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/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/inet.h>
  50 #include <linux/in6.h>
  51 #include <linux/netdevice.h>
  52 #include <linux/if_addr.h>
  53 #include <linux/if_arp.h>
  54 #include <linux/if_arcnet.h>
  55 #include <linux/if_infiniband.h>
  56 #include <linux/route.h>
  57 #include <linux/inetdevice.h>
  58 #include <linux/init.h>
  59 #include <linux/slab.h>
  60 #ifdef CONFIG_SYSCTL
  61 #include <linux/sysctl.h>
  62 #endif
  63 #include <linux/capability.h>
  64 #include <linux/delay.h>
  65 #include <linux/notifier.h>
  66 #include <linux/string.h>
  67 #include <linux/hash.h>
  68
  69 #include <net/net_namespace.h>
  70 #include <net/sock.h>
  71 #include <net/snmp.h>
  72
  73 #include <net/6lowpan.h>
  74 #include <net/firewire.h>
  75 #include <net/ipv6.h>
  76 #include <net/protocol.h>
  77 #include <net/ndisc.h>
  78 #include <net/ip6_route.h>
  79 #include <net/addrconf.h>
  80 #include <net/tcp.h>
  81 #include <net/ip.h>
  82 #include <net/netlink.h>
  83 #include <net/pkt_sched.h>
  84 #include <net/l3mdev.h>
  85 #include <linux/if_tunnel.h>
  86 #include <linux/rtnetlink.h>
  87 #include <linux/netconf.h>
  88 #include <linux/random.h>
  89 #include <linux/uaccess.h>
  90 #include <asm/unaligned.h>
  91
  92 #include <linux/proc_fs.h>
  93 #include <linux/seq_file.h>
  94 #include <linux/export.h>
  95
  96 /* Set to 3 to get tracing... */
  97 #define ACONF_DEBUG 2
  98
  99 #if ACONF_DEBUG >= 3
 100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 101 #else
 102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 103 #endif
 104
 105 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 106
 107 #define IPV6_MAX_STRLEN \
 108         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 109
 110 static inline u32 cstamp_delta(unsigned long cstamp)
 111 {
 112         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 113 }
 114
 115 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 116 {
 117         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 118         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 119         do_div(tmp, 1000000);
 120         return (s32)tmp;
 121 }
 122
 123 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 124 {
 125         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 126         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 127         do_div(tmp, 1000000);
 128         if ((s32)tmp > mrt) {
 129                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 130                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 131                 do_div(tmp, 1000000);
 132         }
 133         return (s32)tmp;
 134 }
 135
 136 #ifdef CONFIG_SYSCTL
 137 static int addrconf_sysctl_register(struct inet6_dev *idev);
 138 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 139 #else
 140 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 141 {
 142         return 0;
 143 }
 144
 145 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 146 {
 147 }
 148 #endif
 149
 150 static void ipv6_regen_rndid(struct inet6_dev *idev);
 151 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 152
 153 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 154 static int ipv6_count_addresses(struct inet6_dev *idev);
 155 static int ipv6_generate_stable_address(struct in6_addr *addr,
 156                                         u8 dad_count,
 157                                         const struct inet6_dev *idev);
 158
 159 /*
 160  *      Configured unicast address hash table
 161  */
 162 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 163 static DEFINE_SPINLOCK(addrconf_hash_lock);
 164
 165 static void addrconf_verify(void);
 166 static void addrconf_verify_rtnl(void);
 167 static void addrconf_verify_work(struct work_struct *);
 168
 169 static struct workqueue_struct *addrconf_wq;
 170 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 171
 172 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 173 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 174
 175 static void addrconf_type_change(struct net_device *dev,
 176                                  unsigned long event);
 177 static int addrconf_ifdown(struct net_device *dev, int how);
 178
 179 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 180                                                   int plen,
 181                                                   const struct net_device *dev,
 182                                                   u32 flags, u32 noflags);
 183
 184 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 185 static void addrconf_dad_work(struct work_struct *w);
 186 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 187 static void addrconf_dad_run(struct inet6_dev *idev);
 188 static void addrconf_rs_timer(unsigned long data);
 189 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 190 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 191
 192 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 193                                 struct prefix_info *pinfo);
 194 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 195                                struct net_device *dev);
 196
 197 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 198         .forwarding             = 0,
 199         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 200         .mtu6                   = IPV6_MIN_MTU,
 201         .accept_ra              = 1,
 202         .accept_redirects       = 1,
 203         .autoconf               = 1,
 204         .force_mld_version      = 0,
 205         .mldv1_unsolicited_report_interval = 10 * HZ,
 206         .mldv2_unsolicited_report_interval = HZ,
 207         .dad_transmits          = 1,
 208         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 209         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 210         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 211         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 212         .use_tempaddr           = 0,
 213         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 214         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 215         .regen_max_retry        = REGEN_MAX_RETRY,
 216         .max_desync_factor      = MAX_DESYNC_FACTOR,
 217         .max_addresses          = IPV6_MAX_ADDRESSES,
 218         .accept_ra_defrtr       = 1,
 219         .accept_ra_from_local   = 0,
 220         .accept_ra_min_hop_limit= 1,
 221         .accept_ra_pinfo        = 1,
 222 #ifdef CONFIG_IPV6_ROUTER_PREF
 223         .accept_ra_rtr_pref     = 1,
 224         .rtr_probe_interval     = 60 * HZ,
 225 #ifdef CONFIG_IPV6_ROUTE_INFO
 226         .accept_ra_rt_info_max_plen = 0,
 227 #endif
 228 #endif
 229         .proxy_ndp              = 0,
 230         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 231         .disable_ipv6           = 0,
 232         .accept_dad             = 1,
 233         .suppress_frag_ndisc    = 1,
 234         .accept_ra_mtu          = 1,
 235         .stable_secret          = {
 236                 .initialized = false,
 237         },
 238         .use_oif_addrs_only     = 0,
 239         .ignore_routes_with_linkdown = 0,
 240         .keep_addr_on_down      = 0,
 241 };
 242
 243 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 244         .forwarding             = 0,
 245         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 246         .mtu6                   = IPV6_MIN_MTU,
 247         .accept_ra              = 1,
 248         .accept_redirects       = 1,
 249         .autoconf               = 1,
 250         .force_mld_version      = 0,
 251         .mldv1_unsolicited_report_interval = 10 * HZ,
 252         .mldv2_unsolicited_report_interval = HZ,
 253         .dad_transmits          = 1,
 254         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 255         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 256         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 257         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 258         .use_tempaddr           = 0,
 259         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 260         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 261         .regen_max_retry        = REGEN_MAX_RETRY,
 262         .max_desync_factor      = MAX_DESYNC_FACTOR,
 263         .max_addresses          = IPV6_MAX_ADDRESSES,
 264         .accept_ra_defrtr       = 1,
 265         .accept_ra_from_local   = 0,
 266         .accept_ra_min_hop_limit= 1,
 267         .accept_ra_pinfo        = 1,
 268 #ifdef CONFIG_IPV6_ROUTER_PREF
 269         .accept_ra_rtr_pref     = 1,
 270         .rtr_probe_interval     = 60 * HZ,
 271 #ifdef CONFIG_IPV6_ROUTE_INFO
 272         .accept_ra_rt_info_max_plen = 0,
 273 #endif
 274 #endif
 275         .proxy_ndp              = 0,
 276         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 277         .disable_ipv6           = 0,
 278         .accept_dad             = 1,
 279         .suppress_frag_ndisc    = 1,
 280         .accept_ra_mtu          = 1,
 281         .stable_secret          = {
 282                 .initialized = false,
 283         },
 284         .use_oif_addrs_only     = 0,
 285         .ignore_routes_with_linkdown = 0,
 286         .keep_addr_on_down      = 0,
 287 };
 288
 289 /* Check if a valid qdisc is available */
 290 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 291 {
 292         return !qdisc_tx_is_noop(dev);
 293 }
 294
 295 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 296 {
 297         if (del_timer(&idev->rs_timer))
 298                 __in6_dev_put(idev);
 299 }
 300
 301 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 302 {
 303         if (cancel_delayed_work(&ifp->dad_work))
 304                 __in6_ifa_put(ifp);
 305 }
 306
 307 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 308                                   unsigned long when)
 309 {
 310         if (!timer_pending(&idev->rs_timer))
 311                 in6_dev_hold(idev);
 312         mod_timer(&idev->rs_timer, jiffies + when);
 313 }
 314
 315 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 316                                    unsigned long delay)
 317 {
 318         if (!delayed_work_pending(&ifp->dad_work))
 319                 in6_ifa_hold(ifp);
 320         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 321 }
 322
 323 static int snmp6_alloc_dev(struct inet6_dev *idev)
 324 {
 325         int i;
 326
 327         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 328         if (!idev->stats.ipv6)
 329                 goto err_ip;
 330
 331         for_each_possible_cpu(i) {
 332                 struct ipstats_mib *addrconf_stats;
 333                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 334                 u64_stats_init(&addrconf_stats->syncp);
 335         }
 336
 337
 338         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 339                                         GFP_KERNEL);
 340         if (!idev->stats.icmpv6dev)
 341                 goto err_icmp;
 342         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 343                                            GFP_KERNEL);
 344         if (!idev->stats.icmpv6msgdev)
 345                 goto err_icmpmsg;
 346
 347         return 0;
 348
 349 err_icmpmsg:
 350         kfree(idev->stats.icmpv6dev);
 351 err_icmp:
 352         free_percpu(idev->stats.ipv6);
 353 err_ip:
 354         return -ENOMEM;
 355 }
 356
 357 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 358 {
 359         struct inet6_dev *ndev;
 360         int err = -ENOMEM;
 361
 362         ASSERT_RTNL();
 363
 364         if (dev->mtu < IPV6_MIN_MTU)
 365                 return ERR_PTR(-EINVAL);
 366
 367         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 368         if (!ndev)
 369                 return ERR_PTR(err);
 370
 371         rwlock_init(&ndev->lock);
 372         ndev->dev = dev;
 373         INIT_LIST_HEAD(&ndev->addr_list);
 374         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 375                     (unsigned long)ndev);
 376         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 377
 378         if (ndev->cnf.stable_secret.initialized)
 379                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 380         else
 381                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
 382
 383         ndev->cnf.mtu6 = dev->mtu;
 384         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 385         if (!ndev->nd_parms) {
 386                 kfree(ndev);
 387                 return ERR_PTR(err);
 388         }
 389         if (ndev->cnf.forwarding)
 390                 dev_disable_lro(dev);
 391         /* We refer to the device */
 392         dev_hold(dev);
 393
 394         if (snmp6_alloc_dev(ndev) < 0) {
 395                 ADBG(KERN_WARNING
 396                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 397                         __func__, dev->name);
 398                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 399                 dev_put(dev);
 400                 kfree(ndev);
 401                 return ERR_PTR(err);
 402         }
 403
 404         if (snmp6_register_dev(ndev) < 0) {
 405                 ADBG(KERN_WARNING
 406                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 407                         __func__, dev->name);
 408                 goto err_release;
 409         }
 410
 411         /* One reference from device. */
 412         in6_dev_hold(ndev);
 413
 414         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 415                 ndev->cnf.accept_dad = -1;
 416
 417 #if IS_ENABLED(CONFIG_IPV6_SIT)
 418         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 419                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 420                 ndev->cnf.rtr_solicits = 0;
 421         }
 422 #endif
 423
 424         INIT_LIST_HEAD(&ndev->tempaddr_list);
 425         if ((dev->flags&IFF_LOOPBACK) ||
 426             dev->type == ARPHRD_TUNNEL ||
 427             dev->type == ARPHRD_TUNNEL6 ||
 428             dev->type == ARPHRD_SIT ||
 429             dev->type == ARPHRD_NONE) {
 430                 ndev->cnf.use_tempaddr = -1;
 431         } else
 432                 ipv6_regen_rndid(ndev);
 433
 434         ndev->token = in6addr_any;
 435
 436         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 437                 ndev->if_flags |= IF_READY;
 438
 439         ipv6_mc_init_dev(ndev);
 440         ndev->tstamp = jiffies;
 441         err = addrconf_sysctl_register(ndev);
 442         if (err) {
 443                 ipv6_mc_destroy_dev(ndev);
 444                 snmp6_unregister_dev(ndev);
 445                 goto err_release;
 446         }
 447         /* protected by rtnl_lock */
 448         rcu_assign_pointer(dev->ip6_ptr, ndev);
 449
 450         /* Join interface-local all-node multicast group */
 451         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 452
 453         /* Join all-node multicast group */
 454         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 455
 456         /* Join all-router multicast group if forwarding is set */
 457         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 458                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 459
 460         return ndev;
 461
 462 err_release:
 463         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 464         ndev->dead = 1;
 465         in6_dev_finish_destroy(ndev);
 466         return ERR_PTR(err);
 467 }
 468
 469 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 470 {
 471         struct inet6_dev *idev;
 472
 473         ASSERT_RTNL();
 474
 475         idev = __in6_dev_get(dev);
 476         if (!idev) {
 477                 idev = ipv6_add_dev(dev);
 478                 if (IS_ERR(idev))
 479                         return NULL;
 480         }
 481
 482         if (dev->flags&IFF_UP)
 483                 ipv6_mc_up(idev);
 484         return idev;
 485 }
 486
 487 static int inet6_netconf_msgsize_devconf(int type)
 488 {
 489         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 490                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 491         bool all = false;
 492
 493         if (type == NETCONFA_ALL)
 494                 all = true;
 495
 496         if (all || type == NETCONFA_FORWARDING)
 497                 size += nla_total_size(4);
 498 #ifdef CONFIG_IPV6_MROUTE
 499         if (all || type == NETCONFA_MC_FORWARDING)
 500                 size += nla_total_size(4);
 501 #endif
 502         if (all || type == NETCONFA_PROXY_NEIGH)
 503                 size += nla_total_size(4);
 504
 505         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 506                 size += nla_total_size(4);
 507
 508         return size;
 509 }
 510
 511 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 512                                       struct ipv6_devconf *devconf, u32 portid,
 513                                       u32 seq, int event, unsigned int flags,
 514                                       int type)
 515 {
 516         struct nlmsghdr  *nlh;
 517         struct netconfmsg *ncm;
 518         bool all = false;
 519
 520         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 521                         flags);
 522         if (!nlh)
 523                 return -EMSGSIZE;
 524
 525         if (type == NETCONFA_ALL)
 526                 all = true;
 527
 528         ncm = nlmsg_data(nlh);
 529         ncm->ncm_family = AF_INET6;
 530
 531         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 532                 goto nla_put_failure;
 533
 534         if ((all || type == NETCONFA_FORWARDING) &&
 535             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 536                 goto nla_put_failure;
 537 #ifdef CONFIG_IPV6_MROUTE
 538         if ((all || type == NETCONFA_MC_FORWARDING) &&
 539             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 540                         devconf->mc_forwarding) < 0)
 541                 goto nla_put_failure;
 542 #endif
 543         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 544             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 545                 goto nla_put_failure;
 546
 547         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 548             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 549                         devconf->ignore_routes_with_linkdown) < 0)
 550                 goto nla_put_failure;
 551
 552         nlmsg_end(skb, nlh);
 553         return 0;
 554
 555 nla_put_failure:
 556         nlmsg_cancel(skb, nlh);
 557         return -EMSGSIZE;
 558 }
 559
 560 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 561                                   struct ipv6_devconf *devconf)
 562 {
 563         struct sk_buff *skb;
 564         int err = -ENOBUFS;
 565
 566         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 567         if (!skb)
 568                 goto errout;
 569
 570         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 571                                          RTM_NEWNETCONF, 0, type);
 572         if (err < 0) {
 573                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 574                 WARN_ON(err == -EMSGSIZE);
 575                 kfree_skb(skb);
 576                 goto errout;
 577         }
 578         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 579         return;
 580 errout:
 581         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 582 }
 583
 584 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 585         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 586         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 587         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 588         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 589 };
 590
 591 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 592                                      struct nlmsghdr *nlh)
 593 {
 594         struct net *net = sock_net(in_skb->sk);
 595         struct nlattr *tb[NETCONFA_MAX+1];
 596         struct netconfmsg *ncm;
 597         struct sk_buff *skb;
 598         struct ipv6_devconf *devconf;
 599         struct inet6_dev *in6_dev;
 600         struct net_device *dev;
 601         int ifindex;
 602         int err;
 603
 604         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 605                           devconf_ipv6_policy);
 606         if (err < 0)
 607                 goto errout;
 608
 609         err = -EINVAL;
 610         if (!tb[NETCONFA_IFINDEX])
 611                 goto errout;
 612
 613         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 614         switch (ifindex) {
 615         case NETCONFA_IFINDEX_ALL:
 616                 devconf = net->ipv6.devconf_all;
 617                 break;
 618         case NETCONFA_IFINDEX_DEFAULT:
 619                 devconf = net->ipv6.devconf_dflt;
 620                 break;
 621         default:
 622                 dev = __dev_get_by_index(net, ifindex);
 623                 if (!dev)
 624                         goto errout;
 625                 in6_dev = __in6_dev_get(dev);
 626                 if (!in6_dev)
 627                         goto errout;
 628                 devconf = &in6_dev->cnf;
 629                 break;
 630         }
 631
 632         err = -ENOBUFS;
 633         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
 634         if (!skb)
 635                 goto errout;
 636
 637         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 638                                          NETLINK_CB(in_skb).portid,
 639                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 640                                          NETCONFA_ALL);
 641         if (err < 0) {
 642                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 643                 WARN_ON(err == -EMSGSIZE);
 644                 kfree_skb(skb);
 645                 goto errout;
 646         }
 647         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 648 errout:
 649         return err;
 650 }
 651
 652 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 653                                       struct netlink_callback *cb)
 654 {
 655         struct net *net = sock_net(skb->sk);
 656         int h, s_h;
 657         int idx, s_idx;
 658         struct net_device *dev;
 659         struct inet6_dev *idev;
 660         struct hlist_head *head;
 661
 662         s_h = cb->args[0];
 663         s_idx = idx = cb->args[1];
 664
 665         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 666                 idx = 0;
 667                 head = &net->dev_index_head[h];
 668                 rcu_read_lock();
 669                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 670                           net->dev_base_seq;
 671                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 672                         if (idx < s_idx)
 673                                 goto cont;
 674                         idev = __in6_dev_get(dev);
 675                         if (!idev)
 676                                 goto cont;
 677
 678                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 679                                                        &idev->cnf,
 680                                                        NETLINK_CB(cb->skb).portid,
 681                                                        cb->nlh->nlmsg_seq,
 682                                                        RTM_NEWNETCONF,
 683                                                        NLM_F_MULTI,
 684                                                        NETCONFA_ALL) < 0) {
 685                                 rcu_read_unlock();
 686                                 goto done;
 687                         }
 688                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 689 cont:
 690                         idx++;
 691                 }
 692                 rcu_read_unlock();
 693         }
 694         if (h == NETDEV_HASHENTRIES) {
 695                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 696                                                net->ipv6.devconf_all,
 697                                                NETLINK_CB(cb->skb).portid,
 698                                                cb->nlh->nlmsg_seq,
 699                                                RTM_NEWNETCONF, NLM_F_MULTI,
 700                                                NETCONFA_ALL) < 0)
 701                         goto done;
 702                 else
 703                         h++;
 704         }
 705         if (h == NETDEV_HASHENTRIES + 1) {
 706                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 707                                                net->ipv6.devconf_dflt,
 708                                                NETLINK_CB(cb->skb).portid,
 709                                                cb->nlh->nlmsg_seq,
 710                                                RTM_NEWNETCONF, NLM_F_MULTI,
 711                                                NETCONFA_ALL) < 0)
 712                         goto done;
 713                 else
 714                         h++;
 715         }
 716 done:
 717         cb->args[0] = h;
 718         cb->args[1] = idx;
 719
 720         return skb->len;
 721 }
 722
 723 #ifdef CONFIG_SYSCTL
 724 static void dev_forward_change(struct inet6_dev *idev)
 725 {
 726         struct net_device *dev;
 727         struct inet6_ifaddr *ifa;
 728
 729         if (!idev)
 730                 return;
 731         dev = idev->dev;
 732         if (idev->cnf.forwarding)
 733                 dev_disable_lro(dev);
 734         if (dev->flags & IFF_MULTICAST) {
 735                 if (idev->cnf.forwarding) {
 736                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 737                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 738                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 739                 } else {
 740                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 741                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 742                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 743                 }
 744         }
 745
 746         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 747                 if (ifa->flags&IFA_F_TENTATIVE)
 748                         continue;
 749                 if (idev->cnf.forwarding)
 750                         addrconf_join_anycast(ifa);
 751                 else
 752                         addrconf_leave_anycast(ifa);
 753         }
 754         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 755                                      dev->ifindex, &idev->cnf);
 756 }
 757
 758
 759 static void addrconf_forward_change(struct net *net, __s32 newf)
 760 {
 761         struct net_device *dev;
 762         struct inet6_dev *idev;
 763
 764         for_each_netdev(net, dev) {
 765                 idev = __in6_dev_get(dev);
 766                 if (idev) {
 767                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 768                         idev->cnf.forwarding = newf;
 769                         if (changed)
 770                                 dev_forward_change(idev);
 771                 }
 772         }
 773 }
 774
 775 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 776 {
 777         struct net *net;
 778         int old;
 779
 780         if (!rtnl_trylock())
 781                 return restart_syscall();
 782
 783         net = (struct net *)table->extra2;
 784         old = *p;
 785         *p = newf;
 786
 787         if (p == &net->ipv6.devconf_dflt->forwarding) {
 788                 if ((!newf) ^ (!old))
 789                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 790                                                      NETCONFA_IFINDEX_DEFAULT,
 791                                                      net->ipv6.devconf_dflt);
 792                 rtnl_unlock();
 793                 return 0;
 794         }
 795
 796         if (p == &net->ipv6.devconf_all->forwarding) {
 797                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 798
 799                 net->ipv6.devconf_dflt->forwarding = newf;
 800                 if ((!newf) ^ (!old_dflt))
 801                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 802                                                      NETCONFA_IFINDEX_DEFAULT,
 803                                                      net->ipv6.devconf_dflt);
 804
 805                 addrconf_forward_change(net, newf);
 806                 if ((!newf) ^ (!old))
 807                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 808                                                      NETCONFA_IFINDEX_ALL,
 809                                                      net->ipv6.devconf_all);
 810         } else if ((!newf) ^ (!old))
 811                 dev_forward_change((struct inet6_dev *)table->extra1);
 812         rtnl_unlock();
 813
 814         if (newf)
 815                 rt6_purge_dflt_routers(net);
 816         return 1;
 817 }
 818
 819 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 820 {
 821         struct net_device *dev;
 822         struct inet6_dev *idev;
 823
 824         for_each_netdev(net, dev) {
 825                 idev = __in6_dev_get(dev);
 826                 if (idev) {
 827                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 828
 829                         idev->cnf.ignore_routes_with_linkdown = newf;
 830                         if (changed)
 831                                 inet6_netconf_notify_devconf(dev_net(dev),
 832                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 833                                                              dev->ifindex,
 834                                                              &idev->cnf);
 835                 }
 836         }
 837 }
 838
 839 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 840 {
 841         struct net *net;
 842         int old;
 843
 844         if (!rtnl_trylock())
 845                 return restart_syscall();
 846
 847         net = (struct net *)table->extra2;
 848         old = *p;
 849         *p = newf;
 850
 851         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 852                 if ((!newf) ^ (!old))
 853                         inet6_netconf_notify_devconf(net,
 854                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 855                                                      NETCONFA_IFINDEX_DEFAULT,
 856                                                      net->ipv6.devconf_dflt);
 857                 rtnl_unlock();
 858                 return 0;
 859         }
 860
 861         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 862                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 863                 addrconf_linkdown_change(net, newf);
 864                 if ((!newf) ^ (!old))
 865                         inet6_netconf_notify_devconf(net,
 866                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 867                                                      NETCONFA_IFINDEX_ALL,
 868                                                      net->ipv6.devconf_all);
 869         }
 870         rtnl_unlock();
 871
 872         return 1;
 873 }
 874
 875 #endif
 876
 877 /* Nobody refers to this ifaddr, destroy it */
 878 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 879 {
 880         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 881
 882 #ifdef NET_REFCNT_DEBUG
 883         pr_debug("%s\n", __func__);
 884 #endif
 885
 886         in6_dev_put(ifp->idev);
 887
 888         if (cancel_delayed_work(&ifp->dad_work))
 889                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 890                           ifp);
 891
 892         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 893                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 894                 return;
 895         }
 896         ip6_rt_put(ifp->rt);
 897
 898         kfree_rcu(ifp, rcu);
 899 }
 900
 901 static void
 902 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 903 {
 904         struct list_head *p;
 905         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 906
 907         /*
 908          * Each device address list is sorted in order of scope -
 909          * global before linklocal.
 910          */
 911         list_for_each(p, &idev->addr_list) {
 912                 struct inet6_ifaddr *ifa
 913                         = list_entry(p, struct inet6_ifaddr, if_list);
 914                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 915                         break;
 916         }
 917
 918         list_add_tail(&ifp->if_list, p);
 919 }
 920
 921 static u32 inet6_addr_hash(const struct in6_addr *addr)
 922 {
 923         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 924 }
 925
 926 /* On success it returns ifp with increased reference count */
 927
 928 static struct inet6_ifaddr *
 929 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 930               const struct in6_addr *peer_addr, int pfxlen,
 931               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 932 {
 933         struct inet6_ifaddr *ifa = NULL;
 934         struct rt6_info *rt;
 935         unsigned int hash;
 936         int err = 0;
 937         int addr_type = ipv6_addr_type(addr);
 938
 939         if (addr_type == IPV6_ADDR_ANY ||
 940             addr_type & IPV6_ADDR_MULTICAST ||
 941             (!(idev->dev->flags & IFF_LOOPBACK) &&
 942              addr_type & IPV6_ADDR_LOOPBACK))
 943                 return ERR_PTR(-EADDRNOTAVAIL);
 944
 945         rcu_read_lock_bh();
 946         if (idev->dead) {
 947                 err = -ENODEV;                  /*XXX*/
 948                 goto out2;
 949         }
 950
 951         if (idev->cnf.disable_ipv6) {
 952                 err = -EACCES;
 953                 goto out2;
 954         }
 955
 956         spin_lock(&addrconf_hash_lock);
 957
 958         /* Ignore adding duplicate addresses on an interface */
 959         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 960                 ADBG("ipv6_add_addr: already assigned\n");
 961                 err = -EEXIST;
 962                 goto out;
 963         }
 964
 965         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 966
 967         if (!ifa) {
 968                 ADBG("ipv6_add_addr: malloc failed\n");
 969                 err = -ENOBUFS;
 970                 goto out;
 971         }
 972
 973         rt = addrconf_dst_alloc(idev, addr, false);
 974         if (IS_ERR(rt)) {
 975                 err = PTR_ERR(rt);
 976                 goto out;
 977         }
 978
 979         neigh_parms_data_state_setall(idev->nd_parms);
 980
 981         ifa->addr = *addr;
 982         if (peer_addr)
 983                 ifa->peer_addr = *peer_addr;
 984
 985         spin_lock_init(&ifa->lock);
 986         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 987         INIT_HLIST_NODE(&ifa->addr_lst);
 988         ifa->scope = scope;
 989         ifa->prefix_len = pfxlen;
 990         ifa->flags = flags | IFA_F_TENTATIVE;
 991         ifa->valid_lft = valid_lft;
 992         ifa->prefered_lft = prefered_lft;
 993         ifa->cstamp = ifa->tstamp = jiffies;
 994         ifa->tokenized = false;
 995
 996         ifa->rt = rt;
 997
 998         ifa->idev = idev;
 999         in6_dev_hold(idev);
1000         /* For caller */
1001         in6_ifa_hold(ifa);
1002
1003         /* Add to big hash table */
1004         hash = inet6_addr_hash(addr);
1005
1006         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1007         spin_unlock(&addrconf_hash_lock);
1008
1009         write_lock(&idev->lock);
1010         /* Add to inet6_dev unicast addr list. */
1011         ipv6_link_dev_addr(idev, ifa);
1012
1013         if (ifa->flags&IFA_F_TEMPORARY) {
1014                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1015                 in6_ifa_hold(ifa);
1016         }
1017
1018         in6_ifa_hold(ifa);
1019         write_unlock(&idev->lock);
1020 out2:
1021         rcu_read_unlock_bh();
1022
1023         if (likely(err == 0))
1024                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1025         else {
1026                 kfree(ifa);
1027                 ifa = ERR_PTR(err);
1028         }
1029
1030         return ifa;
1031 out:
1032         spin_unlock(&addrconf_hash_lock);
1033         goto out2;
1034 }
1035
1036 enum cleanup_prefix_rt_t {
1037         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1038         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1039         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1040 };
1041
1042 /*
1043  * Check, whether the prefix for ifp would still need a prefix route
1044  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1045  * constants.
1046  *
1047  * 1) we don't purge prefix if address was not permanent.
1048  *    prefix is managed by its own lifetime.
1049  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1050  * 3) if there are no addresses, delete prefix.
1051  * 4) if there are still other permanent address(es),
1052  *    corresponding prefix is still permanent.
1053  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1054  *    don't purge the prefix, assume user space is managing it.
1055  * 6) otherwise, update prefix lifetime to the
1056  *    longest valid lifetime among the corresponding
1057  *    addresses on the device.
1058  *    Note: subsequent RA will update lifetime.
1059  **/
1060 static enum cleanup_prefix_rt_t
1061 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1062 {
1063         struct inet6_ifaddr *ifa;
1064         struct inet6_dev *idev = ifp->idev;
1065         unsigned long lifetime;
1066         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1067
1068         *expires = jiffies;
1069
1070         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1071                 if (ifa == ifp)
1072                         continue;
1073                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1074                                        ifp->prefix_len))
1075                         continue;
1076                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1077                         return CLEANUP_PREFIX_RT_NOP;
1078
1079                 action = CLEANUP_PREFIX_RT_EXPIRE;
1080
1081                 spin_lock(&ifa->lock);
1082
1083                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1084                 /*
1085                  * Note: Because this address is
1086                  * not permanent, lifetime <
1087                  * LONG_MAX / HZ here.
1088                  */
1089                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1090                         *expires = ifa->tstamp + lifetime * HZ;
1091                 spin_unlock(&ifa->lock);
1092         }
1093
1094         return action;
1095 }
1096
1097 static void
1098 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1099 {
1100         struct rt6_info *rt;
1101
1102         rt = addrconf_get_prefix_route(&ifp->addr,
1103                                        ifp->prefix_len,
1104                                        ifp->idev->dev,
1105                                        0, RTF_GATEWAY | RTF_DEFAULT);
1106         if (rt) {
1107                 if (del_rt)
1108                         ip6_del_rt(rt);
1109                 else {
1110                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1111                                 rt6_set_expires(rt, expires);
1112                         ip6_rt_put(rt);
1113                 }
1114         }
1115 }
1116
1117
1118 /* This function wants to get referenced ifp and releases it before return */
1119
1120 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1121 {
1122         int state;
1123         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1124         unsigned long expires;
1125
1126         ASSERT_RTNL();
1127
1128         spin_lock_bh(&ifp->lock);
1129         state = ifp->state;
1130         ifp->state = INET6_IFADDR_STATE_DEAD;
1131         spin_unlock_bh(&ifp->lock);
1132
1133         if (state == INET6_IFADDR_STATE_DEAD)
1134                 goto out;
1135
1136         spin_lock_bh(&addrconf_hash_lock);
1137         hlist_del_init_rcu(&ifp->addr_lst);
1138         spin_unlock_bh(&addrconf_hash_lock);
1139
1140         write_lock_bh(&ifp->idev->lock);
1141
1142         if (ifp->flags&IFA_F_TEMPORARY) {
1143                 list_del(&ifp->tmp_list);
1144                 if (ifp->ifpub) {
1145                         in6_ifa_put(ifp->ifpub);
1146                         ifp->ifpub = NULL;
1147                 }
1148                 __in6_ifa_put(ifp);
1149         }
1150
1151         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1152                 action = check_cleanup_prefix_route(ifp, &expires);
1153
1154         list_del_init(&ifp->if_list);
1155         __in6_ifa_put(ifp);
1156
1157         write_unlock_bh(&ifp->idev->lock);
1158
1159         addrconf_del_dad_work(ifp);
1160
1161         ipv6_ifa_notify(RTM_DELADDR, ifp);
1162
1163         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1164
1165         if (action != CLEANUP_PREFIX_RT_NOP) {
1166                 cleanup_prefix_route(ifp, expires,
1167                         action == CLEANUP_PREFIX_RT_DEL);
1168         }
1169
1170         /* clean up prefsrc entries */
1171         rt6_remove_prefsrc(ifp);
1172 out:
1173         in6_ifa_put(ifp);
1174 }
1175
1176 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1177 {
1178         struct inet6_dev *idev = ifp->idev;
1179         struct in6_addr addr, *tmpaddr;
1180         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1181         unsigned long regen_advance;
1182         int tmp_plen;
1183         int ret = 0;
1184         u32 addr_flags;
1185         unsigned long now = jiffies;
1186
1187         write_lock_bh(&idev->lock);
1188         if (ift) {
1189                 spin_lock_bh(&ift->lock);
1190                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1191                 spin_unlock_bh(&ift->lock);
1192                 tmpaddr = &addr;
1193         } else {
1194                 tmpaddr = NULL;
1195         }
1196 retry:
1197         in6_dev_hold(idev);
1198         if (idev->cnf.use_tempaddr <= 0) {
1199                 write_unlock_bh(&idev->lock);
1200                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1201                 in6_dev_put(idev);
1202                 ret = -1;
1203                 goto out;
1204         }
1205         spin_lock_bh(&ifp->lock);
1206         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1207                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1208                 spin_unlock_bh(&ifp->lock);
1209                 write_unlock_bh(&idev->lock);
1210                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1211                         __func__);
1212                 in6_dev_put(idev);
1213                 ret = -1;
1214                 goto out;
1215         }
1216         in6_ifa_hold(ifp);
1217         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1218         ipv6_try_regen_rndid(idev, tmpaddr);
1219         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1220         age = (now - ifp->tstamp) / HZ;
1221         tmp_valid_lft = min_t(__u32,
1222                               ifp->valid_lft,
1223                               idev->cnf.temp_valid_lft + age);
1224         tmp_prefered_lft = min_t(__u32,
1225                                  ifp->prefered_lft,
1226                                  idev->cnf.temp_prefered_lft + age -
1227                                  idev->cnf.max_desync_factor);
1228         tmp_plen = ifp->prefix_len;
1229         tmp_tstamp = ifp->tstamp;
1230         spin_unlock_bh(&ifp->lock);
1231
1232         regen_advance = idev->cnf.regen_max_retry *
1233                         idev->cnf.dad_transmits *
1234                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1235         write_unlock_bh(&idev->lock);
1236
1237         /* A temporary address is created only if this calculated Preferred
1238          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1239          * an implementation must not create a temporary address with a zero
1240          * Preferred Lifetime.
1241          * Use age calculation as in addrconf_verify to avoid unnecessary
1242          * temporary addresses being generated.
1243          */
1244         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1245         if (tmp_prefered_lft <= regen_advance + age) {
1246                 in6_ifa_put(ifp);
1247                 in6_dev_put(idev);
1248                 ret = -1;
1249                 goto out;
1250         }
1251
1252         addr_flags = IFA_F_TEMPORARY;
1253         /* set in addrconf_prefix_rcv() */
1254         if (ifp->flags & IFA_F_OPTIMISTIC)
1255                 addr_flags |= IFA_F_OPTIMISTIC;
1256
1257         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1258                             ipv6_addr_scope(&addr), addr_flags,
1259                             tmp_valid_lft, tmp_prefered_lft);
1260         if (IS_ERR(ift)) {
1261                 in6_ifa_put(ifp);
1262                 in6_dev_put(idev);
1263                 pr_info("%s: retry temporary address regeneration\n", __func__);
1264                 tmpaddr = &addr;
1265                 write_lock_bh(&idev->lock);
1266                 goto retry;
1267         }
1268
1269         spin_lock_bh(&ift->lock);
1270         ift->ifpub = ifp;
1271         ift->cstamp = now;
1272         ift->tstamp = tmp_tstamp;
1273         spin_unlock_bh(&ift->lock);
1274
1275         addrconf_dad_start(ift);
1276         in6_ifa_put(ift);
1277         in6_dev_put(idev);
1278 out:
1279         return ret;
1280 }
1281
1282 /*
1283  *      Choose an appropriate source address (RFC3484)
1284  */
1285 enum {
1286         IPV6_SADDR_RULE_INIT = 0,
1287         IPV6_SADDR_RULE_LOCAL,
1288         IPV6_SADDR_RULE_SCOPE,
1289         IPV6_SADDR_RULE_PREFERRED,
1290 #ifdef CONFIG_IPV6_MIP6
1291         IPV6_SADDR_RULE_HOA,
1292 #endif
1293         IPV6_SADDR_RULE_OIF,
1294         IPV6_SADDR_RULE_LABEL,
1295         IPV6_SADDR_RULE_PRIVACY,
1296         IPV6_SADDR_RULE_ORCHID,
1297         IPV6_SADDR_RULE_PREFIX,
1298 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1299         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1300 #endif
1301         IPV6_SADDR_RULE_MAX
1302 };
1303
1304 struct ipv6_saddr_score {
1305         int                     rule;
1306         int                     addr_type;
1307         struct inet6_ifaddr     *ifa;
1308         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1309         int                     scopedist;
1310         int                     matchlen;
1311 };
1312
1313 struct ipv6_saddr_dst {
1314         const struct in6_addr *addr;
1315         int ifindex;
1316         int scope;
1317         int label;
1318         unsigned int prefs;
1319 };
1320
1321 static inline int ipv6_saddr_preferred(int type)
1322 {
1323         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1324                 return 1;
1325         return 0;
1326 }
1327
1328 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1329 {
1330 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1331         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1332 #else
1333         return false;
1334 #endif
1335 }
1336
1337 static int ipv6_get_saddr_eval(struct net *net,
1338                                struct ipv6_saddr_score *score,
1339                                struct ipv6_saddr_dst *dst,
1340                                int i)
1341 {
1342         int ret;
1343
1344         if (i <= score->rule) {
1345                 switch (i) {
1346                 case IPV6_SADDR_RULE_SCOPE:
1347                         ret = score->scopedist;
1348                         break;
1349                 case IPV6_SADDR_RULE_PREFIX:
1350                         ret = score->matchlen;
1351                         break;
1352                 default:
1353                         ret = !!test_bit(i, score->scorebits);
1354                 }
1355                 goto out;
1356         }
1357
1358         switch (i) {
1359         case IPV6_SADDR_RULE_INIT:
1360                 /* Rule 0: remember if hiscore is not ready yet */
1361                 ret = !!score->ifa;
1362                 break;
1363         case IPV6_SADDR_RULE_LOCAL:
1364                 /* Rule 1: Prefer same address */
1365                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1366                 break;
1367         case IPV6_SADDR_RULE_SCOPE:
1368                 /* Rule 2: Prefer appropriate scope
1369                  *
1370                  *      ret
1371                  *       ^
1372                  *    -1 |  d 15
1373                  *    ---+--+-+---> scope
1374                  *       |
1375                  *       |             d is scope of the destination.
1376                  *  B-d  |  \
1377                  *       |   \      <- smaller scope is better if
1378                  *  B-15 |    \        if scope is enough for destination.
1379                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1380                  * d-C-1 | /
1381                  *       |/         <- greater is better
1382                  *   -C  /             if scope is not enough for destination.
1383                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1384                  *
1385                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1386                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1387                  * Assume B = 0 and we get C > 29.
1388                  */
1389                 ret = __ipv6_addr_src_scope(score->addr_type);
1390                 if (ret >= dst->scope)
1391                         ret = -ret;
1392                 else
1393                         ret -= 128;     /* 30 is enough */
1394                 score->scopedist = ret;
1395                 break;
1396         case IPV6_SADDR_RULE_PREFERRED:
1397             {
1398                 /* Rule 3: Avoid deprecated and optimistic addresses */
1399                 u8 avoid = IFA_F_DEPRECATED;
1400
1401                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1402                         avoid |= IFA_F_OPTIMISTIC;
1403                 ret = ipv6_saddr_preferred(score->addr_type) ||
1404                       !(score->ifa->flags & avoid);
1405                 break;
1406             }
1407 #ifdef CONFIG_IPV6_MIP6
1408         case IPV6_SADDR_RULE_HOA:
1409             {
1410                 /* Rule 4: Prefer home address */
1411                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1412                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1413                 break;
1414             }
1415 #endif
1416         case IPV6_SADDR_RULE_OIF:
1417                 /* Rule 5: Prefer outgoing interface */
1418                 ret = (!dst->ifindex ||
1419                        dst->ifindex == score->ifa->idev->dev->ifindex);
1420                 break;
1421         case IPV6_SADDR_RULE_LABEL:
1422                 /* Rule 6: Prefer matching label */
1423                 ret = ipv6_addr_label(net,
1424                                       &score->ifa->addr, score->addr_type,
1425                                       score->ifa->idev->dev->ifindex) == dst->label;
1426                 break;
1427         case IPV6_SADDR_RULE_PRIVACY:
1428             {
1429                 /* Rule 7: Prefer public address
1430                  * Note: prefer temporary address if use_tempaddr >= 2
1431                  */
1432                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1433                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1434                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1435                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1436                 break;
1437             }
1438         case IPV6_SADDR_RULE_ORCHID:
1439                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1440                  *          non-ORCHID vs non-ORCHID
1441                  */
1442                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1443                         ipv6_addr_orchid(dst->addr));
1444                 break;
1445         case IPV6_SADDR_RULE_PREFIX:
1446                 /* Rule 8: Use longest matching prefix */
1447                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1448                 if (ret > score->ifa->prefix_len)
1449                         ret = score->ifa->prefix_len;
1450                 score->matchlen = ret;
1451                 break;
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1454                 /* Optimistic addresses still have lower precedence than other
1455                  * preferred addresses.
1456                  */
1457                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1458                 break;
1459 #endif
1460         default:
1461                 ret = 0;
1462         }
1463
1464         if (ret)
1465                 __set_bit(i, score->scorebits);
1466         score->rule = i;
1467 out:
1468         return ret;
1469 }
1470
1471 static int __ipv6_dev_get_saddr(struct net *net,
1472                                 struct ipv6_saddr_dst *dst,
1473                                 struct inet6_dev *idev,
1474                                 struct ipv6_saddr_score *scores,
1475                                 int hiscore_idx)
1476 {
1477         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1478
1479         read_lock_bh(&idev->lock);
1480         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1481                 int i;
1482
1483                 /*
1484                  * - Tentative Address (RFC2462 section 5.4)
1485                  *  - A tentative address is not considered
1486                  *    "assigned to an interface" in the traditional
1487                  *    sense, unless it is also flagged as optimistic.
1488                  * - Candidate Source Address (section 4)
1489                  *  - In any case, anycast addresses, multicast
1490                  *    addresses, and the unspecified address MUST
1491                  *    NOT be included in a candidate set.
1492                  */
1493                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1494                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1495                         continue;
1496
1497                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1498
1499                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1500                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1501                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1502                                             idev->dev->name);
1503                         continue;
1504                 }
1505
1506                 score->rule = -1;
1507                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1508
1509                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1510                         int minihiscore, miniscore;
1511
1512                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1513                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1514
1515                         if (minihiscore > miniscore) {
1516                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1517                                     score->scopedist > 0) {
1518                                         /*
1519                                          * special case:
1520                                          * each remaining entry
1521                                          * has too small (not enough)
1522                                          * scope, because ifa entries
1523                                          * are sorted by their scope
1524                                          * values.
1525                                          */
1526                                         goto out;
1527                                 }
1528                                 break;
1529                         } else if (minihiscore < miniscore) {
1530                                 if (hiscore->ifa)
1531                                         in6_ifa_put(hiscore->ifa);
1532
1533                                 in6_ifa_hold(score->ifa);
1534
1535                                 swap(hiscore, score);
1536                                 hiscore_idx = 1 - hiscore_idx;
1537
1538                                 /* restore our iterator */
1539                                 score->ifa = hiscore->ifa;
1540
1541                                 break;
1542                         }
1543                 }
1544         }
1545 out:
1546         read_unlock_bh(&idev->lock);
1547         return hiscore_idx;
1548 }
1549
1550 static int ipv6_get_saddr_master(struct net *net,
1551                                  const struct net_device *dst_dev,
1552                                  const struct net_device *master,
1553                                  struct ipv6_saddr_dst *dst,
1554                                  struct ipv6_saddr_score *scores,
1555                                  int hiscore_idx)
1556 {
1557         struct inet6_dev *idev;
1558
1559         idev = __in6_dev_get(dst_dev);
1560         if (idev)
1561                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1562                                                    scores, hiscore_idx);
1563
1564         idev = __in6_dev_get(master);
1565         if (idev)
1566                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1567                                                    scores, hiscore_idx);
1568
1569         return hiscore_idx;
1570 }
1571
1572 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1573                        const struct in6_addr *daddr, unsigned int prefs,
1574                        struct in6_addr *saddr)
1575 {
1576         struct ipv6_saddr_score scores[2], *hiscore;
1577         struct ipv6_saddr_dst dst;
1578         struct inet6_dev *idev;
1579         struct net_device *dev;
1580         int dst_type;
1581         bool use_oif_addr = false;
1582         int hiscore_idx = 0;
1583
1584         dst_type = __ipv6_addr_type(daddr);
1585         dst.addr = daddr;
1586         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1587         dst.scope = __ipv6_addr_src_scope(dst_type);
1588         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1589         dst.prefs = prefs;
1590
1591         scores[hiscore_idx].rule = -1;
1592         scores[hiscore_idx].ifa = NULL;
1593
1594         rcu_read_lock();
1595
1596         /* Candidate Source Address (section 4)
1597          *  - multicast and link-local destination address,
1598          *    the set of candidate source address MUST only
1599          *    include addresses assigned to interfaces
1600          *    belonging to the same link as the outgoing
1601          *    interface.
1602          * (- For site-local destination addresses, the
1603          *    set of candidate source addresses MUST only
1604          *    include addresses assigned to interfaces
1605          *    belonging to the same site as the outgoing
1606          *    interface.)
1607          *  - "It is RECOMMENDED that the candidate source addresses
1608          *    be the set of unicast addresses assigned to the
1609          *    interface that will be used to send to the destination
1610          *    (the 'outgoing' interface)." (RFC 6724)
1611          */
1612         if (dst_dev) {
1613                 idev = __in6_dev_get(dst_dev);
1614                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1615                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1616                     (idev && idev->cnf.use_oif_addrs_only)) {
1617                         use_oif_addr = true;
1618                 }
1619         }
1620
1621         if (use_oif_addr) {
1622                 if (idev)
1623                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1624         } else {
1625                 const struct net_device *master;
1626                 int master_idx = 0;
1627
1628                 /* if dst_dev exists and is enslaved to an L3 device, then
1629                  * prefer addresses from dst_dev and then the master over
1630                  * any other enslaved devices in the L3 domain.
1631                  */
1632                 master = l3mdev_master_dev_rcu(dst_dev);
1633                 if (master) {
1634                         master_idx = master->ifindex;
1635
1636                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1637                                                             master, &dst,
1638                                                             scores, hiscore_idx);
1639
1640                         if (scores[hiscore_idx].ifa)
1641                                 goto out;
1642                 }
1643
1644                 for_each_netdev_rcu(net, dev) {
1645                         /* only consider addresses on devices in the
1646                          * same L3 domain
1647                          */
1648                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1649                                 continue;
1650                         idev = __in6_dev_get(dev);
1651                         if (!idev)
1652                                 continue;
1653                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1654                 }
1655         }
1656
1657 out:
1658         rcu_read_unlock();
1659
1660         hiscore = &scores[hiscore_idx];
1661         if (!hiscore->ifa)
1662                 return -EADDRNOTAVAIL;
1663
1664         *saddr = hiscore->ifa->addr;
1665         in6_ifa_put(hiscore->ifa);
1666         return 0;
1667 }
1668 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1669
1670 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1671                       u32 banned_flags)
1672 {
1673         struct inet6_ifaddr *ifp;
1674         int err = -EADDRNOTAVAIL;
1675
1676         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1677                 if (ifp->scope > IFA_LINK)
1678                         break;
1679                 if (ifp->scope == IFA_LINK &&
1680                     !(ifp->flags & banned_flags)) {
1681                         *addr = ifp->addr;
1682                         err = 0;
1683                         break;
1684                 }
1685         }
1686         return err;
1687 }
1688
1689 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1690                     u32 banned_flags)
1691 {
1692         struct inet6_dev *idev;
1693         int err = -EADDRNOTAVAIL;
1694
1695         rcu_read_lock();
1696         idev = __in6_dev_get(dev);
1697         if (idev) {
1698                 read_lock_bh(&idev->lock);
1699                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1700                 read_unlock_bh(&idev->lock);
1701         }
1702         rcu_read_unlock();
1703         return err;
1704 }
1705
1706 static int ipv6_count_addresses(struct inet6_dev *idev)
1707 {
1708         int cnt = 0;
1709         struct inet6_ifaddr *ifp;
1710
1711         read_lock_bh(&idev->lock);
1712         list_for_each_entry(ifp, &idev->addr_list, if_list)
1713                 cnt++;
1714         read_unlock_bh(&idev->lock);
1715         return cnt;
1716 }
1717
1718 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1719                   const struct net_device *dev, int strict)
1720 {
1721         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1722 }
1723 EXPORT_SYMBOL(ipv6_chk_addr);
1724
1725 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1726                             const struct net_device *dev, int strict,
1727                             u32 banned_flags)
1728 {
1729         struct inet6_ifaddr *ifp;
1730         unsigned int hash = inet6_addr_hash(addr);
1731         u32 ifp_flags;
1732
1733         rcu_read_lock_bh();
1734         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1735                 if (!net_eq(dev_net(ifp->idev->dev), net))
1736                         continue;
1737                 /* Decouple optimistic from tentative for evaluation here.
1738                  * Ban optimistic addresses explicitly, when required.
1739                  */
1740                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1741                             ? (ifp->flags&~IFA_F_TENTATIVE)
1742                             : ifp->flags;
1743                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1744                     !(ifp_flags&banned_flags) &&
1745                     (!dev || ifp->idev->dev == dev ||
1746                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1747                         rcu_read_unlock_bh();
1748                         return 1;
1749                 }
1750         }
1751
1752         rcu_read_unlock_bh();
1753         return 0;
1754 }
1755 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1756
1757 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1758                                struct net_device *dev)
1759 {
1760         unsigned int hash = inet6_addr_hash(addr);
1761         struct inet6_ifaddr *ifp;
1762
1763         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1764                 if (!net_eq(dev_net(ifp->idev->dev), net))
1765                         continue;
1766                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1767                         if (!dev || ifp->idev->dev == dev)
1768                                 return true;
1769                 }
1770         }
1771         return false;
1772 }
1773
1774 /* Compares an address/prefix_len with addresses on device @dev.
1775  * If one is found it returns true.
1776  */
1777 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1778         const unsigned int prefix_len, struct net_device *dev)
1779 {
1780         struct inet6_dev *idev;
1781         struct inet6_ifaddr *ifa;
1782         bool ret = false;
1783
1784         rcu_read_lock();
1785         idev = __in6_dev_get(dev);
1786         if (idev) {
1787                 read_lock_bh(&idev->lock);
1788                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1789                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1790                         if (ret)
1791                                 break;
1792                 }
1793                 read_unlock_bh(&idev->lock);
1794         }
1795         rcu_read_unlock();
1796
1797         return ret;
1798 }
1799 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1800
1801 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1802 {
1803         struct inet6_dev *idev;
1804         struct inet6_ifaddr *ifa;
1805         int     onlink;
1806
1807         onlink = 0;
1808         rcu_read_lock();
1809         idev = __in6_dev_get(dev);
1810         if (idev) {
1811                 read_lock_bh(&idev->lock);
1812                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1813                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1814                                                    ifa->prefix_len);
1815                         if (onlink)
1816                                 break;
1817                 }
1818                 read_unlock_bh(&idev->lock);
1819         }
1820         rcu_read_unlock();
1821         return onlink;
1822 }
1823 EXPORT_SYMBOL(ipv6_chk_prefix);
1824
1825 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1826                                      struct net_device *dev, int strict)
1827 {
1828         struct inet6_ifaddr *ifp, *result = NULL;
1829         unsigned int hash = inet6_addr_hash(addr);
1830
1831         rcu_read_lock_bh();
1832         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1833                 if (!net_eq(dev_net(ifp->idev->dev), net))
1834                         continue;
1835                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1836                         if (!dev || ifp->idev->dev == dev ||
1837                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1838                                 result = ifp;
1839                                 in6_ifa_hold(ifp);
1840                                 break;
1841                         }
1842                 }
1843         }
1844         rcu_read_unlock_bh();
1845
1846         return result;
1847 }
1848
1849 /* Gets referenced address, destroys ifaddr */
1850
1851 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1852 {
1853         if (dad_failed)
1854                 ifp->flags |= IFA_F_DADFAILED;
1855
1856         if (ifp->flags&IFA_F_PERMANENT) {
1857                 spin_lock_bh(&ifp->lock);
1858                 addrconf_del_dad_work(ifp);
1859                 ifp->flags |= IFA_F_TENTATIVE;
1860                 spin_unlock_bh(&ifp->lock);
1861                 if (dad_failed)
1862                         ipv6_ifa_notify(0, ifp);
1863                 in6_ifa_put(ifp);
1864         } else if (ifp->flags&IFA_F_TEMPORARY) {
1865                 struct inet6_ifaddr *ifpub;
1866                 spin_lock_bh(&ifp->lock);
1867                 ifpub = ifp->ifpub;
1868                 if (ifpub) {
1869                         in6_ifa_hold(ifpub);
1870                         spin_unlock_bh(&ifp->lock);
1871                         ipv6_create_tempaddr(ifpub, ifp);
1872                         in6_ifa_put(ifpub);
1873                 } else {
1874                         spin_unlock_bh(&ifp->lock);
1875                 }
1876                 ipv6_del_addr(ifp);
1877         } else {
1878                 ipv6_del_addr(ifp);
1879         }
1880 }
1881
1882 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1883 {
1884         int err = -ENOENT;
1885
1886         spin_lock_bh(&ifp->lock);
1887         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1888                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1889                 err = 0;
1890         }
1891         spin_unlock_bh(&ifp->lock);
1892
1893         return err;
1894 }
1895
1896 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1897 {
1898         struct inet6_dev *idev = ifp->idev;
1899         struct net *net = dev_net(ifp->idev->dev);
1900
1901         if (addrconf_dad_end(ifp)) {
1902                 in6_ifa_put(ifp);
1903                 return;
1904         }
1905
1906         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1907                              ifp->idev->dev->name, &ifp->addr);
1908
1909         spin_lock_bh(&ifp->lock);
1910
1911         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1912                 int scope = ifp->scope;
1913                 u32 flags = ifp->flags;
1914                 struct in6_addr new_addr;
1915                 struct inet6_ifaddr *ifp2;
1916                 u32 valid_lft, preferred_lft;
1917                 int pfxlen = ifp->prefix_len;
1918                 int retries = ifp->stable_privacy_retry + 1;
1919
1920                 if (retries > net->ipv6.sysctl.idgen_retries) {
1921                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1922                                              ifp->idev->dev->name);
1923                         goto errdad;
1924                 }
1925
1926                 new_addr = ifp->addr;
1927                 if (ipv6_generate_stable_address(&new_addr, retries,
1928                                                  idev))
1929                         goto errdad;
1930
1931                 valid_lft = ifp->valid_lft;
1932                 preferred_lft = ifp->prefered_lft;
1933
1934                 spin_unlock_bh(&ifp->lock);
1935
1936                 if (idev->cnf.max_addresses &&
1937                     ipv6_count_addresses(idev) >=
1938                     idev->cnf.max_addresses)
1939                         goto lock_errdad;
1940
1941                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1942                                      ifp->idev->dev->name);
1943
1944                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1945                                      scope, flags, valid_lft,
1946                                      preferred_lft);
1947                 if (IS_ERR(ifp2))
1948                         goto lock_errdad;
1949
1950                 spin_lock_bh(&ifp2->lock);
1951                 ifp2->stable_privacy_retry = retries;
1952                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1953                 spin_unlock_bh(&ifp2->lock);
1954
1955                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1956                 in6_ifa_put(ifp2);
1957 lock_errdad:
1958                 spin_lock_bh(&ifp->lock);
1959         }
1960
1961 errdad:
1962         /* transition from _POSTDAD to _ERRDAD */
1963         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1964         spin_unlock_bh(&ifp->lock);
1965
1966         addrconf_mod_dad_work(ifp, 0);
1967         in6_ifa_put(ifp);
1968 }
1969
1970 /* Join to solicited addr multicast group.
1971  * caller must hold RTNL */
1972 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1973 {
1974         struct in6_addr maddr;
1975
1976         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1977                 return;
1978
1979         addrconf_addr_solict_mult(addr, &maddr);
1980         ipv6_dev_mc_inc(dev, &maddr);
1981 }
1982
1983 /* caller must hold RTNL */
1984 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1985 {
1986         struct in6_addr maddr;
1987
1988         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1989                 return;
1990
1991         addrconf_addr_solict_mult(addr, &maddr);
1992         __ipv6_dev_mc_dec(idev, &maddr);
1993 }
1994
1995 /* caller must hold RTNL */
1996 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1997 {
1998         struct in6_addr addr;
1999
2000         if (ifp->prefix_len >= 127) /* RFC 6164 */
2001                 return;
2002         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2003         if (ipv6_addr_any(&addr))
2004                 return;
2005         __ipv6_dev_ac_inc(ifp->idev, &addr);
2006 }
2007
2008 /* caller must hold RTNL */
2009 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2010 {
2011         struct in6_addr addr;
2012
2013         if (ifp->prefix_len >= 127) /* RFC 6164 */
2014                 return;
2015         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2016         if (ipv6_addr_any(&addr))
2017                 return;
2018         __ipv6_dev_ac_dec(ifp->idev, &addr);
2019 }
2020
2021 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2022 {
2023         if (dev->addr_len != EUI64_ADDR_LEN)
2024                 return -1;
2025         memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2026         eui[0] ^= 2;
2027         return 0;
2028 }
2029
2030 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2031 {
2032         union fwnet_hwaddr *ha;
2033
2034         if (dev->addr_len != FWNET_ALEN)
2035                 return -1;
2036
2037         ha = (union fwnet_hwaddr *)dev->dev_addr;
2038
2039         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2040         eui[0] ^= 2;
2041         return 0;
2042 }
2043
2044 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2045 {
2046         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2047         if (dev->addr_len != ARCNET_ALEN)
2048                 return -1;
2049         memset(eui, 0, 7);
2050         eui[7] = *(u8 *)dev->dev_addr;
2051         return 0;
2052 }
2053
2054 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2055 {
2056         if (dev->addr_len != INFINIBAND_ALEN)
2057                 return -1;
2058         memcpy(eui, dev->dev_addr + 12, 8);
2059         eui[0] |= 2;
2060         return 0;
2061 }
2062
2063 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2064 {
2065         if (addr == 0)
2066                 return -1;
2067         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2068                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2069                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2070                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2071                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2072                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2073         eui[1] = 0;
2074         eui[2] = 0x5E;
2075         eui[3] = 0xFE;
2076         memcpy(eui + 4, &addr, 4);
2077         return 0;
2078 }
2079
2080 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2081 {
2082         if (dev->priv_flags & IFF_ISATAP)
2083                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2084         return -1;
2085 }
2086
2087 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2088 {
2089         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2090 }
2091
2092 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2093 {
2094         memcpy(eui, dev->perm_addr, 3);
2095         memcpy(eui + 5, dev->perm_addr + 3, 3);
2096         eui[3] = 0xFF;
2097         eui[4] = 0xFE;
2098         eui[0] ^= 2;
2099         return 0;
2100 }
2101
2102 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2103 {
2104         switch (dev->type) {
2105         case ARPHRD_ETHER:
2106         case ARPHRD_FDDI:
2107                 return addrconf_ifid_eui48(eui, dev);
2108         case ARPHRD_ARCNET:
2109                 return addrconf_ifid_arcnet(eui, dev);
2110         case ARPHRD_INFINIBAND:
2111                 return addrconf_ifid_infiniband(eui, dev);
2112         case ARPHRD_SIT:
2113                 return addrconf_ifid_sit(eui, dev);
2114         case ARPHRD_IPGRE:
2115                 return addrconf_ifid_gre(eui, dev);
2116         case ARPHRD_6LOWPAN:
2117                 return addrconf_ifid_eui64(eui, dev);
2118         case ARPHRD_IEEE1394:
2119                 return addrconf_ifid_ieee1394(eui, dev);
2120         case ARPHRD_TUNNEL6:
2121                 return addrconf_ifid_ip6tnl(eui, dev);
2122         }
2123         return -1;
2124 }
2125
2126 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2127 {
2128         int err = -1;
2129         struct inet6_ifaddr *ifp;
2130
2131         read_lock_bh(&idev->lock);
2132         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2133                 if (ifp->scope > IFA_LINK)
2134                         break;
2135                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2136                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2137                         err = 0;
2138                         break;
2139                 }
2140         }
2141         read_unlock_bh(&idev->lock);
2142         return err;
2143 }
2144
2145 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2146 static void ipv6_regen_rndid(struct inet6_dev *idev)
2147 {
2148 regen:
2149         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2150         idev->rndid[0] &= ~0x02;
2151
2152         /*
2153          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2154          * check if generated address is not inappropriate
2155          *
2156          *  - Reserved subnet anycast (RFC 2526)
2157          *      11111101 11....11 1xxxxxxx
2158          *  - ISATAP (RFC4214) 6.1
2159          *      00-00-5E-FE-xx-xx-xx-xx
2160          *  - value 0
2161          *  - XXX: already assigned to an address on the device
2162          */
2163         if (idev->rndid[0] == 0xfd &&
2164             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2165             (idev->rndid[7]&0x80))
2166                 goto regen;
2167         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2168                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2169                         goto regen;
2170                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2171                         goto regen;
2172         }
2173 }
2174
2175 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2176 {
2177         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2178                 ipv6_regen_rndid(idev);
2179 }
2180
2181 /*
2182  *      Add prefix route.
2183  */
2184
2185 static void
2186 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2187                       unsigned long expires, u32 flags)
2188 {
2189         struct fib6_config cfg = {
2190                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2191                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2192                 .fc_ifindex = dev->ifindex,
2193                 .fc_expires = expires,
2194                 .fc_dst_len = plen,
2195                 .fc_flags = RTF_UP | flags,
2196                 .fc_nlinfo.nl_net = dev_net(dev),
2197                 .fc_protocol = RTPROT_KERNEL,
2198         };
2199
2200         cfg.fc_dst = *pfx;
2201
2202         /* Prevent useless cloning on PtP SIT.
2203            This thing is done here expecting that the whole
2204            class of non-broadcast devices need not cloning.
2205          */
2206 #if IS_ENABLED(CONFIG_IPV6_SIT)
2207         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2208                 cfg.fc_flags |= RTF_NONEXTHOP;
2209 #endif
2210
2211         ip6_route_add(&cfg);
2212 }
2213
2214
2215 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2216                                                   int plen,
2217                                                   const struct net_device *dev,
2218                                                   u32 flags, u32 noflags)
2219 {
2220         struct fib6_node *fn;
2221         struct rt6_info *rt = NULL;
2222         struct fib6_table *table;
2223         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2224
2225         table = fib6_get_table(dev_net(dev), tb_id);
2226         if (!table)
2227                 return NULL;
2228
2229         read_lock_bh(&table->tb6_lock);
2230         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2231         if (!fn)
2232                 goto out;
2233
2234         noflags |= RTF_CACHE;
2235         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2236                 if (rt->dst.dev->ifindex != dev->ifindex)
2237                         continue;
2238                 if ((rt->rt6i_flags & flags) != flags)
2239                         continue;
2240                 if ((rt->rt6i_flags & noflags) != 0)
2241                         continue;
2242                 dst_hold(&rt->dst);
2243                 break;
2244         }
2245 out:
2246         read_unlock_bh(&table->tb6_lock);
2247         return rt;
2248 }
2249
2250
2251 /* Create "default" multicast route to the interface */
2252
2253 static void addrconf_add_mroute(struct net_device *dev)
2254 {
2255         struct fib6_config cfg = {
2256                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2257                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2258                 .fc_ifindex = dev->ifindex,
2259                 .fc_dst_len = 8,
2260                 .fc_flags = RTF_UP,
2261                 .fc_nlinfo.nl_net = dev_net(dev),
2262         };
2263
2264         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2265
2266         ip6_route_add(&cfg);
2267 }
2268
2269 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2270 {
2271         struct inet6_dev *idev;
2272
2273         ASSERT_RTNL();
2274
2275         idev = ipv6_find_idev(dev);
2276         if (!idev)
2277                 return ERR_PTR(-ENOBUFS);
2278
2279         if (idev->cnf.disable_ipv6)
2280                 return ERR_PTR(-EACCES);
2281
2282         /* Add default multicast route */
2283         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2284                 addrconf_add_mroute(dev);
2285
2286         return idev;
2287 }
2288
2289 static void manage_tempaddrs(struct inet6_dev *idev,
2290                              struct inet6_ifaddr *ifp,
2291                              __u32 valid_lft, __u32 prefered_lft,
2292                              bool create, unsigned long now)
2293 {
2294         u32 flags;
2295         struct inet6_ifaddr *ift;
2296
2297         read_lock_bh(&idev->lock);
2298         /* update all temporary addresses in the list */
2299         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2300                 int age, max_valid, max_prefered;
2301
2302                 if (ifp != ift->ifpub)
2303                         continue;
2304
2305                 /* RFC 4941 section 3.3:
2306                  * If a received option will extend the lifetime of a public
2307                  * address, the lifetimes of temporary addresses should
2308                  * be extended, subject to the overall constraint that no
2309                  * temporary addresses should ever remain "valid" or "preferred"
2310                  * for a time longer than (TEMP_VALID_LIFETIME) or
2311                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2312                  */
2313                 age = (now - ift->cstamp) / HZ;
2314                 max_valid = idev->cnf.temp_valid_lft - age;
2315                 if (max_valid < 0)
2316                         max_valid = 0;
2317
2318                 max_prefered = idev->cnf.temp_prefered_lft -
2319                                idev->cnf.max_desync_factor - age;
2320                 if (max_prefered < 0)
2321                         max_prefered = 0;
2322
2323                 if (valid_lft > max_valid)
2324                         valid_lft = max_valid;
2325
2326                 if (prefered_lft > max_prefered)
2327                         prefered_lft = max_prefered;
2328
2329                 spin_lock(&ift->lock);
2330                 flags = ift->flags;
2331                 ift->valid_lft = valid_lft;
2332                 ift->prefered_lft = prefered_lft;
2333                 ift->tstamp = now;
2334                 if (prefered_lft > 0)
2335                         ift->flags &= ~IFA_F_DEPRECATED;
2336
2337                 spin_unlock(&ift->lock);
2338                 if (!(flags&IFA_F_TENTATIVE))
2339                         ipv6_ifa_notify(0, ift);
2340         }
2341
2342         if ((create || list_empty(&idev->tempaddr_list)) &&
2343             idev->cnf.use_tempaddr > 0) {
2344                 /* When a new public address is created as described
2345                  * in [ADDRCONF], also create a new temporary address.
2346                  * Also create a temporary address if it's enabled but
2347                  * no temporary address currently exists.
2348                  */
2349                 read_unlock_bh(&idev->lock);
2350                 ipv6_create_tempaddr(ifp, NULL);
2351         } else {
2352                 read_unlock_bh(&idev->lock);
2353         }
2354 }
2355
2356 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2357 {
2358         return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2359                idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2360 }
2361
2362 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2363                                  const struct prefix_info *pinfo,
2364                                  struct inet6_dev *in6_dev,
2365                                  const struct in6_addr *addr, int addr_type,
2366                                  u32 addr_flags, bool sllao, bool tokenized,
2367                                  __u32 valid_lft, u32 prefered_lft)
2368 {
2369         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2370         int create = 0, update_lft = 0;
2371
2372         if (!ifp && valid_lft) {
2373                 int max_addresses = in6_dev->cnf.max_addresses;
2374
2375 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2376                 if (in6_dev->cnf.optimistic_dad &&
2377                     !net->ipv6.devconf_all->forwarding && sllao)
2378                         addr_flags |= IFA_F_OPTIMISTIC;
2379 #endif
2380
2381                 /* Do not allow to create too much of autoconfigured
2382                  * addresses; this would be too easy way to crash kernel.
2383                  */
2384                 if (!max_addresses ||
2385                     ipv6_count_addresses(in6_dev) < max_addresses)
2386                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2387                                             pinfo->prefix_len,
2388                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2389                                             addr_flags, valid_lft,
2390                                             prefered_lft);
2391
2392                 if (IS_ERR_OR_NULL(ifp))
2393                         return -1;
2394
2395                 update_lft = 0;
2396                 create = 1;
2397                 spin_lock_bh(&ifp->lock);
2398                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2399                 ifp->cstamp = jiffies;
2400                 ifp->tokenized = tokenized;
2401                 spin_unlock_bh(&ifp->lock);
2402                 addrconf_dad_start(ifp);
2403         }
2404
2405         if (ifp) {
2406                 u32 flags;
2407                 unsigned long now;
2408                 u32 stored_lft;
2409
2410                 /* update lifetime (RFC2462 5.5.3 e) */
2411                 spin_lock_bh(&ifp->lock);
2412                 now = jiffies;
2413                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2414                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2415                 else
2416                         stored_lft = 0;
2417                 if (!update_lft && !create && stored_lft) {
2418                         const u32 minimum_lft = min_t(u32,
2419                                 stored_lft, MIN_VALID_LIFETIME);
2420                         valid_lft = max(valid_lft, minimum_lft);
2421
2422                         /* RFC4862 Section 5.5.3e:
2423                          * "Note that the preferred lifetime of the
2424                          *  corresponding address is always reset to
2425                          *  the Preferred Lifetime in the received
2426                          *  Prefix Information option, regardless of
2427                          *  whether the valid lifetime is also reset or
2428                          *  ignored."
2429                          *
2430                          * So we should always update prefered_lft here.
2431                          */
2432                         update_lft = 1;
2433                 }
2434
2435                 if (update_lft) {
2436                         ifp->valid_lft = valid_lft;
2437                         ifp->prefered_lft = prefered_lft;
2438                         ifp->tstamp = now;
2439                         flags = ifp->flags;
2440                         ifp->flags &= ~IFA_F_DEPRECATED;
2441                         spin_unlock_bh(&ifp->lock);
2442
2443                         if (!(flags&IFA_F_TENTATIVE))
2444                                 ipv6_ifa_notify(0, ifp);
2445                 } else
2446                         spin_unlock_bh(&ifp->lock);
2447
2448                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2449                                  create, now);
2450
2451                 in6_ifa_put(ifp);
2452                 addrconf_verify();
2453         }
2454
2455         return 0;
2456 }
2457 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2458
2459 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2460 {
2461         struct prefix_info *pinfo;
2462         __u32 valid_lft;
2463         __u32 prefered_lft;
2464         int addr_type, err;
2465         u32 addr_flags = 0;
2466         struct inet6_dev *in6_dev;
2467         struct net *net = dev_net(dev);
2468
2469         pinfo = (struct prefix_info *) opt;
2470
2471         if (len < sizeof(struct prefix_info)) {
2472                 ADBG("addrconf: prefix option too short\n");
2473                 return;
2474         }
2475
2476         /*
2477          *      Validation checks ([ADDRCONF], page 19)
2478          */
2479
2480         addr_type = ipv6_addr_type(&pinfo->prefix);
2481
2482         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2483                 return;
2484
2485         valid_lft = ntohl(pinfo->valid);
2486         prefered_lft = ntohl(pinfo->prefered);
2487
2488         if (prefered_lft > valid_lft) {
2489                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2490                 return;
2491         }
2492
2493         in6_dev = in6_dev_get(dev);
2494
2495         if (!in6_dev) {
2496                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2497                                     dev->name);
2498                 return;
2499         }
2500
2501         /*
2502          *      Two things going on here:
2503          *      1) Add routes for on-link prefixes
2504          *      2) Configure prefixes with the auto flag set
2505          */
2506
2507         if (pinfo->onlink) {
2508                 struct rt6_info *rt;
2509                 unsigned long rt_expires;
2510
2511                 /* Avoid arithmetic overflow. Really, we could
2512                  * save rt_expires in seconds, likely valid_lft,
2513                  * but it would require division in fib gc, that it
2514                  * not good.
2515                  */
2516                 if (HZ > USER_HZ)
2517                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2518                 else
2519                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2520
2521                 if (addrconf_finite_timeout(rt_expires))
2522                         rt_expires *= HZ;
2523
2524                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2525                                                pinfo->prefix_len,
2526                                                dev,
2527                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2528                                                RTF_GATEWAY | RTF_DEFAULT);
2529
2530                 if (rt) {
2531                         /* Autoconf prefix route */
2532                         if (valid_lft == 0) {
2533                                 ip6_del_rt(rt);
2534                                 rt = NULL;
2535                         } else if (addrconf_finite_timeout(rt_expires)) {
2536                                 /* not infinity */
2537                                 rt6_set_expires(rt, jiffies + rt_expires);
2538                         } else {
2539                                 rt6_clean_expires(rt);
2540                         }
2541                 } else if (valid_lft) {
2542                         clock_t expires = 0;
2543                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2544                         if (addrconf_finite_timeout(rt_expires)) {
2545                                 /* not infinity */
2546                                 flags |= RTF_EXPIRES;
2547                                 expires = jiffies_to_clock_t(rt_expires);
2548                         }
2549                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2550                                               dev, expires, flags);
2551                 }
2552                 ip6_rt_put(rt);
2553         }
2554
2555         /* Try to figure out our local address for this prefix */
2556
2557         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2558                 struct in6_addr addr;
2559                 bool tokenized = false, dev_addr_generated = false;
2560
2561                 if (pinfo->prefix_len == 64) {
2562                         memcpy(&addr, &pinfo->prefix, 8);
2563
2564                         if (!ipv6_addr_any(&in6_dev->token)) {
2565                                 read_lock_bh(&in6_dev->lock);
2566                                 memcpy(addr.s6_addr + 8,
2567                                        in6_dev->token.s6_addr + 8, 8);
2568                                 read_unlock_bh(&in6_dev->lock);
2569                                 tokenized = true;
2570                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2571                                    !ipv6_generate_stable_address(&addr, 0,
2572                                                                  in6_dev)) {
2573                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2574                                 goto ok;
2575                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2576                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2577                                 goto put;
2578                         } else {
2579                                 dev_addr_generated = true;
2580                         }
2581                         goto ok;
2582                 }
2583                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2584                                     pinfo->prefix_len);
2585                 goto put;
2586
2587 ok:
2588                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2589                                                    &addr, addr_type,
2590                                                    addr_flags, sllao,
2591                                                    tokenized, valid_lft,
2592                                                    prefered_lft);
2593                 if (err)
2594                         goto put;
2595
2596                 /* Ignore error case here because previous prefix add addr was
2597                  * successful which will be notified.
2598                  */
2599                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2600                                               addr_type, addr_flags, sllao,
2601                                               tokenized, valid_lft,
2602                                               prefered_lft,
2603                                               dev_addr_generated);
2604         }
2605         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2606 put:
2607         in6_dev_put(in6_dev);
2608 }
2609
2610 /*
2611  *      Set destination address.
2612  *      Special case for SIT interfaces where we create a new "virtual"
2613  *      device.
2614  */
2615 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2616 {
2617         struct in6_ifreq ireq;
2618         struct net_device *dev;
2619         int err = -EINVAL;
2620
2621         rtnl_lock();
2622
2623         err = -EFAULT;
2624         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2625                 goto err_exit;
2626
2627         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2628
2629         err = -ENODEV;
2630         if (!dev)
2631                 goto err_exit;
2632
2633 #if IS_ENABLED(CONFIG_IPV6_SIT)
2634         if (dev->type == ARPHRD_SIT) {
2635                 const struct net_device_ops *ops = dev->netdev_ops;
2636                 struct ifreq ifr;
2637                 struct ip_tunnel_parm p;
2638
2639                 err = -EADDRNOTAVAIL;
2640                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2641                         goto err_exit;
2642
2643                 memset(&p, 0, sizeof(p));
2644                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2645                 p.iph.saddr = 0;
2646                 p.iph.version = 4;
2647                 p.iph.ihl = 5;
2648                 p.iph.protocol = IPPROTO_IPV6;
2649                 p.iph.ttl = 64;
2650                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2651
2652                 if (ops->ndo_do_ioctl) {
2653                         mm_segment_t oldfs = get_fs();
2654
2655                         set_fs(KERNEL_DS);
2656                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2657                         set_fs(oldfs);
2658                 } else
2659                         err = -EOPNOTSUPP;
2660
2661                 if (err == 0) {
2662                         err = -ENOBUFS;
2663                         dev = __dev_get_by_name(net, p.name);
2664                         if (!dev)
2665                                 goto err_exit;
2666                         err = dev_open(dev);
2667                 }
2668         }
2669 #endif
2670
2671 err_exit:
2672         rtnl_unlock();
2673         return err;
2674 }
2675
2676 static int ipv6_mc_config(struct sock *sk, bool join,
2677                           const struct in6_addr *addr, int ifindex)
2678 {
2679         int ret;
2680
2681         ASSERT_RTNL();
2682
2683         lock_sock(sk);
2684         if (join)
2685                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2686         else
2687                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2688         release_sock(sk);
2689
2690         return ret;
2691 }
2692
2693 /*
2694  *      Manual configuration of address on an interface
2695  */
2696 static int inet6_addr_add(struct net *net, int ifindex,
2697                           const struct in6_addr *pfx,
2698                           const struct in6_addr *peer_pfx,
2699                           unsigned int plen, __u32 ifa_flags,
2700                           __u32 prefered_lft, __u32 valid_lft)
2701 {
2702         struct inet6_ifaddr *ifp;
2703         struct inet6_dev *idev;
2704         struct net_device *dev;
2705         unsigned long timeout;
2706         clock_t expires;
2707         int scope;
2708         u32 flags;
2709
2710         ASSERT_RTNL();
2711
2712         if (plen > 128)
2713                 return -EINVAL;
2714
2715         /* check the lifetime */
2716         if (!valid_lft || prefered_lft > valid_lft)
2717                 return -EINVAL;
2718
2719         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2720                 return -EINVAL;
2721
2722         dev = __dev_get_by_index(net, ifindex);
2723         if (!dev)
2724                 return -ENODEV;
2725
2726         idev = addrconf_add_dev(dev);
2727         if (IS_ERR(idev))
2728                 return PTR_ERR(idev);
2729
2730         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2731                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2732                                          true, pfx, ifindex);
2733
2734                 if (ret < 0)
2735                         return ret;
2736         }
2737
2738         scope = ipv6_addr_scope(pfx);
2739
2740         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2741         if (addrconf_finite_timeout(timeout)) {
2742                 expires = jiffies_to_clock_t(timeout * HZ);
2743                 valid_lft = timeout;
2744                 flags = RTF_EXPIRES;
2745         } else {
2746                 expires = 0;
2747                 flags = 0;
2748                 ifa_flags |= IFA_F_PERMANENT;
2749         }
2750
2751         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2752         if (addrconf_finite_timeout(timeout)) {
2753                 if (timeout == 0)
2754                         ifa_flags |= IFA_F_DEPRECATED;
2755                 prefered_lft = timeout;
2756         }
2757
2758         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2759                             valid_lft, prefered_lft);
2760
2761         if (!IS_ERR(ifp)) {
2762                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2763                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2764                                               expires, flags);
2765                 }
2766
2767                 /*
2768                  * Note that section 3.1 of RFC 4429 indicates
2769                  * that the Optimistic flag should not be set for
2770                  * manually configured addresses
2771                  */
2772                 addrconf_dad_start(ifp);
2773                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2774                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2775                                          true, jiffies);
2776                 in6_ifa_put(ifp);
2777                 addrconf_verify_rtnl();
2778                 return 0;
2779         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2780                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2781                                false, pfx, ifindex);
2782         }
2783
2784         return PTR_ERR(ifp);
2785 }
2786
2787 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2788                           const struct in6_addr *pfx, unsigned int plen)
2789 {
2790         struct inet6_ifaddr *ifp;
2791         struct inet6_dev *idev;
2792         struct net_device *dev;
2793
2794         if (plen > 128)
2795                 return -EINVAL;
2796
2797         dev = __dev_get_by_index(net, ifindex);
2798         if (!dev)
2799                 return -ENODEV;
2800
2801         idev = __in6_dev_get(dev);
2802         if (!idev)
2803                 return -ENXIO;
2804
2805         read_lock_bh(&idev->lock);
2806         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2807                 if (ifp->prefix_len == plen &&
2808                     ipv6_addr_equal(pfx, &ifp->addr)) {
2809                         in6_ifa_hold(ifp);
2810                         read_unlock_bh(&idev->lock);
2811
2812                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2813                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2814                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2815                                                  jiffies);
2816                         ipv6_del_addr(ifp);
2817                         addrconf_verify_rtnl();
2818                         if (ipv6_addr_is_multicast(pfx)) {
2819                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2820                                                false, pfx, dev->ifindex);
2821                         }
2822                         return 0;
2823                 }
2824         }
2825         read_unlock_bh(&idev->lock);
2826         return -EADDRNOTAVAIL;
2827 }
2828
2829
2830 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2831 {
2832         struct in6_ifreq ireq;
2833         int err;
2834
2835         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2836                 return -EPERM;
2837
2838         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2839                 return -EFAULT;
2840
2841         rtnl_lock();
2842         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2843                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2844                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2845         rtnl_unlock();
2846         return err;
2847 }
2848
2849 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2850 {
2851         struct in6_ifreq ireq;
2852         int err;
2853
2854         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2855                 return -EPERM;
2856
2857         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2858                 return -EFAULT;
2859
2860         rtnl_lock();
2861         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2862                              ireq.ifr6_prefixlen);
2863         rtnl_unlock();
2864         return err;
2865 }
2866
2867 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2868                      int plen, int scope)
2869 {
2870         struct inet6_ifaddr *ifp;
2871
2872         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2873                             scope, IFA_F_PERMANENT,
2874                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2875         if (!IS_ERR(ifp)) {
2876                 spin_lock_bh(&ifp->lock);
2877                 ifp->flags &= ~IFA_F_TENTATIVE;
2878                 spin_unlock_bh(&ifp->lock);
2879                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2880                 in6_ifa_put(ifp);
2881         }
2882 }
2883
2884 #if IS_ENABLED(CONFIG_IPV6_SIT)
2885 static void sit_add_v4_addrs(struct inet6_dev *idev)
2886 {
2887         struct in6_addr addr;
2888         struct net_device *dev;
2889         struct net *net = dev_net(idev->dev);
2890         int scope, plen;
2891         u32 pflags = 0;
2892
2893         ASSERT_RTNL();
2894
2895         memset(&addr, 0, sizeof(struct in6_addr));
2896         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2897
2898         if (idev->dev->flags&IFF_POINTOPOINT) {
2899                 addr.s6_addr32[0] = htonl(0xfe800000);
2900                 scope = IFA_LINK;
2901                 plen = 64;
2902         } else {
2903                 scope = IPV6_ADDR_COMPATv4;
2904                 plen = 96;
2905                 pflags |= RTF_NONEXTHOP;
2906         }
2907
2908         if (addr.s6_addr32[3]) {
2909                 add_addr(idev, &addr, plen, scope);
2910                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2911                 return;
2912         }
2913
2914         for_each_netdev(net, dev) {
2915                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2916                 if (in_dev && (dev->flags & IFF_UP)) {
2917                         struct in_ifaddr *ifa;
2918
2919                         int flag = scope;
2920
2921                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2922
2923                                 addr.s6_addr32[3] = ifa->ifa_local;
2924
2925                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2926                                         continue;
2927                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2928                                         if (idev->dev->flags&IFF_POINTOPOINT)
2929                                                 continue;
2930                                         flag |= IFA_HOST;
2931                                 }
2932
2933                                 add_addr(idev, &addr, plen, flag);
2934                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2935                                                       pflags);
2936                         }
2937                 }
2938         }
2939 }
2940 #endif
2941
2942 static void init_loopback(struct net_device *dev)
2943 {
2944         struct inet6_dev  *idev;
2945         struct net_device *sp_dev;
2946         struct inet6_ifaddr *sp_ifa;
2947         struct rt6_info *sp_rt;
2948
2949         /* ::1 */
2950
2951         ASSERT_RTNL();
2952
2953         idev = ipv6_find_idev(dev);
2954         if (!idev) {
2955                 pr_debug("%s: add_dev failed\n", __func__);
2956                 return;
2957         }
2958
2959         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2960
2961         /* Add routes to other interface's IPv6 addresses */
2962         for_each_netdev(dev_net(dev), sp_dev) {
2963                 if (!strcmp(sp_dev->name, dev->name))
2964                         continue;
2965
2966                 idev = __in6_dev_get(sp_dev);
2967                 if (!idev)
2968                         continue;
2969
2970                 read_lock_bh(&idev->lock);
2971                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2972
2973                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2974                                 continue;
2975
2976                         if (sp_ifa->rt) {
2977                                 /* This dst has been added to garbage list when
2978                                  * lo device down, release this obsolete dst and
2979                                  * reallocate a new router for ifa.
2980                                  */
2981                                 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
2982                                         ip6_rt_put(sp_ifa->rt);
2983                                         sp_ifa->rt = NULL;
2984                                 } else {
2985                                         continue;
2986                                 }
2987                         }
2988
2989                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2990
2991                         /* Failure cases are ignored */
2992                         if (!IS_ERR(sp_rt)) {
2993                                 sp_ifa->rt = sp_rt;
2994                                 ip6_ins_rt(sp_rt);
2995                         }
2996                 }
2997                 read_unlock_bh(&idev->lock);
2998         }
2999 }
3000
3001 void addrconf_add_linklocal(struct inet6_dev *idev,
3002                             const struct in6_addr *addr, u32 flags)
3003 {
3004         struct inet6_ifaddr *ifp;
3005         u32 addr_flags = flags | IFA_F_PERMANENT;
3006
3007 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3008         if (idev->cnf.optimistic_dad &&
3009             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3010                 addr_flags |= IFA_F_OPTIMISTIC;
3011 #endif
3012
3013         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3014                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3015         if (!IS_ERR(ifp)) {
3016                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3017                 addrconf_dad_start(ifp);
3018                 in6_ifa_put(ifp);
3019         }
3020 }
3021 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3022
3023 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3024 {
3025         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3026                 return true;
3027
3028         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3029             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3030                 return true;
3031
3032         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3033             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3034                 return true;
3035
3036         return false;
3037 }
3038
3039 static int ipv6_generate_stable_address(struct in6_addr *address,
3040                                         u8 dad_count,
3041                                         const struct inet6_dev *idev)
3042 {
3043         static DEFINE_SPINLOCK(lock);
3044         static __u32 digest[SHA_DIGEST_WORDS];
3045         static __u32 workspace[SHA_WORKSPACE_WORDS];
3046
3047         static union {
3048                 char __data[SHA_MESSAGE_BYTES];
3049                 struct {
3050                         struct in6_addr secret;
3051                         __be32 prefix[2];
3052                         unsigned char hwaddr[MAX_ADDR_LEN];
3053                         u8 dad_count;
3054                 } __packed;
3055         } data;
3056
3057         struct in6_addr secret;
3058         struct in6_addr temp;
3059         struct net *net = dev_net(idev->dev);
3060
3061         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3062
3063         if (idev->cnf.stable_secret.initialized)
3064                 secret = idev->cnf.stable_secret.secret;
3065         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3066                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3067         else
3068                 return -1;
3069
3070 retry:
3071         spin_lock_bh(&lock);
3072
3073         sha_init(digest);
3074         memset(&data, 0, sizeof(data));
3075         memset(workspace, 0, sizeof(workspace));
3076         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3077         data.prefix[0] = address->s6_addr32[0];
3078         data.prefix[1] = address->s6_addr32[1];
3079         data.secret = secret;
3080         data.dad_count = dad_count;
3081
3082         sha_transform(digest, data.__data, workspace);
3083
3084         temp = *address;
3085         temp.s6_addr32[2] = (__force __be32)digest[0];
3086         temp.s6_addr32[3] = (__force __be32)digest[1];
3087
3088         spin_unlock_bh(&lock);
3089
3090         if (ipv6_reserved_interfaceid(temp)) {
3091                 dad_count++;
3092                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3093                         return -1;
3094                 goto retry;
3095         }
3096
3097         *address = temp;
3098         return 0;
3099 }
3100
3101 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3102 {
3103         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3104
3105         if (s->initialized)
3106                 return;
3107         s = &idev->cnf.stable_secret;
3108         get_random_bytes(&s->secret, sizeof(s->secret));
3109         s->initialized = true;
3110 }
3111
3112 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3113 {
3114         struct in6_addr addr;
3115
3116         /* no link local addresses on L3 master devices */
3117         if (netif_is_l3_master(idev->dev))
3118                 return;
3119
3120         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3121
3122         switch (idev->addr_gen_mode) {
3123         case IN6_ADDR_GEN_MODE_RANDOM:
3124                 ipv6_gen_mode_random_init(idev);
3125                 /* fallthrough */
3126         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3127                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3128                         addrconf_add_linklocal(idev, &addr,
3129                                                IFA_F_STABLE_PRIVACY);
3130                 else if (prefix_route)
3131                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3132                 break;
3133         case IN6_ADDR_GEN_MODE_EUI64:
3134                 /* addrconf_add_linklocal also adds a prefix_route and we
3135                  * only need to care about prefix routes if ipv6_generate_eui64
3136                  * couldn't generate one.
3137                  */
3138                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3139                         addrconf_add_linklocal(idev, &addr, 0);
3140                 else if (prefix_route)
3141                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3142                 break;
3143         case IN6_ADDR_GEN_MODE_NONE:
3144         default:
3145                 /* will not add any link local address */
3146                 break;
3147         }
3148 }
3149
3150 static void addrconf_dev_config(struct net_device *dev)
3151 {
3152         struct inet6_dev *idev;
3153
3154         ASSERT_RTNL();
3155
3156         if ((dev->type != ARPHRD_ETHER) &&
3157             (dev->type != ARPHRD_FDDI) &&
3158             (dev->type != ARPHRD_ARCNET) &&
3159             (dev->type != ARPHRD_INFINIBAND) &&
3160             (dev->type != ARPHRD_IEEE1394) &&
3161             (dev->type != ARPHRD_TUNNEL6) &&
3162             (dev->type != ARPHRD_6LOWPAN) &&
3163             (dev->type != ARPHRD_NONE)) {
3164                 /* Alas, we support only Ethernet autoconfiguration. */
3165                 return;
3166         }
3167
3168         idev = addrconf_add_dev(dev);
3169         if (IS_ERR(idev))
3170                 return;
3171
3172         /* this device type has no EUI support */
3173         if (dev->type == ARPHRD_NONE &&
3174             idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3175                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3176
3177         addrconf_addr_gen(idev, false);
3178 }
3179
3180 #if IS_ENABLED(CONFIG_IPV6_SIT)
3181 static void addrconf_sit_config(struct net_device *dev)
3182 {
3183         struct inet6_dev *idev;
3184
3185         ASSERT_RTNL();
3186
3187         /*
3188          * Configure the tunnel with one of our IPv4
3189          * addresses... we should configure all of
3190          * our v4 addrs in the tunnel
3191          */
3192
3193         idev = ipv6_find_idev(dev);
3194         if (!idev) {
3195                 pr_debug("%s: add_dev failed\n", __func__);
3196                 return;
3197         }
3198
3199         if (dev->priv_flags & IFF_ISATAP) {
3200                 addrconf_addr_gen(idev, false);
3201                 return;
3202         }
3203
3204         sit_add_v4_addrs(idev);
3205
3206         if (dev->flags&IFF_POINTOPOINT)
3207                 addrconf_add_mroute(dev);
3208 }
3209 #endif
3210
3211 #if IS_ENABLED(CONFIG_NET_IPGRE)
3212 static void addrconf_gre_config(struct net_device *dev)
3213 {
3214         struct inet6_dev *idev;
3215
3216         ASSERT_RTNL();
3217
3218         idev = ipv6_find_idev(dev);
3219         if (!idev) {
3220                 pr_debug("%s: add_dev failed\n", __func__);
3221                 return;
3222         }
3223
3224         addrconf_addr_gen(idev, true);
3225         if (dev->flags & IFF_POINTOPOINT)
3226                 addrconf_add_mroute(dev);
3227 }
3228 #endif
3229
3230 static int fixup_permanent_addr(struct inet6_dev *idev,
3231                                 struct inet6_ifaddr *ifp)
3232 {
3233         if (!ifp->rt) {
3234                 struct rt6_info *rt;
3235
3236                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3237                 if (unlikely(IS_ERR(rt)))
3238                         return PTR_ERR(rt);
3239
3240                 ifp->rt = rt;
3241         }
3242
3243         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3244                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3245                                       idev->dev, 0, 0);
3246         }
3247
3248         addrconf_dad_start(ifp);
3249
3250         return 0;
3251 }
3252
3253 static void addrconf_permanent_addr(struct net_device *dev)
3254 {
3255         struct inet6_ifaddr *ifp, *tmp;
3256         struct inet6_dev *idev;
3257
3258         idev = __in6_dev_get(dev);
3259         if (!idev)
3260                 return;
3261
3262         write_lock_bh(&idev->lock);
3263
3264         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3265                 if ((ifp->flags & IFA_F_PERMANENT) &&
3266                     fixup_permanent_addr(idev, ifp) < 0) {
3267                         write_unlock_bh(&idev->lock);
3268                         ipv6_del_addr(ifp);
3269                         write_lock_bh(&idev->lock);
3270
3271                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3272                                              idev->dev->name, &ifp->addr);
3273                 }
3274         }
3275
3276         write_unlock_bh(&idev->lock);
3277 }
3278
3279 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3280                            void *ptr)
3281 {
3282         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3283         struct netdev_notifier_changeupper_info *info;
3284         struct inet6_dev *idev = __in6_dev_get(dev);
3285         int run_pending = 0;
3286         int err;
3287
3288         switch (event) {
3289         case NETDEV_REGISTER:
3290                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3291                         idev = ipv6_add_dev(dev);
3292                         if (IS_ERR(idev))
3293                                 return notifier_from_errno(PTR_ERR(idev));
3294                 }
3295                 break;
3296
3297         case NETDEV_CHANGEMTU:
3298                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3299                 if (dev->mtu < IPV6_MIN_MTU) {
3300                         addrconf_ifdown(dev, 1);
3301                         break;
3302                 }
3303
3304                 if (idev) {
3305                         rt6_mtu_change(dev, dev->mtu);
3306                         idev->cnf.mtu6 = dev->mtu;
3307                         break;
3308                 }
3309
3310                 /* allocate new idev */
3311                 idev = ipv6_add_dev(dev);
3312                 if (IS_ERR(idev))
3313                         break;
3314
3315                 /* device is still not ready */
3316                 if (!(idev->if_flags & IF_READY))
3317                         break;
3318
3319                 run_pending = 1;
3320
3321                 /* fall through */
3322
3323         case NETDEV_UP:
3324         case NETDEV_CHANGE:
3325                 if (dev->flags & IFF_SLAVE)
3326                         break;
3327
3328                 if (idev && idev->cnf.disable_ipv6)
3329                         break;
3330
3331                 if (event == NETDEV_UP) {
3332                         /* restore routes for permanent addresses */
3333                         addrconf_permanent_addr(dev);
3334
3335                         if (!addrconf_qdisc_ok(dev)) {
3336                                 /* device is not ready yet. */
3337                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3338                                         dev->name);
3339                                 break;
3340                         }
3341
3342                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3343                                 idev = ipv6_add_dev(dev);
3344
3345                         if (!IS_ERR_OR_NULL(idev)) {
3346                                 idev->if_flags |= IF_READY;
3347                                 run_pending = 1;
3348                         }
3349                 } else if (event == NETDEV_CHANGE) {
3350                         if (!addrconf_qdisc_ok(dev)) {
3351                                 /* device is still not ready. */
3352                                 break;
3353                         }
3354
3355                         if (idev) {
3356                                 if (idev->if_flags & IF_READY)
3357                                         /* device is already configured. */
3358                                         break;
3359                                 idev->if_flags |= IF_READY;
3360                         }
3361
3362                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3363                                 dev->name);
3364
3365                         run_pending = 1;
3366                 }
3367
3368                 switch (dev->type) {
3369 #if IS_ENABLED(CONFIG_IPV6_SIT)
3370                 case ARPHRD_SIT:
3371                         addrconf_sit_config(dev);
3372                         break;
3373 #endif
3374 #if IS_ENABLED(CONFIG_NET_IPGRE)
3375                 case ARPHRD_IPGRE:
3376                         addrconf_gre_config(dev);
3377                         break;
3378 #endif
3379                 case ARPHRD_LOOPBACK:
3380                         init_loopback(dev);
3381                         break;
3382
3383                 default:
3384                         addrconf_dev_config(dev);
3385                         break;
3386                 }
3387
3388                 if (!IS_ERR_OR_NULL(idev)) {
3389                         if (run_pending)
3390                                 addrconf_dad_run(idev);
3391
3392                         /*
3393                          * If the MTU changed during the interface down,
3394                          * when the interface up, the changed MTU must be
3395                          * reflected in the idev as well as routers.
3396                          */
3397                         if (idev->cnf.mtu6 != dev->mtu &&
3398                             dev->mtu >= IPV6_MIN_MTU) {
3399                                 rt6_mtu_change(dev, dev->mtu);
3400                                 idev->cnf.mtu6 = dev->mtu;
3401                         }
3402                         idev->tstamp = jiffies;
3403                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3404
3405                         /*
3406                          * If the changed mtu during down is lower than
3407                          * IPV6_MIN_MTU stop IPv6 on this interface.
3408                          */
3409                         if (dev->mtu < IPV6_MIN_MTU)
3410                                 addrconf_ifdown(dev, 1);
3411                 }
3412                 break;
3413
3414         case NETDEV_DOWN:
3415         case NETDEV_UNREGISTER:
3416                 /*
3417                  *      Remove all addresses from this interface.
3418                  */
3419                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3420                 break;
3421
3422         case NETDEV_CHANGENAME:
3423                 if (idev) {
3424                         snmp6_unregister_dev(idev);
3425                         addrconf_sysctl_unregister(idev);
3426                         err = addrconf_sysctl_register(idev);
3427                         if (err)
3428                                 return notifier_from_errno(err);
3429                         err = snmp6_register_dev(idev);
3430                         if (err) {
3431                                 addrconf_sysctl_unregister(idev);
3432                                 return notifier_from_errno(err);
3433                         }
3434                 }
3435                 break;
3436
3437         case NETDEV_PRE_TYPE_CHANGE:
3438         case NETDEV_POST_TYPE_CHANGE:
3439                 if (idev)
3440                         addrconf_type_change(dev, event);
3441                 break;
3442
3443         case NETDEV_CHANGEUPPER:
3444                 info = ptr;
3445
3446                 /* flush all routes if dev is linked to or unlinked from
3447                  * an L3 master device (e.g., VRF)
3448                  */
3449                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3450                         addrconf_ifdown(dev, 0);
3451         }
3452
3453         return NOTIFY_OK;
3454 }
3455
3456 /*
3457  *      addrconf module should be notified of a device going up
3458  */
3459 static struct notifier_block ipv6_dev_notf = {
3460         .notifier_call = addrconf_notify,
3461 };
3462
3463 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3464 {
3465         struct inet6_dev *idev;
3466         ASSERT_RTNL();
3467
3468         idev = __in6_dev_get(dev);
3469
3470         if (event == NETDEV_POST_TYPE_CHANGE)
3471                 ipv6_mc_remap(idev);
3472         else if (event == NETDEV_PRE_TYPE_CHANGE)
3473                 ipv6_mc_unmap(idev);
3474 }
3475
3476 static bool addr_is_local(const struct in6_addr *addr)
3477 {
3478         return ipv6_addr_type(addr) &
3479                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3480 }
3481
3482 static int addrconf_ifdown(struct net_device *dev, int how)
3483 {
3484         struct net *net = dev_net(dev);
3485         struct inet6_dev *idev;
3486         struct inet6_ifaddr *ifa, *tmp;
3487         struct list_head del_list;
3488         int _keep_addr;
3489         bool keep_addr;
3490         int state, i;
3491
3492         ASSERT_RTNL();
3493
3494         rt6_ifdown(net, dev);
3495         neigh_ifdown(&nd_tbl, dev);
3496
3497         idev = __in6_dev_get(dev);
3498         if (!idev)
3499                 return -ENODEV;
3500
3501         /*
3502          * Step 1: remove reference to ipv6 device from parent device.
3503          *         Do not dev_put!
3504          */
3505         if (how) {
3506                 idev->dead = 1;
3507
3508                 /* protected by rtnl_lock */
3509                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3510
3511                 /* Step 1.5: remove snmp6 entry */
3512                 snmp6_unregister_dev(idev);
3513
3514         }
3515
3516         /* aggregate the system setting and interface setting */
3517         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3518         if (!_keep_addr)
3519                 _keep_addr = idev->cnf.keep_addr_on_down;
3520
3521         /* combine the user config with event to determine if permanent
3522          * addresses are to be removed from address hash table
3523          */
3524         keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3525
3526         /* Step 2: clear hash table */
3527         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3528                 struct hlist_head *h = &inet6_addr_lst[i];
3529
3530                 spin_lock_bh(&addrconf_hash_lock);
3531 restart:
3532                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3533                         if (ifa->idev == idev) {
3534                                 addrconf_del_dad_work(ifa);
3535                                 /* combined flag + permanent flag decide if
3536                                  * address is retained on a down event
3537                                  */
3538                                 if (!keep_addr ||
3539                                     !(ifa->flags & IFA_F_PERMANENT) ||
3540                                     addr_is_local(&ifa->addr)) {
3541                                         hlist_del_init_rcu(&ifa->addr_lst);
3542                                         goto restart;
3543                                 }
3544                         }
3545                 }
3546                 spin_unlock_bh(&addrconf_hash_lock);
3547         }
3548
3549         write_lock_bh(&idev->lock);
3550
3551         addrconf_del_rs_timer(idev);
3552
3553         /* Step 2: clear flags for stateless addrconf */
3554         if (!how)
3555                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3556
3557         /* Step 3: clear tempaddr list */
3558         while (!list_empty(&idev->tempaddr_list)) {
3559                 ifa = list_first_entry(&idev->tempaddr_list,
3560                                        struct inet6_ifaddr, tmp_list);
3561                 list_del(&ifa->tmp_list);
3562                 write_unlock_bh(&idev->lock);
3563                 spin_lock_bh(&ifa->lock);
3564
3565                 if (ifa->ifpub) {
3566                         in6_ifa_put(ifa->ifpub);
3567                         ifa->ifpub = NULL;
3568                 }
3569                 spin_unlock_bh(&ifa->lock);
3570                 in6_ifa_put(ifa);
3571                 write_lock_bh(&idev->lock);
3572         }
3573
3574         /* re-combine the user config with event to determine if permanent
3575          * addresses are to be removed from the interface list
3576          */
3577         keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3578
3579         INIT_LIST_HEAD(&del_list);
3580         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3581                 struct rt6_info *rt = NULL;
3582
3583                 addrconf_del_dad_work(ifa);
3584
3585                 write_unlock_bh(&idev->lock);
3586                 spin_lock_bh(&ifa->lock);
3587
3588                 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3589                     !addr_is_local(&ifa->addr)) {
3590                         /* set state to skip the notifier below */
3591                         state = INET6_IFADDR_STATE_DEAD;
3592                         ifa->state = 0;
3593                         if (!(ifa->flags & IFA_F_NODAD))
3594                                 ifa->flags |= IFA_F_TENTATIVE;
3595
3596                         rt = ifa->rt;
3597                         ifa->rt = NULL;
3598                 } else {
3599                         state = ifa->state;
3600                         ifa->state = INET6_IFADDR_STATE_DEAD;
3601
3602                         list_move(&ifa->if_list, &del_list);
3603                 }
3604
3605                 spin_unlock_bh(&ifa->lock);
3606
3607                 if (rt)
3608                         ip6_del_rt(rt);
3609
3610                 if (state != INET6_IFADDR_STATE_DEAD) {
3611                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3612                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3613                 } else {
3614                         if (idev->cnf.forwarding)
3615                                 addrconf_leave_anycast(ifa);
3616                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3617                 }
3618
3619                 write_lock_bh(&idev->lock);
3620         }
3621
3622         write_unlock_bh(&idev->lock);
3623
3624         /* now clean up addresses to be removed */
3625         while (!list_empty(&del_list)) {
3626                 ifa = list_first_entry(&del_list,
3627                                        struct inet6_ifaddr, if_list);
3628                 list_del(&ifa->if_list);
3629
3630                 in6_ifa_put(ifa);
3631         }
3632
3633         /* Step 5: Discard anycast and multicast list */
3634         if (how) {
3635                 ipv6_ac_destroy_dev(idev);
3636                 ipv6_mc_destroy_dev(idev);
3637         } else {
3638                 ipv6_mc_down(idev);
3639         }
3640
3641         idev->tstamp = jiffies;
3642
3643         /* Last: Shot the device (if unregistered) */
3644         if (how) {
3645                 addrconf_sysctl_unregister(idev);
3646                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3647                 neigh_ifdown(&nd_tbl, dev);
3648                 in6_dev_put(idev);
3649         }
3650         return 0;
3651 }
3652
3653 static void addrconf_rs_timer(unsigned long data)
3654 {
3655         struct inet6_dev *idev = (struct inet6_dev *)data;
3656         struct net_device *dev = idev->dev;
3657         struct in6_addr lladdr;
3658
3659         write_lock(&idev->lock);
3660         if (idev->dead || !(idev->if_flags & IF_READY))
3661                 goto out;
3662
3663         if (!ipv6_accept_ra(idev))
3664                 goto out;
3665
3666         /* Announcement received after solicitation was sent */
3667         if (idev->if_flags & IF_RA_RCVD)
3668                 goto out;
3669
3670         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3671                 write_unlock(&idev->lock);
3672                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3673                         ndisc_send_rs(dev, &lladdr,
3674                                       &in6addr_linklocal_allrouters);
3675                 else
3676                         goto put;
3677
3678                 write_lock(&idev->lock);
3679                 idev->rs_interval = rfc3315_s14_backoff_update(
3680                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3681                 /* The wait after the last probe can be shorter */
3682                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3683                                              idev->cnf.rtr_solicits) ?
3684                                       idev->cnf.rtr_solicit_delay :
3685                                       idev->rs_interval);
3686         } else {
3687                 /*
3688                  * Note: we do not support deprecated "all on-link"
3689                  * assumption any longer.
3690                  */
3691                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3692         }
3693
3694 out:
3695         write_unlock(&idev->lock);
3696 put:
3697         in6_dev_put(idev);
3698 }
3699
3700 /*
3701  *      Duplicate Address Detection
3702  */
3703 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3704 {
3705         unsigned long rand_num;
3706         struct inet6_dev *idev = ifp->idev;
3707
3708         if (ifp->flags & IFA_F_OPTIMISTIC)
3709                 rand_num = 0;
3710         else
3711                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3712
3713         ifp->dad_probes = idev->cnf.dad_transmits;
3714         addrconf_mod_dad_work(ifp, rand_num);
3715 }
3716
3717 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3718 {
3719         struct inet6_dev *idev = ifp->idev;
3720         struct net_device *dev = idev->dev;
3721         bool notify = false;
3722
3723         addrconf_join_solict(dev, &ifp->addr);
3724
3725         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3726
3727         read_lock_bh(&idev->lock);
3728         spin_lock(&ifp->lock);
3729         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3730                 goto out;
3731
3732         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3733             idev->cnf.accept_dad < 1 ||
3734             !(ifp->flags&IFA_F_TENTATIVE) ||
3735             ifp->flags & IFA_F_NODAD) {
3736                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3737                 spin_unlock(&ifp->lock);
3738                 read_unlock_bh(&idev->lock);
3739
3740                 addrconf_dad_completed(ifp);
3741                 return;
3742         }
3743
3744         if (!(idev->if_flags & IF_READY)) {
3745                 spin_unlock(&ifp->lock);
3746                 read_unlock_bh(&idev->lock);
3747                 /*
3748                  * If the device is not ready:
3749                  * - keep it tentative if it is a permanent address.
3750                  * - otherwise, kill it.
3751                  */
3752                 in6_ifa_hold(ifp);
3753                 addrconf_dad_stop(ifp, 0);
3754                 return;
3755         }
3756
3757         /*
3758          * Optimistic nodes can start receiving
3759          * Frames right away
3760          */
3761         if (ifp->flags & IFA_F_OPTIMISTIC) {
3762                 ip6_ins_rt(ifp->rt);
3763                 if (ipv6_use_optimistic_addr(idev)) {
3764                         /* Because optimistic nodes can use this address,
3765                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3766                          */
3767                         notify = true;
3768                 }
3769         }
3770
3771         addrconf_dad_kick(ifp);
3772 out:
3773         spin_unlock(&ifp->lock);
3774         read_unlock_bh(&idev->lock);
3775         if (notify)
3776                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3777 }
3778
3779 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3780 {
3781         bool begin_dad = false;
3782
3783         spin_lock_bh(&ifp->lock);
3784         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3785                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3786                 begin_dad = true;
3787         }
3788         spin_unlock_bh(&ifp->lock);
3789
3790         if (begin_dad)
3791                 addrconf_mod_dad_work(ifp, 0);
3792 }
3793
3794 static void addrconf_dad_work(struct work_struct *w)
3795 {
3796         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3797                                                 struct inet6_ifaddr,
3798                                                 dad_work);
3799         struct inet6_dev *idev = ifp->idev;
3800         struct in6_addr mcaddr;
3801         bool disable_ipv6 = false;
3802
3803         enum {
3804                 DAD_PROCESS,
3805                 DAD_BEGIN,
3806                 DAD_ABORT,
3807         } action = DAD_PROCESS;
3808
3809         rtnl_lock();
3810
3811         spin_lock_bh(&ifp->lock);
3812         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3813                 action = DAD_BEGIN;
3814                 ifp->state = INET6_IFADDR_STATE_DAD;
3815         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3816                 action = DAD_ABORT;
3817                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3818
3819                 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3820                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3821                         struct in6_addr addr;
3822
3823                         addr.s6_addr32[0] = htonl(0xfe800000);
3824                         addr.s6_addr32[1] = 0;
3825
3826                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3827                             ipv6_addr_equal(&ifp->addr, &addr)) {
3828                                 /* DAD failed for link-local based on MAC */
3829                                 idev->cnf.disable_ipv6 = 1;
3830
3831                                 pr_info("%s: IPv6 being disabled!\n",
3832                                         ifp->idev->dev->name);
3833                                 disable_ipv6 = true;
3834                         }
3835                 }
3836         }
3837         spin_unlock_bh(&ifp->lock);
3838
3839         if (action == DAD_BEGIN) {
3840                 addrconf_dad_begin(ifp);
3841                 goto out;
3842         } else if (action == DAD_ABORT) {
3843                 in6_ifa_hold(ifp);
3844                 addrconf_dad_stop(ifp, 1);
3845                 if (disable_ipv6)
3846                         addrconf_ifdown(idev->dev, 0);
3847                 goto out;
3848         }
3849
3850         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3851                 goto out;
3852
3853         write_lock_bh(&idev->lock);
3854         if (idev->dead || !(idev->if_flags & IF_READY)) {
3855                 write_unlock_bh(&idev->lock);
3856                 goto out;
3857         }
3858
3859         spin_lock(&ifp->lock);
3860         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3861                 spin_unlock(&ifp->lock);
3862                 write_unlock_bh(&idev->lock);
3863                 goto out;
3864         }
3865
3866         if (ifp->dad_probes == 0) {
3867                 /*
3868                  * DAD was successful
3869                  */
3870
3871                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3872                 spin_unlock(&ifp->lock);
3873                 write_unlock_bh(&idev->lock);
3874
3875                 addrconf_dad_completed(ifp);
3876
3877                 goto out;
3878         }
3879
3880         ifp->dad_probes--;
3881         addrconf_mod_dad_work(ifp,
3882                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3883         spin_unlock(&ifp->lock);
3884         write_unlock_bh(&idev->lock);
3885
3886         /* send a neighbour solicitation for our addr */
3887         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3888         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3889 out:
3890         in6_ifa_put(ifp);
3891         rtnl_unlock();
3892 }
3893
3894 /* ifp->idev must be at least read locked */
3895 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3896 {
3897         struct inet6_ifaddr *ifpiter;
3898         struct inet6_dev *idev = ifp->idev;
3899
3900         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3901                 if (ifpiter->scope > IFA_LINK)
3902                         break;
3903                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3904                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3905                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3906                     IFA_F_PERMANENT)
3907                         return false;
3908         }
3909         return true;
3910 }
3911
3912 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3913 {
3914         struct net_device *dev = ifp->idev->dev;
3915         struct in6_addr lladdr;
3916         bool send_rs, send_mld;
3917
3918         addrconf_del_dad_work(ifp);
3919
3920         /*
3921          *      Configure the address for reception. Now it is valid.
3922          */
3923
3924         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3925
3926         /* If added prefix is link local and we are prepared to process
3927            router advertisements, start sending router solicitations.
3928          */
3929
3930         read_lock_bh(&ifp->idev->lock);
3931         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3932         send_rs = send_mld &&
3933                   ipv6_accept_ra(ifp->idev) &&
3934                   ifp->idev->cnf.rtr_solicits != 0 &&
3935                   (dev->flags&IFF_LOOPBACK) == 0;
3936         read_unlock_bh(&ifp->idev->lock);
3937
3938         /* While dad is in progress mld report's source address is in6_addrany.
3939          * Resend with proper ll now.
3940          */
3941         if (send_mld)
3942                 ipv6_mc_dad_complete(ifp->idev);
3943
3944         if (send_rs) {
3945                 /*
3946                  *      If a host as already performed a random delay
3947                  *      [...] as part of DAD [...] there is no need
3948                  *      to delay again before sending the first RS
3949                  */
3950                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3951                         return;
3952                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3953
3954                 write_lock_bh(&ifp->idev->lock);
3955                 spin_lock(&ifp->lock);
3956                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
3957                         ifp->idev->cnf.rtr_solicit_interval);
3958                 ifp->idev->rs_probes = 1;
3959                 ifp->idev->if_flags |= IF_RS_SENT;
3960                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
3961                 spin_unlock(&ifp->lock);
3962                 write_unlock_bh(&ifp->idev->lock);
3963         }
3964 }
3965
3966 static void addrconf_dad_run(struct inet6_dev *idev)
3967 {
3968         struct inet6_ifaddr *ifp;
3969
3970         read_lock_bh(&idev->lock);
3971         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3972                 spin_lock(&ifp->lock);
3973                 if (ifp->flags & IFA_F_TENTATIVE &&
3974                     ifp->state == INET6_IFADDR_STATE_DAD)
3975                         addrconf_dad_kick(ifp);
3976                 spin_unlock(&ifp->lock);
3977         }
3978         read_unlock_bh(&idev->lock);
3979 }
3980
3981 #ifdef CONFIG_PROC_FS
3982 struct if6_iter_state {
3983         struct seq_net_private p;
3984         int bucket;
3985         int offset;
3986 };
3987
3988 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3989 {
3990         struct inet6_ifaddr *ifa = NULL;
3991         struct if6_iter_state *state = seq->private;
3992         struct net *net = seq_file_net(seq);
3993         int p = 0;
3994
3995         /* initial bucket if pos is 0 */
3996         if (pos == 0) {
3997                 state->bucket = 0;
3998                 state->offset = 0;
3999         }
4000
4001         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4002                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4003                                          addr_lst) {
4004                         if (!net_eq(dev_net(ifa->idev->dev), net))
4005                                 continue;
4006                         /* sync with offset */
4007                         if (p < state->offset) {
4008                                 p++;
4009                                 continue;
4010                         }
4011                         state->offset++;
4012                         return ifa;
4013                 }
4014
4015                 /* prepare for next bucket */
4016                 state->offset = 0;
4017                 p = 0;
4018         }
4019         return NULL;
4020 }
4021
4022 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4023                                          struct inet6_ifaddr *ifa)
4024 {
4025         struct if6_iter_state *state = seq->private;
4026         struct net *net = seq_file_net(seq);
4027
4028         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4029                 if (!net_eq(dev_net(ifa->idev->dev), net))
4030                         continue;
4031                 state->offset++;
4032                 return ifa;
4033         }
4034
4035         while (++state->bucket < IN6_ADDR_HSIZE) {
4036                 state->offset = 0;
4037                 hlist_for_each_entry_rcu_bh(ifa,
4038                                      &inet6_addr_lst[state->bucket], addr_lst) {
4039                         if (!net_eq(dev_net(ifa->idev->dev), net))
4040                                 continue;
4041                         state->offset++;
4042                         return ifa;
4043                 }
4044         }
4045
4046         return NULL;
4047 }
4048
4049 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4050         __acquires(rcu_bh)
4051 {
4052         rcu_read_lock_bh();
4053         return if6_get_first(seq, *pos);
4054 }
4055
4056 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4057 {
4058         struct inet6_ifaddr *ifa;
4059
4060         ifa = if6_get_next(seq, v);
4061         ++*pos;
4062         return ifa;
4063 }
4064
4065 static void if6_seq_stop(struct seq_file *seq, void *v)
4066         __releases(rcu_bh)
4067 {
4068         rcu_read_unlock_bh();
4069 }
4070
4071 static int if6_seq_show(struct seq_file *seq, void *v)
4072 {
4073         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4074         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4075                    &ifp->addr,
4076                    ifp->idev->dev->ifindex,
4077                    ifp->prefix_len,
4078                    ifp->scope,
4079                    (u8) ifp->flags,
4080                    ifp->idev->dev->name);
4081         return 0;
4082 }
4083
4084 static const struct seq_operations if6_seq_ops = {
4085         .start  = if6_seq_start,
4086         .next   = if6_seq_next,
4087         .show   = if6_seq_show,
4088         .stop   = if6_seq_stop,
4089 };
4090
4091 static int if6_seq_open(struct inode *inode, struct file *file)
4092 {
4093         return seq_open_net(inode, file, &if6_seq_ops,
4094                             sizeof(struct if6_iter_state));
4095 }
4096
4097 static const struct file_operations if6_fops = {
4098         .owner          = THIS_MODULE,
4099         .open           = if6_seq_open,
4100         .read           = seq_read,
4101         .llseek         = seq_lseek,
4102         .release        = seq_release_net,
4103 };
4104
4105 static int __net_init if6_proc_net_init(struct net *net)
4106 {
4107         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4108                 return -ENOMEM;
4109         return 0;
4110 }
4111
4112 static void __net_exit if6_proc_net_exit(struct net *net)
4113 {
4114         remove_proc_entry("if_inet6", net->proc_net);
4115 }
4116
4117 static struct pernet_operations if6_proc_net_ops = {
4118         .init = if6_proc_net_init,
4119         .exit = if6_proc_net_exit,
4120 };
4121
4122 int __init if6_proc_init(void)
4123 {
4124         return register_pernet_subsys(&if6_proc_net_ops);
4125 }
4126
4127 void if6_proc_exit(void)
4128 {
4129         unregister_pernet_subsys(&if6_proc_net_ops);
4130 }
4131 #endif  /* CONFIG_PROC_FS */
4132
4133 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4134 /* Check if address is a home address configured on any interface. */
4135 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4136 {
4137         int ret = 0;
4138         struct inet6_ifaddr *ifp = NULL;
4139         unsigned int hash = inet6_addr_hash(addr);
4140
4141         rcu_read_lock_bh();
4142         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4143                 if (!net_eq(dev_net(ifp->idev->dev), net))
4144                         continue;
4145                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4146                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4147                         ret = 1;
4148                         break;
4149                 }
4150         }
4151         rcu_read_unlock_bh();
4152         return ret;
4153 }
4154 #endif
4155
4156 /*
4157  *      Periodic address status verification
4158  */
4159
4160 static void addrconf_verify_rtnl(void)
4161 {
4162         unsigned long now, next, next_sec, next_sched;
4163         struct inet6_ifaddr *ifp;
4164         int i;
4165
4166         ASSERT_RTNL();
4167
4168         rcu_read_lock_bh();
4169         now = jiffies;
4170         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4171
4172         cancel_delayed_work(&addr_chk_work);
4173
4174         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4175 restart:
4176                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4177                         unsigned long age;
4178
4179                         /* When setting preferred_lft to a value not zero or
4180                          * infinity, while valid_lft is infinity
4181                          * IFA_F_PERMANENT has a non-infinity life time.
4182                          */
4183                         if ((ifp->flags & IFA_F_PERMANENT) &&
4184                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4185                                 continue;
4186
4187                         spin_lock(&ifp->lock);
4188                         /* We try to batch several events at once. */
4189                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4190
4191                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4192                             age >= ifp->valid_lft) {
4193                                 spin_unlock(&ifp->lock);
4194                                 in6_ifa_hold(ifp);
4195                                 ipv6_del_addr(ifp);
4196                                 goto restart;
4197                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4198                                 spin_unlock(&ifp->lock);
4199                                 continue;
4200                         } else if (age >= ifp->prefered_lft) {
4201                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4202                                 int deprecate = 0;
4203
4204                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4205                                         deprecate = 1;
4206                                         ifp->flags |= IFA_F_DEPRECATED;
4207                                 }
4208
4209                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4210                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4211                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4212
4213                                 spin_unlock(&ifp->lock);
4214
4215                                 if (deprecate) {
4216                                         in6_ifa_hold(ifp);
4217
4218                                         ipv6_ifa_notify(0, ifp);
4219                                         in6_ifa_put(ifp);
4220                                         goto restart;
4221                                 }
4222                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4223                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4224                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4225                                         ifp->idev->cnf.dad_transmits *
4226                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4227
4228                                 if (age >= ifp->prefered_lft - regen_advance) {
4229                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4230                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4231                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4232                                         if (!ifp->regen_count && ifpub) {
4233                                                 ifp->regen_count++;
4234                                                 in6_ifa_hold(ifp);
4235                                                 in6_ifa_hold(ifpub);
4236                                                 spin_unlock(&ifp->lock);
4237
4238                                                 spin_lock(&ifpub->lock);
4239                                                 ifpub->regen_count = 0;
4240                                                 spin_unlock(&ifpub->lock);
4241                                                 ipv6_create_tempaddr(ifpub, ifp);
4242                                                 in6_ifa_put(ifpub);
4243                                                 in6_ifa_put(ifp);
4244                                                 goto restart;
4245                                         }
4246                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4247                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4248                                 spin_unlock(&ifp->lock);
4249                         } else {
4250                                 /* ifp->prefered_lft <= ifp->valid_lft */
4251                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4252                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4253                                 spin_unlock(&ifp->lock);
4254                         }
4255                 }
4256         }
4257
4258         next_sec = round_jiffies_up(next);
4259         next_sched = next;
4260
4261         /* If rounded timeout is accurate enough, accept it. */
4262         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4263                 next_sched = next_sec;
4264
4265         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4266         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4267                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4268
4269         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4270               now, next, next_sec, next_sched);
4271         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4272         rcu_read_unlock_bh();
4273 }
4274
4275 static void addrconf_verify_work(struct work_struct *w)
4276 {
4277         rtnl_lock();
4278         addrconf_verify_rtnl();
4279         rtnl_unlock();
4280 }
4281
4282 static void addrconf_verify(void)
4283 {
4284         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4285 }
4286
4287 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4288                                      struct in6_addr **peer_pfx)
4289 {
4290         struct in6_addr *pfx = NULL;
4291
4292         *peer_pfx = NULL;
4293
4294         if (addr)
4295                 pfx = nla_data(addr);
4296
4297         if (local) {
4298                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4299                         *peer_pfx = pfx;
4300                 pfx = nla_data(local);
4301         }
4302
4303         return pfx;
4304 }
4305
4306 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4307         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4308         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4309         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4310         [IFA_FLAGS]             = { .len = sizeof(u32) },
4311 };
4312
4313 static int
4314 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4315 {
4316         struct net *net = sock_net(skb->sk);
4317         struct ifaddrmsg *ifm;
4318         struct nlattr *tb[IFA_MAX+1];
4319         struct in6_addr *pfx, *peer_pfx;
4320         u32 ifa_flags;
4321         int err;
4322
4323         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4324         if (err < 0)
4325                 return err;
4326
4327         ifm = nlmsg_data(nlh);
4328         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4329         if (!pfx)
4330                 return -EINVAL;
4331
4332         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4333
4334         /* We ignore other flags so far. */
4335         ifa_flags &= IFA_F_MANAGETEMPADDR;
4336
4337         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4338                               ifm->ifa_prefixlen);
4339 }
4340
4341 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4342                              u32 prefered_lft, u32 valid_lft)
4343 {
4344         u32 flags;
4345         clock_t expires;
4346         unsigned long timeout;
4347         bool was_managetempaddr;
4348         bool had_prefixroute;
4349
4350         ASSERT_RTNL();
4351
4352         if (!valid_lft || (prefered_lft > valid_lft))
4353                 return -EINVAL;
4354
4355         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4356             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4357                 return -EINVAL;
4358
4359         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4360         if (addrconf_finite_timeout(timeout)) {
4361                 expires = jiffies_to_clock_t(timeout * HZ);
4362                 valid_lft = timeout;
4363                 flags = RTF_EXPIRES;
4364         } else {
4365                 expires = 0;
4366                 flags = 0;
4367                 ifa_flags |= IFA_F_PERMANENT;
4368         }
4369
4370         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4371         if (addrconf_finite_timeout(timeout)) {
4372                 if (timeout == 0)
4373                         ifa_flags |= IFA_F_DEPRECATED;
4374                 prefered_lft = timeout;
4375         }
4376
4377         spin_lock_bh(&ifp->lock);
4378         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4379         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4380                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4381         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4382                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4383                         IFA_F_NOPREFIXROUTE);
4384         ifp->flags |= ifa_flags;
4385         ifp->tstamp = jiffies;
4386         ifp->valid_lft = valid_lft;
4387         ifp->prefered_lft = prefered_lft;
4388
4389         spin_unlock_bh(&ifp->lock);
4390         if (!(ifp->flags&IFA_F_TENTATIVE))
4391                 ipv6_ifa_notify(0, ifp);
4392
4393         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4394                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4395                                       expires, flags);
4396         } else if (had_prefixroute) {
4397                 enum cleanup_prefix_rt_t action;
4398                 unsigned long rt_expires;
4399
4400                 write_lock_bh(&ifp->idev->lock);
4401                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4402                 write_unlock_bh(&ifp->idev->lock);
4403
4404                 if (action != CLEANUP_PREFIX_RT_NOP) {
4405                         cleanup_prefix_route(ifp, rt_expires,
4406                                 action == CLEANUP_PREFIX_RT_DEL);
4407                 }
4408         }
4409
4410         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4411                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4412                         valid_lft = prefered_lft = 0;
4413                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4414                                  !was_managetempaddr, jiffies);
4415         }
4416
4417         addrconf_verify_rtnl();
4418
4419         return 0;
4420 }
4421
4422 static int
4423 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4424 {
4425         struct net *net = sock_net(skb->sk);
4426         struct ifaddrmsg *ifm;
4427         struct nlattr *tb[IFA_MAX+1];
4428         struct in6_addr *pfx, *peer_pfx;
4429         struct inet6_ifaddr *ifa;
4430         struct net_device *dev;
4431         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4432         u32 ifa_flags;
4433         int err;
4434
4435         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4436         if (err < 0)
4437                 return err;
4438
4439         ifm = nlmsg_data(nlh);
4440         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4441         if (!pfx)
4442                 return -EINVAL;
4443
4444         if (tb[IFA_CACHEINFO]) {
4445                 struct ifa_cacheinfo *ci;
4446
4447                 ci = nla_data(tb[IFA_CACHEINFO]);
4448                 valid_lft = ci->ifa_valid;
4449                 preferred_lft = ci->ifa_prefered;
4450         } else {
4451                 preferred_lft = INFINITY_LIFE_TIME;
4452                 valid_lft = INFINITY_LIFE_TIME;
4453         }
4454
4455         dev =  __dev_get_by_index(net, ifm->ifa_index);
4456         if (!dev)
4457                 return -ENODEV;
4458
4459         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4460
4461         /* We ignore other flags so far. */
4462         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4463                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4464
4465         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4466         if (!ifa) {
4467                 /*
4468                  * It would be best to check for !NLM_F_CREATE here but
4469                  * userspace already relies on not having to provide this.
4470                  */
4471                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4472                                       ifm->ifa_prefixlen, ifa_flags,
4473                                       preferred_lft, valid_lft);
4474         }
4475
4476         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4477             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4478                 err = -EEXIST;
4479         else
4480                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4481
4482         in6_ifa_put(ifa);
4483
4484         return err;
4485 }
4486
4487 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4488                           u8 scope, int ifindex)
4489 {
4490         struct ifaddrmsg *ifm;
4491
4492         ifm = nlmsg_data(nlh);
4493         ifm->ifa_family = AF_INET6;
4494         ifm->ifa_prefixlen = prefixlen;
4495         ifm->ifa_flags = flags;
4496         ifm->ifa_scope = scope;
4497         ifm->ifa_index = ifindex;
4498 }
4499
4500 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4501                          unsigned long tstamp, u32 preferred, u32 valid)
4502 {
4503         struct ifa_cacheinfo ci;
4504
4505         ci.cstamp = cstamp_delta(cstamp);
4506         ci.tstamp = cstamp_delta(tstamp);
4507         ci.ifa_prefered = preferred;
4508         ci.ifa_valid = valid;
4509
4510         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4511 }
4512
4513 static inline int rt_scope(int ifa_scope)
4514 {
4515         if (ifa_scope & IFA_HOST)
4516                 return RT_SCOPE_HOST;
4517         else if (ifa_scope & IFA_LINK)
4518                 return RT_SCOPE_LINK;
4519         else if (ifa_scope & IFA_SITE)
4520                 return RT_SCOPE_SITE;
4521         else
4522                 return RT_SCOPE_UNIVERSE;
4523 }
4524
4525 static inline int inet6_ifaddr_msgsize(void)
4526 {
4527         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4528                + nla_total_size(16) /* IFA_LOCAL */
4529                + nla_total_size(16) /* IFA_ADDRESS */
4530                + nla_total_size(sizeof(struct ifa_cacheinfo))
4531                + nla_total_size(4)  /* IFA_FLAGS */;
4532 }
4533
4534 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4535                              u32 portid, u32 seq, int event, unsigned int flags)
4536 {
4537         struct nlmsghdr  *nlh;
4538         u32 preferred, valid;
4539
4540         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4541         if (!nlh)
4542                 return -EMSGSIZE;
4543
4544         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4545                       ifa->idev->dev->ifindex);
4546
4547         if (!((ifa->flags&IFA_F_PERMANENT) &&
4548               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4549                 preferred = ifa->prefered_lft;
4550                 valid = ifa->valid_lft;
4551                 if (preferred != INFINITY_LIFE_TIME) {
4552                         long tval = (jiffies - ifa->tstamp)/HZ;
4553                         if (preferred > tval)
4554                                 preferred -= tval;
4555                         else
4556                                 preferred = 0;
4557                         if (valid != INFINITY_LIFE_TIME) {
4558                                 if (valid > tval)
4559                                         valid -= tval;
4560                                 else
4561                                         valid = 0;
4562                         }
4563                 }
4564         } else {
4565                 preferred = INFINITY_LIFE_TIME;
4566                 valid = INFINITY_LIFE_TIME;
4567         }
4568
4569         if (!ipv6_addr_any(&ifa->peer_addr)) {
4570                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4571                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4572                         goto error;
4573         } else
4574                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4575                         goto error;
4576
4577         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4578                 goto error;
4579
4580         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4581                 goto error;
4582
4583         nlmsg_end(skb, nlh);
4584         return 0;
4585
4586 error:
4587         nlmsg_cancel(skb, nlh);
4588         return -EMSGSIZE;
4589 }
4590
4591 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4592                                 u32 portid, u32 seq, int event, u16 flags)
4593 {
4594         struct nlmsghdr  *nlh;
4595         u8 scope = RT_SCOPE_UNIVERSE;
4596         int ifindex = ifmca->idev->dev->ifindex;
4597
4598         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4599                 scope = RT_SCOPE_SITE;
4600
4601         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4602         if (!nlh)
4603                 return -EMSGSIZE;
4604
4605         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4606         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4607             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4608                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4609                 nlmsg_cancel(skb, nlh);
4610                 return -EMSGSIZE;
4611         }
4612
4613         nlmsg_end(skb, nlh);
4614         return 0;
4615 }
4616
4617 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4618                                 u32 portid, u32 seq, int event, unsigned int flags)
4619 {
4620         struct nlmsghdr  *nlh;
4621         u8 scope = RT_SCOPE_UNIVERSE;
4622         int ifindex = ifaca->aca_idev->dev->ifindex;
4623
4624         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4625                 scope = RT_SCOPE_SITE;
4626
4627         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4628         if (!nlh)
4629                 return -EMSGSIZE;
4630
4631         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4632         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4633             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4634                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4635                 nlmsg_cancel(skb, nlh);
4636                 return -EMSGSIZE;
4637         }
4638
4639         nlmsg_end(skb, nlh);
4640         return 0;
4641 }
4642
4643 enum addr_type_t {
4644         UNICAST_ADDR,
4645         MULTICAST_ADDR,
4646         ANYCAST_ADDR,
4647 };
4648
4649 /* called with rcu_read_lock() */
4650 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4651                           struct netlink_callback *cb, enum addr_type_t type,
4652                           int s_ip_idx, int *p_ip_idx)
4653 {
4654         struct ifmcaddr6 *ifmca;
4655         struct ifacaddr6 *ifaca;
4656         int err = 1;
4657         int ip_idx = *p_ip_idx;
4658
4659         read_lock_bh(&idev->lock);
4660         switch (type) {
4661         case UNICAST_ADDR: {
4662                 struct inet6_ifaddr *ifa;
4663
4664                 /* unicast address incl. temp addr */
4665                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4666                         if (++ip_idx < s_ip_idx)
4667                                 continue;
4668                         err = inet6_fill_ifaddr(skb, ifa,
4669                                                 NETLINK_CB(cb->skb).portid,
4670                                                 cb->nlh->nlmsg_seq,
4671                                                 RTM_NEWADDR,
4672                                                 NLM_F_MULTI);
4673                         if (err < 0)
4674                                 break;
4675                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4676                 }
4677                 break;
4678         }
4679         case MULTICAST_ADDR:
4680                 /* multicast address */
4681                 for (ifmca = idev->mc_list; ifmca;
4682                      ifmca = ifmca->next, ip_idx++) {
4683                         if (ip_idx < s_ip_idx)
4684                                 continue;
4685                         err = inet6_fill_ifmcaddr(skb, ifmca,
4686                                                   NETLINK_CB(cb->skb).portid,
4687                                                   cb->nlh->nlmsg_seq,
4688                                                   RTM_GETMULTICAST,
4689                                                   NLM_F_MULTI);
4690                         if (err < 0)
4691                                 break;
4692                 }
4693                 break;
4694         case ANYCAST_ADDR:
4695                 /* anycast address */
4696                 for (ifaca = idev->ac_list; ifaca;
4697                      ifaca = ifaca->aca_next, ip_idx++) {
4698                         if (ip_idx < s_ip_idx)
4699                                 continue;
4700                         err = inet6_fill_ifacaddr(skb, ifaca,
4701                                                   NETLINK_CB(cb->skb).portid,
4702                                                   cb->nlh->nlmsg_seq,
4703                                                   RTM_GETANYCAST,
4704                                                   NLM_F_MULTI);
4705                         if (err < 0)
4706                                 break;
4707                 }
4708                 break;
4709         default:
4710                 break;
4711         }
4712         read_unlock_bh(&idev->lock);
4713         *p_ip_idx = ip_idx;
4714         return err;
4715 }
4716
4717 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4718                            enum addr_type_t type)
4719 {
4720         struct net *net = sock_net(skb->sk);
4721         int h, s_h;
4722         int idx, ip_idx;
4723         int s_idx, s_ip_idx;
4724         struct net_device *dev;
4725         struct inet6_dev *idev;
4726         struct hlist_head *head;
4727
4728         s_h = cb->args[0];
4729         s_idx = idx = cb->args[1];
4730         s_ip_idx = ip_idx = cb->args[2];
4731
4732         rcu_read_lock();
4733         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4734         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4735                 idx = 0;
4736                 head = &net->dev_index_head[h];
4737                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4738                         if (idx < s_idx)
4739                                 goto cont;
4740                         if (h > s_h || idx > s_idx)
4741                                 s_ip_idx = 0;
4742                         ip_idx = 0;
4743                         idev = __in6_dev_get(dev);
4744                         if (!idev)
4745                                 goto cont;
4746
4747                         if (in6_dump_addrs(idev, skb, cb, type,
4748                                            s_ip_idx, &ip_idx) < 0)
4749                                 goto done;
4750 cont:
4751                         idx++;
4752                 }
4753         }
4754 done:
4755         rcu_read_unlock();
4756         cb->args[0] = h;
4757         cb->args[1] = idx;
4758         cb->args[2] = ip_idx;
4759
4760         return skb->len;
4761 }
4762
4763 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4764 {
4765         enum addr_type_t type = UNICAST_ADDR;
4766
4767         return inet6_dump_addr(skb, cb, type);
4768 }
4769
4770 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4771 {
4772         enum addr_type_t type = MULTICAST_ADDR;
4773
4774         return inet6_dump_addr(skb, cb, type);
4775 }
4776
4777
4778 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4779 {
4780         enum addr_type_t type = ANYCAST_ADDR;
4781
4782         return inet6_dump_addr(skb, cb, type);
4783 }
4784
4785 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4786 {
4787         struct net *net = sock_net(in_skb->sk);
4788         struct ifaddrmsg *ifm;
4789         struct nlattr *tb[IFA_MAX+1];
4790         struct in6_addr *addr = NULL, *peer;
4791         struct net_device *dev = NULL;
4792         struct inet6_ifaddr *ifa;
4793         struct sk_buff *skb;
4794         int err;
4795
4796         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4797         if (err < 0)
4798                 goto errout;
4799
4800         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4801         if (!addr) {
4802                 err = -EINVAL;
4803                 goto errout;
4804         }
4805
4806         ifm = nlmsg_data(nlh);
4807         if (ifm->ifa_index)
4808                 dev = __dev_get_by_index(net, ifm->ifa_index);
4809
4810         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4811         if (!ifa) {
4812                 err = -EADDRNOTAVAIL;
4813                 goto errout;
4814         }
4815
4816         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4817         if (!skb) {
4818                 err = -ENOBUFS;
4819                 goto errout_ifa;
4820         }
4821
4822         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4823                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4824         if (err < 0) {
4825                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4826                 WARN_ON(err == -EMSGSIZE);
4827                 kfree_skb(skb);
4828                 goto errout_ifa;
4829         }
4830         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4831 errout_ifa:
4832         in6_ifa_put(ifa);
4833 errout:
4834         return err;
4835 }
4836
4837 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4838 {
4839         struct sk_buff *skb;
4840         struct net *net = dev_net(ifa->idev->dev);
4841         int err = -ENOBUFS;
4842
4843         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4844         if (!skb)
4845                 goto errout;
4846
4847         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4848         if (err < 0) {
4849                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4850                 WARN_ON(err == -EMSGSIZE);
4851                 kfree_skb(skb);
4852                 goto errout;
4853         }
4854         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4855         return;
4856 errout:
4857         if (err < 0)
4858                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4859 }
4860
4861 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4862                                 __s32 *array, int bytes)
4863 {
4864         BUG_ON(bytes < (DEVCONF_MAX * 4));
4865
4866         memset(array, 0, bytes);
4867         array[DEVCONF_FORWARDING] = cnf->forwarding;
4868         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4869         array[DEVCONF_MTU6] = cnf->mtu6;
4870         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4871         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4872         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4873         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4874         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4875         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4876                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4877         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4878                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4879         array[DEVCONF_RTR_SOLICIT_DELAY] =
4880                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4881         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4882         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4883                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4884         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4885                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4886         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4887         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4888         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4889         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4890         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4891         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4892         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4893         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4894         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4895 #ifdef CONFIG_IPV6_ROUTER_PREF
4896         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4897         array[DEVCONF_RTR_PROBE_INTERVAL] =
4898                 jiffies_to_msecs(cnf->rtr_probe_interval);
4899 #ifdef CONFIG_IPV6_ROUTE_INFO
4900         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4901 #endif
4902 #endif
4903         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4904         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4905 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4906         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4907         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4908 #endif
4909 #ifdef CONFIG_IPV6_MROUTE
4910         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4911 #endif
4912         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4913         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4914         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4915         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4916         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4917         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4918         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4919         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4920         /* we omit DEVCONF_STABLE_SECRET for now */
4921         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4922         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4923         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4924         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4925 }
4926
4927 static inline size_t inet6_ifla6_size(void)
4928 {
4929         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4930              + nla_total_size(sizeof(struct ifla_cacheinfo))
4931              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4932              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4933              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4934              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4935 }
4936
4937 static inline size_t inet6_if_nlmsg_size(void)
4938 {
4939         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4940                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4941                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4942                + nla_total_size(4) /* IFLA_MTU */
4943                + nla_total_size(4) /* IFLA_LINK */
4944                + nla_total_size(1) /* IFLA_OPERSTATE */
4945                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4946 }
4947
4948 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4949                                         int bytes)
4950 {
4951         int i;
4952         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
4953         BUG_ON(pad < 0);
4954
4955         /* Use put_unaligned() because stats may not be aligned for u64. */
4956         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
4957         for (i = 1; i < ICMP6_MIB_MAX; i++)
4958                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4959
4960         memset(&stats[ICMP6_MIB_MAX], 0, pad);
4961 }
4962
4963 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4964                                         int bytes, size_t syncpoff)
4965 {
4966         int i, c;
4967         u64 buff[IPSTATS_MIB_MAX];
4968         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4969
4970         BUG_ON(pad < 0);
4971
4972         memset(buff, 0, sizeof(buff));
4973         buff[0] = IPSTATS_MIB_MAX;
4974
4975         for_each_possible_cpu(c) {
4976                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4977                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4978         }
4979
4980         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4981         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4982 }
4983
4984 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4985                              int bytes)
4986 {
4987         switch (attrtype) {
4988         case IFLA_INET6_STATS:
4989                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4990                                      offsetof(struct ipstats_mib, syncp));
4991                 break;
4992         case IFLA_INET6_ICMP6STATS:
4993                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
4994                 break;
4995         }
4996 }
4997
4998 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4999                                   u32 ext_filter_mask)
5000 {
5001         struct nlattr *nla;
5002         struct ifla_cacheinfo ci;
5003
5004         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5005                 goto nla_put_failure;
5006         ci.max_reasm_len = IPV6_MAXPLEN;
5007         ci.tstamp = cstamp_delta(idev->tstamp);
5008         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5009         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5010         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5011                 goto nla_put_failure;
5012         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5013         if (!nla)
5014                 goto nla_put_failure;
5015         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5016
5017         /* XXX - MC not implemented */
5018
5019         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5020                 return 0;
5021
5022         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5023         if (!nla)
5024                 goto nla_put_failure;
5025         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5026
5027         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5028         if (!nla)
5029                 goto nla_put_failure;
5030         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5031
5032         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5033         if (!nla)
5034                 goto nla_put_failure;
5035
5036         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5037                 goto nla_put_failure;
5038
5039         read_lock_bh(&idev->lock);
5040         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5041         read_unlock_bh(&idev->lock);
5042
5043         return 0;
5044
5045 nla_put_failure:
5046         return -EMSGSIZE;
5047 }
5048
5049 static size_t inet6_get_link_af_size(const struct net_device *dev,
5050                                      u32 ext_filter_mask)
5051 {
5052         if (!__in6_dev_get(dev))
5053                 return 0;
5054
5055         return inet6_ifla6_size();
5056 }
5057
5058 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5059                               u32 ext_filter_mask)
5060 {
5061         struct inet6_dev *idev = __in6_dev_get(dev);
5062
5063         if (!idev)
5064                 return -ENODATA;
5065
5066         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5067                 return -EMSGSIZE;
5068
5069         return 0;
5070 }
5071
5072 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5073 {
5074         struct inet6_ifaddr *ifp;
5075         struct net_device *dev = idev->dev;
5076         bool clear_token, update_rs = false;
5077         struct in6_addr ll_addr;
5078
5079         ASSERT_RTNL();
5080
5081         if (!token)
5082                 return -EINVAL;
5083         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5084                 return -EINVAL;
5085         if (!ipv6_accept_ra(idev))
5086                 return -EINVAL;
5087         if (idev->cnf.rtr_solicits == 0)
5088                 return -EINVAL;
5089
5090         write_lock_bh(&idev->lock);
5091
5092         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5093         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5094
5095         write_unlock_bh(&idev->lock);
5096
5097         clear_token = ipv6_addr_any(token);
5098         if (clear_token)
5099                 goto update_lft;
5100
5101         if (!idev->dead && (idev->if_flags & IF_READY) &&
5102             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5103                              IFA_F_OPTIMISTIC)) {
5104                 /* If we're not ready, then normal ifup will take care
5105                  * of this. Otherwise, we need to request our rs here.
5106                  */
5107                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5108                 update_rs = true;
5109         }
5110
5111 update_lft:
5112         write_lock_bh(&idev->lock);
5113
5114         if (update_rs) {
5115                 idev->if_flags |= IF_RS_SENT;
5116                 idev->rs_interval = rfc3315_s14_backoff_init(
5117                         idev->cnf.rtr_solicit_interval);
5118                 idev->rs_probes = 1;
5119                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5120         }
5121
5122         /* Well, that's kinda nasty ... */
5123         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5124                 spin_lock(&ifp->lock);
5125                 if (ifp->tokenized) {
5126                         ifp->valid_lft = 0;
5127                         ifp->prefered_lft = 0;
5128                 }
5129                 spin_unlock(&ifp->lock);
5130         }
5131
5132         write_unlock_bh(&idev->lock);
5133         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5134         addrconf_verify_rtnl();
5135         return 0;
5136 }
5137
5138 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5139         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5140         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5141 };
5142
5143 static int inet6_validate_link_af(const struct net_device *dev,
5144                                   const struct nlattr *nla)
5145 {
5146         struct nlattr *tb[IFLA_INET6_MAX + 1];
5147
5148         if (dev && !__in6_dev_get(dev))
5149                 return -EAFNOSUPPORT;
5150
5151         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5152 }
5153
5154 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5155 {
5156         int err = -EINVAL;
5157         struct inet6_dev *idev = __in6_dev_get(dev);
5158         struct nlattr *tb[IFLA_INET6_MAX + 1];
5159
5160         if (!idev)
5161                 return -EAFNOSUPPORT;
5162
5163         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5164                 BUG();
5165
5166         if (tb[IFLA_INET6_TOKEN]) {
5167                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5168                 if (err)
5169                         return err;
5170         }
5171
5172         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5173                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5174
5175                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5176                     mode != IN6_ADDR_GEN_MODE_NONE &&
5177                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5178                     mode != IN6_ADDR_GEN_MODE_RANDOM)
5179                         return -EINVAL;
5180
5181                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5182                     !idev->cnf.stable_secret.initialized &&
5183                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5184                         return -EINVAL;
5185
5186                 idev->addr_gen_mode = mode;
5187                 err = 0;
5188         }
5189
5190         return err;
5191 }
5192
5193 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5194                              u32 portid, u32 seq, int event, unsigned int flags)
5195 {
5196         struct net_device *dev = idev->dev;
5197         struct ifinfomsg *hdr;
5198         struct nlmsghdr *nlh;
5199         void *protoinfo;
5200
5201         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5202         if (!nlh)
5203                 return -EMSGSIZE;
5204
5205         hdr = nlmsg_data(nlh);
5206         hdr->ifi_family = AF_INET6;
5207         hdr->__ifi_pad = 0;
5208         hdr->ifi_type = dev->type;
5209         hdr->ifi_index = dev->ifindex;
5210         hdr->ifi_flags = dev_get_flags(dev);
5211         hdr->ifi_change = 0;
5212
5213         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5214             (dev->addr_len &&
5215              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5216             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5217             (dev->ifindex != dev_get_iflink(dev) &&
5218              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5219             nla_put_u8(skb, IFLA_OPERSTATE,
5220                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5221                 goto nla_put_failure;
5222         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5223         if (!protoinfo)
5224                 goto nla_put_failure;
5225
5226         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5227                 goto nla_put_failure;
5228
5229         nla_nest_end(skb, protoinfo);
5230         nlmsg_end(skb, nlh);
5231         return 0;
5232
5233 nla_put_failure:
5234         nlmsg_cancel(skb, nlh);
5235         return -EMSGSIZE;
5236 }
5237
5238 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5239 {
5240         struct net *net = sock_net(skb->sk);
5241         int h, s_h;
5242         int idx = 0, s_idx;
5243         struct net_device *dev;
5244         struct inet6_dev *idev;
5245         struct hlist_head *head;
5246
5247         s_h = cb->args[0];
5248         s_idx = cb->args[1];
5249
5250         rcu_read_lock();
5251         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5252                 idx = 0;
5253                 head = &net->dev_index_head[h];
5254                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5255                         if (idx < s_idx)
5256                                 goto cont;
5257                         idev = __in6_dev_get(dev);
5258                         if (!idev)
5259                                 goto cont;
5260                         if (inet6_fill_ifinfo(skb, idev,
5261                                               NETLINK_CB(cb->skb).portid,
5262                                               cb->nlh->nlmsg_seq,
5263                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5264                                 goto out;
5265 cont:
5266                         idx++;
5267                 }
5268         }
5269 out:
5270         rcu_read_unlock();
5271         cb->args[1] = idx;
5272         cb->args[0] = h;
5273
5274         return skb->len;
5275 }
5276
5277 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5278 {
5279         struct sk_buff *skb;
5280         struct net *net = dev_net(idev->dev);
5281         int err = -ENOBUFS;
5282
5283         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5284         if (!skb)
5285                 goto errout;
5286
5287         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5288         if (err < 0) {
5289                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5290                 WARN_ON(err == -EMSGSIZE);
5291                 kfree_skb(skb);
5292                 goto errout;
5293         }
5294         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5295         return;
5296 errout:
5297         if (err < 0)
5298                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5299 }
5300
5301 static inline size_t inet6_prefix_nlmsg_size(void)
5302 {
5303         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5304                + nla_total_size(sizeof(struct in6_addr))
5305                + nla_total_size(sizeof(struct prefix_cacheinfo));
5306 }
5307
5308 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5309                              struct prefix_info *pinfo, u32 portid, u32 seq,
5310                              int event, unsigned int flags)
5311 {
5312         struct prefixmsg *pmsg;
5313         struct nlmsghdr *nlh;
5314         struct prefix_cacheinfo ci;
5315
5316         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5317         if (!nlh)
5318                 return -EMSGSIZE;
5319
5320         pmsg = nlmsg_data(nlh);
5321         pmsg->prefix_family = AF_INET6;
5322         pmsg->prefix_pad1 = 0;
5323         pmsg->prefix_pad2 = 0;
5324         pmsg->prefix_ifindex = idev->dev->ifindex;
5325         pmsg->prefix_len = pinfo->prefix_len;
5326         pmsg->prefix_type = pinfo->type;
5327         pmsg->prefix_pad3 = 0;
5328         pmsg->prefix_flags = 0;
5329         if (pinfo->onlink)
5330                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5331         if (pinfo->autoconf)
5332                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5333
5334         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5335                 goto nla_put_failure;
5336         ci.preferred_time = ntohl(pinfo->prefered);
5337         ci.valid_time = ntohl(pinfo->valid);
5338         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5339                 goto nla_put_failure;
5340         nlmsg_end(skb, nlh);
5341         return 0;
5342
5343 nla_put_failure:
5344         nlmsg_cancel(skb, nlh);
5345         return -EMSGSIZE;
5346 }
5347
5348 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5349                          struct prefix_info *pinfo)
5350 {
5351         struct sk_buff *skb;
5352         struct net *net = dev_net(idev->dev);
5353         int err = -ENOBUFS;
5354
5355         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5356         if (!skb)
5357                 goto errout;
5358
5359         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5360         if (err < 0) {
5361                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5362                 WARN_ON(err == -EMSGSIZE);
5363                 kfree_skb(skb);
5364                 goto errout;
5365         }
5366         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5367         return;
5368 errout:
5369         if (err < 0)
5370                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5371 }
5372
5373 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5374 {
5375         struct net *net = dev_net(ifp->idev->dev);
5376
5377         if (event)
5378                 ASSERT_RTNL();
5379
5380         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5381
5382         switch (event) {
5383         case RTM_NEWADDR:
5384                 /*
5385                  * If the address was optimistic
5386                  * we inserted the route at the start of
5387                  * our DAD process, so we don't need
5388                  * to do it again
5389                  */
5390                 if (!(ifp->rt->rt6i_node))
5391                         ip6_ins_rt(ifp->rt);
5392                 if (ifp->idev->cnf.forwarding)
5393                         addrconf_join_anycast(ifp);
5394                 if (!ipv6_addr_any(&ifp->peer_addr))
5395                         addrconf_prefix_route(&ifp->peer_addr, 128,
5396                                               ifp->idev->dev, 0, 0);
5397                 break;
5398         case RTM_DELADDR:
5399                 if (ifp->idev->cnf.forwarding)
5400                         addrconf_leave_anycast(ifp);
5401                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5402                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5403                         struct rt6_info *rt;
5404
5405                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5406                                                        ifp->idev->dev, 0, 0);
5407                         if (rt)
5408                                 ip6_del_rt(rt);
5409                 }
5410                 if (ifp->rt) {
5411                         dst_hold(&ifp->rt->dst);
5412                         ip6_del_rt(ifp->rt);
5413                 }
5414                 rt_genid_bump_ipv6(net);
5415                 break;
5416         }
5417         atomic_inc(&net->ipv6.dev_addr_genid);
5418 }
5419
5420 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5421 {
5422         rcu_read_lock_bh();
5423         if (likely(ifp->idev->dead == 0))
5424                 __ipv6_ifa_notify(event, ifp);
5425         rcu_read_unlock_bh();
5426 }
5427
5428 #ifdef CONFIG_SYSCTL
5429
5430 static
5431 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5432                            void __user *buffer, size_t *lenp, loff_t *ppos)
5433 {
5434         int *valp = ctl->data;
5435         int val = *valp;
5436         loff_t pos = *ppos;
5437         struct ctl_table lctl;
5438         int ret;
5439
5440         /*
5441          * ctl->data points to idev->cnf.forwarding, we should
5442          * not modify it until we get the rtnl lock.
5443          */
5444         lctl = *ctl;
5445         lctl.data = &val;
5446
5447         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5448
5449         if (write)
5450                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5451         if (ret)
5452                 *ppos = pos;
5453         return ret;
5454 }
5455
5456 static
5457 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5458                         void __user *buffer, size_t *lenp, loff_t *ppos)
5459 {
5460         struct inet6_dev *idev = ctl->extra1;
5461         int min_mtu = IPV6_MIN_MTU;
5462         struct ctl_table lctl;
5463
5464         lctl = *ctl;
5465         lctl.extra1 = &min_mtu;
5466         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5467
5468         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5469 }
5470
5471 static void dev_disable_change(struct inet6_dev *idev)
5472 {
5473         struct netdev_notifier_info info;
5474
5475         if (!idev || !idev->dev)
5476                 return;
5477
5478         netdev_notifier_info_init(&info, idev->dev);
5479         if (idev->cnf.disable_ipv6)
5480                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5481         else
5482                 addrconf_notify(NULL, NETDEV_UP, &info);
5483 }
5484
5485 static void addrconf_disable_change(struct net *net, __s32 newf)
5486 {
5487         struct net_device *dev;
5488         struct inet6_dev *idev;
5489
5490         rcu_read_lock();
5491         for_each_netdev_rcu(net, dev) {
5492                 idev = __in6_dev_get(dev);
5493                 if (idev) {
5494                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5495                         idev->cnf.disable_ipv6 = newf;
5496                         if (changed)
5497                                 dev_disable_change(idev);
5498                 }
5499         }
5500         rcu_read_unlock();
5501 }
5502
5503 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5504 {
5505         struct net *net;
5506         int old;
5507
5508         if (!rtnl_trylock())
5509                 return restart_syscall();
5510
5511         net = (struct net *)table->extra2;
5512         old = *p;
5513         *p = newf;
5514
5515         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5516                 rtnl_unlock();
5517                 return 0;
5518         }
5519
5520         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5521                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5522                 addrconf_disable_change(net, newf);
5523         } else if ((!newf) ^ (!old))
5524                 dev_disable_change((struct inet6_dev *)table->extra1);
5525
5526         rtnl_unlock();
5527         return 0;
5528 }
5529
5530 static
5531 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5532                             void __user *buffer, size_t *lenp, loff_t *ppos)
5533 {
5534         int *valp = ctl->data;
5535         int val = *valp;
5536         loff_t pos = *ppos;
5537         struct ctl_table lctl;
5538         int ret;
5539
5540         /*
5541          * ctl->data points to idev->cnf.disable_ipv6, we should
5542          * not modify it until we get the rtnl lock.
5543          */
5544         lctl = *ctl;
5545         lctl.data = &val;
5546
5547         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5548
5549         if (write)
5550                 ret = addrconf_disable_ipv6(ctl, valp, val);
5551         if (ret)
5552                 *ppos = pos;
5553         return ret;
5554 }
5555
5556 static
5557 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5558                               void __user *buffer, size_t *lenp, loff_t *ppos)
5559 {
5560         int *valp = ctl->data;
5561         int ret;
5562         int old, new;
5563
5564         old = *valp;
5565         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5566         new = *valp;
5567
5568         if (write && old != new) {
5569                 struct net *net = ctl->extra2;
5570
5571                 if (!rtnl_trylock())
5572                         return restart_syscall();
5573
5574                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5575                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5576                                                      NETCONFA_IFINDEX_DEFAULT,
5577                                                      net->ipv6.devconf_dflt);
5578                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5579                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5580                                                      NETCONFA_IFINDEX_ALL,
5581                                                      net->ipv6.devconf_all);
5582                 else {
5583                         struct inet6_dev *idev = ctl->extra1;
5584
5585                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5586                                                      idev->dev->ifindex,
5587                                                      &idev->cnf);
5588                 }
5589                 rtnl_unlock();
5590         }
5591
5592         return ret;
5593 }
5594
5595 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5596                                          void __user *buffer, size_t *lenp,
5597                                          loff_t *ppos)
5598 {
5599         int err;
5600         struct in6_addr addr;
5601         char str[IPV6_MAX_STRLEN];
5602         struct ctl_table lctl = *ctl;
5603         struct net *net = ctl->extra2;
5604         struct ipv6_stable_secret *secret = ctl->data;
5605
5606         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5607                 return -EIO;
5608
5609         lctl.maxlen = IPV6_MAX_STRLEN;
5610         lctl.data = str;
5611
5612         if (!rtnl_trylock())
5613                 return restart_syscall();
5614
5615         if (!write && !secret->initialized) {
5616                 err = -EIO;
5617                 goto out;
5618         }
5619
5620         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5621         if (err >= sizeof(str)) {
5622                 err = -EIO;
5623                 goto out;
5624         }
5625
5626         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5627         if (err || !write)
5628                 goto out;
5629
5630         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5631                 err = -EIO;
5632                 goto out;
5633         }
5634
5635         secret->initialized = true;
5636         secret->secret = addr;
5637
5638         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5639                 struct net_device *dev;
5640
5641                 for_each_netdev(net, dev) {
5642                         struct inet6_dev *idev = __in6_dev_get(dev);
5643
5644                         if (idev) {
5645                                 idev->addr_gen_mode =
5646                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5647                         }
5648                 }
5649         } else {
5650                 struct inet6_dev *idev = ctl->extra1;
5651
5652                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5653         }
5654
5655 out:
5656         rtnl_unlock();
5657
5658         return err;
5659 }
5660
5661 static
5662 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5663                                                 int write,
5664                                                 void __user *buffer,
5665                                                 size_t *lenp,
5666                                                 loff_t *ppos)
5667 {
5668         int *valp = ctl->data;
5669         int val = *valp;
5670         loff_t pos = *ppos;
5671         struct ctl_table lctl;
5672         int ret;
5673
5674         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5675          * we should not modify it until we get the rtnl lock.
5676          */
5677         lctl = *ctl;
5678         lctl.data = &val;
5679
5680         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5681
5682         if (write)
5683                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5684         if (ret)
5685                 *ppos = pos;
5686         return ret;
5687 }
5688
5689 static int minus_one = -1;
5690 static const int one = 1;
5691 static const int two_five_five = 255;
5692
5693 static const struct ctl_table addrconf_sysctl[] = {
5694         {
5695                 .procname       = "forwarding",
5696                 .data           = &ipv6_devconf.forwarding,
5697                 .maxlen         = sizeof(int),
5698                 .mode           = 0644,
5699                 .proc_handler   = addrconf_sysctl_forward,
5700         },
5701         {
5702                 .procname       = "hop_limit",
5703                 .data           = &ipv6_devconf.hop_limit,
5704                 .maxlen         = sizeof(int),
5705                 .mode           = 0644,
5706                 .proc_handler   = proc_dointvec_minmax,
5707                 .extra1         = (void *)&one,
5708                 .extra2         = (void *)&two_five_five,
5709         },
5710         {
5711                 .procname       = "mtu",
5712                 .data           = &ipv6_devconf.mtu6,
5713                 .maxlen         = sizeof(int),
5714                 .mode           = 0644,
5715                 .proc_handler   = addrconf_sysctl_mtu,
5716         },
5717         {
5718                 .procname       = "accept_ra",
5719                 .data           = &ipv6_devconf.accept_ra,
5720                 .maxlen         = sizeof(int),
5721                 .mode           = 0644,
5722                 .proc_handler   = proc_dointvec,
5723         },
5724         {
5725                 .procname       = "accept_redirects",
5726                 .data           = &ipv6_devconf.accept_redirects,
5727                 .maxlen         = sizeof(int),
5728                 .mode           = 0644,
5729                 .proc_handler   = proc_dointvec,
5730         },
5731         {
5732                 .procname       = "autoconf",
5733                 .data           = &ipv6_devconf.autoconf,
5734                 .maxlen         = sizeof(int),
5735                 .mode           = 0644,
5736                 .proc_handler   = proc_dointvec,
5737         },
5738         {
5739                 .procname       = "dad_transmits",
5740                 .data           = &ipv6_devconf.dad_transmits,
5741                 .maxlen         = sizeof(int),
5742                 .mode           = 0644,
5743                 .proc_handler   = proc_dointvec,
5744         },
5745         {
5746                 .procname       = "router_solicitations",
5747                 .data           = &ipv6_devconf.rtr_solicits,
5748                 .maxlen         = sizeof(int),
5749                 .mode           = 0644,
5750                 .proc_handler   = proc_dointvec_minmax,
5751                 .extra1         = &minus_one,
5752         },
5753         {
5754                 .procname       = "router_solicitation_interval",
5755                 .data           = &ipv6_devconf.rtr_solicit_interval,
5756                 .maxlen         = sizeof(int),
5757                 .mode           = 0644,
5758                 .proc_handler   = proc_dointvec_jiffies,
5759         },
5760         {
5761                 .procname       = "router_solicitation_max_interval",
5762                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
5763                 .maxlen         = sizeof(int),
5764                 .mode           = 0644,
5765                 .proc_handler   = proc_dointvec_jiffies,
5766         },
5767         {
5768                 .procname       = "router_solicitation_delay",
5769                 .data           = &ipv6_devconf.rtr_solicit_delay,
5770                 .maxlen         = sizeof(int),
5771                 .mode           = 0644,
5772                 .proc_handler   = proc_dointvec_jiffies,
5773         },
5774         {
5775                 .procname       = "force_mld_version",
5776                 .data           = &ipv6_devconf.force_mld_version,
5777                 .maxlen         = sizeof(int),
5778                 .mode           = 0644,
5779                 .proc_handler   = proc_dointvec,
5780         },
5781         {
5782                 .procname       = "mldv1_unsolicited_report_interval",
5783                 .data           =
5784                         &ipv6_devconf.mldv1_unsolicited_report_interval,
5785                 .maxlen         = sizeof(int),
5786                 .mode           = 0644,
5787                 .proc_handler   = proc_dointvec_ms_jiffies,
5788         },
5789         {
5790                 .procname       = "mldv2_unsolicited_report_interval",
5791                 .data           =
5792                         &ipv6_devconf.mldv2_unsolicited_report_interval,
5793                 .maxlen         = sizeof(int),
5794                 .mode           = 0644,
5795                 .proc_handler   = proc_dointvec_ms_jiffies,
5796         },
5797         {
5798                 .procname       = "use_tempaddr",
5799                 .data           = &ipv6_devconf.use_tempaddr,
5800                 .maxlen         = sizeof(int),
5801                 .mode           = 0644,
5802                 .proc_handler   = proc_dointvec,
5803         },
5804         {
5805                 .procname       = "temp_valid_lft",
5806                 .data           = &ipv6_devconf.temp_valid_lft,
5807                 .maxlen         = sizeof(int),
5808                 .mode           = 0644,
5809                 .proc_handler   = proc_dointvec,
5810         },
5811         {
5812                 .procname       = "temp_prefered_lft",
5813                 .data           = &ipv6_devconf.temp_prefered_lft,
5814                 .maxlen         = sizeof(int),
5815                 .mode           = 0644,
5816                 .proc_handler   = proc_dointvec,
5817         },
5818         {
5819                 .procname       = "regen_max_retry",
5820                 .data           = &ipv6_devconf.regen_max_retry,
5821                 .maxlen         = sizeof(int),
5822                 .mode           = 0644,
5823                 .proc_handler   = proc_dointvec,
5824         },
5825         {
5826                 .procname       = "max_desync_factor",
5827                 .data           = &ipv6_devconf.max_desync_factor,
5828                 .maxlen         = sizeof(int),
5829                 .mode           = 0644,
5830                 .proc_handler   = proc_dointvec,
5831         },
5832         {
5833                 .procname       = "max_addresses",
5834                 .data           = &ipv6_devconf.max_addresses,
5835                 .maxlen         = sizeof(int),
5836                 .mode           = 0644,
5837                 .proc_handler   = proc_dointvec,
5838         },
5839         {
5840                 .procname       = "accept_ra_defrtr",
5841                 .data           = &ipv6_devconf.accept_ra_defrtr,
5842                 .maxlen         = sizeof(int),
5843                 .mode           = 0644,
5844                 .proc_handler   = proc_dointvec,
5845         },
5846         {
5847                 .procname       = "accept_ra_min_hop_limit",
5848                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5849                 .maxlen         = sizeof(int),
5850                 .mode           = 0644,
5851                 .proc_handler   = proc_dointvec,
5852         },
5853         {
5854                 .procname       = "accept_ra_pinfo",
5855                 .data           = &ipv6_devconf.accept_ra_pinfo,
5856                 .maxlen         = sizeof(int),
5857                 .mode           = 0644,
5858                 .proc_handler   = proc_dointvec,
5859         },
5860 #ifdef CONFIG_IPV6_ROUTER_PREF
5861         {
5862                 .procname       = "accept_ra_rtr_pref",
5863                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
5864                 .maxlen         = sizeof(int),
5865                 .mode           = 0644,
5866                 .proc_handler   = proc_dointvec,
5867         },
5868         {
5869                 .procname       = "router_probe_interval",
5870                 .data           = &ipv6_devconf.rtr_probe_interval,
5871                 .maxlen         = sizeof(int),
5872                 .mode           = 0644,
5873                 .proc_handler   = proc_dointvec_jiffies,
5874         },
5875 #ifdef CONFIG_IPV6_ROUTE_INFO
5876         {
5877                 .procname       = "accept_ra_rt_info_max_plen",
5878                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5879                 .maxlen         = sizeof(int),
5880                 .mode           = 0644,
5881                 .proc_handler   = proc_dointvec,
5882         },
5883 #endif
5884 #endif
5885         {
5886                 .procname       = "proxy_ndp",
5887                 .data           = &ipv6_devconf.proxy_ndp,
5888                 .maxlen         = sizeof(int),
5889                 .mode           = 0644,
5890                 .proc_handler   = addrconf_sysctl_proxy_ndp,
5891         },
5892         {
5893                 .procname       = "accept_source_route",
5894                 .data           = &ipv6_devconf.accept_source_route,
5895                 .maxlen         = sizeof(int),
5896                 .mode           = 0644,
5897                 .proc_handler   = proc_dointvec,
5898         },
5899 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5900         {
5901                 .procname       = "optimistic_dad",
5902                 .data           = &ipv6_devconf.optimistic_dad,
5903                 .maxlen         = sizeof(int),
5904                 .mode           = 0644,
5905                 .proc_handler   = proc_dointvec,
5906         },
5907         {
5908                 .procname       = "use_optimistic",
5909                 .data           = &ipv6_devconf.use_optimistic,
5910                 .maxlen         = sizeof(int),
5911                 .mode           = 0644,
5912                 .proc_handler   = proc_dointvec,
5913         },
5914 #endif
5915 #ifdef CONFIG_IPV6_MROUTE
5916         {
5917                 .procname       = "mc_forwarding",
5918                 .data           = &ipv6_devconf.mc_forwarding,
5919                 .maxlen         = sizeof(int),
5920                 .mode           = 0444,
5921                 .proc_handler   = proc_dointvec,
5922         },
5923 #endif
5924         {
5925                 .procname       = "disable_ipv6",
5926                 .data           = &ipv6_devconf.disable_ipv6,
5927                 .maxlen         = sizeof(int),
5928                 .mode           = 0644,
5929                 .proc_handler   = addrconf_sysctl_disable,
5930         },
5931         {
5932                 .procname       = "accept_dad",
5933                 .data           = &ipv6_devconf.accept_dad,
5934                 .maxlen         = sizeof(int),
5935                 .mode           = 0644,
5936                 .proc_handler   = proc_dointvec,
5937         },
5938         {
5939                 .procname       = "force_tllao",
5940                 .data           = &ipv6_devconf.force_tllao,
5941                 .maxlen         = sizeof(int),
5942                 .mode           = 0644,
5943                 .proc_handler   = proc_dointvec
5944         },
5945         {
5946                 .procname       = "ndisc_notify",
5947                 .data           = &ipv6_devconf.ndisc_notify,
5948                 .maxlen         = sizeof(int),
5949                 .mode           = 0644,
5950                 .proc_handler   = proc_dointvec
5951         },
5952         {
5953                 .procname       = "suppress_frag_ndisc",
5954                 .data           = &ipv6_devconf.suppress_frag_ndisc,
5955                 .maxlen         = sizeof(int),
5956                 .mode           = 0644,
5957                 .proc_handler   = proc_dointvec
5958         },
5959         {
5960                 .procname       = "accept_ra_from_local",
5961                 .data           = &ipv6_devconf.accept_ra_from_local,
5962                 .maxlen         = sizeof(int),
5963                 .mode           = 0644,
5964                 .proc_handler   = proc_dointvec,
5965         },
5966         {
5967                 .procname       = "accept_ra_mtu",
5968                 .data           = &ipv6_devconf.accept_ra_mtu,
5969                 .maxlen         = sizeof(int),
5970                 .mode           = 0644,
5971                 .proc_handler   = proc_dointvec,
5972         },
5973         {
5974                 .procname       = "stable_secret",
5975                 .data           = &ipv6_devconf.stable_secret,
5976                 .maxlen         = IPV6_MAX_STRLEN,
5977                 .mode           = 0600,
5978                 .proc_handler   = addrconf_sysctl_stable_secret,
5979         },
5980         {
5981                 .procname       = "use_oif_addrs_only",
5982                 .data           = &ipv6_devconf.use_oif_addrs_only,
5983                 .maxlen         = sizeof(int),
5984                 .mode           = 0644,
5985                 .proc_handler   = proc_dointvec,
5986         },
5987         {
5988                 .procname       = "ignore_routes_with_linkdown",
5989                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5990                 .maxlen         = sizeof(int),
5991                 .mode           = 0644,
5992                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5993         },
5994         {
5995                 .procname       = "drop_unicast_in_l2_multicast",
5996                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
5997                 .maxlen         = sizeof(int),
5998                 .mode           = 0644,
5999                 .proc_handler   = proc_dointvec,
6000         },
6001         {
6002                 .procname       = "drop_unsolicited_na",
6003                 .data           = &ipv6_devconf.drop_unsolicited_na,
6004                 .maxlen         = sizeof(int),
6005                 .mode           = 0644,
6006                 .proc_handler   = proc_dointvec,
6007         },
6008         {
6009                 .procname       = "keep_addr_on_down",
6010                 .data           = &ipv6_devconf.keep_addr_on_down,
6011                 .maxlen         = sizeof(int),
6012                 .mode           = 0644,
6013                 .proc_handler   = proc_dointvec,
6014
6015         },
6016         {
6017                 /* sentinel */
6018         }
6019 };
6020
6021 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6022                 struct inet6_dev *idev, struct ipv6_devconf *p)
6023 {
6024         int i, ifindex;
6025         struct ctl_table *table;
6026         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6027
6028         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6029         if (!table)
6030                 goto out;
6031
6032         for (i = 0; table[i].data; i++) {
6033                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6034                 /* If one of these is already set, then it is not safe to
6035                  * overwrite either of them: this makes proc_dointvec_minmax
6036                  * usable.
6037                  */
6038                 if (!table[i].extra1 && !table[i].extra2) {
6039                         table[i].extra1 = idev; /* embedded; no ref */
6040                         table[i].extra2 = net;
6041                 }
6042         }
6043
6044         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6045
6046         p->sysctl_header = register_net_sysctl(net, path, table);
6047         if (!p->sysctl_header)
6048                 goto free;
6049
6050         if (!strcmp(dev_name, "all"))
6051                 ifindex = NETCONFA_IFINDEX_ALL;
6052         else if (!strcmp(dev_name, "default"))
6053                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6054         else
6055                 ifindex = idev->dev->ifindex;
6056         inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6057         return 0;
6058
6059 free:
6060         kfree(table);
6061 out:
6062         return -ENOBUFS;
6063 }
6064
6065 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6066 {
6067         struct ctl_table *table;
6068
6069         if (!p->sysctl_header)
6070                 return;
6071
6072         table = p->sysctl_header->ctl_table_arg;
6073         unregister_net_sysctl_table(p->sysctl_header);
6074         p->sysctl_header = NULL;
6075         kfree(table);
6076 }
6077
6078 static int addrconf_sysctl_register(struct inet6_dev *idev)
6079 {
6080         int err;
6081
6082         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6083                 return -EINVAL;
6084
6085         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6086                                     &ndisc_ifinfo_sysctl_change);
6087         if (err)
6088                 return err;
6089         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6090                                          idev, &idev->cnf);
6091         if (err)
6092                 neigh_sysctl_unregister(idev->nd_parms);
6093
6094         return err;
6095 }
6096
6097 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6098 {
6099         __addrconf_sysctl_unregister(&idev->cnf);
6100         neigh_sysctl_unregister(idev->nd_parms);
6101 }
6102
6103
6104 #endif
6105
6106 static int __net_init addrconf_init_net(struct net *net)
6107 {
6108         int err = -ENOMEM;
6109         struct ipv6_devconf *all, *dflt;
6110
6111         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6112         if (!all)
6113                 goto err_alloc_all;
6114
6115         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6116         if (!dflt)
6117                 goto err_alloc_dflt;
6118
6119         /* these will be inherited by all namespaces */
6120         dflt->autoconf = ipv6_defaults.autoconf;
6121         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6122
6123         dflt->stable_secret.initialized = false;
6124         all->stable_secret.initialized = false;
6125
6126         net->ipv6.devconf_all = all;
6127         net->ipv6.devconf_dflt = dflt;
6128
6129 #ifdef CONFIG_SYSCTL
6130         err = __addrconf_sysctl_register(net, "all", NULL, all);
6131         if (err < 0)
6132                 goto err_reg_all;
6133
6134         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6135         if (err < 0)
6136                 goto err_reg_dflt;
6137 #endif
6138         return 0;
6139
6140 #ifdef CONFIG_SYSCTL
6141 err_reg_dflt:
6142         __addrconf_sysctl_unregister(all);
6143 err_reg_all:
6144         kfree(dflt);
6145 #endif
6146 err_alloc_dflt:
6147         kfree(all);
6148 err_alloc_all:
6149         return err;
6150 }
6151
6152 static void __net_exit addrconf_exit_net(struct net *net)
6153 {
6154 #ifdef CONFIG_SYSCTL
6155         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6156         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6157 #endif
6158         kfree(net->ipv6.devconf_dflt);
6159         kfree(net->ipv6.devconf_all);
6160 }
6161
6162 static struct pernet_operations addrconf_ops = {
6163         .init = addrconf_init_net,
6164         .exit = addrconf_exit_net,
6165 };
6166
6167 static struct rtnl_af_ops inet6_ops __read_mostly = {
6168         .family           = AF_INET6,
6169         .fill_link_af     = inet6_fill_link_af,
6170         .get_link_af_size = inet6_get_link_af_size,
6171         .validate_link_af = inet6_validate_link_af,
6172         .set_link_af      = inet6_set_link_af,
6173 };
6174
6175 /*
6176  *      Init / cleanup code
6177  */
6178
6179 int __init addrconf_init(void)
6180 {
6181         struct inet6_dev *idev;
6182         int i, err;
6183
6184         err = ipv6_addr_label_init();
6185         if (err < 0) {
6186                 pr_crit("%s: cannot initialize default policy table: %d\n",
6187                         __func__, err);
6188                 goto out;
6189         }
6190
6191         err = register_pernet_subsys(&addrconf_ops);
6192         if (err < 0)
6193                 goto out_addrlabel;
6194
6195         addrconf_wq = create_workqueue("ipv6_addrconf");
6196         if (!addrconf_wq) {
6197                 err = -ENOMEM;
6198                 goto out_nowq;
6199         }
6200
6201         /* The addrconf netdev notifier requires that loopback_dev
6202          * has it's ipv6 private information allocated and setup
6203          * before it can bring up and give link-local addresses
6204          * to other devices which are up.
6205          *
6206          * Unfortunately, loopback_dev is not necessarily the first
6207          * entry in the global dev_base list of net devices.  In fact,
6208          * it is likely to be the very last entry on that list.
6209          * So this causes the notifier registry below to try and
6210          * give link-local addresses to all devices besides loopback_dev
6211          * first, then loopback_dev, which cases all the non-loopback_dev
6212          * devices to fail to get a link-local address.
6213          *
6214          * So, as a temporary fix, allocate the ipv6 structure for
6215          * loopback_dev first by hand.
6216          * Longer term, all of the dependencies ipv6 has upon the loopback
6217          * device and it being up should be removed.
6218          */
6219         rtnl_lock();
6220         idev = ipv6_add_dev(init_net.loopback_dev);
6221         rtnl_unlock();
6222         if (IS_ERR(idev)) {
6223                 err = PTR_ERR(idev);
6224                 goto errlo;
6225         }
6226
6227         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6228                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6229
6230         register_netdevice_notifier(&ipv6_dev_notf);
6231
6232         addrconf_verify();
6233
6234         rtnl_af_register(&inet6_ops);
6235
6236         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6237                               NULL);
6238         if (err < 0)
6239                 goto errout;
6240
6241         /* Only the first call to __rtnl_register can fail */
6242         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6243         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6244         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6245                         inet6_dump_ifaddr, NULL);
6246         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6247                         inet6_dump_ifmcaddr, NULL);
6248         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6249                         inet6_dump_ifacaddr, NULL);
6250         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6251                         inet6_netconf_dump_devconf, NULL);
6252
6253         ipv6_addr_label_rtnl_register();
6254
6255         return 0;
6256 errout:
6257         rtnl_af_unregister(&inet6_ops);
6258         unregister_netdevice_notifier(&ipv6_dev_notf);
6259 errlo:
6260         destroy_workqueue(addrconf_wq);
6261 out_nowq:
6262         unregister_pernet_subsys(&addrconf_ops);
6263 out_addrlabel:
6264         ipv6_addr_label_cleanup();
6265 out:
6266         return err;
6267 }
6268
6269 void addrconf_cleanup(void)
6270 {
6271         struct net_device *dev;
6272         int i;
6273
6274         unregister_netdevice_notifier(&ipv6_dev_notf);
6275         unregister_pernet_subsys(&addrconf_ops);
6276         ipv6_addr_label_cleanup();
6277
6278         rtnl_lock();
6279
6280         __rtnl_af_unregister(&inet6_ops);
6281
6282         /* clean dev list */
6283         for_each_netdev(&init_net, dev) {
6284                 if (__in6_dev_get(dev) == NULL)
6285                         continue;
6286                 addrconf_ifdown(dev, 1);
6287         }
6288         addrconf_ifdown(init_net.loopback_dev, 2);
6289
6290         /*
6291          *      Check hash table.
6292          */
6293         spin_lock_bh(&addrconf_hash_lock);
6294         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6295                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6296         spin_unlock_bh(&addrconf_hash_lock);
6297         cancel_delayed_work(&addr_chk_work);
6298         rtnl_unlock();
6299
6300         destroy_workqueue(addrconf_wq);
6301 }