dflt;
6281
6282         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6283         if (!all)
6284                 goto err_alloc_all;
6285
6286         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6287         if (!dflt)
6288                 goto err_alloc_dflt;
6289
6290         /* these will be inherited by all namespaces */
6291         dflt->autoconf = ipv6_defaults.autoconf;
6292         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6293
6294         dflt->stable_secret.initialized = false;
6295         all->stable_secret.initialized = false;
6296
6297         net->ipv6.devconf_all = all;
6298         net->ipv6.devconf_dflt = dflt;
6299
6300 #ifdef CONFIG_SYSCTL
6301         err = __addrconf_sysctl_register(net, "all", NULL, all);
6302         if (err < 0)
6303                 goto err_reg_all;
6304
6305         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6306         if (err < 0)
6307                 goto err_reg_dflt;
6308 #endif
6309         return 0;
6310
6311 #ifdef CONFIG_SYSCTL
6312 err_reg_dflt:
6313         __addrconf_sysctl_unregister(all);
6314 err_reg_all:
6315         kfree(dflt);
6316 #endif
6317 err_alloc_dflt:
6318         kfree(all);
6319 err_alloc_all:
6320         return err;
6321 }
6322
6323 static void __net_exit addrconf_exit_net(struct net *net)
6324 {
6325 #ifdef CONFIG_SYSCTL
6326         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6327         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6328 #endif
6329         kfree(net->ipv6.devconf_dflt);
6330         kfree(net->ipv6.devconf_all);
6331 }
6332
6333 static struct pernet_operations addrconf_ops = {
6334         .init = addrconf_init_net,
6335         .exit = addrconf_exit_net,
6336 };
6337
6338 static struct rtnl_af_ops inet6_ops __read_mostly = {
6339         .family           = AF_INET6,
6340         .fill_link_af     = inet6_fill_link_af,
6341         .get_link_af_size = inet6_get_link_af_size,
6342         .validate_link_af = inet6_validate_link_af,
6343         .set_link_af      = inet6_set_link_af,
6344 };
6345
6346 /*
6347  *      Init / cleanup code
6348  */
6349
6350 int __init addrconf_init(void)
6351 {
6352         struct inet6_dev *idev;
6353         int i, err;
6354
6355         err = ipv6_addr_label_init();
6356         if (err < 0) {
6357                 pr_crit("%s: cannot initialize default policy table: %d\n",
6358                         __func__, err);
6359                 goto out;
6360         }
6361
6362         err = register_pernet_subsys(&addrconf_ops);
6363         if (err < 0)
6364                 goto out_addrlabel;
6365
6366         addrconf_wq = create_workqueue("ipv6_addrconf");
6367         if (!addrconf_wq) {
6368                 err = -ENOMEM;
6369                 goto out_nowq;
6370         }
6371
6372         /* The addrconf netdev notifier requires that loopback_dev
6373          * has it's ipv6 private information allocated and setup
6374          * before it can bring up and give link-local addresses
6375          * to other devices which are up.
6376          *
6377          * Unfortunately, loopback_dev is not necessarily the first
6378          * entry in the global dev_base list of net devices.  In fact,
6379          * it is likely to be the very last entry on that list.
6380          * So this causes the notifier registry below to try and
6381          * give link-local addresses to all devices besides loopback_dev
6382          * first, then loopback_dev, which cases all the non-loopback_dev
6383          * devices to fail to get a link-local address.
6384          *
6385          * So, as a temporary fix, allocate the ipv6 structure for
6386          * loopback_dev first by hand.
6387          * Longer term, all of the dependencies ipv6 has upon the loopback
6388          * device and it being up should be removed.
6389          */
6390         rtnl_lock();
6391         idev = ipv6_add_dev(init_net.loopback_dev);
6392         rtnl_unlock();
6393         if (IS_ERR(idev)) {
6394                 err = PTR_ERR(idev);
6395                 goto errlo;
6396         }
6397
6398         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6399                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6400
6401         register_netdevice_notifier(&ipv6_dev_notf);
6402
6403         addrconf_verify();
6404
6405         rtnl_af_register(&inet6_ops);
6406
6407         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6408                               NULL);
6409         if (err < 0)
6410                 goto errout;
6411
6412         /* Only the first call to __rtnl_register can fail */
6413         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6414         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6415         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6416                         inet6_dump_ifaddr, NULL);
6417         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6418                         inet6_dump_ifmcaddr, NULL);
6419         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6420                         inet6_dump_ifacaddr, NULL);
6421         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6422                         inet6_netconf_dump_devconf, NULL);
6423
6424         ipv6_addr_label_rtnl_register();
6425
6426         return 0;
6427 errout:
6428         rtnl_af_unregister(&inet6_ops);
6429         unregister_netdevice_notifier(&ipv6_dev_notf);
6430 errlo:
6431         destroy_workqueue(addrconf_wq);
6432 out_nowq:
6433         unregister_pernet_subsys(&addrconf_ops);
6434 out_addrlabel:
6435         ipv6_addr_label_cleanup();
6436 out:
6437         return err;
6438 }
6439
6440 void addrconf_cleanup(void)
6441 {
6442         struct net_device *dev;
6443         int i;
6444
6445         unregister_netdevice_notifier(&ipv6_dev_notf);
6446         unregister_pernet_subsys(&addrconf_ops);
6447         ipv6_addr_label_cleanup();
6448
6449         rtnl_lock();
6450
6451         __rtnl_af_unregister(&inet6_ops);
6452
6453         /* clean dev list */
6454         for_each_netdev(&init_net, dev) {
6455                 if (__in6_dev_get(dev) == NULL)
6456                         continue;
6457                 addrconf_ifdown(dev, 1);
6458         }
6459         addrconf_ifdown(init_net.loopback_dev, 2);
6460
6461         /*
6462          *      Check hash table.
6463          */
6464         spin_lock_bh(&addrconf_hash_lock);
6465         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6466                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6467         spin_unlock_bh(&addrconf_hash_lock);
6468         cancel_delayed_work(&addr_chk_work);
6469         rtnl_unlock();
6470
6471         destroy_workqueue(addrconf_wq);
6472 }