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/af_ieee802154.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 #ifdef CONFIG_SYSCTL
 116 static int addrconf_sysctl_register(struct inet6_dev *idev);
 117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 118 #else
 119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 120 {
 121         return 0;
 122 }
 123
 124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 125 {
 126 }
 127 #endif
 128
 129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 131 static void ipv6_regen_rndid(unsigned long data);
 132
 133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 134 static int ipv6_count_addresses(struct inet6_dev *idev);
 135 static int ipv6_generate_stable_address(struct in6_addr *addr,
 136                                         u8 dad_count,
 137                                         const struct inet6_dev *idev);
 138
 139 /*
 140  *      Configured unicast address hash table
 141  */
 142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 143 static DEFINE_SPINLOCK(addrconf_hash_lock);
 144
 145 static void addrconf_verify(void);
 146 static void addrconf_verify_rtnl(void);
 147 static void addrconf_verify_work(struct work_struct *);
 148
 149 static struct workqueue_struct *addrconf_wq;
 150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 151
 152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 154
 155 static void addrconf_type_change(struct net_device *dev,
 156                                  unsigned long event);
 157 static int addrconf_ifdown(struct net_device *dev, int how);
 158
 159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 160                                                   int plen,
 161                                                   const struct net_device *dev,
 162                                                   u32 flags, u32 noflags);
 163
 164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 165 static void addrconf_dad_work(struct work_struct *w);
 166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 167 static void addrconf_dad_run(struct inet6_dev *idev);
 168 static void addrconf_rs_timer(unsigned long data);
 169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 171
 172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 173                                 struct prefix_info *pinfo);
 174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 175                                struct net_device *dev);
 176
 177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 178         .forwarding             = 0,
 179         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 180         .mtu6                   = IPV6_MIN_MTU,
 181         .accept_ra              = 1,
 182         .accept_redirects       = 1,
 183         .autoconf               = 1,
 184         .force_mld_version      = 0,
 185         .mldv1_unsolicited_report_interval = 10 * HZ,
 186         .mldv2_unsolicited_report_interval = HZ,
 187         .dad_transmits          = 1,
 188         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 189         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 190         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 191         .use_tempaddr           = 0,
 192         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 193         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 194         .regen_max_retry        = REGEN_MAX_RETRY,
 195         .max_desync_factor      = MAX_DESYNC_FACTOR,
 196         .max_addresses          = IPV6_MAX_ADDRESSES,
 197         .accept_ra_defrtr       = 1,
 198         .accept_ra_from_local   = 0,
 199         .accept_ra_min_hop_limit= 1,
 200         .accept_ra_pinfo        = 1,
 201 #ifdef CONFIG_IPV6_ROUTER_PREF
 202         .accept_ra_rtr_pref     = 1,
 203         .rtr_probe_interval     = 60 * HZ,
 204 #ifdef CONFIG_IPV6_ROUTE_INFO
 205         .accept_ra_rt_info_max_plen = 0,
 206 #endif
 207 #endif
 208         .proxy_ndp              = 0,
 209         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 210         .disable_ipv6           = 0,
 211         .accept_dad             = 1,
 212         .suppress_frag_ndisc    = 1,
 213         .accept_ra_mtu          = 1,
 214         .stable_secret          = {
 215                 .initialized = false,
 216         },
 217         .use_oif_addrs_only     = 0,
 218         .ignore_routes_with_linkdown = 0,
 219 };
 220
 221 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 222         .forwarding             = 0,
 223         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 224         .mtu6                   = IPV6_MIN_MTU,
 225         .accept_ra              = 1,
 226         .accept_redirects       = 1,
 227         .autoconf               = 1,
 228         .force_mld_version      = 0,
 229         .mldv1_unsolicited_report_interval = 10 * HZ,
 230         .mldv2_unsolicited_report_interval = HZ,
 231         .dad_transmits          = 1,
 232         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 233         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 234         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 235         .use_tempaddr           = 0,
 236         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 237         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 238         .regen_max_retry        = REGEN_MAX_RETRY,
 239         .max_desync_factor      = MAX_DESYNC_FACTOR,
 240         .max_addresses          = IPV6_MAX_ADDRESSES,
 241         .accept_ra_defrtr       = 1,
 242         .accept_ra_from_local   = 0,
 243         .accept_ra_min_hop_limit= 1,
 244         .accept_ra_pinfo        = 1,
 245 #ifdef CONFIG_IPV6_ROUTER_PREF
 246         .accept_ra_rtr_pref     = 1,
 247         .rtr_probe_interval     = 60 * HZ,
 248 #ifdef CONFIG_IPV6_ROUTE_INFO
 249         .accept_ra_rt_info_max_plen = 0,
 250 #endif
 251 #endif
 252         .proxy_ndp              = 0,
 253         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 254         .disable_ipv6           = 0,
 255         .accept_dad             = 1,
 256         .suppress_frag_ndisc    = 1,
 257         .accept_ra_mtu          = 1,
 258         .stable_secret          = {
 259                 .initialized = false,
 260         },
 261         .use_oif_addrs_only     = 0,
 262         .ignore_routes_with_linkdown = 0,
 263 };
 264
 265 /* Check if a valid qdisc is available */
 266 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 267 {
 268         return !qdisc_tx_is_noop(dev);
 269 }
 270
 271 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 272 {
 273         if (del_timer(&idev->rs_timer))
 274                 __in6_dev_put(idev);
 275 }
 276
 277 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 278 {
 279         if (cancel_delayed_work(&ifp->dad_work))
 280                 __in6_ifa_put(ifp);
 281 }
 282
 283 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 284                                   unsigned long when)
 285 {
 286         if (!timer_pending(&idev->rs_timer))
 287                 in6_dev_hold(idev);
 288         mod_timer(&idev->rs_timer, jiffies + when);
 289 }
 290
 291 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 292                                    unsigned long delay)
 293 {
 294         in6_ifa_hold(ifp);
 295         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 296                 in6_ifa_put(ifp);
 297 }
 298
 299 static int snmp6_alloc_dev(struct inet6_dev *idev)
 300 {
 301         int i;
 302
 303         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 304         if (!idev->stats.ipv6)
 305                 goto err_ip;
 306
 307         for_each_possible_cpu(i) {
 308                 struct ipstats_mib *addrconf_stats;
 309                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 310                 u64_stats_init(&addrconf_stats->syncp);
 311         }
 312
 313
 314         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 315                                         GFP_KERNEL);
 316         if (!idev->stats.icmpv6dev)
 317                 goto err_icmp;
 318         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 319                                            GFP_KERNEL);
 320         if (!idev->stats.icmpv6msgdev)
 321                 goto err_icmpmsg;
 322
 323         return 0;
 324
 325 err_icmpmsg:
 326         kfree(idev->stats.icmpv6dev);
 327 err_icmp:
 328         free_percpu(idev->stats.ipv6);
 329 err_ip:
 330         return -ENOMEM;
 331 }
 332
 333 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 334 {
 335         struct inet6_dev *ndev;
 336         int err = -ENOMEM;
 337
 338         ASSERT_RTNL();
 339
 340         if (dev->mtu < IPV6_MIN_MTU)
 341                 return ERR_PTR(-EINVAL);
 342
 343         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 344         if (!ndev)
 345                 return ERR_PTR(err);
 346
 347         rwlock_init(&ndev->lock);
 348         ndev->dev = dev;
 349         INIT_LIST_HEAD(&ndev->addr_list);
 350         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 351                     (unsigned long)ndev);
 352         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 353
 354         if (ndev->cnf.stable_secret.initialized)
 355                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 356         else
 357                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
 358
 359         ndev->cnf.mtu6 = dev->mtu;
 360         ndev->cnf.sysctl = NULL;
 361         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 362         if (!ndev->nd_parms) {
 363                 kfree(ndev);
 364                 return ERR_PTR(err);
 365         }
 366         if (ndev->cnf.forwarding)
 367                 dev_disable_lro(dev);
 368         /* We refer to the device */
 369         dev_hold(dev);
 370
 371         if (snmp6_alloc_dev(ndev) < 0) {
 372                 ADBG(KERN_WARNING
 373                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 374                         __func__, dev->name);
 375                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 376                 dev_put(dev);
 377                 kfree(ndev);
 378                 return ERR_PTR(err);
 379         }
 380
 381         if (snmp6_register_dev(ndev) < 0) {
 382                 ADBG(KERN_WARNING
 383                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 384                         __func__, dev->name);
 385                 goto err_release;
 386         }
 387
 388         /* One reference from device.  We must do this before
 389          * we invoke __ipv6_regen_rndid().
 390          */
 391         in6_dev_hold(ndev);
 392
 393         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 394                 ndev->cnf.accept_dad = -1;
 395
 396 #if IS_ENABLED(CONFIG_IPV6_SIT)
 397         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 398                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 399                 ndev->cnf.rtr_solicits = 0;
 400         }
 401 #endif
 402
 403         INIT_LIST_HEAD(&ndev->tempaddr_list);
 404         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 405         if ((dev->flags&IFF_LOOPBACK) ||
 406             dev->type == ARPHRD_TUNNEL ||
 407             dev->type == ARPHRD_TUNNEL6 ||
 408             dev->type == ARPHRD_SIT ||
 409             dev->type == ARPHRD_NONE) {
 410                 ndev->cnf.use_tempaddr = -1;
 411         } else {
 412                 in6_dev_hold(ndev);
 413                 ipv6_regen_rndid((unsigned long) ndev);
 414         }
 415
 416         ndev->token = in6addr_any;
 417
 418         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 419                 ndev->if_flags |= IF_READY;
 420
 421         ipv6_mc_init_dev(ndev);
 422         ndev->tstamp = jiffies;
 423         err = addrconf_sysctl_register(ndev);
 424         if (err) {
 425                 ipv6_mc_destroy_dev(ndev);
 426                 del_timer(&ndev->regen_timer);
 427                 snmp6_unregister_dev(ndev);
 428                 goto err_release;
 429         }
 430         /* protected by rtnl_lock */
 431         rcu_assign_pointer(dev->ip6_ptr, ndev);
 432
 433         /* Join interface-local all-node multicast group */
 434         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 435
 436         /* Join all-node multicast group */
 437         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 438
 439         /* Join all-router multicast group if forwarding is set */
 440         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 442
 443         return ndev;
 444
 445 err_release:
 446         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 447         ndev->dead = 1;
 448         in6_dev_finish_destroy(ndev);
 449         return ERR_PTR(err);
 450 }
 451
 452 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 453 {
 454         struct inet6_dev *idev;
 455
 456         ASSERT_RTNL();
 457
 458         idev = __in6_dev_get(dev);
 459         if (!idev) {
 460                 idev = ipv6_add_dev(dev);
 461                 if (IS_ERR(idev))
 462                         return NULL;
 463         }
 464
 465         if (dev->flags&IFF_UP)
 466                 ipv6_mc_up(idev);
 467         return idev;
 468 }
 469
 470 static int inet6_netconf_msgsize_devconf(int type)
 471 {
 472         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 473                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 474
 475         /* type -1 is used for ALL */
 476         if (type == -1 || type == NETCONFA_FORWARDING)
 477                 size += nla_total_size(4);
 478 #ifdef CONFIG_IPV6_MROUTE
 479         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 480                 size += nla_total_size(4);
 481 #endif
 482         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 483                 size += nla_total_size(4);
 484
 485         if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 486                 size += nla_total_size(4);
 487
 488         return size;
 489 }
 490
 491 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 492                                       struct ipv6_devconf *devconf, u32 portid,
 493                                       u32 seq, int event, unsigned int flags,
 494                                       int type)
 495 {
 496         struct nlmsghdr  *nlh;
 497         struct netconfmsg *ncm;
 498
 499         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 500                         flags);
 501         if (!nlh)
 502                 return -EMSGSIZE;
 503
 504         ncm = nlmsg_data(nlh);
 505         ncm->ncm_family = AF_INET6;
 506
 507         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 508                 goto nla_put_failure;
 509
 510         /* type -1 is used for ALL */
 511         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 512             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 513                 goto nla_put_failure;
 514 #ifdef CONFIG_IPV6_MROUTE
 515         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 516             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 517                         devconf->mc_forwarding) < 0)
 518                 goto nla_put_failure;
 519 #endif
 520         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 521             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 522                 goto nla_put_failure;
 523
 524         if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 525             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 526                         devconf->ignore_routes_with_linkdown) < 0)
 527                 goto nla_put_failure;
 528
 529         nlmsg_end(skb, nlh);
 530         return 0;
 531
 532 nla_put_failure:
 533         nlmsg_cancel(skb, nlh);
 534         return -EMSGSIZE;
 535 }
 536
 537 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 538                                   struct ipv6_devconf *devconf)
 539 {
 540         struct sk_buff *skb;
 541         int err = -ENOBUFS;
 542
 543         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 544         if (!skb)
 545                 goto errout;
 546
 547         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 548                                          RTM_NEWNETCONF, 0, type);
 549         if (err < 0) {
 550                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 551                 WARN_ON(err == -EMSGSIZE);
 552                 kfree_skb(skb);
 553                 goto errout;
 554         }
 555         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 556         return;
 557 errout:
 558         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 559 }
 560
 561 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 562         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 563         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 564         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 565         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 566 };
 567
 568 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 569                                      struct nlmsghdr *nlh)
 570 {
 571         struct net *net = sock_net(in_skb->sk);
 572         struct nlattr *tb[NETCONFA_MAX+1];
 573         struct netconfmsg *ncm;
 574         struct sk_buff *skb;
 575         struct ipv6_devconf *devconf;
 576         struct inet6_dev *in6_dev;
 577         struct net_device *dev;
 578         int ifindex;
 579         int err;
 580
 581         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 582                           devconf_ipv6_policy);
 583         if (err < 0)
 584                 goto errout;
 585
 586         err = -EINVAL;
 587         if (!tb[NETCONFA_IFINDEX])
 588                 goto errout;
 589
 590         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 591         switch (ifindex) {
 592         case NETCONFA_IFINDEX_ALL:
 593                 devconf = net->ipv6.devconf_all;
 594                 break;
 595         case NETCONFA_IFINDEX_DEFAULT:
 596                 devconf = net->ipv6.devconf_dflt;
 597                 break;
 598         default:
 599                 dev = __dev_get_by_index(net, ifindex);
 600                 if (!dev)
 601                         goto errout;
 602                 in6_dev = __in6_dev_get(dev);
 603                 if (!in6_dev)
 604                         goto errout;
 605                 devconf = &in6_dev->cnf;
 606                 break;
 607         }
 608
 609         err = -ENOBUFS;
 610         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 611         if (!skb)
 612                 goto errout;
 613
 614         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 615                                          NETLINK_CB(in_skb).portid,
 616                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 617                                          -1);
 618         if (err < 0) {
 619                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 620                 WARN_ON(err == -EMSGSIZE);
 621                 kfree_skb(skb);
 622                 goto errout;
 623         }
 624         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 625 errout:
 626         return err;
 627 }
 628
 629 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 630                                       struct netlink_callback *cb)
 631 {
 632         struct net *net = sock_net(skb->sk);
 633         int h, s_h;
 634         int idx, s_idx;
 635         struct net_device *dev;
 636         struct inet6_dev *idev;
 637         struct hlist_head *head;
 638
 639         s_h = cb->args[0];
 640         s_idx = idx = cb->args[1];
 641
 642         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 643                 idx = 0;
 644                 head = &net->dev_index_head[h];
 645                 rcu_read_lock();
 646                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 647                           net->dev_base_seq;
 648                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 649                         if (idx < s_idx)
 650                                 goto cont;
 651                         idev = __in6_dev_get(dev);
 652                         if (!idev)
 653                                 goto cont;
 654
 655                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 656                                                        &idev->cnf,
 657                                                        NETLINK_CB(cb->skb).portid,
 658                                                        cb->nlh->nlmsg_seq,
 659                                                        RTM_NEWNETCONF,
 660                                                        NLM_F_MULTI,
 661                                                        -1) < 0) {
 662                                 rcu_read_unlock();
 663                                 goto done;
 664                         }
 665                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 666 cont:
 667                         idx++;
 668                 }
 669                 rcu_read_unlock();
 670         }
 671         if (h == NETDEV_HASHENTRIES) {
 672                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 673                                                net->ipv6.devconf_all,
 674                                                NETLINK_CB(cb->skb).portid,
 675                                                cb->nlh->nlmsg_seq,
 676                                                RTM_NEWNETCONF, NLM_F_MULTI,
 677                                                -1) < 0)
 678                         goto done;
 679                 else
 680                         h++;
 681         }
 682         if (h == NETDEV_HASHENTRIES + 1) {
 683                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 684                                                net->ipv6.devconf_dflt,
 685                                                NETLINK_CB(cb->skb).portid,
 686                                                cb->nlh->nlmsg_seq,
 687                                                RTM_NEWNETCONF, NLM_F_MULTI,
 688                                                -1) < 0)
 689                         goto done;
 690                 else
 691                         h++;
 692         }
 693 done:
 694         cb->args[0] = h;
 695         cb->args[1] = idx;
 696
 697         return skb->len;
 698 }
 699
 700 #ifdef CONFIG_SYSCTL
 701 static void dev_forward_change(struct inet6_dev *idev)
 702 {
 703         struct net_device *dev;
 704         struct inet6_ifaddr *ifa;
 705
 706         if (!idev)
 707                 return;
 708         dev = idev->dev;
 709         if (idev->cnf.forwarding)
 710                 dev_disable_lro(dev);
 711         if (dev->flags & IFF_MULTICAST) {
 712                 if (idev->cnf.forwarding) {
 713                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 714                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 715                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 716                 } else {
 717                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 718                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 719                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 720                 }
 721         }
 722
 723         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 724                 if (ifa->flags&IFA_F_TENTATIVE)
 725                         continue;
 726                 if (idev->cnf.forwarding)
 727                         addrconf_join_anycast(ifa);
 728                 else
 729                         addrconf_leave_anycast(ifa);
 730         }
 731         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 732                                      dev->ifindex, &idev->cnf);
 733 }
 734
 735
 736 static void addrconf_forward_change(struct net *net, __s32 newf)
 737 {
 738         struct net_device *dev;
 739         struct inet6_dev *idev;
 740
 741         for_each_netdev(net, dev) {
 742                 idev = __in6_dev_get(dev);
 743                 if (idev) {
 744                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 745                         idev->cnf.forwarding = newf;
 746                         if (changed)
 747                                 dev_forward_change(idev);
 748                 }
 749         }
 750 }
 751
 752 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 753 {
 754         struct net *net;
 755         int old;
 756
 757         if (!rtnl_trylock())
 758                 return restart_syscall();
 759
 760         net = (struct net *)table->extra2;
 761         old = *p;
 762         *p = newf;
 763
 764         if (p == &net->ipv6.devconf_dflt->forwarding) {
 765                 if ((!newf) ^ (!old))
 766                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 767                                                      NETCONFA_IFINDEX_DEFAULT,
 768                                                      net->ipv6.devconf_dflt);
 769                 rtnl_unlock();
 770                 return 0;
 771         }
 772
 773         if (p == &net->ipv6.devconf_all->forwarding) {
 774                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 775
 776                 net->ipv6.devconf_dflt->forwarding = newf;
 777                 if ((!newf) ^ (!old_dflt))
 778                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 779                                                      NETCONFA_IFINDEX_DEFAULT,
 780                                                      net->ipv6.devconf_dflt);
 781
 782                 addrconf_forward_change(net, newf);
 783                 if ((!newf) ^ (!old))
 784                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 785                                                      NETCONFA_IFINDEX_ALL,
 786                                                      net->ipv6.devconf_all);
 787         } else if ((!newf) ^ (!old))
 788                 dev_forward_change((struct inet6_dev *)table->extra1);
 789         rtnl_unlock();
 790
 791         if (newf)
 792                 rt6_purge_dflt_routers(net);
 793         return 1;
 794 }
 795
 796 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 797 {
 798         struct net_device *dev;
 799         struct inet6_dev *idev;
 800
 801         for_each_netdev(net, dev) {
 802                 idev = __in6_dev_get(dev);
 803                 if (idev) {
 804                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 805
 806                         idev->cnf.ignore_routes_with_linkdown = newf;
 807                         if (changed)
 808                                 inet6_netconf_notify_devconf(dev_net(dev),
 809                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 810                                                              dev->ifindex,
 811                                                              &idev->cnf);
 812                 }
 813         }
 814 }
 815
 816 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 817 {
 818         struct net *net;
 819         int old;
 820
 821         if (!rtnl_trylock())
 822                 return restart_syscall();
 823
 824         net = (struct net *)table->extra2;
 825         old = *p;
 826         *p = newf;
 827
 828         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 829                 if ((!newf) ^ (!old))
 830                         inet6_netconf_notify_devconf(net,
 831                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 832                                                      NETCONFA_IFINDEX_DEFAULT,
 833                                                      net->ipv6.devconf_dflt);
 834                 rtnl_unlock();
 835                 return 0;
 836         }
 837
 838         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 839                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 840                 addrconf_linkdown_change(net, newf);
 841                 if ((!newf) ^ (!old))
 842                         inet6_netconf_notify_devconf(net,
 843                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 844                                                      NETCONFA_IFINDEX_ALL,
 845                                                      net->ipv6.devconf_all);
 846         }
 847         rtnl_unlock();
 848
 849         return 1;
 850 }
 851
 852 #endif
 853
 854 /* Nobody refers to this ifaddr, destroy it */
 855 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 856 {
 857         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 858
 859 #ifdef NET_REFCNT_DEBUG
 860         pr_debug("%s\n", __func__);
 861 #endif
 862
 863         in6_dev_put(ifp->idev);
 864
 865         if (cancel_delayed_work(&ifp->dad_work))
 866                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 867                           ifp);
 868
 869         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 870                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 871                 return;
 872         }
 873         ip6_rt_put(ifp->rt);
 874
 875         kfree_rcu(ifp, rcu);
 876 }
 877
 878 static void
 879 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 880 {
 881         struct list_head *p;
 882         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 883
 884         /*
 885          * Each device address list is sorted in order of scope -
 886          * global before linklocal.
 887          */
 888         list_for_each(p, &idev->addr_list) {
 889                 struct inet6_ifaddr *ifa
 890                         = list_entry(p, struct inet6_ifaddr, if_list);
 891                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 892                         break;
 893         }
 894
 895         list_add_tail(&ifp->if_list, p);
 896 }
 897
 898 static u32 inet6_addr_hash(const struct in6_addr *addr)
 899 {
 900         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 901 }
 902
 903 /* On success it returns ifp with increased reference count */
 904
 905 static struct inet6_ifaddr *
 906 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 907               const struct in6_addr *peer_addr, int pfxlen,
 908               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 909 {
 910         struct inet6_ifaddr *ifa = NULL;
 911         struct rt6_info *rt;
 912         unsigned int hash;
 913         int err = 0;
 914         int addr_type = ipv6_addr_type(addr);
 915
 916         if (addr_type == IPV6_ADDR_ANY ||
 917             addr_type & IPV6_ADDR_MULTICAST ||
 918             (!(idev->dev->flags & IFF_LOOPBACK) &&
 919              addr_type & IPV6_ADDR_LOOPBACK))
 920                 return ERR_PTR(-EADDRNOTAVAIL);
 921
 922         rcu_read_lock_bh();
 923         if (idev->dead) {
 924                 err = -ENODEV;                  /*XXX*/
 925                 goto out2;
 926         }
 927
 928         if (idev->cnf.disable_ipv6) {
 929                 err = -EACCES;
 930                 goto out2;
 931         }
 932
 933         spin_lock(&addrconf_hash_lock);
 934
 935         /* Ignore adding duplicate addresses on an interface */
 936         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 937                 ADBG("ipv6_add_addr: already assigned\n");
 938                 err = -EEXIST;
 939                 goto out;
 940         }
 941
 942         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 943
 944         if (!ifa) {
 945                 ADBG("ipv6_add_addr: malloc failed\n");
 946                 err = -ENOBUFS;
 947                 goto out;
 948         }
 949
 950         rt = addrconf_dst_alloc(idev, addr, false);
 951         if (IS_ERR(rt)) {
 952                 err = PTR_ERR(rt);
 953                 goto out;
 954         }
 955
 956         neigh_parms_data_state_setall(idev->nd_parms);
 957
 958         ifa->addr = *addr;
 959         if (peer_addr)
 960                 ifa->peer_addr = *peer_addr;
 961
 962         spin_lock_init(&ifa->lock);
 963         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 964         INIT_HLIST_NODE(&ifa->addr_lst);
 965         ifa->scope = scope;
 966         ifa->prefix_len = pfxlen;
 967         ifa->flags = flags;
 968         /* No need to add the TENTATIVE flag for addresses with NODAD */
 969         if (!(flags & IFA_F_NODAD))
 970                 ifa->flags |= IFA_F_TENTATIVE;
 971         ifa->valid_lft = valid_lft;
 972         ifa->prefered_lft = prefered_lft;
 973         ifa->cstamp = ifa->tstamp = jiffies;
 974         ifa->tokenized = false;
 975
 976         ifa->rt = rt;
 977
 978         ifa->idev = idev;
 979         in6_dev_hold(idev);
 980         /* For caller */
 981         in6_ifa_hold(ifa);
 982
 983         /* Add to big hash table */
 984         hash = inet6_addr_hash(addr);
 985
 986         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 987         spin_unlock(&addrconf_hash_lock);
 988
 989         write_lock(&idev->lock);
 990         /* Add to inet6_dev unicast addr list. */
 991         ipv6_link_dev_addr(idev, ifa);
 992
 993         if (ifa->flags&IFA_F_TEMPORARY) {
 994                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 995                 in6_ifa_hold(ifa);
 996         }
 997
 998         in6_ifa_hold(ifa);
 999         write_unlock(&idev->lock);
1000 out2:
1001         rcu_read_unlock_bh();
1002
1003         if (likely(err == 0))
1004                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1005         else {
1006                 kfree(ifa);
1007                 ifa = ERR_PTR(err);
1008         }
1009
1010         return ifa;
1011 out:
1012         spin_unlock(&addrconf_hash_lock);
1013         goto out2;
1014 }
1015
1016 enum cleanup_prefix_rt_t {
1017         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1018         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1019         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1020 };
1021
1022 /*
1023  * Check, whether the prefix for ifp would still need a prefix route
1024  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1025  * constants.
1026  *
1027  * 1) we don't purge prefix if address was not permanent.
1028  *    prefix is managed by its own lifetime.
1029  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1030  * 3) if there are no addresses, delete prefix.
1031  * 4) if there are still other permanent address(es),
1032  *    corresponding prefix is still permanent.
1033  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1034  *    don't purge the prefix, assume user space is managing it.
1035  * 6) otherwise, update prefix lifetime to the
1036  *    longest valid lifetime among the corresponding
1037  *    addresses on the device.
1038  *    Note: subsequent RA will update lifetime.
1039  **/
1040 static enum cleanup_prefix_rt_t
1041 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1042 {
1043         struct inet6_ifaddr *ifa;
1044         struct inet6_dev *idev = ifp->idev;
1045         unsigned long lifetime;
1046         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1047
1048         *expires = jiffies;
1049
1050         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1051                 if (ifa == ifp)
1052                         continue;
1053                 if (ifa->prefix_len != ifp->prefix_len ||
1054                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1055                                        ifp->prefix_len))
1056                         continue;
1057                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1058                         return CLEANUP_PREFIX_RT_NOP;
1059
1060                 action = CLEANUP_PREFIX_RT_EXPIRE;
1061
1062                 spin_lock(&ifa->lock);
1063
1064                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1065                 /*
1066                  * Note: Because this address is
1067                  * not permanent, lifetime <
1068                  * LONG_MAX / HZ here.
1069                  */
1070                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1071                         *expires = ifa->tstamp + lifetime * HZ;
1072                 spin_unlock(&ifa->lock);
1073         }
1074
1075         return action;
1076 }
1077
1078 static void
1079 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1080 {
1081         struct rt6_info *rt;
1082
1083         rt = addrconf_get_prefix_route(&ifp->addr,
1084                                        ifp->prefix_len,
1085                                        ifp->idev->dev,
1086                                        0, RTF_GATEWAY | RTF_DEFAULT);
1087         if (rt) {
1088                 if (del_rt)
1089                         ip6_del_rt(rt);
1090                 else {
1091                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1092                                 rt6_set_expires(rt, expires);
1093                         ip6_rt_put(rt);
1094                 }
1095         }
1096 }
1097
1098
1099 /* This function wants to get referenced ifp and releases it before return */
1100
1101 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1102 {
1103         int state;
1104         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1105         unsigned long expires;
1106
1107         ASSERT_RTNL();
1108
1109         spin_lock_bh(&ifp->lock);
1110         state = ifp->state;
1111         ifp->state = INET6_IFADDR_STATE_DEAD;
1112         spin_unlock_bh(&ifp->lock);
1113
1114         if (state == INET6_IFADDR_STATE_DEAD)
1115                 goto out;
1116
1117         spin_lock_bh(&addrconf_hash_lock);
1118         hlist_del_init_rcu(&ifp->addr_lst);
1119         spin_unlock_bh(&addrconf_hash_lock);
1120
1121         write_lock_bh(&ifp->idev->lock);
1122
1123         if (ifp->flags&IFA_F_TEMPORARY) {
1124                 list_del(&ifp->tmp_list);
1125                 if (ifp->ifpub) {
1126                         in6_ifa_put(ifp->ifpub);
1127                         ifp->ifpub = NULL;
1128                 }
1129                 __in6_ifa_put(ifp);
1130         }
1131
1132         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1133                 action = check_cleanup_prefix_route(ifp, &expires);
1134
1135         list_del_init(&ifp->if_list);
1136         __in6_ifa_put(ifp);
1137
1138         write_unlock_bh(&ifp->idev->lock);
1139
1140         addrconf_del_dad_work(ifp);
1141
1142         ipv6_ifa_notify(RTM_DELADDR, ifp);
1143
1144         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1145
1146         if (action != CLEANUP_PREFIX_RT_NOP) {
1147                 cleanup_prefix_route(ifp, expires,
1148                         action == CLEANUP_PREFIX_RT_DEL);
1149         }
1150
1151         /* clean up prefsrc entries */
1152         rt6_remove_prefsrc(ifp);
1153 out:
1154         in6_ifa_put(ifp);
1155 }
1156
1157 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1158 {
1159         struct inet6_dev *idev = ifp->idev;
1160         struct in6_addr addr, *tmpaddr;
1161         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1162         unsigned long regen_advance;
1163         int tmp_plen;
1164         int ret = 0;
1165         u32 addr_flags;
1166         unsigned long now = jiffies;
1167
1168         write_lock_bh(&idev->lock);
1169         if (ift) {
1170                 spin_lock_bh(&ift->lock);
1171                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1172                 spin_unlock_bh(&ift->lock);
1173                 tmpaddr = &addr;
1174         } else {
1175                 tmpaddr = NULL;
1176         }
1177 retry:
1178         in6_dev_hold(idev);
1179         if (idev->cnf.use_tempaddr <= 0) {
1180                 write_unlock_bh(&idev->lock);
1181                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1182                 in6_dev_put(idev);
1183                 ret = -1;
1184                 goto out;
1185         }
1186         spin_lock_bh(&ifp->lock);
1187         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1188                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1189                 spin_unlock_bh(&ifp->lock);
1190                 write_unlock_bh(&idev->lock);
1191                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1192                         __func__);
1193                 in6_dev_put(idev);
1194                 ret = -1;
1195                 goto out;
1196         }
1197         in6_ifa_hold(ifp);
1198         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1199         __ipv6_try_regen_rndid(idev, tmpaddr);
1200         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1201         age = (now - ifp->tstamp) / HZ;
1202         tmp_valid_lft = min_t(__u32,
1203                               ifp->valid_lft,
1204                               idev->cnf.temp_valid_lft + age);
1205         tmp_prefered_lft = min_t(__u32,
1206                                  ifp->prefered_lft,
1207                                  idev->cnf.temp_prefered_lft + age -
1208                                  idev->cnf.max_desync_factor);
1209         tmp_plen = ifp->prefix_len;
1210         tmp_tstamp = ifp->tstamp;
1211         spin_unlock_bh(&ifp->lock);
1212
1213         regen_advance = idev->cnf.regen_max_retry *
1214                         idev->cnf.dad_transmits *
1215                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1216         write_unlock_bh(&idev->lock);
1217
1218         /* A temporary address is created only if this calculated Preferred
1219          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1220          * an implementation must not create a temporary address with a zero
1221          * Preferred Lifetime.
1222          * Use age calculation as in addrconf_verify to avoid unnecessary
1223          * temporary addresses being generated.
1224          */
1225         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1226         if (tmp_prefered_lft <= regen_advance + age) {
1227                 in6_ifa_put(ifp);
1228                 in6_dev_put(idev);
1229                 ret = -1;
1230                 goto out;
1231         }
1232
1233         addr_flags = IFA_F_TEMPORARY;
1234         /* set in addrconf_prefix_rcv() */
1235         if (ifp->flags & IFA_F_OPTIMISTIC)
1236                 addr_flags |= IFA_F_OPTIMISTIC;
1237
1238         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1239                             ipv6_addr_scope(&addr), addr_flags,
1240                             tmp_valid_lft, tmp_prefered_lft);
1241         if (IS_ERR(ift)) {
1242                 in6_ifa_put(ifp);
1243                 in6_dev_put(idev);
1244                 pr_info("%s: retry temporary address regeneration\n", __func__);
1245                 tmpaddr = &addr;
1246                 write_lock_bh(&idev->lock);
1247                 goto retry;
1248         }
1249
1250         spin_lock_bh(&ift->lock);
1251         ift->ifpub = ifp;
1252         ift->cstamp = now;
1253         ift->tstamp = tmp_tstamp;
1254         spin_unlock_bh(&ift->lock);
1255
1256         addrconf_dad_start(ift);
1257         in6_ifa_put(ift);
1258         in6_dev_put(idev);
1259 out:
1260         return ret;
1261 }
1262
1263 /*
1264  *      Choose an appropriate source address (RFC3484)
1265  */
1266 enum {
1267         IPV6_SADDR_RULE_INIT = 0,
1268         IPV6_SADDR_RULE_LOCAL,
1269         IPV6_SADDR_RULE_SCOPE,
1270         IPV6_SADDR_RULE_PREFERRED,
1271 #ifdef CONFIG_IPV6_MIP6
1272         IPV6_SADDR_RULE_HOA,
1273 #endif
1274         IPV6_SADDR_RULE_OIF,
1275         IPV6_SADDR_RULE_LABEL,
1276         IPV6_SADDR_RULE_PRIVACY,
1277         IPV6_SADDR_RULE_ORCHID,
1278         IPV6_SADDR_RULE_PREFIX,
1279 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1280         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1281 #endif
1282         IPV6_SADDR_RULE_MAX
1283 };
1284
1285 struct ipv6_saddr_score {
1286         int                     rule;
1287         int                     addr_type;
1288         struct inet6_ifaddr     *ifa;
1289         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1290         int                     scopedist;
1291         int                     matchlen;
1292 };
1293
1294 struct ipv6_saddr_dst {
1295         const struct in6_addr *addr;
1296         int ifindex;
1297         int scope;
1298         int label;
1299         unsigned int prefs;
1300 };
1301
1302 static inline int ipv6_saddr_preferred(int type)
1303 {
1304         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1305                 return 1;
1306         return 0;
1307 }
1308
1309 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1310 {
1311 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1312         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1313 #else
1314         return false;
1315 #endif
1316 }
1317
1318 static int ipv6_get_saddr_eval(struct net *net,
1319                                struct ipv6_saddr_score *score,
1320                                struct ipv6_saddr_dst *dst,
1321                                int i)
1322 {
1323         int ret;
1324
1325         if (i <= score->rule) {
1326                 switch (i) {
1327                 case IPV6_SADDR_RULE_SCOPE:
1328                         ret = score->scopedist;
1329                         break;
1330                 case IPV6_SADDR_RULE_PREFIX:
1331                         ret = score->matchlen;
1332                         break;
1333                 default:
1334                         ret = !!test_bit(i, score->scorebits);
1335                 }
1336                 goto out;
1337         }
1338
1339         switch (i) {
1340         case IPV6_SADDR_RULE_INIT:
1341                 /* Rule 0: remember if hiscore is not ready yet */
1342                 ret = !!score->ifa;
1343                 break;
1344         case IPV6_SADDR_RULE_LOCAL:
1345                 /* Rule 1: Prefer same address */
1346                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1347                 break;
1348         case IPV6_SADDR_RULE_SCOPE:
1349                 /* Rule 2: Prefer appropriate scope
1350                  *
1351                  *      ret
1352                  *       ^
1353                  *    -1 |  d 15
1354                  *    ---+--+-+---> scope
1355                  *       |
1356                  *       |             d is scope of the destination.
1357                  *  B-d  |  \
1358                  *       |   \      <- smaller scope is better if
1359                  *  B-15 |    \        if scope is enough for destination.
1360                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1361                  * d-C-1 | /
1362                  *       |/         <- greater is better
1363                  *   -C  /             if scope is not enough for destination.
1364                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1365                  *
1366                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1367                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1368                  * Assume B = 0 and we get C > 29.
1369                  */
1370                 ret = __ipv6_addr_src_scope(score->addr_type);
1371                 if (ret >= dst->scope)
1372                         ret = -ret;
1373                 else
1374                         ret -= 128;     /* 30 is enough */
1375                 score->scopedist = ret;
1376                 break;
1377         case IPV6_SADDR_RULE_PREFERRED:
1378             {
1379                 /* Rule 3: Avoid deprecated and optimistic addresses */
1380                 u8 avoid = IFA_F_DEPRECATED;
1381
1382                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1383                         avoid |= IFA_F_OPTIMISTIC;
1384                 ret = ipv6_saddr_preferred(score->addr_type) ||
1385                       !(score->ifa->flags & avoid);
1386                 break;
1387             }
1388 #ifdef CONFIG_IPV6_MIP6
1389         case IPV6_SADDR_RULE_HOA:
1390             {
1391                 /* Rule 4: Prefer home address */
1392                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1393                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1394                 break;
1395             }
1396 #endif
1397         case IPV6_SADDR_RULE_OIF:
1398                 /* Rule 5: Prefer outgoing interface */
1399                 ret = (!dst->ifindex ||
1400                        dst->ifindex == score->ifa->idev->dev->ifindex);
1401                 break;
1402         case IPV6_SADDR_RULE_LABEL:
1403                 /* Rule 6: Prefer matching label */
1404                 ret = ipv6_addr_label(net,
1405                                       &score->ifa->addr, score->addr_type,
1406                                       score->ifa->idev->dev->ifindex) == dst->label;
1407                 break;
1408         case IPV6_SADDR_RULE_PRIVACY:
1409             {
1410                 /* Rule 7: Prefer public address
1411                  * Note: prefer temporary address if use_tempaddr >= 2
1412                  */
1413                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1414                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1415                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1416                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1417                 break;
1418             }
1419         case IPV6_SADDR_RULE_ORCHID:
1420                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1421                  *          non-ORCHID vs non-ORCHID
1422                  */
1423                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1424                         ipv6_addr_orchid(dst->addr));
1425                 break;
1426         case IPV6_SADDR_RULE_PREFIX:
1427                 /* Rule 8: Use longest matching prefix */
1428                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1429                 if (ret > score->ifa->prefix_len)
1430                         ret = score->ifa->prefix_len;
1431                 score->matchlen = ret;
1432                 break;
1433 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1434         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1435                 /* Optimistic addresses still have lower precedence than other
1436                  * preferred addresses.
1437                  */
1438                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1439                 break;
1440 #endif
1441         default:
1442                 ret = 0;
1443         }
1444
1445         if (ret)
1446                 __set_bit(i, score->scorebits);
1447         score->rule = i;
1448 out:
1449         return ret;
1450 }
1451
1452 static int __ipv6_dev_get_saddr(struct net *net,
1453                                 struct ipv6_saddr_dst *dst,
1454                                 struct inet6_dev *idev,
1455                                 struct ipv6_saddr_score *scores,
1456                                 int hiscore_idx)
1457 {
1458         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1459
1460         read_lock_bh(&idev->lock);
1461         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1462                 int i;
1463
1464                 /*
1465                  * - Tentative Address (RFC2462 section 5.4)
1466                  *  - A tentative address is not considered
1467                  *    "assigned to an interface" in the traditional
1468                  *    sense, unless it is also flagged as optimistic.
1469                  * - Candidate Source Address (section 4)
1470                  *  - In any case, anycast addresses, multicast
1471                  *    addresses, and the unspecified address MUST
1472                  *    NOT be included in a candidate set.
1473                  */
1474                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1475                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1476                         continue;
1477
1478                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1479
1480                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1481                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1482                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1483                                             idev->dev->name);
1484                         continue;
1485                 }
1486
1487                 score->rule = -1;
1488                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1489
1490                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1491                         int minihiscore, miniscore;
1492
1493                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1494                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1495
1496                         if (minihiscore > miniscore) {
1497                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1498                                     score->scopedist > 0) {
1499                                         /*
1500                                          * special case:
1501                                          * each remaining entry
1502                                          * has too small (not enough)
1503                                          * scope, because ifa entries
1504                                          * are sorted by their scope
1505                                          * values.
1506                                          */
1507                                         goto out;
1508                                 }
1509                                 break;
1510                         } else if (minihiscore < miniscore) {
1511                                 if (hiscore->ifa)
1512                                         in6_ifa_put(hiscore->ifa);
1513
1514                                 in6_ifa_hold(score->ifa);
1515
1516                                 swap(hiscore, score);
1517                                 hiscore_idx = 1 - hiscore_idx;
1518
1519                                 /* restore our iterator */
1520                                 score->ifa = hiscore->ifa;
1521
1522                                 break;
1523                         }
1524                 }
1525         }
1526 out:
1527         read_unlock_bh(&idev->lock);
1528         return hiscore_idx;
1529 }
1530
1531 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1532                        const struct in6_addr *daddr, unsigned int prefs,
1533                        struct in6_addr *saddr)
1534 {
1535         struct ipv6_saddr_score scores[2], *hiscore;
1536         struct ipv6_saddr_dst dst;
1537         struct inet6_dev *idev;
1538         struct net_device *dev;
1539         int dst_type;
1540         bool use_oif_addr = false;
1541         int hiscore_idx = 0;
1542
1543         dst_type = __ipv6_addr_type(daddr);
1544         dst.addr = daddr;
1545         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1546         dst.scope = __ipv6_addr_src_scope(dst_type);
1547         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1548         dst.prefs = prefs;
1549
1550         scores[hiscore_idx].rule = -1;
1551         scores[hiscore_idx].ifa = NULL;
1552
1553         rcu_read_lock();
1554
1555         /* Candidate Source Address (section 4)
1556          *  - multicast and link-local destination address,
1557          *    the set of candidate source address MUST only
1558          *    include addresses assigned to interfaces
1559          *    belonging to the same link as the outgoing
1560          *    interface.
1561          * (- For site-local destination addresses, the
1562          *    set of candidate source addresses MUST only
1563          *    include addresses assigned to interfaces
1564          *    belonging to the same site as the outgoing
1565          *    interface.)
1566          *  - "It is RECOMMENDED that the candidate source addresses
1567          *    be the set of unicast addresses assigned to the
1568          *    interface that will be used to send to the destination
1569          *    (the 'outgoing' interface)." (RFC 6724)
1570          */
1571         if (dst_dev) {
1572                 idev = __in6_dev_get(dst_dev);
1573                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1574                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1575                     (idev && idev->cnf.use_oif_addrs_only)) {
1576                         use_oif_addr = true;
1577                 }
1578         }
1579
1580         if (use_oif_addr) {
1581                 if (idev)
1582                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1583         } else {
1584                 for_each_netdev_rcu(net, dev) {
1585                         idev = __in6_dev_get(dev);
1586                         if (!idev)
1587                                 continue;
1588                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1589                 }
1590         }
1591         rcu_read_unlock();
1592
1593         hiscore = &scores[hiscore_idx];
1594         if (!hiscore->ifa)
1595                 return -EADDRNOTAVAIL;
1596
1597         *saddr = hiscore->ifa->addr;
1598         in6_ifa_put(hiscore->ifa);
1599         return 0;
1600 }
1601 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1602
1603 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1604                       u32 banned_flags)
1605 {
1606         struct inet6_ifaddr *ifp;
1607         int err = -EADDRNOTAVAIL;
1608
1609         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1610                 if (ifp->scope > IFA_LINK)
1611                         break;
1612                 if (ifp->scope == IFA_LINK &&
1613                     !(ifp->flags & banned_flags)) {
1614                         *addr = ifp->addr;
1615                         err = 0;
1616                         break;
1617                 }
1618         }
1619         return err;
1620 }
1621
1622 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1623                     u32 banned_flags)
1624 {
1625         struct inet6_dev *idev;
1626         int err = -EADDRNOTAVAIL;
1627
1628         rcu_read_lock();
1629         idev = __in6_dev_get(dev);
1630         if (idev) {
1631                 read_lock_bh(&idev->lock);
1632                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1633                 read_unlock_bh(&idev->lock);
1634         }
1635         rcu_read_unlock();
1636         return err;
1637 }
1638
1639 static int ipv6_count_addresses(struct inet6_dev *idev)
1640 {
1641         int cnt = 0;
1642         struct inet6_ifaddr *ifp;
1643
1644         read_lock_bh(&idev->lock);
1645         list_for_each_entry(ifp, &idev->addr_list, if_list)
1646                 cnt++;
1647         read_unlock_bh(&idev->lock);
1648         return cnt;
1649 }
1650
1651 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1652                   const struct net_device *dev, int strict)
1653 {
1654         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1655 }
1656 EXPORT_SYMBOL(ipv6_chk_addr);
1657
1658 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1659                             const struct net_device *dev, int strict,
1660                             u32 banned_flags)
1661 {
1662         struct inet6_ifaddr *ifp;
1663         unsigned int hash = inet6_addr_hash(addr);
1664         u32 ifp_flags;
1665
1666         rcu_read_lock_bh();
1667         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1668                 if (!net_eq(dev_net(ifp->idev->dev), net))
1669                         continue;
1670                 /* Decouple optimistic from tentative for evaluation here.
1671                  * Ban optimistic addresses explicitly, when required.
1672                  */
1673                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1674                             ? (ifp->flags&~IFA_F_TENTATIVE)
1675                             : ifp->flags;
1676                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1677                     !(ifp_flags&banned_flags) &&
1678                     (!dev || ifp->idev->dev == dev ||
1679                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1680                         rcu_read_unlock_bh();
1681                         return 1;
1682                 }
1683         }
1684
1685         rcu_read_unlock_bh();
1686         return 0;
1687 }
1688 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1689
1690 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1691                                struct net_device *dev)
1692 {
1693         unsigned int hash = inet6_addr_hash(addr);
1694         struct inet6_ifaddr *ifp;
1695
1696         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1697                 if (!net_eq(dev_net(ifp->idev->dev), net))
1698                         continue;
1699                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1700                         if (!dev || ifp->idev->dev == dev)
1701                                 return true;
1702                 }
1703         }
1704         return false;
1705 }
1706
1707 /* Compares an address/prefix_len with addresses on device @dev.
1708  * If one is found it returns true.
1709  */
1710 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1711         const unsigned int prefix_len, struct net_device *dev)
1712 {
1713         struct inet6_dev *idev;
1714         struct inet6_ifaddr *ifa;
1715         bool ret = false;
1716
1717         rcu_read_lock();
1718         idev = __in6_dev_get(dev);
1719         if (idev) {
1720                 read_lock_bh(&idev->lock);
1721                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1722                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1723                         if (ret)
1724                                 break;
1725                 }
1726                 read_unlock_bh(&idev->lock);
1727         }
1728         rcu_read_unlock();
1729
1730         return ret;
1731 }
1732 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1733
1734 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1735 {
1736         struct inet6_dev *idev;
1737         struct inet6_ifaddr *ifa;
1738         int     onlink;
1739
1740         onlink = 0;
1741         rcu_read_lock();
1742         idev = __in6_dev_get(dev);
1743         if (idev) {
1744                 read_lock_bh(&idev->lock);
1745                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1746                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1747                                                    ifa->prefix_len);
1748                         if (onlink)
1749                                 break;
1750                 }
1751                 read_unlock_bh(&idev->lock);
1752         }
1753         rcu_read_unlock();
1754         return onlink;
1755 }
1756 EXPORT_SYMBOL(ipv6_chk_prefix);
1757
1758 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1759                                      struct net_device *dev, int strict)
1760 {
1761         struct inet6_ifaddr *ifp, *result = NULL;
1762         unsigned int hash = inet6_addr_hash(addr);
1763
1764         rcu_read_lock_bh();
1765         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1766                 if (!net_eq(dev_net(ifp->idev->dev), net))
1767                         continue;
1768                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1769                         if (!dev || ifp->idev->dev == dev ||
1770                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1771                                 result = ifp;
1772                                 in6_ifa_hold(ifp);
1773                                 break;
1774                         }
1775                 }
1776         }
1777         rcu_read_unlock_bh();
1778
1779         return result;
1780 }
1781
1782 /* Gets referenced address, destroys ifaddr */
1783
1784 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1785 {
1786         if (ifp->flags&IFA_F_TEMPORARY) {
1787                 struct inet6_ifaddr *ifpub;
1788                 spin_lock_bh(&ifp->lock);
1789                 ifpub = ifp->ifpub;
1790                 if (ifpub) {
1791                         in6_ifa_hold(ifpub);
1792                         spin_unlock_bh(&ifp->lock);
1793                         ipv6_create_tempaddr(ifpub, ifp);
1794                         in6_ifa_put(ifpub);
1795                 } else {
1796                         spin_unlock_bh(&ifp->lock);
1797                 }
1798                 ipv6_del_addr(ifp);
1799         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1800                 spin_lock_bh(&ifp->lock);
1801                 addrconf_del_dad_work(ifp);
1802                 ifp->flags |= IFA_F_TENTATIVE;
1803                 if (dad_failed)
1804                         ifp->flags |= IFA_F_DADFAILED;
1805                 spin_unlock_bh(&ifp->lock);
1806                 if (dad_failed)
1807                         ipv6_ifa_notify(0, ifp);
1808                 in6_ifa_put(ifp);
1809         } else {
1810                 ipv6_del_addr(ifp);
1811         }
1812 }
1813
1814 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1815 {
1816         int err = -ENOENT;
1817
1818         spin_lock_bh(&ifp->lock);
1819         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1820                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1821                 err = 0;
1822         }
1823         spin_unlock_bh(&ifp->lock);
1824
1825         return err;
1826 }
1827
1828 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1829 {
1830         struct in6_addr addr;
1831         struct inet6_dev *idev = ifp->idev;
1832         struct net *net = dev_net(ifp->idev->dev);
1833
1834         if (addrconf_dad_end(ifp)) {
1835                 in6_ifa_put(ifp);
1836                 return;
1837         }
1838
1839         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1840                              ifp->idev->dev->name, &ifp->addr);
1841
1842         spin_lock_bh(&ifp->lock);
1843
1844         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1845                 int scope = ifp->scope;
1846                 u32 flags = ifp->flags;
1847                 struct in6_addr new_addr;
1848                 struct inet6_ifaddr *ifp2;
1849                 u32 valid_lft, preferred_lft;
1850                 int pfxlen = ifp->prefix_len;
1851                 int retries = ifp->stable_privacy_retry + 1;
1852
1853                 if (retries > net->ipv6.sysctl.idgen_retries) {
1854                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1855                                              ifp->idev->dev->name);
1856                         goto errdad;
1857                 }
1858
1859                 new_addr = ifp->addr;
1860                 if (ipv6_generate_stable_address(&new_addr, retries,
1861                                                  idev))
1862                         goto errdad;
1863
1864                 valid_lft = ifp->valid_lft;
1865                 preferred_lft = ifp->prefered_lft;
1866
1867                 spin_unlock_bh(&ifp->lock);
1868
1869                 if (idev->cnf.max_addresses &&
1870                     ipv6_count_addresses(idev) >=
1871                     idev->cnf.max_addresses)
1872                         goto lock_errdad;
1873
1874                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1875                                      ifp->idev->dev->name);
1876
1877                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1878                                      scope, flags, valid_lft,
1879                                      preferred_lft);
1880                 if (IS_ERR(ifp2))
1881                         goto lock_errdad;
1882
1883                 spin_lock_bh(&ifp2->lock);
1884                 ifp2->stable_privacy_retry = retries;
1885                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1886                 spin_unlock_bh(&ifp2->lock);
1887
1888                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1889                 in6_ifa_put(ifp2);
1890 lock_errdad:
1891                 spin_lock_bh(&ifp->lock);
1892         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1893                 addr.s6_addr32[0] = htonl(0xfe800000);
1894                 addr.s6_addr32[1] = 0;
1895
1896                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1897                     ipv6_addr_equal(&ifp->addr, &addr)) {
1898                         /* DAD failed for link-local based on MAC address */
1899                         idev->cnf.disable_ipv6 = 1;
1900
1901                         pr_info("%s: IPv6 being disabled!\n",
1902                                 ifp->idev->dev->name);
1903                 }
1904         }
1905
1906 errdad:
1907         /* transition from _POSTDAD to _ERRDAD */
1908         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1909         spin_unlock_bh(&ifp->lock);
1910
1911         addrconf_mod_dad_work(ifp, 0);
1912         in6_ifa_put(ifp);
1913 }
1914
1915 /* Join to solicited addr multicast group.
1916  * caller must hold RTNL */
1917 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1918 {
1919         struct in6_addr maddr;
1920
1921         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1922                 return;
1923
1924         addrconf_addr_solict_mult(addr, &maddr);
1925         ipv6_dev_mc_inc(dev, &maddr);
1926 }
1927
1928 /* caller must hold RTNL */
1929 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1930 {
1931         struct in6_addr maddr;
1932
1933         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1934                 return;
1935
1936         addrconf_addr_solict_mult(addr, &maddr);
1937         __ipv6_dev_mc_dec(idev, &maddr);
1938 }
1939
1940 /* caller must hold RTNL */
1941 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1942 {
1943         struct in6_addr addr;
1944
1945         if (ifp->prefix_len >= 127) /* RFC 6164 */
1946                 return;
1947         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1948         if (ipv6_addr_any(&addr))
1949                 return;
1950         __ipv6_dev_ac_inc(ifp->idev, &addr);
1951 }
1952
1953 /* caller must hold RTNL */
1954 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1955 {
1956         struct in6_addr addr;
1957
1958         if (ifp->prefix_len >= 127) /* RFC 6164 */
1959                 return;
1960         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1961         if (ipv6_addr_any(&addr))
1962                 return;
1963         __ipv6_dev_ac_dec(ifp->idev, &addr);
1964 }
1965
1966 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1967 {
1968         if (dev->addr_len != IEEE802154_ADDR_LEN)
1969                 return -1;
1970         memcpy(eui, dev->dev_addr, 8);
1971         eui[0] ^= 2;
1972         return 0;
1973 }
1974
1975 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1976 {
1977         union fwnet_hwaddr *ha;
1978
1979         if (dev->addr_len != FWNET_ALEN)
1980                 return -1;
1981
1982         ha = (union fwnet_hwaddr *)dev->dev_addr;
1983
1984         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1985         eui[0] ^= 2;
1986         return 0;
1987 }
1988
1989 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1990 {
1991         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1992         if (dev->addr_len != ARCNET_ALEN)
1993                 return -1;
1994         memset(eui, 0, 7);
1995         eui[7] = *(u8 *)dev->dev_addr;
1996         return 0;
1997 }
1998
1999 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2000 {
2001         if (dev->addr_len != INFINIBAND_ALEN)
2002                 return -1;
2003         memcpy(eui, dev->dev_addr + 12, 8);
2004         eui[0] |= 2;
2005         return 0;
2006 }
2007
2008 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2009 {
2010         if (addr == 0)
2011                 return -1;
2012         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2013                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2014                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2015                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2016                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2017                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2018         eui[1] = 0;
2019         eui[2] = 0x5E;
2020         eui[3] = 0xFE;
2021         memcpy(eui + 4, &addr, 4);
2022         return 0;
2023 }
2024
2025 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2026 {
2027         if (dev->priv_flags & IFF_ISATAP)
2028                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2029         return -1;
2030 }
2031
2032 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2033 {
2034         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2035 }
2036
2037 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2038 {
2039         memcpy(eui, dev->perm_addr, 3);
2040         memcpy(eui + 5, dev->perm_addr + 3, 3);
2041         eui[3] = 0xFF;
2042         eui[4] = 0xFE;
2043         eui[0] ^= 2;
2044         return 0;
2045 }
2046
2047 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2048 {
2049         switch (dev->type) {
2050         case ARPHRD_ETHER:
2051         case ARPHRD_FDDI:
2052                 return addrconf_ifid_eui48(eui, dev);
2053         case ARPHRD_ARCNET:
2054                 return addrconf_ifid_arcnet(eui, dev);
2055         case ARPHRD_INFINIBAND:
2056                 return addrconf_ifid_infiniband(eui, dev);
2057         case ARPHRD_SIT:
2058                 return addrconf_ifid_sit(eui, dev);
2059         case ARPHRD_IPGRE:
2060                 return addrconf_ifid_gre(eui, dev);
2061         case ARPHRD_6LOWPAN:
2062         case ARPHRD_IEEE802154:
2063                 return addrconf_ifid_eui64(eui, dev);
2064         case ARPHRD_IEEE1394:
2065                 return addrconf_ifid_ieee1394(eui, dev);
2066         case ARPHRD_TUNNEL6:
2067                 return addrconf_ifid_ip6tnl(eui, dev);
2068         }
2069         return -1;
2070 }
2071
2072 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2073 {
2074         int err = -1;
2075         struct inet6_ifaddr *ifp;
2076
2077         read_lock_bh(&idev->lock);
2078         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2079                 if (ifp->scope > IFA_LINK)
2080                         break;
2081                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2082                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2083                         err = 0;
2084                         break;
2085                 }
2086         }
2087         read_unlock_bh(&idev->lock);
2088         return err;
2089 }
2090
2091 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2092 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2093 {
2094 regen:
2095         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2096         idev->rndid[0] &= ~0x02;
2097
2098         /*
2099          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2100          * check if generated address is not inappropriate
2101          *
2102          *  - Reserved subnet anycast (RFC 2526)
2103          *      11111101 11....11 1xxxxxxx
2104          *  - ISATAP (RFC4214) 6.1
2105          *      00-00-5E-FE-xx-xx-xx-xx
2106          *  - value 0
2107          *  - XXX: already assigned to an address on the device
2108          */
2109         if (idev->rndid[0] == 0xfd &&
2110             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2111             (idev->rndid[7]&0x80))
2112                 goto regen;
2113         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2114                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2115                         goto regen;
2116                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2117                         goto regen;
2118         }
2119 }
2120
2121 static void ipv6_regen_rndid(unsigned long data)
2122 {
2123         struct inet6_dev *idev = (struct inet6_dev *) data;
2124         unsigned long expires;
2125
2126         rcu_read_lock_bh();
2127         write_lock_bh(&idev->lock);
2128
2129         if (idev->dead)
2130                 goto out;
2131
2132         __ipv6_regen_rndid(idev);
2133
2134         expires = jiffies +
2135                 idev->cnf.temp_prefered_lft * HZ -
2136                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2137                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2138                 idev->cnf.max_desync_factor * HZ;
2139         if (time_before(expires, jiffies)) {
2140                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2141                         __func__, idev->dev->name);
2142                 goto out;
2143         }
2144
2145         if (!mod_timer(&idev->regen_timer, expires))
2146                 in6_dev_hold(idev);
2147
2148 out:
2149         write_unlock_bh(&idev->lock);
2150         rcu_read_unlock_bh();
2151         in6_dev_put(idev);
2152 }
2153
2154 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2155 {
2156         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2157                 __ipv6_regen_rndid(idev);
2158 }
2159
2160 /*
2161  *      Add prefix route.
2162  */
2163
2164 static void
2165 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2166                       unsigned long expires, u32 flags)
2167 {
2168         struct fib6_config cfg = {
2169                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2170                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2171                 .fc_ifindex = dev->ifindex,
2172                 .fc_expires = expires,
2173                 .fc_dst_len = plen,
2174                 .fc_flags = RTF_UP | flags,
2175                 .fc_nlinfo.nl_net = dev_net(dev),
2176                 .fc_protocol = RTPROT_KERNEL,
2177         };
2178
2179         cfg.fc_dst = *pfx;
2180
2181         /* Prevent useless cloning on PtP SIT.
2182            This thing is done here expecting that the whole
2183            class of non-broadcast devices need not cloning.
2184          */
2185 #if IS_ENABLED(CONFIG_IPV6_SIT)
2186         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2187                 cfg.fc_flags |= RTF_NONEXTHOP;
2188 #endif
2189
2190         ip6_route_add(&cfg);
2191 }
2192
2193
2194 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2195                                                   int plen,
2196                                                   const struct net_device *dev,
2197                                                   u32 flags, u32 noflags)
2198 {
2199         struct fib6_node *fn;
2200         struct rt6_info *rt = NULL;
2201         struct fib6_table *table;
2202         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2203
2204         table = fib6_get_table(dev_net(dev), tb_id);
2205         if (!table)
2206                 return NULL;
2207
2208         read_lock_bh(&table->tb6_lock);
2209         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2210         if (!fn)
2211                 goto out;
2212
2213         noflags |= RTF_CACHE;
2214         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2215                 if (rt->dst.dev->ifindex != dev->ifindex)
2216                         continue;
2217                 if ((rt->rt6i_flags & flags) != flags)
2218                         continue;
2219                 if ((rt->rt6i_flags & noflags) != 0)
2220                         continue;
2221                 dst_hold(&rt->dst);
2222                 break;
2223         }
2224 out:
2225         read_unlock_bh(&table->tb6_lock);
2226         return rt;
2227 }
2228
2229
2230 /* Create "default" multicast route to the interface */
2231
2232 static void addrconf_add_mroute(struct net_device *dev)
2233 {
2234         struct fib6_config cfg = {
2235                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2236                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2237                 .fc_ifindex = dev->ifindex,
2238                 .fc_dst_len = 8,
2239                 .fc_flags = RTF_UP,
2240                 .fc_nlinfo.nl_net = dev_net(dev),
2241         };
2242
2243         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2244
2245         ip6_route_add(&cfg);
2246 }
2247
2248 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2249 {
2250         struct inet6_dev *idev;
2251
2252         ASSERT_RTNL();
2253
2254         idev = ipv6_find_idev(dev);
2255         if (!idev)
2256                 return ERR_PTR(-ENOBUFS);
2257
2258         if (idev->cnf.disable_ipv6)
2259                 return ERR_PTR(-EACCES);
2260
2261         /* Add default multicast route */
2262         if (!(dev->flags & IFF_LOOPBACK))
2263                 addrconf_add_mroute(dev);
2264
2265         return idev;
2266 }
2267
2268 static void manage_tempaddrs(struct inet6_dev *idev,
2269                              struct inet6_ifaddr *ifp,
2270                              __u32 valid_lft, __u32 prefered_lft,
2271                              bool create, unsigned long now)
2272 {
2273         u32 flags;
2274         struct inet6_ifaddr *ift;
2275
2276         read_lock_bh(&idev->lock);
2277         /* update all temporary addresses in the list */
2278         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2279                 int age, max_valid, max_prefered;
2280
2281                 if (ifp != ift->ifpub)
2282                         continue;
2283
2284                 /* RFC 4941 section 3.3:
2285                  * If a received option will extend the lifetime of a public
2286                  * address, the lifetimes of temporary addresses should
2287                  * be extended, subject to the overall constraint that no
2288                  * temporary addresses should ever remain "valid" or "preferred"
2289                  * for a time longer than (TEMP_VALID_LIFETIME) or
2290                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2291                  */
2292                 age = (now - ift->cstamp) / HZ;
2293                 max_valid = idev->cnf.temp_valid_lft - age;
2294                 if (max_valid < 0)
2295                         max_valid = 0;
2296
2297                 max_prefered = idev->cnf.temp_prefered_lft -
2298                                idev->cnf.max_desync_factor - age;
2299                 if (max_prefered < 0)
2300                         max_prefered = 0;
2301
2302                 if (valid_lft > max_valid)
2303                         valid_lft = max_valid;
2304
2305                 if (prefered_lft > max_prefered)
2306                         prefered_lft = max_prefered;
2307
2308                 spin_lock(&ift->lock);
2309                 flags = ift->flags;
2310                 ift->valid_lft = valid_lft;
2311                 ift->prefered_lft = prefered_lft;
2312                 ift->tstamp = now;
2313                 if (prefered_lft > 0)
2314                         ift->flags &= ~IFA_F_DEPRECATED;
2315
2316                 spin_unlock(&ift->lock);
2317                 if (!(flags&IFA_F_TENTATIVE))
2318                         ipv6_ifa_notify(0, ift);
2319         }
2320
2321         if ((create || list_empty(&idev->tempaddr_list)) &&
2322             idev->cnf.use_tempaddr > 0) {
2323                 /* When a new public address is created as described
2324                  * in [ADDRCONF], also create a new temporary address.
2325                  * Also create a temporary address if it's enabled but
2326                  * no temporary address currently exists.
2327                  */
2328                 read_unlock_bh(&idev->lock);
2329                 ipv6_create_tempaddr(ifp, NULL);
2330         } else {
2331                 read_unlock_bh(&idev->lock);
2332         }
2333 }
2334
2335 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2336 {
2337         struct prefix_info *pinfo;
2338         __u32 valid_lft;
2339         __u32 prefered_lft;
2340         int addr_type;
2341         u32 addr_flags = 0;
2342         struct inet6_dev *in6_dev;
2343         struct net *net = dev_net(dev);
2344
2345         pinfo = (struct prefix_info *) opt;
2346
2347         if (len < sizeof(struct prefix_info)) {
2348                 ADBG("addrconf: prefix option too short\n");
2349                 return;
2350         }
2351
2352         /*
2353          *      Validation checks ([ADDRCONF], page 19)
2354          */
2355
2356         addr_type = ipv6_addr_type(&pinfo->prefix);
2357
2358         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2359                 return;
2360
2361         valid_lft = ntohl(pinfo->valid);
2362         prefered_lft = ntohl(pinfo->prefered);
2363
2364         if (prefered_lft > valid_lft) {
2365                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2366                 return;
2367         }
2368
2369         in6_dev = in6_dev_get(dev);
2370
2371         if (!in6_dev) {
2372                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2373                                     dev->name);
2374                 return;
2375         }
2376
2377         /*
2378          *      Two things going on here:
2379          *      1) Add routes for on-link prefixes
2380          *      2) Configure prefixes with the auto flag set
2381          */
2382
2383         if (pinfo->onlink) {
2384                 struct rt6_info *rt;
2385                 unsigned long rt_expires;
2386
2387                 /* Avoid arithmetic overflow. Really, we could
2388                  * save rt_expires in seconds, likely valid_lft,
2389                  * but it would require division in fib gc, that it
2390                  * not good.
2391                  */
2392                 if (HZ > USER_HZ)
2393                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2394                 else
2395                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2396
2397                 if (addrconf_finite_timeout(rt_expires))
2398                         rt_expires *= HZ;
2399
2400                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2401                                                pinfo->prefix_len,
2402                                                dev,
2403                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2404                                                RTF_GATEWAY | RTF_DEFAULT);
2405
2406                 if (rt) {
2407                         /* Autoconf prefix route */
2408                         if (valid_lft == 0) {
2409                                 ip6_del_rt(rt);
2410                                 rt = NULL;
2411                         } else if (addrconf_finite_timeout(rt_expires)) {
2412                                 /* not infinity */
2413                                 rt6_set_expires(rt, jiffies + rt_expires);
2414                         } else {
2415                                 rt6_clean_expires(rt);
2416                         }
2417                 } else if (valid_lft) {
2418                         clock_t expires = 0;
2419                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2420                         if (addrconf_finite_timeout(rt_expires)) {
2421                                 /* not infinity */
2422                                 flags |= RTF_EXPIRES;
2423                                 expires = jiffies_to_clock_t(rt_expires);
2424                         }
2425                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2426                                               dev, expires, flags);
2427                 }
2428                 ip6_rt_put(rt);
2429         }
2430
2431         /* Try to figure out our local address for this prefix */
2432
2433         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2434                 struct inet6_ifaddr *ifp;
2435                 struct in6_addr addr;
2436                 int create = 0, update_lft = 0;
2437                 bool tokenized = false;
2438
2439                 if (pinfo->prefix_len == 64) {
2440                         memcpy(&addr, &pinfo->prefix, 8);
2441
2442                         if (!ipv6_addr_any(&in6_dev->token)) {
2443                                 read_lock_bh(&in6_dev->lock);
2444                                 memcpy(addr.s6_addr + 8,
2445                                        in6_dev->token.s6_addr + 8, 8);
2446                                 read_unlock_bh(&in6_dev->lock);
2447                                 tokenized = true;
2448                         } else if (in6_dev->addr_gen_mode ==
2449                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2450                                    !ipv6_generate_stable_address(&addr, 0,
2451                                                                  in6_dev)) {
2452                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2453                                 goto ok;
2454                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2455                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2456                                 in6_dev_put(in6_dev);
2457                                 return;
2458                         }
2459                         goto ok;
2460                 }
2461                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2462                                     pinfo->prefix_len);
2463                 in6_dev_put(in6_dev);
2464                 return;
2465
2466 ok:
2467
2468                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2469
2470                 if (!ifp && valid_lft) {
2471                         int max_addresses = in6_dev->cnf.max_addresses;
2472
2473 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2474                         if (in6_dev->cnf.optimistic_dad &&
2475                             !net->ipv6.devconf_all->forwarding && sllao)
2476                                 addr_flags |= IFA_F_OPTIMISTIC;
2477 #endif
2478
2479                         /* Do not allow to create too much of autoconfigured
2480                          * addresses; this would be too easy way to crash kernel.
2481                          */
2482                         if (!max_addresses ||
2483                             ipv6_count_addresses(in6_dev) < max_addresses)
2484                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2485                                                     pinfo->prefix_len,
2486                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2487                                                     addr_flags, valid_lft,
2488                                                     prefered_lft);
2489
2490                         if (IS_ERR_OR_NULL(ifp)) {
2491                                 in6_dev_put(in6_dev);
2492                                 return;
2493                         }
2494
2495                         update_lft = 0;
2496                         create = 1;
2497                         spin_lock_bh(&ifp->lock);
2498                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2499                         ifp->cstamp = jiffies;
2500                         ifp->tokenized = tokenized;
2501                         spin_unlock_bh(&ifp->lock);
2502                         addrconf_dad_start(ifp);
2503                 }
2504
2505                 if (ifp) {
2506                         u32 flags;
2507                         unsigned long now;
2508                         u32 stored_lft;
2509
2510                         /* update lifetime (RFC2462 5.5.3 e) */
2511                         spin_lock_bh(&ifp->lock);
2512                         now = jiffies;
2513                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2514                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2515                         else
2516                                 stored_lft = 0;
2517                         if (!update_lft && !create && stored_lft) {
2518                                 const u32 minimum_lft = min_t(u32,
2519                                         stored_lft, MIN_VALID_LIFETIME);
2520                                 valid_lft = max(valid_lft, minimum_lft);
2521
2522                                 /* RFC4862 Section 5.5.3e:
2523                                  * "Note that the preferred lifetime of the
2524                                  *  corresponding address is always reset to
2525                                  *  the Preferred Lifetime in the received
2526                                  *  Prefix Information option, regardless of
2527                                  *  whether the valid lifetime is also reset or
2528                                  *  ignored."
2529                                  *
2530                                  * So we should always update prefered_lft here.
2531                                  */
2532                                 update_lft = 1;
2533                         }
2534
2535                         if (update_lft) {
2536                                 ifp->valid_lft = valid_lft;
2537                                 ifp->prefered_lft = prefered_lft;
2538                                 ifp->tstamp = now;
2539                                 flags = ifp->flags;
2540                                 ifp->flags &= ~IFA_F_DEPRECATED;
2541                                 spin_unlock_bh(&ifp->lock);
2542
2543                                 if (!(flags&IFA_F_TENTATIVE))
2544                                         ipv6_ifa_notify(0, ifp);
2545                         } else
2546                                 spin_unlock_bh(&ifp->lock);
2547
2548                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2549                                          create, now);
2550
2551                         in6_ifa_put(ifp);
2552                         addrconf_verify();
2553                 }
2554         }
2555         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2556         in6_dev_put(in6_dev);
2557 }
2558
2559 /*
2560  *      Set destination address.
2561  *      Special case for SIT interfaces where we create a new "virtual"
2562  *      device.
2563  */
2564 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2565 {
2566         struct in6_ifreq ireq;
2567         struct net_device *dev;
2568         int err = -EINVAL;
2569
2570         rtnl_lock();
2571
2572         err = -EFAULT;
2573         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2574                 goto err_exit;
2575
2576         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2577
2578         err = -ENODEV;
2579         if (!dev)
2580                 goto err_exit;
2581
2582 #if IS_ENABLED(CONFIG_IPV6_SIT)
2583         if (dev->type == ARPHRD_SIT) {
2584                 const struct net_device_ops *ops = dev->netdev_ops;
2585                 struct ifreq ifr;
2586                 struct ip_tunnel_parm p;
2587
2588                 err = -EADDRNOTAVAIL;
2589                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2590                         goto err_exit;
2591
2592                 memset(&p, 0, sizeof(p));
2593                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2594                 p.iph.saddr = 0;
2595                 p.iph.version = 4;
2596                 p.iph.ihl = 5;
2597                 p.iph.protocol = IPPROTO_IPV6;
2598                 p.iph.ttl = 64;
2599                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2600
2601                 if (ops->ndo_do_ioctl) {
2602                         mm_segment_t oldfs = get_fs();
2603
2604                         set_fs(KERNEL_DS);
2605                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2606                         set_fs(oldfs);
2607                 } else
2608                         err = -EOPNOTSUPP;
2609
2610                 if (err == 0) {
2611                         err = -ENOBUFS;
2612                         dev = __dev_get_by_name(net, p.name);
2613                         if (!dev)
2614                                 goto err_exit;
2615                         err = dev_open(dev);
2616                 }
2617         }
2618 #endif
2619
2620 err_exit:
2621         rtnl_unlock();
2622         return err;
2623 }
2624
2625 static int ipv6_mc_config(struct sock *sk, bool join,
2626                           const struct in6_addr *addr, int ifindex)
2627 {
2628         int ret;
2629
2630         ASSERT_RTNL();
2631
2632         lock_sock(sk);
2633         if (join)
2634                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2635         else
2636                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2637         release_sock(sk);
2638
2639         return ret;
2640 }
2641
2642 /*
2643  *      Manual configuration of address on an interface
2644  */
2645 static int inet6_addr_add(struct net *net, int ifindex,
2646                           const struct in6_addr *pfx,
2647                           const struct in6_addr *peer_pfx,
2648                           unsigned int plen, __u32 ifa_flags,
2649                           __u32 prefered_lft, __u32 valid_lft)
2650 {
2651         struct inet6_ifaddr *ifp;
2652         struct inet6_dev *idev;
2653         struct net_device *dev;
2654         unsigned long timeout;
2655         clock_t expires;
2656         int scope;
2657         u32 flags;
2658
2659         ASSERT_RTNL();
2660
2661         if (plen > 128)
2662                 return -EINVAL;
2663
2664         /* check the lifetime */
2665         if (!valid_lft || prefered_lft > valid_lft)
2666                 return -EINVAL;
2667
2668         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2669                 return -EINVAL;
2670
2671         dev = __dev_get_by_index(net, ifindex);
2672         if (!dev)
2673                 return -ENODEV;
2674
2675         idev = addrconf_add_dev(dev);
2676         if (IS_ERR(idev))
2677                 return PTR_ERR(idev);
2678
2679         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2680                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2681                                          true, pfx, ifindex);
2682
2683                 if (ret < 0)
2684                         return ret;
2685         }
2686
2687         scope = ipv6_addr_scope(pfx);
2688
2689         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2690         if (addrconf_finite_timeout(timeout)) {
2691                 expires = jiffies_to_clock_t(timeout * HZ);
2692                 valid_lft = timeout;
2693                 flags = RTF_EXPIRES;
2694         } else {
2695                 expires = 0;
2696                 flags = 0;
2697                 ifa_flags |= IFA_F_PERMANENT;
2698         }
2699
2700         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2701         if (addrconf_finite_timeout(timeout)) {
2702                 if (timeout == 0)
2703                         ifa_flags |= IFA_F_DEPRECATED;
2704                 prefered_lft = timeout;
2705         }
2706
2707         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2708                             valid_lft, prefered_lft);
2709
2710         if (!IS_ERR(ifp)) {
2711                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2712                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2713                                               expires, flags);
2714                 }
2715
2716                 /*
2717                  * Note that section 3.1 of RFC 4429 indicates
2718                  * that the Optimistic flag should not be set for
2719                  * manually configured addresses
2720                  */
2721                 addrconf_dad_start(ifp);
2722                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2723                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2724                                          true, jiffies);
2725                 in6_ifa_put(ifp);
2726                 addrconf_verify_rtnl();
2727                 return 0;
2728         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2729                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2730                                false, pfx, ifindex);
2731         }
2732
2733         return PTR_ERR(ifp);
2734 }
2735
2736 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2737                           const struct in6_addr *pfx, unsigned int plen)
2738 {
2739         struct inet6_ifaddr *ifp;
2740         struct inet6_dev *idev;
2741         struct net_device *dev;
2742
2743         if (plen > 128)
2744                 return -EINVAL;
2745
2746         dev = __dev_get_by_index(net, ifindex);
2747         if (!dev)
2748                 return -ENODEV;
2749
2750         idev = __in6_dev_get(dev);
2751         if (!idev)
2752                 return -ENXIO;
2753
2754         read_lock_bh(&idev->lock);
2755         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2756                 if (ifp->prefix_len == plen &&
2757                     ipv6_addr_equal(pfx, &ifp->addr)) {
2758                         in6_ifa_hold(ifp);
2759                         read_unlock_bh(&idev->lock);
2760
2761                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2762                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2763                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2764                                                  jiffies);
2765                         ipv6_del_addr(ifp);
2766                         addrconf_verify_rtnl();
2767                         if (ipv6_addr_is_multicast(pfx)) {
2768                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2769                                                false, pfx, dev->ifindex);
2770                         }
2771                         return 0;
2772                 }
2773         }
2774         read_unlock_bh(&idev->lock);
2775         return -EADDRNOTAVAIL;
2776 }
2777
2778
2779 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2780 {
2781         struct in6_ifreq ireq;
2782         int err;
2783
2784         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2785                 return -EPERM;
2786
2787         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2788                 return -EFAULT;
2789
2790         rtnl_lock();
2791         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2792                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2793                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2794         rtnl_unlock();
2795         return err;
2796 }
2797
2798 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2799 {
2800         struct in6_ifreq ireq;
2801         int err;
2802
2803         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2804                 return -EPERM;
2805
2806         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2807                 return -EFAULT;
2808
2809         rtnl_lock();
2810         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2811                              ireq.ifr6_prefixlen);
2812         rtnl_unlock();
2813         return err;
2814 }
2815
2816 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2817                      int plen, int scope)
2818 {
2819         struct inet6_ifaddr *ifp;
2820
2821         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2822                             scope, IFA_F_PERMANENT,
2823                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2824         if (!IS_ERR(ifp)) {
2825                 spin_lock_bh(&ifp->lock);
2826                 ifp->flags &= ~IFA_F_TENTATIVE;
2827                 spin_unlock_bh(&ifp->lock);
2828                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2829                 in6_ifa_put(ifp);
2830         }
2831 }
2832
2833 #if IS_ENABLED(CONFIG_IPV6_SIT)
2834 static void sit_add_v4_addrs(struct inet6_dev *idev)
2835 {
2836         struct in6_addr addr;
2837         struct net_device *dev;
2838         struct net *net = dev_net(idev->dev);
2839         int scope, plen;
2840         u32 pflags = 0;
2841
2842         ASSERT_RTNL();
2843
2844         memset(&addr, 0, sizeof(struct in6_addr));
2845         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2846
2847         if (idev->dev->flags&IFF_POINTOPOINT) {
2848                 addr.s6_addr32[0] = htonl(0xfe800000);
2849                 scope = IFA_LINK;
2850                 plen = 64;
2851         } else {
2852                 scope = IPV6_ADDR_COMPATv4;
2853                 plen = 96;
2854                 pflags |= RTF_NONEXTHOP;
2855         }
2856
2857         if (addr.s6_addr32[3]) {
2858                 add_addr(idev, &addr, plen, scope);
2859                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2860                 return;
2861         }
2862
2863         for_each_netdev(net, dev) {
2864                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2865                 if (in_dev && (dev->flags & IFF_UP)) {
2866                         struct in_ifaddr *ifa;
2867
2868                         int flag = scope;
2869
2870                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2871
2872                                 addr.s6_addr32[3] = ifa->ifa_local;
2873
2874                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2875                                         continue;
2876                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2877                                         if (idev->dev->flags&IFF_POINTOPOINT)
2878                                                 continue;
2879                                         flag |= IFA_HOST;
2880                                 }
2881
2882                                 add_addr(idev, &addr, plen, flag);
2883                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2884                                                       pflags);
2885                         }
2886                 }
2887         }
2888 }
2889 #endif
2890
2891 static void init_loopback(struct net_device *dev)
2892 {
2893         struct inet6_dev  *idev;
2894         struct net_device *sp_dev;
2895         struct inet6_ifaddr *sp_ifa;
2896         struct rt6_info *sp_rt;
2897
2898         /* ::1 */
2899
2900         ASSERT_RTNL();
2901
2902         idev = ipv6_find_idev(dev);
2903         if (!idev) {
2904                 pr_debug("%s: add_dev failed\n", __func__);
2905                 return;
2906         }
2907
2908         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2909
2910         /* Add routes to other interface's IPv6 addresses */
2911         for_each_netdev(dev_net(dev), sp_dev) {
2912                 if (!strcmp(sp_dev->name, dev->name))
2913                         continue;
2914
2915                 idev = __in6_dev_get(sp_dev);
2916                 if (!idev)
2917                         continue;
2918
2919                 read_lock_bh(&idev->lock);
2920                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2921
2922                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2923                                 continue;
2924
2925                         if (sp_ifa->rt) {
2926                                 /* This dst has been added to garbage list when
2927                                  * lo device down, release this obsolete dst and
2928                                  * reallocate a new router for ifa.
2929                                  */
2930                                 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
2931                                         ip6_rt_put(sp_ifa->rt);
2932                                         sp_ifa->rt = NULL;
2933                                 } else {
2934                                         continue;
2935                                 }
2936                         }
2937
2938                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2939
2940                         /* Failure cases are ignored */
2941                         if (!IS_ERR(sp_rt)) {
2942                                 sp_ifa->rt = sp_rt;
2943                                 ip6_ins_rt(sp_rt);
2944                         }
2945                 }
2946                 read_unlock_bh(&idev->lock);
2947         }
2948 }
2949
2950 static void addrconf_add_linklocal(struct inet6_dev *idev,
2951                                    const struct in6_addr *addr, u32 flags)
2952 {
2953         struct inet6_ifaddr *ifp;
2954         u32 addr_flags = flags | IFA_F_PERMANENT;
2955
2956 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2957         if (idev->cnf.optimistic_dad &&
2958             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2959                 addr_flags |= IFA_F_OPTIMISTIC;
2960 #endif
2961
2962         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2963                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2964         if (!IS_ERR(ifp)) {
2965                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2966                 addrconf_dad_start(ifp);
2967                 in6_ifa_put(ifp);
2968         }
2969 }
2970
2971 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2972 {
2973         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2974                 return true;
2975
2976         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2977             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2978                 return true;
2979
2980         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2981             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2982                 return true;
2983
2984         return false;
2985 }
2986
2987 static int ipv6_generate_stable_address(struct in6_addr *address,
2988                                         u8 dad_count,
2989                                         const struct inet6_dev *idev)
2990 {
2991         static DEFINE_SPINLOCK(lock);
2992         static __u32 digest[SHA_DIGEST_WORDS];
2993         static __u32 workspace[SHA_WORKSPACE_WORDS];
2994
2995         static union {
2996                 char __data[SHA_MESSAGE_BYTES];
2997                 struct {
2998                         struct in6_addr secret;
2999                         __be32 prefix[2];
3000                         unsigned char hwaddr[MAX_ADDR_LEN];
3001                         u8 dad_count;
3002                 } __packed;
3003         } data;
3004
3005         struct in6_addr secret;
3006         struct in6_addr temp;
3007         struct net *net = dev_net(idev->dev);
3008
3009         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3010
3011         if (idev->cnf.stable_secret.initialized)
3012                 secret = idev->cnf.stable_secret.secret;
3013         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3014                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3015         else
3016                 return -1;
3017
3018 retry:
3019         spin_lock_bh(&lock);
3020
3021         sha_init(digest);
3022         memset(&data, 0, sizeof(data));
3023         memset(workspace, 0, sizeof(workspace));
3024         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3025         data.prefix[0] = address->s6_addr32[0];
3026         data.prefix[1] = address->s6_addr32[1];
3027         data.secret = secret;
3028         data.dad_count = dad_count;
3029
3030         sha_transform(digest, data.__data, workspace);
3031
3032         temp = *address;
3033         temp.s6_addr32[2] = (__force __be32)digest[0];
3034         temp.s6_addr32[3] = (__force __be32)digest[1];
3035
3036         spin_unlock_bh(&lock);
3037
3038         if (ipv6_reserved_interfaceid(temp)) {
3039                 dad_count++;
3040                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3041                         return -1;
3042                 goto retry;
3043         }
3044
3045         *address = temp;
3046         return 0;
3047 }
3048
3049 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3050 {
3051         struct in6_addr addr;
3052
3053         /* no link local addresses on L3 master devices */
3054         if (netif_is_l3_master(idev->dev))
3055                 return;
3056
3057         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3058
3059         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3060                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3061                         addrconf_add_linklocal(idev, &addr,
3062                                                IFA_F_STABLE_PRIVACY);
3063                 else if (prefix_route)
3064                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3065         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3066                 /* addrconf_add_linklocal also adds a prefix_route and we
3067                  * only need to care about prefix routes if ipv6_generate_eui64
3068                  * couldn't generate one.
3069                  */
3070                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3071                         addrconf_add_linklocal(idev, &addr, 0);
3072                 else if (prefix_route)
3073                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3074         }
3075 }
3076
3077 static void addrconf_dev_config(struct net_device *dev)
3078 {
3079         struct inet6_dev *idev;
3080
3081         ASSERT_RTNL();
3082
3083         if ((dev->type != ARPHRD_ETHER) &&
3084             (dev->type != ARPHRD_FDDI) &&
3085             (dev->type != ARPHRD_ARCNET) &&
3086             (dev->type != ARPHRD_INFINIBAND) &&
3087             (dev->type != ARPHRD_IEEE802154) &&
3088             (dev->type != ARPHRD_IEEE1394) &&
3089             (dev->type != ARPHRD_TUNNEL6) &&
3090             (dev->type != ARPHRD_6LOWPAN)) {
3091                 /* Alas, we support only Ethernet autoconfiguration. */
3092                 return;
3093         }
3094
3095         idev = addrconf_add_dev(dev);
3096         if (IS_ERR(idev))
3097                 return;
3098
3099         addrconf_addr_gen(idev, false);
3100 }
3101
3102 #if IS_ENABLED(CONFIG_IPV6_SIT)
3103 static void addrconf_sit_config(struct net_device *dev)
3104 {
3105         struct inet6_dev *idev;
3106
3107         ASSERT_RTNL();
3108
3109         /*
3110          * Configure the tunnel with one of our IPv4
3111          * addresses... we should configure all of
3112          * our v4 addrs in the tunnel
3113          */
3114
3115         idev = ipv6_find_idev(dev);
3116         if (!idev) {
3117                 pr_debug("%s: add_dev failed\n", __func__);
3118                 return;
3119         }
3120
3121         if (dev->priv_flags & IFF_ISATAP) {
3122                 addrconf_addr_gen(idev, false);
3123                 return;
3124         }
3125
3126         sit_add_v4_addrs(idev);
3127
3128         if (dev->flags&IFF_POINTOPOINT)
3129                 addrconf_add_mroute(dev);
3130 }
3131 #endif
3132
3133 #if IS_ENABLED(CONFIG_NET_IPGRE)
3134 static void addrconf_gre_config(struct net_device *dev)
3135 {
3136         struct inet6_dev *idev;
3137
3138         ASSERT_RTNL();
3139
3140         idev = ipv6_find_idev(dev);
3141         if (!idev) {
3142                 pr_debug("%s: add_dev failed\n", __func__);
3143                 return;
3144         }
3145
3146         addrconf_addr_gen(idev, true);
3147         if (dev->flags & IFF_POINTOPOINT)
3148                 addrconf_add_mroute(dev);
3149 }
3150 #endif
3151
3152 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3153                            void *ptr)
3154 {
3155         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3156         struct netdev_notifier_changeupper_info *info;
3157         struct inet6_dev *idev = __in6_dev_get(dev);
3158         struct net *net = dev_net(dev);
3159         int run_pending = 0;
3160         int err;
3161
3162         switch (event) {
3163         case NETDEV_REGISTER:
3164                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3165                         idev = ipv6_add_dev(dev);
3166                         if (IS_ERR(idev))
3167                                 return notifier_from_errno(PTR_ERR(idev));
3168                 }
3169                 break;
3170
3171         case NETDEV_CHANGEMTU:
3172                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3173                 if (dev->mtu < IPV6_MIN_MTU) {
3174                         addrconf_ifdown(dev, dev != net->loopback_dev);
3175                         break;
3176                 }
3177
3178                 if (idev) {
3179                         rt6_mtu_change(dev, dev->mtu);
3180                         idev->cnf.mtu6 = dev->mtu;
3181                         break;
3182                 }
3183
3184                 /* allocate new idev */
3185                 idev = ipv6_add_dev(dev);
3186                 if (IS_ERR(idev))
3187                         break;
3188
3189                 /* device is still not ready */
3190                 if (!(idev->if_flags & IF_READY))
3191                         break;
3192
3193                 run_pending = 1;
3194
3195                 /* fall through */
3196
3197         case NETDEV_UP:
3198         case NETDEV_CHANGE:
3199                 if (dev->flags & IFF_SLAVE)
3200                         break;
3201
3202                 if (idev && idev->cnf.disable_ipv6)
3203                         break;
3204
3205                 if (event == NETDEV_UP) {
3206                         if (!addrconf_qdisc_ok(dev)) {
3207                                 /* device is not ready yet. */
3208                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3209                                         dev->name);
3210                                 break;
3211                         }
3212
3213                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3214                                 idev = ipv6_add_dev(dev);
3215
3216                         if (!IS_ERR_OR_NULL(idev)) {
3217                                 idev->if_flags |= IF_READY;
3218                                 run_pending = 1;
3219                         }
3220                 } else if (event == NETDEV_CHANGE) {
3221                         if (!addrconf_qdisc_ok(dev)) {
3222                                 /* device is still not ready. */
3223                                 break;
3224                         }
3225
3226                         if (idev) {
3227                                 if (idev->if_flags & IF_READY) {
3228                                         /* device is already configured -
3229                                          * but resend MLD reports, we might
3230                                          * have roamed and need to update
3231                                          * multicast snooping switches
3232                                          */
3233                                         ipv6_mc_up(idev);
3234                                         break;
3235                                 }
3236                                 idev->if_flags |= IF_READY;
3237                         }
3238
3239                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3240                                 dev->name);
3241
3242                         run_pending = 1;
3243                 }
3244
3245                 switch (dev->type) {
3246 #if IS_ENABLED(CONFIG_IPV6_SIT)
3247                 case ARPHRD_SIT:
3248                         addrconf_sit_config(dev);
3249                         break;
3250 #endif
3251 #if IS_ENABLED(CONFIG_NET_IPGRE)
3252                 case ARPHRD_IPGRE:
3253                         addrconf_gre_config(dev);
3254                         break;
3255 #endif
3256                 case ARPHRD_LOOPBACK:
3257                         init_loopback(dev);
3258                         break;
3259
3260                 default:
3261                         addrconf_dev_config(dev);
3262                         break;
3263                 }
3264
3265                 if (!IS_ERR_OR_NULL(idev)) {
3266                         if (run_pending)
3267                                 addrconf_dad_run(idev);
3268
3269                         /*
3270                          * If the MTU changed during the interface down,
3271                          * when the interface up, the changed MTU must be
3272                          * reflected in the idev as well as routers.
3273                          */
3274                         if (idev->cnf.mtu6 != dev->mtu &&
3275                             dev->mtu >= IPV6_MIN_MTU) {
3276                                 rt6_mtu_change(dev, dev->mtu);
3277                                 idev->cnf.mtu6 = dev->mtu;
3278                         }
3279                         idev->tstamp = jiffies;
3280                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3281
3282                         /*
3283                          * If the changed mtu during down is lower than
3284                          * IPV6_MIN_MTU stop IPv6 on this interface.
3285                          */
3286                         if (dev->mtu < IPV6_MIN_MTU)
3287                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3288                 }
3289                 break;
3290
3291         case NETDEV_DOWN:
3292         case NETDEV_UNREGISTER:
3293                 /*
3294                  *      Remove all addresses from this interface.
3295                  */
3296                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3297                 break;
3298
3299         case NETDEV_CHANGENAME:
3300                 if (idev) {
3301                         snmp6_unregister_dev(idev);
3302                         addrconf_sysctl_unregister(idev);
3303                         err = addrconf_sysctl_register(idev);
3304                         if (err)
3305                                 return notifier_from_errno(err);
3306                         err = snmp6_register_dev(idev);
3307                         if (err) {
3308                                 addrconf_sysctl_unregister(idev);
3309                                 return notifier_from_errno(err);
3310                         }
3311                 }
3312                 break;
3313
3314         case NETDEV_PRE_TYPE_CHANGE:
3315         case NETDEV_POST_TYPE_CHANGE:
3316                 addrconf_type_change(dev, event);
3317                 break;
3318
3319         case NETDEV_CHANGEUPPER:
3320                 info = ptr;
3321
3322                 /* flush all routes if dev is linked to or unlinked from
3323                  * an L3 master device (e.g., VRF)
3324                  */
3325                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3326                         addrconf_ifdown(dev, 0);
3327         }
3328
3329         return NOTIFY_OK;
3330 }
3331
3332 /*
3333  *      addrconf module should be notified of a device going up
3334  */
3335 static struct notifier_block ipv6_dev_notf = {
3336         .notifier_call = addrconf_notify,
3337         .priority = ADDRCONF_NOTIFY_PRIORITY,
3338 };
3339
3340 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3341 {
3342         struct inet6_dev *idev;
3343         ASSERT_RTNL();
3344
3345         idev = __in6_dev_get(dev);
3346
3347         if (event == NETDEV_POST_TYPE_CHANGE)
3348                 ipv6_mc_remap(idev);
3349         else if (event == NETDEV_PRE_TYPE_CHANGE)
3350                 ipv6_mc_unmap(idev);
3351 }
3352
3353 static int addrconf_ifdown(struct net_device *dev, int how)
3354 {
3355         struct net *net = dev_net(dev);
3356         struct inet6_dev *idev;
3357         struct inet6_ifaddr *ifa;
3358         int state, i;
3359
3360         ASSERT_RTNL();
3361
3362         rt6_ifdown(net, dev);
3363         neigh_ifdown(&nd_tbl, dev);
3364
3365         idev = __in6_dev_get(dev);
3366         if (!idev)
3367                 return -ENODEV;
3368
3369         /*
3370          * Step 1: remove reference to ipv6 device from parent device.
3371          *         Do not dev_put!
3372          */
3373         if (how) {
3374                 idev->dead = 1;
3375
3376                 /* protected by rtnl_lock */
3377                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3378
3379                 /* Step 1.5: remove snmp6 entry */
3380                 snmp6_unregister_dev(idev);
3381
3382         }
3383
3384         /* Step 2: clear hash table */
3385         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3386                 struct hlist_head *h = &inet6_addr_lst[i];
3387
3388                 spin_lock_bh(&addrconf_hash_lock);
3389 restart:
3390                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3391                         if (ifa->idev == idev) {
3392                                 hlist_del_init_rcu(&ifa->addr_lst);
3393                                 addrconf_del_dad_work(ifa);
3394                                 goto restart;
3395                         }
3396                 }
3397                 spin_unlock_bh(&addrconf_hash_lock);
3398         }
3399
3400         write_lock_bh(&idev->lock);
3401
3402         addrconf_del_rs_timer(idev);
3403
3404         /* Step 2: clear flags for stateless addrconf */
3405         if (!how)
3406                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3407
3408         if (how && del_timer(&idev->regen_timer))
3409                 in6_dev_put(idev);
3410
3411         /* Step 3: clear tempaddr list */
3412         while (!list_empty(&idev->tempaddr_list)) {
3413                 ifa = list_first_entry(&idev->tempaddr_list,
3414                                        struct inet6_ifaddr, tmp_list);
3415                 list_del(&ifa->tmp_list);
3416                 write_unlock_bh(&idev->lock);
3417                 spin_lock_bh(&ifa->lock);
3418
3419                 if (ifa->ifpub) {
3420                         in6_ifa_put(ifa->ifpub);
3421                         ifa->ifpub = NULL;
3422                 }
3423                 spin_unlock_bh(&ifa->lock);
3424                 in6_ifa_put(ifa);
3425                 write_lock_bh(&idev->lock);
3426         }
3427
3428         while (!list_empty(&idev->addr_list)) {
3429                 ifa = list_first_entry(&idev->addr_list,
3430                                        struct inet6_ifaddr, if_list);
3431                 addrconf_del_dad_work(ifa);
3432
3433                 list_del(&ifa->if_list);
3434
3435                 write_unlock_bh(&idev->lock);
3436
3437                 spin_lock_bh(&ifa->lock);
3438                 state = ifa->state;
3439                 ifa->state = INET6_IFADDR_STATE_DEAD;
3440                 spin_unlock_bh(&ifa->lock);
3441
3442                 if (state != INET6_IFADDR_STATE_DEAD) {
3443                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3444                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3445                 }
3446                 in6_ifa_put(ifa);
3447
3448                 write_lock_bh(&idev->lock);
3449         }
3450
3451         write_unlock_bh(&idev->lock);
3452
3453         /* Step 5: Discard anycast and multicast list */
3454         if (how) {
3455                 ipv6_ac_destroy_dev(idev);
3456                 ipv6_mc_destroy_dev(idev);
3457         } else {
3458                 ipv6_mc_down(idev);
3459         }
3460
3461         idev->tstamp = jiffies;
3462
3463         /* Last: Shot the device (if unregistered) */
3464         if (how) {
3465                 addrconf_sysctl_unregister(idev);
3466                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3467                 neigh_ifdown(&nd_tbl, dev);
3468                 in6_dev_put(idev);
3469         }
3470         return 0;
3471 }
3472
3473 static void addrconf_rs_timer(unsigned long data)
3474 {
3475         struct inet6_dev *idev = (struct inet6_dev *)data;
3476         struct net_device *dev = idev->dev;
3477         struct in6_addr lladdr;
3478
3479         write_lock(&idev->lock);
3480         if (idev->dead || !(idev->if_flags & IF_READY))
3481                 goto out;
3482
3483         if (!ipv6_accept_ra(idev))
3484                 goto out;
3485
3486         /* Announcement received after solicitation was sent */
3487         if (idev->if_flags & IF_RA_RCVD)
3488                 goto out;
3489
3490         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3491                 write_unlock(&idev->lock);
3492                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3493                         ndisc_send_rs(dev, &lladdr,
3494                                       &in6addr_linklocal_allrouters);
3495                 else
3496                         goto put;
3497
3498                 write_lock(&idev->lock);
3499                 /* The wait after the last probe can be shorter */
3500                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3501                                              idev->cnf.rtr_solicits) ?
3502                                       idev->cnf.rtr_solicit_delay :
3503                                       idev->cnf.rtr_solicit_interval);
3504         } else {
3505                 /*
3506                  * Note: we do not support deprecated "all on-link"
3507                  * assumption any longer.
3508                  */
3509                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3510         }
3511
3512 out:
3513         write_unlock(&idev->lock);
3514 put:
3515         in6_dev_put(idev);
3516 }
3517
3518 /*
3519  *      Duplicate Address Detection
3520  */
3521 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3522 {
3523         unsigned long rand_num;
3524         struct inet6_dev *idev = ifp->idev;
3525
3526         if (ifp->flags & IFA_F_OPTIMISTIC)
3527                 rand_num = 0;
3528         else
3529                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3530
3531         ifp->dad_probes = idev->cnf.dad_transmits;
3532         addrconf_mod_dad_work(ifp, rand_num);
3533 }
3534
3535 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3536 {
3537         struct inet6_dev *idev = ifp->idev;
3538         struct net_device *dev = idev->dev;
3539         bool notify = false;
3540
3541         addrconf_join_solict(dev, &ifp->addr);
3542
3543         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3544
3545         read_lock_bh(&idev->lock);
3546         spin_lock(&ifp->lock);
3547         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3548                 goto out;
3549
3550         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3551             idev->cnf.accept_dad < 1 ||
3552             !(ifp->flags&IFA_F_TENTATIVE) ||
3553             ifp->flags & IFA_F_NODAD) {
3554                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3555                 spin_unlock(&ifp->lock);
3556                 read_unlock_bh(&idev->lock);
3557
3558                 addrconf_dad_completed(ifp);
3559                 return;
3560         }
3561
3562         if (!(idev->if_flags & IF_READY)) {
3563                 spin_unlock(&ifp->lock);
3564                 read_unlock_bh(&idev->lock);
3565                 /*
3566                  * If the device is not ready:
3567                  * - keep it tentative if it is a permanent address.
3568                  * - otherwise, kill it.
3569                  */
3570                 in6_ifa_hold(ifp);
3571                 addrconf_dad_stop(ifp, 0);
3572                 return;
3573         }
3574
3575         /*
3576          * Optimistic nodes can start receiving
3577          * Frames right away
3578          */
3579         if (ifp->flags & IFA_F_OPTIMISTIC) {
3580                 ip6_ins_rt(ifp->rt);
3581                 if (ipv6_use_optimistic_addr(idev)) {
3582                         /* Because optimistic nodes can use this address,
3583                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3584                          */
3585                         notify = true;
3586                 }
3587         }
3588
3589         addrconf_dad_kick(ifp);
3590 out:
3591         spin_unlock(&ifp->lock);
3592         read_unlock_bh(&idev->lock);
3593         if (notify)
3594                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3595 }
3596
3597 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3598 {
3599         bool begin_dad = false;
3600
3601         spin_lock_bh(&ifp->lock);
3602         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3603                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3604                 begin_dad = true;
3605         }
3606         spin_unlock_bh(&ifp->lock);
3607
3608         if (begin_dad)
3609                 addrconf_mod_dad_work(ifp, 0);
3610 }
3611
3612 static void addrconf_dad_work(struct work_struct *w)
3613 {
3614         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3615                                                 struct inet6_ifaddr,
3616                                                 dad_work);
3617         struct inet6_dev *idev = ifp->idev;
3618         struct in6_addr mcaddr;
3619
3620         enum {
3621                 DAD_PROCESS,
3622                 DAD_BEGIN,
3623                 DAD_ABORT,
3624         } action = DAD_PROCESS;
3625
3626         rtnl_lock();
3627
3628         spin_lock_bh(&ifp->lock);
3629         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3630                 action = DAD_BEGIN;
3631                 ifp->state = INET6_IFADDR_STATE_DAD;
3632         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3633                 action = DAD_ABORT;
3634                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3635         }
3636         spin_unlock_bh(&ifp->lock);
3637
3638         if (action == DAD_BEGIN) {
3639                 addrconf_dad_begin(ifp);
3640                 goto out;
3641         } else if (action == DAD_ABORT) {
3642                 in6_ifa_hold(ifp);
3643                 addrconf_dad_stop(ifp, 1);
3644                 goto out;
3645         }
3646
3647         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3648                 goto out;
3649
3650         write_lock_bh(&idev->lock);
3651         if (idev->dead || !(idev->if_flags & IF_READY)) {
3652                 write_unlock_bh(&idev->lock);
3653                 goto out;
3654         }
3655
3656         spin_lock(&ifp->lock);
3657         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3658                 spin_unlock(&ifp->lock);
3659                 write_unlock_bh(&idev->lock);
3660                 goto out;
3661         }
3662
3663         if (ifp->dad_probes == 0) {
3664                 /*
3665                  * DAD was successful
3666                  */
3667
3668                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3669                 spin_unlock(&ifp->lock);
3670                 write_unlock_bh(&idev->lock);
3671
3672                 addrconf_dad_completed(ifp);
3673
3674                 goto out;
3675         }
3676
3677         ifp->dad_probes--;
3678         addrconf_mod_dad_work(ifp,
3679                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3680         spin_unlock(&ifp->lock);
3681         write_unlock_bh(&idev->lock);
3682
3683         /* send a neighbour solicitation for our addr */
3684         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3685         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3686 out:
3687         in6_ifa_put(ifp);
3688         rtnl_unlock();
3689 }
3690
3691 /* ifp->idev must be at least read locked */
3692 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3693 {
3694         struct inet6_ifaddr *ifpiter;
3695         struct inet6_dev *idev = ifp->idev;
3696
3697         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3698                 if (ifpiter->scope > IFA_LINK)
3699                         break;
3700                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3701                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3702                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3703                     IFA_F_PERMANENT)
3704                         return false;
3705         }
3706         return true;
3707 }
3708
3709 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3710 {
3711         struct net_device *dev = ifp->idev->dev;
3712         struct in6_addr lladdr;
3713         bool send_rs, send_mld;
3714
3715         addrconf_del_dad_work(ifp);
3716
3717         /*
3718          *      Configure the address for reception. Now it is valid.
3719          */
3720
3721         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3722
3723         /* If added prefix is link local and we are prepared to process
3724            router advertisements, start sending router solicitations.
3725          */
3726
3727         read_lock_bh(&ifp->idev->lock);
3728         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3729         send_rs = send_mld &&
3730                   ipv6_accept_ra(ifp->idev) &&
3731                   ifp->idev->cnf.rtr_solicits > 0 &&
3732                   (dev->flags&IFF_LOOPBACK) == 0;
3733         read_unlock_bh(&ifp->idev->lock);
3734
3735         /* While dad is in progress mld report's source address is in6_addrany.
3736          * Resend with proper ll now.
3737          */
3738         if (send_mld)
3739                 ipv6_mc_dad_complete(ifp->idev);
3740
3741         if (send_rs) {
3742                 /*
3743                  *      If a host as already performed a random delay
3744                  *      [...] as part of DAD [...] there is no need
3745                  *      to delay again before sending the first RS
3746                  */
3747                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3748                         return;
3749                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3750
3751                 write_lock_bh(&ifp->idev->lock);
3752                 spin_lock(&ifp->lock);
3753                 ifp->idev->rs_probes = 1;
3754                 ifp->idev->if_flags |= IF_RS_SENT;
3755                 addrconf_mod_rs_timer(ifp->idev,
3756                                       ifp->idev->cnf.rtr_solicit_interval);
3757                 spin_unlock(&ifp->lock);
3758                 write_unlock_bh(&ifp->idev->lock);
3759         }
3760 }
3761
3762 static void addrconf_dad_run(struct inet6_dev *idev)
3763 {
3764         struct inet6_ifaddr *ifp;
3765
3766         read_lock_bh(&idev->lock);
3767         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3768                 spin_lock(&ifp->lock);
3769                 if (ifp->flags & IFA_F_TENTATIVE &&
3770                     ifp->state == INET6_IFADDR_STATE_DAD)
3771                         addrconf_dad_kick(ifp);
3772                 spin_unlock(&ifp->lock);
3773         }
3774         read_unlock_bh(&idev->lock);
3775 }
3776
3777 #ifdef CONFIG_PROC_FS
3778 struct if6_iter_state {
3779         struct seq_net_private p;
3780         int bucket;
3781         int offset;
3782 };
3783
3784 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3785 {
3786         struct inet6_ifaddr *ifa = NULL;
3787         struct if6_iter_state *state = seq->private;
3788         struct net *net = seq_file_net(seq);
3789         int p = 0;
3790
3791         /* initial bucket if pos is 0 */
3792         if (pos == 0) {
3793                 state->bucket = 0;
3794                 state->offset = 0;
3795         }
3796
3797         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3798                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3799                                          addr_lst) {
3800                         if (!net_eq(dev_net(ifa->idev->dev), net))
3801                                 continue;
3802                         /* sync with offset */
3803                         if (p < state->offset) {
3804                                 p++;
3805                                 continue;
3806                         }
3807                         return ifa;
3808                 }
3809
3810                 /* prepare for next bucket */
3811                 state->offset = 0;
3812                 p = 0;
3813         }
3814         return NULL;
3815 }
3816
3817 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3818                                          struct inet6_ifaddr *ifa)
3819 {
3820         struct if6_iter_state *state = seq->private;
3821         struct net *net = seq_file_net(seq);
3822
3823         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3824                 if (!net_eq(dev_net(ifa->idev->dev), net))
3825                         continue;
3826                 state->offset++;
3827                 return ifa;
3828         }
3829
3830         state->offset = 0;
3831         while (++state->bucket < IN6_ADDR_HSIZE) {
3832                 hlist_for_each_entry_rcu_bh(ifa,
3833                                      &inet6_addr_lst[state->bucket], addr_lst) {
3834                         if (!net_eq(dev_net(ifa->idev->dev), net))
3835                                 continue;
3836                         return ifa;
3837                 }
3838         }
3839
3840         return NULL;
3841 }
3842
3843 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3844         __acquires(rcu_bh)
3845 {
3846         rcu_read_lock_bh();
3847         return if6_get_first(seq, *pos);
3848 }
3849
3850 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3851 {
3852         struct inet6_ifaddr *ifa;
3853
3854         ifa = if6_get_next(seq, v);
3855         ++*pos;
3856         return ifa;
3857 }
3858
3859 static void if6_seq_stop(struct seq_file *seq, void *v)
3860         __releases(rcu_bh)
3861 {
3862         rcu_read_unlock_bh();
3863 }
3864
3865 static int if6_seq_show(struct seq_file *seq, void *v)
3866 {
3867         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3868         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3869                    &ifp->addr,
3870                    ifp->idev->dev->ifindex,
3871                    ifp->prefix_len,
3872                    ifp->scope,
3873                    (u8) ifp->flags,
3874                    ifp->idev->dev->name);
3875         return 0;
3876 }
3877
3878 static const struct seq_operations if6_seq_ops = {
3879         .start  = if6_seq_start,
3880         .next   = if6_seq_next,
3881         .show   = if6_seq_show,
3882         .stop   = if6_seq_stop,
3883 };
3884
3885 static int if6_seq_open(struct inode *inode, struct file *file)
3886 {
3887         return seq_open_net(inode, file, &if6_seq_ops,
3888                             sizeof(struct if6_iter_state));
3889 }
3890
3891 static const struct file_operations if6_fops = {
3892         .owner          = THIS_MODULE,
3893         .open           = if6_seq_open,
3894         .read           = seq_read,
3895         .llseek         = seq_lseek,
3896         .release        = seq_release_net,
3897 };
3898
3899 static int __net_init if6_proc_net_init(struct net *net)
3900 {
3901         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3902                 return -ENOMEM;
3903         return 0;
3904 }
3905
3906 static void __net_exit if6_proc_net_exit(struct net *net)
3907 {
3908         remove_proc_entry("if_inet6", net->proc_net);
3909 }
3910
3911 static struct pernet_operations if6_proc_net_ops = {
3912         .init = if6_proc_net_init,
3913         .exit = if6_proc_net_exit,
3914 };
3915
3916 int __init if6_proc_init(void)
3917 {
3918         return register_pernet_subsys(&if6_proc_net_ops);
3919 }
3920
3921 void if6_proc_exit(void)
3922 {
3923         unregister_pernet_subsys(&if6_proc_net_ops);
3924 }
3925 #endif  /* CONFIG_PROC_FS */
3926
3927 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3928 /* Check if address is a home address configured on any interface. */
3929 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3930 {
3931         int ret = 0;
3932         struct inet6_ifaddr *ifp = NULL;
3933         unsigned int hash = inet6_addr_hash(addr);
3934
3935         rcu_read_lock_bh();
3936         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3937                 if (!net_eq(dev_net(ifp->idev->dev), net))
3938                         continue;
3939                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3940                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3941                         ret = 1;
3942                         break;
3943                 }
3944         }
3945         rcu_read_unlock_bh();
3946         return ret;
3947 }
3948 #endif
3949
3950 /*
3951  *      Periodic address status verification
3952  */
3953
3954 static void addrconf_verify_rtnl(void)
3955 {
3956         unsigned long now, next, next_sec, next_sched;
3957         struct inet6_ifaddr *ifp;
3958         int i;
3959
3960         ASSERT_RTNL();
3961
3962         rcu_read_lock_bh();
3963         now = jiffies;
3964         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3965
3966         cancel_delayed_work(&addr_chk_work);
3967
3968         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3969 restart:
3970                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3971                         unsigned long age;
3972
3973                         /* When setting preferred_lft to a value not zero or
3974                          * infinity, while valid_lft is infinity
3975                          * IFA_F_PERMANENT has a non-infinity life time.
3976                          */
3977                         if ((ifp->flags & IFA_F_PERMANENT) &&
3978                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3979                                 continue;
3980
3981                         spin_lock(&ifp->lock);
3982                         /* We try to batch several events at once. */
3983                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3984
3985                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3986                             age >= ifp->valid_lft) {
3987                                 spin_unlock(&ifp->lock);
3988                                 in6_ifa_hold(ifp);
3989                                 ipv6_del_addr(ifp);
3990                                 goto restart;
3991                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3992                                 spin_unlock(&ifp->lock);
3993                                 continue;
3994                         } else if (age >= ifp->prefered_lft) {
3995                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3996                                 int deprecate = 0;
3997
3998                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3999                                         deprecate = 1;
4000                                         ifp->flags |= IFA_F_DEPRECATED;
4001                                 }
4002
4003                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4004                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4005                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4006
4007                                 spin_unlock(&ifp->lock);
4008
4009                                 if (deprecate) {
4010                                         in6_ifa_hold(ifp);
4011
4012                                         ipv6_ifa_notify(0, ifp);
4013                                         in6_ifa_put(ifp);
4014                                         goto restart;
4015                                 }
4016                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4017                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4018                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4019                                         ifp->idev->cnf.dad_transmits *
4020                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4021
4022                                 if (age >= ifp->prefered_lft - regen_advance) {
4023                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4024                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4025                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4026                                         if (!ifp->regen_count && ifpub) {
4027                                                 ifp->regen_count++;
4028                                                 in6_ifa_hold(ifp);
4029                                                 in6_ifa_hold(ifpub);
4030                                                 spin_unlock(&ifp->lock);
4031
4032                                                 spin_lock(&ifpub->lock);
4033                                                 ifpub->regen_count = 0;
4034                                                 spin_unlock(&ifpub->lock);
4035                                                 ipv6_create_tempaddr(ifpub, ifp);
4036                                                 in6_ifa_put(ifpub);
4037                                                 in6_ifa_put(ifp);
4038                                                 goto restart;
4039                                         }
4040                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4041                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4042                                 spin_unlock(&ifp->lock);
4043                         } else {
4044                                 /* ifp->prefered_lft <= ifp->valid_lft */
4045                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4046                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4047                                 spin_unlock(&ifp->lock);
4048                         }
4049                 }
4050         }
4051
4052         next_sec = round_jiffies_up(next);
4053         next_sched = next;
4054
4055         /* If rounded timeout is accurate enough, accept it. */
4056         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4057                 next_sched = next_sec;
4058
4059         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4060         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4061                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4062
4063         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4064               now, next, next_sec, next_sched);
4065         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4066         rcu_read_unlock_bh();
4067 }
4068
4069 static void addrconf_verify_work(struct work_struct *w)
4070 {
4071         rtnl_lock();
4072         addrconf_verify_rtnl();
4073         rtnl_unlock();
4074 }
4075
4076 static void addrconf_verify(void)
4077 {
4078         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4079 }
4080
4081 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4082                                      struct in6_addr **peer_pfx)
4083 {
4084         struct in6_addr *pfx = NULL;
4085
4086         *peer_pfx = NULL;
4087
4088         if (addr)
4089                 pfx = nla_data(addr);
4090
4091         if (local) {
4092                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4093                         *peer_pfx = pfx;
4094                 pfx = nla_data(local);
4095         }
4096
4097         return pfx;
4098 }
4099
4100 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4101         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4102         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4103         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4104         [IFA_FLAGS]             = { .len = sizeof(u32) },
4105 };
4106
4107 static int
4108 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4109 {
4110         struct net *net = sock_net(skb->sk);
4111         struct ifaddrmsg *ifm;
4112         struct nlattr *tb[IFA_MAX+1];
4113         struct in6_addr *pfx, *peer_pfx;
4114         u32 ifa_flags;
4115         int err;
4116
4117         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4118         if (err < 0)
4119                 return err;
4120
4121         ifm = nlmsg_data(nlh);
4122         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4123         if (!pfx)
4124                 return -EINVAL;
4125
4126         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4127
4128         /* We ignore other flags so far. */
4129         ifa_flags &= IFA_F_MANAGETEMPADDR;
4130
4131         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4132                               ifm->ifa_prefixlen);
4133 }
4134
4135 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4136                              u32 prefered_lft, u32 valid_lft)
4137 {
4138         u32 flags;
4139         clock_t expires;
4140         unsigned long timeout;
4141         bool was_managetempaddr;
4142         bool had_prefixroute;
4143
4144         ASSERT_RTNL();
4145
4146         if (!valid_lft || (prefered_lft > valid_lft))
4147                 return -EINVAL;
4148
4149         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4150             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4151                 return -EINVAL;
4152
4153         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4154         if (addrconf_finite_timeout(timeout)) {
4155                 expires = jiffies_to_clock_t(timeout * HZ);
4156                 valid_lft = timeout;
4157                 flags = RTF_EXPIRES;
4158         } else {
4159                 expires = 0;
4160                 flags = 0;
4161                 ifa_flags |= IFA_F_PERMANENT;
4162         }
4163
4164         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4165         if (addrconf_finite_timeout(timeout)) {
4166                 if (timeout == 0)
4167                         ifa_flags |= IFA_F_DEPRECATED;
4168                 prefered_lft = timeout;
4169         }
4170
4171         spin_lock_bh(&ifp->lock);
4172         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4173         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4174                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4175         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4176                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4177                         IFA_F_NOPREFIXROUTE);
4178         ifp->flags |= ifa_flags;
4179         ifp->tstamp = jiffies;
4180         ifp->valid_lft = valid_lft;
4181         ifp->prefered_lft = prefered_lft;
4182
4183         spin_unlock_bh(&ifp->lock);
4184         if (!(ifp->flags&IFA_F_TENTATIVE))
4185                 ipv6_ifa_notify(0, ifp);
4186
4187         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4188                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4189                                       expires, flags);
4190         } else if (had_prefixroute) {
4191                 enum cleanup_prefix_rt_t action;
4192                 unsigned long rt_expires;
4193
4194                 write_lock_bh(&ifp->idev->lock);
4195                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4196                 write_unlock_bh(&ifp->idev->lock);
4197
4198                 if (action != CLEANUP_PREFIX_RT_NOP) {
4199                         cleanup_prefix_route(ifp, rt_expires,
4200                                 action == CLEANUP_PREFIX_RT_DEL);
4201                 }
4202         }
4203
4204         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4205                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4206                         valid_lft = prefered_lft = 0;
4207                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4208                                  !was_managetempaddr, jiffies);
4209         }
4210
4211         addrconf_verify_rtnl();
4212
4213         return 0;
4214 }
4215
4216 static int
4217 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4218 {
4219         struct net *net = sock_net(skb->sk);
4220         struct ifaddrmsg *ifm;
4221         struct nlattr *tb[IFA_MAX+1];
4222         struct in6_addr *pfx, *peer_pfx;
4223         struct inet6_ifaddr *ifa;
4224         struct net_device *dev;
4225         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4226         u32 ifa_flags;
4227         int err;
4228
4229         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4230         if (err < 0)
4231                 return err;
4232
4233         ifm = nlmsg_data(nlh);
4234         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4235         if (!pfx)
4236                 return -EINVAL;
4237
4238         if (tb[IFA_CACHEINFO]) {
4239                 struct ifa_cacheinfo *ci;
4240
4241                 ci = nla_data(tb[IFA_CACHEINFO]);
4242                 valid_lft = ci->ifa_valid;
4243                 preferred_lft = ci->ifa_prefered;
4244         } else {
4245                 preferred_lft = INFINITY_LIFE_TIME;
4246                 valid_lft = INFINITY_LIFE_TIME;
4247         }
4248
4249         dev =  __dev_get_by_index(net, ifm->ifa_index);
4250         if (!dev)
4251                 return -ENODEV;
4252
4253         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4254
4255         /* We ignore other flags so far. */
4256         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4257                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4258
4259         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4260         if (!ifa) {
4261                 /*
4262                  * It would be best to check for !NLM_F_CREATE here but
4263                  * userspace already relies on not having to provide this.
4264                  */
4265                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4266                                       ifm->ifa_prefixlen, ifa_flags,
4267                                       preferred_lft, valid_lft);
4268         }
4269
4270         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4271             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4272                 err = -EEXIST;
4273         else
4274                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4275
4276         in6_ifa_put(ifa);
4277
4278         return err;
4279 }
4280
4281 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4282                           u8 scope, int ifindex)
4283 {
4284         struct ifaddrmsg *ifm;
4285
4286         ifm = nlmsg_data(nlh);
4287         ifm->ifa_family = AF_INET6;
4288         ifm->ifa_prefixlen = prefixlen;
4289         ifm->ifa_flags = flags;
4290         ifm->ifa_scope = scope;
4291         ifm->ifa_index = ifindex;
4292 }
4293
4294 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4295                          unsigned long tstamp, u32 preferred, u32 valid)
4296 {
4297         struct ifa_cacheinfo ci;
4298
4299         ci.cstamp = cstamp_delta(cstamp);
4300         ci.tstamp = cstamp_delta(tstamp);
4301         ci.ifa_prefered = preferred;
4302         ci.ifa_valid = valid;
4303
4304         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4305 }
4306
4307 static inline int rt_scope(int ifa_scope)
4308 {
4309         if (ifa_scope & IFA_HOST)
4310                 return RT_SCOPE_HOST;
4311         else if (ifa_scope & IFA_LINK)
4312                 return RT_SCOPE_LINK;
4313         else if (ifa_scope & IFA_SITE)
4314                 return RT_SCOPE_SITE;
4315         else
4316                 return RT_SCOPE_UNIVERSE;
4317 }
4318
4319 static inline int inet6_ifaddr_msgsize(void)
4320 {
4321         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4322                + nla_total_size(16) /* IFA_LOCAL */
4323                + nla_total_size(16) /* IFA_ADDRESS */
4324                + nla_total_size(sizeof(struct ifa_cacheinfo))
4325                + nla_total_size(4)  /* IFA_FLAGS */;
4326 }
4327
4328 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4329                              u32 portid, u32 seq, int event, unsigned int flags)
4330 {
4331         struct nlmsghdr  *nlh;
4332         u32 preferred, valid;
4333
4334         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4335         if (!nlh)
4336                 return -EMSGSIZE;
4337
4338         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4339                       ifa->idev->dev->ifindex);
4340
4341         if (!((ifa->flags&IFA_F_PERMANENT) &&
4342               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4343                 preferred = ifa->prefered_lft;
4344                 valid = ifa->valid_lft;
4345                 if (preferred != INFINITY_LIFE_TIME) {
4346                         long tval = (jiffies - ifa->tstamp)/HZ;
4347                         if (preferred > tval)
4348                                 preferred -= tval;
4349                         else
4350                                 preferred = 0;
4351                         if (valid != INFINITY_LIFE_TIME) {
4352                                 if (valid > tval)
4353                                         valid -= tval;
4354                                 else
4355                                         valid = 0;
4356                         }
4357                 }
4358         } else {
4359                 preferred = INFINITY_LIFE_TIME;
4360                 valid = INFINITY_LIFE_TIME;
4361         }
4362
4363         if (!ipv6_addr_any(&ifa->peer_addr)) {
4364                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4365                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4366                         goto error;
4367         } else
4368                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4369                         goto error;
4370
4371         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4372                 goto error;
4373
4374         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4375                 goto error;
4376
4377         nlmsg_end(skb, nlh);
4378         return 0;
4379
4380 error:
4381         nlmsg_cancel(skb, nlh);
4382         return -EMSGSIZE;
4383 }
4384
4385 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4386                                 u32 portid, u32 seq, int event, u16 flags)
4387 {
4388         struct nlmsghdr  *nlh;
4389         u8 scope = RT_SCOPE_UNIVERSE;
4390         int ifindex = ifmca->idev->dev->ifindex;
4391
4392         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4393                 scope = RT_SCOPE_SITE;
4394
4395         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4396         if (!nlh)
4397                 return -EMSGSIZE;
4398
4399         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4400         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4401             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4402                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4403                 nlmsg_cancel(skb, nlh);
4404                 return -EMSGSIZE;
4405         }
4406
4407         nlmsg_end(skb, nlh);
4408         return 0;
4409 }
4410
4411 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4412                                 u32 portid, u32 seq, int event, unsigned int flags)
4413 {
4414         struct nlmsghdr  *nlh;
4415         u8 scope = RT_SCOPE_UNIVERSE;
4416         int ifindex = ifaca->aca_idev->dev->ifindex;
4417
4418         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4419                 scope = RT_SCOPE_SITE;
4420
4421         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4422         if (!nlh)
4423                 return -EMSGSIZE;
4424
4425         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4426         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4427             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4428                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4429                 nlmsg_cancel(skb, nlh);
4430                 return -EMSGSIZE;
4431         }
4432
4433         nlmsg_end(skb, nlh);
4434         return 0;
4435 }
4436
4437 enum addr_type_t {
4438         UNICAST_ADDR,
4439         MULTICAST_ADDR,
4440         ANYCAST_ADDR,
4441 };
4442
4443 /* called with rcu_read_lock() */
4444 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4445                           struct netlink_callback *cb, enum addr_type_t type,
4446                           int s_ip_idx, int *p_ip_idx)
4447 {
4448         struct ifmcaddr6 *ifmca;
4449         struct ifacaddr6 *ifaca;
4450         int err = 1;
4451         int ip_idx = *p_ip_idx;
4452
4453         read_lock_bh(&idev->lock);
4454         switch (type) {
4455         case UNICAST_ADDR: {
4456                 struct inet6_ifaddr *ifa;
4457
4458                 /* unicast address incl. temp addr */
4459                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4460                         if (ip_idx < s_ip_idx)
4461                                 goto next;
4462                         err = inet6_fill_ifaddr(skb, ifa,
4463                                                 NETLINK_CB(cb->skb).portid,
4464                                                 cb->nlh->nlmsg_seq,
4465                                                 RTM_NEWADDR,
4466                                                 NLM_F_MULTI);
4467                         if (err < 0)
4468                                 break;
4469                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4470 next:
4471                         ip_idx++;
4472                 }
4473                 break;
4474         }
4475         case MULTICAST_ADDR:
4476                 /* multicast address */
4477                 for (ifmca = idev->mc_list; ifmca;
4478                      ifmca = ifmca->next, ip_idx++) {
4479                         if (ip_idx < s_ip_idx)
4480                                 continue;
4481                         err = inet6_fill_ifmcaddr(skb, ifmca,
4482                                                   NETLINK_CB(cb->skb).portid,
4483                                                   cb->nlh->nlmsg_seq,
4484                                                   RTM_GETMULTICAST,
4485                                                   NLM_F_MULTI);
4486                         if (err < 0)
4487                                 break;
4488                 }
4489                 break;
4490         case ANYCAST_ADDR:
4491                 /* anycast address */
4492                 for (ifaca = idev->ac_list; ifaca;
4493                      ifaca = ifaca->aca_next, ip_idx++) {
4494                         if (ip_idx < s_ip_idx)
4495                                 continue;
4496                         err = inet6_fill_ifacaddr(skb, ifaca,
4497                                                   NETLINK_CB(cb->skb).portid,
4498                                                   cb->nlh->nlmsg_seq,
4499                                                   RTM_GETANYCAST,
4500                                                   NLM_F_MULTI);
4501                         if (err < 0)
4502                                 break;
4503                 }
4504                 break;
4505         default:
4506                 break;
4507         }
4508         read_unlock_bh(&idev->lock);
4509         *p_ip_idx = ip_idx;
4510         return err;
4511 }
4512
4513 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4514                            enum addr_type_t type)
4515 {
4516         struct net *net = sock_net(skb->sk);
4517         int h, s_h;
4518         int idx, ip_idx;
4519         int s_idx, s_ip_idx;
4520         struct net_device *dev;
4521         struct inet6_dev *idev;
4522         struct hlist_head *head;
4523
4524         s_h = cb->args[0];
4525         s_idx = idx = cb->args[1];
4526         s_ip_idx = ip_idx = cb->args[2];
4527
4528         rcu_read_lock();
4529         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4530         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4531                 idx = 0;
4532                 head = &net->dev_index_head[h];
4533                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4534                         if (idx < s_idx)
4535                                 goto cont;
4536                         if (h > s_h || idx > s_idx)
4537                                 s_ip_idx = 0;
4538                         ip_idx = 0;
4539                         idev = __in6_dev_get(dev);
4540                         if (!idev)
4541                                 goto cont;
4542
4543                         if (in6_dump_addrs(idev, skb, cb, type,
4544                                            s_ip_idx, &ip_idx) < 0)
4545                                 goto done;
4546 cont:
4547                         idx++;
4548                 }
4549         }
4550 done:
4551         rcu_read_unlock();
4552         cb->args[0] = h;
4553         cb->args[1] = idx;
4554         cb->args[2] = ip_idx;
4555
4556         return skb->len;
4557 }
4558
4559 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4560 {
4561         enum addr_type_t type = UNICAST_ADDR;
4562
4563         return inet6_dump_addr(skb, cb, type);
4564 }
4565
4566 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4567 {
4568         enum addr_type_t type = MULTICAST_ADDR;
4569
4570         return inet6_dump_addr(skb, cb, type);
4571 }
4572
4573
4574 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4575 {
4576         enum addr_type_t type = ANYCAST_ADDR;
4577
4578         return inet6_dump_addr(skb, cb, type);
4579 }
4580
4581 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4582 {
4583         struct net *net = sock_net(in_skb->sk);
4584         struct ifaddrmsg *ifm;
4585         struct nlattr *tb[IFA_MAX+1];
4586         struct in6_addr *addr = NULL, *peer;
4587         struct net_device *dev = NULL;
4588         struct inet6_ifaddr *ifa;
4589         struct sk_buff *skb;
4590         int err;
4591
4592         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4593         if (err < 0)
4594                 goto errout;
4595
4596         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4597         if (!addr) {
4598                 err = -EINVAL;
4599                 goto errout;
4600         }
4601
4602         ifm = nlmsg_data(nlh);
4603         if (ifm->ifa_index)
4604                 dev = __dev_get_by_index(net, ifm->ifa_index);
4605
4606         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4607         if (!ifa) {
4608                 err = -EADDRNOTAVAIL;
4609                 goto errout;
4610         }
4611
4612         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4613         if (!skb) {
4614                 err = -ENOBUFS;
4615                 goto errout_ifa;
4616         }
4617
4618         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4619                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4620         if (err < 0) {
4621                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4622                 WARN_ON(err == -EMSGSIZE);
4623                 kfree_skb(skb);
4624                 goto errout_ifa;
4625         }
4626         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4627 errout_ifa:
4628         in6_ifa_put(ifa);
4629 errout:
4630         return err;
4631 }
4632
4633 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4634 {
4635         struct sk_buff *skb;
4636         struct net *net = dev_net(ifa->idev->dev);
4637         int err = -ENOBUFS;
4638
4639         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4640         if (!skb)
4641                 goto errout;
4642
4643         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4644         if (err < 0) {
4645                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4646                 WARN_ON(err == -EMSGSIZE);
4647                 kfree_skb(skb);
4648                 goto errout;
4649         }
4650         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4651         return;
4652 errout:
4653         if (err < 0)
4654                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4655 }
4656
4657 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4658                                 __s32 *array, int bytes)
4659 {
4660         BUG_ON(bytes < (DEVCONF_MAX * 4));
4661
4662         memset(array, 0, bytes);
4663         array[DEVCONF_FORWARDING] = cnf->forwarding;
4664         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4665         array[DEVCONF_MTU6] = cnf->mtu6;
4666         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4667         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4668         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4669         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4670         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4671         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4672                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4673         array[DEVCONF_RTR_SOLICIT_DELAY] =
4674                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4675         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4676         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4677                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4678         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4679                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4680         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4681         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4682         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4683         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4684         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4685         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4686         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4687         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4688         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4689 #ifdef CONFIG_IPV6_ROUTER_PREF
4690         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4691         array[DEVCONF_RTR_PROBE_INTERVAL] =
4692                 jiffies_to_msecs(cnf->rtr_probe_interval);
4693 #ifdef CONFIG_IPV6_ROUTE_INFO
4694         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4695 #endif
4696 #endif
4697         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4698         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4699 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4700         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4701         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4702 #endif
4703 #ifdef CONFIG_IPV6_MROUTE
4704         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4705 #endif
4706         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4707         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4708         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4709         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4710         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4711         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4712         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4713         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4714         /* we omit DEVCONF_STABLE_SECRET for now */
4715         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4716 }
4717
4718 static inline size_t inet6_ifla6_size(void)
4719 {
4720         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4721              + nla_total_size(sizeof(struct ifla_cacheinfo))
4722              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4723              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4724              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4725              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4726 }
4727
4728 static inline size_t inet6_if_nlmsg_size(void)
4729 {
4730         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4731                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4732                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4733                + nla_total_size(4) /* IFLA_MTU */
4734                + nla_total_size(4) /* IFLA_LINK */
4735                + nla_total_size(1) /* IFLA_OPERSTATE */
4736                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4737 }
4738
4739 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4740                                       int items, int bytes)
4741 {
4742         int i;
4743         int pad = bytes - sizeof(u64) * items;
4744         BUG_ON(pad < 0);
4745
4746         /* Use put_unaligned() because stats may not be aligned for u64. */
4747         put_unaligned(items, &stats[0]);
4748         for (i = 1; i < items; i++)
4749                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4750
4751         memset(&stats[items], 0, pad);
4752 }
4753
4754 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4755                                         int bytes, size_t syncpoff)
4756 {
4757         int i, c;
4758         u64 buff[IPSTATS_MIB_MAX];
4759         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4760
4761         BUG_ON(pad < 0);
4762
4763         memset(buff, 0, sizeof(buff));
4764         buff[0] = IPSTATS_MIB_MAX;
4765
4766         for_each_possible_cpu(c) {
4767                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4768                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4769         }
4770
4771         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4772         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4773 }
4774
4775 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4776                              int bytes)
4777 {
4778         switch (attrtype) {
4779         case IFLA_INET6_STATS:
4780                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4781                                      offsetof(struct ipstats_mib, syncp));
4782                 break;
4783         case IFLA_INET6_ICMP6STATS:
4784                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4785                 break;
4786         }
4787 }
4788
4789 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4790                                   u32 ext_filter_mask)
4791 {
4792         struct nlattr *nla;
4793         struct ifla_cacheinfo ci;
4794
4795         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4796                 goto nla_put_failure;
4797         ci.max_reasm_len = IPV6_MAXPLEN;
4798         ci.tstamp = cstamp_delta(idev->tstamp);
4799         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4800         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4801         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4802                 goto nla_put_failure;
4803         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4804         if (!nla)
4805                 goto nla_put_failure;
4806         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4807
4808         /* XXX - MC not implemented */
4809
4810         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4811                 return 0;
4812
4813         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4814         if (!nla)
4815                 goto nla_put_failure;
4816         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4817
4818         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4819         if (!nla)
4820                 goto nla_put_failure;
4821         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4822
4823         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4824         if (!nla)
4825                 goto nla_put_failure;
4826
4827         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4828                 goto nla_put_failure;
4829
4830         read_lock_bh(&idev->lock);
4831         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4832         read_unlock_bh(&idev->lock);
4833
4834         return 0;
4835
4836 nla_put_failure:
4837         return -EMSGSIZE;
4838 }
4839
4840 static size_t inet6_get_link_af_size(const struct net_device *dev,
4841                                      u32 ext_filter_mask)
4842 {
4843         if (!__in6_dev_get(dev))
4844                 return 0;
4845
4846         return inet6_ifla6_size();
4847 }
4848
4849 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4850                               u32 ext_filter_mask)
4851 {
4852         struct inet6_dev *idev = __in6_dev_get(dev);
4853
4854         if (!idev)
4855                 return -ENODATA;
4856
4857         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4858                 return -EMSGSIZE;
4859
4860         return 0;
4861 }
4862
4863 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4864 {
4865         struct inet6_ifaddr *ifp;
4866         struct net_device *dev = idev->dev;
4867         bool update_rs = false;
4868         struct in6_addr ll_addr;
4869
4870         ASSERT_RTNL();
4871
4872         if (!token)
4873                 return -EINVAL;
4874         if (ipv6_addr_any(token))
4875                 return -EINVAL;
4876         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4877                 return -EINVAL;
4878         if (!ipv6_accept_ra(idev))
4879                 return -EINVAL;
4880         if (idev->cnf.rtr_solicits <= 0)
4881                 return -EINVAL;
4882
4883         write_lock_bh(&idev->lock);
4884
4885         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4886         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4887
4888         write_unlock_bh(&idev->lock);
4889
4890         if (!idev->dead && (idev->if_flags & IF_READY) &&
4891             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4892                              IFA_F_OPTIMISTIC)) {
4893
4894                 /* If we're not ready, then normal ifup will take care
4895                  * of this. Otherwise, we need to request our rs here.
4896                  */
4897                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4898                 update_rs = true;
4899         }
4900
4901         write_lock_bh(&idev->lock);
4902
4903         if (update_rs) {
4904                 idev->if_flags |= IF_RS_SENT;
4905                 idev->rs_probes = 1;
4906                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4907         }
4908
4909         /* Well, that's kinda nasty ... */
4910         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4911                 spin_lock(&ifp->lock);
4912                 if (ifp->tokenized) {
4913                         ifp->valid_lft = 0;
4914                         ifp->prefered_lft = 0;
4915                 }
4916                 spin_unlock(&ifp->lock);
4917         }
4918
4919         write_unlock_bh(&idev->lock);
4920         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4921         addrconf_verify_rtnl();
4922         return 0;
4923 }
4924
4925 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4926         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4927         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4928 };
4929
4930 static int inet6_validate_link_af(const struct net_device *dev,
4931                                   const struct nlattr *nla)
4932 {
4933         struct nlattr *tb[IFLA_INET6_MAX + 1];
4934
4935         if (dev && !__in6_dev_get(dev))
4936                 return -EAFNOSUPPORT;
4937
4938         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4939 }
4940
4941 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4942 {
4943         int err = -EINVAL;
4944         struct inet6_dev *idev = __in6_dev_get(dev);
4945         struct nlattr *tb[IFLA_INET6_MAX + 1];
4946
4947         if (!idev)
4948                 return -EAFNOSUPPORT;
4949
4950         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4951                 BUG();
4952
4953         if (tb[IFLA_INET6_TOKEN]) {
4954                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4955                 if (err)
4956                         return err;
4957         }
4958
4959         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4960                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4961
4962                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4963                     mode != IN6_ADDR_GEN_MODE_NONE &&
4964                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4965                         return -EINVAL;
4966
4967                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4968                     !idev->cnf.stable_secret.initialized &&
4969                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4970                         return -EINVAL;
4971
4972                 idev->addr_gen_mode = mode;
4973                 err = 0;
4974         }
4975
4976         return err;
4977 }
4978
4979 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4980                              u32 portid, u32 seq, int event, unsigned int flags)
4981 {
4982         struct net_device *dev = idev->dev;
4983         struct ifinfomsg *hdr;
4984         struct nlmsghdr *nlh;
4985         void *protoinfo;
4986
4987         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4988         if (!nlh)
4989                 return -EMSGSIZE;
4990
4991         hdr = nlmsg_data(nlh);
4992         hdr->ifi_family = AF_INET6;
4993         hdr->__ifi_pad = 0;
4994         hdr->ifi_type = dev->type;
4995         hdr->ifi_index = dev->ifindex;
4996         hdr->ifi_flags = dev_get_flags(dev);
4997         hdr->ifi_change = 0;
4998
4999         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5000             (dev->addr_len &&
5001              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5002             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5003             (dev->ifindex != dev_get_iflink(dev) &&
5004              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5005             nla_put_u8(skb, IFLA_OPERSTATE,
5006                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5007                 goto nla_put_failure;
5008         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5009         if (!protoinfo)
5010                 goto nla_put_failure;
5011
5012         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5013                 goto nla_put_failure;
5014
5015         nla_nest_end(skb, protoinfo);
5016         nlmsg_end(skb, nlh);
5017         return 0;
5018
5019 nla_put_failure:
5020         nlmsg_cancel(skb, nlh);
5021         return -EMSGSIZE;
5022 }
5023
5024 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5025 {
5026         struct net *net = sock_net(skb->sk);
5027         int h, s_h;
5028         int idx = 0, s_idx;
5029         struct net_device *dev;
5030         struct inet6_dev *idev;
5031         struct hlist_head *head;
5032
5033         s_h = cb->args[0];
5034         s_idx = cb->args[1];
5035
5036         rcu_read_lock();
5037         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5038                 idx = 0;
5039                 head = &net->dev_index_head[h];
5040                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5041                         if (idx < s_idx)
5042                                 goto cont;
5043                         idev = __in6_dev_get(dev);
5044                         if (!idev)
5045                                 goto cont;
5046                         if (inet6_fill_ifinfo(skb, idev,
5047                                               NETLINK_CB(cb->skb).portid,
5048                                               cb->nlh->nlmsg_seq,
5049                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5050                                 goto out;
5051 cont:
5052                         idx++;
5053                 }
5054         }
5055 out:
5056         rcu_read_unlock();
5057         cb->args[1] = idx;
5058         cb->args[0] = h;
5059
5060         return skb->len;
5061 }
5062
5063 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5064 {
5065         struct sk_buff *skb;
5066         struct net *net = dev_net(idev->dev);
5067         int err = -ENOBUFS;
5068
5069         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5070         if (!skb)
5071                 goto errout;
5072
5073         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5074         if (err < 0) {
5075                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5076                 WARN_ON(err == -EMSGSIZE);
5077                 kfree_skb(skb);
5078                 goto errout;
5079         }
5080         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5081         return;
5082 errout:
5083         if (err < 0)
5084                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5085 }
5086
5087 static inline size_t inet6_prefix_nlmsg_size(void)
5088 {
5089         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5090                + nla_total_size(sizeof(struct in6_addr))
5091                + nla_total_size(sizeof(struct prefix_cacheinfo));
5092 }
5093
5094 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5095                              struct prefix_info *pinfo, u32 portid, u32 seq,
5096                              int event, unsigned int flags)
5097 {
5098         struct prefixmsg *pmsg;
5099         struct nlmsghdr *nlh;
5100         struct prefix_cacheinfo ci;
5101
5102         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5103         if (!nlh)
5104                 return -EMSGSIZE;
5105
5106         pmsg = nlmsg_data(nlh);
5107         pmsg->prefix_family = AF_INET6;
5108         pmsg->prefix_pad1 = 0;
5109         pmsg->prefix_pad2 = 0;
5110         pmsg->prefix_ifindex = idev->dev->ifindex;
5111         pmsg->prefix_len = pinfo->prefix_len;
5112         pmsg->prefix_type = pinfo->type;
5113         pmsg->prefix_pad3 = 0;
5114         pmsg->prefix_flags = 0;
5115         if (pinfo->onlink)
5116                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5117         if (pinfo->autoconf)
5118                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5119
5120         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5121                 goto nla_put_failure;
5122         ci.preferred_time = ntohl(pinfo->prefered);
5123         ci.valid_time = ntohl(pinfo->valid);
5124         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5125                 goto nla_put_failure;
5126         nlmsg_end(skb, nlh);
5127         return 0;
5128
5129 nla_put_failure:
5130         nlmsg_cancel(skb, nlh);
5131         return -EMSGSIZE;
5132 }
5133
5134 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5135                          struct prefix_info *pinfo)
5136 {
5137         struct sk_buff *skb;
5138         struct net *net = dev_net(idev->dev);
5139         int err = -ENOBUFS;
5140
5141         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5142         if (!skb)
5143                 goto errout;
5144
5145         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5146         if (err < 0) {
5147                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5148                 WARN_ON(err == -EMSGSIZE);
5149                 kfree_skb(skb);
5150                 goto errout;
5151         }
5152         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5153         return;
5154 errout:
5155         if (err < 0)
5156                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5157 }
5158
5159 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5160 {
5161         struct net *net = dev_net(ifp->idev->dev);
5162
5163         if (event)
5164                 ASSERT_RTNL();
5165
5166         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5167
5168         switch (event) {
5169         case RTM_NEWADDR:
5170                 /*
5171                  * If the address was optimistic
5172                  * we inserted the route at the start of
5173                  * our DAD process, so we don't need
5174                  * to do it again
5175                  */
5176                 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5177                         ip6_ins_rt(ifp->rt);
5178                 if (ifp->idev->cnf.forwarding)
5179                         addrconf_join_anycast(ifp);
5180                 if (!ipv6_addr_any(&ifp->peer_addr))
5181                         addrconf_prefix_route(&ifp->peer_addr, 128,
5182                                               ifp->idev->dev, 0, 0);
5183                 break;
5184         case RTM_DELADDR:
5185                 if (ifp->idev->cnf.forwarding)
5186                         addrconf_leave_anycast(ifp);
5187                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5188                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5189                         struct rt6_info *rt;
5190
5191                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5192                                                        ifp->idev->dev, 0, 0);
5193                         if (rt)
5194                                 ip6_del_rt(rt);
5195                 }
5196                 dst_hold(&ifp->rt->dst);
5197
5198                 ip6_del_rt(ifp->rt);
5199
5200                 rt_genid_bump_ipv6(net);
5201                 break;
5202         }
5203         atomic_inc(&net->ipv6.dev_addr_genid);
5204 }
5205
5206 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5207 {
5208         rcu_read_lock_bh();
5209         if (likely(ifp->idev->dead == 0))
5210                 __ipv6_ifa_notify(event, ifp);
5211         rcu_read_unlock_bh();
5212 }
5213
5214 #ifdef CONFIG_SYSCTL
5215
5216 static
5217 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5218                            void __user *buffer, size_t *lenp, loff_t *ppos)
5219 {
5220         int *valp = ctl->data;
5221         int val = *valp;
5222         loff_t pos = *ppos;
5223         struct ctl_table lctl;
5224         int ret;
5225
5226         /*
5227          * ctl->data points to idev->cnf.forwarding, we should
5228          * not modify it until we get the rtnl lock.
5229          */
5230         lctl = *ctl;
5231         lctl.data = &val;
5232
5233         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5234
5235         if (write)
5236                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5237         if (ret)
5238                 *ppos = pos;
5239         return ret;
5240 }
5241
5242 static
5243 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5244                         void __user *buffer, size_t *lenp, loff_t *ppos)
5245 {
5246         struct inet6_dev *idev = ctl->extra1;
5247         int min_mtu = IPV6_MIN_MTU;
5248         struct ctl_table lctl;
5249
5250         lctl = *ctl;
5251         lctl.extra1 = &min_mtu;
5252         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5253
5254         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5255 }
5256
5257 static void dev_disable_change(struct inet6_dev *idev)
5258 {
5259         struct netdev_notifier_info info;
5260
5261         if (!idev || !idev->dev)
5262                 return;
5263
5264         netdev_notifier_info_init(&info, idev->dev);
5265         if (idev->cnf.disable_ipv6)
5266                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5267         else
5268                 addrconf_notify(NULL, NETDEV_UP, &info);
5269 }
5270
5271 static void addrconf_disable_change(struct net *net, __s32 newf)
5272 {
5273         struct net_device *dev;
5274         struct inet6_dev *idev;
5275
5276         for_each_netdev(net, dev) {
5277                 idev = __in6_dev_get(dev);
5278                 if (idev) {
5279                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5280                         idev->cnf.disable_ipv6 = newf;
5281                         if (changed)
5282                                 dev_disable_change(idev);
5283                 }
5284         }
5285 }
5286
5287 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5288 {
5289         struct net *net;
5290         int old;
5291
5292         if (!rtnl_trylock())
5293                 return restart_syscall();
5294
5295         net = (struct net *)table->extra2;
5296         old = *p;
5297         *p = newf;
5298
5299         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5300                 rtnl_unlock();
5301                 return 0;
5302         }
5303
5304         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5305                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5306                 addrconf_disable_change(net, newf);
5307         } else if ((!newf) ^ (!old))
5308                 dev_disable_change((struct inet6_dev *)table->extra1);
5309
5310         rtnl_unlock();
5311         return 0;
5312 }
5313
5314 static
5315 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5316                             void __user *buffer, size_t *lenp, loff_t *ppos)
5317 {
5318         int *valp = ctl->data;
5319         int val = *valp;
5320         loff_t pos = *ppos;
5321         struct ctl_table lctl;
5322         int ret;
5323
5324         /*
5325          * ctl->data points to idev->cnf.disable_ipv6, we should
5326          * not modify it until we get the rtnl lock.
5327          */
5328         lctl = *ctl;
5329         lctl.data = &val;
5330
5331         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5332
5333         if (write)
5334                 ret = addrconf_disable_ipv6(ctl, valp, val);
5335         if (ret)
5336                 *ppos = pos;
5337         return ret;
5338 }
5339
5340 static
5341 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5342                               void __user *buffer, size_t *lenp, loff_t *ppos)
5343 {
5344         int *valp = ctl->data;
5345         int ret;
5346         int old, new;
5347
5348         old = *valp;
5349         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5350         new = *valp;
5351
5352         if (write && old != new) {
5353                 struct net *net = ctl->extra2;
5354
5355                 if (!rtnl_trylock())
5356                         return restart_syscall();
5357
5358                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5359                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5360                                                      NETCONFA_IFINDEX_DEFAULT,
5361                                                      net->ipv6.devconf_dflt);
5362                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5363                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5364                                                      NETCONFA_IFINDEX_ALL,
5365                                                      net->ipv6.devconf_all);
5366                 else {
5367                         struct inet6_dev *idev = ctl->extra1;
5368
5369                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5370                                                      idev->dev->ifindex,
5371                                                      &idev->cnf);
5372                 }
5373                 rtnl_unlock();
5374         }
5375
5376         return ret;
5377 }
5378
5379 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5380                                          void __user *buffer, size_t *lenp,
5381                                          loff_t *ppos)
5382 {
5383         int err;
5384         struct in6_addr addr;
5385         char str[IPV6_MAX_STRLEN];
5386         struct ctl_table lctl = *ctl;
5387         struct net *net = ctl->extra2;
5388         struct ipv6_stable_secret *secret = ctl->data;
5389
5390         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5391                 return -EIO;
5392
5393         lctl.maxlen = IPV6_MAX_STRLEN;
5394         lctl.data = str;
5395
5396         if (!rtnl_trylock())
5397                 return restart_syscall();
5398
5399         if (!write && !secret->initialized) {
5400                 err = -EIO;
5401                 goto out;
5402         }
5403
5404         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5405         if (err >= sizeof(str)) {
5406                 err = -EIO;
5407                 goto out;
5408         }
5409
5410         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5411         if (err || !write)
5412                 goto out;
5413
5414         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5415                 err = -EIO;
5416                 goto out;
5417         }
5418
5419         secret->initialized = true;
5420         secret->secret = addr;
5421
5422         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5423                 struct net_device *dev;
5424
5425                 for_each_netdev(net, dev) {
5426                         struct inet6_dev *idev = __in6_dev_get(dev);
5427
5428                         if (idev) {
5429                                 idev->addr_gen_mode =
5430                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5431                         }
5432                 }
5433         } else {
5434                 struct inet6_dev *idev = ctl->extra1;
5435
5436                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5437         }
5438
5439 out:
5440         rtnl_unlock();
5441
5442         return err;
5443 }
5444
5445 static
5446 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5447                                                 int write,
5448                                                 void __user *buffer,
5449                                                 size_t *lenp,
5450                                                 loff_t *ppos)
5451 {
5452         int *valp = ctl->data;
5453         int val = *valp;
5454         loff_t pos = *ppos;
5455         struct ctl_table lctl;
5456         int ret;
5457
5458         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5459          * we should not modify it until we get the rtnl lock.
5460          */
5461         lctl = *ctl;
5462         lctl.data = &val;
5463
5464         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5465
5466         if (write)
5467                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5468         if (ret)
5469                 *ppos = pos;
5470         return ret;
5471 }
5472
5473 static struct addrconf_sysctl_table
5474 {
5475         struct ctl_table_header *sysctl_header;
5476         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5477 } addrconf_sysctl __read_mostly = {
5478         .sysctl_header = NULL,
5479         .addrconf_vars = {
5480                 {
5481                         .procname       = "forwarding",
5482                         .data           = &ipv6_devconf.forwarding,
5483                         .maxlen         = sizeof(int),
5484                         .mode           = 0644,
5485                         .proc_handler   = addrconf_sysctl_forward,
5486                 },
5487                 {
5488                         .procname       = "hop_limit",
5489                         .data           = &ipv6_devconf.hop_limit,
5490                         .maxlen         = sizeof(int),
5491                         .mode           = 0644,
5492                         .proc_handler   = proc_dointvec,
5493                 },
5494                 {
5495                         .procname       = "mtu",
5496                         .data           = &ipv6_devconf.mtu6,
5497                         .maxlen         = sizeof(int),
5498                         .mode           = 0644,
5499                         .proc_handler   = addrconf_sysctl_mtu,
5500                 },
5501                 {
5502                         .procname       = "accept_ra",
5503                         .data           = &ipv6_devconf.accept_ra,
5504                         .maxlen         = sizeof(int),
5505                         .mode           = 0644,
5506                         .proc_handler   = proc_dointvec,
5507                 },
5508                 {
5509                         .procname       = "accept_redirects",
5510                         .data           = &ipv6_devconf.accept_redirects,
5511                         .maxlen         = sizeof(int),
5512                         .mode           = 0644,
5513                         .proc_handler   = proc_dointvec,
5514                 },
5515                 {
5516                         .procname       = "autoconf",
5517                         .data           = &ipv6_devconf.autoconf,
5518                         .maxlen         = sizeof(int),
5519                         .mode           = 0644,
5520                         .proc_handler   = proc_dointvec,
5521                 },
5522                 {
5523                         .procname       = "dad_transmits",
5524                         .data           = &ipv6_devconf.dad_transmits,
5525                         .maxlen         = sizeof(int),
5526                         .mode           = 0644,
5527                         .proc_handler   = proc_dointvec,
5528                 },
5529                 {
5530                         .procname       = "router_solicitations",
5531                         .data           = &ipv6_devconf.rtr_solicits,
5532                         .maxlen         = sizeof(int),
5533                         .mode           = 0644,
5534                         .proc_handler   = proc_dointvec,
5535                 },
5536                 {
5537                         .procname       = "router_solicitation_interval",
5538                         .data           = &ipv6_devconf.rtr_solicit_interval,
5539                         .maxlen         = sizeof(int),
5540                         .mode           = 0644,
5541                         .proc_handler   = proc_dointvec_jiffies,
5542                 },
5543                 {
5544                         .procname       = "router_solicitation_delay",
5545                         .data           = &ipv6_devconf.rtr_solicit_delay,
5546                         .maxlen         = sizeof(int),
5547                         .mode           = 0644,
5548                         .proc_handler   = proc_dointvec_jiffies,
5549                 },
5550                 {
5551                         .procname       = "force_mld_version",
5552                         .data           = &ipv6_devconf.force_mld_version,
5553                         .maxlen         = sizeof(int),
5554                         .mode           = 0644,
5555                         .proc_handler   = proc_dointvec,
5556                 },
5557                 {
5558                         .procname       = "mldv1_unsolicited_report_interval",
5559                         .data           =
5560                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5561                         .maxlen         = sizeof(int),
5562                         .mode           = 0644,
5563                         .proc_handler   = proc_dointvec_ms_jiffies,
5564                 },
5565                 {
5566                         .procname       = "mldv2_unsolicited_report_interval",
5567                         .data           =
5568                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5569                         .maxlen         = sizeof(int),
5570                         .mode           = 0644,
5571                         .proc_handler   = proc_dointvec_ms_jiffies,
5572                 },
5573                 {
5574                         .procname       = "use_tempaddr",
5575                         .data           = &ipv6_devconf.use_tempaddr,
5576                         .maxlen         = sizeof(int),
5577                         .mode           = 0644,
5578                         .proc_handler   = proc_dointvec,
5579                 },
5580                 {
5581                         .procname       = "temp_valid_lft",
5582                         .data           = &ipv6_devconf.temp_valid_lft,
5583                         .maxlen         = sizeof(int),
5584                         .mode           = 0644,
5585                         .proc_handler   = proc_dointvec,
5586                 },
5587                 {
5588                         .procname       = "temp_prefered_lft",
5589                         .data           = &ipv6_devconf.temp_prefered_lft,
5590                         .maxlen         = sizeof(int),
5591                         .mode           = 0644,
5592                         .proc_handler   = proc_dointvec,
5593                 },
5594                 {
5595                         .procname       = "regen_max_retry",
5596                         .data           = &ipv6_devconf.regen_max_retry,
5597                         .maxlen         = sizeof(int),
5598                         .mode           = 0644,
5599                         .proc_handler   = proc_dointvec,
5600                 },
5601                 {
5602                         .procname       = "max_desync_factor",
5603                         .data           = &ipv6_devconf.max_desync_factor,
5604                         .maxlen         = sizeof(int),
5605                         .mode           = 0644,
5606                         .proc_handler   = proc_dointvec,
5607                 },
5608                 {
5609                         .procname       = "max_addresses",
5610                         .data           = &ipv6_devconf.max_addresses,
5611                         .maxlen         = sizeof(int),
5612                         .mode           = 0644,
5613                         .proc_handler   = proc_dointvec,
5614                 },
5615                 {
5616                         .procname       = "accept_ra_defrtr",
5617                         .data           = &ipv6_devconf.accept_ra_defrtr,
5618                         .maxlen         = sizeof(int),
5619                         .mode           = 0644,
5620                         .proc_handler   = proc_dointvec,
5621                 },
5622                 {
5623                         .procname       = "accept_ra_min_hop_limit",
5624                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5625                         .maxlen         = sizeof(int),
5626                         .mode           = 0644,
5627                         .proc_handler   = proc_dointvec,
5628                 },
5629                 {
5630                         .procname       = "accept_ra_pinfo",
5631                         .data           = &ipv6_devconf.accept_ra_pinfo,
5632                         .maxlen         = sizeof(int),
5633                         .mode           = 0644,
5634                         .proc_handler   = proc_dointvec,
5635                 },
5636 #ifdef CONFIG_IPV6_ROUTER_PREF
5637                 {
5638                         .procname       = "accept_ra_rtr_pref",
5639                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5640                         .maxlen         = sizeof(int),
5641                         .mode           = 0644,
5642                         .proc_handler   = proc_dointvec,
5643                 },
5644                 {
5645                         .procname       = "router_probe_interval",
5646                         .data           = &ipv6_devconf.rtr_probe_interval,
5647                         .maxlen         = sizeof(int),
5648                         .mode           = 0644,
5649                         .proc_handler   = proc_dointvec_jiffies,
5650                 },
5651 #ifdef CONFIG_IPV6_ROUTE_INFO
5652                 {
5653                         .procname       = "accept_ra_rt_info_max_plen",
5654                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5655                         .maxlen         = sizeof(int),
5656                         .mode           = 0644,
5657                         .proc_handler   = proc_dointvec,
5658                 },
5659 #endif
5660 #endif
5661                 {
5662                         .procname       = "proxy_ndp",
5663                         .data           = &ipv6_devconf.proxy_ndp,
5664                         .maxlen         = sizeof(int),
5665                         .mode           = 0644,
5666                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5667                 },
5668                 {
5669                         .procname       = "accept_source_route",
5670                         .data           = &ipv6_devconf.accept_source_route,
5671                         .maxlen         = sizeof(int),
5672                         .mode           = 0644,
5673                         .proc_handler   = proc_dointvec,
5674                 },
5675 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5676                 {
5677                         .procname       = "optimistic_dad",
5678                         .data           = &ipv6_devconf.optimistic_dad,
5679                         .maxlen         = sizeof(int),
5680                         .mode           = 0644,
5681                         .proc_handler   = proc_dointvec,
5682
5683                 },
5684                 {
5685                         .procname       = "use_optimistic",
5686                         .data           = &ipv6_devconf.use_optimistic,
5687                         .maxlen         = sizeof(int),
5688                         .mode           = 0644,
5689                         .proc_handler   = proc_dointvec,
5690
5691                 },
5692 #endif
5693 #ifdef CONFIG_IPV6_MROUTE
5694                 {
5695                         .procname       = "mc_forwarding",
5696                         .data           = &ipv6_devconf.mc_forwarding,
5697                         .maxlen         = sizeof(int),
5698                         .mode           = 0444,
5699                         .proc_handler   = proc_dointvec,
5700                 },
5701 #endif
5702                 {
5703                         .procname       = "disable_ipv6",
5704                         .data           = &ipv6_devconf.disable_ipv6,
5705                         .maxlen         = sizeof(int),
5706                         .mode           = 0644,
5707                         .proc_handler   = addrconf_sysctl_disable,
5708                 },
5709                 {
5710                         .procname       = "accept_dad",
5711                         .data           = &ipv6_devconf.accept_dad,
5712                         .maxlen         = sizeof(int),
5713                         .mode           = 0644,
5714                         .proc_handler   = proc_dointvec,
5715                 },
5716                 {
5717                         .procname       = "force_tllao",
5718                         .data           = &ipv6_devconf.force_tllao,
5719                         .maxlen         = sizeof(int),
5720                         .mode           = 0644,
5721                         .proc_handler   = proc_dointvec
5722                 },
5723                 {
5724                         .procname       = "ndisc_notify",
5725                         .data           = &ipv6_devconf.ndisc_notify,
5726                         .maxlen         = sizeof(int),
5727                         .mode           = 0644,
5728                         .proc_handler   = proc_dointvec
5729                 },
5730                 {
5731                         .procname       = "suppress_frag_ndisc",
5732                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5733                         .maxlen         = sizeof(int),
5734                         .mode           = 0644,
5735                         .proc_handler   = proc_dointvec
5736                 },
5737                 {
5738                         .procname       = "accept_ra_from_local",
5739                         .data           = &ipv6_devconf.accept_ra_from_local,
5740                         .maxlen         = sizeof(int),
5741                         .mode           = 0644,
5742                         .proc_handler   = proc_dointvec,
5743                 },
5744                 {
5745                         .procname       = "accept_ra_mtu",
5746                         .data           = &ipv6_devconf.accept_ra_mtu,
5747                         .maxlen         = sizeof(int),
5748                         .mode           = 0644,
5749                         .proc_handler   = proc_dointvec,
5750                 },
5751                 {
5752                         .procname       = "stable_secret",
5753                         .data           = &ipv6_devconf.stable_secret,
5754                         .maxlen         = IPV6_MAX_STRLEN,
5755                         .mode           = 0600,
5756                         .proc_handler   = addrconf_sysctl_stable_secret,
5757                 },
5758                 {
5759                         .procname       = "use_oif_addrs_only",
5760                         .data           = &ipv6_devconf.use_oif_addrs_only,
5761                         .maxlen         = sizeof(int),
5762                         .mode           = 0644,
5763                         .proc_handler   = proc_dointvec,
5764                 },
5765                 {
5766                         .procname       = "ignore_routes_with_linkdown",
5767                         .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5768                         .maxlen         = sizeof(int),
5769                         .mode           = 0644,
5770                         .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5771                 },
5772                 {
5773                         /* sentinel */
5774                 }
5775         },
5776 };
5777
5778 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5779                 struct inet6_dev *idev, struct ipv6_devconf *p)
5780 {
5781         int i;
5782         struct addrconf_sysctl_table *t;
5783         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5784
5785         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5786         if (!t)
5787                 goto out;
5788
5789         for (i = 0; t->addrconf_vars[i].data; i++) {
5790                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5791                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5792                 t->addrconf_vars[i].extra2 = net;
5793         }
5794
5795         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5796
5797         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5798         if (!t->sysctl_header)
5799                 goto free;
5800
5801         p->sysctl = t;
5802         return 0;
5803
5804 free:
5805         kfree(t);
5806 out:
5807         return -ENOBUFS;
5808 }
5809
5810 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5811 {
5812         struct addrconf_sysctl_table *t;
5813
5814         if (!p->sysctl)
5815                 return;
5816
5817         t = p->sysctl;
5818         p->sysctl = NULL;
5819         unregister_net_sysctl_table(t->sysctl_header);
5820         kfree(t);
5821 }
5822
5823 static int addrconf_sysctl_register(struct inet6_dev *idev)
5824 {
5825         int err;
5826
5827         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5828                 return -EINVAL;
5829
5830         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5831                                     &ndisc_ifinfo_sysctl_change);
5832         if (err)
5833                 return err;
5834         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5835                                          idev, &idev->cnf);
5836         if (err)
5837                 neigh_sysctl_unregister(idev->nd_parms);
5838
5839         return err;
5840 }
5841
5842 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5843 {
5844         __addrconf_sysctl_unregister(&idev->cnf);
5845         neigh_sysctl_unregister(idev->nd_parms);
5846 }
5847
5848
5849 #endif
5850
5851 static int __net_init addrconf_init_net(struct net *net)
5852 {
5853         int err = -ENOMEM;
5854         struct ipv6_devconf *all, *dflt;
5855
5856         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5857         if (!all)
5858                 goto err_alloc_all;
5859
5860         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5861         if (!dflt)
5862                 goto err_alloc_dflt;
5863
5864         /* these will be inherited by all namespaces */
5865         dflt->autoconf = ipv6_defaults.autoconf;
5866         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5867
5868         dflt->stable_secret.initialized = false;
5869         all->stable_secret.initialized = false;
5870
5871         net->ipv6.devconf_all = all;
5872         net->ipv6.devconf_dflt = dflt;
5873
5874 #ifdef CONFIG_SYSCTL
5875         err = __addrconf_sysctl_register(net, "all", NULL, all);
5876         if (err < 0)
5877                 goto err_reg_all;
5878
5879         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5880         if (err < 0)
5881                 goto err_reg_dflt;
5882 #endif
5883         return 0;
5884
5885 #ifdef CONFIG_SYSCTL
5886 err_reg_dflt:
5887         __addrconf_sysctl_unregister(all);
5888 err_reg_all:
5889         kfree(dflt);
5890 #endif
5891 err_alloc_dflt:
5892         kfree(all);
5893 err_alloc_all:
5894         return err;
5895 }
5896
5897 static void __net_exit addrconf_exit_net(struct net *net)
5898 {
5899 #ifdef CONFIG_SYSCTL
5900         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5901         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5902 #endif
5903         kfree(net->ipv6.devconf_dflt);
5904         kfree(net->ipv6.devconf_all);
5905 }
5906
5907 static struct pernet_operations addrconf_ops = {
5908         .init = addrconf_init_net,
5909         .exit = addrconf_exit_net,
5910 };
5911
5912 static struct rtnl_af_ops inet6_ops __read_mostly = {
5913         .family           = AF_INET6,
5914         .fill_link_af     = inet6_fill_link_af,
5915         .get_link_af_size = inet6_get_link_af_size,
5916         .validate_link_af = inet6_validate_link_af,
5917         .set_link_af      = inet6_set_link_af,
5918 };
5919
5920 /*
5921  *      Init / cleanup code
5922  */
5923
5924 int __init addrconf_init(void)
5925 {
5926         struct inet6_dev *idev;
5927         int i, err;
5928
5929         err = ipv6_addr_label_init();
5930         if (err < 0) {
5931                 pr_crit("%s: cannot initialize default policy table: %d\n",
5932                         __func__, err);
5933                 goto out;
5934         }
5935
5936         err = register_pernet_subsys(&addrconf_ops);
5937         if (err < 0)
5938                 goto out_addrlabel;
5939
5940         addrconf_wq = create_workqueue("ipv6_addrconf");
5941         if (!addrconf_wq) {
5942                 err = -ENOMEM;
5943                 goto out_nowq;
5944         }
5945
5946         /* The addrconf netdev notifier requires that loopback_dev
5947          * has it's ipv6 private information allocated and setup
5948          * before it can bring up and give link-local addresses
5949          * to other devices which are up.
5950          *
5951          * Unfortunately, loopback_dev is not necessarily the first
5952          * entry in the global dev_base list of net devices.  In fact,
5953          * it is likely to be the very last entry on that list.
5954          * So this causes the notifier registry below to try and
5955          * give link-local addresses to all devices besides loopback_dev
5956          * first, then loopback_dev, which cases all the non-loopback_dev
5957          * devices to fail to get a link-local address.
5958          *
5959          * So, as a temporary fix, allocate the ipv6 structure for
5960          * loopback_dev first by hand.
5961          * Longer term, all of the dependencies ipv6 has upon the loopback
5962          * device and it being up should be removed.
5963          */
5964         rtnl_lock();
5965         idev = ipv6_add_dev(init_net.loopback_dev);
5966         rtnl_unlock();
5967         if (IS_ERR(idev)) {
5968                 err = PTR_ERR(idev);
5969                 goto errlo;
5970         }
5971
5972         ip6_route_init_special_entries();
5973
5974         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5975                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5976
5977         register_netdevice_notifier(&ipv6_dev_notf);
5978
5979         addrconf_verify();
5980
5981         rtnl_af_register(&inet6_ops);
5982
5983         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5984                               NULL);
5985         if (err < 0)
5986                 goto errout;
5987
5988         /* Only the first call to __rtnl_register can fail */
5989         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5990         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5991         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5992                         inet6_dump_ifaddr, NULL);
5993         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5994                         inet6_dump_ifmcaddr, NULL);
5995         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5996                         inet6_dump_ifacaddr, NULL);
5997         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5998                         inet6_netconf_dump_devconf, NULL);
5999
6000         ipv6_addr_label_rtnl_register();
6001
6002         return 0;
6003 errout:
6004         rtnl_af_unregister(&inet6_ops);
6005         unregister_netdevice_notifier(&ipv6_dev_notf);
6006 errlo:
6007         destroy_workqueue(addrconf_wq);
6008 out_nowq:
6009         unregister_pernet_subsys(&addrconf_ops);
6010 out_addrlabel:
6011         ipv6_addr_label_cleanup();
6012 out:
6013         return err;
6014 }
6015
6016 void addrconf_cleanup(void)
6017 {
6018         struct net_device *dev;
6019         int i;
6020
6021         unregister_netdevice_notifier(&ipv6_dev_notf);
6022         unregister_pernet_subsys(&addrconf_ops);
6023         ipv6_addr_label_cleanup();
6024
6025         rtnl_lock();
6026
6027         __rtnl_af_unregister(&inet6_ops);
6028
6029         /* clean dev list */
6030         for_each_netdev(&init_net, dev) {
6031                 if (__in6_dev_get(dev) == NULL)
6032                         continue;
6033                 addrconf_ifdown(dev, 1);
6034         }
6035         addrconf_ifdown(init_net.loopback_dev, 2);
6036
6037         /*
6038          *      Check hash table.
6039          */
6040         spin_lock_bh(&addrconf_hash_lock);
6041         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6042                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6043         spin_unlock_bh(&addrconf_hash_lock);
6044         cancel_delayed_work(&addr_chk_work);
6045         rtnl_unlock();
6046
6047         destroy_workqueue(addrconf_wq);
6048 }