net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/inet.h>
  50 #include <linux/in6.h>
  51 #include <linux/netdevice.h>
  52 #include <linux/if_addr.h>
  53 #include <linux/if_arp.h>
  54 #include <linux/if_arcnet.h>
  55 #include <linux/if_infiniband.h>
  56 #include <linux/route.h>
  57 #include <linux/inetdevice.h>
  58 #include <linux/init.h>
  59 #include <linux/slab.h>
  60 #ifdef CONFIG_SYSCTL
  61 #include <linux/sysctl.h>
  62 #endif
  63 #include <linux/capability.h>
  64 #include <linux/delay.h>
  65 #include <linux/notifier.h>
  66 #include <linux/string.h>
  67 #include <linux/hash.h>
  68
  69 #include <net/net_namespace.h>
  70 #include <net/sock.h>
  71 #include <net/snmp.h>
  72
  73 #include <net/6lowpan.h>
  74 #include <net/firewire.h>
  75 #include <net/ipv6.h>
  76 #include <net/protocol.h>
  77 #include <net/ndisc.h>
  78 #include <net/ip6_route.h>
  79 #include <net/addrconf.h>
  80 #include <net/tcp.h>
  81 #include <net/ip.h>
  82 #include <net/netlink.h>
  83 #include <net/pkt_sched.h>
  84 #include <net/l3mdev.h>
  85 #include <linux/if_tunnel.h>
  86 #include <linux/rtnetlink.h>
  87 #include <linux/netconf.h>
  88 #include <linux/random.h>
  89 #include <linux/uaccess.h>
  90 #include <asm/unaligned.h>
  91
  92 #include <linux/proc_fs.h>
  93 #include <linux/seq_file.h>
  94 #include <linux/export.h>
  95
  96 /* Set to 3 to get tracing... */
  97 #define ACONF_DEBUG 2
  98
  99 #if ACONF_DEBUG >= 3
 100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 101 #else
 102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 103 #endif
 104
 105 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 106
 107 #define IPV6_MAX_STRLEN \
 108         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 109
 110 static inline u32 cstamp_delta(unsigned long cstamp)
 111 {
 112         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 113 }
 114
 115 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 116 {
 117         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 118         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 119         do_div(tmp, 1000000);
 120         return (s32)tmp;
 121 }
 122
 123 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 124 {
 125         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 126         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 127         do_div(tmp, 1000000);
 128         if ((s32)tmp > mrt) {
 129                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 130                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 131                 do_div(tmp, 1000000);
 132         }
 133         return (s32)tmp;
 134 }
 135
 136 #ifdef CONFIG_SYSCTL
 137 static int addrconf_sysctl_register(struct inet6_dev *idev);
 138 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 139 #else
 140 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 141 {
 142         return 0;
 143 }
 144
 145 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 146 {
 147 }
 148 #endif
 149
 150 static void ipv6_regen_rndid(struct inet6_dev *idev);
 151 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 152
 153 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 154 static int ipv6_count_addresses(struct inet6_dev *idev);
 155 static int ipv6_generate_stable_address(struct in6_addr *addr,
 156                                         u8 dad_count,
 157                                         const struct inet6_dev *idev);
 158
 159 /*
 160  *      Configured unicast address hash table
 161  */
 162 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 163 static DEFINE_SPINLOCK(addrconf_hash_lock);
 164
 165 static void addrconf_verify(void);
 166 static void addrconf_verify_rtnl(void);
 167 static void addrconf_verify_work(struct work_struct *);
 168
 169 static struct workqueue_struct *addrconf_wq;
 170 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 171
 172 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 173 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 174
 175 static void addrconf_type_change(struct net_device *dev,
 176                                  unsigned long event);
 177 static int addrconf_ifdown(struct net_device *dev, int how);
 178
 179 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 180                                                   int plen,
 181                                                   const struct net_device *dev,
 182                                                   u32 flags, u32 noflags);
 183
 184 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 185 static void addrconf_dad_work(struct work_struct *w);
 186 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id);
 187 static void addrconf_dad_run(struct inet6_dev *idev);
 188 static void addrconf_rs_timer(unsigned long data);
 189 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 190 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 191
 192 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 193                                 struct prefix_info *pinfo);
 194 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 195                                struct net_device *dev);
 196
 197 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 198         .forwarding             = 0,
 199         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 200         .mtu6                   = IPV6_MIN_MTU,
 201         .accept_ra              = 1,
 202         .accept_redirects       = 1,
 203         .autoconf               = 1,
 204         .force_mld_version      = 0,
 205         .mldv1_unsolicited_report_interval = 10 * HZ,
 206         .mldv2_unsolicited_report_interval = HZ,
 207         .dad_transmits          = 1,
 208         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 209         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 210         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 211         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 212         .use_tempaddr           = 0,
 213         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 214         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 215         .regen_max_retry        = REGEN_MAX_RETRY,
 216         .max_desync_factor      = MAX_DESYNC_FACTOR,
 217         .max_addresses          = IPV6_MAX_ADDRESSES,
 218         .accept_ra_defrtr       = 1,
 219         .accept_ra_from_local   = 0,
 220         .accept_ra_min_hop_limit= 1,
 221         .accept_ra_pinfo        = 1,
 222 #ifdef CONFIG_IPV6_ROUTER_PREF
 223         .accept_ra_rtr_pref     = 1,
 224         .rtr_probe_interval     = 60 * HZ,
 225 #ifdef CONFIG_IPV6_ROUTE_INFO
 226         .accept_ra_rt_info_max_plen = 0,
 227 #endif
 228 #endif
 229         .proxy_ndp              = 0,
 230         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 231         .disable_ipv6           = 0,
 232         .accept_dad             = 1,
 233         .suppress_frag_ndisc    = 1,
 234         .accept_ra_mtu          = 1,
 235         .stable_secret          = {
 236                 .initialized = false,
 237         },
 238         .use_oif_addrs_only     = 0,
 239         .ignore_routes_with_linkdown = 0,
 240         .keep_addr_on_down      = 0,
 241         .seg6_enabled           = 0,
 242 #ifdef CONFIG_IPV6_SEG6_HMAC
 243         .seg6_require_hmac      = 0,
 244 #endif
 245         .enhanced_dad           = 1,
 246 };
 247
 248 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 249         .forwarding             = 0,
 250         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 251         .mtu6                   = IPV6_MIN_MTU,
 252         .accept_ra              = 1,
 253         .accept_redirects       = 1,
 254         .autoconf               = 1,
 255         .force_mld_version      = 0,
 256         .mldv1_unsolicited_report_interval = 10 * HZ,
 257         .mldv2_unsolicited_report_interval = HZ,
 258         .dad_transmits          = 1,
 259         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 260         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 261         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 262         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 263         .use_tempaddr           = 0,
 264         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 265         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 266         .regen_max_retry        = REGEN_MAX_RETRY,
 267         .max_desync_factor      = MAX_DESYNC_FACTOR,
 268         .max_addresses          = IPV6_MAX_ADDRESSES,
 269         .accept_ra_defrtr       = 1,
 270         .accept_ra_from_local   = 0,
 271         .accept_ra_min_hop_limit= 1,
 272         .accept_ra_pinfo        = 1,
 273 #ifdef CONFIG_IPV6_ROUTER_PREF
 274         .accept_ra_rtr_pref     = 1,
 275         .rtr_probe_interval     = 60 * HZ,
 276 #ifdef CONFIG_IPV6_ROUTE_INFO
 277         .accept_ra_rt_info_max_plen = 0,
 278 #endif
 279 #endif
 280         .proxy_ndp              = 0,
 281         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 282         .disable_ipv6           = 0,
 283         .accept_dad             = 1,
 284         .suppress_frag_ndisc    = 1,
 285         .accept_ra_mtu          = 1,
 286         .stable_secret          = {
 287                 .initialized = false,
 288         },
 289         .use_oif_addrs_only     = 0,
 290         .ignore_routes_with_linkdown = 0,
 291         .keep_addr_on_down      = 0,
 292         .seg6_enabled           = 0,
 293 #ifdef CONFIG_IPV6_SEG6_HMAC
 294         .seg6_require_hmac      = 0,
 295 #endif
 296         .enhanced_dad           = 1,
 297 };
 298
 299 /* Check if a valid qdisc is available */
 300 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 301 {
 302         return !qdisc_tx_is_noop(dev);
 303 }
 304
 305 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 306 {
 307         if (del_timer(&idev->rs_timer))
 308                 __in6_dev_put(idev);
 309 }
 310
 311 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 312 {
 313         if (cancel_delayed_work(&ifp->dad_work))
 314                 __in6_ifa_put(ifp);
 315 }
 316
 317 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 318                                   unsigned long when)
 319 {
 320         if (!timer_pending(&idev->rs_timer))
 321                 in6_dev_hold(idev);
 322         mod_timer(&idev->rs_timer, jiffies + when);
 323 }
 324
 325 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 326                                    unsigned long delay)
 327 {
 328         if (!delayed_work_pending(&ifp->dad_work))
 329                 in6_ifa_hold(ifp);
 330         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 331 }
 332
 333 static int snmp6_alloc_dev(struct inet6_dev *idev)
 334 {
 335         int i;
 336
 337         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 338         if (!idev->stats.ipv6)
 339                 goto err_ip;
 340
 341         for_each_possible_cpu(i) {
 342                 struct ipstats_mib *addrconf_stats;
 343                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 344                 u64_stats_init(&addrconf_stats->syncp);
 345         }
 346
 347
 348         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 349                                         GFP_KERNEL);
 350         if (!idev->stats.icmpv6dev)
 351                 goto err_icmp;
 352         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 353                                            GFP_KERNEL);
 354         if (!idev->stats.icmpv6msgdev)
 355                 goto err_icmpmsg;
 356
 357         return 0;
 358
 359 err_icmpmsg:
 360         kfree(idev->stats.icmpv6dev);
 361 err_icmp:
 362         free_percpu(idev->stats.ipv6);
 363 err_ip:
 364         return -ENOMEM;
 365 }
 366
 367 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 368 {
 369         struct inet6_dev *ndev;
 370         int err = -ENOMEM;
 371
 372         ASSERT_RTNL();
 373
 374         if (dev->mtu < IPV6_MIN_MTU)
 375                 return ERR_PTR(-EINVAL);
 376
 377         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 378         if (!ndev)
 379                 return ERR_PTR(err);
 380
 381         rwlock_init(&ndev->lock);
 382         ndev->dev = dev;
 383         INIT_LIST_HEAD(&ndev->addr_list);
 384         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 385                     (unsigned long)ndev);
 386         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 387
 388         if (ndev->cnf.stable_secret.initialized)
 389                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 390         else
 391                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
 392
 393         ndev->cnf.mtu6 = dev->mtu;
 394         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 395         if (!ndev->nd_parms) {
 396                 kfree(ndev);
 397                 return ERR_PTR(err);
 398         }
 399         if (ndev->cnf.forwarding)
 400                 dev_disable_lro(dev);
 401         /* We refer to the device */
 402         dev_hold(dev);
 403
 404         if (snmp6_alloc_dev(ndev) < 0) {
 405                 ADBG(KERN_WARNING
 406                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 407                         __func__, dev->name);
 408                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 409                 dev_put(dev);
 410                 kfree(ndev);
 411                 return ERR_PTR(err);
 412         }
 413
 414         if (snmp6_register_dev(ndev) < 0) {
 415                 ADBG(KERN_WARNING
 416                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 417                         __func__, dev->name);
 418                 goto err_release;
 419         }
 420
 421         /* One reference from device. */
 422         in6_dev_hold(ndev);
 423
 424         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 425                 ndev->cnf.accept_dad = -1;
 426
 427 #if IS_ENABLED(CONFIG_IPV6_SIT)
 428         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 429                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 430                 ndev->cnf.rtr_solicits = 0;
 431         }
 432 #endif
 433
 434         INIT_LIST_HEAD(&ndev->tempaddr_list);
 435         ndev->desync_factor = U32_MAX;
 436         if ((dev->flags&IFF_LOOPBACK) ||
 437             dev->type == ARPHRD_TUNNEL ||
 438             dev->type == ARPHRD_TUNNEL6 ||
 439             dev->type == ARPHRD_SIT ||
 440             dev->type == ARPHRD_NONE) {
 441                 ndev->cnf.use_tempaddr = -1;
 442         } else
 443                 ipv6_regen_rndid(ndev);
 444
 445         ndev->token = in6addr_any;
 446
 447         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 448                 ndev->if_flags |= IF_READY;
 449
 450         ipv6_mc_init_dev(ndev);
 451         ndev->tstamp = jiffies;
 452         err = addrconf_sysctl_register(ndev);
 453         if (err) {
 454                 ipv6_mc_destroy_dev(ndev);
 455                 snmp6_unregister_dev(ndev);
 456                 goto err_release;
 457         }
 458         /* protected by rtnl_lock */
 459         rcu_assign_pointer(dev->ip6_ptr, ndev);
 460
 461         /* Join interface-local all-node multicast group */
 462         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 463
 464         /* Join all-node multicast group */
 465         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 466
 467         /* Join all-router multicast group if forwarding is set */
 468         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 469                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 470
 471         return ndev;
 472
 473 err_release:
 474         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 475         ndev->dead = 1;
 476         in6_dev_finish_destroy(ndev);
 477         return ERR_PTR(err);
 478 }
 479
 480 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 481 {
 482         struct inet6_dev *idev;
 483
 484         ASSERT_RTNL();
 485
 486         idev = __in6_dev_get(dev);
 487         if (!idev) {
 488                 idev = ipv6_add_dev(dev);
 489                 if (IS_ERR(idev))
 490                         return NULL;
 491         }
 492
 493         if (dev->flags&IFF_UP)
 494                 ipv6_mc_up(idev);
 495         return idev;
 496 }
 497
 498 static int inet6_netconf_msgsize_devconf(int type)
 499 {
 500         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 501                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 502         bool all = false;
 503
 504         if (type == NETCONFA_ALL)
 505                 all = true;
 506
 507         if (all || type == NETCONFA_FORWARDING)
 508                 size += nla_total_size(4);
 509 #ifdef CONFIG_IPV6_MROUTE
 510         if (all || type == NETCONFA_MC_FORWARDING)
 511                 size += nla_total_size(4);
 512 #endif
 513         if (all || type == NETCONFA_PROXY_NEIGH)
 514                 size += nla_total_size(4);
 515
 516         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 517                 size += nla_total_size(4);
 518
 519         return size;
 520 }
 521
 522 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 523                                       struct ipv6_devconf *devconf, u32 portid,
 524                                       u32 seq, int event, unsigned int flags,
 525                                       int type)
 526 {
 527         struct nlmsghdr  *nlh;
 528         struct netconfmsg *ncm;
 529         bool all = false;
 530
 531         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 532                         flags);
 533         if (!nlh)
 534                 return -EMSGSIZE;
 535
 536         if (type == NETCONFA_ALL)
 537                 all = true;
 538
 539         ncm = nlmsg_data(nlh);
 540         ncm->ncm_family = AF_INET6;
 541
 542         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 543                 goto nla_put_failure;
 544
 545         if ((all || type == NETCONFA_FORWARDING) &&
 546             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 547                 goto nla_put_failure;
 548 #ifdef CONFIG_IPV6_MROUTE
 549         if ((all || type == NETCONFA_MC_FORWARDING) &&
 550             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 551                         devconf->mc_forwarding) < 0)
 552                 goto nla_put_failure;
 553 #endif
 554         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 555             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 556                 goto nla_put_failure;
 557
 558         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 559             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 560                         devconf->ignore_routes_with_linkdown) < 0)
 561                 goto nla_put_failure;
 562
 563         nlmsg_end(skb, nlh);
 564         return 0;
 565
 566 nla_put_failure:
 567         nlmsg_cancel(skb, nlh);
 568         return -EMSGSIZE;
 569 }
 570
 571 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 572                                   struct ipv6_devconf *devconf)
 573 {
 574         struct sk_buff *skb;
 575         int err = -ENOBUFS;
 576
 577         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 578         if (!skb)
 579                 goto errout;
 580
 581         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 582                                          RTM_NEWNETCONF, 0, type);
 583         if (err < 0) {
 584                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 585                 WARN_ON(err == -EMSGSIZE);
 586                 kfree_skb(skb);
 587                 goto errout;
 588         }
 589         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 590         return;
 591 errout:
 592         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 593 }
 594
 595 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 596         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 597         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 598         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 599         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 600 };
 601
 602 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 603                                      struct nlmsghdr *nlh)
 604 {
 605         struct net *net = sock_net(in_skb->sk);
 606         struct nlattr *tb[NETCONFA_MAX+1];
 607         struct netconfmsg *ncm;
 608         struct sk_buff *skb;
 609         struct ipv6_devconf *devconf;
 610         struct inet6_dev *in6_dev;
 611         struct net_device *dev;
 612         int ifindex;
 613         int err;
 614
 615         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 616                           devconf_ipv6_policy);
 617         if (err < 0)
 618                 goto errout;
 619
 620         err = -EINVAL;
 621         if (!tb[NETCONFA_IFINDEX])
 622                 goto errout;
 623
 624         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 625         switch (ifindex) {
 626         case NETCONFA_IFINDEX_ALL:
 627                 devconf = net->ipv6.devconf_all;
 628                 break;
 629         case NETCONFA_IFINDEX_DEFAULT:
 630                 devconf = net->ipv6.devconf_dflt;
 631                 break;
 632         default:
 633                 dev = __dev_get_by_index(net, ifindex);
 634                 if (!dev)
 635                         goto errout;
 636                 in6_dev = __in6_dev_get(dev);
 637                 if (!in6_dev)
 638                         goto errout;
 639                 devconf = &in6_dev->cnf;
 640                 break;
 641         }
 642
 643         err = -ENOBUFS;
 644         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
 645         if (!skb)
 646                 goto errout;
 647
 648         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 649                                          NETLINK_CB(in_skb).portid,
 650                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 651                                          NETCONFA_ALL);
 652         if (err < 0) {
 653                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 654                 WARN_ON(err == -EMSGSIZE);
 655                 kfree_skb(skb);
 656                 goto errout;
 657         }
 658         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 659 errout:
 660         return err;
 661 }
 662
 663 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 664                                       struct netlink_callback *cb)
 665 {
 666         struct net *net = sock_net(skb->sk);
 667         int h, s_h;
 668         int idx, s_idx;
 669         struct net_device *dev;
 670         struct inet6_dev *idev;
 671         struct hlist_head *head;
 672
 673         s_h = cb->args[0];
 674         s_idx = idx = cb->args[1];
 675
 676         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 677                 idx = 0;
 678                 head = &net->dev_index_head[h];
 679                 rcu_read_lock();
 680                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 681                           net->dev_base_seq;
 682                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 683                         if (idx < s_idx)
 684                                 goto cont;
 685                         idev = __in6_dev_get(dev);
 686                         if (!idev)
 687                                 goto cont;
 688
 689                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 690                                                        &idev->cnf,
 691                                                        NETLINK_CB(cb->skb).portid,
 692                                                        cb->nlh->nlmsg_seq,
 693                                                        RTM_NEWNETCONF,
 694                                                        NLM_F_MULTI,
 695                                                        NETCONFA_ALL) < 0) {
 696                                 rcu_read_unlock();
 697                                 goto done;
 698                         }
 699                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 700 cont:
 701                         idx++;
 702                 }
 703                 rcu_read_unlock();
 704         }
 705         if (h == NETDEV_HASHENTRIES) {
 706                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 707                                                net->ipv6.devconf_all,
 708                                                NETLINK_CB(cb->skb).portid,
 709                                                cb->nlh->nlmsg_seq,
 710                                                RTM_NEWNETCONF, NLM_F_MULTI,
 711                                                NETCONFA_ALL) < 0)
 712                         goto done;
 713                 else
 714                         h++;
 715         }
 716         if (h == NETDEV_HASHENTRIES + 1) {
 717                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 718                                                net->ipv6.devconf_dflt,
 719                                                NETLINK_CB(cb->skb).portid,
 720                                                cb->nlh->nlmsg_seq,
 721                                                RTM_NEWNETCONF, NLM_F_MULTI,
 722                                                NETCONFA_ALL) < 0)
 723                         goto done;
 724                 else
 725                         h++;
 726         }
 727 done:
 728         cb->args[0] = h;
 729         cb->args[1] = idx;
 730
 731         return skb->len;
 732 }
 733
 734 #ifdef CONFIG_SYSCTL
 735 static void dev_forward_change(struct inet6_dev *idev)
 736 {
 737         struct net_device *dev;
 738         struct inet6_ifaddr *ifa;
 739
 740         if (!idev)
 741                 return;
 742         dev = idev->dev;
 743         if (idev->cnf.forwarding)
 744                 dev_disable_lro(dev);
 745         if (dev->flags & IFF_MULTICAST) {
 746                 if (idev->cnf.forwarding) {
 747                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 748                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 749                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 750                 } else {
 751                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 752                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 753                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 754                 }
 755         }
 756
 757         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 758                 if (ifa->flags&IFA_F_TENTATIVE)
 759                         continue;
 760                 if (idev->cnf.forwarding)
 761                         addrconf_join_anycast(ifa);
 762                 else
 763                         addrconf_leave_anycast(ifa);
 764         }
 765         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 766                                      dev->ifindex, &idev->cnf);
 767 }
 768
 769
 770 static void addrconf_forward_change(struct net *net, __s32 newf)
 771 {
 772         struct net_device *dev;
 773         struct inet6_dev *idev;
 774
 775         for_each_netdev(net, dev) {
 776                 idev = __in6_dev_get(dev);
 777                 if (idev) {
 778                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 779                         idev->cnf.forwarding = newf;
 780                         if (changed)
 781                                 dev_forward_change(idev);
 782                 }
 783         }
 784 }
 785
 786 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 787 {
 788         struct net *net;
 789         int old;
 790
 791         if (!rtnl_trylock())
 792                 return restart_syscall();
 793
 794         net = (struct net *)table->extra2;
 795         old = *p;
 796         *p = newf;
 797
 798         if (p == &net->ipv6.devconf_dflt->forwarding) {
 799                 if ((!newf) ^ (!old))
 800                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 801                                                      NETCONFA_IFINDEX_DEFAULT,
 802                                                      net->ipv6.devconf_dflt);
 803                 rtnl_unlock();
 804                 return 0;
 805         }
 806
 807         if (p == &net->ipv6.devconf_all->forwarding) {
 808                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 809
 810                 net->ipv6.devconf_dflt->forwarding = newf;
 811                 if ((!newf) ^ (!old_dflt))
 812                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 813                                                      NETCONFA_IFINDEX_DEFAULT,
 814                                                      net->ipv6.devconf_dflt);
 815
 816                 addrconf_forward_change(net, newf);
 817                 if ((!newf) ^ (!old))
 818                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 819                                                      NETCONFA_IFINDEX_ALL,
 820                                                      net->ipv6.devconf_all);
 821         } else if ((!newf) ^ (!old))
 822                 dev_forward_change((struct inet6_dev *)table->extra1);
 823         rtnl_unlock();
 824
 825         if (newf)
 826                 rt6_purge_dflt_routers(net);
 827         return 1;
 828 }
 829
 830 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 831 {
 832         struct net_device *dev;
 833         struct inet6_dev *idev;
 834
 835         for_each_netdev(net, dev) {
 836                 idev = __in6_dev_get(dev);
 837                 if (idev) {
 838                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 839
 840                         idev->cnf.ignore_routes_with_linkdown = newf;
 841                         if (changed)
 842                                 inet6_netconf_notify_devconf(dev_net(dev),
 843                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 844                                                              dev->ifindex,
 845                                                              &idev->cnf);
 846                 }
 847         }
 848 }
 849
 850 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 851 {
 852         struct net *net;
 853         int old;
 854
 855         if (!rtnl_trylock())
 856                 return restart_syscall();
 857
 858         net = (struct net *)table->extra2;
 859         old = *p;
 860         *p = newf;
 861
 862         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 863                 if ((!newf) ^ (!old))
 864                         inet6_netconf_notify_devconf(net,
 865                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 866                                                      NETCONFA_IFINDEX_DEFAULT,
 867                                                      net->ipv6.devconf_dflt);
 868                 rtnl_unlock();
 869                 return 0;
 870         }
 871
 872         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 873                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 874                 addrconf_linkdown_change(net, newf);
 875                 if ((!newf) ^ (!old))
 876                         inet6_netconf_notify_devconf(net,
 877                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 878                                                      NETCONFA_IFINDEX_ALL,
 879                                                      net->ipv6.devconf_all);
 880         }
 881         rtnl_unlock();
 882
 883         return 1;
 884 }
 885
 886 #endif
 887
 888 /* Nobody refers to this ifaddr, destroy it */
 889 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 890 {
 891         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 892
 893 #ifdef NET_REFCNT_DEBUG
 894         pr_debug("%s\n", __func__);
 895 #endif
 896
 897         in6_dev_put(ifp->idev);
 898
 899         if (cancel_delayed_work(&ifp->dad_work))
 900                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 901                           ifp);
 902
 903         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 904                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 905                 return;
 906         }
 907         ip6_rt_put(ifp->rt);
 908
 909         kfree_rcu(ifp, rcu);
 910 }
 911
 912 static void
 913 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 914 {
 915         struct list_head *p;
 916         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 917
 918         /*
 919          * Each device address list is sorted in order of scope -
 920          * global before linklocal.
 921          */
 922         list_for_each(p, &idev->addr_list) {
 923                 struct inet6_ifaddr *ifa
 924                         = list_entry(p, struct inet6_ifaddr, if_list);
 925                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 926                         break;
 927         }
 928
 929         list_add_tail(&ifp->if_list, p);
 930 }
 931
 932 static u32 inet6_addr_hash(const struct in6_addr *addr)
 933 {
 934         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 935 }
 936
 937 /* On success it returns ifp with increased reference count */
 938
 939 static struct inet6_ifaddr *
 940 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 941               const struct in6_addr *peer_addr, int pfxlen,
 942               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 943 {
 944         struct inet6_ifaddr *ifa = NULL;
 945         struct rt6_info *rt;
 946         unsigned int hash;
 947         int err = 0;
 948         int addr_type = ipv6_addr_type(addr);
 949
 950         if (addr_type == IPV6_ADDR_ANY ||
 951             addr_type & IPV6_ADDR_MULTICAST ||
 952             (!(idev->dev->flags & IFF_LOOPBACK) &&
 953              addr_type & IPV6_ADDR_LOOPBACK))
 954                 return ERR_PTR(-EADDRNOTAVAIL);
 955
 956         rcu_read_lock_bh();
 957         if (idev->dead) {
 958                 err = -ENODEV;                  /*XXX*/
 959                 goto out2;
 960         }
 961
 962         if (idev->cnf.disable_ipv6) {
 963                 err = -EACCES;
 964                 goto out2;
 965         }
 966
 967         spin_lock(&addrconf_hash_lock);
 968
 969         /* Ignore adding duplicate addresses on an interface */
 970         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 971                 ADBG("ipv6_add_addr: already assigned\n");
 972                 err = -EEXIST;
 973                 goto out;
 974         }
 975
 976         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 977
 978         if (!ifa) {
 979                 ADBG("ipv6_add_addr: malloc failed\n");
 980                 err = -ENOBUFS;
 981                 goto out;
 982         }
 983
 984         rt = addrconf_dst_alloc(idev, addr, false);
 985         if (IS_ERR(rt)) {
 986                 err = PTR_ERR(rt);
 987                 goto out;
 988         }
 989
 990         neigh_parms_data_state_setall(idev->nd_parms);
 991
 992         ifa->addr = *addr;
 993         if (peer_addr)
 994                 ifa->peer_addr = *peer_addr;
 995
 996         spin_lock_init(&ifa->lock);
 997         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 998         INIT_HLIST_NODE(&ifa->addr_lst);
 999         ifa->scope = scope;
1000         ifa->prefix_len = pfxlen;
1001         ifa->flags = flags | IFA_F_TENTATIVE;
1002         ifa->valid_lft = valid_lft;
1003         ifa->prefered_lft = prefered_lft;
1004         ifa->cstamp = ifa->tstamp = jiffies;
1005         ifa->tokenized = false;
1006
1007         ifa->rt = rt;
1008
1009         ifa->idev = idev;
1010         in6_dev_hold(idev);
1011         /* For caller */
1012         in6_ifa_hold(ifa);
1013
1014         /* Add to big hash table */
1015         hash = inet6_addr_hash(addr);
1016
1017         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1018         spin_unlock(&addrconf_hash_lock);
1019
1020         write_lock(&idev->lock);
1021         /* Add to inet6_dev unicast addr list. */
1022         ipv6_link_dev_addr(idev, ifa);
1023
1024         if (ifa->flags&IFA_F_TEMPORARY) {
1025                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1026                 in6_ifa_hold(ifa);
1027         }
1028
1029         in6_ifa_hold(ifa);
1030         write_unlock(&idev->lock);
1031 out2:
1032         rcu_read_unlock_bh();
1033
1034         if (likely(err == 0))
1035                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1036         else {
1037                 kfree(ifa);
1038                 ifa = ERR_PTR(err);
1039         }
1040
1041         return ifa;
1042 out:
1043         spin_unlock(&addrconf_hash_lock);
1044         goto out2;
1045 }
1046
1047 enum cleanup_prefix_rt_t {
1048         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1049         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1050         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1051 };
1052
1053 /*
1054  * Check, whether the prefix for ifp would still need a prefix route
1055  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1056  * constants.
1057  *
1058  * 1) we don't purge prefix if address was not permanent.
1059  *    prefix is managed by its own lifetime.
1060  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1061  * 3) if there are no addresses, delete prefix.
1062  * 4) if there are still other permanent address(es),
1063  *    corresponding prefix is still permanent.
1064  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1065  *    don't purge the prefix, assume user space is managing it.
1066  * 6) otherwise, update prefix lifetime to the
1067  *    longest valid lifetime among the corresponding
1068  *    addresses on the device.
1069  *    Note: subsequent RA will update lifetime.
1070  **/
1071 static enum cleanup_prefix_rt_t
1072 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1073 {
1074         struct inet6_ifaddr *ifa;
1075         struct inet6_dev *idev = ifp->idev;
1076         unsigned long lifetime;
1077         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1078
1079         *expires = jiffies;
1080
1081         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1082                 if (ifa == ifp)
1083                         continue;
1084                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1085                                        ifp->prefix_len))
1086                         continue;
1087                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1088                         return CLEANUP_PREFIX_RT_NOP;
1089
1090                 action = CLEANUP_PREFIX_RT_EXPIRE;
1091
1092                 spin_lock(&ifa->lock);
1093
1094                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1095                 /*
1096                  * Note: Because this address is
1097                  * not permanent, lifetime <
1098                  * LONG_MAX / HZ here.
1099                  */
1100                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1101                         *expires = ifa->tstamp + lifetime * HZ;
1102                 spin_unlock(&ifa->lock);
1103         }
1104
1105         return action;
1106 }
1107
1108 static void
1109 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1110 {
1111         struct rt6_info *rt;
1112
1113         rt = addrconf_get_prefix_route(&ifp->addr,
1114                                        ifp->prefix_len,
1115                                        ifp->idev->dev,
1116                                        0, RTF_GATEWAY | RTF_DEFAULT);
1117         if (rt) {
1118                 if (del_rt)
1119                         ip6_del_rt(rt);
1120                 else {
1121                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1122                                 rt6_set_expires(rt, expires);
1123                         ip6_rt_put(rt);
1124                 }
1125         }
1126 }
1127
1128
1129 /* This function wants to get referenced ifp and releases it before return */
1130
1131 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1132 {
1133         int state;
1134         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1135         unsigned long expires;
1136
1137         ASSERT_RTNL();
1138
1139         spin_lock_bh(&ifp->lock);
1140         state = ifp->state;
1141         ifp->state = INET6_IFADDR_STATE_DEAD;
1142         spin_unlock_bh(&ifp->lock);
1143
1144         if (state == INET6_IFADDR_STATE_DEAD)
1145                 goto out;
1146
1147         spin_lock_bh(&addrconf_hash_lock);
1148         hlist_del_init_rcu(&ifp->addr_lst);
1149         spin_unlock_bh(&addrconf_hash_lock);
1150
1151         write_lock_bh(&ifp->idev->lock);
1152
1153         if (ifp->flags&IFA_F_TEMPORARY) {
1154                 list_del(&ifp->tmp_list);
1155                 if (ifp->ifpub) {
1156                         in6_ifa_put(ifp->ifpub);
1157                         ifp->ifpub = NULL;
1158                 }
1159                 __in6_ifa_put(ifp);
1160         }
1161
1162         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1163                 action = check_cleanup_prefix_route(ifp, &expires);
1164
1165         list_del_init(&ifp->if_list);
1166         __in6_ifa_put(ifp);
1167
1168         write_unlock_bh(&ifp->idev->lock);
1169
1170         addrconf_del_dad_work(ifp);
1171
1172         ipv6_ifa_notify(RTM_DELADDR, ifp);
1173
1174         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1175
1176         if (action != CLEANUP_PREFIX_RT_NOP) {
1177                 cleanup_prefix_route(ifp, expires,
1178                         action == CLEANUP_PREFIX_RT_DEL);
1179         }
1180
1181         /* clean up prefsrc entries */
1182         rt6_remove_prefsrc(ifp);
1183 out:
1184         in6_ifa_put(ifp);
1185 }
1186
1187 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1188 {
1189         struct inet6_dev *idev = ifp->idev;
1190         struct in6_addr addr, *tmpaddr;
1191         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1192         unsigned long regen_advance;
1193         int tmp_plen;
1194         int ret = 0;
1195         u32 addr_flags;
1196         unsigned long now = jiffies;
1197         long max_desync_factor;
1198         s32 cnf_temp_preferred_lft;
1199
1200         write_lock_bh(&idev->lock);
1201         if (ift) {
1202                 spin_lock_bh(&ift->lock);
1203                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1204                 spin_unlock_bh(&ift->lock);
1205                 tmpaddr = &addr;
1206         } else {
1207                 tmpaddr = NULL;
1208         }
1209 retry:
1210         in6_dev_hold(idev);
1211         if (idev->cnf.use_tempaddr <= 0) {
1212                 write_unlock_bh(&idev->lock);
1213                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1214                 in6_dev_put(idev);
1215                 ret = -1;
1216                 goto out;
1217         }
1218         spin_lock_bh(&ifp->lock);
1219         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1220                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1221                 spin_unlock_bh(&ifp->lock);
1222                 write_unlock_bh(&idev->lock);
1223                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1224                         __func__);
1225                 in6_dev_put(idev);
1226                 ret = -1;
1227                 goto out;
1228         }
1229         in6_ifa_hold(ifp);
1230         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1231         ipv6_try_regen_rndid(idev, tmpaddr);
1232         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1233         age = (now - ifp->tstamp) / HZ;
1234
1235         regen_advance = idev->cnf.regen_max_retry *
1236                         idev->cnf.dad_transmits *
1237                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1238
1239         /* recalculate max_desync_factor each time and update
1240          * idev->desync_factor if it's larger
1241          */
1242         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1243         max_desync_factor = min_t(__u32,
1244                                   idev->cnf.max_desync_factor,
1245                                   cnf_temp_preferred_lft - regen_advance);
1246
1247         if (unlikely(idev->desync_factor > max_desync_factor)) {
1248                 if (max_desync_factor > 0) {
1249                         get_random_bytes(&idev->desync_factor,
1250                                          sizeof(idev->desync_factor));
1251                         idev->desync_factor %= max_desync_factor;
1252                 } else {
1253                         idev->desync_factor = 0;
1254                 }
1255         }
1256
1257         tmp_valid_lft = min_t(__u32,
1258                               ifp->valid_lft,
1259                               idev->cnf.temp_valid_lft + age);
1260         tmp_prefered_lft = cnf_temp_preferred_lft + age -
1261                             idev->desync_factor;
1262         tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1263         tmp_plen = ifp->prefix_len;
1264         tmp_tstamp = ifp->tstamp;
1265         spin_unlock_bh(&ifp->lock);
1266
1267         write_unlock_bh(&idev->lock);
1268
1269         /* A temporary address is created only if this calculated Preferred
1270          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1271          * an implementation must not create a temporary address with a zero
1272          * Preferred Lifetime.
1273          * Use age calculation as in addrconf_verify to avoid unnecessary
1274          * temporary addresses being generated.
1275          */
1276         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1277         if (tmp_prefered_lft <= regen_advance + age) {
1278                 in6_ifa_put(ifp);
1279                 in6_dev_put(idev);
1280                 ret = -1;
1281                 goto out;
1282         }
1283
1284         addr_flags = IFA_F_TEMPORARY;
1285         /* set in addrconf_prefix_rcv() */
1286         if (ifp->flags & IFA_F_OPTIMISTIC)
1287                 addr_flags |= IFA_F_OPTIMISTIC;
1288
1289         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1290                             ipv6_addr_scope(&addr), addr_flags,
1291                             tmp_valid_lft, tmp_prefered_lft);
1292         if (IS_ERR(ift)) {
1293                 in6_ifa_put(ifp);
1294                 in6_dev_put(idev);
1295                 pr_info("%s: retry temporary address regeneration\n", __func__);
1296                 tmpaddr = &addr;
1297                 write_lock_bh(&idev->lock);
1298                 goto retry;
1299         }
1300
1301         spin_lock_bh(&ift->lock);
1302         ift->ifpub = ifp;
1303         ift->cstamp = now;
1304         ift->tstamp = tmp_tstamp;
1305         spin_unlock_bh(&ift->lock);
1306
1307         addrconf_dad_start(ift);
1308         in6_ifa_put(ift);
1309         in6_dev_put(idev);
1310 out:
1311         return ret;
1312 }
1313
1314 /*
1315  *      Choose an appropriate source address (RFC3484)
1316  */
1317 enum {
1318         IPV6_SADDR_RULE_INIT = 0,
1319         IPV6_SADDR_RULE_LOCAL,
1320         IPV6_SADDR_RULE_SCOPE,
1321         IPV6_SADDR_RULE_PREFERRED,
1322 #ifdef CONFIG_IPV6_MIP6
1323         IPV6_SADDR_RULE_HOA,
1324 #endif
1325         IPV6_SADDR_RULE_OIF,
1326         IPV6_SADDR_RULE_LABEL,
1327         IPV6_SADDR_RULE_PRIVACY,
1328         IPV6_SADDR_RULE_ORCHID,
1329         IPV6_SADDR_RULE_PREFIX,
1330 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1331         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1332 #endif
1333         IPV6_SADDR_RULE_MAX
1334 };
1335
1336 struct ipv6_saddr_score {
1337         int                     rule;
1338         int                     addr_type;
1339         struct inet6_ifaddr     *ifa;
1340         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1341         int                     scopedist;
1342         int                     matchlen;
1343 };
1344
1345 struct ipv6_saddr_dst {
1346         const struct in6_addr *addr;
1347         int ifindex;
1348         int scope;
1349         int label;
1350         unsigned int prefs;
1351 };
1352
1353 static inline int ipv6_saddr_preferred(int type)
1354 {
1355         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1356                 return 1;
1357         return 0;
1358 }
1359
1360 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1361 {
1362 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1363         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1364 #else
1365         return false;
1366 #endif
1367 }
1368
1369 static int ipv6_get_saddr_eval(struct net *net,
1370                                struct ipv6_saddr_score *score,
1371                                struct ipv6_saddr_dst *dst,
1372                                int i)
1373 {
1374         int ret;
1375
1376         if (i <= score->rule) {
1377                 switch (i) {
1378                 case IPV6_SADDR_RULE_SCOPE:
1379                         ret = score->scopedist;
1380                         break;
1381                 case IPV6_SADDR_RULE_PREFIX:
1382                         ret = score->matchlen;
1383                         break;
1384                 default:
1385                         ret = !!test_bit(i, score->scorebits);
1386                 }
1387                 goto out;
1388         }
1389
1390         switch (i) {
1391         case IPV6_SADDR_RULE_INIT:
1392                 /* Rule 0: remember if hiscore is not ready yet */
1393                 ret = !!score->ifa;
1394                 break;
1395         case IPV6_SADDR_RULE_LOCAL:
1396                 /* Rule 1: Prefer same address */
1397                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1398                 break;
1399         case IPV6_SADDR_RULE_SCOPE:
1400                 /* Rule 2: Prefer appropriate scope
1401                  *
1402                  *      ret
1403                  *       ^
1404                  *    -1 |  d 15
1405                  *    ---+--+-+---> scope
1406                  *       |
1407                  *       |             d is scope of the destination.
1408                  *  B-d  |  \
1409                  *       |   \      <- smaller scope is better if
1410                  *  B-15 |    \        if scope is enough for destination.
1411                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1412                  * d-C-1 | /
1413                  *       |/         <- greater is better
1414                  *   -C  /             if scope is not enough for destination.
1415                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1416                  *
1417                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1418                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1419                  * Assume B = 0 and we get C > 29.
1420                  */
1421                 ret = __ipv6_addr_src_scope(score->addr_type);
1422                 if (ret >= dst->scope)
1423                         ret = -ret;
1424                 else
1425                         ret -= 128;     /* 30 is enough */
1426                 score->scopedist = ret;
1427                 break;
1428         case IPV6_SADDR_RULE_PREFERRED:
1429             {
1430                 /* Rule 3: Avoid deprecated and optimistic addresses */
1431                 u8 avoid = IFA_F_DEPRECATED;
1432
1433                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1434                         avoid |= IFA_F_OPTIMISTIC;
1435                 ret = ipv6_saddr_preferred(score->addr_type) ||
1436                       !(score->ifa->flags & avoid);
1437                 break;
1438             }
1439 #ifdef CONFIG_IPV6_MIP6
1440         case IPV6_SADDR_RULE_HOA:
1441             {
1442                 /* Rule 4: Prefer home address */
1443                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1444                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1445                 break;
1446             }
1447 #endif
1448         case IPV6_SADDR_RULE_OIF:
1449                 /* Rule 5: Prefer outgoing interface */
1450                 ret = (!dst->ifindex ||
1451                        dst->ifindex == score->ifa->idev->dev->ifindex);
1452                 break;
1453         case IPV6_SADDR_RULE_LABEL:
1454                 /* Rule 6: Prefer matching label */
1455                 ret = ipv6_addr_label(net,
1456                                       &score->ifa->addr, score->addr_type,
1457                                       score->ifa->idev->dev->ifindex) == dst->label;
1458                 break;
1459         case IPV6_SADDR_RULE_PRIVACY:
1460             {
1461                 /* Rule 7: Prefer public address
1462                  * Note: prefer temporary address if use_tempaddr >= 2
1463                  */
1464                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1465                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1466                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1467                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1468                 break;
1469             }
1470         case IPV6_SADDR_RULE_ORCHID:
1471                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1472                  *          non-ORCHID vs non-ORCHID
1473                  */
1474                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1475                         ipv6_addr_orchid(dst->addr));
1476                 break;
1477         case IPV6_SADDR_RULE_PREFIX:
1478                 /* Rule 8: Use longest matching prefix */
1479                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1480                 if (ret > score->ifa->prefix_len)
1481                         ret = score->ifa->prefix_len;
1482                 score->matchlen = ret;
1483                 break;
1484 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1485         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1486                 /* Optimistic addresses still have lower precedence than other
1487                  * preferred addresses.
1488                  */
1489                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1490                 break;
1491 #endif
1492         default:
1493                 ret = 0;
1494         }
1495
1496         if (ret)
1497                 __set_bit(i, score->scorebits);
1498         score->rule = i;
1499 out:
1500         return ret;
1501 }
1502
1503 static int __ipv6_dev_get_saddr(struct net *net,
1504                                 struct ipv6_saddr_dst *dst,
1505                                 struct inet6_dev *idev,
1506                                 struct ipv6_saddr_score *scores,
1507                                 int hiscore_idx)
1508 {
1509         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1510
1511         read_lock_bh(&idev->lock);
1512         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1513                 int i;
1514
1515                 /*
1516                  * - Tentative Address (RFC2462 section 5.4)
1517                  *  - A tentative address is not considered
1518                  *    "assigned to an interface" in the traditional
1519                  *    sense, unless it is also flagged as optimistic.
1520                  * - Candidate Source Address (section 4)
1521                  *  - In any case, anycast addresses, multicast
1522                  *    addresses, and the unspecified address MUST
1523                  *    NOT be included in a candidate set.
1524                  */
1525                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1526                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1527                         continue;
1528
1529                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1530
1531                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1532                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1533                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1534                                             idev->dev->name);
1535                         continue;
1536                 }
1537
1538                 score->rule = -1;
1539                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1540
1541                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1542                         int minihiscore, miniscore;
1543
1544                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1545                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1546
1547                         if (minihiscore > miniscore) {
1548                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1549                                     score->scopedist > 0) {
1550                                         /*
1551                                          * special case:
1552                                          * each remaining entry
1553                                          * has too small (not enough)
1554                                          * scope, because ifa entries
1555                                          * are sorted by their scope
1556                                          * values.
1557                                          */
1558                                         goto out;
1559                                 }
1560                                 break;
1561                         } else if (minihiscore < miniscore) {
1562                                 if (hiscore->ifa)
1563                                         in6_ifa_put(hiscore->ifa);
1564
1565                                 in6_ifa_hold(score->ifa);
1566
1567                                 swap(hiscore, score);
1568                                 hiscore_idx = 1 - hiscore_idx;
1569
1570                                 /* restore our iterator */
1571                                 score->ifa = hiscore->ifa;
1572
1573                                 break;
1574                         }
1575                 }
1576         }
1577 out:
1578         read_unlock_bh(&idev->lock);
1579         return hiscore_idx;
1580 }
1581
1582 static int ipv6_get_saddr_master(struct net *net,
1583                                  const struct net_device *dst_dev,
1584                                  const struct net_device *master,
1585                                  struct ipv6_saddr_dst *dst,
1586                                  struct ipv6_saddr_score *scores,
1587                                  int hiscore_idx)
1588 {
1589         struct inet6_dev *idev;
1590
1591         idev = __in6_dev_get(dst_dev);
1592         if (idev)
1593                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1594                                                    scores, hiscore_idx);
1595
1596         idev = __in6_dev_get(master);
1597         if (idev)
1598                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1599                                                    scores, hiscore_idx);
1600
1601         return hiscore_idx;
1602 }
1603
1604 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1605                        const struct in6_addr *daddr, unsigned int prefs,
1606                        struct in6_addr *saddr)
1607 {
1608         struct ipv6_saddr_score scores[2], *hiscore;
1609         struct ipv6_saddr_dst dst;
1610         struct inet6_dev *idev;
1611         struct net_device *dev;
1612         int dst_type;
1613         bool use_oif_addr = false;
1614         int hiscore_idx = 0;
1615
1616         dst_type = __ipv6_addr_type(daddr);
1617         dst.addr = daddr;
1618         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1619         dst.scope = __ipv6_addr_src_scope(dst_type);
1620         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1621         dst.prefs = prefs;
1622
1623         scores[hiscore_idx].rule = -1;
1624         scores[hiscore_idx].ifa = NULL;
1625
1626         rcu_read_lock();
1627
1628         /* Candidate Source Address (section 4)
1629          *  - multicast and link-local destination address,
1630          *    the set of candidate source address MUST only
1631          *    include addresses assigned to interfaces
1632          *    belonging to the same link as the outgoing
1633          *    interface.
1634          * (- For site-local destination addresses, the
1635          *    set of candidate source addresses MUST only
1636          *    include addresses assigned to interfaces
1637          *    belonging to the same site as the outgoing
1638          *    interface.)
1639          *  - "It is RECOMMENDED that the candidate source addresses
1640          *    be the set of unicast addresses assigned to the
1641          *    interface that will be used to send to the destination
1642          *    (the 'outgoing' interface)." (RFC 6724)
1643          */
1644         if (dst_dev) {
1645                 idev = __in6_dev_get(dst_dev);
1646                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1647                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1648                     (idev && idev->cnf.use_oif_addrs_only)) {
1649                         use_oif_addr = true;
1650                 }
1651         }
1652
1653         if (use_oif_addr) {
1654                 if (idev)
1655                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1656         } else {
1657                 const struct net_device *master;
1658                 int master_idx = 0;
1659
1660                 /* if dst_dev exists and is enslaved to an L3 device, then
1661                  * prefer addresses from dst_dev and then the master over
1662                  * any other enslaved devices in the L3 domain.
1663                  */
1664                 master = l3mdev_master_dev_rcu(dst_dev);
1665                 if (master) {
1666                         master_idx = master->ifindex;
1667
1668                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1669                                                             master, &dst,
1670                                                             scores, hiscore_idx);
1671
1672                         if (scores[hiscore_idx].ifa)
1673                                 goto out;
1674                 }
1675
1676                 for_each_netdev_rcu(net, dev) {
1677                         /* only consider addresses on devices in the
1678                          * same L3 domain
1679                          */
1680                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1681                                 continue;
1682                         idev = __in6_dev_get(dev);
1683                         if (!idev)
1684                                 continue;
1685                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1686                 }
1687         }
1688
1689 out:
1690         rcu_read_unlock();
1691
1692         hiscore = &scores[hiscore_idx];
1693         if (!hiscore->ifa)
1694                 return -EADDRNOTAVAIL;
1695
1696         *saddr = hiscore->ifa->addr;
1697         in6_ifa_put(hiscore->ifa);
1698         return 0;
1699 }
1700 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1701
1702 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1703                       u32 banned_flags)
1704 {
1705         struct inet6_ifaddr *ifp;
1706         int err = -EADDRNOTAVAIL;
1707
1708         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1709                 if (ifp->scope > IFA_LINK)
1710                         break;
1711                 if (ifp->scope == IFA_LINK &&
1712                     !(ifp->flags & banned_flags)) {
1713                         *addr = ifp->addr;
1714                         err = 0;
1715                         break;
1716                 }
1717         }
1718         return err;
1719 }
1720
1721 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1722                     u32 banned_flags)
1723 {
1724         struct inet6_dev *idev;
1725         int err = -EADDRNOTAVAIL;
1726
1727         rcu_read_lock();
1728         idev = __in6_dev_get(dev);
1729         if (idev) {
1730                 read_lock_bh(&idev->lock);
1731                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1732                 read_unlock_bh(&idev->lock);
1733         }
1734         rcu_read_unlock();
1735         return err;
1736 }
1737
1738 static int ipv6_count_addresses(struct inet6_dev *idev)
1739 {
1740         int cnt = 0;
1741         struct inet6_ifaddr *ifp;
1742
1743         read_lock_bh(&idev->lock);
1744         list_for_each_entry(ifp, &idev->addr_list, if_list)
1745                 cnt++;
1746         read_unlock_bh(&idev->lock);
1747         return cnt;
1748 }
1749
1750 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1751                   const struct net_device *dev, int strict)
1752 {
1753         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1754 }
1755 EXPORT_SYMBOL(ipv6_chk_addr);
1756
1757 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1758                             const struct net_device *dev, int strict,
1759                             u32 banned_flags)
1760 {
1761         struct inet6_ifaddr *ifp;
1762         unsigned int hash = inet6_addr_hash(addr);
1763         u32 ifp_flags;
1764
1765         rcu_read_lock_bh();
1766         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1767                 if (!net_eq(dev_net(ifp->idev->dev), net))
1768                         continue;
1769                 /* Decouple optimistic from tentative for evaluation here.
1770                  * Ban optimistic addresses explicitly, when required.
1771                  */
1772                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1773                             ? (ifp->flags&~IFA_F_TENTATIVE)
1774                             : ifp->flags;
1775                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1776                     !(ifp_flags&banned_flags) &&
1777                     (!dev || ifp->idev->dev == dev ||
1778                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1779                         rcu_read_unlock_bh();
1780                         return 1;
1781                 }
1782         }
1783
1784         rcu_read_unlock_bh();
1785         return 0;
1786 }
1787 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1788
1789 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1790                                struct net_device *dev)
1791 {
1792         unsigned int hash = inet6_addr_hash(addr);
1793         struct inet6_ifaddr *ifp;
1794
1795         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1796                 if (!net_eq(dev_net(ifp->idev->dev), net))
1797                         continue;
1798                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1799                         if (!dev || ifp->idev->dev == dev)
1800                                 return true;
1801                 }
1802         }
1803         return false;
1804 }
1805
1806 /* Compares an address/prefix_len with addresses on device @dev.
1807  * If one is found it returns true.
1808  */
1809 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1810         const unsigned int prefix_len, struct net_device *dev)
1811 {
1812         struct inet6_dev *idev;
1813         struct inet6_ifaddr *ifa;
1814         bool ret = false;
1815
1816         rcu_read_lock();
1817         idev = __in6_dev_get(dev);
1818         if (idev) {
1819                 read_lock_bh(&idev->lock);
1820                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1821                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1822                         if (ret)
1823                                 break;
1824                 }
1825                 read_unlock_bh(&idev->lock);
1826         }
1827         rcu_read_unlock();
1828
1829         return ret;
1830 }
1831 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1832
1833 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1834 {
1835         struct inet6_dev *idev;
1836         struct inet6_ifaddr *ifa;
1837         int     onlink;
1838
1839         onlink = 0;
1840         rcu_read_lock();
1841         idev = __in6_dev_get(dev);
1842         if (idev) {
1843                 read_lock_bh(&idev->lock);
1844                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1845                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1846                                                    ifa->prefix_len);
1847                         if (onlink)
1848                                 break;
1849                 }
1850                 read_unlock_bh(&idev->lock);
1851         }
1852         rcu_read_unlock();
1853         return onlink;
1854 }
1855 EXPORT_SYMBOL(ipv6_chk_prefix);
1856
1857 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1858                                      struct net_device *dev, int strict)
1859 {
1860         struct inet6_ifaddr *ifp, *result = NULL;
1861         unsigned int hash = inet6_addr_hash(addr);
1862
1863         rcu_read_lock_bh();
1864         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1865                 if (!net_eq(dev_net(ifp->idev->dev), net))
1866                         continue;
1867                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1868                         if (!dev || ifp->idev->dev == dev ||
1869                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1870                                 result = ifp;
1871                                 in6_ifa_hold(ifp);
1872                                 break;
1873                         }
1874                 }
1875         }
1876         rcu_read_unlock_bh();
1877
1878         return result;
1879 }
1880
1881 /* Gets referenced address, destroys ifaddr */
1882
1883 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1884 {
1885         if (dad_failed)
1886                 ifp->flags |= IFA_F_DADFAILED;
1887
1888         if (ifp->flags&IFA_F_PERMANENT) {
1889                 spin_lock_bh(&ifp->lock);
1890                 addrconf_del_dad_work(ifp);
1891                 ifp->flags |= IFA_F_TENTATIVE;
1892                 spin_unlock_bh(&ifp->lock);
1893                 if (dad_failed)
1894                         ipv6_ifa_notify(0, ifp);
1895                 in6_ifa_put(ifp);
1896         } else if (ifp->flags&IFA_F_TEMPORARY) {
1897                 struct inet6_ifaddr *ifpub;
1898                 spin_lock_bh(&ifp->lock);
1899                 ifpub = ifp->ifpub;
1900                 if (ifpub) {
1901                         in6_ifa_hold(ifpub);
1902                         spin_unlock_bh(&ifp->lock);
1903                         ipv6_create_tempaddr(ifpub, ifp);
1904                         in6_ifa_put(ifpub);
1905                 } else {
1906                         spin_unlock_bh(&ifp->lock);
1907                 }
1908                 ipv6_del_addr(ifp);
1909         } else {
1910                 ipv6_del_addr(ifp);
1911         }
1912 }
1913
1914 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1915 {
1916         int err = -ENOENT;
1917
1918         spin_lock_bh(&ifp->lock);
1919         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1920                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1921                 err = 0;
1922         }
1923         spin_unlock_bh(&ifp->lock);
1924
1925         return err;
1926 }
1927
1928 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1929 {
1930         struct inet6_dev *idev = ifp->idev;
1931         struct net *net = dev_net(ifp->idev->dev);
1932
1933         if (addrconf_dad_end(ifp)) {
1934                 in6_ifa_put(ifp);
1935                 return;
1936         }
1937
1938         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1939                              ifp->idev->dev->name, &ifp->addr);
1940
1941         spin_lock_bh(&ifp->lock);
1942
1943         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1944                 int scope = ifp->scope;
1945                 u32 flags = ifp->flags;
1946                 struct in6_addr new_addr;
1947                 struct inet6_ifaddr *ifp2;
1948                 u32 valid_lft, preferred_lft;
1949                 int pfxlen = ifp->prefix_len;
1950                 int retries = ifp->stable_privacy_retry + 1;
1951
1952                 if (retries > net->ipv6.sysctl.idgen_retries) {
1953                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1954                                              ifp->idev->dev->name);
1955                         goto errdad;
1956                 }
1957
1958                 new_addr = ifp->addr;
1959                 if (ipv6_generate_stable_address(&new_addr, retries,
1960                                                  idev))
1961                         goto errdad;
1962
1963                 valid_lft = ifp->valid_lft;
1964                 preferred_lft = ifp->prefered_lft;
1965
1966                 spin_unlock_bh(&ifp->lock);
1967
1968                 if (idev->cnf.max_addresses &&
1969                     ipv6_count_addresses(idev) >=
1970                     idev->cnf.max_addresses)
1971                         goto lock_errdad;
1972
1973                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1974                                      ifp->idev->dev->name);
1975
1976                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1977                                      scope, flags, valid_lft,
1978                                      preferred_lft);
1979                 if (IS_ERR(ifp2))
1980                         goto lock_errdad;
1981
1982                 spin_lock_bh(&ifp2->lock);
1983                 ifp2->stable_privacy_retry = retries;
1984                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1985                 spin_unlock_bh(&ifp2->lock);
1986
1987                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1988                 in6_ifa_put(ifp2);
1989 lock_errdad:
1990                 spin_lock_bh(&ifp->lock);
1991         }
1992
1993 errdad:
1994         /* transition from _POSTDAD to _ERRDAD */
1995         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1996         spin_unlock_bh(&ifp->lock);
1997
1998         addrconf_mod_dad_work(ifp, 0);
1999         in6_ifa_put(ifp);
2000 }
2001
2002 /* Join to solicited addr multicast group.
2003  * caller must hold RTNL */
2004 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2005 {
2006         struct in6_addr maddr;
2007
2008         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2009                 return;
2010
2011         addrconf_addr_solict_mult(addr, &maddr);
2012         ipv6_dev_mc_inc(dev, &maddr);
2013 }
2014
2015 /* caller must hold RTNL */
2016 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2017 {
2018         struct in6_addr maddr;
2019
2020         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2021                 return;
2022
2023         addrconf_addr_solict_mult(addr, &maddr);
2024         __ipv6_dev_mc_dec(idev, &maddr);
2025 }
2026
2027 /* caller must hold RTNL */
2028 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2029 {
2030         struct in6_addr addr;
2031
2032         if (ifp->prefix_len >= 127) /* RFC 6164 */
2033                 return;
2034         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2035         if (ipv6_addr_any(&addr))
2036                 return;
2037         __ipv6_dev_ac_inc(ifp->idev, &addr);
2038 }
2039
2040 /* caller must hold RTNL */
2041 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2042 {
2043         struct in6_addr addr;
2044
2045         if (ifp->prefix_len >= 127) /* RFC 6164 */
2046                 return;
2047         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2048         if (ipv6_addr_any(&addr))
2049                 return;
2050         __ipv6_dev_ac_dec(ifp->idev, &addr);
2051 }
2052
2053 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2054 {
2055         if (dev->addr_len != EUI64_ADDR_LEN)
2056                 return -1;
2057         memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2058         eui[0] ^= 2;
2059         return 0;
2060 }
2061
2062 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2063 {
2064         union fwnet_hwaddr *ha;
2065
2066         if (dev->addr_len != FWNET_ALEN)
2067                 return -1;
2068
2069         ha = (union fwnet_hwaddr *)dev->dev_addr;
2070
2071         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2072         eui[0] ^= 2;
2073         return 0;
2074 }
2075
2076 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2077 {
2078         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2079         if (dev->addr_len != ARCNET_ALEN)
2080                 return -1;
2081         memset(eui, 0, 7);
2082         eui[7] = *(u8 *)dev->dev_addr;
2083         return 0;
2084 }
2085
2086 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2087 {
2088         if (dev->addr_len != INFINIBAND_ALEN)
2089                 return -1;
2090         memcpy(eui, dev->dev_addr + 12, 8);
2091         eui[0] |= 2;
2092         return 0;
2093 }
2094
2095 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2096 {
2097         if (addr == 0)
2098                 return -1;
2099         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2100                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2101                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2102                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2103                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2104                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2105         eui[1] = 0;
2106         eui[2] = 0x5E;
2107         eui[3] = 0xFE;
2108         memcpy(eui + 4, &addr, 4);
2109         return 0;
2110 }
2111
2112 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2113 {
2114         if (dev->priv_flags & IFF_ISATAP)
2115                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2116         return -1;
2117 }
2118
2119 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2120 {
2121         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2122 }
2123
2124 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2125 {
2126         memcpy(eui, dev->perm_addr, 3);
2127         memcpy(eui + 5, dev->perm_addr + 3, 3);
2128         eui[3] = 0xFF;
2129         eui[4] = 0xFE;
2130         eui[0] ^= 2;
2131         return 0;
2132 }
2133
2134 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2135 {
2136         switch (dev->type) {
2137         case ARPHRD_ETHER:
2138         case ARPHRD_FDDI:
2139                 return addrconf_ifid_eui48(eui, dev);
2140         case ARPHRD_ARCNET:
2141                 return addrconf_ifid_arcnet(eui, dev);
2142         case ARPHRD_INFINIBAND:
2143                 return addrconf_ifid_infiniband(eui, dev);
2144         case ARPHRD_SIT:
2145                 return addrconf_ifid_sit(eui, dev);
2146         case ARPHRD_IPGRE:
2147                 return addrconf_ifid_gre(eui, dev);
2148         case ARPHRD_6LOWPAN:
2149                 return addrconf_ifid_eui64(eui, dev);
2150         case ARPHRD_IEEE1394:
2151                 return addrconf_ifid_ieee1394(eui, dev);
2152         case ARPHRD_TUNNEL6:
2153                 return addrconf_ifid_ip6tnl(eui, dev);
2154         }
2155         return -1;
2156 }
2157
2158 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2159 {
2160         int err = -1;
2161         struct inet6_ifaddr *ifp;
2162
2163         read_lock_bh(&idev->lock);
2164         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2165                 if (ifp->scope > IFA_LINK)
2166                         break;
2167                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2168                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2169                         err = 0;
2170                         break;
2171                 }
2172         }
2173         read_unlock_bh(&idev->lock);
2174         return err;
2175 }
2176
2177 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2178 static void ipv6_regen_rndid(struct inet6_dev *idev)
2179 {
2180 regen:
2181         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2182         idev->rndid[0] &= ~0x02;
2183
2184         /*
2185          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2186          * check if generated address is not inappropriate
2187          *
2188          *  - Reserved subnet anycast (RFC 2526)
2189          *      11111101 11....11 1xxxxxxx
2190          *  - ISATAP (RFC4214) 6.1
2191          *      00-00-5E-FE-xx-xx-xx-xx
2192          *  - value 0
2193          *  - XXX: already assigned to an address on the device
2194          */
2195         if (idev->rndid[0] == 0xfd &&
2196             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2197             (idev->rndid[7]&0x80))
2198                 goto regen;
2199         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2200                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2201                         goto regen;
2202                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2203                         goto regen;
2204         }
2205 }
2206
2207 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2208 {
2209         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2210                 ipv6_regen_rndid(idev);
2211 }
2212
2213 /*
2214  *      Add prefix route.
2215  */
2216
2217 static void
2218 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2219                       unsigned long expires, u32 flags)
2220 {
2221         struct fib6_config cfg = {
2222                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2223                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2224                 .fc_ifindex = dev->ifindex,
2225                 .fc_expires = expires,
2226                 .fc_dst_len = plen,
2227                 .fc_flags = RTF_UP | flags,
2228                 .fc_nlinfo.nl_net = dev_net(dev),
2229                 .fc_protocol = RTPROT_KERNEL,
2230         };
2231
2232         cfg.fc_dst = *pfx;
2233
2234         /* Prevent useless cloning on PtP SIT.
2235            This thing is done here expecting that the whole
2236            class of non-broadcast devices need not cloning.
2237          */
2238 #if IS_ENABLED(CONFIG_IPV6_SIT)
2239         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2240                 cfg.fc_flags |= RTF_NONEXTHOP;
2241 #endif
2242
2243         ip6_route_add(&cfg);
2244 }
2245
2246
2247 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2248                                                   int plen,
2249                                                   const struct net_device *dev,
2250                                                   u32 flags, u32 noflags)
2251 {
2252         struct fib6_node *fn;
2253         struct rt6_info *rt = NULL;
2254         struct fib6_table *table;
2255         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2256
2257         table = fib6_get_table(dev_net(dev), tb_id);
2258         if (!table)
2259                 return NULL;
2260
2261         read_lock_bh(&table->tb6_lock);
2262         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2263         if (!fn)
2264                 goto out;
2265
2266         noflags |= RTF_CACHE;
2267         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2268                 if (rt->dst.dev->ifindex != dev->ifindex)
2269                         continue;
2270                 if ((rt->rt6i_flags & flags) != flags)
2271                         continue;
2272                 if ((rt->rt6i_flags & noflags) != 0)
2273                         continue;
2274                 dst_hold(&rt->dst);
2275                 break;
2276         }
2277 out:
2278         read_unlock_bh(&table->tb6_lock);
2279         return rt;
2280 }
2281
2282
2283 /* Create "default" multicast route to the interface */
2284
2285 static void addrconf_add_mroute(struct net_device *dev)
2286 {
2287         struct fib6_config cfg = {
2288                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2289                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2290                 .fc_ifindex = dev->ifindex,
2291                 .fc_dst_len = 8,
2292                 .fc_flags = RTF_UP,
2293                 .fc_nlinfo.nl_net = dev_net(dev),
2294         };
2295
2296         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2297
2298         ip6_route_add(&cfg);
2299 }
2300
2301 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2302 {
2303         struct inet6_dev *idev;
2304
2305         ASSERT_RTNL();
2306
2307         idev = ipv6_find_idev(dev);
2308         if (!idev)
2309                 return ERR_PTR(-ENOBUFS);
2310
2311         if (idev->cnf.disable_ipv6)
2312                 return ERR_PTR(-EACCES);
2313
2314         /* Add default multicast route */
2315         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2316                 addrconf_add_mroute(dev);
2317
2318         return idev;
2319 }
2320
2321 static void manage_tempaddrs(struct inet6_dev *idev,
2322                              struct inet6_ifaddr *ifp,
2323                              __u32 valid_lft, __u32 prefered_lft,
2324                              bool create, unsigned long now)
2325 {
2326         u32 flags;
2327         struct inet6_ifaddr *ift;
2328
2329         read_lock_bh(&idev->lock);
2330         /* update all temporary addresses in the list */
2331         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2332                 int age, max_valid, max_prefered;
2333
2334                 if (ifp != ift->ifpub)
2335                         continue;
2336
2337                 /* RFC 4941 section 3.3:
2338                  * If a received option will extend the lifetime of a public
2339                  * address, the lifetimes of temporary addresses should
2340                  * be extended, subject to the overall constraint that no
2341                  * temporary addresses should ever remain "valid" or "preferred"
2342                  * for a time longer than (TEMP_VALID_LIFETIME) or
2343                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2344                  */
2345                 age = (now - ift->cstamp) / HZ;
2346                 max_valid = idev->cnf.temp_valid_lft - age;
2347                 if (max_valid < 0)
2348                         max_valid = 0;
2349
2350                 max_prefered = idev->cnf.temp_prefered_lft -
2351                                idev->desync_factor - age;
2352                 if (max_prefered < 0)
2353                         max_prefered = 0;
2354
2355                 if (valid_lft > max_valid)
2356                         valid_lft = max_valid;
2357
2358                 if (prefered_lft > max_prefered)
2359                         prefered_lft = max_prefered;
2360
2361                 spin_lock(&ift->lock);
2362                 flags = ift->flags;
2363                 ift->valid_lft = valid_lft;
2364                 ift->prefered_lft = prefered_lft;
2365                 ift->tstamp = now;
2366                 if (prefered_lft > 0)
2367                         ift->flags &= ~IFA_F_DEPRECATED;
2368
2369                 spin_unlock(&ift->lock);
2370                 if (!(flags&IFA_F_TENTATIVE))
2371                         ipv6_ifa_notify(0, ift);
2372         }
2373
2374         if ((create || list_empty(&idev->tempaddr_list)) &&
2375             idev->cnf.use_tempaddr > 0) {
2376                 /* When a new public address is created as described
2377                  * in [ADDRCONF], also create a new temporary address.
2378                  * Also create a temporary address if it's enabled but
2379                  * no temporary address currently exists.
2380                  */
2381                 read_unlock_bh(&idev->lock);
2382                 ipv6_create_tempaddr(ifp, NULL);
2383         } else {
2384                 read_unlock_bh(&idev->lock);
2385         }
2386 }
2387
2388 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2389 {
2390         return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2391                idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2392 }
2393
2394 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2395                                  const struct prefix_info *pinfo,
2396                                  struct inet6_dev *in6_dev,
2397                                  const struct in6_addr *addr, int addr_type,
2398                                  u32 addr_flags, bool sllao, bool tokenized,
2399                                  __u32 valid_lft, u32 prefered_lft)
2400 {
2401         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2402         int create = 0, update_lft = 0;
2403
2404         if (!ifp && valid_lft) {
2405                 int max_addresses = in6_dev->cnf.max_addresses;
2406
2407 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2408                 if (in6_dev->cnf.optimistic_dad &&
2409                     !net->ipv6.devconf_all->forwarding && sllao)
2410                         addr_flags |= IFA_F_OPTIMISTIC;
2411 #endif
2412
2413                 /* Do not allow to create too much of autoconfigured
2414                  * addresses; this would be too easy way to crash kernel.
2415                  */
2416                 if (!max_addresses ||
2417                     ipv6_count_addresses(in6_dev) < max_addresses)
2418                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2419                                             pinfo->prefix_len,
2420                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2421                                             addr_flags, valid_lft,
2422                                             prefered_lft);
2423
2424                 if (IS_ERR_OR_NULL(ifp))
2425                         return -1;
2426
2427                 update_lft = 0;
2428                 create = 1;
2429                 spin_lock_bh(&ifp->lock);
2430                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2431                 ifp->cstamp = jiffies;
2432                 ifp->tokenized = tokenized;
2433                 spin_unlock_bh(&ifp->lock);
2434                 addrconf_dad_start(ifp);
2435         }
2436
2437         if (ifp) {
2438                 u32 flags;
2439                 unsigned long now;
2440                 u32 stored_lft;
2441
2442                 /* update lifetime (RFC2462 5.5.3 e) */
2443                 spin_lock_bh(&ifp->lock);
2444                 now = jiffies;
2445                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2446                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2447                 else
2448                         stored_lft = 0;
2449                 if (!update_lft && !create && stored_lft) {
2450                         const u32 minimum_lft = min_t(u32,
2451                                 stored_lft, MIN_VALID_LIFETIME);
2452                         valid_lft = max(valid_lft, minimum_lft);
2453
2454                         /* RFC4862 Section 5.5.3e:
2455                          * "Note that the preferred lifetime of the
2456                          *  corresponding address is always reset to
2457                          *  the Preferred Lifetime in the received
2458                          *  Prefix Information option, regardless of
2459                          *  whether the valid lifetime is also reset or
2460                          *  ignored."
2461                          *
2462                          * So we should always update prefered_lft here.
2463                          */
2464                         update_lft = 1;
2465                 }
2466
2467                 if (update_lft) {
2468                         ifp->valid_lft = valid_lft;
2469                         ifp->prefered_lft = prefered_lft;
2470                         ifp->tstamp = now;
2471                         flags = ifp->flags;
2472                         ifp->flags &= ~IFA_F_DEPRECATED;
2473                         spin_unlock_bh(&ifp->lock);
2474
2475                         if (!(flags&IFA_F_TENTATIVE))
2476                                 ipv6_ifa_notify(0, ifp);
2477                 } else
2478                         spin_unlock_bh(&ifp->lock);
2479
2480                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2481                                  create, now);
2482
2483                 in6_ifa_put(ifp);
2484                 addrconf_verify();
2485         }
2486
2487         return 0;
2488 }
2489 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2490
2491 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2492 {
2493         struct prefix_info *pinfo;
2494         __u32 valid_lft;
2495         __u32 prefered_lft;
2496         int addr_type, err;
2497         u32 addr_flags = 0;
2498         struct inet6_dev *in6_dev;
2499         struct net *net = dev_net(dev);
2500
2501         pinfo = (struct prefix_info *) opt;
2502
2503         if (len < sizeof(struct prefix_info)) {
2504                 ADBG("addrconf: prefix option too short\n");
2505                 return;
2506         }
2507
2508         /*
2509          *      Validation checks ([ADDRCONF], page 19)
2510          */
2511
2512         addr_type = ipv6_addr_type(&pinfo->prefix);
2513
2514         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2515                 return;
2516
2517         valid_lft = ntohl(pinfo->valid);
2518         prefered_lft = ntohl(pinfo->prefered);
2519
2520         if (prefered_lft > valid_lft) {
2521                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2522                 return;
2523         }
2524
2525         in6_dev = in6_dev_get(dev);
2526
2527         if (!in6_dev) {
2528                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2529                                     dev->name);
2530                 return;
2531         }
2532
2533         /*
2534          *      Two things going on here:
2535          *      1) Add routes for on-link prefixes
2536          *      2) Configure prefixes with the auto flag set
2537          */
2538
2539         if (pinfo->onlink) {
2540                 struct rt6_info *rt;
2541                 unsigned long rt_expires;
2542
2543                 /* Avoid arithmetic overflow. Really, we could
2544                  * save rt_expires in seconds, likely valid_lft,
2545                  * but it would require division in fib gc, that it
2546                  * not good.
2547                  */
2548                 if (HZ > USER_HZ)
2549                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2550                 else
2551                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2552
2553                 if (addrconf_finite_timeout(rt_expires))
2554                         rt_expires *= HZ;
2555
2556                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2557                                                pinfo->prefix_len,
2558                                                dev,
2559                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2560                                                RTF_GATEWAY | RTF_DEFAULT);
2561
2562                 if (rt) {
2563                         /* Autoconf prefix route */
2564                         if (valid_lft == 0) {
2565                                 ip6_del_rt(rt);
2566                                 rt = NULL;
2567                         } else if (addrconf_finite_timeout(rt_expires)) {
2568                                 /* not infinity */
2569                                 rt6_set_expires(rt, jiffies + rt_expires);
2570                         } else {
2571                                 rt6_clean_expires(rt);
2572                         }
2573                 } else if (valid_lft) {
2574                         clock_t expires = 0;
2575                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2576                         if (addrconf_finite_timeout(rt_expires)) {
2577                                 /* not infinity */
2578                                 flags |= RTF_EXPIRES;
2579                                 expires = jiffies_to_clock_t(rt_expires);
2580                         }
2581                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2582                                               dev, expires, flags);
2583                 }
2584                 ip6_rt_put(rt);
2585         }
2586
2587         /* Try to figure out our local address for this prefix */
2588
2589         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2590                 struct in6_addr addr;
2591                 bool tokenized = false, dev_addr_generated = false;
2592
2593                 if (pinfo->prefix_len == 64) {
2594                         memcpy(&addr, &pinfo->prefix, 8);
2595
2596                         if (!ipv6_addr_any(&in6_dev->token)) {
2597                                 read_lock_bh(&in6_dev->lock);
2598                                 memcpy(addr.s6_addr + 8,
2599                                        in6_dev->token.s6_addr + 8, 8);
2600                                 read_unlock_bh(&in6_dev->lock);
2601                                 tokenized = true;
2602                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2603                                    !ipv6_generate_stable_address(&addr, 0,
2604                                                                  in6_dev)) {
2605                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2606                                 goto ok;
2607                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2608                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2609                                 goto put;
2610                         } else {
2611                                 dev_addr_generated = true;
2612                         }
2613                         goto ok;
2614                 }
2615                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2616                                     pinfo->prefix_len);
2617                 goto put;
2618
2619 ok:
2620                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2621                                                    &addr, addr_type,
2622                                                    addr_flags, sllao,
2623                                                    tokenized, valid_lft,
2624                                                    prefered_lft);
2625                 if (err)
2626                         goto put;
2627
2628                 /* Ignore error case here because previous prefix add addr was
2629                  * successful which will be notified.
2630                  */
2631                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2632                                               addr_type, addr_flags, sllao,
2633                                               tokenized, valid_lft,
2634                                               prefered_lft,
2635                                               dev_addr_generated);
2636         }
2637         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2638 put:
2639         in6_dev_put(in6_dev);
2640 }
2641
2642 /*
2643  *      Set destination address.
2644  *      Special case for SIT interfaces where we create a new "virtual"
2645  *      device.
2646  */
2647 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2648 {
2649         struct in6_ifreq ireq;
2650         struct net_device *dev;
2651         int err = -EINVAL;
2652
2653         rtnl_lock();
2654
2655         err = -EFAULT;
2656         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2657                 goto err_exit;
2658
2659         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2660
2661         err = -ENODEV;
2662         if (!dev)
2663                 goto err_exit;
2664
2665 #if IS_ENABLED(CONFIG_IPV6_SIT)
2666         if (dev->type == ARPHRD_SIT) {
2667                 const struct net_device_ops *ops = dev->netdev_ops;
2668                 struct ifreq ifr;
2669                 struct ip_tunnel_parm p;
2670
2671                 err = -EADDRNOTAVAIL;
2672                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2673                         goto err_exit;
2674
2675                 memset(&p, 0, sizeof(p));
2676                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2677                 p.iph.saddr = 0;
2678                 p.iph.version = 4;
2679                 p.iph.ihl = 5;
2680                 p.iph.protocol = IPPROTO_IPV6;
2681                 p.iph.ttl = 64;
2682                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2683
2684                 if (ops->ndo_do_ioctl) {
2685                         mm_segment_t oldfs = get_fs();
2686
2687                         set_fs(KERNEL_DS);
2688                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2689                         set_fs(oldfs);
2690                 } else
2691                         err = -EOPNOTSUPP;
2692
2693                 if (err == 0) {
2694                         err = -ENOBUFS;
2695                         dev = __dev_get_by_name(net, p.name);
2696                         if (!dev)
2697                                 goto err_exit;
2698                         err = dev_open(dev);
2699                 }
2700         }
2701 #endif
2702
2703 err_exit:
2704         rtnl_unlock();
2705         return err;
2706 }
2707
2708 static int ipv6_mc_config(struct sock *sk, bool join,
2709                           const struct in6_addr *addr, int ifindex)
2710 {
2711         int ret;
2712
2713         ASSERT_RTNL();
2714
2715         lock_sock(sk);
2716         if (join)
2717                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2718         else
2719                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2720         release_sock(sk);
2721
2722         return ret;
2723 }
2724
2725 /*
2726  *      Manual configuration of address on an interface
2727  */
2728 static int inet6_addr_add(struct net *net, int ifindex,
2729                           const struct in6_addr *pfx,
2730                           const struct in6_addr *peer_pfx,
2731                           unsigned int plen, __u32 ifa_flags,
2732                           __u32 prefered_lft, __u32 valid_lft)
2733 {
2734         struct inet6_ifaddr *ifp;
2735         struct inet6_dev *idev;
2736         struct net_device *dev;
2737         unsigned long timeout;
2738         clock_t expires;
2739         int scope;
2740         u32 flags;
2741
2742         ASSERT_RTNL();
2743
2744         if (plen > 128)
2745                 return -EINVAL;
2746
2747         /* check the lifetime */
2748         if (!valid_lft || prefered_lft > valid_lft)
2749                 return -EINVAL;
2750
2751         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2752                 return -EINVAL;
2753
2754         dev = __dev_get_by_index(net, ifindex);
2755         if (!dev)
2756                 return -ENODEV;
2757
2758         idev = addrconf_add_dev(dev);
2759         if (IS_ERR(idev))
2760                 return PTR_ERR(idev);
2761
2762         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2763                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2764                                          true, pfx, ifindex);
2765
2766                 if (ret < 0)
2767                         return ret;
2768         }
2769
2770         scope = ipv6_addr_scope(pfx);
2771
2772         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2773         if (addrconf_finite_timeout(timeout)) {
2774                 expires = jiffies_to_clock_t(timeout * HZ);
2775                 valid_lft = timeout;
2776                 flags = RTF_EXPIRES;
2777         } else {
2778                 expires = 0;
2779                 flags = 0;
2780                 ifa_flags |= IFA_F_PERMANENT;
2781         }
2782
2783         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2784         if (addrconf_finite_timeout(timeout)) {
2785                 if (timeout == 0)
2786                         ifa_flags |= IFA_F_DEPRECATED;
2787                 prefered_lft = timeout;
2788         }
2789
2790         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2791                             valid_lft, prefered_lft);
2792
2793         if (!IS_ERR(ifp)) {
2794                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2795                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2796                                               expires, flags);
2797                 }
2798
2799                 /*
2800                  * Note that section 3.1 of RFC 4429 indicates
2801                  * that the Optimistic flag should not be set for
2802                  * manually configured addresses
2803                  */
2804                 addrconf_dad_start(ifp);
2805                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2806                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2807                                          true, jiffies);
2808                 in6_ifa_put(ifp);
2809                 addrconf_verify_rtnl();
2810                 return 0;
2811         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2812                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2813                                false, pfx, ifindex);
2814         }
2815
2816         return PTR_ERR(ifp);
2817 }
2818
2819 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2820                           const struct in6_addr *pfx, unsigned int plen)
2821 {
2822         struct inet6_ifaddr *ifp;
2823         struct inet6_dev *idev;
2824         struct net_device *dev;
2825
2826         if (plen > 128)
2827                 return -EINVAL;
2828
2829         dev = __dev_get_by_index(net, ifindex);
2830         if (!dev)
2831                 return -ENODEV;
2832
2833         idev = __in6_dev_get(dev);
2834         if (!idev)
2835                 return -ENXIO;
2836
2837         read_lock_bh(&idev->lock);
2838         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2839                 if (ifp->prefix_len == plen &&
2840                     ipv6_addr_equal(pfx, &ifp->addr)) {
2841                         in6_ifa_hold(ifp);
2842                         read_unlock_bh(&idev->lock);
2843
2844                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2845                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2846                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2847                                                  jiffies);
2848                         ipv6_del_addr(ifp);
2849                         addrconf_verify_rtnl();
2850                         if (ipv6_addr_is_multicast(pfx)) {
2851                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2852                                                false, pfx, dev->ifindex);
2853                         }
2854                         return 0;
2855                 }
2856         }
2857         read_unlock_bh(&idev->lock);
2858         return -EADDRNOTAVAIL;
2859 }
2860
2861
2862 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2863 {
2864         struct in6_ifreq ireq;
2865         int err;
2866
2867         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2868                 return -EPERM;
2869
2870         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2871                 return -EFAULT;
2872
2873         rtnl_lock();
2874         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2875                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2876                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2877         rtnl_unlock();
2878         return err;
2879 }
2880
2881 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2882 {
2883         struct in6_ifreq ireq;
2884         int err;
2885
2886         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2887                 return -EPERM;
2888
2889         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2890                 return -EFAULT;
2891
2892         rtnl_lock();
2893         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2894                              ireq.ifr6_prefixlen);
2895         rtnl_unlock();
2896         return err;
2897 }
2898
2899 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2900                      int plen, int scope)
2901 {
2902         struct inet6_ifaddr *ifp;
2903
2904         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2905                             scope, IFA_F_PERMANENT,
2906                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2907         if (!IS_ERR(ifp)) {
2908                 spin_lock_bh(&ifp->lock);
2909                 ifp->flags &= ~IFA_F_TENTATIVE;
2910                 spin_unlock_bh(&ifp->lock);
2911                 rt_genid_bump_ipv6(dev_net(idev->dev));
2912                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2913                 in6_ifa_put(ifp);
2914         }
2915 }
2916
2917 #if IS_ENABLED(CONFIG_IPV6_SIT)
2918 static void sit_add_v4_addrs(struct inet6_dev *idev)
2919 {
2920         struct in6_addr addr;
2921         struct net_device *dev;
2922         struct net *net = dev_net(idev->dev);
2923         int scope, plen;
2924         u32 pflags = 0;
2925
2926         ASSERT_RTNL();
2927
2928         memset(&addr, 0, sizeof(struct in6_addr));
2929         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2930
2931         if (idev->dev->flags&IFF_POINTOPOINT) {
2932                 addr.s6_addr32[0] = htonl(0xfe800000);
2933                 scope = IFA_LINK;
2934                 plen = 64;
2935         } else {
2936                 scope = IPV6_ADDR_COMPATv4;
2937                 plen = 96;
2938                 pflags |= RTF_NONEXTHOP;
2939         }
2940
2941         if (addr.s6_addr32[3]) {
2942                 add_addr(idev, &addr, plen, scope);
2943                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2944                 return;
2945         }
2946
2947         for_each_netdev(net, dev) {
2948                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2949                 if (in_dev && (dev->flags & IFF_UP)) {
2950                         struct in_ifaddr *ifa;
2951
2952                         int flag = scope;
2953
2954                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2955
2956                                 addr.s6_addr32[3] = ifa->ifa_local;
2957
2958                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2959                                         continue;
2960                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2961                                         if (idev->dev->flags&IFF_POINTOPOINT)
2962                                                 continue;
2963                                         flag |= IFA_HOST;
2964                                 }
2965
2966                                 add_addr(idev, &addr, plen, flag);
2967                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2968                                                       pflags);
2969                         }
2970                 }
2971         }
2972 }
2973 #endif
2974
2975 static void init_loopback(struct net_device *dev)
2976 {
2977         struct inet6_dev  *idev;
2978         struct net_device *sp_dev;
2979         struct inet6_ifaddr *sp_ifa;
2980         struct rt6_info *sp_rt;
2981
2982         /* ::1 */
2983
2984         ASSERT_RTNL();
2985
2986         idev = ipv6_find_idev(dev);
2987         if (!idev) {
2988                 pr_debug("%s: add_dev failed\n", __func__);
2989                 return;
2990         }
2991
2992         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2993
2994         /* Add routes to other interface's IPv6 addresses */
2995         for_each_netdev(dev_net(dev), sp_dev) {
2996                 if (!strcmp(sp_dev->name, dev->name))
2997                         continue;
2998
2999                 idev = __in6_dev_get(sp_dev);
3000                 if (!idev)
3001                         continue;
3002
3003                 read_lock_bh(&idev->lock);
3004                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
3005
3006                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
3007                                 continue;
3008
3009                         if (sp_ifa->rt) {
3010                                 /* This dst has been added to garbage list when
3011                                  * lo device down, release this obsolete dst and
3012                                  * reallocate a new router for ifa.
3013                                  */
3014                                 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3015                                         ip6_rt_put(sp_ifa->rt);
3016                                         sp_ifa->rt = NULL;
3017                                 } else {
3018                                         continue;
3019                                 }
3020                         }
3021
3022                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3023
3024                         /* Failure cases are ignored */
3025                         if (!IS_ERR(sp_rt)) {
3026                                 sp_ifa->rt = sp_rt;
3027                                 ip6_ins_rt(sp_rt);
3028                         }
3029                 }
3030                 read_unlock_bh(&idev->lock);
3031         }
3032 }
3033
3034 void addrconf_add_linklocal(struct inet6_dev *idev,
3035                             const struct in6_addr *addr, u32 flags)
3036 {
3037         struct inet6_ifaddr *ifp;
3038         u32 addr_flags = flags | IFA_F_PERMANENT;
3039
3040 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3041         if (idev->cnf.optimistic_dad &&
3042             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3043                 addr_flags |= IFA_F_OPTIMISTIC;
3044 #endif
3045
3046         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3047                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3048         if (!IS_ERR(ifp)) {
3049                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3050                 addrconf_dad_start(ifp);
3051                 in6_ifa_put(ifp);
3052         }
3053 }
3054 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3055
3056 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3057 {
3058         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3059                 return true;
3060
3061         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3062             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3063                 return true;
3064
3065         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3066             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3067                 return true;
3068
3069         return false;
3070 }
3071
3072 static int ipv6_generate_stable_address(struct in6_addr *address,
3073                                         u8 dad_count,
3074                                         const struct inet6_dev *idev)
3075 {
3076         static DEFINE_SPINLOCK(lock);
3077         static __u32 digest[SHA_DIGEST_WORDS];
3078         static __u32 workspace[SHA_WORKSPACE_WORDS];
3079
3080         static union {
3081                 char __data[SHA_MESSAGE_BYTES];
3082                 struct {
3083                         struct in6_addr secret;
3084                         __be32 prefix[2];
3085                         unsigned char hwaddr[MAX_ADDR_LEN];
3086                         u8 dad_count;
3087                 } __packed;
3088         } data;
3089
3090         struct in6_addr secret;
3091         struct in6_addr temp;
3092         struct net *net = dev_net(idev->dev);
3093
3094         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3095
3096         if (idev->cnf.stable_secret.initialized)
3097                 secret = idev->cnf.stable_secret.secret;
3098         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3099                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3100         else
3101                 return -1;
3102
3103 retry:
3104         spin_lock_bh(&lock);
3105
3106         sha_init(digest);
3107         memset(&data, 0, sizeof(data));
3108         memset(workspace, 0, sizeof(workspace));
3109         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3110         data.prefix[0] = address->s6_addr32[0];
3111         data.prefix[1] = address->s6_addr32[1];
3112         data.secret = secret;
3113         data.dad_count = dad_count;
3114
3115         sha_transform(digest, data.__data, workspace);
3116
3117         temp = *address;
3118         temp.s6_addr32[2] = (__force __be32)digest[0];
3119         temp.s6_addr32[3] = (__force __be32)digest[1];
3120
3121         spin_unlock_bh(&lock);
3122
3123         if (ipv6_reserved_interfaceid(temp)) {
3124                 dad_count++;
3125                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3126                         return -1;
3127                 goto retry;
3128         }
3129
3130         *address = temp;
3131         return 0;
3132 }
3133
3134 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3135 {
3136         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3137
3138         if (s->initialized)
3139                 return;
3140         s = &idev->cnf.stable_secret;
3141         get_random_bytes(&s->secret, sizeof(s->secret));
3142         s->initialized = true;
3143 }
3144
3145 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3146 {
3147         struct in6_addr addr;
3148
3149         /* no link local addresses on L3 master devices */
3150         if (netif_is_l3_master(idev->dev))
3151                 return;
3152
3153         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3154
3155         switch (idev->addr_gen_mode) {
3156         case IN6_ADDR_GEN_MODE_RANDOM:
3157                 ipv6_gen_mode_random_init(idev);
3158                 /* fallthrough */
3159         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3160                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3161                         addrconf_add_linklocal(idev, &addr,
3162                                                IFA_F_STABLE_PRIVACY);
3163                 else if (prefix_route)
3164                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3165                 break;
3166         case IN6_ADDR_GEN_MODE_EUI64:
3167                 /* addrconf_add_linklocal also adds a prefix_route and we
3168                  * only need to care about prefix routes if ipv6_generate_eui64
3169                  * couldn't generate one.
3170                  */
3171                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3172                         addrconf_add_linklocal(idev, &addr, 0);
3173                 else if (prefix_route)
3174                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3175                 break;
3176         case IN6_ADDR_GEN_MODE_NONE:
3177         default:
3178                 /* will not add any link local address */
3179                 break;
3180         }
3181 }
3182
3183 static void addrconf_dev_config(struct net_device *dev)
3184 {
3185         struct inet6_dev *idev;
3186
3187         ASSERT_RTNL();
3188
3189         if ((dev->type != ARPHRD_ETHER) &&
3190             (dev->type != ARPHRD_FDDI) &&
3191             (dev->type != ARPHRD_ARCNET) &&
3192             (dev->type != ARPHRD_INFINIBAND) &&
3193             (dev->type != ARPHRD_IEEE1394) &&
3194             (dev->type != ARPHRD_TUNNEL6) &&
3195             (dev->type != ARPHRD_6LOWPAN) &&
3196             (dev->type != ARPHRD_NONE)) {
3197                 /* Alas, we support only Ethernet autoconfiguration. */
3198                 return;
3199         }
3200
3201         idev = addrconf_add_dev(dev);
3202         if (IS_ERR(idev))
3203                 return;
3204
3205         /* this device type has no EUI support */
3206         if (dev->type == ARPHRD_NONE &&
3207             idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3208                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3209
3210         addrconf_addr_gen(idev, false);
3211 }
3212
3213 #if IS_ENABLED(CONFIG_IPV6_SIT)
3214 static void addrconf_sit_config(struct net_device *dev)
3215 {
3216         struct inet6_dev *idev;
3217
3218         ASSERT_RTNL();
3219
3220         /*
3221          * Configure the tunnel with one of our IPv4
3222          * addresses... we should configure all of
3223          * our v4 addrs in the tunnel
3224          */
3225
3226         idev = ipv6_find_idev(dev);
3227         if (!idev) {
3228                 pr_debug("%s: add_dev failed\n", __func__);
3229                 return;
3230         }
3231
3232         if (dev->priv_flags & IFF_ISATAP) {
3233                 addrconf_addr_gen(idev, false);
3234                 return;
3235         }
3236
3237         sit_add_v4_addrs(idev);
3238
3239         if (dev->flags&IFF_POINTOPOINT)
3240                 addrconf_add_mroute(dev);
3241 }
3242 #endif
3243
3244 #if IS_ENABLED(CONFIG_NET_IPGRE)
3245 static void addrconf_gre_config(struct net_device *dev)
3246 {
3247         struct inet6_dev *idev;
3248
3249         ASSERT_RTNL();
3250
3251         idev = ipv6_find_idev(dev);
3252         if (!idev) {
3253                 pr_debug("%s: add_dev failed\n", __func__);
3254                 return;
3255         }
3256
3257         addrconf_addr_gen(idev, true);
3258         if (dev->flags & IFF_POINTOPOINT)
3259                 addrconf_add_mroute(dev);
3260 }
3261 #endif
3262
3263 static int fixup_permanent_addr(struct inet6_dev *idev,
3264                                 struct inet6_ifaddr *ifp)
3265 {
3266         if (!ifp->rt) {
3267                 struct rt6_info *rt;
3268
3269                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3270                 if (unlikely(IS_ERR(rt)))
3271                         return PTR_ERR(rt);
3272
3273                 ifp->rt = rt;
3274         }
3275
3276         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3277                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3278                                       idev->dev, 0, 0);
3279         }
3280
3281         addrconf_dad_start(ifp);
3282
3283         return 0;
3284 }
3285
3286 static void addrconf_permanent_addr(struct net_device *dev)
3287 {
3288         struct inet6_ifaddr *ifp, *tmp;
3289         struct inet6_dev *idev;
3290
3291         idev = __in6_dev_get(dev);
3292         if (!idev)
3293                 return;
3294
3295         write_lock_bh(&idev->lock);
3296
3297         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3298                 if ((ifp->flags & IFA_F_PERMANENT) &&
3299                     fixup_permanent_addr(idev, ifp) < 0) {
3300                         write_unlock_bh(&idev->lock);
3301                         ipv6_del_addr(ifp);
3302                         write_lock_bh(&idev->lock);
3303
3304                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3305                                              idev->dev->name, &ifp->addr);
3306                 }
3307         }
3308
3309         write_unlock_bh(&idev->lock);
3310 }
3311
3312 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3313                            void *ptr)
3314 {
3315         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3316         struct netdev_notifier_changeupper_info *info;
3317         struct inet6_dev *idev = __in6_dev_get(dev);
3318         int run_pending = 0;
3319         int err;
3320
3321         switch (event) {
3322         case NETDEV_REGISTER:
3323                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3324                         idev = ipv6_add_dev(dev);
3325                         if (IS_ERR(idev))
3326                                 return notifier_from_errno(PTR_ERR(idev));
3327                 }
3328                 break;
3329
3330         case NETDEV_CHANGEMTU:
3331                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3332                 if (dev->mtu < IPV6_MIN_MTU) {
3333                         addrconf_ifdown(dev, 1);
3334                         break;
3335                 }
3336
3337                 if (idev) {
3338                         rt6_mtu_change(dev, dev->mtu);
3339                         idev->cnf.mtu6 = dev->mtu;
3340                         break;
3341                 }
3342
3343                 /* allocate new idev */
3344                 idev = ipv6_add_dev(dev);
3345                 if (IS_ERR(idev))
3346                         break;
3347
3348                 /* device is still not ready */
3349                 if (!(idev->if_flags & IF_READY))
3350                         break;
3351
3352                 run_pending = 1;
3353
3354                 /* fall through */
3355
3356         case NETDEV_UP:
3357         case NETDEV_CHANGE:
3358                 if (dev->flags & IFF_SLAVE)
3359                         break;
3360
3361                 if (idev && idev->cnf.disable_ipv6)
3362                         break;
3363
3364                 if (event == NETDEV_UP) {
3365                         /* restore routes for permanent addresses */
3366                         addrconf_permanent_addr(dev);
3367
3368                         if (!addrconf_qdisc_ok(dev)) {
3369                                 /* device is not ready yet. */
3370                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3371                                         dev->name);
3372                                 break;
3373                         }
3374
3375                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3376                                 idev = ipv6_add_dev(dev);
3377
3378                         if (!IS_ERR_OR_NULL(idev)) {
3379                                 idev->if_flags |= IF_READY;
3380                                 run_pending = 1;
3381                         }
3382                 } else if (event == NETDEV_CHANGE) {
3383                         if (!addrconf_qdisc_ok(dev)) {
3384                                 /* device is still not ready. */
3385                                 break;
3386                         }
3387
3388                         if (idev) {
3389                                 if (idev->if_flags & IF_READY) {
3390                                         /* device is already configured -
3391                                          * but resend MLD reports, we might
3392                                          * have roamed and need to update
3393                                          * multicast snooping switches
3394                                          */
3395                                         ipv6_mc_up(idev);
3396                                         break;
3397                                 }
3398                                 idev->if_flags |= IF_READY;
3399                         }
3400
3401                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3402                                 dev->name);
3403
3404                         run_pending = 1;
3405                 }
3406
3407                 switch (dev->type) {
3408 #if IS_ENABLED(CONFIG_IPV6_SIT)
3409                 case ARPHRD_SIT:
3410                         addrconf_sit_config(dev);
3411                         break;
3412 #endif
3413 #if IS_ENABLED(CONFIG_NET_IPGRE)
3414                 case ARPHRD_IPGRE:
3415                         addrconf_gre_config(dev);
3416                         break;
3417 #endif
3418                 case ARPHRD_LOOPBACK:
3419                         init_loopback(dev);
3420                         break;
3421
3422                 default:
3423                         addrconf_dev_config(dev);
3424                         break;
3425                 }
3426
3427                 if (!IS_ERR_OR_NULL(idev)) {
3428                         if (run_pending)
3429                                 addrconf_dad_run(idev);
3430
3431                         /*
3432                          * If the MTU changed during the interface down,
3433                          * when the interface up, the changed MTU must be
3434                          * reflected in the idev as well as routers.
3435                          */
3436                         if (idev->cnf.mtu6 != dev->mtu &&
3437                             dev->mtu >= IPV6_MIN_MTU) {
3438                                 rt6_mtu_change(dev, dev->mtu);
3439                                 idev->cnf.mtu6 = dev->mtu;
3440                         }
3441                         idev->tstamp = jiffies;
3442                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3443
3444                         /*
3445                          * If the changed mtu during down is lower than
3446                          * IPV6_MIN_MTU stop IPv6 on this interface.
3447                          */
3448                         if (dev->mtu < IPV6_MIN_MTU)
3449                                 addrconf_ifdown(dev, 1);
3450                 }
3451                 break;
3452
3453         case NETDEV_DOWN:
3454         case NETDEV_UNREGISTER:
3455                 /*
3456                  *      Remove all addresses from this interface.
3457                  */
3458                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3459                 break;
3460
3461         case NETDEV_CHANGENAME:
3462                 if (idev) {
3463                         snmp6_unregister_dev(idev);
3464                         addrconf_sysctl_unregister(idev);
3465                         err = addrconf_sysctl_register(idev);
3466                         if (err)
3467                                 return notifier_from_errno(err);
3468                         err = snmp6_register_dev(idev);
3469                         if (err) {
3470                                 addrconf_sysctl_unregister(idev);
3471                                 return notifier_from_errno(err);
3472                         }
3473                 }
3474                 break;
3475
3476         case NETDEV_PRE_TYPE_CHANGE:
3477         case NETDEV_POST_TYPE_CHANGE:
3478                 if (idev)
3479                         addrconf_type_change(dev, event);
3480                 break;
3481
3482         case NETDEV_CHANGEUPPER:
3483                 info = ptr;
3484
3485                 /* flush all routes if dev is linked to or unlinked from
3486                  * an L3 master device (e.g., VRF)
3487                  */
3488                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3489                         addrconf_ifdown(dev, 0);
3490         }
3491
3492         return NOTIFY_OK;
3493 }
3494
3495 /*
3496  *      addrconf module should be notified of a device going up
3497  */
3498 static struct notifier_block ipv6_dev_notf = {
3499         .notifier_call = addrconf_notify,
3500 };
3501
3502 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3503 {
3504         struct inet6_dev *idev;
3505         ASSERT_RTNL();
3506
3507         idev = __in6_dev_get(dev);
3508
3509         if (event == NETDEV_POST_TYPE_CHANGE)
3510                 ipv6_mc_remap(idev);
3511         else if (event == NETDEV_PRE_TYPE_CHANGE)
3512                 ipv6_mc_unmap(idev);
3513 }
3514
3515 static bool addr_is_local(const struct in6_addr *addr)
3516 {
3517         return ipv6_addr_type(addr) &
3518                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3519 }
3520
3521 static int addrconf_ifdown(struct net_device *dev, int how)
3522 {
3523         struct net *net = dev_net(dev);
3524         struct inet6_dev *idev;
3525         struct inet6_ifaddr *ifa, *tmp;
3526         struct list_head del_list;
3527         int _keep_addr;
3528         bool keep_addr;
3529         int state, i;
3530
3531         ASSERT_RTNL();
3532
3533         rt6_ifdown(net, dev);
3534         neigh_ifdown(&nd_tbl, dev);
3535
3536         idev = __in6_dev_get(dev);
3537         if (!idev)
3538                 return -ENODEV;
3539
3540         /*
3541          * Step 1: remove reference to ipv6 device from parent device.
3542          *         Do not dev_put!
3543          */
3544         if (how) {
3545                 idev->dead = 1;
3546
3547                 /* protected by rtnl_lock */
3548                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3549
3550                 /* Step 1.5: remove snmp6 entry */
3551                 snmp6_unregister_dev(idev);
3552
3553         }
3554
3555         /* aggregate the system setting and interface setting */
3556         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3557         if (!_keep_addr)
3558                 _keep_addr = idev->cnf.keep_addr_on_down;
3559
3560         /* combine the user config with event to determine if permanent
3561          * addresses are to be removed from address hash table
3562          */
3563         keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3564
3565         /* Step 2: clear hash table */
3566         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3567                 struct hlist_head *h = &inet6_addr_lst[i];
3568
3569                 spin_lock_bh(&addrconf_hash_lock);
3570 restart:
3571                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3572                         if (ifa->idev == idev) {
3573                                 addrconf_del_dad_work(ifa);
3574                                 /* combined flag + permanent flag decide if
3575                                  * address is retained on a down event
3576                                  */
3577                                 if (!keep_addr ||
3578                                     !(ifa->flags & IFA_F_PERMANENT) ||
3579                                     addr_is_local(&ifa->addr)) {
3580                                         hlist_del_init_rcu(&ifa->addr_lst);
3581                                         goto restart;
3582                                 }
3583                         }
3584                 }
3585                 spin_unlock_bh(&addrconf_hash_lock);
3586         }
3587
3588         write_lock_bh(&idev->lock);
3589
3590         addrconf_del_rs_timer(idev);
3591
3592         /* Step 2: clear flags for stateless addrconf */
3593         if (!how)
3594                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3595
3596         /* Step 3: clear tempaddr list */
3597         while (!list_empty(&idev->tempaddr_list)) {
3598                 ifa = list_first_entry(&idev->tempaddr_list,
3599                                        struct inet6_ifaddr, tmp_list);
3600                 list_del(&ifa->tmp_list);
3601                 write_unlock_bh(&idev->lock);
3602                 spin_lock_bh(&ifa->lock);
3603
3604                 if (ifa->ifpub) {
3605                         in6_ifa_put(ifa->ifpub);
3606                         ifa->ifpub = NULL;
3607                 }
3608                 spin_unlock_bh(&ifa->lock);
3609                 in6_ifa_put(ifa);
3610                 write_lock_bh(&idev->lock);
3611         }
3612
3613         /* re-combine the user config with event to determine if permanent
3614          * addresses are to be removed from the interface list
3615          */
3616         keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3617
3618         INIT_LIST_HEAD(&del_list);
3619         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3620                 struct rt6_info *rt = NULL;
3621
3622                 addrconf_del_dad_work(ifa);
3623
3624                 write_unlock_bh(&idev->lock);
3625                 spin_lock_bh(&ifa->lock);
3626
3627                 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3628                     !addr_is_local(&ifa->addr)) {
3629                         /* set state to skip the notifier below */
3630                         state = INET6_IFADDR_STATE_DEAD;
3631                         ifa->state = 0;
3632                         if (!(ifa->flags & IFA_F_NODAD))
3633                                 ifa->flags |= IFA_F_TENTATIVE;
3634
3635                         rt = ifa->rt;
3636                         ifa->rt = NULL;
3637                 } else {
3638                         state = ifa->state;
3639                         ifa->state = INET6_IFADDR_STATE_DEAD;
3640
3641                         list_move(&ifa->if_list, &del_list);
3642                 }
3643
3644                 spin_unlock_bh(&ifa->lock);
3645
3646                 if (rt)
3647                         ip6_del_rt(rt);
3648
3649                 if (state != INET6_IFADDR_STATE_DEAD) {
3650                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3651                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3652                 } else {
3653                         if (idev->cnf.forwarding)
3654                                 addrconf_leave_anycast(ifa);
3655                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3656                 }
3657
3658                 write_lock_bh(&idev->lock);
3659         }
3660
3661         write_unlock_bh(&idev->lock);
3662
3663         /* now clean up addresses to be removed */
3664         while (!list_empty(&del_list)) {
3665                 ifa = list_first_entry(&del_list,
3666                                        struct inet6_ifaddr, if_list);
3667                 list_del(&ifa->if_list);
3668
3669                 in6_ifa_put(ifa);
3670         }
3671
3672         /* Step 5: Discard anycast and multicast list */
3673         if (how) {
3674                 ipv6_ac_destroy_dev(idev);
3675                 ipv6_mc_destroy_dev(idev);
3676         } else {
3677                 ipv6_mc_down(idev);
3678         }
3679
3680         idev->tstamp = jiffies;
3681
3682         /* Last: Shot the device (if unregistered) */
3683         if (how) {
3684                 addrconf_sysctl_unregister(idev);
3685                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3686                 neigh_ifdown(&nd_tbl, dev);
3687                 in6_dev_put(idev);
3688         }
3689         return 0;
3690 }
3691
3692 static void addrconf_rs_timer(unsigned long data)
3693 {
3694         struct inet6_dev *idev = (struct inet6_dev *)data;
3695         struct net_device *dev = idev->dev;
3696         struct in6_addr lladdr;
3697
3698         write_lock(&idev->lock);
3699         if (idev->dead || !(idev->if_flags & IF_READY))
3700                 goto out;
3701
3702         if (!ipv6_accept_ra(idev))
3703                 goto out;
3704
3705         /* Announcement received after solicitation was sent */
3706         if (idev->if_flags & IF_RA_RCVD)
3707                 goto out;
3708
3709         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3710                 write_unlock(&idev->lock);
3711                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3712                         ndisc_send_rs(dev, &lladdr,
3713                                       &in6addr_linklocal_allrouters);
3714                 else
3715                         goto put;
3716
3717                 write_lock(&idev->lock);
3718                 idev->rs_interval = rfc3315_s14_backoff_update(
3719                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3720                 /* The wait after the last probe can be shorter */
3721                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3722                                              idev->cnf.rtr_solicits) ?
3723                                       idev->cnf.rtr_solicit_delay :
3724                                       idev->rs_interval);
3725         } else {
3726                 /*
3727                  * Note: we do not support deprecated "all on-link"
3728                  * assumption any longer.
3729                  */
3730                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3731         }
3732
3733 out:
3734         write_unlock(&idev->lock);
3735 put:
3736         in6_dev_put(idev);
3737 }
3738
3739 /*
3740  *      Duplicate Address Detection
3741  */
3742 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3743 {
3744         unsigned long rand_num;
3745         struct inet6_dev *idev = ifp->idev;
3746         u64 nonce;
3747
3748         if (ifp->flags & IFA_F_OPTIMISTIC)
3749                 rand_num = 0;
3750         else
3751                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3752
3753         nonce = 0;
3754         if (idev->cnf.enhanced_dad ||
3755             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3756                 do
3757                         get_random_bytes(&nonce, 6);
3758                 while (nonce == 0);
3759         }
3760         ifp->dad_nonce = nonce;
3761         ifp->dad_probes = idev->cnf.dad_transmits;
3762         addrconf_mod_dad_work(ifp, rand_num);
3763 }
3764
3765 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3766 {
3767         struct inet6_dev *idev = ifp->idev;
3768         struct net_device *dev = idev->dev;
3769         bool bump_id, notify = false;
3770
3771         addrconf_join_solict(dev, &ifp->addr);
3772
3773         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3774
3775         read_lock_bh(&idev->lock);
3776         spin_lock(&ifp->lock);
3777         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3778                 goto out;
3779
3780         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3781             idev->cnf.accept_dad < 1 ||
3782             !(ifp->flags&IFA_F_TENTATIVE) ||
3783             ifp->flags & IFA_F_NODAD) {
3784                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3785                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3786                 spin_unlock(&ifp->lock);
3787                 read_unlock_bh(&idev->lock);
3788
3789                 addrconf_dad_completed(ifp, bump_id);
3790                 return;
3791         }
3792
3793         if (!(idev->if_flags & IF_READY)) {
3794                 spin_unlock(&ifp->lock);
3795                 read_unlock_bh(&idev->lock);
3796                 /*
3797                  * If the device is not ready:
3798                  * - keep it tentative if it is a permanent address.
3799                  * - otherwise, kill it.
3800                  */
3801                 in6_ifa_hold(ifp);
3802                 addrconf_dad_stop(ifp, 0);
3803                 return;
3804         }
3805
3806         /*
3807          * Optimistic nodes can start receiving
3808          * Frames right away
3809          */
3810         if (ifp->flags & IFA_F_OPTIMISTIC) {
3811                 ip6_ins_rt(ifp->rt);
3812                 if (ipv6_use_optimistic_addr(idev)) {
3813                         /* Because optimistic nodes can use this address,
3814                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3815                          */
3816                         notify = true;
3817                 }
3818         }
3819
3820         addrconf_dad_kick(ifp);
3821 out:
3822         spin_unlock(&ifp->lock);
3823         read_unlock_bh(&idev->lock);
3824         if (notify)
3825                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3826 }
3827
3828 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3829 {
3830         bool begin_dad = false;
3831
3832         spin_lock_bh(&ifp->lock);
3833         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3834                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3835                 begin_dad = true;
3836         }
3837         spin_unlock_bh(&ifp->lock);
3838
3839         if (begin_dad)
3840                 addrconf_mod_dad_work(ifp, 0);
3841 }
3842
3843 static void addrconf_dad_work(struct work_struct *w)
3844 {
3845         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3846                                                 struct inet6_ifaddr,
3847                                                 dad_work);
3848         struct inet6_dev *idev = ifp->idev;
3849         bool bump_id, disable_ipv6 = false;
3850         struct in6_addr mcaddr;
3851
3852         enum {
3853                 DAD_PROCESS,
3854                 DAD_BEGIN,
3855                 DAD_ABORT,
3856         } action = DAD_PROCESS;
3857
3858         rtnl_lock();
3859
3860         spin_lock_bh(&ifp->lock);
3861         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3862                 action = DAD_BEGIN;
3863                 ifp->state = INET6_IFADDR_STATE_DAD;
3864         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3865                 action = DAD_ABORT;
3866                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3867
3868                 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3869                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3870                         struct in6_addr addr;
3871
3872                         addr.s6_addr32[0] = htonl(0xfe800000);
3873                         addr.s6_addr32[1] = 0;
3874
3875                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3876                             ipv6_addr_equal(&ifp->addr, &addr)) {
3877                                 /* DAD failed for link-local based on MAC */
3878                                 idev->cnf.disable_ipv6 = 1;
3879
3880                                 pr_info("%s: IPv6 being disabled!\n",
3881                                         ifp->idev->dev->name);
3882                                 disable_ipv6 = true;
3883                         }
3884                 }
3885         }
3886         spin_unlock_bh(&ifp->lock);
3887
3888         if (action == DAD_BEGIN) {
3889                 addrconf_dad_begin(ifp);
3890                 goto out;
3891         } else if (action == DAD_ABORT) {
3892                 in6_ifa_hold(ifp);
3893                 addrconf_dad_stop(ifp, 1);
3894                 if (disable_ipv6)
3895                         addrconf_ifdown(idev->dev, 0);
3896                 goto out;
3897         }
3898
3899         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3900                 goto out;
3901
3902         write_lock_bh(&idev->lock);
3903         if (idev->dead || !(idev->if_flags & IF_READY)) {
3904                 write_unlock_bh(&idev->lock);
3905                 goto out;
3906         }
3907
3908         spin_lock(&ifp->lock);
3909         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3910                 spin_unlock(&ifp->lock);
3911                 write_unlock_bh(&idev->lock);
3912                 goto out;
3913         }
3914
3915         if (ifp->dad_probes == 0) {
3916                 /*
3917                  * DAD was successful
3918                  */
3919
3920                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3921                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3922                 spin_unlock(&ifp->lock);
3923                 write_unlock_bh(&idev->lock);
3924
3925                 addrconf_dad_completed(ifp, bump_id);
3926
3927                 goto out;
3928         }
3929
3930         ifp->dad_probes--;
3931         addrconf_mod_dad_work(ifp,
3932                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3933         spin_unlock(&ifp->lock);
3934         write_unlock_bh(&idev->lock);
3935
3936         /* send a neighbour solicitation for our addr */
3937         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3938         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
3939                       ifp->dad_nonce);
3940 out:
3941         in6_ifa_put(ifp);
3942         rtnl_unlock();
3943 }
3944
3945 /* ifp->idev must be at least read locked */
3946 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3947 {
3948         struct inet6_ifaddr *ifpiter;
3949         struct inet6_dev *idev = ifp->idev;
3950
3951         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3952                 if (ifpiter->scope > IFA_LINK)
3953                         break;
3954                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3955                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3956                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3957                     IFA_F_PERMANENT)
3958                         return false;
3959         }
3960         return true;
3961 }
3962
3963 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
3964 {
3965         struct net_device *dev = ifp->idev->dev;
3966         struct in6_addr lladdr;
3967         bool send_rs, send_mld;
3968
3969         addrconf_del_dad_work(ifp);
3970
3971         /*
3972          *      Configure the address for reception. Now it is valid.
3973          */
3974
3975         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3976
3977         /* If added prefix is link local and we are prepared to process
3978            router advertisements, start sending router solicitations.
3979          */
3980
3981         read_lock_bh(&ifp->idev->lock);
3982         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3983         send_rs = send_mld &&
3984                   ipv6_accept_ra(ifp->idev) &&
3985                   ifp->idev->cnf.rtr_solicits != 0 &&
3986                   (dev->flags&IFF_LOOPBACK) == 0;
3987         read_unlock_bh(&ifp->idev->lock);
3988
3989         /* While dad is in progress mld report's source address is in6_addrany.
3990          * Resend with proper ll now.
3991          */
3992         if (send_mld)
3993                 ipv6_mc_dad_complete(ifp->idev);
3994
3995         if (send_rs) {
3996                 /*
3997                  *      If a host as already performed a random delay
3998                  *      [...] as part of DAD [...] there is no need
3999                  *      to delay again before sending the first RS
4000                  */
4001                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4002                         return;
4003                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4004
4005                 write_lock_bh(&ifp->idev->lock);
4006                 spin_lock(&ifp->lock);
4007                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4008                         ifp->idev->cnf.rtr_solicit_interval);
4009                 ifp->idev->rs_probes = 1;
4010                 ifp->idev->if_flags |= IF_RS_SENT;
4011                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4012                 spin_unlock(&ifp->lock);
4013                 write_unlock_bh(&ifp->idev->lock);
4014         }
4015
4016         if (bump_id)
4017                 rt_genid_bump_ipv6(dev_net(dev));
4018 }
4019
4020 static void addrconf_dad_run(struct inet6_dev *idev)
4021 {
4022         struct inet6_ifaddr *ifp;
4023
4024         read_lock_bh(&idev->lock);
4025         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4026                 spin_lock(&ifp->lock);
4027                 if (ifp->flags & IFA_F_TENTATIVE &&
4028                     ifp->state == INET6_IFADDR_STATE_DAD)
4029                         addrconf_dad_kick(ifp);
4030                 spin_unlock(&ifp->lock);
4031         }
4032         read_unlock_bh(&idev->lock);
4033 }
4034
4035 #ifdef CONFIG_PROC_FS
4036 struct if6_iter_state {
4037         struct seq_net_private p;
4038         int bucket;
4039         int offset;
4040 };
4041
4042 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4043 {
4044         struct inet6_ifaddr *ifa = NULL;
4045         struct if6_iter_state *state = seq->private;
4046         struct net *net = seq_file_net(seq);
4047         int p = 0;
4048
4049         /* initial bucket if pos is 0 */
4050         if (pos == 0) {
4051                 state->bucket = 0;
4052                 state->offset = 0;
4053         }
4054
4055         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4056                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4057                                          addr_lst) {
4058                         if (!net_eq(dev_net(ifa->idev->dev), net))
4059                                 continue;
4060                         /* sync with offset */
4061                         if (p < state->offset) {
4062                                 p++;
4063                                 continue;
4064                         }
4065                         state->offset++;
4066                         return ifa;
4067                 }
4068
4069                 /* prepare for next bucket */
4070                 state->offset = 0;
4071                 p = 0;
4072         }
4073         return NULL;
4074 }
4075
4076 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4077                                          struct inet6_ifaddr *ifa)
4078 {
4079         struct if6_iter_state *state = seq->private;
4080         struct net *net = seq_file_net(seq);
4081
4082         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4083                 if (!net_eq(dev_net(ifa->idev->dev), net))
4084                         continue;
4085                 state->offset++;
4086                 return ifa;
4087         }
4088
4089         while (++state->bucket < IN6_ADDR_HSIZE) {
4090                 state->offset = 0;
4091                 hlist_for_each_entry_rcu_bh(ifa,
4092                                      &inet6_addr_lst[state->bucket], addr_lst) {
4093                         if (!net_eq(dev_net(ifa->idev->dev), net))
4094                                 continue;
4095                         state->offset++;
4096                         return ifa;
4097                 }
4098         }
4099
4100         return NULL;
4101 }
4102
4103 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4104         __acquires(rcu_bh)
4105 {
4106         rcu_read_lock_bh();
4107         return if6_get_first(seq, *pos);
4108 }
4109
4110 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4111 {
4112         struct inet6_ifaddr *ifa;
4113
4114         ifa = if6_get_next(seq, v);
4115         ++*pos;
4116         return ifa;
4117 }
4118
4119 static void if6_seq_stop(struct seq_file *seq, void *v)
4120         __releases(rcu_bh)
4121 {
4122         rcu_read_unlock_bh();
4123 }
4124
4125 static int if6_seq_show(struct seq_file *seq, void *v)
4126 {
4127         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4128         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4129                    &ifp->addr,
4130                    ifp->idev->dev->ifindex,
4131                    ifp->prefix_len,
4132                    ifp->scope,
4133                    (u8) ifp->flags,
4134                    ifp->idev->dev->name);
4135         return 0;
4136 }
4137
4138 static const struct seq_operations if6_seq_ops = {
4139         .start  = if6_seq_start,
4140         .next   = if6_seq_next,
4141         .show   = if6_seq_show,
4142         .stop   = if6_seq_stop,
4143 };
4144
4145 static int if6_seq_open(struct inode *inode, struct file *file)
4146 {
4147         return seq_open_net(inode, file, &if6_seq_ops,
4148                             sizeof(struct if6_iter_state));
4149 }
4150
4151 static const struct file_operations if6_fops = {
4152         .owner          = THIS_MODULE,
4153         .open           = if6_seq_open,
4154         .read           = seq_read,
4155         .llseek         = seq_lseek,
4156         .release        = seq_release_net,
4157 };
4158
4159 static int __net_init if6_proc_net_init(struct net *net)
4160 {
4161         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4162                 return -ENOMEM;
4163         return 0;
4164 }
4165
4166 static void __net_exit if6_proc_net_exit(struct net *net)
4167 {
4168         remove_proc_entry("if_inet6", net->proc_net);
4169 }
4170
4171 static struct pernet_operations if6_proc_net_ops = {
4172         .init = if6_proc_net_init,
4173         .exit = if6_proc_net_exit,
4174 };
4175
4176 int __init if6_proc_init(void)
4177 {
4178         return register_pernet_subsys(&if6_proc_net_ops);
4179 }
4180
4181 void if6_proc_exit(void)
4182 {
4183         unregister_pernet_subsys(&if6_proc_net_ops);
4184 }
4185 #endif  /* CONFIG_PROC_FS */
4186
4187 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4188 /* Check if address is a home address configured on any interface. */
4189 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4190 {
4191         int ret = 0;
4192         struct inet6_ifaddr *ifp = NULL;
4193         unsigned int hash = inet6_addr_hash(addr);
4194
4195         rcu_read_lock_bh();
4196         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4197                 if (!net_eq(dev_net(ifp->idev->dev), net))
4198                         continue;
4199                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4200                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4201                         ret = 1;
4202                         break;
4203                 }
4204         }
4205         rcu_read_unlock_bh();
4206         return ret;
4207 }
4208 #endif
4209
4210 /*
4211  *      Periodic address status verification
4212  */
4213
4214 static void addrconf_verify_rtnl(void)
4215 {
4216         unsigned long now, next, next_sec, next_sched;
4217         struct inet6_ifaddr *ifp;
4218         int i;
4219
4220         ASSERT_RTNL();
4221
4222         rcu_read_lock_bh();
4223         now = jiffies;
4224         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4225
4226         cancel_delayed_work(&addr_chk_work);
4227
4228         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4229 restart:
4230                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4231                         unsigned long age;
4232
4233                         /* When setting preferred_lft to a value not zero or
4234                          * infinity, while valid_lft is infinity
4235                          * IFA_F_PERMANENT has a non-infinity life time.
4236                          */
4237                         if ((ifp->flags & IFA_F_PERMANENT) &&
4238                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4239                                 continue;
4240
4241                         spin_lock(&ifp->lock);
4242                         /* We try to batch several events at once. */
4243                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4244
4245                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4246                             age >= ifp->valid_lft) {
4247                                 spin_unlock(&ifp->lock);
4248                                 in6_ifa_hold(ifp);
4249                                 ipv6_del_addr(ifp);
4250                                 goto restart;
4251                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4252                                 spin_unlock(&ifp->lock);
4253                                 continue;
4254                         } else if (age >= ifp->prefered_lft) {
4255                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4256                                 int deprecate = 0;
4257
4258                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4259                                         deprecate = 1;
4260                                         ifp->flags |= IFA_F_DEPRECATED;
4261                                 }
4262
4263                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4264                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4265                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4266
4267                                 spin_unlock(&ifp->lock);
4268
4269                                 if (deprecate) {
4270                                         in6_ifa_hold(ifp);
4271
4272                                         ipv6_ifa_notify(0, ifp);
4273                                         in6_ifa_put(ifp);
4274                                         goto restart;
4275                                 }
4276                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4277                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4278                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4279                                         ifp->idev->cnf.dad_transmits *
4280                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4281
4282                                 if (age >= ifp->prefered_lft - regen_advance) {
4283                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4284                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4285                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4286                                         if (!ifp->regen_count && ifpub) {
4287                                                 ifp->regen_count++;
4288                                                 in6_ifa_hold(ifp);
4289                                                 in6_ifa_hold(ifpub);
4290                                                 spin_unlock(&ifp->lock);
4291
4292                                                 spin_lock(&ifpub->lock);
4293                                                 ifpub->regen_count = 0;
4294                                                 spin_unlock(&ifpub->lock);
4295                                                 ipv6_create_tempaddr(ifpub, ifp);
4296                                                 in6_ifa_put(ifpub);
4297                                                 in6_ifa_put(ifp);
4298                                                 goto restart;
4299                                         }
4300                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4301                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4302                                 spin_unlock(&ifp->lock);
4303                         } else {
4304                                 /* ifp->prefered_lft <= ifp->valid_lft */
4305                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4306                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4307                                 spin_unlock(&ifp->lock);
4308                         }
4309                 }
4310         }
4311
4312         next_sec = round_jiffies_up(next);
4313         next_sched = next;
4314
4315         /* If rounded timeout is accurate enough, accept it. */
4316         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4317                 next_sched = next_sec;
4318
4319         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4320         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4321                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4322
4323         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4324               now, next, next_sec, next_sched);
4325         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4326         rcu_read_unlock_bh();
4327 }
4328
4329 static void addrconf_verify_work(struct work_struct *w)
4330 {
4331         rtnl_lock();
4332         addrconf_verify_rtnl();
4333         rtnl_unlock();
4334 }
4335
4336 static void addrconf_verify(void)
4337 {
4338         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4339 }
4340
4341 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4342                                      struct in6_addr **peer_pfx)
4343 {
4344         struct in6_addr *pfx = NULL;
4345
4346         *peer_pfx = NULL;
4347
4348         if (addr)
4349                 pfx = nla_data(addr);
4350
4351         if (local) {
4352                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4353                         *peer_pfx = pfx;
4354                 pfx = nla_data(local);
4355         }
4356
4357         return pfx;
4358 }
4359
4360 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4361         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4362         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4363         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4364         [IFA_FLAGS]             = { .len = sizeof(u32) },
4365 };
4366
4367 static int
4368 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4369 {
4370         struct net *net = sock_net(skb->sk);
4371         struct ifaddrmsg *ifm;
4372         struct nlattr *tb[IFA_MAX+1];
4373         struct in6_addr *pfx, *peer_pfx;
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         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4387
4388         /* We ignore other flags so far. */
4389         ifa_flags &= IFA_F_MANAGETEMPADDR;
4390
4391         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4392                               ifm->ifa_prefixlen);
4393 }
4394
4395 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4396                              u32 prefered_lft, u32 valid_lft)
4397 {
4398         u32 flags;
4399         clock_t expires;
4400         unsigned long timeout;
4401         bool was_managetempaddr;
4402         bool had_prefixroute;
4403
4404         ASSERT_RTNL();
4405
4406         if (!valid_lft || (prefered_lft > valid_lft))
4407                 return -EINVAL;
4408
4409         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4410             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4411                 return -EINVAL;
4412
4413         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4414         if (addrconf_finite_timeout(timeout)) {
4415                 expires = jiffies_to_clock_t(timeout * HZ);
4416                 valid_lft = timeout;
4417                 flags = RTF_EXPIRES;
4418         } else {
4419                 expires = 0;
4420                 flags = 0;
4421                 ifa_flags |= IFA_F_PERMANENT;
4422         }
4423
4424         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4425         if (addrconf_finite_timeout(timeout)) {
4426                 if (timeout == 0)
4427                         ifa_flags |= IFA_F_DEPRECATED;
4428                 prefered_lft = timeout;
4429         }
4430
4431         spin_lock_bh(&ifp->lock);
4432         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4433         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4434                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4435         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4436                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4437                         IFA_F_NOPREFIXROUTE);
4438         ifp->flags |= ifa_flags;
4439         ifp->tstamp = jiffies;
4440         ifp->valid_lft = valid_lft;
4441         ifp->prefered_lft = prefered_lft;
4442
4443         spin_unlock_bh(&ifp->lock);
4444         if (!(ifp->flags&IFA_F_TENTATIVE))
4445                 ipv6_ifa_notify(0, ifp);
4446
4447         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4448                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4449                                       expires, flags);
4450         } else if (had_prefixroute) {
4451                 enum cleanup_prefix_rt_t action;
4452                 unsigned long rt_expires;
4453
4454                 write_lock_bh(&ifp->idev->lock);
4455                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4456                 write_unlock_bh(&ifp->idev->lock);
4457
4458                 if (action != CLEANUP_PREFIX_RT_NOP) {
4459                         cleanup_prefix_route(ifp, rt_expires,
4460                                 action == CLEANUP_PREFIX_RT_DEL);
4461                 }
4462         }
4463
4464         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4465                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4466                         valid_lft = prefered_lft = 0;
4467                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4468                                  !was_managetempaddr, jiffies);
4469         }
4470
4471         addrconf_verify_rtnl();
4472
4473         return 0;
4474 }
4475
4476 static int
4477 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4478 {
4479         struct net *net = sock_net(skb->sk);
4480         struct ifaddrmsg *ifm;
4481         struct nlattr *tb[IFA_MAX+1];
4482         struct in6_addr *pfx, *peer_pfx;
4483         struct inet6_ifaddr *ifa;
4484         struct net_device *dev;
4485         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4486         u32 ifa_flags;
4487         int err;
4488
4489         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4490         if (err < 0)
4491                 return err;
4492
4493         ifm = nlmsg_data(nlh);
4494         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4495         if (!pfx)
4496                 return -EINVAL;
4497
4498         if (tb[IFA_CACHEINFO]) {
4499                 struct ifa_cacheinfo *ci;
4500
4501                 ci = nla_data(tb[IFA_CACHEINFO]);
4502                 valid_lft = ci->ifa_valid;
4503                 preferred_lft = ci->ifa_prefered;
4504         } else {
4505                 preferred_lft = INFINITY_LIFE_TIME;
4506                 valid_lft = INFINITY_LIFE_TIME;
4507         }
4508
4509         dev =  __dev_get_by_index(net, ifm->ifa_index);
4510         if (!dev)
4511                 return -ENODEV;
4512
4513         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4514
4515         /* We ignore other flags so far. */
4516         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4517                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4518
4519         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4520         if (!ifa) {
4521                 /*
4522                  * It would be best to check for !NLM_F_CREATE here but
4523                  * userspace already relies on not having to provide this.
4524                  */
4525                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4526                                       ifm->ifa_prefixlen, ifa_flags,
4527                                       preferred_lft, valid_lft);
4528         }
4529
4530         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4531             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4532                 err = -EEXIST;
4533         else
4534                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4535
4536         in6_ifa_put(ifa);
4537
4538         return err;
4539 }
4540
4541 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4542                           u8 scope, int ifindex)
4543 {
4544         struct ifaddrmsg *ifm;
4545
4546         ifm = nlmsg_data(nlh);
4547         ifm->ifa_family = AF_INET6;
4548         ifm->ifa_prefixlen = prefixlen;
4549         ifm->ifa_flags = flags;
4550         ifm->ifa_scope = scope;
4551         ifm->ifa_index = ifindex;
4552 }
4553
4554 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4555                          unsigned long tstamp, u32 preferred, u32 valid)
4556 {
4557         struct ifa_cacheinfo ci;
4558
4559         ci.cstamp = cstamp_delta(cstamp);
4560         ci.tstamp = cstamp_delta(tstamp);
4561         ci.ifa_prefered = preferred;
4562         ci.ifa_valid = valid;
4563
4564         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4565 }
4566
4567 static inline int rt_scope(int ifa_scope)
4568 {
4569         if (ifa_scope & IFA_HOST)
4570                 return RT_SCOPE_HOST;
4571         else if (ifa_scope & IFA_LINK)
4572                 return RT_SCOPE_LINK;
4573         else if (ifa_scope & IFA_SITE)
4574                 return RT_SCOPE_SITE;
4575         else
4576                 return RT_SCOPE_UNIVERSE;
4577 }
4578
4579 static inline int inet6_ifaddr_msgsize(void)
4580 {
4581         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4582                + nla_total_size(16) /* IFA_LOCAL */
4583                + nla_total_size(16) /* IFA_ADDRESS */
4584                + nla_total_size(sizeof(struct ifa_cacheinfo))
4585                + nla_total_size(4)  /* IFA_FLAGS */;
4586 }
4587
4588 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4589                              u32 portid, u32 seq, int event, unsigned int flags)
4590 {
4591         struct nlmsghdr  *nlh;
4592         u32 preferred, valid;
4593
4594         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4595         if (!nlh)
4596                 return -EMSGSIZE;
4597
4598         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4599                       ifa->idev->dev->ifindex);
4600
4601         if (!((ifa->flags&IFA_F_PERMANENT) &&
4602               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4603                 preferred = ifa->prefered_lft;
4604                 valid = ifa->valid_lft;
4605                 if (preferred != INFINITY_LIFE_TIME) {
4606                         long tval = (jiffies - ifa->tstamp)/HZ;
4607                         if (preferred > tval)
4608                                 preferred -= tval;
4609                         else
4610                                 preferred = 0;
4611                         if (valid != INFINITY_LIFE_TIME) {
4612                                 if (valid > tval)
4613                                         valid -= tval;
4614                                 else
4615                                         valid = 0;
4616                         }
4617                 }
4618         } else {
4619                 preferred = INFINITY_LIFE_TIME;
4620                 valid = INFINITY_LIFE_TIME;
4621         }
4622
4623         if (!ipv6_addr_any(&ifa->peer_addr)) {
4624                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4625                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4626                         goto error;
4627         } else
4628                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4629                         goto error;
4630
4631         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4632                 goto error;
4633
4634         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4635                 goto error;
4636
4637         nlmsg_end(skb, nlh);
4638         return 0;
4639
4640 error:
4641         nlmsg_cancel(skb, nlh);
4642         return -EMSGSIZE;
4643 }
4644
4645 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4646                                 u32 portid, u32 seq, int event, u16 flags)
4647 {
4648         struct nlmsghdr  *nlh;
4649         u8 scope = RT_SCOPE_UNIVERSE;
4650         int ifindex = ifmca->idev->dev->ifindex;
4651
4652         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4653                 scope = RT_SCOPE_SITE;
4654
4655         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4656         if (!nlh)
4657                 return -EMSGSIZE;
4658
4659         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4660         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4661             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4662                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4663                 nlmsg_cancel(skb, nlh);
4664                 return -EMSGSIZE;
4665         }
4666
4667         nlmsg_end(skb, nlh);
4668         return 0;
4669 }
4670
4671 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4672                                 u32 portid, u32 seq, int event, unsigned int flags)
4673 {
4674         struct nlmsghdr  *nlh;
4675         u8 scope = RT_SCOPE_UNIVERSE;
4676         int ifindex = ifaca->aca_idev->dev->ifindex;
4677
4678         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4679                 scope = RT_SCOPE_SITE;
4680
4681         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4682         if (!nlh)
4683                 return -EMSGSIZE;
4684
4685         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4686         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4687             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4688                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4689                 nlmsg_cancel(skb, nlh);
4690                 return -EMSGSIZE;
4691         }
4692
4693         nlmsg_end(skb, nlh);
4694         return 0;
4695 }
4696
4697 enum addr_type_t {
4698         UNICAST_ADDR,
4699         MULTICAST_ADDR,
4700         ANYCAST_ADDR,
4701 };
4702
4703 /* called with rcu_read_lock() */
4704 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4705                           struct netlink_callback *cb, enum addr_type_t type,
4706                           int s_ip_idx, int *p_ip_idx)
4707 {
4708         struct ifmcaddr6 *ifmca;
4709         struct ifacaddr6 *ifaca;
4710         int err = 1;
4711         int ip_idx = *p_ip_idx;
4712
4713         read_lock_bh(&idev->lock);
4714         switch (type) {
4715         case UNICAST_ADDR: {
4716                 struct inet6_ifaddr *ifa;
4717
4718                 /* unicast address incl. temp addr */
4719                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4720                         if (++ip_idx < s_ip_idx)
4721                                 continue;
4722                         err = inet6_fill_ifaddr(skb, ifa,
4723                                                 NETLINK_CB(cb->skb).portid,
4724                                                 cb->nlh->nlmsg_seq,
4725                                                 RTM_NEWADDR,
4726                                                 NLM_F_MULTI);
4727                         if (err < 0)
4728                                 break;
4729                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4730                 }
4731                 break;
4732         }
4733         case MULTICAST_ADDR:
4734                 /* multicast address */
4735                 for (ifmca = idev->mc_list; ifmca;
4736                      ifmca = ifmca->next, ip_idx++) {
4737                         if (ip_idx < s_ip_idx)
4738                                 continue;
4739                         err = inet6_fill_ifmcaddr(skb, ifmca,
4740                                                   NETLINK_CB(cb->skb).portid,
4741                                                   cb->nlh->nlmsg_seq,
4742                                                   RTM_GETMULTICAST,
4743                                                   NLM_F_MULTI);
4744                         if (err < 0)
4745                                 break;
4746                 }
4747                 break;
4748         case ANYCAST_ADDR:
4749                 /* anycast address */
4750                 for (ifaca = idev->ac_list; ifaca;
4751                      ifaca = ifaca->aca_next, ip_idx++) {
4752                         if (ip_idx < s_ip_idx)
4753                                 continue;
4754                         err = inet6_fill_ifacaddr(skb, ifaca,
4755                                                   NETLINK_CB(cb->skb).portid,
4756                                                   cb->nlh->nlmsg_seq,
4757                                                   RTM_GETANYCAST,
4758                                                   NLM_F_MULTI);
4759                         if (err < 0)
4760                                 break;
4761                 }
4762                 break;
4763         default:
4764                 break;
4765         }
4766         read_unlock_bh(&idev->lock);
4767         *p_ip_idx = ip_idx;
4768         return err;
4769 }
4770
4771 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4772                            enum addr_type_t type)
4773 {
4774         struct net *net = sock_net(skb->sk);
4775         int h, s_h;
4776         int idx, ip_idx;
4777         int s_idx, s_ip_idx;
4778         struct net_device *dev;
4779         struct inet6_dev *idev;
4780         struct hlist_head *head;
4781
4782         s_h = cb->args[0];
4783         s_idx = idx = cb->args[1];
4784         s_ip_idx = ip_idx = cb->args[2];
4785
4786         rcu_read_lock();
4787         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4788         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4789                 idx = 0;
4790                 head = &net->dev_index_head[h];
4791                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4792                         if (idx < s_idx)
4793                                 goto cont;
4794                         if (h > s_h || idx > s_idx)
4795                                 s_ip_idx = 0;
4796                         ip_idx = 0;
4797                         idev = __in6_dev_get(dev);
4798                         if (!idev)
4799                                 goto cont;
4800
4801                         if (in6_dump_addrs(idev, skb, cb, type,
4802                                            s_ip_idx, &ip_idx) < 0)
4803                                 goto done;
4804 cont:
4805                         idx++;
4806                 }
4807         }
4808 done:
4809         rcu_read_unlock();
4810         cb->args[0] = h;
4811         cb->args[1] = idx;
4812         cb->args[2] = ip_idx;
4813
4814         return skb->len;
4815 }
4816
4817 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4818 {
4819         enum addr_type_t type = UNICAST_ADDR;
4820
4821         return inet6_dump_addr(skb, cb, type);
4822 }
4823
4824 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4825 {
4826         enum addr_type_t type = MULTICAST_ADDR;
4827
4828         return inet6_dump_addr(skb, cb, type);
4829 }
4830
4831
4832 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4833 {
4834         enum addr_type_t type = ANYCAST_ADDR;
4835
4836         return inet6_dump_addr(skb, cb, type);
4837 }
4838
4839 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4840 {
4841         struct net *net = sock_net(in_skb->sk);
4842         struct ifaddrmsg *ifm;
4843         struct nlattr *tb[IFA_MAX+1];
4844         struct in6_addr *addr = NULL, *peer;
4845         struct net_device *dev = NULL;
4846         struct inet6_ifaddr *ifa;
4847         struct sk_buff *skb;
4848         int err;
4849
4850         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4851         if (err < 0)
4852                 goto errout;
4853
4854         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4855         if (!addr) {
4856                 err = -EINVAL;
4857                 goto errout;
4858         }
4859
4860         ifm = nlmsg_data(nlh);
4861         if (ifm->ifa_index)
4862                 dev = __dev_get_by_index(net, ifm->ifa_index);
4863
4864         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4865         if (!ifa) {
4866                 err = -EADDRNOTAVAIL;
4867                 goto errout;
4868         }
4869
4870         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4871         if (!skb) {
4872                 err = -ENOBUFS;
4873                 goto errout_ifa;
4874         }
4875
4876         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4877                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4878         if (err < 0) {
4879                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4880                 WARN_ON(err == -EMSGSIZE);
4881                 kfree_skb(skb);
4882                 goto errout_ifa;
4883         }
4884         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4885 errout_ifa:
4886         in6_ifa_put(ifa);
4887 errout:
4888         return err;
4889 }
4890
4891 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4892 {
4893         struct sk_buff *skb;
4894         struct net *net = dev_net(ifa->idev->dev);
4895         int err = -ENOBUFS;
4896
4897         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4898         if (!skb)
4899                 goto errout;
4900
4901         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4902         if (err < 0) {
4903                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4904                 WARN_ON(err == -EMSGSIZE);
4905                 kfree_skb(skb);
4906                 goto errout;
4907         }
4908         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4909         return;
4910 errout:
4911         if (err < 0)
4912                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4913 }
4914
4915 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4916                                 __s32 *array, int bytes)
4917 {
4918         BUG_ON(bytes < (DEVCONF_MAX * 4));
4919
4920         memset(array, 0, bytes);
4921         array[DEVCONF_FORWARDING] = cnf->forwarding;
4922         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4923         array[DEVCONF_MTU6] = cnf->mtu6;
4924         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4925         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4926         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4927         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4928         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4929         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4930                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4931         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4932                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4933         array[DEVCONF_RTR_SOLICIT_DELAY] =
4934                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4935         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4936         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4937                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4938         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4939                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4940         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4941         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4942         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4943         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4944         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4945         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4946         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4947         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4948         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4949 #ifdef CONFIG_IPV6_ROUTER_PREF
4950         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4951         array[DEVCONF_RTR_PROBE_INTERVAL] =
4952                 jiffies_to_msecs(cnf->rtr_probe_interval);
4953 #ifdef CONFIG_IPV6_ROUTE_INFO
4954         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4955 #endif
4956 #endif
4957         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4958         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4959 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4960         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4961         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4962 #endif
4963 #ifdef CONFIG_IPV6_MROUTE
4964         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4965 #endif
4966         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4967         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4968         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4969         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4970         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4971         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4972         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4973         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4974         /* we omit DEVCONF_STABLE_SECRET for now */
4975         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4976         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4977         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4978         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4979         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
4980 #ifdef CONFIG_IPV6_SEG6_HMAC
4981         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
4982 #endif
4983         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
4984 }
4985
4986 static inline size_t inet6_ifla6_size(void)
4987 {
4988         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4989              + nla_total_size(sizeof(struct ifla_cacheinfo))
4990              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4991              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4992              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4993              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4994 }
4995
4996 static inline size_t inet6_if_nlmsg_size(void)
4997 {
4998         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4999                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5000                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5001                + nla_total_size(4) /* IFLA_MTU */
5002                + nla_total_size(4) /* IFLA_LINK */
5003                + nla_total_size(1) /* IFLA_OPERSTATE */
5004                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5005 }
5006
5007 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5008                                         int bytes)
5009 {
5010         int i;
5011         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5012         BUG_ON(pad < 0);
5013
5014         /* Use put_unaligned() because stats may not be aligned for u64. */
5015         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5016         for (i = 1; i < ICMP6_MIB_MAX; i++)
5017                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5018
5019         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5020 }
5021
5022 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5023                                         int bytes, size_t syncpoff)
5024 {
5025         int i, c;
5026         u64 buff[IPSTATS_MIB_MAX];
5027         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5028
5029         BUG_ON(pad < 0);
5030
5031         memset(buff, 0, sizeof(buff));
5032         buff[0] = IPSTATS_MIB_MAX;
5033
5034         for_each_possible_cpu(c) {
5035                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5036                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5037         }
5038
5039         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5040         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5041 }
5042
5043 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5044                              int bytes)
5045 {
5046         switch (attrtype) {
5047         case IFLA_INET6_STATS:
5048                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5049                                      offsetof(struct ipstats_mib, syncp));
5050                 break;
5051         case IFLA_INET6_ICMP6STATS:
5052                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5053                 break;
5054         }
5055 }
5056
5057 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5058                                   u32 ext_filter_mask)
5059 {
5060         struct nlattr *nla;
5061         struct ifla_cacheinfo ci;
5062
5063         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5064                 goto nla_put_failure;
5065         ci.max_reasm_len = IPV6_MAXPLEN;
5066         ci.tstamp = cstamp_delta(idev->tstamp);
5067         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5068         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5069         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5070                 goto nla_put_failure;
5071         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5072         if (!nla)
5073                 goto nla_put_failure;
5074         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5075
5076         /* XXX - MC not implemented */
5077
5078         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5079                 return 0;
5080
5081         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5082         if (!nla)
5083                 goto nla_put_failure;
5084         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5085
5086         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5087         if (!nla)
5088                 goto nla_put_failure;
5089         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5090
5091         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5092         if (!nla)
5093                 goto nla_put_failure;
5094
5095         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5096                 goto nla_put_failure;
5097
5098         read_lock_bh(&idev->lock);
5099         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5100         read_unlock_bh(&idev->lock);
5101
5102         return 0;
5103
5104 nla_put_failure:
5105         return -EMSGSIZE;
5106 }
5107
5108 static size_t inet6_get_link_af_size(const struct net_device *dev,
5109                                      u32 ext_filter_mask)
5110 {
5111         if (!__in6_dev_get(dev))
5112                 return 0;
5113
5114         return inet6_ifla6_size();
5115 }
5116
5117 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5118                               u32 ext_filter_mask)
5119 {
5120         struct inet6_dev *idev = __in6_dev_get(dev);
5121
5122         if (!idev)
5123                 return -ENODATA;
5124
5125         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5126                 return -EMSGSIZE;
5127
5128         return 0;
5129 }
5130
5131 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5132 {
5133         struct inet6_ifaddr *ifp;
5134         struct net_device *dev = idev->dev;
5135         bool clear_token, update_rs = false;
5136         struct in6_addr ll_addr;
5137
5138         ASSERT_RTNL();
5139
5140         if (!token)
5141                 return -EINVAL;
5142         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5143                 return -EINVAL;
5144         if (!ipv6_accept_ra(idev))
5145                 return -EINVAL;
5146         if (idev->cnf.rtr_solicits == 0)
5147                 return -EINVAL;
5148
5149         write_lock_bh(&idev->lock);
5150
5151         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5152         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5153
5154         write_unlock_bh(&idev->lock);
5155
5156         clear_token = ipv6_addr_any(token);
5157         if (clear_token)
5158                 goto update_lft;
5159
5160         if (!idev->dead && (idev->if_flags & IF_READY) &&
5161             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5162                              IFA_F_OPTIMISTIC)) {
5163                 /* If we're not ready, then normal ifup will take care
5164                  * of this. Otherwise, we need to request our rs here.
5165                  */
5166                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5167                 update_rs = true;
5168         }
5169
5170 update_lft:
5171         write_lock_bh(&idev->lock);
5172
5173         if (update_rs) {
5174                 idev->if_flags |= IF_RS_SENT;
5175                 idev->rs_interval = rfc3315_s14_backoff_init(
5176                         idev->cnf.rtr_solicit_interval);
5177                 idev->rs_probes = 1;
5178                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5179         }
5180
5181         /* Well, that's kinda nasty ... */
5182         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5183                 spin_lock(&ifp->lock);
5184                 if (ifp->tokenized) {
5185                         ifp->valid_lft = 0;
5186                         ifp->prefered_lft = 0;
5187                 }
5188                 spin_unlock(&ifp->lock);
5189         }
5190
5191         write_unlock_bh(&idev->lock);
5192         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5193         addrconf_verify_rtnl();
5194         return 0;
5195 }
5196
5197 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5198         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5199         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5200 };
5201
5202 static int inet6_validate_link_af(const struct net_device *dev,
5203                                   const struct nlattr *nla)
5204 {
5205         struct nlattr *tb[IFLA_INET6_MAX + 1];
5206
5207         if (dev && !__in6_dev_get(dev))
5208                 return -EAFNOSUPPORT;
5209
5210         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5211 }
5212
5213 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5214 {
5215         int err = -EINVAL;
5216         struct inet6_dev *idev = __in6_dev_get(dev);
5217         struct nlattr *tb[IFLA_INET6_MAX + 1];
5218
5219         if (!idev)
5220                 return -EAFNOSUPPORT;
5221
5222         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5223                 BUG();
5224
5225         if (tb[IFLA_INET6_TOKEN]) {
5226                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5227                 if (err)
5228                         return err;
5229         }
5230
5231         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5232                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5233
5234                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5235                     mode != IN6_ADDR_GEN_MODE_NONE &&
5236                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5237                     mode != IN6_ADDR_GEN_MODE_RANDOM)
5238                         return -EINVAL;
5239
5240                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5241                     !idev->cnf.stable_secret.initialized &&
5242                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5243                         return -EINVAL;
5244
5245                 idev->addr_gen_mode = mode;
5246                 err = 0;
5247         }
5248
5249         return err;
5250 }
5251
5252 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5253                              u32 portid, u32 seq, int event, unsigned int flags)
5254 {
5255         struct net_device *dev = idev->dev;
5256         struct ifinfomsg *hdr;
5257         struct nlmsghdr *nlh;
5258         void *protoinfo;
5259
5260         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5261         if (!nlh)
5262                 return -EMSGSIZE;
5263
5264         hdr = nlmsg_data(nlh);
5265         hdr->ifi_family = AF_INET6;
5266         hdr->__ifi_pad = 0;
5267         hdr->ifi_type = dev->type;
5268         hdr->ifi_index = dev->ifindex;
5269         hdr->ifi_flags = dev_get_flags(dev);
5270         hdr->ifi_change = 0;
5271
5272         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5273             (dev->addr_len &&
5274              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5275             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5276             (dev->ifindex != dev_get_iflink(dev) &&
5277              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5278             nla_put_u8(skb, IFLA_OPERSTATE,
5279                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5280                 goto nla_put_failure;
5281         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5282         if (!protoinfo)
5283                 goto nla_put_failure;
5284
5285         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5286                 goto nla_put_failure;
5287
5288         nla_nest_end(skb, protoinfo);
5289         nlmsg_end(skb, nlh);
5290         return 0;
5291
5292 nla_put_failure:
5293         nlmsg_cancel(skb, nlh);
5294         return -EMSGSIZE;
5295 }
5296
5297 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5298 {
5299         struct net *net = sock_net(skb->sk);
5300         int h, s_h;
5301         int idx = 0, s_idx;
5302         struct net_device *dev;
5303         struct inet6_dev *idev;
5304         struct hlist_head *head;
5305
5306         s_h = cb->args[0];
5307         s_idx = cb->args[1];
5308
5309         rcu_read_lock();
5310         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5311                 idx = 0;
5312                 head = &net->dev_index_head[h];
5313                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5314                         if (idx < s_idx)
5315                                 goto cont;
5316                         idev = __in6_dev_get(dev);
5317                         if (!idev)
5318                                 goto cont;
5319                         if (inet6_fill_ifinfo(skb, idev,
5320                                               NETLINK_CB(cb->skb).portid,
5321                                               cb->nlh->nlmsg_seq,
5322                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5323                                 goto out;
5324 cont:
5325                         idx++;
5326                 }
5327         }
5328 out:
5329         rcu_read_unlock();
5330         cb->args[1] = idx;
5331         cb->args[0] = h;
5332
5333         return skb->len;
5334 }
5335
5336 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5337 {
5338         struct sk_buff *skb;
5339         struct net *net = dev_net(idev->dev);
5340         int err = -ENOBUFS;
5341
5342         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5343         if (!skb)
5344                 goto errout;
5345
5346         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5347         if (err < 0) {
5348                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5349                 WARN_ON(err == -EMSGSIZE);
5350                 kfree_skb(skb);
5351                 goto errout;
5352         }
5353         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5354         return;
5355 errout:
5356         if (err < 0)
5357                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5358 }
5359
5360 static inline size_t inet6_prefix_nlmsg_size(void)
5361 {
5362         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5363                + nla_total_size(sizeof(struct in6_addr))
5364                + nla_total_size(sizeof(struct prefix_cacheinfo));
5365 }
5366
5367 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5368                              struct prefix_info *pinfo, u32 portid, u32 seq,
5369                              int event, unsigned int flags)
5370 {
5371         struct prefixmsg *pmsg;
5372         struct nlmsghdr *nlh;
5373         struct prefix_cacheinfo ci;
5374
5375         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5376         if (!nlh)
5377                 return -EMSGSIZE;
5378
5379         pmsg = nlmsg_data(nlh);
5380         pmsg->prefix_family = AF_INET6;
5381         pmsg->prefix_pad1 = 0;
5382         pmsg->prefix_pad2 = 0;
5383         pmsg->prefix_ifindex = idev->dev->ifindex;
5384         pmsg->prefix_len = pinfo->prefix_len;
5385         pmsg->prefix_type = pinfo->type;
5386         pmsg->prefix_pad3 = 0;
5387         pmsg->prefix_flags = 0;
5388         if (pinfo->onlink)
5389                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5390         if (pinfo->autoconf)
5391                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5392
5393         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5394                 goto nla_put_failure;
5395         ci.preferred_time = ntohl(pinfo->prefered);
5396         ci.valid_time = ntohl(pinfo->valid);
5397         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5398                 goto nla_put_failure;
5399         nlmsg_end(skb, nlh);
5400         return 0;
5401
5402 nla_put_failure:
5403         nlmsg_cancel(skb, nlh);
5404         return -EMSGSIZE;
5405 }
5406
5407 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5408                          struct prefix_info *pinfo)
5409 {
5410         struct sk_buff *skb;
5411         struct net *net = dev_net(idev->dev);
5412         int err = -ENOBUFS;
5413
5414         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5415         if (!skb)
5416                 goto errout;
5417
5418         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5419         if (err < 0) {
5420                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5421                 WARN_ON(err == -EMSGSIZE);
5422                 kfree_skb(skb);
5423                 goto errout;
5424         }
5425         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5426         return;
5427 errout:
5428         if (err < 0)
5429                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5430 }
5431
5432 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5433 {
5434         struct net *net = dev_net(ifp->idev->dev);
5435
5436         if (event)
5437                 ASSERT_RTNL();
5438
5439         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5440
5441         switch (event) {
5442         case RTM_NEWADDR:
5443                 /*
5444                  * If the address was optimistic
5445                  * we inserted the route at the start of
5446                  * our DAD process, so we don't need
5447                  * to do it again
5448                  */
5449                 if (!(ifp->rt->rt6i_node))
5450                         ip6_ins_rt(ifp->rt);
5451                 if (ifp->idev->cnf.forwarding)
5452                         addrconf_join_anycast(ifp);
5453                 if (!ipv6_addr_any(&ifp->peer_addr))
5454                         addrconf_prefix_route(&ifp->peer_addr, 128,
5455                                               ifp->idev->dev, 0, 0);
5456                 break;
5457         case RTM_DELADDR:
5458                 if (ifp->idev->cnf.forwarding)
5459                         addrconf_leave_anycast(ifp);
5460                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5461                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5462                         struct rt6_info *rt;
5463
5464                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5465                                                        ifp->idev->dev, 0, 0);
5466                         if (rt)
5467                                 ip6_del_rt(rt);
5468                 }
5469                 if (ifp->rt) {
5470                         dst_hold(&ifp->rt->dst);
5471                         ip6_del_rt(ifp->rt);
5472                 }
5473                 rt_genid_bump_ipv6(net);
5474                 break;
5475         }
5476         atomic_inc(&net->ipv6.dev_addr_genid);
5477 }
5478
5479 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5480 {
5481         rcu_read_lock_bh();
5482         if (likely(ifp->idev->dead == 0))
5483                 __ipv6_ifa_notify(event, ifp);
5484         rcu_read_unlock_bh();
5485 }
5486
5487 #ifdef CONFIG_SYSCTL
5488
5489 static
5490 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5491                            void __user *buffer, size_t *lenp, loff_t *ppos)
5492 {
5493         int *valp = ctl->data;
5494         int val = *valp;
5495         loff_t pos = *ppos;
5496         struct ctl_table lctl;
5497         int ret;
5498
5499         /*
5500          * ctl->data points to idev->cnf.forwarding, we should
5501          * not modify it until we get the rtnl lock.
5502          */
5503         lctl = *ctl;
5504         lctl.data = &val;
5505
5506         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5507
5508         if (write)
5509                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5510         if (ret)
5511                 *ppos = pos;
5512         return ret;
5513 }
5514
5515 static
5516 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5517                         void __user *buffer, size_t *lenp, loff_t *ppos)
5518 {
5519         struct inet6_dev *idev = ctl->extra1;
5520         int min_mtu = IPV6_MIN_MTU;
5521         struct ctl_table lctl;
5522
5523         lctl = *ctl;
5524         lctl.extra1 = &min_mtu;
5525         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5526
5527         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5528 }
5529
5530 static void dev_disable_change(struct inet6_dev *idev)
5531 {
5532         struct netdev_notifier_info info;
5533
5534         if (!idev || !idev->dev)
5535                 return;
5536
5537         netdev_notifier_info_init(&info, idev->dev);
5538         if (idev->cnf.disable_ipv6)
5539                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5540         else
5541                 addrconf_notify(NULL, NETDEV_UP, &info);
5542 }
5543
5544 static void addrconf_disable_change(struct net *net, __s32 newf)
5545 {
5546         struct net_device *dev;
5547         struct inet6_dev *idev;
5548
5549         for_each_netdev(net, dev) {
5550                 idev = __in6_dev_get(dev);
5551                 if (idev) {
5552                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5553                         idev->cnf.disable_ipv6 = newf;
5554                         if (changed)
5555                                 dev_disable_change(idev);
5556                 }
5557         }
5558 }
5559
5560 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5561 {
5562         struct net *net;
5563         int old;
5564
5565         if (!rtnl_trylock())
5566                 return restart_syscall();
5567
5568         net = (struct net *)table->extra2;
5569         old = *p;
5570         *p = newf;
5571
5572         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5573                 rtnl_unlock();
5574                 return 0;
5575         }
5576
5577         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5578                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5579                 addrconf_disable_change(net, newf);
5580         } else if ((!newf) ^ (!old))
5581                 dev_disable_change((struct inet6_dev *)table->extra1);
5582
5583         rtnl_unlock();
5584         return 0;
5585 }
5586
5587 static
5588 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5589                             void __user *buffer, size_t *lenp, loff_t *ppos)
5590 {
5591         int *valp = ctl->data;
5592         int val = *valp;
5593         loff_t pos = *ppos;
5594         struct ctl_table lctl;
5595         int ret;
5596
5597         /*
5598          * ctl->data points to idev->cnf.disable_ipv6, we should
5599          * not modify it until we get the rtnl lock.
5600          */
5601         lctl = *ctl;
5602         lctl.data = &val;
5603
5604         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5605
5606         if (write)
5607                 ret = addrconf_disable_ipv6(ctl, valp, val);
5608         if (ret)
5609                 *ppos = pos;
5610         return ret;
5611 }
5612
5613 static
5614 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5615                               void __user *buffer, size_t *lenp, loff_t *ppos)
5616 {
5617         int *valp = ctl->data;
5618         int ret;
5619         int old, new;
5620
5621         old = *valp;
5622         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5623         new = *valp;
5624
5625         if (write && old != new) {
5626                 struct net *net = ctl->extra2;
5627
5628                 if (!rtnl_trylock())
5629                         return restart_syscall();
5630
5631                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5632                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5633                                                      NETCONFA_IFINDEX_DEFAULT,
5634                                                      net->ipv6.devconf_dflt);
5635                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5636                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5637                                                      NETCONFA_IFINDEX_ALL,
5638                                                      net->ipv6.devconf_all);
5639                 else {
5640                         struct inet6_dev *idev = ctl->extra1;
5641
5642                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5643                                                      idev->dev->ifindex,
5644                                                      &idev->cnf);
5645                 }
5646                 rtnl_unlock();
5647         }
5648
5649         return ret;
5650 }
5651
5652 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5653                                          void __user *buffer, size_t *lenp,
5654                                          loff_t *ppos)
5655 {
5656         int err;
5657         struct in6_addr addr;
5658         char str[IPV6_MAX_STRLEN];
5659         struct ctl_table lctl = *ctl;
5660         struct net *net = ctl->extra2;
5661         struct ipv6_stable_secret *secret = ctl->data;
5662
5663         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5664                 return -EIO;
5665
5666         lctl.maxlen = IPV6_MAX_STRLEN;
5667         lctl.data = str;
5668
5669         if (!rtnl_trylock())
5670                 return restart_syscall();
5671
5672         if (!write && !secret->initialized) {
5673                 err = -EIO;
5674                 goto out;
5675         }
5676
5677         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5678         if (err >= sizeof(str)) {
5679                 err = -EIO;
5680                 goto out;
5681         }
5682
5683         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5684         if (err || !write)
5685                 goto out;
5686
5687         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5688                 err = -EIO;
5689                 goto out;
5690         }
5691
5692         secret->initialized = true;
5693         secret->secret = addr;
5694
5695         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5696                 struct net_device *dev;
5697
5698                 for_each_netdev(net, dev) {
5699                         struct inet6_dev *idev = __in6_dev_get(dev);
5700
5701                         if (idev) {
5702                                 idev->addr_gen_mode =
5703                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5704                         }
5705                 }
5706         } else {
5707                 struct inet6_dev *idev = ctl->extra1;
5708
5709                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5710         }
5711
5712 out:
5713         rtnl_unlock();
5714
5715         return err;
5716 }
5717
5718 static
5719 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5720                                                 int write,
5721                                                 void __user *buffer,
5722                                                 size_t *lenp,
5723                                                 loff_t *ppos)
5724 {
5725         int *valp = ctl->data;
5726         int val = *valp;
5727         loff_t pos = *ppos;
5728         struct ctl_table lctl;
5729         int ret;
5730
5731         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5732          * we should not modify it until we get the rtnl lock.
5733          */
5734         lctl = *ctl;
5735         lctl.data = &val;
5736
5737         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5738
5739         if (write)
5740                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5741         if (ret)
5742                 *ppos = pos;
5743         return ret;
5744 }
5745
5746 static int minus_one = -1;
5747 static const int one = 1;
5748 static const int two_five_five = 255;
5749
5750 static const struct ctl_table addrconf_sysctl[] = {
5751         {
5752                 .procname       = "forwarding",
5753                 .data           = &ipv6_devconf.forwarding,
5754                 .maxlen         = sizeof(int),
5755                 .mode           = 0644,
5756                 .proc_handler   = addrconf_sysctl_forward,
5757         },
5758         {
5759                 .procname       = "hop_limit",
5760                 .data           = &ipv6_devconf.hop_limit,
5761                 .maxlen         = sizeof(int),
5762                 .mode           = 0644,
5763                 .proc_handler   = proc_dointvec_minmax,
5764                 .extra1         = (void *)&one,
5765                 .extra2         = (void *)&two_five_five,
5766         },
5767         {
5768                 .procname       = "mtu",
5769                 .data           = &ipv6_devconf.mtu6,
5770                 .maxlen         = sizeof(int),
5771                 .mode           = 0644,
5772                 .proc_handler   = addrconf_sysctl_mtu,
5773         },
5774         {
5775                 .procname       = "accept_ra",
5776                 .data           = &ipv6_devconf.accept_ra,
5777                 .maxlen         = sizeof(int),
5778                 .mode           = 0644,
5779                 .proc_handler   = proc_dointvec,
5780         },
5781         {
5782                 .procname       = "accept_redirects",
5783                 .data           = &ipv6_devconf.accept_redirects,
5784                 .maxlen         = sizeof(int),
5785                 .mode           = 0644,
5786                 .proc_handler   = proc_dointvec,
5787         },
5788         {
5789                 .procname       = "autoconf",
5790                 .data           = &ipv6_devconf.autoconf,
5791                 .maxlen         = sizeof(int),
5792                 .mode           = 0644,
5793                 .proc_handler   = proc_dointvec,
5794         },
5795         {
5796                 .procname       = "dad_transmits",
5797                 .data           = &ipv6_devconf.dad_transmits,
5798                 .maxlen         = sizeof(int),
5799                 .mode           = 0644,
5800                 .proc_handler   = proc_dointvec,
5801         },
5802         {
5803                 .procname       = "router_solicitations",
5804                 .data           = &ipv6_devconf.rtr_solicits,
5805                 .maxlen         = sizeof(int),
5806                 .mode           = 0644,
5807                 .proc_handler   = proc_dointvec_minmax,
5808                 .extra1         = &minus_one,
5809         },
5810         {
5811                 .procname       = "router_solicitation_interval",
5812                 .data           = &ipv6_devconf.rtr_solicit_interval,
5813                 .maxlen         = sizeof(int),
5814                 .mode           = 0644,
5815                 .proc_handler   = proc_dointvec_jiffies,
5816         },
5817         {
5818                 .procname       = "router_solicitation_max_interval",
5819                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
5820                 .maxlen         = sizeof(int),
5821                 .mode           = 0644,
5822                 .proc_handler   = proc_dointvec_jiffies,
5823         },
5824         {
5825                 .procname       = "router_solicitation_delay",
5826                 .data           = &ipv6_devconf.rtr_solicit_delay,
5827                 .maxlen         = sizeof(int),
5828                 .mode           = 0644,
5829                 .proc_handler   = proc_dointvec_jiffies,
5830         },
5831         {
5832                 .procname       = "force_mld_version",
5833                 .data           = &ipv6_devconf.force_mld_version,
5834                 .maxlen         = sizeof(int),
5835                 .mode           = 0644,
5836                 .proc_handler   = proc_dointvec,
5837         },
5838         {
5839                 .procname       = "mldv1_unsolicited_report_interval",
5840                 .data           =
5841                         &ipv6_devconf.mldv1_unsolicited_report_interval,
5842                 .maxlen         = sizeof(int),
5843                 .mode           = 0644,
5844                 .proc_handler   = proc_dointvec_ms_jiffies,
5845         },
5846         {
5847                 .procname       = "mldv2_unsolicited_report_interval",
5848                 .data           =
5849                         &ipv6_devconf.mldv2_unsolicited_report_interval,
5850                 .maxlen         = sizeof(int),
5851                 .mode           = 0644,
5852                 .proc_handler   = proc_dointvec_ms_jiffies,
5853         },
5854         {
5855                 .procname       = "use_tempaddr",
5856                 .data           = &ipv6_devconf.use_tempaddr,
5857                 .maxlen         = sizeof(int),
5858                 .mode           = 0644,
5859                 .proc_handler   = proc_dointvec,
5860         },
5861         {
5862                 .procname       = "temp_valid_lft",
5863                 .data           = &ipv6_devconf.temp_valid_lft,
5864                 .maxlen         = sizeof(int),
5865                 .mode           = 0644,
5866                 .proc_handler   = proc_dointvec,
5867         },
5868         {
5869                 .procname       = "temp_prefered_lft",
5870                 .data           = &ipv6_devconf.temp_prefered_lft,
5871                 .maxlen         = sizeof(int),
5872                 .mode           = 0644,
5873                 .proc_handler   = proc_dointvec,
5874         },
5875         {
5876                 .procname       = "regen_max_retry",
5877                 .data           = &ipv6_devconf.regen_max_retry,
5878                 .maxlen         = sizeof(int),
5879                 .mode           = 0644,
5880                 .proc_handler   = proc_dointvec,
5881         },
5882         {
5883                 .procname       = "max_desync_factor",
5884                 .data           = &ipv6_devconf.max_desync_factor,
5885                 .maxlen         = sizeof(int),
5886                 .mode           = 0644,
5887                 .proc_handler   = proc_dointvec,
5888         },
5889         {
5890                 .procname       = "max_addresses",
5891                 .data           = &ipv6_devconf.max_addresses,
5892                 .maxlen         = sizeof(int),
5893                 .mode           = 0644,
5894                 .proc_handler   = proc_dointvec,
5895         },
5896         {
5897                 .procname       = "accept_ra_defrtr",
5898                 .data           = &ipv6_devconf.accept_ra_defrtr,
5899                 .maxlen         = sizeof(int),
5900                 .mode           = 0644,
5901                 .proc_handler   = proc_dointvec,
5902         },
5903         {
5904                 .procname       = "accept_ra_min_hop_limit",
5905                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5906                 .maxlen         = sizeof(int),
5907                 .mode           = 0644,
5908                 .proc_handler   = proc_dointvec,
5909         },
5910         {
5911                 .procname       = "accept_ra_pinfo",
5912                 .data           = &ipv6_devconf.accept_ra_pinfo,
5913                 .maxlen         = sizeof(int),
5914                 .mode           = 0644,
5915                 .proc_handler   = proc_dointvec,
5916         },
5917 #ifdef CONFIG_IPV6_ROUTER_PREF
5918         {
5919                 .procname       = "accept_ra_rtr_pref",
5920                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
5921                 .maxlen         = sizeof(int),
5922                 .mode           = 0644,
5923                 .proc_handler   = proc_dointvec,
5924         },
5925         {
5926                 .procname       = "router_probe_interval",
5927                 .data           = &ipv6_devconf.rtr_probe_interval,
5928                 .maxlen         = sizeof(int),
5929                 .mode           = 0644,
5930                 .proc_handler   = proc_dointvec_jiffies,
5931         },
5932 #ifdef CONFIG_IPV6_ROUTE_INFO
5933         {
5934                 .procname       = "accept_ra_rt_info_max_plen",
5935                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5936                 .maxlen         = sizeof(int),
5937                 .mode           = 0644,
5938                 .proc_handler   = proc_dointvec,
5939         },
5940 #endif
5941 #endif
5942         {
5943                 .procname       = "proxy_ndp",
5944                 .data           = &ipv6_devconf.proxy_ndp,
5945                 .maxlen         = sizeof(int),
5946                 .mode           = 0644,
5947                 .proc_handler   = addrconf_sysctl_proxy_ndp,
5948         },
5949         {
5950                 .procname       = "accept_source_route",
5951                 .data           = &ipv6_devconf.accept_source_route,
5952                 .maxlen         = sizeof(int),
5953                 .mode           = 0644,
5954                 .proc_handler   = proc_dointvec,
5955         },
5956 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5957         {
5958                 .procname       = "optimistic_dad",
5959                 .data           = &ipv6_devconf.optimistic_dad,
5960                 .maxlen         = sizeof(int),
5961                 .mode           = 0644,
5962                 .proc_handler   = proc_dointvec,
5963         },
5964         {
5965                 .procname       = "use_optimistic",
5966                 .data           = &ipv6_devconf.use_optimistic,
5967                 .maxlen         = sizeof(int),
5968                 .mode           = 0644,
5969                 .proc_handler   = proc_dointvec,
5970         },
5971 #endif
5972 #ifdef CONFIG_IPV6_MROUTE
5973         {
5974                 .procname       = "mc_forwarding",
5975                 .data           = &ipv6_devconf.mc_forwarding,
5976                 .maxlen         = sizeof(int),
5977                 .mode           = 0444,
5978                 .proc_handler   = proc_dointvec,
5979         },
5980 #endif
5981         {
5982                 .procname       = "disable_ipv6",
5983                 .data           = &ipv6_devconf.disable_ipv6,
5984                 .maxlen         = sizeof(int),
5985                 .mode           = 0644,
5986                 .proc_handler   = addrconf_sysctl_disable,
5987         },
5988         {
5989                 .procname       = "accept_dad",
5990                 .data           = &ipv6_devconf.accept_dad,
5991                 .maxlen         = sizeof(int),
5992                 .mode           = 0644,
5993                 .proc_handler   = proc_dointvec,
5994         },
5995         {
5996                 .procname       = "force_tllao",
5997                 .data           = &ipv6_devconf.force_tllao,
5998                 .maxlen         = sizeof(int),
5999                 .mode           = 0644,
6000                 .proc_handler   = proc_dointvec
6001         },
6002         {
6003                 .procname       = "ndisc_notify",
6004                 .data           = &ipv6_devconf.ndisc_notify,
6005                 .maxlen         = sizeof(int),
6006                 .mode           = 0644,
6007                 .proc_handler   = proc_dointvec
6008         },
6009         {
6010                 .procname       = "suppress_frag_ndisc",
6011                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6012                 .maxlen         = sizeof(int),
6013                 .mode           = 0644,
6014                 .proc_handler   = proc_dointvec
6015         },
6016         {
6017                 .procname       = "accept_ra_from_local",
6018                 .data           = &ipv6_devconf.accept_ra_from_local,
6019                 .maxlen         = sizeof(int),
6020                 .mode           = 0644,
6021                 .proc_handler   = proc_dointvec,
6022         },
6023         {
6024                 .procname       = "accept_ra_mtu",
6025                 .data           = &ipv6_devconf.accept_ra_mtu,
6026                 .maxlen         = sizeof(int),
6027                 .mode           = 0644,
6028                 .proc_handler   = proc_dointvec,
6029         },
6030         {
6031                 .procname       = "stable_secret",
6032                 .data           = &ipv6_devconf.stable_secret,
6033                 .maxlen         = IPV6_MAX_STRLEN,
6034                 .mode           = 0600,
6035                 .proc_handler   = addrconf_sysctl_stable_secret,
6036         },
6037         {
6038                 .procname       = "use_oif_addrs_only",
6039                 .data           = &ipv6_devconf.use_oif_addrs_only,
6040                 .maxlen         = sizeof(int),
6041                 .mode           = 0644,
6042                 .proc_handler   = proc_dointvec,
6043         },
6044         {
6045                 .procname       = "ignore_routes_with_linkdown",
6046                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6047                 .maxlen         = sizeof(int),
6048                 .mode           = 0644,
6049                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6050         },
6051         {
6052                 .procname       = "drop_unicast_in_l2_multicast",
6053                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6054                 .maxlen         = sizeof(int),
6055                 .mode           = 0644,
6056                 .proc_handler   = proc_dointvec,
6057         },
6058         {
6059                 .procname       = "drop_unsolicited_na",
6060                 .data           = &ipv6_devconf.drop_unsolicited_na,
6061                 .maxlen         = sizeof(int),
6062                 .mode           = 0644,
6063                 .proc_handler   = proc_dointvec,
6064         },
6065         {
6066                 .procname       = "keep_addr_on_down",
6067                 .data           = &ipv6_devconf.keep_addr_on_down,
6068                 .maxlen         = sizeof(int),
6069                 .mode           = 0644,
6070                 .proc_handler   = proc_dointvec,
6071
6072         },
6073         {
6074                 .procname       = "seg6_enabled",
6075                 .data           = &ipv6_devconf.seg6_enabled,
6076                 .maxlen         = sizeof(int),
6077                 .mode           = 0644,
6078                 .proc_handler   = proc_dointvec,
6079         },
6080 #ifdef CONFIG_IPV6_SEG6_HMAC
6081         {
6082                 .procname       = "seg6_require_hmac",
6083                 .data           = &ipv6_devconf.seg6_require_hmac,
6084                 .maxlen         = sizeof(int),
6085                 .mode           = 0644,
6086                 .proc_handler   = proc_dointvec,
6087         },
6088 #endif
6089         {
6090                 .procname       = "enhanced_dad",
6091                 .data           = &ipv6_devconf.enhanced_dad,
6092                 .maxlen         = sizeof(int),
6093                 .mode           = 0644,
6094                 .proc_handler   = proc_dointvec,
6095         },
6096         {
6097                 /* sentinel */
6098         }
6099 };
6100
6101 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6102                 struct inet6_dev *idev, struct ipv6_devconf *p)
6103 {
6104         int i, ifindex;
6105         struct ctl_table *table;
6106         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6107
6108         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6109         if (!table)
6110                 goto out;
6111
6112         for (i = 0; table[i].data; i++) {
6113                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6114                 /* If one of these is already set, then it is not safe to
6115                  * overwrite either of them: this makes proc_dointvec_minmax
6116                  * usable.
6117                  */
6118                 if (!table[i].extra1 && !table[i].extra2) {
6119                         table[i].extra1 = idev; /* embedded; no ref */
6120                         table[i].extra2 = net;
6121                 }
6122         }
6123
6124         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6125
6126         p->sysctl_header = register_net_sysctl(net, path, table);
6127         if (!p->sysctl_header)
6128                 goto free;
6129
6130         if (!strcmp(dev_name, "all"))
6131                 ifindex = NETCONFA_IFINDEX_ALL;
6132         else if (!strcmp(dev_name, "default"))
6133                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6134         else
6135                 ifindex = idev->dev->ifindex;
6136         inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6137         return 0;
6138
6139 free:
6140         kfree(table);
6141 out:
6142         return -ENOBUFS;
6143 }
6144
6145 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6146 {
6147         struct ctl_table *table;
6148
6149         if (!p->sysctl_header)
6150                 return;
6151
6152         table = p->sysctl_header->ctl_table_arg;
6153         unregister_net_sysctl_table(p->sysctl_header);
6154         p->sysctl_header = NULL;
6155         kfree(table);
6156 }
6157
6158 static int addrconf_sysctl_register(struct inet6_dev *idev)
6159 {
6160         int err;
6161
6162         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6163                 return -EINVAL;
6164
6165         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6166                                     &ndisc_ifinfo_sysctl_change);
6167         if (err)
6168                 return err;
6169         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6170                                          idev, &idev->cnf);
6171         if (err)
6172                 neigh_sysctl_unregister(idev->nd_parms);
6173
6174         return err;
6175 }
6176
6177 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6178 {
6179         __addrconf_sysctl_unregister(&idev->cnf);
6180         neigh_sysctl_unregister(idev->nd_parms);
6181 }
6182
6183
6184 #endif
6185
6186 static int __net_init addrconf_init_net(struct net *net)
6187 {
6188         int err = -ENOMEM;
6189         struct ipv6_devconf *all, *dflt;
6190
6191         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6192         if (!all)
6193                 goto err_alloc_all;
6194
6195         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6196         if (!dflt)
6197                 goto err_alloc_dflt;
6198
6199         /* these will be inherited by all namespaces */
6200         dflt->autoconf = ipv6_defaults.autoconf;
6201         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6202
6203         dflt->stable_secret.initialized = false;
6204         all->stable_secret.initialized = false;
6205
6206         net->ipv6.devconf_all = all;
6207         net->ipv6.devconf_dflt = dflt;
6208
6209 #ifdef CONFIG_SYSCTL
6210         err = __addrconf_sysctl_register(net, "all", NULL, all);
6211         if (err < 0)
6212                 goto err_reg_all;
6213
6214         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6215         if (err < 0)
6216                 goto err_reg_dflt;
6217 #endif
6218         return 0;
6219
6220 #ifdef CONFIG_SYSCTL
6221 err_reg_dflt:
6222         __addrconf_sysctl_unregister(all);
6223 err_reg_all:
6224         kfree(dflt);
6225 #endif
6226 err_alloc_dflt:
6227         kfree(all);
6228 err_alloc_all:
6229         return err;
6230 }
6231
6232 static void __net_exit addrconf_exit_net(struct net *net)
6233 {
6234 #ifdef CONFIG_SYSCTL
6235         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6236         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6237 #endif
6238         kfree(net->ipv6.devconf_dflt);
6239         kfree(net->ipv6.devconf_all);
6240 }
6241
6242 static struct pernet_operations addrconf_ops = {
6243         .init = addrconf_init_net,
6244         .exit = addrconf_exit_net,
6245 };
6246
6247 static struct rtnl_af_ops inet6_ops __read_mostly = {
6248         .family           = AF_INET6,
6249         .fill_link_af     = inet6_fill_link_af,
6250         .get_link_af_size = inet6_get_link_af_size,
6251         .validate_link_af = inet6_validate_link_af,
6252         .set_link_af      = inet6_set_link_af,
6253 };
6254
6255 /*
6256  *      Init / cleanup code
6257  */
6258
6259 int __init addrconf_init(void)
6260 {
6261         struct inet6_dev *idev;
6262         int i, err;
6263
6264         err = ipv6_addr_label_init();
6265         if (err < 0) {
6266                 pr_crit("%s: cannot initialize default policy table: %d\n",
6267                         __func__, err);
6268                 goto out;
6269         }
6270
6271         err = register_pernet_subsys(&addrconf_ops);
6272         if (err < 0)
6273                 goto out_addrlabel;
6274
6275         addrconf_wq = create_workqueue("ipv6_addrconf");
6276         if (!addrconf_wq) {
6277                 err = -ENOMEM;
6278                 goto out_nowq;
6279         }
6280
6281         /* The addrconf netdev notifier requires that loopback_dev
6282          * has it's ipv6 private information allocated and setup
6283          * before it can bring up and give link-local addresses
6284          * to other devices which are up.
6285          *
6286          * Unfortunately, loopback_dev is not necessarily the first
6287          * entry in the global dev_base list of net devices.  In fact,
6288          * it is likely to be the very last entry on that list.
6289          * So this causes the notifier registry below to try and
6290          * give link-local addresses to all devices besides loopback_dev
6291          * first, then loopback_dev, which cases all the non-loopback_dev
6292          * devices to fail to get a link-local address.
6293          *
6294          * So, as a temporary fix, allocate the ipv6 structure for
6295          * loopback_dev first by hand.
6296          * Longer term, all of the dependencies ipv6 has upon the loopback
6297          * device and it being up should be removed.
6298          */
6299         rtnl_lock();
6300         idev = ipv6_add_dev(init_net.loopback_dev);
6301         rtnl_unlock();
6302         if (IS_ERR(idev)) {
6303                 err = PTR_ERR(idev);
6304                 goto errlo;
6305         }
6306
6307         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6308                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6309
6310         register_netdevice_notifier(&ipv6_dev_notf);
6311
6312         addrconf_verify();
6313
6314         rtnl_af_register(&inet6_ops);
6315
6316         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6317                               NULL);
6318         if (err < 0)
6319                 goto errout;
6320
6321         /* Only the first call to __rtnl_register can fail */
6322         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6323         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6324         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6325                         inet6_dump_ifaddr, NULL);
6326         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6327                         inet6_dump_ifmcaddr, NULL);
6328         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6329                         inet6_dump_ifacaddr, NULL);
6330         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6331                         inet6_netconf_dump_devconf, NULL);
6332
6333         ipv6_addr_label_rtnl_register();
6334
6335         return 0;
6336 errout:
6337         rtnl_af_unregister(&inet6_ops);
6338         unregister_netdevice_notifier(&ipv6_dev_notf);
6339 errlo:
6340         destroy_workqueue(addrconf_wq);
6341 out_nowq:
6342         unregister_pernet_subsys(&addrconf_ops);
6343 out_addrlabel:
6344         ipv6_addr_label_cleanup();
6345 out:
6346         return err;
6347 }
6348
6349 void addrconf_cleanup(void)
6350 {
6351         struct net_device *dev;
6352         int i;
6353
6354         unregister_netdevice_notifier(&ipv6_dev_notf);
6355         unregister_pernet_subsys(&addrconf_ops);
6356         ipv6_addr_label_cleanup();
6357
6358         rtnl_lock();
6359
6360         __rtnl_af_unregister(&inet6_ops);
6361
6362         /* clean dev list */
6363         for_each_netdev(&init_net, dev) {
6364                 if (__in6_dev_get(dev) == NULL)
6365                         continue;
6366                 addrconf_ifdown(dev, 1);
6367         }
6368         addrconf_ifdown(init_net.loopback_dev, 2);
6369
6370         /*
6371          *      Check hash table.
6372          */
6373         spin_lock_bh(&addrconf_hash_lock);
6374         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6375                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6376         spin_unlock_bh(&addrconf_hash_lock);
6377         cancel_delayed_work(&addr_chk_work);
6378         rtnl_unlock();
6379
6380         destroy_workqueue(addrconf_wq);
6381 }