net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/inet.h>
  50 #include <linux/in6.h>
  51 #include <linux/netdevice.h>
  52 #include <linux/if_addr.h>
  53 #include <linux/if_arp.h>
  54 #include <linux/if_arcnet.h>
  55 #include <linux/if_infiniband.h>
  56 #include <linux/route.h>
  57 #include <linux/inetdevice.h>
  58 #include <linux/init.h>
  59 #include <linux/slab.h>
  60 #ifdef CONFIG_SYSCTL
  61 #include <linux/sysctl.h>
  62 #endif
  63 #include <linux/capability.h>
  64 #include <linux/delay.h>
  65 #include <linux/notifier.h>
  66 #include <linux/string.h>
  67 #include <linux/hash.h>
  68
  69 #include <net/net_namespace.h>
  70 #include <net/sock.h>
  71 #include <net/snmp.h>
  72
  73 #include <net/6lowpan.h>
  74 #include <net/firewire.h>
  75 #include <net/ipv6.h>
  76 #include <net/protocol.h>
  77 #include <net/ndisc.h>
  78 #include <net/ip6_route.h>
  79 #include <net/addrconf.h>
  80 #include <net/tcp.h>
  81 #include <net/ip.h>
  82 #include <net/netlink.h>
  83 #include <net/pkt_sched.h>
  84 #include <net/l3mdev.h>
  85 #include <linux/if_tunnel.h>
  86 #include <linux/rtnetlink.h>
  87 #include <linux/netconf.h>
  88 #include <linux/random.h>
  89 #include <linux/uaccess.h>
  90 #include <asm/unaligned.h>
  91
  92 #include <linux/proc_fs.h>
  93 #include <linux/seq_file.h>
  94 #include <linux/export.h>
  95
  96 /* Set to 3 to get tracing... */
  97 #define ACONF_DEBUG 2
  98
  99 #if ACONF_DEBUG >= 3
 100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 101 #else
 102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 103 #endif
 104
 105 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 106
 107 #define IPV6_MAX_STRLEN \
 108         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 109
 110 static inline u32 cstamp_delta(unsigned long cstamp)
 111 {
 112         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 113 }
 114
 115 #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         .keep_addr_on_down      = 0,
 220 };
 221
 222 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 223         .forwarding             = 0,
 224         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 225         .mtu6                   = IPV6_MIN_MTU,
 226         .accept_ra              = 1,
 227         .accept_redirects       = 1,
 228         .autoconf               = 1,
 229         .force_mld_version      = 0,
 230         .mldv1_unsolicited_report_interval = 10 * HZ,
 231         .mldv2_unsolicited_report_interval = HZ,
 232         .dad_transmits          = 1,
 233         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 234         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 235         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 236         .use_tempaddr           = 0,
 237         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 238         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 239         .regen_max_retry        = REGEN_MAX_RETRY,
 240         .max_desync_factor      = MAX_DESYNC_FACTOR,
 241         .max_addresses          = IPV6_MAX_ADDRESSES,
 242         .accept_ra_defrtr       = 1,
 243         .accept_ra_from_local   = 0,
 244         .accept_ra_min_hop_limit= 1,
 245         .accept_ra_pinfo        = 1,
 246 #ifdef CONFIG_IPV6_ROUTER_PREF
 247         .accept_ra_rtr_pref     = 1,
 248         .rtr_probe_interval     = 60 * HZ,
 249 #ifdef CONFIG_IPV6_ROUTE_INFO
 250         .accept_ra_rt_info_max_plen = 0,
 251 #endif
 252 #endif
 253         .proxy_ndp              = 0,
 254         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 255         .disable_ipv6           = 0,
 256         .accept_dad             = 1,
 257         .suppress_frag_ndisc    = 1,
 258         .accept_ra_mtu          = 1,
 259         .stable_secret          = {
 260                 .initialized = false,
 261         },
 262         .use_oif_addrs_only     = 0,
 263         .ignore_routes_with_linkdown = 0,
 264         .keep_addr_on_down      = 0,
 265 };
 266
 267 /* Check if a valid qdisc is available */
 268 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 269 {
 270         return !qdisc_tx_is_noop(dev);
 271 }
 272
 273 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 274 {
 275         if (del_timer(&idev->rs_timer))
 276                 __in6_dev_put(idev);
 277 }
 278
 279 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 280 {
 281         if (cancel_delayed_work(&ifp->dad_work))
 282                 __in6_ifa_put(ifp);
 283 }
 284
 285 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 286                                   unsigned long when)
 287 {
 288         if (!timer_pending(&idev->rs_timer))
 289                 in6_dev_hold(idev);
 290         mod_timer(&idev->rs_timer, jiffies + when);
 291 }
 292
 293 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 294                                    unsigned long delay)
 295 {
 296         if (!delayed_work_pending(&ifp->dad_work))
 297                 in6_ifa_hold(ifp);
 298         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 299 }
 300
 301 static int snmp6_alloc_dev(struct inet6_dev *idev)
 302 {
 303         int i;
 304
 305         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 306         if (!idev->stats.ipv6)
 307                 goto err_ip;
 308
 309         for_each_possible_cpu(i) {
 310                 struct ipstats_mib *addrconf_stats;
 311                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 312                 u64_stats_init(&addrconf_stats->syncp);
 313         }
 314
 315
 316         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 317                                         GFP_KERNEL);
 318         if (!idev->stats.icmpv6dev)
 319                 goto err_icmp;
 320         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 321                                            GFP_KERNEL);
 322         if (!idev->stats.icmpv6msgdev)
 323                 goto err_icmpmsg;
 324
 325         return 0;
 326
 327 err_icmpmsg:
 328         kfree(idev->stats.icmpv6dev);
 329 err_icmp:
 330         free_percpu(idev->stats.ipv6);
 331 err_ip:
 332         return -ENOMEM;
 333 }
 334
 335 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 336 {
 337         struct inet6_dev *ndev;
 338         int err = -ENOMEM;
 339
 340         ASSERT_RTNL();
 341
 342         if (dev->mtu < IPV6_MIN_MTU)
 343                 return ERR_PTR(-EINVAL);
 344
 345         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 346         if (!ndev)
 347                 return ERR_PTR(err);
 348
 349         rwlock_init(&ndev->lock);
 350         ndev->dev = dev;
 351         INIT_LIST_HEAD(&ndev->addr_list);
 352         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 353                     (unsigned long)ndev);
 354         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 355
 356         if (ndev->cnf.stable_secret.initialized)
 357                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 358         else
 359                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
 360
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (!ndev->nd_parms) {
 365                 kfree(ndev);
 366                 return ERR_PTR(err);
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG(KERN_WARNING
 375                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name);
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 dev_put(dev);
 379                 kfree(ndev);
 380                 return ERR_PTR(err);
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG(KERN_WARNING
 385                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name);
 387                 goto err_release;
 388         }
 389
 390         /* One reference from device.  We must do this before
 391          * we invoke __ipv6_regen_rndid().
 392          */
 393         in6_dev_hold(ndev);
 394
 395         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 396                 ndev->cnf.accept_dad = -1;
 397
 398 #if IS_ENABLED(CONFIG_IPV6_SIT)
 399         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 400                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 401                 ndev->cnf.rtr_solicits = 0;
 402         }
 403 #endif
 404
 405         INIT_LIST_HEAD(&ndev->tempaddr_list);
 406         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 407         if ((dev->flags&IFF_LOOPBACK) ||
 408             dev->type == ARPHRD_TUNNEL ||
 409             dev->type == ARPHRD_TUNNEL6 ||
 410             dev->type == ARPHRD_SIT ||
 411             dev->type == ARPHRD_NONE) {
 412                 ndev->cnf.use_tempaddr = -1;
 413         } else {
 414                 in6_dev_hold(ndev);
 415                 ipv6_regen_rndid((unsigned long) ndev);
 416         }
 417
 418         ndev->token = in6addr_any;
 419
 420         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 421                 ndev->if_flags |= IF_READY;
 422
 423         ipv6_mc_init_dev(ndev);
 424         ndev->tstamp = jiffies;
 425         err = addrconf_sysctl_register(ndev);
 426         if (err) {
 427                 ipv6_mc_destroy_dev(ndev);
 428                 del_timer(&ndev->regen_timer);
 429                 snmp6_unregister_dev(ndev);
 430                 goto err_release;
 431         }
 432         /* protected by rtnl_lock */
 433         rcu_assign_pointer(dev->ip6_ptr, ndev);
 434
 435         /* Join interface-local all-node multicast group */
 436         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 437
 438         /* Join all-node multicast group */
 439         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 440
 441         /* Join all-router multicast group if forwarding is set */
 442         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 443                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 444
 445         return ndev;
 446
 447 err_release:
 448         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 449         ndev->dead = 1;
 450         in6_dev_finish_destroy(ndev);
 451         return ERR_PTR(err);
 452 }
 453
 454 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 455 {
 456         struct inet6_dev *idev;
 457
 458         ASSERT_RTNL();
 459
 460         idev = __in6_dev_get(dev);
 461         if (!idev) {
 462                 idev = ipv6_add_dev(dev);
 463                 if (IS_ERR(idev))
 464                         return NULL;
 465         }
 466
 467         if (dev->flags&IFF_UP)
 468                 ipv6_mc_up(idev);
 469         return idev;
 470 }
 471
 472 static int inet6_netconf_msgsize_devconf(int type)
 473 {
 474         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 475                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 476         bool all = false;
 477
 478         if (type == NETCONFA_ALL)
 479                 all = true;
 480
 481         if (all || type == NETCONFA_FORWARDING)
 482                 size += nla_total_size(4);
 483 #ifdef CONFIG_IPV6_MROUTE
 484         if (all || type == NETCONFA_MC_FORWARDING)
 485                 size += nla_total_size(4);
 486 #endif
 487         if (all || type == NETCONFA_PROXY_NEIGH)
 488                 size += nla_total_size(4);
 489
 490         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 491                 size += nla_total_size(4);
 492
 493         return size;
 494 }
 495
 496 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 497                                       struct ipv6_devconf *devconf, u32 portid,
 498                                       u32 seq, int event, unsigned int flags,
 499                                       int type)
 500 {
 501         struct nlmsghdr  *nlh;
 502         struct netconfmsg *ncm;
 503         bool all = false;
 504
 505         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 506                         flags);
 507         if (!nlh)
 508                 return -EMSGSIZE;
 509
 510         if (type == NETCONFA_ALL)
 511                 all = true;
 512
 513         ncm = nlmsg_data(nlh);
 514         ncm->ncm_family = AF_INET6;
 515
 516         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 517                 goto nla_put_failure;
 518
 519         if ((all || type == NETCONFA_FORWARDING) &&
 520             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 521                 goto nla_put_failure;
 522 #ifdef CONFIG_IPV6_MROUTE
 523         if ((all || type == NETCONFA_MC_FORWARDING) &&
 524             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 525                         devconf->mc_forwarding) < 0)
 526                 goto nla_put_failure;
 527 #endif
 528         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 529             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 530                 goto nla_put_failure;
 531
 532         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 533             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 534                         devconf->ignore_routes_with_linkdown) < 0)
 535                 goto nla_put_failure;
 536
 537         nlmsg_end(skb, nlh);
 538         return 0;
 539
 540 nla_put_failure:
 541         nlmsg_cancel(skb, nlh);
 542         return -EMSGSIZE;
 543 }
 544
 545 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 546                                   struct ipv6_devconf *devconf)
 547 {
 548         struct sk_buff *skb;
 549         int err = -ENOBUFS;
 550
 551         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 552         if (!skb)
 553                 goto errout;
 554
 555         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 556                                          RTM_NEWNETCONF, 0, type);
 557         if (err < 0) {
 558                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 559                 WARN_ON(err == -EMSGSIZE);
 560                 kfree_skb(skb);
 561                 goto errout;
 562         }
 563         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 564         return;
 565 errout:
 566         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 567 }
 568
 569 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 570         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 571         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 572         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 573         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 574 };
 575
 576 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 577                                      struct nlmsghdr *nlh)
 578 {
 579         struct net *net = sock_net(in_skb->sk);
 580         struct nlattr *tb[NETCONFA_MAX+1];
 581         struct netconfmsg *ncm;
 582         struct sk_buff *skb;
 583         struct ipv6_devconf *devconf;
 584         struct inet6_dev *in6_dev;
 585         struct net_device *dev;
 586         int ifindex;
 587         int err;
 588
 589         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 590                           devconf_ipv6_policy);
 591         if (err < 0)
 592                 goto errout;
 593
 594         err = -EINVAL;
 595         if (!tb[NETCONFA_IFINDEX])
 596                 goto errout;
 597
 598         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 599         switch (ifindex) {
 600         case NETCONFA_IFINDEX_ALL:
 601                 devconf = net->ipv6.devconf_all;
 602                 break;
 603         case NETCONFA_IFINDEX_DEFAULT:
 604                 devconf = net->ipv6.devconf_dflt;
 605                 break;
 606         default:
 607                 dev = __dev_get_by_index(net, ifindex);
 608                 if (!dev)
 609                         goto errout;
 610                 in6_dev = __in6_dev_get(dev);
 611                 if (!in6_dev)
 612                         goto errout;
 613                 devconf = &in6_dev->cnf;
 614                 break;
 615         }
 616
 617         err = -ENOBUFS;
 618         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
 619         if (!skb)
 620                 goto errout;
 621
 622         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 623                                          NETLINK_CB(in_skb).portid,
 624                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 625                                          NETCONFA_ALL);
 626         if (err < 0) {
 627                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 628                 WARN_ON(err == -EMSGSIZE);
 629                 kfree_skb(skb);
 630                 goto errout;
 631         }
 632         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 633 errout:
 634         return err;
 635 }
 636
 637 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 638                                       struct netlink_callback *cb)
 639 {
 640         struct net *net = sock_net(skb->sk);
 641         int h, s_h;
 642         int idx, s_idx;
 643         struct net_device *dev;
 644         struct inet6_dev *idev;
 645         struct hlist_head *head;
 646
 647         s_h = cb->args[0];
 648         s_idx = idx = cb->args[1];
 649
 650         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 651                 idx = 0;
 652                 head = &net->dev_index_head[h];
 653                 rcu_read_lock();
 654                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 655                           net->dev_base_seq;
 656                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 657                         if (idx < s_idx)
 658                                 goto cont;
 659                         idev = __in6_dev_get(dev);
 660                         if (!idev)
 661                                 goto cont;
 662
 663                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 664                                                        &idev->cnf,
 665                                                        NETLINK_CB(cb->skb).portid,
 666                                                        cb->nlh->nlmsg_seq,
 667                                                        RTM_NEWNETCONF,
 668                                                        NLM_F_MULTI,
 669                                                        NETCONFA_ALL) < 0) {
 670                                 rcu_read_unlock();
 671                                 goto done;
 672                         }
 673                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 674 cont:
 675                         idx++;
 676                 }
 677                 rcu_read_unlock();
 678         }
 679         if (h == NETDEV_HASHENTRIES) {
 680                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 681                                                net->ipv6.devconf_all,
 682                                                NETLINK_CB(cb->skb).portid,
 683                                                cb->nlh->nlmsg_seq,
 684                                                RTM_NEWNETCONF, NLM_F_MULTI,
 685                                                NETCONFA_ALL) < 0)
 686                         goto done;
 687                 else
 688                         h++;
 689         }
 690         if (h == NETDEV_HASHENTRIES + 1) {
 691                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 692                                                net->ipv6.devconf_dflt,
 693                                                NETLINK_CB(cb->skb).portid,
 694                                                cb->nlh->nlmsg_seq,
 695                                                RTM_NEWNETCONF, NLM_F_MULTI,
 696                                                NETCONFA_ALL) < 0)
 697                         goto done;
 698                 else
 699                         h++;
 700         }
 701 done:
 702         cb->args[0] = h;
 703         cb->args[1] = idx;
 704
 705         return skb->len;
 706 }
 707
 708 #ifdef CONFIG_SYSCTL
 709 static void dev_forward_change(struct inet6_dev *idev)
 710 {
 711         struct net_device *dev;
 712         struct inet6_ifaddr *ifa;
 713
 714         if (!idev)
 715                 return;
 716         dev = idev->dev;
 717         if (idev->cnf.forwarding)
 718                 dev_disable_lro(dev);
 719         if (dev->flags & IFF_MULTICAST) {
 720                 if (idev->cnf.forwarding) {
 721                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 722                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 723                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 724                 } else {
 725                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 726                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 727                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 728                 }
 729         }
 730
 731         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 732                 if (ifa->flags&IFA_F_TENTATIVE)
 733                         continue;
 734                 if (idev->cnf.forwarding)
 735                         addrconf_join_anycast(ifa);
 736                 else
 737                         addrconf_leave_anycast(ifa);
 738         }
 739         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 740                                      dev->ifindex, &idev->cnf);
 741 }
 742
 743
 744 static void addrconf_forward_change(struct net *net, __s32 newf)
 745 {
 746         struct net_device *dev;
 747         struct inet6_dev *idev;
 748
 749         for_each_netdev(net, dev) {
 750                 idev = __in6_dev_get(dev);
 751                 if (idev) {
 752                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 753                         idev->cnf.forwarding = newf;
 754                         if (changed)
 755                                 dev_forward_change(idev);
 756                 }
 757         }
 758 }
 759
 760 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 761 {
 762         struct net *net;
 763         int old;
 764
 765         if (!rtnl_trylock())
 766                 return restart_syscall();
 767
 768         net = (struct net *)table->extra2;
 769         old = *p;
 770         *p = newf;
 771
 772         if (p == &net->ipv6.devconf_dflt->forwarding) {
 773                 if ((!newf) ^ (!old))
 774                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 775                                                      NETCONFA_IFINDEX_DEFAULT,
 776                                                      net->ipv6.devconf_dflt);
 777                 rtnl_unlock();
 778                 return 0;
 779         }
 780
 781         if (p == &net->ipv6.devconf_all->forwarding) {
 782                 net->ipv6.devconf_dflt->forwarding = newf;
 783                 addrconf_forward_change(net, newf);
 784                 if ((!newf) ^ (!old))
 785                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 786                                                      NETCONFA_IFINDEX_ALL,
 787                                                      net->ipv6.devconf_all);
 788         } else if ((!newf) ^ (!old))
 789                 dev_forward_change((struct inet6_dev *)table->extra1);
 790         rtnl_unlock();
 791
 792         if (newf)
 793                 rt6_purge_dflt_routers(net);
 794         return 1;
 795 }
 796
 797 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 798 {
 799         struct net_device *dev;
 800         struct inet6_dev *idev;
 801
 802         for_each_netdev(net, dev) {
 803                 idev = __in6_dev_get(dev);
 804                 if (idev) {
 805                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 806
 807                         idev->cnf.ignore_routes_with_linkdown = newf;
 808                         if (changed)
 809                                 inet6_netconf_notify_devconf(dev_net(dev),
 810                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 811                                                              dev->ifindex,
 812                                                              &idev->cnf);
 813                 }
 814         }
 815 }
 816
 817 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 818 {
 819         struct net *net;
 820         int old;
 821
 822         if (!rtnl_trylock())
 823                 return restart_syscall();
 824
 825         net = (struct net *)table->extra2;
 826         old = *p;
 827         *p = newf;
 828
 829         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 830                 if ((!newf) ^ (!old))
 831                         inet6_netconf_notify_devconf(net,
 832                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 833                                                      NETCONFA_IFINDEX_DEFAULT,
 834                                                      net->ipv6.devconf_dflt);
 835                 rtnl_unlock();
 836                 return 0;
 837         }
 838
 839         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 840                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 841                 addrconf_linkdown_change(net, newf);
 842                 if ((!newf) ^ (!old))
 843                         inet6_netconf_notify_devconf(net,
 844                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 845                                                      NETCONFA_IFINDEX_ALL,
 846                                                      net->ipv6.devconf_all);
 847         }
 848         rtnl_unlock();
 849
 850         return 1;
 851 }
 852
 853 #endif
 854
 855 /* Nobody refers to this ifaddr, destroy it */
 856 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 857 {
 858         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 859
 860 #ifdef NET_REFCNT_DEBUG
 861         pr_debug("%s\n", __func__);
 862 #endif
 863
 864         in6_dev_put(ifp->idev);
 865
 866         if (cancel_delayed_work(&ifp->dad_work))
 867                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 868                           ifp);
 869
 870         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 871                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 872                 return;
 873         }
 874         ip6_rt_put(ifp->rt);
 875
 876         kfree_rcu(ifp, rcu);
 877 }
 878
 879 static void
 880 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 881 {
 882         struct list_head *p;
 883         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 884
 885         /*
 886          * Each device address list is sorted in order of scope -
 887          * global before linklocal.
 888          */
 889         list_for_each(p, &idev->addr_list) {
 890                 struct inet6_ifaddr *ifa
 891                         = list_entry(p, struct inet6_ifaddr, if_list);
 892                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 893                         break;
 894         }
 895
 896         list_add_tail(&ifp->if_list, p);
 897 }
 898
 899 static u32 inet6_addr_hash(const struct in6_addr *addr)
 900 {
 901         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 902 }
 903
 904 /* On success it returns ifp with increased reference count */
 905
 906 static struct inet6_ifaddr *
 907 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 908               const struct in6_addr *peer_addr, int pfxlen,
 909               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 910 {
 911         struct inet6_ifaddr *ifa = NULL;
 912         struct rt6_info *rt;
 913         unsigned int hash;
 914         int err = 0;
 915         int addr_type = ipv6_addr_type(addr);
 916
 917         if (addr_type == IPV6_ADDR_ANY ||
 918             addr_type & IPV6_ADDR_MULTICAST ||
 919             (!(idev->dev->flags & IFF_LOOPBACK) &&
 920              addr_type & IPV6_ADDR_LOOPBACK))
 921                 return ERR_PTR(-EADDRNOTAVAIL);
 922
 923         rcu_read_lock_bh();
 924         if (idev->dead) {
 925                 err = -ENODEV;                  /*XXX*/
 926                 goto out2;
 927         }
 928
 929         if (idev->cnf.disable_ipv6) {
 930                 err = -EACCES;
 931                 goto out2;
 932         }
 933
 934         spin_lock(&addrconf_hash_lock);
 935
 936         /* Ignore adding duplicate addresses on an interface */
 937         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 938                 ADBG("ipv6_add_addr: already assigned\n");
 939                 err = -EEXIST;
 940                 goto out;
 941         }
 942
 943         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 944
 945         if (!ifa) {
 946                 ADBG("ipv6_add_addr: malloc failed\n");
 947                 err = -ENOBUFS;
 948                 goto out;
 949         }
 950
 951         rt = addrconf_dst_alloc(idev, addr, false);
 952         if (IS_ERR(rt)) {
 953                 err = PTR_ERR(rt);
 954                 goto out;
 955         }
 956
 957         neigh_parms_data_state_setall(idev->nd_parms);
 958
 959         ifa->addr = *addr;
 960         if (peer_addr)
 961                 ifa->peer_addr = *peer_addr;
 962
 963         spin_lock_init(&ifa->lock);
 964         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 965         INIT_HLIST_NODE(&ifa->addr_lst);
 966         ifa->scope = scope;
 967         ifa->prefix_len = pfxlen;
 968         ifa->flags = flags | IFA_F_TENTATIVE;
 969         ifa->valid_lft = valid_lft;
 970         ifa->prefered_lft = prefered_lft;
 971         ifa->cstamp = ifa->tstamp = jiffies;
 972         ifa->tokenized = false;
 973
 974         ifa->rt = rt;
 975
 976         ifa->idev = idev;
 977         in6_dev_hold(idev);
 978         /* For caller */
 979         in6_ifa_hold(ifa);
 980
 981         /* Add to big hash table */
 982         hash = inet6_addr_hash(addr);
 983
 984         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 985         spin_unlock(&addrconf_hash_lock);
 986
 987         write_lock(&idev->lock);
 988         /* Add to inet6_dev unicast addr list. */
 989         ipv6_link_dev_addr(idev, ifa);
 990
 991         if (ifa->flags&IFA_F_TEMPORARY) {
 992                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 993                 in6_ifa_hold(ifa);
 994         }
 995
 996         in6_ifa_hold(ifa);
 997         write_unlock(&idev->lock);
 998 out2:
 999         rcu_read_unlock_bh();
1000
1001         if (likely(err == 0))
1002                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1003         else {
1004                 kfree(ifa);
1005                 ifa = ERR_PTR(err);
1006         }
1007
1008         return ifa;
1009 out:
1010         spin_unlock(&addrconf_hash_lock);
1011         goto out2;
1012 }
1013
1014 enum cleanup_prefix_rt_t {
1015         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1016         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1017         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1018 };
1019
1020 /*
1021  * Check, whether the prefix for ifp would still need a prefix route
1022  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1023  * constants.
1024  *
1025  * 1) we don't purge prefix if address was not permanent.
1026  *    prefix is managed by its own lifetime.
1027  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1028  * 3) if there are no addresses, delete prefix.
1029  * 4) if there are still other permanent address(es),
1030  *    corresponding prefix is still permanent.
1031  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1032  *    don't purge the prefix, assume user space is managing it.
1033  * 6) otherwise, update prefix lifetime to the
1034  *    longest valid lifetime among the corresponding
1035  *    addresses on the device.
1036  *    Note: subsequent RA will update lifetime.
1037  **/
1038 static enum cleanup_prefix_rt_t
1039 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1040 {
1041         struct inet6_ifaddr *ifa;
1042         struct inet6_dev *idev = ifp->idev;
1043         unsigned long lifetime;
1044         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1045
1046         *expires = jiffies;
1047
1048         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1049                 if (ifa == ifp)
1050                         continue;
1051                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1052                                        ifp->prefix_len))
1053                         continue;
1054                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1055                         return CLEANUP_PREFIX_RT_NOP;
1056
1057                 action = CLEANUP_PREFIX_RT_EXPIRE;
1058
1059                 spin_lock(&ifa->lock);
1060
1061                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1062                 /*
1063                  * Note: Because this address is
1064                  * not permanent, lifetime <
1065                  * LONG_MAX / HZ here.
1066                  */
1067                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1068                         *expires = ifa->tstamp + lifetime * HZ;
1069                 spin_unlock(&ifa->lock);
1070         }
1071
1072         return action;
1073 }
1074
1075 static void
1076 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1077 {
1078         struct rt6_info *rt;
1079
1080         rt = addrconf_get_prefix_route(&ifp->addr,
1081                                        ifp->prefix_len,
1082                                        ifp->idev->dev,
1083                                        0, RTF_GATEWAY | RTF_DEFAULT);
1084         if (rt) {
1085                 if (del_rt)
1086                         ip6_del_rt(rt);
1087                 else {
1088                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1089                                 rt6_set_expires(rt, expires);
1090                         ip6_rt_put(rt);
1091                 }
1092         }
1093 }
1094
1095
1096 /* This function wants to get referenced ifp and releases it before return */
1097
1098 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1099 {
1100         int state;
1101         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1102         unsigned long expires;
1103
1104         ASSERT_RTNL();
1105
1106         spin_lock_bh(&ifp->lock);
1107         state = ifp->state;
1108         ifp->state = INET6_IFADDR_STATE_DEAD;
1109         spin_unlock_bh(&ifp->lock);
1110
1111         if (state == INET6_IFADDR_STATE_DEAD)
1112                 goto out;
1113
1114         spin_lock_bh(&addrconf_hash_lock);
1115         hlist_del_init_rcu(&ifp->addr_lst);
1116         spin_unlock_bh(&addrconf_hash_lock);
1117
1118         write_lock_bh(&ifp->idev->lock);
1119
1120         if (ifp->flags&IFA_F_TEMPORARY) {
1121                 list_del(&ifp->tmp_list);
1122                 if (ifp->ifpub) {
1123                         in6_ifa_put(ifp->ifpub);
1124                         ifp->ifpub = NULL;
1125                 }
1126                 __in6_ifa_put(ifp);
1127         }
1128
1129         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1130                 action = check_cleanup_prefix_route(ifp, &expires);
1131
1132         list_del_init(&ifp->if_list);
1133         __in6_ifa_put(ifp);
1134
1135         write_unlock_bh(&ifp->idev->lock);
1136
1137         addrconf_del_dad_work(ifp);
1138
1139         ipv6_ifa_notify(RTM_DELADDR, ifp);
1140
1141         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1142
1143         if (action != CLEANUP_PREFIX_RT_NOP) {
1144                 cleanup_prefix_route(ifp, expires,
1145                         action == CLEANUP_PREFIX_RT_DEL);
1146         }
1147
1148         /* clean up prefsrc entries */
1149         rt6_remove_prefsrc(ifp);
1150 out:
1151         in6_ifa_put(ifp);
1152 }
1153
1154 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1155 {
1156         struct inet6_dev *idev = ifp->idev;
1157         struct in6_addr addr, *tmpaddr;
1158         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1159         unsigned long regen_advance;
1160         int tmp_plen;
1161         int ret = 0;
1162         u32 addr_flags;
1163         unsigned long now = jiffies;
1164
1165         write_lock_bh(&idev->lock);
1166         if (ift) {
1167                 spin_lock_bh(&ift->lock);
1168                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1169                 spin_unlock_bh(&ift->lock);
1170                 tmpaddr = &addr;
1171         } else {
1172                 tmpaddr = NULL;
1173         }
1174 retry:
1175         in6_dev_hold(idev);
1176         if (idev->cnf.use_tempaddr <= 0) {
1177                 write_unlock_bh(&idev->lock);
1178                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1179                 in6_dev_put(idev);
1180                 ret = -1;
1181                 goto out;
1182         }
1183         spin_lock_bh(&ifp->lock);
1184         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1185                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1186                 spin_unlock_bh(&ifp->lock);
1187                 write_unlock_bh(&idev->lock);
1188                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1189                         __func__);
1190                 in6_dev_put(idev);
1191                 ret = -1;
1192                 goto out;
1193         }
1194         in6_ifa_hold(ifp);
1195         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1196         __ipv6_try_regen_rndid(idev, tmpaddr);
1197         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1198         age = (now - ifp->tstamp) / HZ;
1199         tmp_valid_lft = min_t(__u32,
1200                               ifp->valid_lft,
1201                               idev->cnf.temp_valid_lft + age);
1202         tmp_prefered_lft = min_t(__u32,
1203                                  ifp->prefered_lft,
1204                                  idev->cnf.temp_prefered_lft + age -
1205                                  idev->cnf.max_desync_factor);
1206         tmp_plen = ifp->prefix_len;
1207         tmp_tstamp = ifp->tstamp;
1208         spin_unlock_bh(&ifp->lock);
1209
1210         regen_advance = idev->cnf.regen_max_retry *
1211                         idev->cnf.dad_transmits *
1212                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1213         write_unlock_bh(&idev->lock);
1214
1215         /* A temporary address is created only if this calculated Preferred
1216          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1217          * an implementation must not create a temporary address with a zero
1218          * Preferred Lifetime.
1219          * Use age calculation as in addrconf_verify to avoid unnecessary
1220          * temporary addresses being generated.
1221          */
1222         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1223         if (tmp_prefered_lft <= regen_advance + age) {
1224                 in6_ifa_put(ifp);
1225                 in6_dev_put(idev);
1226                 ret = -1;
1227                 goto out;
1228         }
1229
1230         addr_flags = IFA_F_TEMPORARY;
1231         /* set in addrconf_prefix_rcv() */
1232         if (ifp->flags & IFA_F_OPTIMISTIC)
1233                 addr_flags |= IFA_F_OPTIMISTIC;
1234
1235         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1236                             ipv6_addr_scope(&addr), addr_flags,
1237                             tmp_valid_lft, tmp_prefered_lft);
1238         if (IS_ERR(ift)) {
1239                 in6_ifa_put(ifp);
1240                 in6_dev_put(idev);
1241                 pr_info("%s: retry temporary address regeneration\n", __func__);
1242                 tmpaddr = &addr;
1243                 write_lock_bh(&idev->lock);
1244                 goto retry;
1245         }
1246
1247         spin_lock_bh(&ift->lock);
1248         ift->ifpub = ifp;
1249         ift->cstamp = now;
1250         ift->tstamp = tmp_tstamp;
1251         spin_unlock_bh(&ift->lock);
1252
1253         addrconf_dad_start(ift);
1254         in6_ifa_put(ift);
1255         in6_dev_put(idev);
1256 out:
1257         return ret;
1258 }
1259
1260 /*
1261  *      Choose an appropriate source address (RFC3484)
1262  */
1263 enum {
1264         IPV6_SADDR_RULE_INIT = 0,
1265         IPV6_SADDR_RULE_LOCAL,
1266         IPV6_SADDR_RULE_SCOPE,
1267         IPV6_SADDR_RULE_PREFERRED,
1268 #ifdef CONFIG_IPV6_MIP6
1269         IPV6_SADDR_RULE_HOA,
1270 #endif
1271         IPV6_SADDR_RULE_OIF,
1272         IPV6_SADDR_RULE_LABEL,
1273         IPV6_SADDR_RULE_PRIVACY,
1274         IPV6_SADDR_RULE_ORCHID,
1275         IPV6_SADDR_RULE_PREFIX,
1276 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1277         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1278 #endif
1279         IPV6_SADDR_RULE_MAX
1280 };
1281
1282 struct ipv6_saddr_score {
1283         int                     rule;
1284         int                     addr_type;
1285         struct inet6_ifaddr     *ifa;
1286         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1287         int                     scopedist;
1288         int                     matchlen;
1289 };
1290
1291 struct ipv6_saddr_dst {
1292         const struct in6_addr *addr;
1293         int ifindex;
1294         int scope;
1295         int label;
1296         unsigned int prefs;
1297 };
1298
1299 static inline int ipv6_saddr_preferred(int type)
1300 {
1301         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1302                 return 1;
1303         return 0;
1304 }
1305
1306 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1307 {
1308 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1309         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1310 #else
1311         return false;
1312 #endif
1313 }
1314
1315 static int ipv6_get_saddr_eval(struct net *net,
1316                                struct ipv6_saddr_score *score,
1317                                struct ipv6_saddr_dst *dst,
1318                                int i)
1319 {
1320         int ret;
1321
1322         if (i <= score->rule) {
1323                 switch (i) {
1324                 case IPV6_SADDR_RULE_SCOPE:
1325                         ret = score->scopedist;
1326                         break;
1327                 case IPV6_SADDR_RULE_PREFIX:
1328                         ret = score->matchlen;
1329                         break;
1330                 default:
1331                         ret = !!test_bit(i, score->scorebits);
1332                 }
1333                 goto out;
1334         }
1335
1336         switch (i) {
1337         case IPV6_SADDR_RULE_INIT:
1338                 /* Rule 0: remember if hiscore is not ready yet */
1339                 ret = !!score->ifa;
1340                 break;
1341         case IPV6_SADDR_RULE_LOCAL:
1342                 /* Rule 1: Prefer same address */
1343                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1344                 break;
1345         case IPV6_SADDR_RULE_SCOPE:
1346                 /* Rule 2: Prefer appropriate scope
1347                  *
1348                  *      ret
1349                  *       ^
1350                  *    -1 |  d 15
1351                  *    ---+--+-+---> scope
1352                  *       |
1353                  *       |             d is scope of the destination.
1354                  *  B-d  |  \
1355                  *       |   \      <- smaller scope is better if
1356                  *  B-15 |    \        if scope is enough for destination.
1357                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1358                  * d-C-1 | /
1359                  *       |/         <- greater is better
1360                  *   -C  /             if scope is not enough for destination.
1361                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1362                  *
1363                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1364                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1365                  * Assume B = 0 and we get C > 29.
1366                  */
1367                 ret = __ipv6_addr_src_scope(score->addr_type);
1368                 if (ret >= dst->scope)
1369                         ret = -ret;
1370                 else
1371                         ret -= 128;     /* 30 is enough */
1372                 score->scopedist = ret;
1373                 break;
1374         case IPV6_SADDR_RULE_PREFERRED:
1375             {
1376                 /* Rule 3: Avoid deprecated and optimistic addresses */
1377                 u8 avoid = IFA_F_DEPRECATED;
1378
1379                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1380                         avoid |= IFA_F_OPTIMISTIC;
1381                 ret = ipv6_saddr_preferred(score->addr_type) ||
1382                       !(score->ifa->flags & avoid);
1383                 break;
1384             }
1385 #ifdef CONFIG_IPV6_MIP6
1386         case IPV6_SADDR_RULE_HOA:
1387             {
1388                 /* Rule 4: Prefer home address */
1389                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1390                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1391                 break;
1392             }
1393 #endif
1394         case IPV6_SADDR_RULE_OIF:
1395                 /* Rule 5: Prefer outgoing interface */
1396                 ret = (!dst->ifindex ||
1397                        dst->ifindex == score->ifa->idev->dev->ifindex);
1398                 break;
1399         case IPV6_SADDR_RULE_LABEL:
1400                 /* Rule 6: Prefer matching label */
1401                 ret = ipv6_addr_label(net,
1402                                       &score->ifa->addr, score->addr_type,
1403                                       score->ifa->idev->dev->ifindex) == dst->label;
1404                 break;
1405         case IPV6_SADDR_RULE_PRIVACY:
1406             {
1407                 /* Rule 7: Prefer public address
1408                  * Note: prefer temporary address if use_tempaddr >= 2
1409                  */
1410                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1411                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1412                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1413                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1414                 break;
1415             }
1416         case IPV6_SADDR_RULE_ORCHID:
1417                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1418                  *          non-ORCHID vs non-ORCHID
1419                  */
1420                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1421                         ipv6_addr_orchid(dst->addr));
1422                 break;
1423         case IPV6_SADDR_RULE_PREFIX:
1424                 /* Rule 8: Use longest matching prefix */
1425                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1426                 if (ret > score->ifa->prefix_len)
1427                         ret = score->ifa->prefix_len;
1428                 score->matchlen = ret;
1429                 break;
1430 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1431         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1432                 /* Optimistic addresses still have lower precedence than other
1433                  * preferred addresses.
1434                  */
1435                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1436                 break;
1437 #endif
1438         default:
1439                 ret = 0;
1440         }
1441
1442         if (ret)
1443                 __set_bit(i, score->scorebits);
1444         score->rule = i;
1445 out:
1446         return ret;
1447 }
1448
1449 static int __ipv6_dev_get_saddr(struct net *net,
1450                                 struct ipv6_saddr_dst *dst,
1451                                 struct inet6_dev *idev,
1452                                 struct ipv6_saddr_score *scores,
1453                                 int hiscore_idx)
1454 {
1455         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1456
1457         read_lock_bh(&idev->lock);
1458         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1459                 int i;
1460
1461                 /*
1462                  * - Tentative Address (RFC2462 section 5.4)
1463                  *  - A tentative address is not considered
1464                  *    "assigned to an interface" in the traditional
1465                  *    sense, unless it is also flagged as optimistic.
1466                  * - Candidate Source Address (section 4)
1467                  *  - In any case, anycast addresses, multicast
1468                  *    addresses, and the unspecified address MUST
1469                  *    NOT be included in a candidate set.
1470                  */
1471                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1472                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1473                         continue;
1474
1475                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1476
1477                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1478                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1479                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1480                                             idev->dev->name);
1481                         continue;
1482                 }
1483
1484                 score->rule = -1;
1485                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1486
1487                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1488                         int minihiscore, miniscore;
1489
1490                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1491                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1492
1493                         if (minihiscore > miniscore) {
1494                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1495                                     score->scopedist > 0) {
1496                                         /*
1497                                          * special case:
1498                                          * each remaining entry
1499                                          * has too small (not enough)
1500                                          * scope, because ifa entries
1501                                          * are sorted by their scope
1502                                          * values.
1503                                          */
1504                                         goto out;
1505                                 }
1506                                 break;
1507                         } else if (minihiscore < miniscore) {
1508                                 if (hiscore->ifa)
1509                                         in6_ifa_put(hiscore->ifa);
1510
1511                                 in6_ifa_hold(score->ifa);
1512
1513                                 swap(hiscore, score);
1514                                 hiscore_idx = 1 - hiscore_idx;
1515
1516                                 /* restore our iterator */
1517                                 score->ifa = hiscore->ifa;
1518
1519                                 break;
1520                         }
1521                 }
1522         }
1523 out:
1524         read_unlock_bh(&idev->lock);
1525         return hiscore_idx;
1526 }
1527
1528 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1529                        const struct in6_addr *daddr, unsigned int prefs,
1530                        struct in6_addr *saddr)
1531 {
1532         struct ipv6_saddr_score scores[2], *hiscore;
1533         struct ipv6_saddr_dst dst;
1534         struct inet6_dev *idev;
1535         struct net_device *dev;
1536         int dst_type;
1537         bool use_oif_addr = false;
1538         int hiscore_idx = 0;
1539
1540         dst_type = __ipv6_addr_type(daddr);
1541         dst.addr = daddr;
1542         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1543         dst.scope = __ipv6_addr_src_scope(dst_type);
1544         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1545         dst.prefs = prefs;
1546
1547         scores[hiscore_idx].rule = -1;
1548         scores[hiscore_idx].ifa = NULL;
1549
1550         rcu_read_lock();
1551
1552         /* Candidate Source Address (section 4)
1553          *  - multicast and link-local destination address,
1554          *    the set of candidate source address MUST only
1555          *    include addresses assigned to interfaces
1556          *    belonging to the same link as the outgoing
1557          *    interface.
1558          * (- For site-local destination addresses, the
1559          *    set of candidate source addresses MUST only
1560          *    include addresses assigned to interfaces
1561          *    belonging to the same site as the outgoing
1562          *    interface.)
1563          *  - "It is RECOMMENDED that the candidate source addresses
1564          *    be the set of unicast addresses assigned to the
1565          *    interface that will be used to send to the destination
1566          *    (the 'outgoing' interface)." (RFC 6724)
1567          */
1568         if (dst_dev) {
1569                 idev = __in6_dev_get(dst_dev);
1570                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1571                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1572                     (idev && idev->cnf.use_oif_addrs_only)) {
1573                         use_oif_addr = true;
1574                 }
1575         }
1576
1577         if (use_oif_addr) {
1578                 if (idev)
1579                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1580         } else {
1581                 for_each_netdev_rcu(net, dev) {
1582                         idev = __in6_dev_get(dev);
1583                         if (!idev)
1584                                 continue;
1585                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1586                 }
1587         }
1588         rcu_read_unlock();
1589
1590         hiscore = &scores[hiscore_idx];
1591         if (!hiscore->ifa)
1592                 return -EADDRNOTAVAIL;
1593
1594         *saddr = hiscore->ifa->addr;
1595         in6_ifa_put(hiscore->ifa);
1596         return 0;
1597 }
1598 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1599
1600 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1601                       u32 banned_flags)
1602 {
1603         struct inet6_ifaddr *ifp;
1604         int err = -EADDRNOTAVAIL;
1605
1606         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1607                 if (ifp->scope > IFA_LINK)
1608                         break;
1609                 if (ifp->scope == IFA_LINK &&
1610                     !(ifp->flags & banned_flags)) {
1611                         *addr = ifp->addr;
1612                         err = 0;
1613                         break;
1614                 }
1615         }
1616         return err;
1617 }
1618
1619 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1620                     u32 banned_flags)
1621 {
1622         struct inet6_dev *idev;
1623         int err = -EADDRNOTAVAIL;
1624
1625         rcu_read_lock();
1626         idev = __in6_dev_get(dev);
1627         if (idev) {
1628                 read_lock_bh(&idev->lock);
1629                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1630                 read_unlock_bh(&idev->lock);
1631         }
1632         rcu_read_unlock();
1633         return err;
1634 }
1635
1636 static int ipv6_count_addresses(struct inet6_dev *idev)
1637 {
1638         int cnt = 0;
1639         struct inet6_ifaddr *ifp;
1640
1641         read_lock_bh(&idev->lock);
1642         list_for_each_entry(ifp, &idev->addr_list, if_list)
1643                 cnt++;
1644         read_unlock_bh(&idev->lock);
1645         return cnt;
1646 }
1647
1648 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1649                   const struct net_device *dev, int strict)
1650 {
1651         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1652 }
1653 EXPORT_SYMBOL(ipv6_chk_addr);
1654
1655 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1656                             const struct net_device *dev, int strict,
1657                             u32 banned_flags)
1658 {
1659         struct inet6_ifaddr *ifp;
1660         unsigned int hash = inet6_addr_hash(addr);
1661         u32 ifp_flags;
1662
1663         rcu_read_lock_bh();
1664         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1665                 if (!net_eq(dev_net(ifp->idev->dev), net))
1666                         continue;
1667                 /* Decouple optimistic from tentative for evaluation here.
1668                  * Ban optimistic addresses explicitly, when required.
1669                  */
1670                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1671                             ? (ifp->flags&~IFA_F_TENTATIVE)
1672                             : ifp->flags;
1673                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1674                     !(ifp_flags&banned_flags) &&
1675                     (!dev || ifp->idev->dev == dev ||
1676                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1677                         rcu_read_unlock_bh();
1678                         return 1;
1679                 }
1680         }
1681
1682         rcu_read_unlock_bh();
1683         return 0;
1684 }
1685 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1686
1687 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1688                                struct net_device *dev)
1689 {
1690         unsigned int hash = inet6_addr_hash(addr);
1691         struct inet6_ifaddr *ifp;
1692
1693         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1694                 if (!net_eq(dev_net(ifp->idev->dev), net))
1695                         continue;
1696                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1697                         if (!dev || ifp->idev->dev == dev)
1698                                 return true;
1699                 }
1700         }
1701         return false;
1702 }
1703
1704 /* Compares an address/prefix_len with addresses on device @dev.
1705  * If one is found it returns true.
1706  */
1707 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1708         const unsigned int prefix_len, struct net_device *dev)
1709 {
1710         struct inet6_dev *idev;
1711         struct inet6_ifaddr *ifa;
1712         bool ret = false;
1713
1714         rcu_read_lock();
1715         idev = __in6_dev_get(dev);
1716         if (idev) {
1717                 read_lock_bh(&idev->lock);
1718                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1719                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1720                         if (ret)
1721                                 break;
1722                 }
1723                 read_unlock_bh(&idev->lock);
1724         }
1725         rcu_read_unlock();
1726
1727         return ret;
1728 }
1729 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1730
1731 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1732 {
1733         struct inet6_dev *idev;
1734         struct inet6_ifaddr *ifa;
1735         int     onlink;
1736
1737         onlink = 0;
1738         rcu_read_lock();
1739         idev = __in6_dev_get(dev);
1740         if (idev) {
1741                 read_lock_bh(&idev->lock);
1742                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1743                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1744                                                    ifa->prefix_len);
1745                         if (onlink)
1746                                 break;
1747                 }
1748                 read_unlock_bh(&idev->lock);
1749         }
1750         rcu_read_unlock();
1751         return onlink;
1752 }
1753 EXPORT_SYMBOL(ipv6_chk_prefix);
1754
1755 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1756                                      struct net_device *dev, int strict)
1757 {
1758         struct inet6_ifaddr *ifp, *result = NULL;
1759         unsigned int hash = inet6_addr_hash(addr);
1760
1761         rcu_read_lock_bh();
1762         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1763                 if (!net_eq(dev_net(ifp->idev->dev), net))
1764                         continue;
1765                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1766                         if (!dev || ifp->idev->dev == dev ||
1767                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1768                                 result = ifp;
1769                                 in6_ifa_hold(ifp);
1770                                 break;
1771                         }
1772                 }
1773         }
1774         rcu_read_unlock_bh();
1775
1776         return result;
1777 }
1778
1779 /* Gets referenced address, destroys ifaddr */
1780
1781 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1782 {
1783         if (dad_failed)
1784                 ifp->flags |= IFA_F_DADFAILED;
1785
1786         if (ifp->flags&IFA_F_PERMANENT) {
1787                 spin_lock_bh(&ifp->lock);
1788                 addrconf_del_dad_work(ifp);
1789                 ifp->flags |= IFA_F_TENTATIVE;
1790                 spin_unlock_bh(&ifp->lock);
1791                 if (dad_failed)
1792                         ipv6_ifa_notify(0, ifp);
1793                 in6_ifa_put(ifp);
1794         } else if (ifp->flags&IFA_F_TEMPORARY) {
1795                 struct inet6_ifaddr *ifpub;
1796                 spin_lock_bh(&ifp->lock);
1797                 ifpub = ifp->ifpub;
1798                 if (ifpub) {
1799                         in6_ifa_hold(ifpub);
1800                         spin_unlock_bh(&ifp->lock);
1801                         ipv6_create_tempaddr(ifpub, ifp);
1802                         in6_ifa_put(ifpub);
1803                 } else {
1804                         spin_unlock_bh(&ifp->lock);
1805                 }
1806                 ipv6_del_addr(ifp);
1807         } else {
1808                 ipv6_del_addr(ifp);
1809         }
1810 }
1811
1812 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1813 {
1814         int err = -ENOENT;
1815
1816         spin_lock_bh(&ifp->lock);
1817         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1818                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1819                 err = 0;
1820         }
1821         spin_unlock_bh(&ifp->lock);
1822
1823         return err;
1824 }
1825
1826 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1827 {
1828         struct in6_addr addr;
1829         struct inet6_dev *idev = ifp->idev;
1830         struct net *net = dev_net(ifp->idev->dev);
1831
1832         if (addrconf_dad_end(ifp)) {
1833                 in6_ifa_put(ifp);
1834                 return;
1835         }
1836
1837         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1838                              ifp->idev->dev->name, &ifp->addr);
1839
1840         spin_lock_bh(&ifp->lock);
1841
1842         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1843                 int scope = ifp->scope;
1844                 u32 flags = ifp->flags;
1845                 struct in6_addr new_addr;
1846                 struct inet6_ifaddr *ifp2;
1847                 u32 valid_lft, preferred_lft;
1848                 int pfxlen = ifp->prefix_len;
1849                 int retries = ifp->stable_privacy_retry + 1;
1850
1851                 if (retries > net->ipv6.sysctl.idgen_retries) {
1852                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1853                                              ifp->idev->dev->name);
1854                         goto errdad;
1855                 }
1856
1857                 new_addr = ifp->addr;
1858                 if (ipv6_generate_stable_address(&new_addr, retries,
1859                                                  idev))
1860                         goto errdad;
1861
1862                 valid_lft = ifp->valid_lft;
1863                 preferred_lft = ifp->prefered_lft;
1864
1865                 spin_unlock_bh(&ifp->lock);
1866
1867                 if (idev->cnf.max_addresses &&
1868                     ipv6_count_addresses(idev) >=
1869                     idev->cnf.max_addresses)
1870                         goto lock_errdad;
1871
1872                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1873                                      ifp->idev->dev->name);
1874
1875                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1876                                      scope, flags, valid_lft,
1877                                      preferred_lft);
1878                 if (IS_ERR(ifp2))
1879                         goto lock_errdad;
1880
1881                 spin_lock_bh(&ifp2->lock);
1882                 ifp2->stable_privacy_retry = retries;
1883                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1884                 spin_unlock_bh(&ifp2->lock);
1885
1886                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1887                 in6_ifa_put(ifp2);
1888 lock_errdad:
1889                 spin_lock_bh(&ifp->lock);
1890         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1891                 addr.s6_addr32[0] = htonl(0xfe800000);
1892                 addr.s6_addr32[1] = 0;
1893
1894                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1895                     ipv6_addr_equal(&ifp->addr, &addr)) {
1896                         /* DAD failed for link-local based on MAC address */
1897                         idev->cnf.disable_ipv6 = 1;
1898
1899                         pr_info("%s: IPv6 being disabled!\n",
1900                                 ifp->idev->dev->name);
1901                 }
1902         }
1903
1904 errdad:
1905         /* transition from _POSTDAD to _ERRDAD */
1906         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1907         spin_unlock_bh(&ifp->lock);
1908
1909         addrconf_mod_dad_work(ifp, 0);
1910 }
1911
1912 /* Join to solicited addr multicast group.
1913  * caller must hold RTNL */
1914 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1915 {
1916         struct in6_addr maddr;
1917
1918         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1919                 return;
1920
1921         addrconf_addr_solict_mult(addr, &maddr);
1922         ipv6_dev_mc_inc(dev, &maddr);
1923 }
1924
1925 /* caller must hold RTNL */
1926 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1927 {
1928         struct in6_addr maddr;
1929
1930         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1931                 return;
1932
1933         addrconf_addr_solict_mult(addr, &maddr);
1934         __ipv6_dev_mc_dec(idev, &maddr);
1935 }
1936
1937 /* caller must hold RTNL */
1938 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1939 {
1940         struct in6_addr addr;
1941
1942         if (ifp->prefix_len >= 127) /* RFC 6164 */
1943                 return;
1944         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1945         if (ipv6_addr_any(&addr))
1946                 return;
1947         __ipv6_dev_ac_inc(ifp->idev, &addr);
1948 }
1949
1950 /* caller must hold RTNL */
1951 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1952 {
1953         struct in6_addr addr;
1954
1955         if (ifp->prefix_len >= 127) /* RFC 6164 */
1956                 return;
1957         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1958         if (ipv6_addr_any(&addr))
1959                 return;
1960         __ipv6_dev_ac_dec(ifp->idev, &addr);
1961 }
1962
1963 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1964 {
1965         if (dev->addr_len != EUI64_ADDR_LEN)
1966                 return -1;
1967         memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
1968         eui[0] ^= 2;
1969         return 0;
1970 }
1971
1972 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1973 {
1974         union fwnet_hwaddr *ha;
1975
1976         if (dev->addr_len != FWNET_ALEN)
1977                 return -1;
1978
1979         ha = (union fwnet_hwaddr *)dev->dev_addr;
1980
1981         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1982         eui[0] ^= 2;
1983         return 0;
1984 }
1985
1986 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1987 {
1988         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1989         if (dev->addr_len != ARCNET_ALEN)
1990                 return -1;
1991         memset(eui, 0, 7);
1992         eui[7] = *(u8 *)dev->dev_addr;
1993         return 0;
1994 }
1995
1996 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1997 {
1998         if (dev->addr_len != INFINIBAND_ALEN)
1999                 return -1;
2000         memcpy(eui, dev->dev_addr + 12, 8);
2001         eui[0] |= 2;
2002         return 0;
2003 }
2004
2005 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2006 {
2007         if (addr == 0)
2008                 return -1;
2009         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2010                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2011                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2012                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2013                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2014                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2015         eui[1] = 0;
2016         eui[2] = 0x5E;
2017         eui[3] = 0xFE;
2018         memcpy(eui + 4, &addr, 4);
2019         return 0;
2020 }
2021
2022 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2023 {
2024         if (dev->priv_flags & IFF_ISATAP)
2025                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2026         return -1;
2027 }
2028
2029 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2030 {
2031         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2032 }
2033
2034 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2035 {
2036         memcpy(eui, dev->perm_addr, 3);
2037         memcpy(eui + 5, dev->perm_addr + 3, 3);
2038         eui[3] = 0xFF;
2039         eui[4] = 0xFE;
2040         eui[0] ^= 2;
2041         return 0;
2042 }
2043
2044 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2045 {
2046         switch (dev->type) {
2047         case ARPHRD_ETHER:
2048         case ARPHRD_FDDI:
2049                 return addrconf_ifid_eui48(eui, dev);
2050         case ARPHRD_ARCNET:
2051                 return addrconf_ifid_arcnet(eui, dev);
2052         case ARPHRD_INFINIBAND:
2053                 return addrconf_ifid_infiniband(eui, dev);
2054         case ARPHRD_SIT:
2055                 return addrconf_ifid_sit(eui, dev);
2056         case ARPHRD_IPGRE:
2057                 return addrconf_ifid_gre(eui, dev);
2058         case ARPHRD_6LOWPAN:
2059                 return addrconf_ifid_eui64(eui, dev);
2060         case ARPHRD_IEEE1394:
2061                 return addrconf_ifid_ieee1394(eui, dev);
2062         case ARPHRD_TUNNEL6:
2063                 return addrconf_ifid_ip6tnl(eui, dev);
2064         }
2065         return -1;
2066 }
2067
2068 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2069 {
2070         int err = -1;
2071         struct inet6_ifaddr *ifp;
2072
2073         read_lock_bh(&idev->lock);
2074         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2075                 if (ifp->scope > IFA_LINK)
2076                         break;
2077                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2078                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2079                         err = 0;
2080                         break;
2081                 }
2082         }
2083         read_unlock_bh(&idev->lock);
2084         return err;
2085 }
2086
2087 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2088 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2089 {
2090 regen:
2091         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2092         idev->rndid[0] &= ~0x02;
2093
2094         /*
2095          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2096          * check if generated address is not inappropriate
2097          *
2098          *  - Reserved subnet anycast (RFC 2526)
2099          *      11111101 11....11 1xxxxxxx
2100          *  - ISATAP (RFC4214) 6.1
2101          *      00-00-5E-FE-xx-xx-xx-xx
2102          *  - value 0
2103          *  - XXX: already assigned to an address on the device
2104          */
2105         if (idev->rndid[0] == 0xfd &&
2106             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2107             (idev->rndid[7]&0x80))
2108                 goto regen;
2109         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2110                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2111                         goto regen;
2112                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2113                         goto regen;
2114         }
2115 }
2116
2117 static void ipv6_regen_rndid(unsigned long data)
2118 {
2119         struct inet6_dev *idev = (struct inet6_dev *) data;
2120         unsigned long expires;
2121
2122         rcu_read_lock_bh();
2123         write_lock_bh(&idev->lock);
2124
2125         if (idev->dead)
2126                 goto out;
2127
2128         __ipv6_regen_rndid(idev);
2129
2130         expires = jiffies +
2131                 idev->cnf.temp_prefered_lft * HZ -
2132                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2133                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2134                 idev->cnf.max_desync_factor * HZ;
2135         if (time_before(expires, jiffies)) {
2136                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2137                         __func__, idev->dev->name);
2138                 goto out;
2139         }
2140
2141         if (!mod_timer(&idev->regen_timer, expires))
2142                 in6_dev_hold(idev);
2143
2144 out:
2145         write_unlock_bh(&idev->lock);
2146         rcu_read_unlock_bh();
2147         in6_dev_put(idev);
2148 }
2149
2150 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2151 {
2152         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2153                 __ipv6_regen_rndid(idev);
2154 }
2155
2156 /*
2157  *      Add prefix route.
2158  */
2159
2160 static void
2161 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2162                       unsigned long expires, u32 flags)
2163 {
2164         struct fib6_config cfg = {
2165                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2166                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2167                 .fc_ifindex = dev->ifindex,
2168                 .fc_expires = expires,
2169                 .fc_dst_len = plen,
2170                 .fc_flags = RTF_UP | flags,
2171                 .fc_nlinfo.nl_net = dev_net(dev),
2172                 .fc_protocol = RTPROT_KERNEL,
2173         };
2174
2175         cfg.fc_dst = *pfx;
2176
2177         /* Prevent useless cloning on PtP SIT.
2178            This thing is done here expecting that the whole
2179            class of non-broadcast devices need not cloning.
2180          */
2181 #if IS_ENABLED(CONFIG_IPV6_SIT)
2182         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2183                 cfg.fc_flags |= RTF_NONEXTHOP;
2184 #endif
2185
2186         ip6_route_add(&cfg);
2187 }
2188
2189
2190 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2191                                                   int plen,
2192                                                   const struct net_device *dev,
2193                                                   u32 flags, u32 noflags)
2194 {
2195         struct fib6_node *fn;
2196         struct rt6_info *rt = NULL;
2197         struct fib6_table *table;
2198         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2199
2200         table = fib6_get_table(dev_net(dev), tb_id);
2201         if (!table)
2202                 return NULL;
2203
2204         read_lock_bh(&table->tb6_lock);
2205         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2206         if (!fn)
2207                 goto out;
2208
2209         noflags |= RTF_CACHE;
2210         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2211                 if (rt->dst.dev->ifindex != dev->ifindex)
2212                         continue;
2213                 if ((rt->rt6i_flags & flags) != flags)
2214                         continue;
2215                 if ((rt->rt6i_flags & noflags) != 0)
2216                         continue;
2217                 dst_hold(&rt->dst);
2218                 break;
2219         }
2220 out:
2221         read_unlock_bh(&table->tb6_lock);
2222         return rt;
2223 }
2224
2225
2226 /* Create "default" multicast route to the interface */
2227
2228 static void addrconf_add_mroute(struct net_device *dev)
2229 {
2230         struct fib6_config cfg = {
2231                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2232                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2233                 .fc_ifindex = dev->ifindex,
2234                 .fc_dst_len = 8,
2235                 .fc_flags = RTF_UP,
2236                 .fc_nlinfo.nl_net = dev_net(dev),
2237         };
2238
2239         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2240
2241         ip6_route_add(&cfg);
2242 }
2243
2244 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2245 {
2246         struct inet6_dev *idev;
2247
2248         ASSERT_RTNL();
2249
2250         idev = ipv6_find_idev(dev);
2251         if (!idev)
2252                 return ERR_PTR(-ENOBUFS);
2253
2254         if (idev->cnf.disable_ipv6)
2255                 return ERR_PTR(-EACCES);
2256
2257         /* Add default multicast route */
2258         if (!(dev->flags & IFF_LOOPBACK))
2259                 addrconf_add_mroute(dev);
2260
2261         return idev;
2262 }
2263
2264 static void manage_tempaddrs(struct inet6_dev *idev,
2265                              struct inet6_ifaddr *ifp,
2266                              __u32 valid_lft, __u32 prefered_lft,
2267                              bool create, unsigned long now)
2268 {
2269         u32 flags;
2270         struct inet6_ifaddr *ift;
2271
2272         read_lock_bh(&idev->lock);
2273         /* update all temporary addresses in the list */
2274         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2275                 int age, max_valid, max_prefered;
2276
2277                 if (ifp != ift->ifpub)
2278                         continue;
2279
2280                 /* RFC 4941 section 3.3:
2281                  * If a received option will extend the lifetime of a public
2282                  * address, the lifetimes of temporary addresses should
2283                  * be extended, subject to the overall constraint that no
2284                  * temporary addresses should ever remain "valid" or "preferred"
2285                  * for a time longer than (TEMP_VALID_LIFETIME) or
2286                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2287                  */
2288                 age = (now - ift->cstamp) / HZ;
2289                 max_valid = idev->cnf.temp_valid_lft - age;
2290                 if (max_valid < 0)
2291                         max_valid = 0;
2292
2293                 max_prefered = idev->cnf.temp_prefered_lft -
2294                                idev->cnf.max_desync_factor - age;
2295                 if (max_prefered < 0)
2296                         max_prefered = 0;
2297
2298                 if (valid_lft > max_valid)
2299                         valid_lft = max_valid;
2300
2301                 if (prefered_lft > max_prefered)
2302                         prefered_lft = max_prefered;
2303
2304                 spin_lock(&ift->lock);
2305                 flags = ift->flags;
2306                 ift->valid_lft = valid_lft;
2307                 ift->prefered_lft = prefered_lft;
2308                 ift->tstamp = now;
2309                 if (prefered_lft > 0)
2310                         ift->flags &= ~IFA_F_DEPRECATED;
2311
2312                 spin_unlock(&ift->lock);
2313                 if (!(flags&IFA_F_TENTATIVE))
2314                         ipv6_ifa_notify(0, ift);
2315         }
2316
2317         if ((create || list_empty(&idev->tempaddr_list)) &&
2318             idev->cnf.use_tempaddr > 0) {
2319                 /* When a new public address is created as described
2320                  * in [ADDRCONF], also create a new temporary address.
2321                  * Also create a temporary address if it's enabled but
2322                  * no temporary address currently exists.
2323                  */
2324                 read_unlock_bh(&idev->lock);
2325                 ipv6_create_tempaddr(ifp, NULL);
2326         } else {
2327                 read_unlock_bh(&idev->lock);
2328         }
2329 }
2330
2331 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2332 {
2333         return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2334                idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2335 }
2336
2337 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2338 {
2339         struct prefix_info *pinfo;
2340         __u32 valid_lft;
2341         __u32 prefered_lft;
2342         int addr_type;
2343         u32 addr_flags = 0;
2344         struct inet6_dev *in6_dev;
2345         struct net *net = dev_net(dev);
2346
2347         pinfo = (struct prefix_info *) opt;
2348
2349         if (len < sizeof(struct prefix_info)) {
2350                 ADBG("addrconf: prefix option too short\n");
2351                 return;
2352         }
2353
2354         /*
2355          *      Validation checks ([ADDRCONF], page 19)
2356          */
2357
2358         addr_type = ipv6_addr_type(&pinfo->prefix);
2359
2360         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2361                 return;
2362
2363         valid_lft = ntohl(pinfo->valid);
2364         prefered_lft = ntohl(pinfo->prefered);
2365
2366         if (prefered_lft > valid_lft) {
2367                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2368                 return;
2369         }
2370
2371         in6_dev = in6_dev_get(dev);
2372
2373         if (!in6_dev) {
2374                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2375                                     dev->name);
2376                 return;
2377         }
2378
2379         /*
2380          *      Two things going on here:
2381          *      1) Add routes for on-link prefixes
2382          *      2) Configure prefixes with the auto flag set
2383          */
2384
2385         if (pinfo->onlink) {
2386                 struct rt6_info *rt;
2387                 unsigned long rt_expires;
2388
2389                 /* Avoid arithmetic overflow. Really, we could
2390                  * save rt_expires in seconds, likely valid_lft,
2391                  * but it would require division in fib gc, that it
2392                  * not good.
2393                  */
2394                 if (HZ > USER_HZ)
2395                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2396                 else
2397                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2398
2399                 if (addrconf_finite_timeout(rt_expires))
2400                         rt_expires *= HZ;
2401
2402                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2403                                                pinfo->prefix_len,
2404                                                dev,
2405                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2406                                                RTF_GATEWAY | RTF_DEFAULT);
2407
2408                 if (rt) {
2409                         /* Autoconf prefix route */
2410                         if (valid_lft == 0) {
2411                                 ip6_del_rt(rt);
2412                                 rt = NULL;
2413                         } else if (addrconf_finite_timeout(rt_expires)) {
2414                                 /* not infinity */
2415                                 rt6_set_expires(rt, jiffies + rt_expires);
2416                         } else {
2417                                 rt6_clean_expires(rt);
2418                         }
2419                 } else if (valid_lft) {
2420                         clock_t expires = 0;
2421                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2422                         if (addrconf_finite_timeout(rt_expires)) {
2423                                 /* not infinity */
2424                                 flags |= RTF_EXPIRES;
2425                                 expires = jiffies_to_clock_t(rt_expires);
2426                         }
2427                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2428                                               dev, expires, flags);
2429                 }
2430                 ip6_rt_put(rt);
2431         }
2432
2433         /* Try to figure out our local address for this prefix */
2434
2435         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2436                 struct inet6_ifaddr *ifp;
2437                 struct in6_addr addr;
2438                 int create = 0, update_lft = 0;
2439                 bool tokenized = false;
2440
2441                 if (pinfo->prefix_len == 64) {
2442                         memcpy(&addr, &pinfo->prefix, 8);
2443
2444                         if (!ipv6_addr_any(&in6_dev->token)) {
2445                                 read_lock_bh(&in6_dev->lock);
2446                                 memcpy(addr.s6_addr + 8,
2447                                        in6_dev->token.s6_addr + 8, 8);
2448                                 read_unlock_bh(&in6_dev->lock);
2449                                 tokenized = true;
2450                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2451                                    !ipv6_generate_stable_address(&addr, 0,
2452                                                                  in6_dev)) {
2453                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2454                                 goto ok;
2455                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2456                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2457                                 in6_dev_put(in6_dev);
2458                                 return;
2459                         }
2460                         goto ok;
2461                 }
2462                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2463                                     pinfo->prefix_len);
2464                 in6_dev_put(in6_dev);
2465                 return;
2466
2467 ok:
2468
2469                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2470
2471                 if (!ifp && valid_lft) {
2472                         int max_addresses = in6_dev->cnf.max_addresses;
2473
2474 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2475                         if (in6_dev->cnf.optimistic_dad &&
2476                             !net->ipv6.devconf_all->forwarding && sllao)
2477                                 addr_flags |= IFA_F_OPTIMISTIC;
2478 #endif
2479
2480                         /* Do not allow to create too much of autoconfigured
2481                          * addresses; this would be too easy way to crash kernel.
2482                          */
2483                         if (!max_addresses ||
2484                             ipv6_count_addresses(in6_dev) < max_addresses)
2485                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2486                                                     pinfo->prefix_len,
2487                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2488                                                     addr_flags, valid_lft,
2489                                                     prefered_lft);
2490
2491                         if (IS_ERR_OR_NULL(ifp)) {
2492                                 in6_dev_put(in6_dev);
2493                                 return;
2494                         }
2495
2496                         update_lft = 0;
2497                         create = 1;
2498                         spin_lock_bh(&ifp->lock);
2499                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2500                         ifp->cstamp = jiffies;
2501                         ifp->tokenized = tokenized;
2502                         spin_unlock_bh(&ifp->lock);
2503                         addrconf_dad_start(ifp);
2504                 }
2505
2506                 if (ifp) {
2507                         u32 flags;
2508                         unsigned long now;
2509                         u32 stored_lft;
2510
2511                         /* update lifetime (RFC2462 5.5.3 e) */
2512                         spin_lock_bh(&ifp->lock);
2513                         now = jiffies;
2514                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2515                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2516                         else
2517                                 stored_lft = 0;
2518                         if (!update_lft && !create && stored_lft) {
2519                                 const u32 minimum_lft = min_t(u32,
2520                                         stored_lft, MIN_VALID_LIFETIME);
2521                                 valid_lft = max(valid_lft, minimum_lft);
2522
2523                                 /* RFC4862 Section 5.5.3e:
2524                                  * "Note that the preferred lifetime of the
2525                                  *  corresponding address is always reset to
2526                                  *  the Preferred Lifetime in the received
2527                                  *  Prefix Information option, regardless of
2528                                  *  whether the valid lifetime is also reset or
2529                                  *  ignored."
2530                                  *
2531                                  * So we should always update prefered_lft here.
2532                                  */
2533                                 update_lft = 1;
2534                         }
2535
2536                         if (update_lft) {
2537                                 ifp->valid_lft = valid_lft;
2538                                 ifp->prefered_lft = prefered_lft;
2539                                 ifp->tstamp = now;
2540                                 flags = ifp->flags;
2541                                 ifp->flags &= ~IFA_F_DEPRECATED;
2542                                 spin_unlock_bh(&ifp->lock);
2543
2544                                 if (!(flags&IFA_F_TENTATIVE))
2545                                         ipv6_ifa_notify(0, ifp);
2546                         } else
2547                                 spin_unlock_bh(&ifp->lock);
2548
2549                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2550                                          create, now);
2551
2552                         in6_ifa_put(ifp);
2553                         addrconf_verify();
2554                 }
2555         }
2556         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2557         in6_dev_put(in6_dev);
2558 }
2559
2560 /*
2561  *      Set destination address.
2562  *      Special case for SIT interfaces where we create a new "virtual"
2563  *      device.
2564  */
2565 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2566 {
2567         struct in6_ifreq ireq;
2568         struct net_device *dev;
2569         int err = -EINVAL;
2570
2571         rtnl_lock();
2572
2573         err = -EFAULT;
2574         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2575                 goto err_exit;
2576
2577         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2578
2579         err = -ENODEV;
2580         if (!dev)
2581                 goto err_exit;
2582
2583 #if IS_ENABLED(CONFIG_IPV6_SIT)
2584         if (dev->type == ARPHRD_SIT) {
2585                 const struct net_device_ops *ops = dev->netdev_ops;
2586                 struct ifreq ifr;
2587                 struct ip_tunnel_parm p;
2588
2589                 err = -EADDRNOTAVAIL;
2590                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2591                         goto err_exit;
2592
2593                 memset(&p, 0, sizeof(p));
2594                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2595                 p.iph.saddr = 0;
2596                 p.iph.version = 4;
2597                 p.iph.ihl = 5;
2598                 p.iph.protocol = IPPROTO_IPV6;
2599                 p.iph.ttl = 64;
2600                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2601
2602                 if (ops->ndo_do_ioctl) {
2603                         mm_segment_t oldfs = get_fs();
2604
2605                         set_fs(KERNEL_DS);
2606                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2607                         set_fs(oldfs);
2608                 } else
2609                         err = -EOPNOTSUPP;
2610
2611                 if (err == 0) {
2612                         err = -ENOBUFS;
2613                         dev = __dev_get_by_name(net, p.name);
2614                         if (!dev)
2615                                 goto err_exit;
2616                         err = dev_open(dev);
2617                 }
2618         }
2619 #endif
2620
2621 err_exit:
2622         rtnl_unlock();
2623         return err;
2624 }
2625
2626 static int ipv6_mc_config(struct sock *sk, bool join,
2627                           const struct in6_addr *addr, int ifindex)
2628 {
2629         int ret;
2630
2631         ASSERT_RTNL();
2632
2633         lock_sock(sk);
2634         if (join)
2635                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2636         else
2637                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2638         release_sock(sk);
2639
2640         return ret;
2641 }
2642
2643 /*
2644  *      Manual configuration of address on an interface
2645  */
2646 static int inet6_addr_add(struct net *net, int ifindex,
2647                           const struct in6_addr *pfx,
2648                           const struct in6_addr *peer_pfx,
2649                           unsigned int plen, __u32 ifa_flags,
2650                           __u32 prefered_lft, __u32 valid_lft)
2651 {
2652         struct inet6_ifaddr *ifp;
2653         struct inet6_dev *idev;
2654         struct net_device *dev;
2655         unsigned long timeout;
2656         clock_t expires;
2657         int scope;
2658         u32 flags;
2659
2660         ASSERT_RTNL();
2661
2662         if (plen > 128)
2663                 return -EINVAL;
2664
2665         /* check the lifetime */
2666         if (!valid_lft || prefered_lft > valid_lft)
2667                 return -EINVAL;
2668
2669         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2670                 return -EINVAL;
2671
2672         dev = __dev_get_by_index(net, ifindex);
2673         if (!dev)
2674                 return -ENODEV;
2675
2676         idev = addrconf_add_dev(dev);
2677         if (IS_ERR(idev))
2678                 return PTR_ERR(idev);
2679
2680         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2681                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2682                                          true, pfx, ifindex);
2683
2684                 if (ret < 0)
2685                         return ret;
2686         }
2687
2688         scope = ipv6_addr_scope(pfx);
2689
2690         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2691         if (addrconf_finite_timeout(timeout)) {
2692                 expires = jiffies_to_clock_t(timeout * HZ);
2693                 valid_lft = timeout;
2694                 flags = RTF_EXPIRES;
2695         } else {
2696                 expires = 0;
2697                 flags = 0;
2698                 ifa_flags |= IFA_F_PERMANENT;
2699         }
2700
2701         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2702         if (addrconf_finite_timeout(timeout)) {
2703                 if (timeout == 0)
2704                         ifa_flags |= IFA_F_DEPRECATED;
2705                 prefered_lft = timeout;
2706         }
2707
2708         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2709                             valid_lft, prefered_lft);
2710
2711         if (!IS_ERR(ifp)) {
2712                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2713                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2714                                               expires, flags);
2715                 }
2716
2717                 /*
2718                  * Note that section 3.1 of RFC 4429 indicates
2719                  * that the Optimistic flag should not be set for
2720                  * manually configured addresses
2721                  */
2722                 addrconf_dad_start(ifp);
2723                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2724                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2725                                          true, jiffies);
2726                 in6_ifa_put(ifp);
2727                 addrconf_verify_rtnl();
2728                 return 0;
2729         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2730                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2731                                false, pfx, ifindex);
2732         }
2733
2734         return PTR_ERR(ifp);
2735 }
2736
2737 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2738                           const struct in6_addr *pfx, unsigned int plen)
2739 {
2740         struct inet6_ifaddr *ifp;
2741         struct inet6_dev *idev;
2742         struct net_device *dev;
2743
2744         if (plen > 128)
2745                 return -EINVAL;
2746
2747         dev = __dev_get_by_index(net, ifindex);
2748         if (!dev)
2749                 return -ENODEV;
2750
2751         idev = __in6_dev_get(dev);
2752         if (!idev)
2753                 return -ENXIO;
2754
2755         read_lock_bh(&idev->lock);
2756         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2757                 if (ifp->prefix_len == plen &&
2758                     ipv6_addr_equal(pfx, &ifp->addr)) {
2759                         in6_ifa_hold(ifp);
2760                         read_unlock_bh(&idev->lock);
2761
2762                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2763                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2764                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2765                                                  jiffies);
2766                         ipv6_del_addr(ifp);
2767                         addrconf_verify_rtnl();
2768                         if (ipv6_addr_is_multicast(pfx)) {
2769                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2770                                                false, pfx, dev->ifindex);
2771                         }
2772                         return 0;
2773                 }
2774         }
2775         read_unlock_bh(&idev->lock);
2776         return -EADDRNOTAVAIL;
2777 }
2778
2779
2780 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2781 {
2782         struct in6_ifreq ireq;
2783         int err;
2784
2785         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2786                 return -EPERM;
2787
2788         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2789                 return -EFAULT;
2790
2791         rtnl_lock();
2792         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2793                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2794                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2795         rtnl_unlock();
2796         return err;
2797 }
2798
2799 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2800 {
2801         struct in6_ifreq ireq;
2802         int err;
2803
2804         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2805                 return -EPERM;
2806
2807         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2808                 return -EFAULT;
2809
2810         rtnl_lock();
2811         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2812                              ireq.ifr6_prefixlen);
2813         rtnl_unlock();
2814         return err;
2815 }
2816
2817 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2818                      int plen, int scope)
2819 {
2820         struct inet6_ifaddr *ifp;
2821
2822         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2823                             scope, IFA_F_PERMANENT,
2824                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2825         if (!IS_ERR(ifp)) {
2826                 spin_lock_bh(&ifp->lock);
2827                 ifp->flags &= ~IFA_F_TENTATIVE;
2828                 spin_unlock_bh(&ifp->lock);
2829                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2830                 in6_ifa_put(ifp);
2831         }
2832 }
2833
2834 #if IS_ENABLED(CONFIG_IPV6_SIT)
2835 static void sit_add_v4_addrs(struct inet6_dev *idev)
2836 {
2837         struct in6_addr addr;
2838         struct net_device *dev;
2839         struct net *net = dev_net(idev->dev);
2840         int scope, plen;
2841         u32 pflags = 0;
2842
2843         ASSERT_RTNL();
2844
2845         memset(&addr, 0, sizeof(struct in6_addr));
2846         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2847
2848         if (idev->dev->flags&IFF_POINTOPOINT) {
2849                 addr.s6_addr32[0] = htonl(0xfe800000);
2850                 scope = IFA_LINK;
2851                 plen = 64;
2852         } else {
2853                 scope = IPV6_ADDR_COMPATv4;
2854                 plen = 96;
2855                 pflags |= RTF_NONEXTHOP;
2856         }
2857
2858         if (addr.s6_addr32[3]) {
2859                 add_addr(idev, &addr, plen, scope);
2860                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2861                 return;
2862         }
2863
2864         for_each_netdev(net, dev) {
2865                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2866                 if (in_dev && (dev->flags & IFF_UP)) {
2867                         struct in_ifaddr *ifa;
2868
2869                         int flag = scope;
2870
2871                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2872
2873                                 addr.s6_addr32[3] = ifa->ifa_local;
2874
2875                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2876                                         continue;
2877                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2878                                         if (idev->dev->flags&IFF_POINTOPOINT)
2879                                                 continue;
2880                                         flag |= IFA_HOST;
2881                                 }
2882
2883                                 add_addr(idev, &addr, plen, flag);
2884                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2885                                                       pflags);
2886                         }
2887                 }
2888         }
2889 }
2890 #endif
2891
2892 static void init_loopback(struct net_device *dev)
2893 {
2894         struct inet6_dev  *idev;
2895         struct net_device *sp_dev;
2896         struct inet6_ifaddr *sp_ifa;
2897         struct rt6_info *sp_rt;
2898
2899         /* ::1 */
2900
2901         ASSERT_RTNL();
2902
2903         idev = ipv6_find_idev(dev);
2904         if (!idev) {
2905                 pr_debug("%s: add_dev failed\n", __func__);
2906                 return;
2907         }
2908
2909         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2910
2911         /* Add routes to other interface's IPv6 addresses */
2912         for_each_netdev(dev_net(dev), sp_dev) {
2913                 if (!strcmp(sp_dev->name, dev->name))
2914                         continue;
2915
2916                 idev = __in6_dev_get(sp_dev);
2917                 if (!idev)
2918                         continue;
2919
2920                 read_lock_bh(&idev->lock);
2921                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2922
2923                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2924                                 continue;
2925
2926                         if (sp_ifa->rt) {
2927                                 /* This dst has been added to garbage list when
2928                                  * lo device down, release this obsolete dst and
2929                                  * reallocate a new router for ifa.
2930                                  */
2931                                 if (sp_ifa->rt->dst.obsolete > 0) {
2932                                         ip6_rt_put(sp_ifa->rt);
2933                                         sp_ifa->rt = NULL;
2934                                 } else {
2935                                         continue;
2936                                 }
2937                         }
2938
2939                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2940
2941                         /* Failure cases are ignored */
2942                         if (!IS_ERR(sp_rt)) {
2943                                 sp_ifa->rt = sp_rt;
2944                                 ip6_ins_rt(sp_rt);
2945                         }
2946                 }
2947                 read_unlock_bh(&idev->lock);
2948         }
2949 }
2950
2951 static void addrconf_add_linklocal(struct inet6_dev *idev,
2952                                    const struct in6_addr *addr, u32 flags)
2953 {
2954         struct inet6_ifaddr *ifp;
2955         u32 addr_flags = flags | IFA_F_PERMANENT;
2956
2957 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2958         if (idev->cnf.optimistic_dad &&
2959             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2960                 addr_flags |= IFA_F_OPTIMISTIC;
2961 #endif
2962
2963         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2964                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2965         if (!IS_ERR(ifp)) {
2966                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2967                 addrconf_dad_start(ifp);
2968                 in6_ifa_put(ifp);
2969         }
2970 }
2971
2972 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2973 {
2974         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2975                 return true;
2976
2977         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2978             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2979                 return true;
2980
2981         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2982             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2983                 return true;
2984
2985         return false;
2986 }
2987
2988 static int ipv6_generate_stable_address(struct in6_addr *address,
2989                                         u8 dad_count,
2990                                         const struct inet6_dev *idev)
2991 {
2992         static DEFINE_SPINLOCK(lock);
2993         static __u32 digest[SHA_DIGEST_WORDS];
2994         static __u32 workspace[SHA_WORKSPACE_WORDS];
2995
2996         static union {
2997                 char __data[SHA_MESSAGE_BYTES];
2998                 struct {
2999                         struct in6_addr secret;
3000                         __be32 prefix[2];
3001                         unsigned char hwaddr[MAX_ADDR_LEN];
3002                         u8 dad_count;
3003                 } __packed;
3004         } data;
3005
3006         struct in6_addr secret;
3007         struct in6_addr temp;
3008         struct net *net = dev_net(idev->dev);
3009
3010         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3011
3012         if (idev->cnf.stable_secret.initialized)
3013                 secret = idev->cnf.stable_secret.secret;
3014         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3015                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3016         else
3017                 return -1;
3018
3019 retry:
3020         spin_lock_bh(&lock);
3021
3022         sha_init(digest);
3023         memset(&data, 0, sizeof(data));
3024         memset(workspace, 0, sizeof(workspace));
3025         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3026         data.prefix[0] = address->s6_addr32[0];
3027         data.prefix[1] = address->s6_addr32[1];
3028         data.secret = secret;
3029         data.dad_count = dad_count;
3030
3031         sha_transform(digest, data.__data, workspace);
3032
3033         temp = *address;
3034         temp.s6_addr32[2] = (__force __be32)digest[0];
3035         temp.s6_addr32[3] = (__force __be32)digest[1];
3036
3037         spin_unlock_bh(&lock);
3038
3039         if (ipv6_reserved_interfaceid(temp)) {
3040                 dad_count++;
3041                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3042                         return -1;
3043                 goto retry;
3044         }
3045
3046         *address = temp;
3047         return 0;
3048 }
3049
3050 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3051 {
3052         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3053
3054         if (s->initialized)
3055                 return;
3056         s = &idev->cnf.stable_secret;
3057         get_random_bytes(&s->secret, sizeof(s->secret));
3058         s->initialized = true;
3059 }
3060
3061 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3062 {
3063         struct in6_addr addr;
3064
3065         /* no link local addresses on L3 master devices */
3066         if (netif_is_l3_master(idev->dev))
3067                 return;
3068
3069         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3070
3071         switch (idev->addr_gen_mode) {
3072         case IN6_ADDR_GEN_MODE_RANDOM:
3073                 ipv6_gen_mode_random_init(idev);
3074                 /* fallthrough */
3075         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3076                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3077                         addrconf_add_linklocal(idev, &addr,
3078                                                IFA_F_STABLE_PRIVACY);
3079                 else if (prefix_route)
3080                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3081                 break;
3082         case IN6_ADDR_GEN_MODE_EUI64:
3083                 /* addrconf_add_linklocal also adds a prefix_route and we
3084                  * only need to care about prefix routes if ipv6_generate_eui64
3085                  * couldn't generate one.
3086                  */
3087                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3088                         addrconf_add_linklocal(idev, &addr, 0);
3089                 else if (prefix_route)
3090                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3091                 break;
3092         case IN6_ADDR_GEN_MODE_NONE:
3093         default:
3094                 /* will not add any link local address */
3095                 break;
3096         }
3097 }
3098
3099 static void addrconf_dev_config(struct net_device *dev)
3100 {
3101         struct inet6_dev *idev;
3102
3103         ASSERT_RTNL();
3104
3105         if ((dev->type != ARPHRD_ETHER) &&
3106             (dev->type != ARPHRD_FDDI) &&
3107             (dev->type != ARPHRD_ARCNET) &&
3108             (dev->type != ARPHRD_INFINIBAND) &&
3109             (dev->type != ARPHRD_IEEE1394) &&
3110             (dev->type != ARPHRD_TUNNEL6) &&
3111             (dev->type != ARPHRD_6LOWPAN) &&
3112             (dev->type != ARPHRD_NONE)) {
3113                 /* Alas, we support only Ethernet autoconfiguration. */
3114                 return;
3115         }
3116
3117         idev = addrconf_add_dev(dev);
3118         if (IS_ERR(idev))
3119                 return;
3120
3121         /* this device type has no EUI support */
3122         if (dev->type == ARPHRD_NONE &&
3123             idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3124                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3125
3126         addrconf_addr_gen(idev, false);
3127 }
3128
3129 #if IS_ENABLED(CONFIG_IPV6_SIT)
3130 static void addrconf_sit_config(struct net_device *dev)
3131 {
3132         struct inet6_dev *idev;
3133
3134         ASSERT_RTNL();
3135
3136         /*
3137          * Configure the tunnel with one of our IPv4
3138          * addresses... we should configure all of
3139          * our v4 addrs in the tunnel
3140          */
3141
3142         idev = ipv6_find_idev(dev);
3143         if (!idev) {
3144                 pr_debug("%s: add_dev failed\n", __func__);
3145                 return;
3146         }
3147
3148         if (dev->priv_flags & IFF_ISATAP) {
3149                 addrconf_addr_gen(idev, false);
3150                 return;
3151         }
3152
3153         sit_add_v4_addrs(idev);
3154
3155         if (dev->flags&IFF_POINTOPOINT)
3156                 addrconf_add_mroute(dev);
3157 }
3158 #endif
3159
3160 #if IS_ENABLED(CONFIG_NET_IPGRE)
3161 static void addrconf_gre_config(struct net_device *dev)
3162 {
3163         struct inet6_dev *idev;
3164
3165         ASSERT_RTNL();
3166
3167         idev = ipv6_find_idev(dev);
3168         if (!idev) {
3169                 pr_debug("%s: add_dev failed\n", __func__);
3170                 return;
3171         }
3172
3173         addrconf_addr_gen(idev, true);
3174         if (dev->flags & IFF_POINTOPOINT)
3175                 addrconf_add_mroute(dev);
3176 }
3177 #endif
3178
3179 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
3180 /* If the host route is cached on the addr struct make sure it is associated
3181  * with the proper table. e.g., enslavement can change and if so the cached
3182  * host route needs to move to the new table.
3183  */
3184 static void l3mdev_check_host_rt(struct inet6_dev *idev,
3185                                   struct inet6_ifaddr *ifp)
3186 {
3187         if (ifp->rt) {
3188                 u32 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
3189
3190                 if (tb_id != ifp->rt->rt6i_table->tb6_id) {
3191                         ip6_del_rt(ifp->rt);
3192                         ifp->rt = NULL;
3193                 }
3194         }
3195 }
3196 #else
3197 static void l3mdev_check_host_rt(struct inet6_dev *idev,
3198                                   struct inet6_ifaddr *ifp)
3199 {
3200 }
3201 #endif
3202
3203 static int fixup_permanent_addr(struct inet6_dev *idev,
3204                                 struct inet6_ifaddr *ifp)
3205 {
3206         l3mdev_check_host_rt(idev, ifp);
3207
3208         if (!ifp->rt) {
3209                 struct rt6_info *rt;
3210
3211                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3212                 if (unlikely(IS_ERR(rt)))
3213                         return PTR_ERR(rt);
3214
3215                 ifp->rt = rt;
3216         }
3217
3218         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3219                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3220                                       idev->dev, 0, 0);
3221         }
3222
3223         addrconf_dad_start(ifp);
3224
3225         return 0;
3226 }
3227
3228 static void addrconf_permanent_addr(struct net_device *dev)
3229 {
3230         struct inet6_ifaddr *ifp, *tmp;
3231         struct inet6_dev *idev;
3232
3233         idev = __in6_dev_get(dev);
3234         if (!idev)
3235                 return;
3236
3237         write_lock_bh(&idev->lock);
3238
3239         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3240                 if ((ifp->flags & IFA_F_PERMANENT) &&
3241                     fixup_permanent_addr(idev, ifp) < 0) {
3242                         write_unlock_bh(&idev->lock);
3243                         ipv6_del_addr(ifp);
3244                         write_lock_bh(&idev->lock);
3245
3246                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3247                                              idev->dev->name, &ifp->addr);
3248                 }
3249         }
3250
3251         write_unlock_bh(&idev->lock);
3252 }
3253
3254 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3255                            void *ptr)
3256 {
3257         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3258         struct netdev_notifier_changeupper_info *info;
3259         struct inet6_dev *idev = __in6_dev_get(dev);
3260         int run_pending = 0;
3261         int err;
3262
3263         switch (event) {
3264         case NETDEV_REGISTER:
3265                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3266                         idev = ipv6_add_dev(dev);
3267                         if (IS_ERR(idev))
3268                                 return notifier_from_errno(PTR_ERR(idev));
3269                 }
3270                 break;
3271
3272         case NETDEV_CHANGEMTU:
3273                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3274                 if (dev->mtu < IPV6_MIN_MTU) {
3275                         addrconf_ifdown(dev, 1);
3276                         break;
3277                 }
3278
3279                 if (idev) {
3280                         rt6_mtu_change(dev, dev->mtu);
3281                         idev->cnf.mtu6 = dev->mtu;
3282                         break;
3283                 }
3284
3285                 /* allocate new idev */
3286                 idev = ipv6_add_dev(dev);
3287                 if (IS_ERR(idev))
3288                         break;
3289
3290                 /* device is still not ready */
3291                 if (!(idev->if_flags & IF_READY))
3292                         break;
3293
3294                 run_pending = 1;
3295
3296                 /* fall through */
3297
3298         case NETDEV_UP:
3299         case NETDEV_CHANGE:
3300                 if (dev->flags & IFF_SLAVE)
3301                         break;
3302
3303                 if (idev && idev->cnf.disable_ipv6)
3304                         break;
3305
3306                 if (event == NETDEV_UP) {
3307                         if (!addrconf_qdisc_ok(dev)) {
3308                                 /* device is not ready yet. */
3309                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3310                                         dev->name);
3311                                 break;
3312                         }
3313
3314                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3315                                 idev = ipv6_add_dev(dev);
3316
3317                         if (!IS_ERR_OR_NULL(idev)) {
3318                                 idev->if_flags |= IF_READY;
3319                                 run_pending = 1;
3320                         }
3321                 } else if (event == NETDEV_CHANGE) {
3322                         if (!addrconf_qdisc_ok(dev)) {
3323                                 /* device is still not ready. */
3324                                 break;
3325                         }
3326
3327                         if (idev) {
3328                                 if (idev->if_flags & IF_READY)
3329                                         /* device is already configured. */
3330                                         break;
3331                                 idev->if_flags |= IF_READY;
3332                         }
3333
3334                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3335                                 dev->name);
3336
3337                         run_pending = 1;
3338                 }
3339
3340                 /* restore routes for permanent addresses */
3341                 addrconf_permanent_addr(dev);
3342
3343                 switch (dev->type) {
3344 #if IS_ENABLED(CONFIG_IPV6_SIT)
3345                 case ARPHRD_SIT:
3346                         addrconf_sit_config(dev);
3347                         break;
3348 #endif
3349 #if IS_ENABLED(CONFIG_NET_IPGRE)
3350                 case ARPHRD_IPGRE:
3351                         addrconf_gre_config(dev);
3352                         break;
3353 #endif
3354                 case ARPHRD_LOOPBACK:
3355                         init_loopback(dev);
3356                         break;
3357
3358                 default:
3359                         addrconf_dev_config(dev);
3360                         break;
3361                 }
3362
3363                 if (!IS_ERR_OR_NULL(idev)) {
3364                         if (run_pending)
3365                                 addrconf_dad_run(idev);
3366
3367                         /*
3368                          * If the MTU changed during the interface down,
3369                          * when the interface up, the changed MTU must be
3370                          * reflected in the idev as well as routers.
3371                          */
3372                         if (idev->cnf.mtu6 != dev->mtu &&
3373                             dev->mtu >= IPV6_MIN_MTU) {
3374                                 rt6_mtu_change(dev, dev->mtu);
3375                                 idev->cnf.mtu6 = dev->mtu;
3376                         }
3377                         idev->tstamp = jiffies;
3378                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3379
3380                         /*
3381                          * If the changed mtu during down is lower than
3382                          * IPV6_MIN_MTU stop IPv6 on this interface.
3383                          */
3384                         if (dev->mtu < IPV6_MIN_MTU)
3385                                 addrconf_ifdown(dev, 1);
3386                 }
3387                 break;
3388
3389         case NETDEV_DOWN:
3390         case NETDEV_UNREGISTER:
3391                 /*
3392                  *      Remove all addresses from this interface.
3393                  */
3394                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3395                 break;
3396
3397         case NETDEV_CHANGENAME:
3398                 if (idev) {
3399                         snmp6_unregister_dev(idev);
3400                         addrconf_sysctl_unregister(idev);
3401                         err = addrconf_sysctl_register(idev);
3402                         if (err)
3403                                 return notifier_from_errno(err);
3404                         err = snmp6_register_dev(idev);
3405                         if (err) {
3406                                 addrconf_sysctl_unregister(idev);
3407                                 return notifier_from_errno(err);
3408                         }
3409                 }
3410                 break;
3411
3412         case NETDEV_PRE_TYPE_CHANGE:
3413         case NETDEV_POST_TYPE_CHANGE:
3414                 if (idev)
3415                         addrconf_type_change(dev, event);
3416                 break;
3417
3418         case NETDEV_CHANGEUPPER:
3419                 info = ptr;
3420
3421                 /* flush all routes if dev is linked to or unlinked from
3422                  * an L3 master device (e.g., VRF)
3423                  */
3424                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3425                         addrconf_ifdown(dev, 0);
3426         }
3427
3428         return NOTIFY_OK;
3429 }
3430
3431 /*
3432  *      addrconf module should be notified of a device going up
3433  */
3434 static struct notifier_block ipv6_dev_notf = {
3435         .notifier_call = addrconf_notify,
3436 };
3437
3438 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3439 {
3440         struct inet6_dev *idev;
3441         ASSERT_RTNL();
3442
3443         idev = __in6_dev_get(dev);
3444
3445         if (event == NETDEV_POST_TYPE_CHANGE)
3446                 ipv6_mc_remap(idev);
3447         else if (event == NETDEV_PRE_TYPE_CHANGE)
3448                 ipv6_mc_unmap(idev);
3449 }
3450
3451 static bool addr_is_local(const struct in6_addr *addr)
3452 {
3453         return ipv6_addr_type(addr) &
3454                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3455 }
3456
3457 static int addrconf_ifdown(struct net_device *dev, int how)
3458 {
3459         struct net *net = dev_net(dev);
3460         struct inet6_dev *idev;
3461         struct inet6_ifaddr *ifa, *tmp;
3462         struct list_head del_list;
3463         int _keep_addr;
3464         bool keep_addr;
3465         int state, i;
3466
3467         ASSERT_RTNL();
3468
3469         rt6_ifdown(net, dev);
3470         neigh_ifdown(&nd_tbl, dev);
3471
3472         idev = __in6_dev_get(dev);
3473         if (!idev)
3474                 return -ENODEV;
3475
3476         /*
3477          * Step 1: remove reference to ipv6 device from parent device.
3478          *         Do not dev_put!
3479          */
3480         if (how) {
3481                 idev->dead = 1;
3482
3483                 /* protected by rtnl_lock */
3484                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3485
3486                 /* Step 1.5: remove snmp6 entry */
3487                 snmp6_unregister_dev(idev);
3488
3489         }
3490
3491         /* aggregate the system setting and interface setting */
3492         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3493         if (!_keep_addr)
3494                 _keep_addr = idev->cnf.keep_addr_on_down;
3495
3496         /* combine the user config with event to determine if permanent
3497          * addresses are to be removed from address hash table
3498          */
3499         keep_addr = !(how || _keep_addr <= 0);
3500
3501         /* Step 2: clear hash table */
3502         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3503                 struct hlist_head *h = &inet6_addr_lst[i];
3504
3505                 spin_lock_bh(&addrconf_hash_lock);
3506 restart:
3507                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3508                         if (ifa->idev == idev) {
3509                                 addrconf_del_dad_work(ifa);
3510                                 /* combined flag + permanent flag decide if
3511                                  * address is retained on a down event
3512                                  */
3513                                 if (!keep_addr ||
3514                                     !(ifa->flags & IFA_F_PERMANENT) ||
3515                                     addr_is_local(&ifa->addr)) {
3516                                         hlist_del_init_rcu(&ifa->addr_lst);
3517                                         goto restart;
3518                                 }
3519                         }
3520                 }
3521                 spin_unlock_bh(&addrconf_hash_lock);
3522         }
3523
3524         write_lock_bh(&idev->lock);
3525
3526         addrconf_del_rs_timer(idev);
3527
3528         /* Step 2: clear flags for stateless addrconf */
3529         if (!how)
3530                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3531
3532         if (how && del_timer(&idev->regen_timer))
3533                 in6_dev_put(idev);
3534
3535         /* Step 3: clear tempaddr list */
3536         while (!list_empty(&idev->tempaddr_list)) {
3537                 ifa = list_first_entry(&idev->tempaddr_list,
3538                                        struct inet6_ifaddr, tmp_list);
3539                 list_del(&ifa->tmp_list);
3540                 write_unlock_bh(&idev->lock);
3541                 spin_lock_bh(&ifa->lock);
3542
3543                 if (ifa->ifpub) {
3544                         in6_ifa_put(ifa->ifpub);
3545                         ifa->ifpub = NULL;
3546                 }
3547                 spin_unlock_bh(&ifa->lock);
3548                 in6_ifa_put(ifa);
3549                 write_lock_bh(&idev->lock);
3550         }
3551
3552         /* re-combine the user config with event to determine if permanent
3553          * addresses are to be removed from the interface list
3554          */
3555         keep_addr = (!how && _keep_addr > 0);
3556
3557         INIT_LIST_HEAD(&del_list);
3558         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3559                 addrconf_del_dad_work(ifa);
3560
3561                 write_unlock_bh(&idev->lock);
3562                 spin_lock_bh(&ifa->lock);
3563
3564                 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3565                     !addr_is_local(&ifa->addr)) {
3566                         /* set state to skip the notifier below */
3567                         state = INET6_IFADDR_STATE_DEAD;
3568                         ifa->state = 0;
3569                         if (!(ifa->flags & IFA_F_NODAD))
3570                                 ifa->flags |= IFA_F_TENTATIVE;
3571                 } else {
3572                         state = ifa->state;
3573                         ifa->state = INET6_IFADDR_STATE_DEAD;
3574
3575                         list_del(&ifa->if_list);
3576                         list_add(&ifa->if_list, &del_list);
3577                 }
3578
3579                 spin_unlock_bh(&ifa->lock);
3580
3581                 if (state != INET6_IFADDR_STATE_DEAD) {
3582                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3583                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3584                 }
3585
3586                 write_lock_bh(&idev->lock);
3587         }
3588
3589         write_unlock_bh(&idev->lock);
3590
3591         /* now clean up addresses to be removed */
3592         while (!list_empty(&del_list)) {
3593                 ifa = list_first_entry(&del_list,
3594                                        struct inet6_ifaddr, if_list);
3595                 list_del(&ifa->if_list);
3596
3597                 in6_ifa_put(ifa);
3598         }
3599
3600         /* Step 5: Discard anycast and multicast list */
3601         if (how) {
3602                 ipv6_ac_destroy_dev(idev);
3603                 ipv6_mc_destroy_dev(idev);
3604         } else {
3605                 ipv6_mc_down(idev);
3606         }
3607
3608         idev->tstamp = jiffies;
3609
3610         /* Last: Shot the device (if unregistered) */
3611         if (how) {
3612                 addrconf_sysctl_unregister(idev);
3613                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3614                 neigh_ifdown(&nd_tbl, dev);
3615                 in6_dev_put(idev);
3616         }
3617         return 0;
3618 }
3619
3620 static void addrconf_rs_timer(unsigned long data)
3621 {
3622         struct inet6_dev *idev = (struct inet6_dev *)data;
3623         struct net_device *dev = idev->dev;
3624         struct in6_addr lladdr;
3625
3626         write_lock(&idev->lock);
3627         if (idev->dead || !(idev->if_flags & IF_READY))
3628                 goto out;
3629
3630         if (!ipv6_accept_ra(idev))
3631                 goto out;
3632
3633         /* Announcement received after solicitation was sent */
3634         if (idev->if_flags & IF_RA_RCVD)
3635                 goto out;
3636
3637         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3638                 write_unlock(&idev->lock);
3639                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3640                         ndisc_send_rs(dev, &lladdr,
3641                                       &in6addr_linklocal_allrouters);
3642                 else
3643                         goto put;
3644
3645                 write_lock(&idev->lock);
3646                 /* The wait after the last probe can be shorter */
3647                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3648                                              idev->cnf.rtr_solicits) ?
3649                                       idev->cnf.rtr_solicit_delay :
3650                                       idev->cnf.rtr_solicit_interval);
3651         } else {
3652                 /*
3653                  * Note: we do not support deprecated "all on-link"
3654                  * assumption any longer.
3655                  */
3656                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3657         }
3658
3659 out:
3660         write_unlock(&idev->lock);
3661 put:
3662         in6_dev_put(idev);
3663 }
3664
3665 /*
3666  *      Duplicate Address Detection
3667  */
3668 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3669 {
3670         unsigned long rand_num;
3671         struct inet6_dev *idev = ifp->idev;
3672
3673         if (ifp->flags & IFA_F_OPTIMISTIC)
3674                 rand_num = 0;
3675         else
3676                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3677
3678         ifp->dad_probes = idev->cnf.dad_transmits;
3679         addrconf_mod_dad_work(ifp, rand_num);
3680 }
3681
3682 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3683 {
3684         struct inet6_dev *idev = ifp->idev;
3685         struct net_device *dev = idev->dev;
3686         bool notify = false;
3687
3688         addrconf_join_solict(dev, &ifp->addr);
3689
3690         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3691
3692         read_lock_bh(&idev->lock);
3693         spin_lock(&ifp->lock);
3694         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3695                 goto out;
3696
3697         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3698             idev->cnf.accept_dad < 1 ||
3699             !(ifp->flags&IFA_F_TENTATIVE) ||
3700             ifp->flags & IFA_F_NODAD) {
3701                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3702                 spin_unlock(&ifp->lock);
3703                 read_unlock_bh(&idev->lock);
3704
3705                 addrconf_dad_completed(ifp);
3706                 return;
3707         }
3708
3709         if (!(idev->if_flags & IF_READY)) {
3710                 spin_unlock(&ifp->lock);
3711                 read_unlock_bh(&idev->lock);
3712                 /*
3713                  * If the device is not ready:
3714                  * - keep it tentative if it is a permanent address.
3715                  * - otherwise, kill it.
3716                  */
3717                 in6_ifa_hold(ifp);
3718                 addrconf_dad_stop(ifp, 0);
3719                 return;
3720         }
3721
3722         /*
3723          * Optimistic nodes can start receiving
3724          * Frames right away
3725          */
3726         if (ifp->flags & IFA_F_OPTIMISTIC) {
3727                 ip6_ins_rt(ifp->rt);
3728                 if (ipv6_use_optimistic_addr(idev)) {
3729                         /* Because optimistic nodes can use this address,
3730                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3731                          */
3732                         notify = true;
3733                 }
3734         }
3735
3736         addrconf_dad_kick(ifp);
3737 out:
3738         spin_unlock(&ifp->lock);
3739         read_unlock_bh(&idev->lock);
3740         if (notify)
3741                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3742 }
3743
3744 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3745 {
3746         bool begin_dad = false;
3747
3748         spin_lock_bh(&ifp->lock);
3749         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3750                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3751                 begin_dad = true;
3752         }
3753         spin_unlock_bh(&ifp->lock);
3754
3755         if (begin_dad)
3756                 addrconf_mod_dad_work(ifp, 0);
3757 }
3758
3759 static void addrconf_dad_work(struct work_struct *w)
3760 {
3761         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3762                                                 struct inet6_ifaddr,
3763                                                 dad_work);
3764         struct inet6_dev *idev = ifp->idev;
3765         struct in6_addr mcaddr;
3766
3767         enum {
3768                 DAD_PROCESS,
3769                 DAD_BEGIN,
3770                 DAD_ABORT,
3771         } action = DAD_PROCESS;
3772
3773         rtnl_lock();
3774
3775         spin_lock_bh(&ifp->lock);
3776         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3777                 action = DAD_BEGIN;
3778                 ifp->state = INET6_IFADDR_STATE_DAD;
3779         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3780                 action = DAD_ABORT;
3781                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3782         }
3783         spin_unlock_bh(&ifp->lock);
3784
3785         if (action == DAD_BEGIN) {
3786                 addrconf_dad_begin(ifp);
3787                 goto out;
3788         } else if (action == DAD_ABORT) {
3789                 addrconf_dad_stop(ifp, 1);
3790                 goto out;
3791         }
3792
3793         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3794                 goto out;
3795
3796         write_lock_bh(&idev->lock);
3797         if (idev->dead || !(idev->if_flags & IF_READY)) {
3798                 write_unlock_bh(&idev->lock);
3799                 goto out;
3800         }
3801
3802         spin_lock(&ifp->lock);
3803         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3804                 spin_unlock(&ifp->lock);
3805                 write_unlock_bh(&idev->lock);
3806                 goto out;
3807         }
3808
3809         if (ifp->dad_probes == 0) {
3810                 /*
3811                  * DAD was successful
3812                  */
3813
3814                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3815                 spin_unlock(&ifp->lock);
3816                 write_unlock_bh(&idev->lock);
3817
3818                 addrconf_dad_completed(ifp);
3819
3820                 goto out;
3821         }
3822
3823         ifp->dad_probes--;
3824         addrconf_mod_dad_work(ifp,
3825                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3826         spin_unlock(&ifp->lock);
3827         write_unlock_bh(&idev->lock);
3828
3829         /* send a neighbour solicitation for our addr */
3830         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3831         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3832 out:
3833         in6_ifa_put(ifp);
3834         rtnl_unlock();
3835 }
3836
3837 /* ifp->idev must be at least read locked */
3838 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3839 {
3840         struct inet6_ifaddr *ifpiter;
3841         struct inet6_dev *idev = ifp->idev;
3842
3843         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3844                 if (ifpiter->scope > IFA_LINK)
3845                         break;
3846                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3847                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3848                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3849                     IFA_F_PERMANENT)
3850                         return false;
3851         }
3852         return true;
3853 }
3854
3855 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3856 {
3857         struct net_device *dev = ifp->idev->dev;
3858         struct in6_addr lladdr;
3859         bool send_rs, send_mld;
3860
3861         addrconf_del_dad_work(ifp);
3862
3863         /*
3864          *      Configure the address for reception. Now it is valid.
3865          */
3866
3867         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3868
3869         /* If added prefix is link local and we are prepared to process
3870            router advertisements, start sending router solicitations.
3871          */
3872
3873         read_lock_bh(&ifp->idev->lock);
3874         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3875         send_rs = send_mld &&
3876                   ipv6_accept_ra(ifp->idev) &&
3877                   ifp->idev->cnf.rtr_solicits > 0 &&
3878                   (dev->flags&IFF_LOOPBACK) == 0;
3879         read_unlock_bh(&ifp->idev->lock);
3880
3881         /* While dad is in progress mld report's source address is in6_addrany.
3882          * Resend with proper ll now.
3883          */
3884         if (send_mld)
3885                 ipv6_mc_dad_complete(ifp->idev);
3886
3887         if (send_rs) {
3888                 /*
3889                  *      If a host as already performed a random delay
3890                  *      [...] as part of DAD [...] there is no need
3891                  *      to delay again before sending the first RS
3892                  */
3893                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3894                         return;
3895                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3896
3897                 write_lock_bh(&ifp->idev->lock);
3898                 spin_lock(&ifp->lock);
3899                 ifp->idev->rs_probes = 1;
3900                 ifp->idev->if_flags |= IF_RS_SENT;
3901                 addrconf_mod_rs_timer(ifp->idev,
3902                                       ifp->idev->cnf.rtr_solicit_interval);
3903                 spin_unlock(&ifp->lock);
3904                 write_unlock_bh(&ifp->idev->lock);
3905         }
3906 }
3907
3908 static void addrconf_dad_run(struct inet6_dev *idev)
3909 {
3910         struct inet6_ifaddr *ifp;
3911
3912         read_lock_bh(&idev->lock);
3913         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3914                 spin_lock(&ifp->lock);
3915                 if (ifp->flags & IFA_F_TENTATIVE &&
3916                     ifp->state == INET6_IFADDR_STATE_DAD)
3917                         addrconf_dad_kick(ifp);
3918                 spin_unlock(&ifp->lock);
3919         }
3920         read_unlock_bh(&idev->lock);
3921 }
3922
3923 #ifdef CONFIG_PROC_FS
3924 struct if6_iter_state {
3925         struct seq_net_private p;
3926         int bucket;
3927         int offset;
3928 };
3929
3930 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3931 {
3932         struct inet6_ifaddr *ifa = NULL;
3933         struct if6_iter_state *state = seq->private;
3934         struct net *net = seq_file_net(seq);
3935         int p = 0;
3936
3937         /* initial bucket if pos is 0 */
3938         if (pos == 0) {
3939                 state->bucket = 0;
3940                 state->offset = 0;
3941         }
3942
3943         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3944                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3945                                          addr_lst) {
3946                         if (!net_eq(dev_net(ifa->idev->dev), net))
3947                                 continue;
3948                         /* sync with offset */
3949                         if (p < state->offset) {
3950                                 p++;
3951                                 continue;
3952                         }
3953                         state->offset++;
3954                         return ifa;
3955                 }
3956
3957                 /* prepare for next bucket */
3958                 state->offset = 0;
3959                 p = 0;
3960         }
3961         return NULL;
3962 }
3963
3964 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3965                                          struct inet6_ifaddr *ifa)
3966 {
3967         struct if6_iter_state *state = seq->private;
3968         struct net *net = seq_file_net(seq);
3969
3970         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3971                 if (!net_eq(dev_net(ifa->idev->dev), net))
3972                         continue;
3973                 state->offset++;
3974                 return ifa;
3975         }
3976
3977         while (++state->bucket < IN6_ADDR_HSIZE) {
3978                 state->offset = 0;
3979                 hlist_for_each_entry_rcu_bh(ifa,
3980                                      &inet6_addr_lst[state->bucket], addr_lst) {
3981                         if (!net_eq(dev_net(ifa->idev->dev), net))
3982                                 continue;
3983                         state->offset++;
3984                         return ifa;
3985                 }
3986         }
3987
3988         return NULL;
3989 }
3990
3991 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3992         __acquires(rcu_bh)
3993 {
3994         rcu_read_lock_bh();
3995         return if6_get_first(seq, *pos);
3996 }
3997
3998 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3999 {
4000         struct inet6_ifaddr *ifa;
4001
4002         ifa = if6_get_next(seq, v);
4003         ++*pos;
4004         return ifa;
4005 }
4006
4007 static void if6_seq_stop(struct seq_file *seq, void *v)
4008         __releases(rcu_bh)
4009 {
4010         rcu_read_unlock_bh();
4011 }
4012
4013 static int if6_seq_show(struct seq_file *seq, void *v)
4014 {
4015         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4016         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4017                    &ifp->addr,
4018                    ifp->idev->dev->ifindex,
4019                    ifp->prefix_len,
4020                    ifp->scope,
4021                    (u8) ifp->flags,
4022                    ifp->idev->dev->name);
4023         return 0;
4024 }
4025
4026 static const struct seq_operations if6_seq_ops = {
4027         .start  = if6_seq_start,
4028         .next   = if6_seq_next,
4029         .show   = if6_seq_show,
4030         .stop   = if6_seq_stop,
4031 };
4032
4033 static int if6_seq_open(struct inode *inode, struct file *file)
4034 {
4035         return seq_open_net(inode, file, &if6_seq_ops,
4036                             sizeof(struct if6_iter_state));
4037 }
4038
4039 static const struct file_operations if6_fops = {
4040         .owner          = THIS_MODULE,
4041         .open           = if6_seq_open,
4042         .read           = seq_read,
4043         .llseek         = seq_lseek,
4044         .release        = seq_release_net,
4045 };
4046
4047 static int __net_init if6_proc_net_init(struct net *net)
4048 {
4049         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4050                 return -ENOMEM;
4051         return 0;
4052 }
4053
4054 static void __net_exit if6_proc_net_exit(struct net *net)
4055 {
4056         remove_proc_entry("if_inet6", net->proc_net);
4057 }
4058
4059 static struct pernet_operations if6_proc_net_ops = {
4060         .init = if6_proc_net_init,
4061         .exit = if6_proc_net_exit,
4062 };
4063
4064 int __init if6_proc_init(void)
4065 {
4066         return register_pernet_subsys(&if6_proc_net_ops);
4067 }
4068
4069 void if6_proc_exit(void)
4070 {
4071         unregister_pernet_subsys(&if6_proc_net_ops);
4072 }
4073 #endif  /* CONFIG_PROC_FS */
4074
4075 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4076 /* Check if address is a home address configured on any interface. */
4077 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4078 {
4079         int ret = 0;
4080         struct inet6_ifaddr *ifp = NULL;
4081         unsigned int hash = inet6_addr_hash(addr);
4082
4083         rcu_read_lock_bh();
4084         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4085                 if (!net_eq(dev_net(ifp->idev->dev), net))
4086                         continue;
4087                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4088                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4089                         ret = 1;
4090                         break;
4091                 }
4092         }
4093         rcu_read_unlock_bh();
4094         return ret;
4095 }
4096 #endif
4097
4098 /*
4099  *      Periodic address status verification
4100  */
4101
4102 static void addrconf_verify_rtnl(void)
4103 {
4104         unsigned long now, next, next_sec, next_sched;
4105         struct inet6_ifaddr *ifp;
4106         int i;
4107
4108         ASSERT_RTNL();
4109
4110         rcu_read_lock_bh();
4111         now = jiffies;
4112         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4113
4114         cancel_delayed_work(&addr_chk_work);
4115
4116         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4117 restart:
4118                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4119                         unsigned long age;
4120
4121                         /* When setting preferred_lft to a value not zero or
4122                          * infinity, while valid_lft is infinity
4123                          * IFA_F_PERMANENT has a non-infinity life time.
4124                          */
4125                         if ((ifp->flags & IFA_F_PERMANENT) &&
4126                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4127                                 continue;
4128
4129                         spin_lock(&ifp->lock);
4130                         /* We try to batch several events at once. */
4131                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4132
4133                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4134                             age >= ifp->valid_lft) {
4135                                 spin_unlock(&ifp->lock);
4136                                 in6_ifa_hold(ifp);
4137                                 ipv6_del_addr(ifp);
4138                                 goto restart;
4139                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4140                                 spin_unlock(&ifp->lock);
4141                                 continue;
4142                         } else if (age >= ifp->prefered_lft) {
4143                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4144                                 int deprecate = 0;
4145
4146                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4147                                         deprecate = 1;
4148                                         ifp->flags |= IFA_F_DEPRECATED;
4149                                 }
4150
4151                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4152                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4153                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4154
4155                                 spin_unlock(&ifp->lock);
4156
4157                                 if (deprecate) {
4158                                         in6_ifa_hold(ifp);
4159
4160                                         ipv6_ifa_notify(0, ifp);
4161                                         in6_ifa_put(ifp);
4162                                         goto restart;
4163                                 }
4164                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4165                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4166                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4167                                         ifp->idev->cnf.dad_transmits *
4168                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4169
4170                                 if (age >= ifp->prefered_lft - regen_advance) {
4171                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4172                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4173                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4174                                         if (!ifp->regen_count && ifpub) {
4175                                                 ifp->regen_count++;
4176                                                 in6_ifa_hold(ifp);
4177                                                 in6_ifa_hold(ifpub);
4178                                                 spin_unlock(&ifp->lock);
4179
4180                                                 spin_lock(&ifpub->lock);
4181                                                 ifpub->regen_count = 0;
4182                                                 spin_unlock(&ifpub->lock);
4183                                                 ipv6_create_tempaddr(ifpub, ifp);
4184                                                 in6_ifa_put(ifpub);
4185                                                 in6_ifa_put(ifp);
4186                                                 goto restart;
4187                                         }
4188                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4189                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4190                                 spin_unlock(&ifp->lock);
4191                         } else {
4192                                 /* ifp->prefered_lft <= ifp->valid_lft */
4193                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4194                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4195                                 spin_unlock(&ifp->lock);
4196                         }
4197                 }
4198         }
4199
4200         next_sec = round_jiffies_up(next);
4201         next_sched = next;
4202
4203         /* If rounded timeout is accurate enough, accept it. */
4204         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4205                 next_sched = next_sec;
4206
4207         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4208         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4209                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4210
4211         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4212               now, next, next_sec, next_sched);
4213         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4214         rcu_read_unlock_bh();
4215 }
4216
4217 static void addrconf_verify_work(struct work_struct *w)
4218 {
4219         rtnl_lock();
4220         addrconf_verify_rtnl();
4221         rtnl_unlock();
4222 }
4223
4224 static void addrconf_verify(void)
4225 {
4226         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4227 }
4228
4229 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4230                                      struct in6_addr **peer_pfx)
4231 {
4232         struct in6_addr *pfx = NULL;
4233
4234         *peer_pfx = NULL;
4235
4236         if (addr)
4237                 pfx = nla_data(addr);
4238
4239         if (local) {
4240                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4241                         *peer_pfx = pfx;
4242                 pfx = nla_data(local);
4243         }
4244
4245         return pfx;
4246 }
4247
4248 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4249         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4250         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4251         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4252         [IFA_FLAGS]             = { .len = sizeof(u32) },
4253 };
4254
4255 static int
4256 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4257 {
4258         struct net *net = sock_net(skb->sk);
4259         struct ifaddrmsg *ifm;
4260         struct nlattr *tb[IFA_MAX+1];
4261         struct in6_addr *pfx, *peer_pfx;
4262         u32 ifa_flags;
4263         int err;
4264
4265         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4266         if (err < 0)
4267                 return err;
4268
4269         ifm = nlmsg_data(nlh);
4270         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4271         if (!pfx)
4272                 return -EINVAL;
4273
4274         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4275
4276         /* We ignore other flags so far. */
4277         ifa_flags &= IFA_F_MANAGETEMPADDR;
4278
4279         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4280                               ifm->ifa_prefixlen);
4281 }
4282
4283 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4284                              u32 prefered_lft, u32 valid_lft)
4285 {
4286         u32 flags;
4287         clock_t expires;
4288         unsigned long timeout;
4289         bool was_managetempaddr;
4290         bool had_prefixroute;
4291
4292         ASSERT_RTNL();
4293
4294         if (!valid_lft || (prefered_lft > valid_lft))
4295                 return -EINVAL;
4296
4297         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4298             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4299                 return -EINVAL;
4300
4301         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4302         if (addrconf_finite_timeout(timeout)) {
4303                 expires = jiffies_to_clock_t(timeout * HZ);
4304                 valid_lft = timeout;
4305                 flags = RTF_EXPIRES;
4306         } else {
4307                 expires = 0;
4308                 flags = 0;
4309                 ifa_flags |= IFA_F_PERMANENT;
4310         }
4311
4312         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4313         if (addrconf_finite_timeout(timeout)) {
4314                 if (timeout == 0)
4315                         ifa_flags |= IFA_F_DEPRECATED;
4316                 prefered_lft = timeout;
4317         }
4318
4319         spin_lock_bh(&ifp->lock);
4320         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4321         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4322                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4323         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4324                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4325                         IFA_F_NOPREFIXROUTE);
4326         ifp->flags |= ifa_flags;
4327         ifp->tstamp = jiffies;
4328         ifp->valid_lft = valid_lft;
4329         ifp->prefered_lft = prefered_lft;
4330
4331         spin_unlock_bh(&ifp->lock);
4332         if (!(ifp->flags&IFA_F_TENTATIVE))
4333                 ipv6_ifa_notify(0, ifp);
4334
4335         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4336                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4337                                       expires, flags);
4338         } else if (had_prefixroute) {
4339                 enum cleanup_prefix_rt_t action;
4340                 unsigned long rt_expires;
4341
4342                 write_lock_bh(&ifp->idev->lock);
4343                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4344                 write_unlock_bh(&ifp->idev->lock);
4345
4346                 if (action != CLEANUP_PREFIX_RT_NOP) {
4347                         cleanup_prefix_route(ifp, rt_expires,
4348                                 action == CLEANUP_PREFIX_RT_DEL);
4349                 }
4350         }
4351
4352         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4353                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4354                         valid_lft = prefered_lft = 0;
4355                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4356                                  !was_managetempaddr, jiffies);
4357         }
4358
4359         addrconf_verify_rtnl();
4360
4361         return 0;
4362 }
4363
4364 static int
4365 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4366 {
4367         struct net *net = sock_net(skb->sk);
4368         struct ifaddrmsg *ifm;
4369         struct nlattr *tb[IFA_MAX+1];
4370         struct in6_addr *pfx, *peer_pfx;
4371         struct inet6_ifaddr *ifa;
4372         struct net_device *dev;
4373         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4374         u32 ifa_flags;
4375         int err;
4376
4377         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4378         if (err < 0)
4379                 return err;
4380
4381         ifm = nlmsg_data(nlh);
4382         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4383         if (!pfx)
4384                 return -EINVAL;
4385
4386         if (tb[IFA_CACHEINFO]) {
4387                 struct ifa_cacheinfo *ci;
4388
4389                 ci = nla_data(tb[IFA_CACHEINFO]);
4390                 valid_lft = ci->ifa_valid;
4391                 preferred_lft = ci->ifa_prefered;
4392         } else {
4393                 preferred_lft = INFINITY_LIFE_TIME;
4394                 valid_lft = INFINITY_LIFE_TIME;
4395         }
4396
4397         dev =  __dev_get_by_index(net, ifm->ifa_index);
4398         if (!dev)
4399                 return -ENODEV;
4400
4401         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4402
4403         /* We ignore other flags so far. */
4404         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4405                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4406
4407         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4408         if (!ifa) {
4409                 /*
4410                  * It would be best to check for !NLM_F_CREATE here but
4411                  * userspace already relies on not having to provide this.
4412                  */
4413                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4414                                       ifm->ifa_prefixlen, ifa_flags,
4415                                       preferred_lft, valid_lft);
4416         }
4417
4418         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4419             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4420                 err = -EEXIST;
4421         else
4422                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4423
4424         in6_ifa_put(ifa);
4425
4426         return err;
4427 }
4428
4429 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4430                           u8 scope, int ifindex)
4431 {
4432         struct ifaddrmsg *ifm;
4433
4434         ifm = nlmsg_data(nlh);
4435         ifm->ifa_family = AF_INET6;
4436         ifm->ifa_prefixlen = prefixlen;
4437         ifm->ifa_flags = flags;
4438         ifm->ifa_scope = scope;
4439         ifm->ifa_index = ifindex;
4440 }
4441
4442 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4443                          unsigned long tstamp, u32 preferred, u32 valid)
4444 {
4445         struct ifa_cacheinfo ci;
4446
4447         ci.cstamp = cstamp_delta(cstamp);
4448         ci.tstamp = cstamp_delta(tstamp);
4449         ci.ifa_prefered = preferred;
4450         ci.ifa_valid = valid;
4451
4452         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4453 }
4454
4455 static inline int rt_scope(int ifa_scope)
4456 {
4457         if (ifa_scope & IFA_HOST)
4458                 return RT_SCOPE_HOST;
4459         else if (ifa_scope & IFA_LINK)
4460                 return RT_SCOPE_LINK;
4461         else if (ifa_scope & IFA_SITE)
4462                 return RT_SCOPE_SITE;
4463         else
4464                 return RT_SCOPE_UNIVERSE;
4465 }
4466
4467 static inline int inet6_ifaddr_msgsize(void)
4468 {
4469         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4470                + nla_total_size(16) /* IFA_LOCAL */
4471                + nla_total_size(16) /* IFA_ADDRESS */
4472                + nla_total_size(sizeof(struct ifa_cacheinfo))
4473                + nla_total_size(4)  /* IFA_FLAGS */;
4474 }
4475
4476 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4477                              u32 portid, u32 seq, int event, unsigned int flags)
4478 {
4479         struct nlmsghdr  *nlh;
4480         u32 preferred, valid;
4481
4482         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4483         if (!nlh)
4484                 return -EMSGSIZE;
4485
4486         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4487                       ifa->idev->dev->ifindex);
4488
4489         if (!((ifa->flags&IFA_F_PERMANENT) &&
4490               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4491                 preferred = ifa->prefered_lft;
4492                 valid = ifa->valid_lft;
4493                 if (preferred != INFINITY_LIFE_TIME) {
4494                         long tval = (jiffies - ifa->tstamp)/HZ;
4495                         if (preferred > tval)
4496                                 preferred -= tval;
4497                         else
4498                                 preferred = 0;
4499                         if (valid != INFINITY_LIFE_TIME) {
4500                                 if (valid > tval)
4501                                         valid -= tval;
4502                                 else
4503                                         valid = 0;
4504                         }
4505                 }
4506         } else {
4507                 preferred = INFINITY_LIFE_TIME;
4508                 valid = INFINITY_LIFE_TIME;
4509         }
4510
4511         if (!ipv6_addr_any(&ifa->peer_addr)) {
4512                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4513                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4514                         goto error;
4515         } else
4516                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4517                         goto error;
4518
4519         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4520                 goto error;
4521
4522         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4523                 goto error;
4524
4525         nlmsg_end(skb, nlh);
4526         return 0;
4527
4528 error:
4529         nlmsg_cancel(skb, nlh);
4530         return -EMSGSIZE;
4531 }
4532
4533 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4534                                 u32 portid, u32 seq, int event, u16 flags)
4535 {
4536         struct nlmsghdr  *nlh;
4537         u8 scope = RT_SCOPE_UNIVERSE;
4538         int ifindex = ifmca->idev->dev->ifindex;
4539
4540         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4541                 scope = RT_SCOPE_SITE;
4542
4543         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4544         if (!nlh)
4545                 return -EMSGSIZE;
4546
4547         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4548         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4549             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4550                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4551                 nlmsg_cancel(skb, nlh);
4552                 return -EMSGSIZE;
4553         }
4554
4555         nlmsg_end(skb, nlh);
4556         return 0;
4557 }
4558
4559 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4560                                 u32 portid, u32 seq, int event, unsigned int flags)
4561 {
4562         struct nlmsghdr  *nlh;
4563         u8 scope = RT_SCOPE_UNIVERSE;
4564         int ifindex = ifaca->aca_idev->dev->ifindex;
4565
4566         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4567                 scope = RT_SCOPE_SITE;
4568
4569         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4570         if (!nlh)
4571                 return -EMSGSIZE;
4572
4573         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4574         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4575             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4576                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4577                 nlmsg_cancel(skb, nlh);
4578                 return -EMSGSIZE;
4579         }
4580
4581         nlmsg_end(skb, nlh);
4582         return 0;
4583 }
4584
4585 enum addr_type_t {
4586         UNICAST_ADDR,
4587         MULTICAST_ADDR,
4588         ANYCAST_ADDR,
4589 };
4590
4591 /* called with rcu_read_lock() */
4592 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4593                           struct netlink_callback *cb, enum addr_type_t type,
4594                           int s_ip_idx, int *p_ip_idx)
4595 {
4596         struct ifmcaddr6 *ifmca;
4597         struct ifacaddr6 *ifaca;
4598         int err = 1;
4599         int ip_idx = *p_ip_idx;
4600
4601         read_lock_bh(&idev->lock);
4602         switch (type) {
4603         case UNICAST_ADDR: {
4604                 struct inet6_ifaddr *ifa;
4605
4606                 /* unicast address incl. temp addr */
4607                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4608                         if (++ip_idx < s_ip_idx)
4609                                 continue;
4610                         err = inet6_fill_ifaddr(skb, ifa,
4611                                                 NETLINK_CB(cb->skb).portid,
4612                                                 cb->nlh->nlmsg_seq,
4613                                                 RTM_NEWADDR,
4614                                                 NLM_F_MULTI);
4615                         if (err < 0)
4616                                 break;
4617                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4618                 }
4619                 break;
4620         }
4621         case MULTICAST_ADDR:
4622                 /* multicast address */
4623                 for (ifmca = idev->mc_list; ifmca;
4624                      ifmca = ifmca->next, ip_idx++) {
4625                         if (ip_idx < s_ip_idx)
4626                                 continue;
4627                         err = inet6_fill_ifmcaddr(skb, ifmca,
4628                                                   NETLINK_CB(cb->skb).portid,
4629                                                   cb->nlh->nlmsg_seq,
4630                                                   RTM_GETMULTICAST,
4631                                                   NLM_F_MULTI);
4632                         if (err < 0)
4633                                 break;
4634                 }
4635                 break;
4636         case ANYCAST_ADDR:
4637                 /* anycast address */
4638                 for (ifaca = idev->ac_list; ifaca;
4639                      ifaca = ifaca->aca_next, ip_idx++) {
4640                         if (ip_idx < s_ip_idx)
4641                                 continue;
4642                         err = inet6_fill_ifacaddr(skb, ifaca,
4643                                                   NETLINK_CB(cb->skb).portid,
4644                                                   cb->nlh->nlmsg_seq,
4645                                                   RTM_GETANYCAST,
4646                                                   NLM_F_MULTI);
4647                         if (err < 0)
4648                                 break;
4649                 }
4650                 break;
4651         default:
4652                 break;
4653         }
4654         read_unlock_bh(&idev->lock);
4655         *p_ip_idx = ip_idx;
4656         return err;
4657 }
4658
4659 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4660                            enum addr_type_t type)
4661 {
4662         struct net *net = sock_net(skb->sk);
4663         int h, s_h;
4664         int idx, ip_idx;
4665         int s_idx, s_ip_idx;
4666         struct net_device *dev;
4667         struct inet6_dev *idev;
4668         struct hlist_head *head;
4669
4670         s_h = cb->args[0];
4671         s_idx = idx = cb->args[1];
4672         s_ip_idx = ip_idx = cb->args[2];
4673
4674         rcu_read_lock();
4675         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4676         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4677                 idx = 0;
4678                 head = &net->dev_index_head[h];
4679                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4680                         if (idx < s_idx)
4681                                 goto cont;
4682                         if (h > s_h || idx > s_idx)
4683                                 s_ip_idx = 0;
4684                         ip_idx = 0;
4685                         idev = __in6_dev_get(dev);
4686                         if (!idev)
4687                                 goto cont;
4688
4689                         if (in6_dump_addrs(idev, skb, cb, type,
4690                                            s_ip_idx, &ip_idx) < 0)
4691                                 goto done;
4692 cont:
4693                         idx++;
4694                 }
4695         }
4696 done:
4697         rcu_read_unlock();
4698         cb->args[0] = h;
4699         cb->args[1] = idx;
4700         cb->args[2] = ip_idx;
4701
4702         return skb->len;
4703 }
4704
4705 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4706 {
4707         enum addr_type_t type = UNICAST_ADDR;
4708
4709         return inet6_dump_addr(skb, cb, type);
4710 }
4711
4712 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4713 {
4714         enum addr_type_t type = MULTICAST_ADDR;
4715
4716         return inet6_dump_addr(skb, cb, type);
4717 }
4718
4719
4720 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4721 {
4722         enum addr_type_t type = ANYCAST_ADDR;
4723
4724         return inet6_dump_addr(skb, cb, type);
4725 }
4726
4727 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4728 {
4729         struct net *net = sock_net(in_skb->sk);
4730         struct ifaddrmsg *ifm;
4731         struct nlattr *tb[IFA_MAX+1];
4732         struct in6_addr *addr = NULL, *peer;
4733         struct net_device *dev = NULL;
4734         struct inet6_ifaddr *ifa;
4735         struct sk_buff *skb;
4736         int err;
4737
4738         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4739         if (err < 0)
4740                 goto errout;
4741
4742         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4743         if (!addr) {
4744                 err = -EINVAL;
4745                 goto errout;
4746         }
4747
4748         ifm = nlmsg_data(nlh);
4749         if (ifm->ifa_index)
4750                 dev = __dev_get_by_index(net, ifm->ifa_index);
4751
4752         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4753         if (!ifa) {
4754                 err = -EADDRNOTAVAIL;
4755                 goto errout;
4756         }
4757
4758         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4759         if (!skb) {
4760                 err = -ENOBUFS;
4761                 goto errout_ifa;
4762         }
4763
4764         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4765                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4766         if (err < 0) {
4767                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4768                 WARN_ON(err == -EMSGSIZE);
4769                 kfree_skb(skb);
4770                 goto errout_ifa;
4771         }
4772         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4773 errout_ifa:
4774         in6_ifa_put(ifa);
4775 errout:
4776         return err;
4777 }
4778
4779 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4780 {
4781         struct sk_buff *skb;
4782         struct net *net = dev_net(ifa->idev->dev);
4783         int err = -ENOBUFS;
4784
4785         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4786         if (!skb)
4787                 goto errout;
4788
4789         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4790         if (err < 0) {
4791                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4792                 WARN_ON(err == -EMSGSIZE);
4793                 kfree_skb(skb);
4794                 goto errout;
4795         }
4796         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4797         return;
4798 errout:
4799         if (err < 0)
4800                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4801 }
4802
4803 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4804                                 __s32 *array, int bytes)
4805 {
4806         BUG_ON(bytes < (DEVCONF_MAX * 4));
4807
4808         memset(array, 0, bytes);
4809         array[DEVCONF_FORWARDING] = cnf->forwarding;
4810         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4811         array[DEVCONF_MTU6] = cnf->mtu6;
4812         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4813         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4814         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4815         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4816         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4817         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4818                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4819         array[DEVCONF_RTR_SOLICIT_DELAY] =
4820                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4821         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4822         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4823                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4824         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4825                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4826         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4827         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4828         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4829         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4830         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4831         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4832         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4833         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4834         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4835 #ifdef CONFIG_IPV6_ROUTER_PREF
4836         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4837         array[DEVCONF_RTR_PROBE_INTERVAL] =
4838                 jiffies_to_msecs(cnf->rtr_probe_interval);
4839 #ifdef CONFIG_IPV6_ROUTE_INFO
4840         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4841 #endif
4842 #endif
4843         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4844         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4845 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4846         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4847         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4848 #endif
4849 #ifdef CONFIG_IPV6_MROUTE
4850         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4851 #endif
4852         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4853         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4854         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4855         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4856         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4857         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4858         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4859         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4860         /* we omit DEVCONF_STABLE_SECRET for now */
4861         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4862         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4863         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4864         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4865 }
4866
4867 static inline size_t inet6_ifla6_size(void)
4868 {
4869         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4870              + nla_total_size(sizeof(struct ifla_cacheinfo))
4871              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4872              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4873              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4874              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4875 }
4876
4877 static inline size_t inet6_if_nlmsg_size(void)
4878 {
4879         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4880                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4881                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4882                + nla_total_size(4) /* IFLA_MTU */
4883                + nla_total_size(4) /* IFLA_LINK */
4884                + nla_total_size(1) /* IFLA_OPERSTATE */
4885                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4886 }
4887
4888 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4889                                       int items, int bytes)
4890 {
4891         int i;
4892         int pad = bytes - sizeof(u64) * items;
4893         BUG_ON(pad < 0);
4894
4895         /* Use put_unaligned() because stats may not be aligned for u64. */
4896         put_unaligned(items, &stats[0]);
4897         for (i = 1; i < items; i++)
4898                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4899
4900         memset(&stats[items], 0, pad);
4901 }
4902
4903 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4904                                         int bytes, size_t syncpoff)
4905 {
4906         int i, c;
4907         u64 buff[IPSTATS_MIB_MAX];
4908         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4909
4910         BUG_ON(pad < 0);
4911
4912         memset(buff, 0, sizeof(buff));
4913         buff[0] = IPSTATS_MIB_MAX;
4914
4915         for_each_possible_cpu(c) {
4916                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4917                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4918         }
4919
4920         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4921         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4922 }
4923
4924 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4925                              int bytes)
4926 {
4927         switch (attrtype) {
4928         case IFLA_INET6_STATS:
4929                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4930                                      offsetof(struct ipstats_mib, syncp));
4931                 break;
4932         case IFLA_INET6_ICMP6STATS:
4933                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4934                 break;
4935         }
4936 }
4937
4938 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4939                                   u32 ext_filter_mask)
4940 {
4941         struct nlattr *nla;
4942         struct ifla_cacheinfo ci;
4943
4944         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4945                 goto nla_put_failure;
4946         ci.max_reasm_len = IPV6_MAXPLEN;
4947         ci.tstamp = cstamp_delta(idev->tstamp);
4948         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4949         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4950         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4951                 goto nla_put_failure;
4952         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4953         if (!nla)
4954                 goto nla_put_failure;
4955         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4956
4957         /* XXX - MC not implemented */
4958
4959         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4960                 return 0;
4961
4962         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4963         if (!nla)
4964                 goto nla_put_failure;
4965         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4966
4967         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4968         if (!nla)
4969                 goto nla_put_failure;
4970         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4971
4972         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4973         if (!nla)
4974                 goto nla_put_failure;
4975
4976         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4977                 goto nla_put_failure;
4978
4979         read_lock_bh(&idev->lock);
4980         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4981         read_unlock_bh(&idev->lock);
4982
4983         return 0;
4984
4985 nla_put_failure:
4986         return -EMSGSIZE;
4987 }
4988
4989 static size_t inet6_get_link_af_size(const struct net_device *dev,
4990                                      u32 ext_filter_mask)
4991 {
4992         if (!__in6_dev_get(dev))
4993                 return 0;
4994
4995         return inet6_ifla6_size();
4996 }
4997
4998 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4999                               u32 ext_filter_mask)
5000 {
5001         struct inet6_dev *idev = __in6_dev_get(dev);
5002
5003         if (!idev)
5004                 return -ENODATA;
5005
5006         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5007                 return -EMSGSIZE;
5008
5009         return 0;
5010 }
5011
5012 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5013 {
5014         struct inet6_ifaddr *ifp;
5015         struct net_device *dev = idev->dev;
5016         bool update_rs = false;
5017         struct in6_addr ll_addr;
5018
5019         ASSERT_RTNL();
5020
5021         if (!token)
5022                 return -EINVAL;
5023         if (ipv6_addr_any(token))
5024                 return -EINVAL;
5025         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5026                 return -EINVAL;
5027         if (!ipv6_accept_ra(idev))
5028                 return -EINVAL;
5029         if (idev->cnf.rtr_solicits <= 0)
5030                 return -EINVAL;
5031
5032         write_lock_bh(&idev->lock);
5033
5034         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5035         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5036
5037         write_unlock_bh(&idev->lock);
5038
5039         if (!idev->dead && (idev->if_flags & IF_READY) &&
5040             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5041                              IFA_F_OPTIMISTIC)) {
5042
5043                 /* If we're not ready, then normal ifup will take care
5044                  * of this. Otherwise, we need to request our rs here.
5045                  */
5046                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5047                 update_rs = true;
5048         }
5049
5050         write_lock_bh(&idev->lock);
5051
5052         if (update_rs) {
5053                 idev->if_flags |= IF_RS_SENT;
5054                 idev->rs_probes = 1;
5055                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
5056         }
5057
5058         /* Well, that's kinda nasty ... */
5059         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5060                 spin_lock(&ifp->lock);
5061                 if (ifp->tokenized) {
5062                         ifp->valid_lft = 0;
5063                         ifp->prefered_lft = 0;
5064                 }
5065                 spin_unlock(&ifp->lock);
5066         }
5067
5068         write_unlock_bh(&idev->lock);
5069         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5070         addrconf_verify_rtnl();
5071         return 0;
5072 }
5073
5074 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5075         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5076         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5077 };
5078
5079 static int inet6_validate_link_af(const struct net_device *dev,
5080                                   const struct nlattr *nla)
5081 {
5082         struct nlattr *tb[IFLA_INET6_MAX + 1];
5083
5084         if (dev && !__in6_dev_get(dev))
5085                 return -EAFNOSUPPORT;
5086
5087         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5088 }
5089
5090 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5091 {
5092         int err = -EINVAL;
5093         struct inet6_dev *idev = __in6_dev_get(dev);
5094         struct nlattr *tb[IFLA_INET6_MAX + 1];
5095
5096         if (!idev)
5097                 return -EAFNOSUPPORT;
5098
5099         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5100                 BUG();
5101
5102         if (tb[IFLA_INET6_TOKEN]) {
5103                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5104                 if (err)
5105                         return err;
5106         }
5107
5108         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5109                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5110
5111                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5112                     mode != IN6_ADDR_GEN_MODE_NONE &&
5113                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5114                     mode != IN6_ADDR_GEN_MODE_RANDOM)
5115                         return -EINVAL;
5116
5117                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5118                     !idev->cnf.stable_secret.initialized &&
5119                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5120                         return -EINVAL;
5121
5122                 idev->addr_gen_mode = mode;
5123                 err = 0;
5124         }
5125
5126         return err;
5127 }
5128
5129 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5130                              u32 portid, u32 seq, int event, unsigned int flags)
5131 {
5132         struct net_device *dev = idev->dev;
5133         struct ifinfomsg *hdr;
5134         struct nlmsghdr *nlh;
5135         void *protoinfo;
5136
5137         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5138         if (!nlh)
5139                 return -EMSGSIZE;
5140
5141         hdr = nlmsg_data(nlh);
5142         hdr->ifi_family = AF_INET6;
5143         hdr->__ifi_pad = 0;
5144         hdr->ifi_type = dev->type;
5145         hdr->ifi_index = dev->ifindex;
5146         hdr->ifi_flags = dev_get_flags(dev);
5147         hdr->ifi_change = 0;
5148
5149         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5150             (dev->addr_len &&
5151              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5152             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5153             (dev->ifindex != dev_get_iflink(dev) &&
5154              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5155             nla_put_u8(skb, IFLA_OPERSTATE,
5156                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5157                 goto nla_put_failure;
5158         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5159         if (!protoinfo)
5160                 goto nla_put_failure;
5161
5162         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5163                 goto nla_put_failure;
5164
5165         nla_nest_end(skb, protoinfo);
5166         nlmsg_end(skb, nlh);
5167         return 0;
5168
5169 nla_put_failure:
5170         nlmsg_cancel(skb, nlh);
5171         return -EMSGSIZE;
5172 }
5173
5174 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5175 {
5176         struct net *net = sock_net(skb->sk);
5177         int h, s_h;
5178         int idx = 0, s_idx;
5179         struct net_device *dev;
5180         struct inet6_dev *idev;
5181         struct hlist_head *head;
5182
5183         s_h = cb->args[0];
5184         s_idx = cb->args[1];
5185
5186         rcu_read_lock();
5187         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5188                 idx = 0;
5189                 head = &net->dev_index_head[h];
5190                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5191                         if (idx < s_idx)
5192                                 goto cont;
5193                         idev = __in6_dev_get(dev);
5194                         if (!idev)
5195                                 goto cont;
5196                         if (inet6_fill_ifinfo(skb, idev,
5197                                               NETLINK_CB(cb->skb).portid,
5198                                               cb->nlh->nlmsg_seq,
5199                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5200                                 goto out;
5201 cont:
5202                         idx++;
5203                 }
5204         }
5205 out:
5206         rcu_read_unlock();
5207         cb->args[1] = idx;
5208         cb->args[0] = h;
5209
5210         return skb->len;
5211 }
5212
5213 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5214 {
5215         struct sk_buff *skb;
5216         struct net *net = dev_net(idev->dev);
5217         int err = -ENOBUFS;
5218
5219         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5220         if (!skb)
5221                 goto errout;
5222
5223         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5224         if (err < 0) {
5225                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5226                 WARN_ON(err == -EMSGSIZE);
5227                 kfree_skb(skb);
5228                 goto errout;
5229         }
5230         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5231         return;
5232 errout:
5233         if (err < 0)
5234                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5235 }
5236
5237 static inline size_t inet6_prefix_nlmsg_size(void)
5238 {
5239         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5240                + nla_total_size(sizeof(struct in6_addr))
5241                + nla_total_size(sizeof(struct prefix_cacheinfo));
5242 }
5243
5244 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5245                              struct prefix_info *pinfo, u32 portid, u32 seq,
5246                              int event, unsigned int flags)
5247 {
5248         struct prefixmsg *pmsg;
5249         struct nlmsghdr *nlh;
5250         struct prefix_cacheinfo ci;
5251
5252         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5253         if (!nlh)
5254                 return -EMSGSIZE;
5255
5256         pmsg = nlmsg_data(nlh);
5257         pmsg->prefix_family = AF_INET6;
5258         pmsg->prefix_pad1 = 0;
5259         pmsg->prefix_pad2 = 0;
5260         pmsg->prefix_ifindex = idev->dev->ifindex;
5261         pmsg->prefix_len = pinfo->prefix_len;
5262         pmsg->prefix_type = pinfo->type;
5263         pmsg->prefix_pad3 = 0;
5264         pmsg->prefix_flags = 0;
5265         if (pinfo->onlink)
5266                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5267         if (pinfo->autoconf)
5268                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5269
5270         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5271                 goto nla_put_failure;
5272         ci.preferred_time = ntohl(pinfo->prefered);
5273         ci.valid_time = ntohl(pinfo->valid);
5274         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5275                 goto nla_put_failure;
5276         nlmsg_end(skb, nlh);
5277         return 0;
5278
5279 nla_put_failure:
5280         nlmsg_cancel(skb, nlh);
5281         return -EMSGSIZE;
5282 }
5283
5284 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5285                          struct prefix_info *pinfo)
5286 {
5287         struct sk_buff *skb;
5288         struct net *net = dev_net(idev->dev);
5289         int err = -ENOBUFS;
5290
5291         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5292         if (!skb)
5293                 goto errout;
5294
5295         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5296         if (err < 0) {
5297                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5298                 WARN_ON(err == -EMSGSIZE);
5299                 kfree_skb(skb);
5300                 goto errout;
5301         }
5302         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5303         return;
5304 errout:
5305         if (err < 0)
5306                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5307 }
5308
5309 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5310 {
5311         struct net *net = dev_net(ifp->idev->dev);
5312
5313         if (event)
5314                 ASSERT_RTNL();
5315
5316         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5317
5318         switch (event) {
5319         case RTM_NEWADDR:
5320                 /*
5321                  * If the address was optimistic
5322                  * we inserted the route at the start of
5323                  * our DAD process, so we don't need
5324                  * to do it again
5325                  */
5326                 if (!(ifp->rt->rt6i_node))
5327                         ip6_ins_rt(ifp->rt);
5328                 if (ifp->idev->cnf.forwarding)
5329                         addrconf_join_anycast(ifp);
5330                 if (!ipv6_addr_any(&ifp->peer_addr))
5331                         addrconf_prefix_route(&ifp->peer_addr, 128,
5332                                               ifp->idev->dev, 0, 0);
5333                 break;
5334         case RTM_DELADDR:
5335                 if (ifp->idev->cnf.forwarding)
5336                         addrconf_leave_anycast(ifp);
5337                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5338                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5339                         struct rt6_info *rt;
5340
5341                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5342                                                        ifp->idev->dev, 0, 0);
5343                         if (rt)
5344                                 ip6_del_rt(rt);
5345                 }
5346                 dst_hold(&ifp->rt->dst);
5347
5348                 ip6_del_rt(ifp->rt);
5349
5350                 rt_genid_bump_ipv6(net);
5351                 break;
5352         }
5353         atomic_inc(&net->ipv6.dev_addr_genid);
5354 }
5355
5356 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5357 {
5358         rcu_read_lock_bh();
5359         if (likely(ifp->idev->dead == 0))
5360                 __ipv6_ifa_notify(event, ifp);
5361         rcu_read_unlock_bh();
5362 }
5363
5364 #ifdef CONFIG_SYSCTL
5365
5366 static
5367 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5368                            void __user *buffer, size_t *lenp, loff_t *ppos)
5369 {
5370         int *valp = ctl->data;
5371         int val = *valp;
5372         loff_t pos = *ppos;
5373         struct ctl_table lctl;
5374         int ret;
5375
5376         /*
5377          * ctl->data points to idev->cnf.forwarding, we should
5378          * not modify it until we get the rtnl lock.
5379          */
5380         lctl = *ctl;
5381         lctl.data = &val;
5382
5383         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5384
5385         if (write)
5386                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5387         if (ret)
5388                 *ppos = pos;
5389         return ret;
5390 }
5391
5392 static
5393 int addrconf_sysctl_hop_limit(struct ctl_table *ctl, int write,
5394                               void __user *buffer, size_t *lenp, loff_t *ppos)
5395 {
5396         struct ctl_table lctl;
5397         int min_hl = 1, max_hl = 255;
5398
5399         lctl = *ctl;
5400         lctl.extra1 = &min_hl;
5401         lctl.extra2 = &max_hl;
5402
5403         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5404 }
5405
5406 static
5407 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5408                         void __user *buffer, size_t *lenp, loff_t *ppos)
5409 {
5410         struct inet6_dev *idev = ctl->extra1;
5411         int min_mtu = IPV6_MIN_MTU;
5412         struct ctl_table lctl;
5413
5414         lctl = *ctl;
5415         lctl.extra1 = &min_mtu;
5416         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5417
5418         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5419 }
5420
5421 static void dev_disable_change(struct inet6_dev *idev)
5422 {
5423         struct netdev_notifier_info info;
5424
5425         if (!idev || !idev->dev)
5426                 return;
5427
5428         netdev_notifier_info_init(&info, idev->dev);
5429         if (idev->cnf.disable_ipv6)
5430                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5431         else
5432                 addrconf_notify(NULL, NETDEV_UP, &info);
5433 }
5434
5435 static void addrconf_disable_change(struct net *net, __s32 newf)
5436 {
5437         struct net_device *dev;
5438         struct inet6_dev *idev;
5439
5440         rcu_read_lock();
5441         for_each_netdev_rcu(net, dev) {
5442                 idev = __in6_dev_get(dev);
5443                 if (idev) {
5444                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5445                         idev->cnf.disable_ipv6 = newf;
5446                         if (changed)
5447                                 dev_disable_change(idev);
5448                 }
5449         }
5450         rcu_read_unlock();
5451 }
5452
5453 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5454 {
5455         struct net *net;
5456         int old;
5457
5458         if (!rtnl_trylock())
5459                 return restart_syscall();
5460
5461         net = (struct net *)table->extra2;
5462         old = *p;
5463         *p = newf;
5464
5465         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5466                 rtnl_unlock();
5467                 return 0;
5468         }
5469
5470         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5471                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5472                 addrconf_disable_change(net, newf);
5473         } else if ((!newf) ^ (!old))
5474                 dev_disable_change((struct inet6_dev *)table->extra1);
5475
5476         rtnl_unlock();
5477         return 0;
5478 }
5479
5480 static
5481 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5482                             void __user *buffer, size_t *lenp, loff_t *ppos)
5483 {
5484         int *valp = ctl->data;
5485         int val = *valp;
5486         loff_t pos = *ppos;
5487         struct ctl_table lctl;
5488         int ret;
5489
5490         /*
5491          * ctl->data points to idev->cnf.disable_ipv6, we should
5492          * not modify it until we get the rtnl lock.
5493          */
5494         lctl = *ctl;
5495         lctl.data = &val;
5496
5497         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5498
5499         if (write)
5500                 ret = addrconf_disable_ipv6(ctl, valp, val);
5501         if (ret)
5502                 *ppos = pos;
5503         return ret;
5504 }
5505
5506 static
5507 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5508                               void __user *buffer, size_t *lenp, loff_t *ppos)
5509 {
5510         int *valp = ctl->data;
5511         int ret;
5512         int old, new;
5513
5514         old = *valp;
5515         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5516         new = *valp;
5517
5518         if (write && old != new) {
5519                 struct net *net = ctl->extra2;
5520
5521                 if (!rtnl_trylock())
5522                         return restart_syscall();
5523
5524                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5525                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5526                                                      NETCONFA_IFINDEX_DEFAULT,
5527                                                      net->ipv6.devconf_dflt);
5528                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5529                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5530                                                      NETCONFA_IFINDEX_ALL,
5531                                                      net->ipv6.devconf_all);
5532                 else {
5533                         struct inet6_dev *idev = ctl->extra1;
5534
5535                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5536                                                      idev->dev->ifindex,
5537                                                      &idev->cnf);
5538                 }
5539                 rtnl_unlock();
5540         }
5541
5542         return ret;
5543 }
5544
5545 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5546                                          void __user *buffer, size_t *lenp,
5547                                          loff_t *ppos)
5548 {
5549         int err;
5550         struct in6_addr addr;
5551         char str[IPV6_MAX_STRLEN];
5552         struct ctl_table lctl = *ctl;
5553         struct net *net = ctl->extra2;
5554         struct ipv6_stable_secret *secret = ctl->data;
5555
5556         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5557                 return -EIO;
5558
5559         lctl.maxlen = IPV6_MAX_STRLEN;
5560         lctl.data = str;
5561
5562         if (!rtnl_trylock())
5563                 return restart_syscall();
5564
5565         if (!write && !secret->initialized) {
5566                 err = -EIO;
5567                 goto out;
5568         }
5569
5570         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5571         if (err >= sizeof(str)) {
5572                 err = -EIO;
5573                 goto out;
5574         }
5575
5576         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5577         if (err || !write)
5578                 goto out;
5579
5580         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5581                 err = -EIO;
5582                 goto out;
5583         }
5584
5585         secret->initialized = true;
5586         secret->secret = addr;
5587
5588         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5589                 struct net_device *dev;
5590
5591                 for_each_netdev(net, dev) {
5592                         struct inet6_dev *idev = __in6_dev_get(dev);
5593
5594                         if (idev) {
5595                                 idev->addr_gen_mode =
5596                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5597                         }
5598                 }
5599         } else {
5600                 struct inet6_dev *idev = ctl->extra1;
5601
5602                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5603         }
5604
5605 out:
5606         rtnl_unlock();
5607
5608         return err;
5609 }
5610
5611 static
5612 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5613                                                 int write,
5614                                                 void __user *buffer,
5615                                                 size_t *lenp,
5616                                                 loff_t *ppos)
5617 {
5618         int *valp = ctl->data;
5619         int val = *valp;
5620         loff_t pos = *ppos;
5621         struct ctl_table lctl;
5622         int ret;
5623
5624         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5625          * we should not modify it until we get the rtnl lock.
5626          */
5627         lctl = *ctl;
5628         lctl.data = &val;
5629
5630         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5631
5632         if (write)
5633                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5634         if (ret)
5635                 *ppos = pos;
5636         return ret;
5637 }
5638
5639 static struct addrconf_sysctl_table
5640 {
5641         struct ctl_table_header *sysctl_header;
5642         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5643 } addrconf_sysctl __read_mostly = {
5644         .sysctl_header = NULL,
5645         .addrconf_vars = {
5646                 {
5647                         .procname       = "forwarding",
5648                         .data           = &ipv6_devconf.forwarding,
5649                         .maxlen         = sizeof(int),
5650                         .mode           = 0644,
5651                         .proc_handler   = addrconf_sysctl_forward,
5652                 },
5653                 {
5654                         .procname       = "hop_limit",
5655                         .data           = &ipv6_devconf.hop_limit,
5656                         .maxlen         = sizeof(int),
5657                         .mode           = 0644,
5658                         .proc_handler   = addrconf_sysctl_hop_limit,
5659                 },
5660                 {
5661                         .procname       = "mtu",
5662                         .data           = &ipv6_devconf.mtu6,
5663                         .maxlen         = sizeof(int),
5664                         .mode           = 0644,
5665                         .proc_handler   = addrconf_sysctl_mtu,
5666                 },
5667                 {
5668                         .procname       = "accept_ra",
5669                         .data           = &ipv6_devconf.accept_ra,
5670                         .maxlen         = sizeof(int),
5671                         .mode           = 0644,
5672                         .proc_handler   = proc_dointvec,
5673                 },
5674                 {
5675                         .procname       = "accept_redirects",
5676                         .data           = &ipv6_devconf.accept_redirects,
5677                         .maxlen         = sizeof(int),
5678                         .mode           = 0644,
5679                         .proc_handler   = proc_dointvec,
5680                 },
5681                 {
5682                         .procname       = "autoconf",
5683                         .data           = &ipv6_devconf.autoconf,
5684                         .maxlen         = sizeof(int),
5685                         .mode           = 0644,
5686                         .proc_handler   = proc_dointvec,
5687                 },
5688                 {
5689                         .procname       = "dad_transmits",
5690                         .data           = &ipv6_devconf.dad_transmits,
5691                         .maxlen         = sizeof(int),
5692                         .mode           = 0644,
5693                         .proc_handler   = proc_dointvec,
5694                 },
5695                 {
5696                         .procname       = "router_solicitations",
5697                         .data           = &ipv6_devconf.rtr_solicits,
5698                         .maxlen         = sizeof(int),
5699                         .mode           = 0644,
5700                         .proc_handler   = proc_dointvec,
5701                 },
5702                 {
5703                         .procname       = "router_solicitation_interval",
5704                         .data           = &ipv6_devconf.rtr_solicit_interval,
5705                         .maxlen         = sizeof(int),
5706                         .mode           = 0644,
5707                         .proc_handler   = proc_dointvec_jiffies,
5708                 },
5709                 {
5710                         .procname       = "router_solicitation_delay",
5711                         .data           = &ipv6_devconf.rtr_solicit_delay,
5712                         .maxlen         = sizeof(int),
5713                         .mode           = 0644,
5714                         .proc_handler   = proc_dointvec_jiffies,
5715                 },
5716                 {
5717                         .procname       = "force_mld_version",
5718                         .data           = &ipv6_devconf.force_mld_version,
5719                         .maxlen         = sizeof(int),
5720                         .mode           = 0644,
5721                         .proc_handler   = proc_dointvec,
5722                 },
5723                 {
5724                         .procname       = "mldv1_unsolicited_report_interval",
5725                         .data           =
5726                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5727                         .maxlen         = sizeof(int),
5728                         .mode           = 0644,
5729                         .proc_handler   = proc_dointvec_ms_jiffies,
5730                 },
5731                 {
5732                         .procname       = "mldv2_unsolicited_report_interval",
5733                         .data           =
5734                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5735                         .maxlen         = sizeof(int),
5736                         .mode           = 0644,
5737                         .proc_handler   = proc_dointvec_ms_jiffies,
5738                 },
5739                 {
5740                         .procname       = "use_tempaddr",
5741                         .data           = &ipv6_devconf.use_tempaddr,
5742                         .maxlen         = sizeof(int),
5743                         .mode           = 0644,
5744                         .proc_handler   = proc_dointvec,
5745                 },
5746                 {
5747                         .procname       = "temp_valid_lft",
5748                         .data           = &ipv6_devconf.temp_valid_lft,
5749                         .maxlen         = sizeof(int),
5750                         .mode           = 0644,
5751                         .proc_handler   = proc_dointvec,
5752                 },
5753                 {
5754                         .procname       = "temp_prefered_lft",
5755                         .data           = &ipv6_devconf.temp_prefered_lft,
5756                         .maxlen         = sizeof(int),
5757                         .mode           = 0644,
5758                         .proc_handler   = proc_dointvec,
5759                 },
5760                 {
5761                         .procname       = "regen_max_retry",
5762                         .data           = &ipv6_devconf.regen_max_retry,
5763                         .maxlen         = sizeof(int),
5764                         .mode           = 0644,
5765                         .proc_handler   = proc_dointvec,
5766                 },
5767                 {
5768                         .procname       = "max_desync_factor",
5769                         .data           = &ipv6_devconf.max_desync_factor,
5770                         .maxlen         = sizeof(int),
5771                         .mode           = 0644,
5772                         .proc_handler   = proc_dointvec,
5773                 },
5774                 {
5775                         .procname       = "max_addresses",
5776                         .data           = &ipv6_devconf.max_addresses,
5777                         .maxlen         = sizeof(int),
5778                         .mode           = 0644,
5779                         .proc_handler   = proc_dointvec,
5780                 },
5781                 {
5782                         .procname       = "accept_ra_defrtr",
5783                         .data           = &ipv6_devconf.accept_ra_defrtr,
5784                         .maxlen         = sizeof(int),
5785                         .mode           = 0644,
5786                         .proc_handler   = proc_dointvec,
5787                 },
5788                 {
5789                         .procname       = "accept_ra_min_hop_limit",
5790                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5791                         .maxlen         = sizeof(int),
5792                         .mode           = 0644,
5793                         .proc_handler   = proc_dointvec,
5794                 },
5795                 {
5796                         .procname       = "accept_ra_pinfo",
5797                         .data           = &ipv6_devconf.accept_ra_pinfo,
5798                         .maxlen         = sizeof(int),
5799                         .mode           = 0644,
5800                         .proc_handler   = proc_dointvec,
5801                 },
5802 #ifdef CONFIG_IPV6_ROUTER_PREF
5803                 {
5804                         .procname       = "accept_ra_rtr_pref",
5805                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5806                         .maxlen         = sizeof(int),
5807                         .mode           = 0644,
5808                         .proc_handler   = proc_dointvec,
5809                 },
5810                 {
5811                         .procname       = "router_probe_interval",
5812                         .data           = &ipv6_devconf.rtr_probe_interval,
5813                         .maxlen         = sizeof(int),
5814                         .mode           = 0644,
5815                         .proc_handler   = proc_dointvec_jiffies,
5816                 },
5817 #ifdef CONFIG_IPV6_ROUTE_INFO
5818                 {
5819                         .procname       = "accept_ra_rt_info_max_plen",
5820                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5821                         .maxlen         = sizeof(int),
5822                         .mode           = 0644,
5823                         .proc_handler   = proc_dointvec,
5824                 },
5825 #endif
5826 #endif
5827                 {
5828                         .procname       = "proxy_ndp",
5829                         .data           = &ipv6_devconf.proxy_ndp,
5830                         .maxlen         = sizeof(int),
5831                         .mode           = 0644,
5832                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5833                 },
5834                 {
5835                         .procname       = "accept_source_route",
5836                         .data           = &ipv6_devconf.accept_source_route,
5837                         .maxlen         = sizeof(int),
5838                         .mode           = 0644,
5839                         .proc_handler   = proc_dointvec,
5840                 },
5841 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5842                 {
5843                         .procname       = "optimistic_dad",
5844                         .data           = &ipv6_devconf.optimistic_dad,
5845                         .maxlen         = sizeof(int),
5846                         .mode           = 0644,
5847                         .proc_handler   = proc_dointvec,
5848
5849                 },
5850                 {
5851                         .procname       = "use_optimistic",
5852                         .data           = &ipv6_devconf.use_optimistic,
5853                         .maxlen         = sizeof(int),
5854                         .mode           = 0644,
5855                         .proc_handler   = proc_dointvec,
5856
5857                 },
5858 #endif
5859 #ifdef CONFIG_IPV6_MROUTE
5860                 {
5861                         .procname       = "mc_forwarding",
5862                         .data           = &ipv6_devconf.mc_forwarding,
5863                         .maxlen         = sizeof(int),
5864                         .mode           = 0444,
5865                         .proc_handler   = proc_dointvec,
5866                 },
5867 #endif
5868                 {
5869                         .procname       = "disable_ipv6",
5870                         .data           = &ipv6_devconf.disable_ipv6,
5871                         .maxlen         = sizeof(int),
5872                         .mode           = 0644,
5873                         .proc_handler   = addrconf_sysctl_disable,
5874                 },
5875                 {
5876                         .procname       = "accept_dad",
5877                         .data           = &ipv6_devconf.accept_dad,
5878                         .maxlen         = sizeof(int),
5879                         .mode           = 0644,
5880                         .proc_handler   = proc_dointvec,
5881                 },
5882                 {
5883                         .procname       = "force_tllao",
5884                         .data           = &ipv6_devconf.force_tllao,
5885                         .maxlen         = sizeof(int),
5886                         .mode           = 0644,
5887                         .proc_handler   = proc_dointvec
5888                 },
5889                 {
5890                         .procname       = "ndisc_notify",
5891                         .data           = &ipv6_devconf.ndisc_notify,
5892                         .maxlen         = sizeof(int),
5893                         .mode           = 0644,
5894                         .proc_handler   = proc_dointvec
5895                 },
5896                 {
5897                         .procname       = "suppress_frag_ndisc",
5898                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5899                         .maxlen         = sizeof(int),
5900                         .mode           = 0644,
5901                         .proc_handler   = proc_dointvec
5902                 },
5903                 {
5904                         .procname       = "accept_ra_from_local",
5905                         .data           = &ipv6_devconf.accept_ra_from_local,
5906                         .maxlen         = sizeof(int),
5907                         .mode           = 0644,
5908                         .proc_handler   = proc_dointvec,
5909                 },
5910                 {
5911                         .procname       = "accept_ra_mtu",
5912                         .data           = &ipv6_devconf.accept_ra_mtu,
5913                         .maxlen         = sizeof(int),
5914                         .mode           = 0644,
5915                         .proc_handler   = proc_dointvec,
5916                 },
5917                 {
5918                         .procname       = "stable_secret",
5919                         .data           = &ipv6_devconf.stable_secret,
5920                         .maxlen         = IPV6_MAX_STRLEN,
5921                         .mode           = 0600,
5922                         .proc_handler   = addrconf_sysctl_stable_secret,
5923                 },
5924                 {
5925                         .procname       = "use_oif_addrs_only",
5926                         .data           = &ipv6_devconf.use_oif_addrs_only,
5927                         .maxlen         = sizeof(int),
5928                         .mode           = 0644,
5929                         .proc_handler   = proc_dointvec,
5930                 },
5931                 {
5932                         .procname       = "ignore_routes_with_linkdown",
5933                         .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5934                         .maxlen         = sizeof(int),
5935                         .mode           = 0644,
5936                         .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5937                 },
5938                 {
5939                         .procname       = "drop_unicast_in_l2_multicast",
5940                         .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
5941                         .maxlen         = sizeof(int),
5942                         .mode           = 0644,
5943                         .proc_handler   = proc_dointvec,
5944                 },
5945                 {
5946                         .procname       = "drop_unsolicited_na",
5947                         .data           = &ipv6_devconf.drop_unsolicited_na,
5948                         .maxlen         = sizeof(int),
5949                         .mode           = 0644,
5950                         .proc_handler   = proc_dointvec,
5951                 },
5952                 {
5953                         .procname       = "keep_addr_on_down",
5954                         .data           = &ipv6_devconf.keep_addr_on_down,
5955                         .maxlen         = sizeof(int),
5956                         .mode           = 0644,
5957                         .proc_handler   = proc_dointvec,
5958
5959                 },
5960                 {
5961                         /* sentinel */
5962                 }
5963         },
5964 };
5965
5966 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5967                 struct inet6_dev *idev, struct ipv6_devconf *p)
5968 {
5969         int i;
5970         struct addrconf_sysctl_table *t;
5971         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5972
5973         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5974         if (!t)
5975                 goto out;
5976
5977         for (i = 0; t->addrconf_vars[i].data; i++) {
5978                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5979                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5980                 t->addrconf_vars[i].extra2 = net;
5981         }
5982
5983         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5984
5985         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5986         if (!t->sysctl_header)
5987                 goto free;
5988
5989         p->sysctl = t;
5990         return 0;
5991
5992 free:
5993         kfree(t);
5994 out:
5995         return -ENOBUFS;
5996 }
5997
5998 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5999 {
6000         struct addrconf_sysctl_table *t;
6001
6002         if (!p->sysctl)
6003                 return;
6004
6005         t = p->sysctl;
6006         p->sysctl = NULL;
6007         unregister_net_sysctl_table(t->sysctl_header);
6008         kfree(t);
6009 }
6010
6011 static int addrconf_sysctl_register(struct inet6_dev *idev)
6012 {
6013         int err;
6014
6015         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6016                 return -EINVAL;
6017
6018         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6019                                     &ndisc_ifinfo_sysctl_change);
6020         if (err)
6021                 return err;
6022         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6023                                          idev, &idev->cnf);
6024         if (err)
6025                 neigh_sysctl_unregister(idev->nd_parms);
6026
6027         return err;
6028 }
6029
6030 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6031 {
6032         __addrconf_sysctl_unregister(&idev->cnf);
6033         neigh_sysctl_unregister(idev->nd_parms);
6034 }
6035
6036
6037 #endif
6038
6039 static int __net_init addrconf_init_net(struct net *net)
6040 {
6041         int err = -ENOMEM;
6042         struct ipv6_devconf *all, *dflt;
6043
6044         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6045         if (!all)
6046                 goto err_alloc_all;
6047
6048         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6049         if (!dflt)
6050                 goto err_alloc_dflt;
6051
6052         /* these will be inherited by all namespaces */
6053         dflt->autoconf = ipv6_defaults.autoconf;
6054         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6055
6056         dflt->stable_secret.initialized = false;
6057         all->stable_secret.initialized = false;
6058
6059         net->ipv6.devconf_all = all;
6060         net->ipv6.devconf_dflt = dflt;
6061
6062 #ifdef CONFIG_SYSCTL
6063         err = __addrconf_sysctl_register(net, "all", NULL, all);
6064         if (err < 0)
6065                 goto err_reg_all;
6066
6067         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6068         if (err < 0)
6069                 goto err_reg_dflt;
6070 #endif
6071         return 0;
6072
6073 #ifdef CONFIG_SYSCTL
6074 err_reg_dflt:
6075         __addrconf_sysctl_unregister(all);
6076 err_reg_all:
6077         kfree(dflt);
6078 #endif
6079 err_alloc_dflt:
6080         kfree(all);
6081 err_alloc_all:
6082         return err;
6083 }
6084
6085 static void __net_exit addrconf_exit_net(struct net *net)
6086 {
6087 #ifdef CONFIG_SYSCTL
6088         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6089         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6090 #endif
6091         kfree(net->ipv6.devconf_dflt);
6092         kfree(net->ipv6.devconf_all);
6093 }
6094
6095 static struct pernet_operations addrconf_ops = {
6096         .init = addrconf_init_net,
6097         .exit = addrconf_exit_net,
6098 };
6099
6100 static struct rtnl_af_ops inet6_ops __read_mostly = {
6101         .family           = AF_INET6,
6102         .fill_link_af     = inet6_fill_link_af,
6103         .get_link_af_size = inet6_get_link_af_size,
6104         .validate_link_af = inet6_validate_link_af,
6105         .set_link_af      = inet6_set_link_af,
6106 };
6107
6108 /*
6109  *      Init / cleanup code
6110  */
6111
6112 int __init addrconf_init(void)
6113 {
6114         struct inet6_dev *idev;
6115         int i, err;
6116
6117         err = ipv6_addr_label_init();
6118         if (err < 0) {
6119                 pr_crit("%s: cannot initialize default policy table: %d\n",
6120                         __func__, err);
6121                 goto out;
6122         }
6123
6124         err = register_pernet_subsys(&addrconf_ops);
6125         if (err < 0)
6126                 goto out_addrlabel;
6127
6128         addrconf_wq = create_workqueue("ipv6_addrconf");
6129         if (!addrconf_wq) {
6130                 err = -ENOMEM;
6131                 goto out_nowq;
6132         }
6133
6134         /* The addrconf netdev notifier requires that loopback_dev
6135          * has it's ipv6 private information allocated and setup
6136          * before it can bring up and give link-local addresses
6137          * to other devices which are up.
6138          *
6139          * Unfortunately, loopback_dev is not necessarily the first
6140          * entry in the global dev_base list of net devices.  In fact,
6141          * it is likely to be the very last entry on that list.
6142          * So this causes the notifier registry below to try and
6143          * give link-local addresses to all devices besides loopback_dev
6144          * first, then loopback_dev, which cases all the non-loopback_dev
6145          * devices to fail to get a link-local address.
6146          *
6147          * So, as a temporary fix, allocate the ipv6 structure for
6148          * loopback_dev first by hand.
6149          * Longer term, all of the dependencies ipv6 has upon the loopback
6150          * device and it being up should be removed.
6151          */
6152         rtnl_lock();
6153         idev = ipv6_add_dev(init_net.loopback_dev);
6154         rtnl_unlock();
6155         if (IS_ERR(idev)) {
6156                 err = PTR_ERR(idev);
6157                 goto errlo;
6158         }
6159
6160         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6161                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6162
6163         register_netdevice_notifier(&ipv6_dev_notf);
6164
6165         addrconf_verify();
6166
6167         rtnl_af_register(&inet6_ops);
6168
6169         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6170                               NULL);
6171         if (err < 0)
6172                 goto errout;
6173
6174         /* Only the first call to __rtnl_register can fail */
6175         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6176         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6177         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6178                         inet6_dump_ifaddr, NULL);
6179         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6180                         inet6_dump_ifmcaddr, NULL);
6181         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6182                         inet6_dump_ifacaddr, NULL);
6183         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6184                         inet6_netconf_dump_devconf, NULL);
6185
6186         ipv6_addr_label_rtnl_register();
6187
6188         return 0;
6189 errout:
6190         rtnl_af_unregister(&inet6_ops);
6191         unregister_netdevice_notifier(&ipv6_dev_notf);
6192 errlo:
6193         destroy_workqueue(addrconf_wq);
6194 out_nowq:
6195         unregister_pernet_subsys(&addrconf_ops);
6196 out_addrlabel:
6197         ipv6_addr_label_cleanup();
6198 out:
6199         return err;
6200 }
6201
6202 void addrconf_cleanup(void)
6203 {
6204         struct net_device *dev;
6205         int i;
6206
6207         unregister_netdevice_notifier(&ipv6_dev_notf);
6208         unregister_pernet_subsys(&addrconf_ops);
6209         ipv6_addr_label_cleanup();
6210
6211         rtnl_lock();
6212
6213         __rtnl_af_unregister(&inet6_ops);
6214
6215         /* clean dev list */
6216         for_each_netdev(&init_net, dev) {
6217                 if (__in6_dev_get(dev) == NULL)
6218                         continue;
6219                 addrconf_ifdown(dev, 1);
6220         }
6221         addrconf_ifdown(init_net.loopback_dev, 2);
6222
6223         /*
6224          *      Check hash table.
6225          */
6226         spin_lock_bh(&addrconf_hash_lock);
6227         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6228                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6229         spin_unlock_bh(&addrconf_hash_lock);
6230         cancel_delayed_work(&addr_chk_work);
6231         rtnl_unlock();
6232
6233         destroy_workqueue(addrconf_wq);
6234 }