net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/sched/signal.h>
  47 #include <linux/socket.h>
  48 #include <linux/sockios.h>
  49 #include <linux/net.h>
  50 #include <linux/inet.h>
  51 #include <linux/in6.h>
  52 #include <linux/netdevice.h>
  53 #include <linux/if_addr.h>
  54 #include <linux/if_arp.h>
  55 #include <linux/if_arcnet.h>
  56 #include <linux/if_infiniband.h>
  57 #include <linux/route.h>
  58 #include <linux/inetdevice.h>
  59 #include <linux/init.h>
  60 #include <linux/slab.h>
  61 #ifdef CONFIG_SYSCTL
  62 #include <linux/sysctl.h>
  63 #endif
  64 #include <linux/capability.h>
  65 #include <linux/delay.h>
  66 #include <linux/notifier.h>
  67 #include <linux/string.h>
  68 #include <linux/hash.h>
  69
  70 #include <net/net_namespace.h>
  71 #include <net/sock.h>
  72 #include <net/snmp.h>
  73
  74 #include <net/6lowpan.h>
  75 #include <net/firewire.h>
  76 #include <net/ipv6.h>
  77 #include <net/protocol.h>
  78 #include <net/ndisc.h>
  79 #include <net/ip6_route.h>
  80 #include <net/addrconf.h>
  81 #include <net/tcp.h>
  82 #include <net/ip.h>
  83 #include <net/netlink.h>
  84 #include <net/pkt_sched.h>
  85 #include <net/l3mdev.h>
  86 #include <linux/if_tunnel.h>
  87 #include <linux/rtnetlink.h>
  88 #include <linux/netconf.h>
  89 #include <linux/random.h>
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 /* Set to 3 to get tracing... */
  98 #define ACONF_DEBUG 2
  99
 100 #if ACONF_DEBUG >= 3
 101 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 102 #else
 103 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 104 #endif
 105
 106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 107
 108 #define IPV6_MAX_STRLEN \
 109         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 110
 111 static inline u32 cstamp_delta(unsigned long cstamp)
 112 {
 113         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 114 }
 115
 116 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 117 {
 118         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 119         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 120         do_div(tmp, 1000000);
 121         return (s32)tmp;
 122 }
 123
 124 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 125 {
 126         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 127         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 128         do_div(tmp, 1000000);
 129         if ((s32)tmp > mrt) {
 130                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 131                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 132                 do_div(tmp, 1000000);
 133         }
 134         return (s32)tmp;
 135 }
 136
 137 #ifdef CONFIG_SYSCTL
 138 static int addrconf_sysctl_register(struct inet6_dev *idev);
 139 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 140 #else
 141 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 142 {
 143         return 0;
 144 }
 145
 146 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 147 {
 148 }
 149 #endif
 150
 151 static void ipv6_regen_rndid(struct inet6_dev *idev);
 152 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 153
 154 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 155 static int ipv6_count_addresses(struct inet6_dev *idev);
 156 static int ipv6_generate_stable_address(struct in6_addr *addr,
 157                                         u8 dad_count,
 158                                         const struct inet6_dev *idev);
 159
 160 /*
 161  *      Configured unicast address hash table
 162  */
 163 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 164 static DEFINE_SPINLOCK(addrconf_hash_lock);
 165
 166 static void addrconf_verify(void);
 167 static void addrconf_verify_rtnl(void);
 168 static void addrconf_verify_work(struct work_struct *);
 169
 170 static struct workqueue_struct *addrconf_wq;
 171 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 172
 173 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 174 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 175
 176 static void addrconf_type_change(struct net_device *dev,
 177                                  unsigned long event);
 178 static int addrconf_ifdown(struct net_device *dev, int how);
 179
 180 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 181                                                   int plen,
 182                                                   const struct net_device *dev,
 183                                                   u32 flags, u32 noflags);
 184
 185 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 186 static void addrconf_dad_work(struct work_struct *w);
 187 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 188                                    bool send_na);
 189 static void addrconf_dad_run(struct inet6_dev *idev);
 190 static void addrconf_rs_timer(unsigned long data);
 191 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 192 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 193
 194 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 195                                 struct prefix_info *pinfo);
 196 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 197                                struct net_device *dev);
 198
 199 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 200         .forwarding             = 0,
 201         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 202         .mtu6                   = IPV6_MIN_MTU,
 203         .accept_ra              = 1,
 204         .accept_redirects       = 1,
 205         .autoconf               = 1,
 206         .force_mld_version      = 0,
 207         .mldv1_unsolicited_report_interval = 10 * HZ,
 208         .mldv2_unsolicited_report_interval = HZ,
 209         .dad_transmits          = 1,
 210         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 211         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 212         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 213         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 214         .use_tempaddr           = 0,
 215         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 216         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 217         .regen_max_retry        = REGEN_MAX_RETRY,
 218         .max_desync_factor      = MAX_DESYNC_FACTOR,
 219         .max_addresses          = IPV6_MAX_ADDRESSES,
 220         .accept_ra_defrtr       = 1,
 221         .accept_ra_from_local   = 0,
 222         .accept_ra_min_hop_limit= 1,
 223         .accept_ra_pinfo        = 1,
 224 #ifdef CONFIG_IPV6_ROUTER_PREF
 225         .accept_ra_rtr_pref     = 1,
 226         .rtr_probe_interval     = 60 * HZ,
 227 #ifdef CONFIG_IPV6_ROUTE_INFO
 228         .accept_ra_rt_info_min_plen = 0,
 229         .accept_ra_rt_info_max_plen = 0,
 230 #endif
 231 #endif
 232         .proxy_ndp              = 0,
 233         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 234         .disable_ipv6           = 0,
 235         .accept_dad             = 0,
 236         .suppress_frag_ndisc    = 1,
 237         .accept_ra_mtu          = 1,
 238         .stable_secret          = {
 239                 .initialized = false,
 240         },
 241         .use_oif_addrs_only     = 0,
 242         .ignore_routes_with_linkdown = 0,
 243         .keep_addr_on_down      = 0,
 244         .seg6_enabled           = 0,
 245 #ifdef CONFIG_IPV6_SEG6_HMAC
 246         .seg6_require_hmac      = 0,
 247 #endif
 248         .enhanced_dad           = 1,
 249         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 250         .disable_policy         = 0,
 251 };
 252
 253 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 254         .forwarding             = 0,
 255         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 256         .mtu6                   = IPV6_MIN_MTU,
 257         .accept_ra              = 1,
 258         .accept_redirects       = 1,
 259         .autoconf               = 1,
 260         .force_mld_version      = 0,
 261         .mldv1_unsolicited_report_interval = 10 * HZ,
 262         .mldv2_unsolicited_report_interval = HZ,
 263         .dad_transmits          = 1,
 264         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 265         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 266         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 267         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 268         .use_tempaddr           = 0,
 269         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 270         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 271         .regen_max_retry        = REGEN_MAX_RETRY,
 272         .max_desync_factor      = MAX_DESYNC_FACTOR,
 273         .max_addresses          = IPV6_MAX_ADDRESSES,
 274         .accept_ra_defrtr       = 1,
 275         .accept_ra_from_local   = 0,
 276         .accept_ra_min_hop_limit= 1,
 277         .accept_ra_pinfo        = 1,
 278 #ifdef CONFIG_IPV6_ROUTER_PREF
 279         .accept_ra_rtr_pref     = 1,
 280         .rtr_probe_interval     = 60 * HZ,
 281 #ifdef CONFIG_IPV6_ROUTE_INFO
 282         .accept_ra_rt_info_min_plen = 0,
 283         .accept_ra_rt_info_max_plen = 0,
 284 #endif
 285 #endif
 286         .proxy_ndp              = 0,
 287         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 288         .disable_ipv6           = 0,
 289         .accept_dad             = 1,
 290         .suppress_frag_ndisc    = 1,
 291         .accept_ra_mtu          = 1,
 292         .stable_secret          = {
 293                 .initialized = false,
 294         },
 295         .use_oif_addrs_only     = 0,
 296         .ignore_routes_with_linkdown = 0,
 297         .keep_addr_on_down      = 0,
 298         .seg6_enabled           = 0,
 299 #ifdef CONFIG_IPV6_SEG6_HMAC
 300         .seg6_require_hmac      = 0,
 301 #endif
 302         .enhanced_dad           = 1,
 303         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 304         .disable_policy         = 0,
 305 };
 306
 307 /* Check if link is ready: is it up and is a valid qdisc available */
 308 static inline bool addrconf_link_ready(const struct net_device *dev)
 309 {
 310         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 311 }
 312
 313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 314 {
 315         if (del_timer(&idev->rs_timer))
 316                 __in6_dev_put(idev);
 317 }
 318
 319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 320 {
 321         if (cancel_delayed_work(&ifp->dad_work))
 322                 __in6_ifa_put(ifp);
 323 }
 324
 325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 326                                   unsigned long when)
 327 {
 328         if (!timer_pending(&idev->rs_timer))
 329                 in6_dev_hold(idev);
 330         mod_timer(&idev->rs_timer, jiffies + when);
 331 }
 332
 333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 334                                    unsigned long delay)
 335 {
 336         in6_ifa_hold(ifp);
 337         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 338                 in6_ifa_put(ifp);
 339 }
 340
 341 static int snmp6_alloc_dev(struct inet6_dev *idev)
 342 {
 343         int i;
 344
 345         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 346         if (!idev->stats.ipv6)
 347                 goto err_ip;
 348
 349         for_each_possible_cpu(i) {
 350                 struct ipstats_mib *addrconf_stats;
 351                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 352                 u64_stats_init(&addrconf_stats->syncp);
 353         }
 354
 355
 356         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 357                                         GFP_KERNEL);
 358         if (!idev->stats.icmpv6dev)
 359                 goto err_icmp;
 360         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 361                                            GFP_KERNEL);
 362         if (!idev->stats.icmpv6msgdev)
 363                 goto err_icmpmsg;
 364
 365         return 0;
 366
 367 err_icmpmsg:
 368         kfree(idev->stats.icmpv6dev);
 369 err_icmp:
 370         free_percpu(idev->stats.ipv6);
 371 err_ip:
 372         return -ENOMEM;
 373 }
 374
 375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 376 {
 377         struct inet6_dev *ndev;
 378         int err = -ENOMEM;
 379
 380         ASSERT_RTNL();
 381
 382         if (dev->mtu < IPV6_MIN_MTU)
 383                 return ERR_PTR(-EINVAL);
 384
 385         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 386         if (!ndev)
 387                 return ERR_PTR(err);
 388
 389         rwlock_init(&ndev->lock);
 390         ndev->dev = dev;
 391         INIT_LIST_HEAD(&ndev->addr_list);
 392         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 393                     (unsigned long)ndev);
 394         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 395
 396         if (ndev->cnf.stable_secret.initialized)
 397                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 398         else
 399                 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
 400
 401         ndev->cnf.mtu6 = dev->mtu;
 402         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 403         if (!ndev->nd_parms) {
 404                 kfree(ndev);
 405                 return ERR_PTR(err);
 406         }
 407         if (ndev->cnf.forwarding)
 408                 dev_disable_lro(dev);
 409         /* We refer to the device */
 410         dev_hold(dev);
 411
 412         if (snmp6_alloc_dev(ndev) < 0) {
 413                 ADBG(KERN_WARNING
 414                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 415                         __func__, dev->name);
 416                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 417                 dev_put(dev);
 418                 kfree(ndev);
 419                 return ERR_PTR(err);
 420         }
 421
 422         if (snmp6_register_dev(ndev) < 0) {
 423                 ADBG(KERN_WARNING
 424                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 425                         __func__, dev->name);
 426                 goto err_release;
 427         }
 428
 429         /* One reference from device. */
 430         refcount_set(&ndev->refcnt, 1);
 431
 432         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 433                 ndev->cnf.accept_dad = -1;
 434
 435 #if IS_ENABLED(CONFIG_IPV6_SIT)
 436         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 437                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 438                 ndev->cnf.rtr_solicits = 0;
 439         }
 440 #endif
 441
 442         INIT_LIST_HEAD(&ndev->tempaddr_list);
 443         ndev->desync_factor = U32_MAX;
 444         if ((dev->flags&IFF_LOOPBACK) ||
 445             dev->type == ARPHRD_TUNNEL ||
 446             dev->type == ARPHRD_TUNNEL6 ||
 447             dev->type == ARPHRD_SIT ||
 448             dev->type == ARPHRD_NONE) {
 449                 ndev->cnf.use_tempaddr = -1;
 450         } else
 451                 ipv6_regen_rndid(ndev);
 452
 453         ndev->token = in6addr_any;
 454
 455         if (netif_running(dev) && addrconf_link_ready(dev))
 456                 ndev->if_flags |= IF_READY;
 457
 458         ipv6_mc_init_dev(ndev);
 459         ndev->tstamp = jiffies;
 460         err = addrconf_sysctl_register(ndev);
 461         if (err) {
 462                 ipv6_mc_destroy_dev(ndev);
 463                 snmp6_unregister_dev(ndev);
 464                 goto err_release;
 465         }
 466         /* protected by rtnl_lock */
 467         rcu_assign_pointer(dev->ip6_ptr, ndev);
 468
 469         /* Join interface-local all-node multicast group */
 470         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 471
 472         /* Join all-node multicast group */
 473         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 474
 475         /* Join all-router multicast group if forwarding is set */
 476         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 477                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 478
 479         return ndev;
 480
 481 err_release:
 482         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 483         ndev->dead = 1;
 484         in6_dev_finish_destroy(ndev);
 485         return ERR_PTR(err);
 486 }
 487
 488 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 489 {
 490         struct inet6_dev *idev;
 491
 492         ASSERT_RTNL();
 493
 494         idev = __in6_dev_get(dev);
 495         if (!idev) {
 496                 idev = ipv6_add_dev(dev);
 497                 if (IS_ERR(idev))
 498                         return NULL;
 499         }
 500
 501         if (dev->flags&IFF_UP)
 502                 ipv6_mc_up(idev);
 503         return idev;
 504 }
 505
 506 static int inet6_netconf_msgsize_devconf(int type)
 507 {
 508         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 509                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 510         bool all = false;
 511
 512         if (type == NETCONFA_ALL)
 513                 all = true;
 514
 515         if (all || type == NETCONFA_FORWARDING)
 516                 size += nla_total_size(4);
 517 #ifdef CONFIG_IPV6_MROUTE
 518         if (all || type == NETCONFA_MC_FORWARDING)
 519                 size += nla_total_size(4);
 520 #endif
 521         if (all || type == NETCONFA_PROXY_NEIGH)
 522                 size += nla_total_size(4);
 523
 524         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 525                 size += nla_total_size(4);
 526
 527         return size;
 528 }
 529
 530 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 531                                       struct ipv6_devconf *devconf, u32 portid,
 532                                       u32 seq, int event, unsigned int flags,
 533                                       int type)
 534 {
 535         struct nlmsghdr  *nlh;
 536         struct netconfmsg *ncm;
 537         bool all = false;
 538
 539         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 540                         flags);
 541         if (!nlh)
 542                 return -EMSGSIZE;
 543
 544         if (type == NETCONFA_ALL)
 545                 all = true;
 546
 547         ncm = nlmsg_data(nlh);
 548         ncm->ncm_family = AF_INET6;
 549
 550         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 551                 goto nla_put_failure;
 552
 553         if (!devconf)
 554                 goto out;
 555
 556         if ((all || type == NETCONFA_FORWARDING) &&
 557             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 558                 goto nla_put_failure;
 559 #ifdef CONFIG_IPV6_MROUTE
 560         if ((all || type == NETCONFA_MC_FORWARDING) &&
 561             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 562                         devconf->mc_forwarding) < 0)
 563                 goto nla_put_failure;
 564 #endif
 565         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 566             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 567                 goto nla_put_failure;
 568
 569         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 570             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 571                         devconf->ignore_routes_with_linkdown) < 0)
 572                 goto nla_put_failure;
 573
 574 out:
 575         nlmsg_end(skb, nlh);
 576         return 0;
 577
 578 nla_put_failure:
 579         nlmsg_cancel(skb, nlh);
 580         return -EMSGSIZE;
 581 }
 582
 583 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 584                                   int ifindex, struct ipv6_devconf *devconf)
 585 {
 586         struct sk_buff *skb;
 587         int err = -ENOBUFS;
 588
 589         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 590         if (!skb)
 591                 goto errout;
 592
 593         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 594                                          event, 0, type);
 595         if (err < 0) {
 596                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 597                 WARN_ON(err == -EMSGSIZE);
 598                 kfree_skb(skb);
 599                 goto errout;
 600         }
 601         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 602         return;
 603 errout:
 604         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 605 }
 606
 607 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 608         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 609         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 610         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 611         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 612 };
 613
 614 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 615                                      struct nlmsghdr *nlh,
 616                                      struct netlink_ext_ack *extack)
 617 {
 618         struct net *net = sock_net(in_skb->sk);
 619         struct nlattr *tb[NETCONFA_MAX+1];
 620         struct netconfmsg *ncm;
 621         struct sk_buff *skb;
 622         struct ipv6_devconf *devconf;
 623         struct inet6_dev *in6_dev;
 624         struct net_device *dev;
 625         int ifindex;
 626         int err;
 627
 628         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 629                           devconf_ipv6_policy, extack);
 630         if (err < 0)
 631                 goto errout;
 632
 633         err = -EINVAL;
 634         if (!tb[NETCONFA_IFINDEX])
 635                 goto errout;
 636
 637         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 638         switch (ifindex) {
 639         case NETCONFA_IFINDEX_ALL:
 640                 devconf = net->ipv6.devconf_all;
 641                 break;
 642         case NETCONFA_IFINDEX_DEFAULT:
 643                 devconf = net->ipv6.devconf_dflt;
 644                 break;
 645         default:
 646                 dev = __dev_get_by_index(net, ifindex);
 647                 if (!dev)
 648                         goto errout;
 649                 in6_dev = __in6_dev_get(dev);
 650                 if (!in6_dev)
 651                         goto errout;
 652                 devconf = &in6_dev->cnf;
 653                 break;
 654         }
 655
 656         err = -ENOBUFS;
 657         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
 658         if (!skb)
 659                 goto errout;
 660
 661         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 662                                          NETLINK_CB(in_skb).portid,
 663                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 664                                          NETCONFA_ALL);
 665         if (err < 0) {
 666                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 667                 WARN_ON(err == -EMSGSIZE);
 668                 kfree_skb(skb);
 669                 goto errout;
 670         }
 671         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 672 errout:
 673         return err;
 674 }
 675
 676 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 677                                       struct netlink_callback *cb)
 678 {
 679         struct net *net = sock_net(skb->sk);
 680         int h, s_h;
 681         int idx, s_idx;
 682         struct net_device *dev;
 683         struct inet6_dev *idev;
 684         struct hlist_head *head;
 685
 686         s_h = cb->args[0];
 687         s_idx = idx = cb->args[1];
 688
 689         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 690                 idx = 0;
 691                 head = &net->dev_index_head[h];
 692                 rcu_read_lock();
 693                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 694                           net->dev_base_seq;
 695                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 696                         if (idx < s_idx)
 697                                 goto cont;
 698                         idev = __in6_dev_get(dev);
 699                         if (!idev)
 700                                 goto cont;
 701
 702                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 703                                                        &idev->cnf,
 704                                                        NETLINK_CB(cb->skb).portid,
 705                                                        cb->nlh->nlmsg_seq,
 706                                                        RTM_NEWNETCONF,
 707                                                        NLM_F_MULTI,
 708                                                        NETCONFA_ALL) < 0) {
 709                                 rcu_read_unlock();
 710                                 goto done;
 711                         }
 712                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 713 cont:
 714                         idx++;
 715                 }
 716                 rcu_read_unlock();
 717         }
 718         if (h == NETDEV_HASHENTRIES) {
 719                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 720                                                net->ipv6.devconf_all,
 721                                                NETLINK_CB(cb->skb).portid,
 722                                                cb->nlh->nlmsg_seq,
 723                                                RTM_NEWNETCONF, NLM_F_MULTI,
 724                                                NETCONFA_ALL) < 0)
 725                         goto done;
 726                 else
 727                         h++;
 728         }
 729         if (h == NETDEV_HASHENTRIES + 1) {
 730                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 731                                                net->ipv6.devconf_dflt,
 732                                                NETLINK_CB(cb->skb).portid,
 733                                                cb->nlh->nlmsg_seq,
 734                                                RTM_NEWNETCONF, NLM_F_MULTI,
 735                                                NETCONFA_ALL) < 0)
 736                         goto done;
 737                 else
 738                         h++;
 739         }
 740 done:
 741         cb->args[0] = h;
 742         cb->args[1] = idx;
 743
 744         return skb->len;
 745 }
 746
 747 #ifdef CONFIG_SYSCTL
 748 static void dev_forward_change(struct inet6_dev *idev)
 749 {
 750         struct net_device *dev;
 751         struct inet6_ifaddr *ifa;
 752
 753         if (!idev)
 754                 return;
 755         dev = idev->dev;
 756         if (idev->cnf.forwarding)
 757                 dev_disable_lro(dev);
 758         if (dev->flags & IFF_MULTICAST) {
 759                 if (idev->cnf.forwarding) {
 760                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 761                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 762                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 763                 } else {
 764                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 765                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 766                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 767                 }
 768         }
 769
 770         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 771                 if (ifa->flags&IFA_F_TENTATIVE)
 772                         continue;
 773                 if (idev->cnf.forwarding)
 774                         addrconf_join_anycast(ifa);
 775                 else
 776                         addrconf_leave_anycast(ifa);
 777         }
 778         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 779                                      NETCONFA_FORWARDING,
 780                                      dev->ifindex, &idev->cnf);
 781 }
 782
 783
 784 static void addrconf_forward_change(struct net *net, __s32 newf)
 785 {
 786         struct net_device *dev;
 787         struct inet6_dev *idev;
 788
 789         for_each_netdev(net, dev) {
 790                 idev = __in6_dev_get(dev);
 791                 if (idev) {
 792                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 793                         idev->cnf.forwarding = newf;
 794                         if (changed)
 795                                 dev_forward_change(idev);
 796                 }
 797         }
 798 }
 799
 800 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 801 {
 802         struct net *net;
 803         int old;
 804
 805         if (!rtnl_trylock())
 806                 return restart_syscall();
 807
 808         net = (struct net *)table->extra2;
 809         old = *p;
 810         *p = newf;
 811
 812         if (p == &net->ipv6.devconf_dflt->forwarding) {
 813                 if ((!newf) ^ (!old))
 814                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 815                                                      NETCONFA_FORWARDING,
 816                                                      NETCONFA_IFINDEX_DEFAULT,
 817                                                      net->ipv6.devconf_dflt);
 818                 rtnl_unlock();
 819                 return 0;
 820         }
 821
 822         if (p == &net->ipv6.devconf_all->forwarding) {
 823                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 824
 825                 net->ipv6.devconf_dflt->forwarding = newf;
 826                 if ((!newf) ^ (!old_dflt))
 827                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 828                                                      NETCONFA_FORWARDING,
 829                                                      NETCONFA_IFINDEX_DEFAULT,
 830                                                      net->ipv6.devconf_dflt);
 831
 832                 addrconf_forward_change(net, newf);
 833                 if ((!newf) ^ (!old))
 834                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 835                                                      NETCONFA_FORWARDING,
 836                                                      NETCONFA_IFINDEX_ALL,
 837                                                      net->ipv6.devconf_all);
 838         } else if ((!newf) ^ (!old))
 839                 dev_forward_change((struct inet6_dev *)table->extra1);
 840         rtnl_unlock();
 841
 842         if (newf)
 843                 rt6_purge_dflt_routers(net);
 844         return 1;
 845 }
 846
 847 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 848 {
 849         struct net_device *dev;
 850         struct inet6_dev *idev;
 851
 852         for_each_netdev(net, dev) {
 853                 idev = __in6_dev_get(dev);
 854                 if (idev) {
 855                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 856
 857                         idev->cnf.ignore_routes_with_linkdown = newf;
 858                         if (changed)
 859                                 inet6_netconf_notify_devconf(dev_net(dev),
 860                                                              RTM_NEWNETCONF,
 861                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 862                                                              dev->ifindex,
 863                                                              &idev->cnf);
 864                 }
 865         }
 866 }
 867
 868 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 869 {
 870         struct net *net;
 871         int old;
 872
 873         if (!rtnl_trylock())
 874                 return restart_syscall();
 875
 876         net = (struct net *)table->extra2;
 877         old = *p;
 878         *p = newf;
 879
 880         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 881                 if ((!newf) ^ (!old))
 882                         inet6_netconf_notify_devconf(net,
 883                                                      RTM_NEWNETCONF,
 884                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 885                                                      NETCONFA_IFINDEX_DEFAULT,
 886                                                      net->ipv6.devconf_dflt);
 887                 rtnl_unlock();
 888                 return 0;
 889         }
 890
 891         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 892                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 893                 addrconf_linkdown_change(net, newf);
 894                 if ((!newf) ^ (!old))
 895                         inet6_netconf_notify_devconf(net,
 896                                                      RTM_NEWNETCONF,
 897                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 898                                                      NETCONFA_IFINDEX_ALL,
 899                                                      net->ipv6.devconf_all);
 900         }
 901         rtnl_unlock();
 902
 903         return 1;
 904 }
 905
 906 #endif
 907
 908 /* Nobody refers to this ifaddr, destroy it */
 909 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 910 {
 911         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 912
 913 #ifdef NET_REFCNT_DEBUG
 914         pr_debug("%s\n", __func__);
 915 #endif
 916
 917         in6_dev_put(ifp->idev);
 918
 919         if (cancel_delayed_work(&ifp->dad_work))
 920                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 921                           ifp);
 922
 923         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 924                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 925                 return;
 926         }
 927         ip6_rt_put(ifp->rt);
 928
 929         kfree_rcu(ifp, rcu);
 930 }
 931
 932 static void
 933 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 934 {
 935         struct list_head *p;
 936         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 937
 938         /*
 939          * Each device address list is sorted in order of scope -
 940          * global before linklocal.
 941          */
 942         list_for_each(p, &idev->addr_list) {
 943                 struct inet6_ifaddr *ifa
 944                         = list_entry(p, struct inet6_ifaddr, if_list);
 945                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 946                         break;
 947         }
 948
 949         list_add_tail(&ifp->if_list, p);
 950 }
 951
 952 static u32 inet6_addr_hash(const struct in6_addr *addr)
 953 {
 954         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 955 }
 956
 957 /* On success it returns ifp with increased reference count */
 958
 959 static struct inet6_ifaddr *
 960 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 961               const struct in6_addr *peer_addr, int pfxlen,
 962               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 963 {
 964         struct net *net = dev_net(idev->dev);
 965         struct inet6_ifaddr *ifa = NULL;
 966         struct rt6_info *rt;
 967         struct in6_validator_info i6vi;
 968         unsigned int hash;
 969         int err = 0;
 970         int addr_type = ipv6_addr_type(addr);
 971
 972         if (addr_type == IPV6_ADDR_ANY ||
 973             addr_type & IPV6_ADDR_MULTICAST ||
 974             (!(idev->dev->flags & IFF_LOOPBACK) &&
 975              addr_type & IPV6_ADDR_LOOPBACK))
 976                 return ERR_PTR(-EADDRNOTAVAIL);
 977
 978         rcu_read_lock_bh();
 979
 980         in6_dev_hold(idev);
 981
 982         if (idev->dead) {
 983                 err = -ENODEV;                  /*XXX*/
 984                 goto out2;
 985         }
 986
 987         if (idev->cnf.disable_ipv6) {
 988                 err = -EACCES;
 989                 goto out2;
 990         }
 991
 992         i6vi.i6vi_addr = *addr;
 993         i6vi.i6vi_dev = idev;
 994         rcu_read_unlock_bh();
 995
 996         err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
 997
 998         rcu_read_lock_bh();
 999         err = notifier_to_errno(err);
1000         if (err)
1001                 goto out2;
1002
1003         spin_lock(&addrconf_hash_lock);
1004
1005         /* Ignore adding duplicate addresses on an interface */
1006         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
1007                 ADBG("ipv6_add_addr: already assigned\n");
1008                 err = -EEXIST;
1009                 goto out;
1010         }
1011
1012         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
1013
1014         if (!ifa) {
1015                 ADBG("ipv6_add_addr: malloc failed\n");
1016                 err = -ENOBUFS;
1017                 goto out;
1018         }
1019
1020         rt = addrconf_dst_alloc(idev, addr, false);
1021         if (IS_ERR(rt)) {
1022                 err = PTR_ERR(rt);
1023                 goto out;
1024         }
1025
1026         if (net->ipv6.devconf_all->disable_policy ||
1027             idev->cnf.disable_policy)
1028                 rt->dst.flags |= DST_NOPOLICY;
1029
1030         neigh_parms_data_state_setall(idev->nd_parms);
1031
1032         ifa->addr = *addr;
1033         if (peer_addr)
1034                 ifa->peer_addr = *peer_addr;
1035
1036         spin_lock_init(&ifa->lock);
1037         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1038         INIT_HLIST_NODE(&ifa->addr_lst);
1039         ifa->scope = scope;
1040         ifa->prefix_len = pfxlen;
1041         ifa->flags = flags;
1042         /* No need to add the TENTATIVE flag for addresses with NODAD */
1043         if (!(flags & IFA_F_NODAD))
1044                 ifa->flags |= IFA_F_TENTATIVE;
1045         ifa->valid_lft = valid_lft;
1046         ifa->prefered_lft = prefered_lft;
1047         ifa->cstamp = ifa->tstamp = jiffies;
1048         ifa->tokenized = false;
1049
1050         ifa->rt = rt;
1051
1052         ifa->idev = idev;
1053         /* For caller */
1054         refcount_set(&ifa->refcnt, 1);
1055
1056         /* Add to big hash table */
1057         hash = inet6_addr_hash(addr);
1058
1059         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1060         spin_unlock(&addrconf_hash_lock);
1061
1062         write_lock(&idev->lock);
1063         /* Add to inet6_dev unicast addr list. */
1064         ipv6_link_dev_addr(idev, ifa);
1065
1066         if (ifa->flags&IFA_F_TEMPORARY) {
1067                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1068                 in6_ifa_hold(ifa);
1069         }
1070
1071         in6_ifa_hold(ifa);
1072         write_unlock(&idev->lock);
1073 out2:
1074         rcu_read_unlock_bh();
1075
1076         if (likely(err == 0))
1077                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1078         else {
1079                 kfree(ifa);
1080                 in6_dev_put(idev);
1081                 ifa = ERR_PTR(err);
1082         }
1083
1084         return ifa;
1085 out:
1086         spin_unlock(&addrconf_hash_lock);
1087         goto out2;
1088 }
1089
1090 enum cleanup_prefix_rt_t {
1091         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1092         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1093         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1094 };
1095
1096 /*
1097  * Check, whether the prefix for ifp would still need a prefix route
1098  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1099  * constants.
1100  *
1101  * 1) we don't purge prefix if address was not permanent.
1102  *    prefix is managed by its own lifetime.
1103  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1104  * 3) if there are no addresses, delete prefix.
1105  * 4) if there are still other permanent address(es),
1106  *    corresponding prefix is still permanent.
1107  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1108  *    don't purge the prefix, assume user space is managing it.
1109  * 6) otherwise, update prefix lifetime to the
1110  *    longest valid lifetime among the corresponding
1111  *    addresses on the device.
1112  *    Note: subsequent RA will update lifetime.
1113  **/
1114 static enum cleanup_prefix_rt_t
1115 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1116 {
1117         struct inet6_ifaddr *ifa;
1118         struct inet6_dev *idev = ifp->idev;
1119         unsigned long lifetime;
1120         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1121
1122         *expires = jiffies;
1123
1124         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1125                 if (ifa == ifp)
1126                         continue;
1127                 if (ifa->prefix_len != ifp->prefix_len ||
1128                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1129                                        ifp->prefix_len))
1130                         continue;
1131                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1132                         return CLEANUP_PREFIX_RT_NOP;
1133
1134                 action = CLEANUP_PREFIX_RT_EXPIRE;
1135
1136                 spin_lock(&ifa->lock);
1137
1138                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1139                 /*
1140                  * Note: Because this address is
1141                  * not permanent, lifetime <
1142                  * LONG_MAX / HZ here.
1143                  */
1144                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1145                         *expires = ifa->tstamp + lifetime * HZ;
1146                 spin_unlock(&ifa->lock);
1147         }
1148
1149         return action;
1150 }
1151
1152 static void
1153 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1154 {
1155         struct rt6_info *rt;
1156
1157         rt = addrconf_get_prefix_route(&ifp->addr,
1158                                        ifp->prefix_len,
1159                                        ifp->idev->dev,
1160                                        0, RTF_GATEWAY | RTF_DEFAULT);
1161         if (rt) {
1162                 if (del_rt)
1163                         ip6_del_rt(rt);
1164                 else {
1165                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1166                                 rt6_set_expires(rt, expires);
1167                         ip6_rt_put(rt);
1168                 }
1169         }
1170 }
1171
1172
1173 /* This function wants to get referenced ifp and releases it before return */
1174
1175 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1176 {
1177         int state;
1178         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1179         unsigned long expires;
1180
1181         ASSERT_RTNL();
1182
1183         spin_lock_bh(&ifp->lock);
1184         state = ifp->state;
1185         ifp->state = INET6_IFADDR_STATE_DEAD;
1186         spin_unlock_bh(&ifp->lock);
1187
1188         if (state == INET6_IFADDR_STATE_DEAD)
1189                 goto out;
1190
1191         spin_lock_bh(&addrconf_hash_lock);
1192         hlist_del_init_rcu(&ifp->addr_lst);
1193         spin_unlock_bh(&addrconf_hash_lock);
1194
1195         write_lock_bh(&ifp->idev->lock);
1196
1197         if (ifp->flags&IFA_F_TEMPORARY) {
1198                 list_del(&ifp->tmp_list);
1199                 if (ifp->ifpub) {
1200                         in6_ifa_put(ifp->ifpub);
1201                         ifp->ifpub = NULL;
1202                 }
1203                 __in6_ifa_put(ifp);
1204         }
1205
1206         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1207                 action = check_cleanup_prefix_route(ifp, &expires);
1208
1209         list_del_init(&ifp->if_list);
1210         __in6_ifa_put(ifp);
1211
1212         write_unlock_bh(&ifp->idev->lock);
1213
1214         addrconf_del_dad_work(ifp);
1215
1216         ipv6_ifa_notify(RTM_DELADDR, ifp);
1217
1218         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1219
1220         if (action != CLEANUP_PREFIX_RT_NOP) {
1221                 cleanup_prefix_route(ifp, expires,
1222                         action == CLEANUP_PREFIX_RT_DEL);
1223         }
1224
1225         /* clean up prefsrc entries */
1226         rt6_remove_prefsrc(ifp);
1227 out:
1228         in6_ifa_put(ifp);
1229 }
1230
1231 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1232 {
1233         struct inet6_dev *idev = ifp->idev;
1234         struct in6_addr addr, *tmpaddr;
1235         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1236         unsigned long regen_advance;
1237         int tmp_plen;
1238         int ret = 0;
1239         u32 addr_flags;
1240         unsigned long now = jiffies;
1241         long max_desync_factor;
1242         s32 cnf_temp_preferred_lft;
1243
1244         write_lock_bh(&idev->lock);
1245         if (ift) {
1246                 spin_lock_bh(&ift->lock);
1247                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1248                 spin_unlock_bh(&ift->lock);
1249                 tmpaddr = &addr;
1250         } else {
1251                 tmpaddr = NULL;
1252         }
1253 retry:
1254         in6_dev_hold(idev);
1255         if (idev->cnf.use_tempaddr <= 0) {
1256                 write_unlock_bh(&idev->lock);
1257                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1258                 in6_dev_put(idev);
1259                 ret = -1;
1260                 goto out;
1261         }
1262         spin_lock_bh(&ifp->lock);
1263         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1264                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1265                 spin_unlock_bh(&ifp->lock);
1266                 write_unlock_bh(&idev->lock);
1267                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1268                         __func__);
1269                 in6_dev_put(idev);
1270                 ret = -1;
1271                 goto out;
1272         }
1273         in6_ifa_hold(ifp);
1274         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1275         ipv6_try_regen_rndid(idev, tmpaddr);
1276         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1277         age = (now - ifp->tstamp) / HZ;
1278
1279         regen_advance = idev->cnf.regen_max_retry *
1280                         idev->cnf.dad_transmits *
1281                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1282
1283         /* recalculate max_desync_factor each time and update
1284          * idev->desync_factor if it's larger
1285          */
1286         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1287         max_desync_factor = min_t(__u32,
1288                                   idev->cnf.max_desync_factor,
1289                                   cnf_temp_preferred_lft - regen_advance);
1290
1291         if (unlikely(idev->desync_factor > max_desync_factor)) {
1292                 if (max_desync_factor > 0) {
1293                         get_random_bytes(&idev->desync_factor,
1294                                          sizeof(idev->desync_factor));
1295                         idev->desync_factor %= max_desync_factor;
1296                 } else {
1297                         idev->desync_factor = 0;
1298                 }
1299         }
1300
1301         tmp_valid_lft = min_t(__u32,
1302                               ifp->valid_lft,
1303                               idev->cnf.temp_valid_lft + age);
1304         tmp_prefered_lft = cnf_temp_preferred_lft + age -
1305                             idev->desync_factor;
1306         tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1307         tmp_plen = ifp->prefix_len;
1308         tmp_tstamp = ifp->tstamp;
1309         spin_unlock_bh(&ifp->lock);
1310
1311         write_unlock_bh(&idev->lock);
1312
1313         /* A temporary address is created only if this calculated Preferred
1314          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1315          * an implementation must not create a temporary address with a zero
1316          * Preferred Lifetime.
1317          * Use age calculation as in addrconf_verify to avoid unnecessary
1318          * temporary addresses being generated.
1319          */
1320         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1321         if (tmp_prefered_lft <= regen_advance + age) {
1322                 in6_ifa_put(ifp);
1323                 in6_dev_put(idev);
1324                 ret = -1;
1325                 goto out;
1326         }
1327
1328         addr_flags = IFA_F_TEMPORARY;
1329         /* set in addrconf_prefix_rcv() */
1330         if (ifp->flags & IFA_F_OPTIMISTIC)
1331                 addr_flags |= IFA_F_OPTIMISTIC;
1332
1333         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1334                             ipv6_addr_scope(&addr), addr_flags,
1335                             tmp_valid_lft, tmp_prefered_lft);
1336         if (IS_ERR(ift)) {
1337                 in6_ifa_put(ifp);
1338                 in6_dev_put(idev);
1339                 pr_info("%s: retry temporary address regeneration\n", __func__);
1340                 tmpaddr = &addr;
1341                 write_lock_bh(&idev->lock);
1342                 goto retry;
1343         }
1344
1345         spin_lock_bh(&ift->lock);
1346         ift->ifpub = ifp;
1347         ift->cstamp = now;
1348         ift->tstamp = tmp_tstamp;
1349         spin_unlock_bh(&ift->lock);
1350
1351         addrconf_dad_start(ift);
1352         in6_ifa_put(ift);
1353         in6_dev_put(idev);
1354 out:
1355         return ret;
1356 }
1357
1358 /*
1359  *      Choose an appropriate source address (RFC3484)
1360  */
1361 enum {
1362         IPV6_SADDR_RULE_INIT = 0,
1363         IPV6_SADDR_RULE_LOCAL,
1364         IPV6_SADDR_RULE_SCOPE,
1365         IPV6_SADDR_RULE_PREFERRED,
1366 #ifdef CONFIG_IPV6_MIP6
1367         IPV6_SADDR_RULE_HOA,
1368 #endif
1369         IPV6_SADDR_RULE_OIF,
1370         IPV6_SADDR_RULE_LABEL,
1371         IPV6_SADDR_RULE_PRIVACY,
1372         IPV6_SADDR_RULE_ORCHID,
1373         IPV6_SADDR_RULE_PREFIX,
1374 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1375         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1376 #endif
1377         IPV6_SADDR_RULE_MAX
1378 };
1379
1380 struct ipv6_saddr_score {
1381         int                     rule;
1382         int                     addr_type;
1383         struct inet6_ifaddr     *ifa;
1384         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1385         int                     scopedist;
1386         int                     matchlen;
1387 };
1388
1389 struct ipv6_saddr_dst {
1390         const struct in6_addr *addr;
1391         int ifindex;
1392         int scope;
1393         int label;
1394         unsigned int prefs;
1395 };
1396
1397 static inline int ipv6_saddr_preferred(int type)
1398 {
1399         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1400                 return 1;
1401         return 0;
1402 }
1403
1404 static bool ipv6_use_optimistic_addr(struct net *net,
1405                                      struct inet6_dev *idev)
1406 {
1407 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1408         if (!idev)
1409                 return false;
1410         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1411                 return false;
1412         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1413                 return false;
1414
1415         return true;
1416 #else
1417         return false;
1418 #endif
1419 }
1420
1421 static int ipv6_get_saddr_eval(struct net *net,
1422                                struct ipv6_saddr_score *score,
1423                                struct ipv6_saddr_dst *dst,
1424                                int i)
1425 {
1426         int ret;
1427
1428         if (i <= score->rule) {
1429                 switch (i) {
1430                 case IPV6_SADDR_RULE_SCOPE:
1431                         ret = score->scopedist;
1432                         break;
1433                 case IPV6_SADDR_RULE_PREFIX:
1434                         ret = score->matchlen;
1435                         break;
1436                 default:
1437                         ret = !!test_bit(i, score->scorebits);
1438                 }
1439                 goto out;
1440         }
1441
1442         switch (i) {
1443         case IPV6_SADDR_RULE_INIT:
1444                 /* Rule 0: remember if hiscore is not ready yet */
1445                 ret = !!score->ifa;
1446                 break;
1447         case IPV6_SADDR_RULE_LOCAL:
1448                 /* Rule 1: Prefer same address */
1449                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1450                 break;
1451         case IPV6_SADDR_RULE_SCOPE:
1452                 /* Rule 2: Prefer appropriate scope
1453                  *
1454                  *      ret
1455                  *       ^
1456                  *    -1 |  d 15
1457                  *    ---+--+-+---> scope
1458                  *       |
1459                  *       |             d is scope of the destination.
1460                  *  B-d  |  \
1461                  *       |   \      <- smaller scope is better if
1462                  *  B-15 |    \        if scope is enough for destination.
1463                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1464                  * d-C-1 | /
1465                  *       |/         <- greater is better
1466                  *   -C  /             if scope is not enough for destination.
1467                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1468                  *
1469                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1470                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1471                  * Assume B = 0 and we get C > 29.
1472                  */
1473                 ret = __ipv6_addr_src_scope(score->addr_type);
1474                 if (ret >= dst->scope)
1475                         ret = -ret;
1476                 else
1477                         ret -= 128;     /* 30 is enough */
1478                 score->scopedist = ret;
1479                 break;
1480         case IPV6_SADDR_RULE_PREFERRED:
1481             {
1482                 /* Rule 3: Avoid deprecated and optimistic addresses */
1483                 u8 avoid = IFA_F_DEPRECATED;
1484
1485                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1486                         avoid |= IFA_F_OPTIMISTIC;
1487                 ret = ipv6_saddr_preferred(score->addr_type) ||
1488                       !(score->ifa->flags & avoid);
1489                 break;
1490             }
1491 #ifdef CONFIG_IPV6_MIP6
1492         case IPV6_SADDR_RULE_HOA:
1493             {
1494                 /* Rule 4: Prefer home address */
1495                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1496                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1497                 break;
1498             }
1499 #endif
1500         case IPV6_SADDR_RULE_OIF:
1501                 /* Rule 5: Prefer outgoing interface */
1502                 ret = (!dst->ifindex ||
1503                        dst->ifindex == score->ifa->idev->dev->ifindex);
1504                 break;
1505         case IPV6_SADDR_RULE_LABEL:
1506                 /* Rule 6: Prefer matching label */
1507                 ret = ipv6_addr_label(net,
1508                                       &score->ifa->addr, score->addr_type,
1509                                       score->ifa->idev->dev->ifindex) == dst->label;
1510                 break;
1511         case IPV6_SADDR_RULE_PRIVACY:
1512             {
1513                 /* Rule 7: Prefer public address
1514                  * Note: prefer temporary address if use_tempaddr >= 2
1515                  */
1516                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1517                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1518                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1519                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1520                 break;
1521             }
1522         case IPV6_SADDR_RULE_ORCHID:
1523                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1524                  *          non-ORCHID vs non-ORCHID
1525                  */
1526                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1527                         ipv6_addr_orchid(dst->addr));
1528                 break;
1529         case IPV6_SADDR_RULE_PREFIX:
1530                 /* Rule 8: Use longest matching prefix */
1531                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1532                 if (ret > score->ifa->prefix_len)
1533                         ret = score->ifa->prefix_len;
1534                 score->matchlen = ret;
1535                 break;
1536 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1537         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1538                 /* Optimistic addresses still have lower precedence than other
1539                  * preferred addresses.
1540                  */
1541                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1542                 break;
1543 #endif
1544         default:
1545                 ret = 0;
1546         }
1547
1548         if (ret)
1549                 __set_bit(i, score->scorebits);
1550         score->rule = i;
1551 out:
1552         return ret;
1553 }
1554
1555 static int __ipv6_dev_get_saddr(struct net *net,
1556                                 struct ipv6_saddr_dst *dst,
1557                                 struct inet6_dev *idev,
1558                                 struct ipv6_saddr_score *scores,
1559                                 int hiscore_idx)
1560 {
1561         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1562
1563         read_lock_bh(&idev->lock);
1564         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1565                 int i;
1566
1567                 /*
1568                  * - Tentative Address (RFC2462 section 5.4)
1569                  *  - A tentative address is not considered
1570                  *    "assigned to an interface" in the traditional
1571                  *    sense, unless it is also flagged as optimistic.
1572                  * - Candidate Source Address (section 4)
1573                  *  - In any case, anycast addresses, multicast
1574                  *    addresses, and the unspecified address MUST
1575                  *    NOT be included in a candidate set.
1576                  */
1577                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1578                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1579                         continue;
1580
1581                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1582
1583                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1584                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1585                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1586                                             idev->dev->name);
1587                         continue;
1588                 }
1589
1590                 score->rule = -1;
1591                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1592
1593                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1594                         int minihiscore, miniscore;
1595
1596                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1597                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1598
1599                         if (minihiscore > miniscore) {
1600                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1601                                     score->scopedist > 0) {
1602                                         /*
1603                                          * special case:
1604                                          * each remaining entry
1605                                          * has too small (not enough)
1606                                          * scope, because ifa entries
1607                                          * are sorted by their scope
1608                                          * values.
1609                                          */
1610                                         goto out;
1611                                 }
1612                                 break;
1613                         } else if (minihiscore < miniscore) {
1614                                 if (hiscore->ifa)
1615                                         in6_ifa_put(hiscore->ifa);
1616
1617                                 in6_ifa_hold(score->ifa);
1618
1619                                 swap(hiscore, score);
1620                                 hiscore_idx = 1 - hiscore_idx;
1621
1622                                 /* restore our iterator */
1623                                 score->ifa = hiscore->ifa;
1624
1625                                 break;
1626                         }
1627                 }
1628         }
1629 out:
1630         read_unlock_bh(&idev->lock);
1631         return hiscore_idx;
1632 }
1633
1634 static int ipv6_get_saddr_master(struct net *net,
1635                                  const struct net_device *dst_dev,
1636                                  const struct net_device *master,
1637                                  struct ipv6_saddr_dst *dst,
1638                                  struct ipv6_saddr_score *scores,
1639                                  int hiscore_idx)
1640 {
1641         struct inet6_dev *idev;
1642
1643         idev = __in6_dev_get(dst_dev);
1644         if (idev)
1645                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1646                                                    scores, hiscore_idx);
1647
1648         idev = __in6_dev_get(master);
1649         if (idev)
1650                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1651                                                    scores, hiscore_idx);
1652
1653         return hiscore_idx;
1654 }
1655
1656 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1657                        const struct in6_addr *daddr, unsigned int prefs,
1658                        struct in6_addr *saddr)
1659 {
1660         struct ipv6_saddr_score scores[2], *hiscore;
1661         struct ipv6_saddr_dst dst;
1662         struct inet6_dev *idev;
1663         struct net_device *dev;
1664         int dst_type;
1665         bool use_oif_addr = false;
1666         int hiscore_idx = 0;
1667
1668         dst_type = __ipv6_addr_type(daddr);
1669         dst.addr = daddr;
1670         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1671         dst.scope = __ipv6_addr_src_scope(dst_type);
1672         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1673         dst.prefs = prefs;
1674
1675         scores[hiscore_idx].rule = -1;
1676         scores[hiscore_idx].ifa = NULL;
1677
1678         rcu_read_lock();
1679
1680         /* Candidate Source Address (section 4)
1681          *  - multicast and link-local destination address,
1682          *    the set of candidate source address MUST only
1683          *    include addresses assigned to interfaces
1684          *    belonging to the same link as the outgoing
1685          *    interface.
1686          * (- For site-local destination addresses, the
1687          *    set of candidate source addresses MUST only
1688          *    include addresses assigned to interfaces
1689          *    belonging to the same site as the outgoing
1690          *    interface.)
1691          *  - "It is RECOMMENDED that the candidate source addresses
1692          *    be the set of unicast addresses assigned to the
1693          *    interface that will be used to send to the destination
1694          *    (the 'outgoing' interface)." (RFC 6724)
1695          */
1696         if (dst_dev) {
1697                 idev = __in6_dev_get(dst_dev);
1698                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1699                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1700                     (idev && idev->cnf.use_oif_addrs_only)) {
1701                         use_oif_addr = true;
1702                 }
1703         }
1704
1705         if (use_oif_addr) {
1706                 if (idev)
1707                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1708         } else {
1709                 const struct net_device *master;
1710                 int master_idx = 0;
1711
1712                 /* if dst_dev exists and is enslaved to an L3 device, then
1713                  * prefer addresses from dst_dev and then the master over
1714                  * any other enslaved devices in the L3 domain.
1715                  */
1716                 master = l3mdev_master_dev_rcu(dst_dev);
1717                 if (master) {
1718                         master_idx = master->ifindex;
1719
1720                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1721                                                             master, &dst,
1722                                                             scores, hiscore_idx);
1723
1724                         if (scores[hiscore_idx].ifa)
1725                                 goto out;
1726                 }
1727
1728                 for_each_netdev_rcu(net, dev) {
1729                         /* only consider addresses on devices in the
1730                          * same L3 domain
1731                          */
1732                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1733                                 continue;
1734                         idev = __in6_dev_get(dev);
1735                         if (!idev)
1736                                 continue;
1737                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1738                 }
1739         }
1740
1741 out:
1742         rcu_read_unlock();
1743
1744         hiscore = &scores[hiscore_idx];
1745         if (!hiscore->ifa)
1746                 return -EADDRNOTAVAIL;
1747
1748         *saddr = hiscore->ifa->addr;
1749         in6_ifa_put(hiscore->ifa);
1750         return 0;
1751 }
1752 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1753
1754 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1755                       u32 banned_flags)
1756 {
1757         struct inet6_ifaddr *ifp;
1758         int err = -EADDRNOTAVAIL;
1759
1760         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1761                 if (ifp->scope > IFA_LINK)
1762                         break;
1763                 if (ifp->scope == IFA_LINK &&
1764                     !(ifp->flags & banned_flags)) {
1765                         *addr = ifp->addr;
1766                         err = 0;
1767                         break;
1768                 }
1769         }
1770         return err;
1771 }
1772
1773 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1774                     u32 banned_flags)
1775 {
1776         struct inet6_dev *idev;
1777         int err = -EADDRNOTAVAIL;
1778
1779         rcu_read_lock();
1780         idev = __in6_dev_get(dev);
1781         if (idev) {
1782                 read_lock_bh(&idev->lock);
1783                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1784                 read_unlock_bh(&idev->lock);
1785         }
1786         rcu_read_unlock();
1787         return err;
1788 }
1789
1790 static int ipv6_count_addresses(struct inet6_dev *idev)
1791 {
1792         int cnt = 0;
1793         struct inet6_ifaddr *ifp;
1794
1795         read_lock_bh(&idev->lock);
1796         list_for_each_entry(ifp, &idev->addr_list, if_list)
1797                 cnt++;
1798         read_unlock_bh(&idev->lock);
1799         return cnt;
1800 }
1801
1802 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1803                   const struct net_device *dev, int strict)
1804 {
1805         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1806 }
1807 EXPORT_SYMBOL(ipv6_chk_addr);
1808
1809 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1810                             const struct net_device *dev, int strict,
1811                             u32 banned_flags)
1812 {
1813         struct inet6_ifaddr *ifp;
1814         unsigned int hash = inet6_addr_hash(addr);
1815         u32 ifp_flags;
1816
1817         rcu_read_lock_bh();
1818         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1819                 if (!net_eq(dev_net(ifp->idev->dev), net))
1820                         continue;
1821                 /* Decouple optimistic from tentative for evaluation here.
1822                  * Ban optimistic addresses explicitly, when required.
1823                  */
1824                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1825                             ? (ifp->flags&~IFA_F_TENTATIVE)
1826                             : ifp->flags;
1827                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1828                     !(ifp_flags&banned_flags) &&
1829                     (!dev || ifp->idev->dev == dev ||
1830                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1831                         rcu_read_unlock_bh();
1832                         return 1;
1833                 }
1834         }
1835
1836         rcu_read_unlock_bh();
1837         return 0;
1838 }
1839 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1840
1841 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1842                                struct net_device *dev)
1843 {
1844         unsigned int hash = inet6_addr_hash(addr);
1845         struct inet6_ifaddr *ifp;
1846
1847         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1848                 if (!net_eq(dev_net(ifp->idev->dev), net))
1849                         continue;
1850                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1851                         if (!dev || ifp->idev->dev == dev)
1852                                 return true;
1853                 }
1854         }
1855         return false;
1856 }
1857
1858 /* Compares an address/prefix_len with addresses on device @dev.
1859  * If one is found it returns true.
1860  */
1861 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1862         const unsigned int prefix_len, struct net_device *dev)
1863 {
1864         struct inet6_dev *idev;
1865         struct inet6_ifaddr *ifa;
1866         bool ret = false;
1867
1868         rcu_read_lock();
1869         idev = __in6_dev_get(dev);
1870         if (idev) {
1871                 read_lock_bh(&idev->lock);
1872                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1873                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1874                         if (ret)
1875                                 break;
1876                 }
1877                 read_unlock_bh(&idev->lock);
1878         }
1879         rcu_read_unlock();
1880
1881         return ret;
1882 }
1883 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1884
1885 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1886 {
1887         struct inet6_dev *idev;
1888         struct inet6_ifaddr *ifa;
1889         int     onlink;
1890
1891         onlink = 0;
1892         rcu_read_lock();
1893         idev = __in6_dev_get(dev);
1894         if (idev) {
1895                 read_lock_bh(&idev->lock);
1896                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1897                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1898                                                    ifa->prefix_len);
1899                         if (onlink)
1900                                 break;
1901                 }
1902                 read_unlock_bh(&idev->lock);
1903         }
1904         rcu_read_unlock();
1905         return onlink;
1906 }
1907 EXPORT_SYMBOL(ipv6_chk_prefix);
1908
1909 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1910                                      struct net_device *dev, int strict)
1911 {
1912         struct inet6_ifaddr *ifp, *result = NULL;
1913         unsigned int hash = inet6_addr_hash(addr);
1914
1915         rcu_read_lock_bh();
1916         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1917                 if (!net_eq(dev_net(ifp->idev->dev), net))
1918                         continue;
1919                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1920                         if (!dev || ifp->idev->dev == dev ||
1921                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1922                                 result = ifp;
1923                                 in6_ifa_hold(ifp);
1924                                 break;
1925                         }
1926                 }
1927         }
1928         rcu_read_unlock_bh();
1929
1930         return result;
1931 }
1932
1933 /* Gets referenced address, destroys ifaddr */
1934
1935 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1936 {
1937         if (dad_failed)
1938                 ifp->flags |= IFA_F_DADFAILED;
1939
1940         if (ifp->flags&IFA_F_TEMPORARY) {
1941                 struct inet6_ifaddr *ifpub;
1942                 spin_lock_bh(&ifp->lock);
1943                 ifpub = ifp->ifpub;
1944                 if (ifpub) {
1945                         in6_ifa_hold(ifpub);
1946                         spin_unlock_bh(&ifp->lock);
1947                         ipv6_create_tempaddr(ifpub, ifp);
1948                         in6_ifa_put(ifpub);
1949                 } else {
1950                         spin_unlock_bh(&ifp->lock);
1951                 }
1952                 ipv6_del_addr(ifp);
1953         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1954                 spin_lock_bh(&ifp->lock);
1955                 addrconf_del_dad_work(ifp);
1956                 ifp->flags |= IFA_F_TENTATIVE;
1957                 spin_unlock_bh(&ifp->lock);
1958                 if (dad_failed)
1959                         ipv6_ifa_notify(0, ifp);
1960                 in6_ifa_put(ifp);
1961         } else {
1962                 ipv6_del_addr(ifp);
1963         }
1964 }
1965
1966 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1967 {
1968         int err = -ENOENT;
1969
1970         spin_lock_bh(&ifp->lock);
1971         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1972                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1973                 err = 0;
1974         }
1975         spin_unlock_bh(&ifp->lock);
1976
1977         return err;
1978 }
1979
1980 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1981 {
1982         struct inet6_dev *idev = ifp->idev;
1983         struct net *net = dev_net(ifp->idev->dev);
1984
1985         if (addrconf_dad_end(ifp)) {
1986                 in6_ifa_put(ifp);
1987                 return;
1988         }
1989
1990         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1991                              ifp->idev->dev->name, &ifp->addr);
1992
1993         spin_lock_bh(&ifp->lock);
1994
1995         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1996                 int scope = ifp->scope;
1997                 u32 flags = ifp->flags;
1998                 struct in6_addr new_addr;
1999                 struct inet6_ifaddr *ifp2;
2000                 u32 valid_lft, preferred_lft;
2001                 int pfxlen = ifp->prefix_len;
2002                 int retries = ifp->stable_privacy_retry + 1;
2003
2004                 if (retries > net->ipv6.sysctl.idgen_retries) {
2005                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2006                                              ifp->idev->dev->name);
2007                         goto errdad;
2008                 }
2009
2010                 new_addr = ifp->addr;
2011                 if (ipv6_generate_stable_address(&new_addr, retries,
2012                                                  idev))
2013                         goto errdad;
2014
2015                 valid_lft = ifp->valid_lft;
2016                 preferred_lft = ifp->prefered_lft;
2017
2018                 spin_unlock_bh(&ifp->lock);
2019
2020                 if (idev->cnf.max_addresses &&
2021                     ipv6_count_addresses(idev) >=
2022                     idev->cnf.max_addresses)
2023                         goto lock_errdad;
2024
2025                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2026                                      ifp->idev->dev->name);
2027
2028                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2029                                      scope, flags, valid_lft,
2030                                      preferred_lft);
2031                 if (IS_ERR(ifp2))
2032                         goto lock_errdad;
2033
2034                 spin_lock_bh(&ifp2->lock);
2035                 ifp2->stable_privacy_retry = retries;
2036                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2037                 spin_unlock_bh(&ifp2->lock);
2038
2039                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2040                 in6_ifa_put(ifp2);
2041 lock_errdad:
2042                 spin_lock_bh(&ifp->lock);
2043         }
2044
2045 errdad:
2046         /* transition from _POSTDAD to _ERRDAD */
2047         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2048         spin_unlock_bh(&ifp->lock);
2049
2050         addrconf_mod_dad_work(ifp, 0);
2051         in6_ifa_put(ifp);
2052 }
2053
2054 /* Join to solicited addr multicast group.
2055  * caller must hold RTNL */
2056 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2057 {
2058         struct in6_addr maddr;
2059
2060         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2061                 return;
2062
2063         addrconf_addr_solict_mult(addr, &maddr);
2064         ipv6_dev_mc_inc(dev, &maddr);
2065 }
2066
2067 /* caller must hold RTNL */
2068 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2069 {
2070         struct in6_addr maddr;
2071
2072         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2073                 return;
2074
2075         addrconf_addr_solict_mult(addr, &maddr);
2076         __ipv6_dev_mc_dec(idev, &maddr);
2077 }
2078
2079 /* caller must hold RTNL */
2080 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2081 {
2082         struct in6_addr addr;
2083
2084         if (ifp->prefix_len >= 127) /* RFC 6164 */
2085                 return;
2086         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2087         if (ipv6_addr_any(&addr))
2088                 return;
2089         __ipv6_dev_ac_inc(ifp->idev, &addr);
2090 }
2091
2092 /* caller must hold RTNL */
2093 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2094 {
2095         struct in6_addr addr;
2096
2097         if (ifp->prefix_len >= 127) /* RFC 6164 */
2098                 return;
2099         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2100         if (ipv6_addr_any(&addr))
2101                 return;
2102         __ipv6_dev_ac_dec(ifp->idev, &addr);
2103 }
2104
2105 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2106 {
2107         switch (dev->addr_len) {
2108         case ETH_ALEN:
2109                 memcpy(eui, dev->dev_addr, 3);
2110                 eui[3] = 0xFF;
2111                 eui[4] = 0xFE;
2112                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2113                 break;
2114         case EUI64_ADDR_LEN:
2115                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2116                 eui[0] ^= 2;
2117                 break;
2118         default:
2119                 return -1;
2120         }
2121
2122         return 0;
2123 }
2124
2125 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2126 {
2127         union fwnet_hwaddr *ha;
2128
2129         if (dev->addr_len != FWNET_ALEN)
2130                 return -1;
2131
2132         ha = (union fwnet_hwaddr *)dev->dev_addr;
2133
2134         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2135         eui[0] ^= 2;
2136         return 0;
2137 }
2138
2139 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2140 {
2141         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2142         if (dev->addr_len != ARCNET_ALEN)
2143                 return -1;
2144         memset(eui, 0, 7);
2145         eui[7] = *(u8 *)dev->dev_addr;
2146         return 0;
2147 }
2148
2149 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2150 {
2151         if (dev->addr_len != INFINIBAND_ALEN)
2152                 return -1;
2153         memcpy(eui, dev->dev_addr + 12, 8);
2154         eui[0] |= 2;
2155         return 0;
2156 }
2157
2158 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2159 {
2160         if (addr == 0)
2161                 return -1;
2162         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2163                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2164                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2165                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2166                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2167                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2168         eui[1] = 0;
2169         eui[2] = 0x5E;
2170         eui[3] = 0xFE;
2171         memcpy(eui + 4, &addr, 4);
2172         return 0;
2173 }
2174
2175 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2176 {
2177         if (dev->priv_flags & IFF_ISATAP)
2178                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2179         return -1;
2180 }
2181
2182 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2183 {
2184         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2185 }
2186
2187 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2188 {
2189         memcpy(eui, dev->perm_addr, 3);
2190         memcpy(eui + 5, dev->perm_addr + 3, 3);
2191         eui[3] = 0xFF;
2192         eui[4] = 0xFE;
2193         eui[0] ^= 2;
2194         return 0;
2195 }
2196
2197 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2198 {
2199         switch (dev->type) {
2200         case ARPHRD_ETHER:
2201         case ARPHRD_FDDI:
2202                 return addrconf_ifid_eui48(eui, dev);
2203         case ARPHRD_ARCNET:
2204                 return addrconf_ifid_arcnet(eui, dev);
2205         case ARPHRD_INFINIBAND:
2206                 return addrconf_ifid_infiniband(eui, dev);
2207         case ARPHRD_SIT:
2208                 return addrconf_ifid_sit(eui, dev);
2209         case ARPHRD_IPGRE:
2210         case ARPHRD_TUNNEL:
2211                 return addrconf_ifid_gre(eui, dev);
2212         case ARPHRD_6LOWPAN:
2213                 return addrconf_ifid_6lowpan(eui, dev);
2214         case ARPHRD_IEEE1394:
2215                 return addrconf_ifid_ieee1394(eui, dev);
2216         case ARPHRD_TUNNEL6:
2217         case ARPHRD_IP6GRE:
2218                 return addrconf_ifid_ip6tnl(eui, dev);
2219         }
2220         return -1;
2221 }
2222
2223 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2224 {
2225         int err = -1;
2226         struct inet6_ifaddr *ifp;
2227
2228         read_lock_bh(&idev->lock);
2229         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2230                 if (ifp->scope > IFA_LINK)
2231                         break;
2232                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2233                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2234                         err = 0;
2235                         break;
2236                 }
2237         }
2238         read_unlock_bh(&idev->lock);
2239         return err;
2240 }
2241
2242 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2243 static void ipv6_regen_rndid(struct inet6_dev *idev)
2244 {
2245 regen:
2246         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2247         idev->rndid[0] &= ~0x02;
2248
2249         /*
2250          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2251          * check if generated address is not inappropriate
2252          *
2253          *  - Reserved subnet anycast (RFC 2526)
2254          *      11111101 11....11 1xxxxxxx
2255          *  - ISATAP (RFC4214) 6.1
2256          *      00-00-5E-FE-xx-xx-xx-xx
2257          *  - value 0
2258          *  - XXX: already assigned to an address on the device
2259          */
2260         if (idev->rndid[0] == 0xfd &&
2261             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2262             (idev->rndid[7]&0x80))
2263                 goto regen;
2264         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2265                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2266                         goto regen;
2267                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2268                         goto regen;
2269         }
2270 }
2271
2272 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2273 {
2274         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2275                 ipv6_regen_rndid(idev);
2276 }
2277
2278 /*
2279  *      Add prefix route.
2280  */
2281
2282 static void
2283 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2284                       unsigned long expires, u32 flags)
2285 {
2286         struct fib6_config cfg = {
2287                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2288                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2289                 .fc_ifindex = dev->ifindex,
2290                 .fc_expires = expires,
2291                 .fc_dst_len = plen,
2292                 .fc_flags = RTF_UP | flags,
2293                 .fc_nlinfo.nl_net = dev_net(dev),
2294                 .fc_protocol = RTPROT_KERNEL,
2295         };
2296
2297         cfg.fc_dst = *pfx;
2298
2299         /* Prevent useless cloning on PtP SIT.
2300            This thing is done here expecting that the whole
2301            class of non-broadcast devices need not cloning.
2302          */
2303 #if IS_ENABLED(CONFIG_IPV6_SIT)
2304         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2305                 cfg.fc_flags |= RTF_NONEXTHOP;
2306 #endif
2307
2308         ip6_route_add(&cfg, NULL);
2309 }
2310
2311
2312 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2313                                                   int plen,
2314                                                   const struct net_device *dev,
2315                                                   u32 flags, u32 noflags)
2316 {
2317         struct fib6_node *fn;
2318         struct rt6_info *rt = NULL;
2319         struct fib6_table *table;
2320         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2321
2322         table = fib6_get_table(dev_net(dev), tb_id);
2323         if (!table)
2324                 return NULL;
2325
2326         read_lock_bh(&table->tb6_lock);
2327         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2328         if (!fn)
2329                 goto out;
2330
2331         noflags |= RTF_CACHE;
2332         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2333                 if (rt->dst.dev->ifindex != dev->ifindex)
2334                         continue;
2335                 if ((rt->rt6i_flags & flags) != flags)
2336                         continue;
2337                 if ((rt->rt6i_flags & noflags) != 0)
2338                         continue;
2339                 dst_hold(&rt->dst);
2340                 break;
2341         }
2342 out:
2343         read_unlock_bh(&table->tb6_lock);
2344         return rt;
2345 }
2346
2347
2348 /* Create "default" multicast route to the interface */
2349
2350 static void addrconf_add_mroute(struct net_device *dev)
2351 {
2352         struct fib6_config cfg = {
2353                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2354                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2355                 .fc_ifindex = dev->ifindex,
2356                 .fc_dst_len = 8,
2357                 .fc_flags = RTF_UP,
2358                 .fc_nlinfo.nl_net = dev_net(dev),
2359                 .fc_protocol = RTPROT_KERNEL,
2360         };
2361
2362         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2363
2364         ip6_route_add(&cfg, NULL);
2365 }
2366
2367 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2368 {
2369         struct inet6_dev *idev;
2370
2371         ASSERT_RTNL();
2372
2373         idev = ipv6_find_idev(dev);
2374         if (!idev)
2375                 return ERR_PTR(-ENOBUFS);
2376
2377         if (idev->cnf.disable_ipv6)
2378                 return ERR_PTR(-EACCES);
2379
2380         /* Add default multicast route */
2381         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2382                 addrconf_add_mroute(dev);
2383
2384         return idev;
2385 }
2386
2387 static void manage_tempaddrs(struct inet6_dev *idev,
2388                              struct inet6_ifaddr *ifp,
2389                              __u32 valid_lft, __u32 prefered_lft,
2390                              bool create, unsigned long now)
2391 {
2392         u32 flags;
2393         struct inet6_ifaddr *ift;
2394
2395         read_lock_bh(&idev->lock);
2396         /* update all temporary addresses in the list */
2397         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2398                 int age, max_valid, max_prefered;
2399
2400                 if (ifp != ift->ifpub)
2401                         continue;
2402
2403                 /* RFC 4941 section 3.3:
2404                  * If a received option will extend the lifetime of a public
2405                  * address, the lifetimes of temporary addresses should
2406                  * be extended, subject to the overall constraint that no
2407                  * temporary addresses should ever remain "valid" or "preferred"
2408                  * for a time longer than (TEMP_VALID_LIFETIME) or
2409                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2410                  */
2411                 age = (now - ift->cstamp) / HZ;
2412                 max_valid = idev->cnf.temp_valid_lft - age;
2413                 if (max_valid < 0)
2414                         max_valid = 0;
2415
2416                 max_prefered = idev->cnf.temp_prefered_lft -
2417                                idev->desync_factor - age;
2418                 if (max_prefered < 0)
2419                         max_prefered = 0;
2420
2421                 if (valid_lft > max_valid)
2422                         valid_lft = max_valid;
2423
2424                 if (prefered_lft > max_prefered)
2425                         prefered_lft = max_prefered;
2426
2427                 spin_lock(&ift->lock);
2428                 flags = ift->flags;
2429                 ift->valid_lft = valid_lft;
2430                 ift->prefered_lft = prefered_lft;
2431                 ift->tstamp = now;
2432                 if (prefered_lft > 0)
2433                         ift->flags &= ~IFA_F_DEPRECATED;
2434
2435                 spin_unlock(&ift->lock);
2436                 if (!(flags&IFA_F_TENTATIVE))
2437                         ipv6_ifa_notify(0, ift);
2438         }
2439
2440         if ((create || list_empty(&idev->tempaddr_list)) &&
2441             idev->cnf.use_tempaddr > 0) {
2442                 /* When a new public address is created as described
2443                  * in [ADDRCONF], also create a new temporary address.
2444                  * Also create a temporary address if it's enabled but
2445                  * no temporary address currently exists.
2446                  */
2447                 read_unlock_bh(&idev->lock);
2448                 ipv6_create_tempaddr(ifp, NULL);
2449         } else {
2450                 read_unlock_bh(&idev->lock);
2451         }
2452 }
2453
2454 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2455 {
2456         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2457                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2458 }
2459
2460 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2461                                  const struct prefix_info *pinfo,
2462                                  struct inet6_dev *in6_dev,
2463                                  const struct in6_addr *addr, int addr_type,
2464                                  u32 addr_flags, bool sllao, bool tokenized,
2465                                  __u32 valid_lft, u32 prefered_lft)
2466 {
2467         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2468         int create = 0, update_lft = 0;
2469
2470         if (!ifp && valid_lft) {
2471                 int max_addresses = in6_dev->cnf.max_addresses;
2472
2473 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2474                 if ((net->ipv6.devconf_all->optimistic_dad ||
2475                      in6_dev->cnf.optimistic_dad) &&
2476                     !net->ipv6.devconf_all->forwarding && sllao)
2477                         addr_flags |= IFA_F_OPTIMISTIC;
2478 #endif
2479
2480                 /* Do not allow to create too much of autoconfigured
2481                  * addresses; this would be too easy way to crash kernel.
2482                  */
2483                 if (!max_addresses ||
2484                     ipv6_count_addresses(in6_dev) < max_addresses)
2485                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2486                                             pinfo->prefix_len,
2487                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2488                                             addr_flags, valid_lft,
2489                                             prefered_lft);
2490
2491                 if (IS_ERR_OR_NULL(ifp))
2492                         return -1;
2493
2494                 update_lft = 0;
2495                 create = 1;
2496                 spin_lock_bh(&ifp->lock);
2497                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2498                 ifp->cstamp = jiffies;
2499                 ifp->tokenized = tokenized;
2500                 spin_unlock_bh(&ifp->lock);
2501                 addrconf_dad_start(ifp);
2502         }
2503
2504         if (ifp) {
2505                 u32 flags;
2506                 unsigned long now;
2507                 u32 stored_lft;
2508
2509                 /* update lifetime (RFC2462 5.5.3 e) */
2510                 spin_lock_bh(&ifp->lock);
2511                 now = jiffies;
2512                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2513                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2514                 else
2515                         stored_lft = 0;
2516                 if (!update_lft && !create && stored_lft) {
2517                         const u32 minimum_lft = min_t(u32,
2518                                 stored_lft, MIN_VALID_LIFETIME);
2519                         valid_lft = max(valid_lft, minimum_lft);
2520
2521                         /* RFC4862 Section 5.5.3e:
2522                          * "Note that the preferred lifetime of the
2523                          *  corresponding address is always reset to
2524                          *  the Preferred Lifetime in the received
2525                          *  Prefix Information option, regardless of
2526                          *  whether the valid lifetime is also reset or
2527                          *  ignored."
2528                          *
2529                          * So we should always update prefered_lft here.
2530                          */
2531                         update_lft = 1;
2532                 }
2533
2534                 if (update_lft) {
2535                         ifp->valid_lft = valid_lft;
2536                         ifp->prefered_lft = prefered_lft;
2537                         ifp->tstamp = now;
2538                         flags = ifp->flags;
2539                         ifp->flags &= ~IFA_F_DEPRECATED;
2540                         spin_unlock_bh(&ifp->lock);
2541
2542                         if (!(flags&IFA_F_TENTATIVE))
2543                                 ipv6_ifa_notify(0, ifp);
2544                 } else
2545                         spin_unlock_bh(&ifp->lock);
2546
2547                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2548                                  create, now);
2549
2550                 in6_ifa_put(ifp);
2551                 addrconf_verify();
2552         }
2553
2554         return 0;
2555 }
2556 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2557
2558 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2559 {
2560         struct prefix_info *pinfo;
2561         __u32 valid_lft;
2562         __u32 prefered_lft;
2563         int addr_type, err;
2564         u32 addr_flags = 0;
2565         struct inet6_dev *in6_dev;
2566         struct net *net = dev_net(dev);
2567
2568         pinfo = (struct prefix_info *) opt;
2569
2570         if (len < sizeof(struct prefix_info)) {
2571                 ADBG("addrconf: prefix option too short\n");
2572                 return;
2573         }
2574
2575         /*
2576          *      Validation checks ([ADDRCONF], page 19)
2577          */
2578
2579         addr_type = ipv6_addr_type(&pinfo->prefix);
2580
2581         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2582                 return;
2583
2584         valid_lft = ntohl(pinfo->valid);
2585         prefered_lft = ntohl(pinfo->prefered);
2586
2587         if (prefered_lft > valid_lft) {
2588                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2589                 return;
2590         }
2591
2592         in6_dev = in6_dev_get(dev);
2593
2594         if (!in6_dev) {
2595                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2596                                     dev->name);
2597                 return;
2598         }
2599
2600         /*
2601          *      Two things going on here:
2602          *      1) Add routes for on-link prefixes
2603          *      2) Configure prefixes with the auto flag set
2604          */
2605
2606         if (pinfo->onlink) {
2607                 struct rt6_info *rt;
2608                 unsigned long rt_expires;
2609
2610                 /* Avoid arithmetic overflow. Really, we could
2611                  * save rt_expires in seconds, likely valid_lft,
2612                  * but it would require division in fib gc, that it
2613                  * not good.
2614                  */
2615                 if (HZ > USER_HZ)
2616                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2617                 else
2618                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2619
2620                 if (addrconf_finite_timeout(rt_expires))
2621                         rt_expires *= HZ;
2622
2623                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2624                                                pinfo->prefix_len,
2625                                                dev,
2626                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2627                                                RTF_GATEWAY | RTF_DEFAULT);
2628
2629                 if (rt) {
2630                         /* Autoconf prefix route */
2631                         if (valid_lft == 0) {
2632                                 ip6_del_rt(rt);
2633                                 rt = NULL;
2634                         } else if (addrconf_finite_timeout(rt_expires)) {
2635                                 /* not infinity */
2636                                 rt6_set_expires(rt, jiffies + rt_expires);
2637                         } else {
2638                                 rt6_clean_expires(rt);
2639                         }
2640                 } else if (valid_lft) {
2641                         clock_t expires = 0;
2642                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2643                         if (addrconf_finite_timeout(rt_expires)) {
2644                                 /* not infinity */
2645                                 flags |= RTF_EXPIRES;
2646                                 expires = jiffies_to_clock_t(rt_expires);
2647                         }
2648                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2649                                               dev, expires, flags);
2650                 }
2651                 ip6_rt_put(rt);
2652         }
2653
2654         /* Try to figure out our local address for this prefix */
2655
2656         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2657                 struct in6_addr addr;
2658                 bool tokenized = false, dev_addr_generated = false;
2659
2660                 if (pinfo->prefix_len == 64) {
2661                         memcpy(&addr, &pinfo->prefix, 8);
2662
2663                         if (!ipv6_addr_any(&in6_dev->token)) {
2664                                 read_lock_bh(&in6_dev->lock);
2665                                 memcpy(addr.s6_addr + 8,
2666                                        in6_dev->token.s6_addr + 8, 8);
2667                                 read_unlock_bh(&in6_dev->lock);
2668                                 tokenized = true;
2669                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2670                                    !ipv6_generate_stable_address(&addr, 0,
2671                                                                  in6_dev)) {
2672                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2673                                 goto ok;
2674                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2675                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2676                                 goto put;
2677                         } else {
2678                                 dev_addr_generated = true;
2679                         }
2680                         goto ok;
2681                 }
2682                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2683                                     pinfo->prefix_len);
2684                 goto put;
2685
2686 ok:
2687                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2688                                                    &addr, addr_type,
2689                                                    addr_flags, sllao,
2690                                                    tokenized, valid_lft,
2691                                                    prefered_lft);
2692                 if (err)
2693                         goto put;
2694
2695                 /* Ignore error case here because previous prefix add addr was
2696                  * successful which will be notified.
2697                  */
2698                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2699                                               addr_type, addr_flags, sllao,
2700                                               tokenized, valid_lft,
2701                                               prefered_lft,
2702                                               dev_addr_generated);
2703         }
2704         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2705 put:
2706         in6_dev_put(in6_dev);
2707 }
2708
2709 /*
2710  *      Set destination address.
2711  *      Special case for SIT interfaces where we create a new "virtual"
2712  *      device.
2713  */
2714 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2715 {
2716         struct in6_ifreq ireq;
2717         struct net_device *dev;
2718         int err = -EINVAL;
2719
2720         rtnl_lock();
2721
2722         err = -EFAULT;
2723         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2724                 goto err_exit;
2725
2726         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2727
2728         err = -ENODEV;
2729         if (!dev)
2730                 goto err_exit;
2731
2732 #if IS_ENABLED(CONFIG_IPV6_SIT)
2733         if (dev->type == ARPHRD_SIT) {
2734                 const struct net_device_ops *ops = dev->netdev_ops;
2735                 struct ifreq ifr;
2736                 struct ip_tunnel_parm p;
2737
2738                 err = -EADDRNOTAVAIL;
2739                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2740                         goto err_exit;
2741
2742                 memset(&p, 0, sizeof(p));
2743                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2744                 p.iph.saddr = 0;
2745                 p.iph.version = 4;
2746                 p.iph.ihl = 5;
2747                 p.iph.protocol = IPPROTO_IPV6;
2748                 p.iph.ttl = 64;
2749                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2750
2751                 if (ops->ndo_do_ioctl) {
2752                         mm_segment_t oldfs = get_fs();
2753
2754                         set_fs(KERNEL_DS);
2755                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2756                         set_fs(oldfs);
2757                 } else
2758                         err = -EOPNOTSUPP;
2759
2760                 if (err == 0) {
2761                         err = -ENOBUFS;
2762                         dev = __dev_get_by_name(net, p.name);
2763                         if (!dev)
2764                                 goto err_exit;
2765                         err = dev_open(dev);
2766                 }
2767         }
2768 #endif
2769
2770 err_exit:
2771         rtnl_unlock();
2772         return err;
2773 }
2774
2775 static int ipv6_mc_config(struct sock *sk, bool join,
2776                           const struct in6_addr *addr, int ifindex)
2777 {
2778         int ret;
2779
2780         ASSERT_RTNL();
2781
2782         lock_sock(sk);
2783         if (join)
2784                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2785         else
2786                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2787         release_sock(sk);
2788
2789         return ret;
2790 }
2791
2792 /*
2793  *      Manual configuration of address on an interface
2794  */
2795 static int inet6_addr_add(struct net *net, int ifindex,
2796                           const struct in6_addr *pfx,
2797                           const struct in6_addr *peer_pfx,
2798                           unsigned int plen, __u32 ifa_flags,
2799                           __u32 prefered_lft, __u32 valid_lft)
2800 {
2801         struct inet6_ifaddr *ifp;
2802         struct inet6_dev *idev;
2803         struct net_device *dev;
2804         unsigned long timeout;
2805         clock_t expires;
2806         int scope;
2807         u32 flags;
2808
2809         ASSERT_RTNL();
2810
2811         if (plen > 128)
2812                 return -EINVAL;
2813
2814         /* check the lifetime */
2815         if (!valid_lft || prefered_lft > valid_lft)
2816                 return -EINVAL;
2817
2818         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2819                 return -EINVAL;
2820
2821         dev = __dev_get_by_index(net, ifindex);
2822         if (!dev)
2823                 return -ENODEV;
2824
2825         idev = addrconf_add_dev(dev);
2826         if (IS_ERR(idev))
2827                 return PTR_ERR(idev);
2828
2829         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2830                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2831                                          true, pfx, ifindex);
2832
2833                 if (ret < 0)
2834                         return ret;
2835         }
2836
2837         scope = ipv6_addr_scope(pfx);
2838
2839         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2840         if (addrconf_finite_timeout(timeout)) {
2841                 expires = jiffies_to_clock_t(timeout * HZ);
2842                 valid_lft = timeout;
2843                 flags = RTF_EXPIRES;
2844         } else {
2845                 expires = 0;
2846                 flags = 0;
2847                 ifa_flags |= IFA_F_PERMANENT;
2848         }
2849
2850         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2851         if (addrconf_finite_timeout(timeout)) {
2852                 if (timeout == 0)
2853                         ifa_flags |= IFA_F_DEPRECATED;
2854                 prefered_lft = timeout;
2855         }
2856
2857         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2858                             valid_lft, prefered_lft);
2859
2860         if (!IS_ERR(ifp)) {
2861                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2862                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2863                                               expires, flags);
2864                 }
2865
2866                 /*
2867                  * Note that section 3.1 of RFC 4429 indicates
2868                  * that the Optimistic flag should not be set for
2869                  * manually configured addresses
2870                  */
2871                 addrconf_dad_start(ifp);
2872                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2873                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2874                                          true, jiffies);
2875                 in6_ifa_put(ifp);
2876                 addrconf_verify_rtnl();
2877                 return 0;
2878         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2879                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2880                                false, pfx, ifindex);
2881         }
2882
2883         return PTR_ERR(ifp);
2884 }
2885
2886 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2887                           const struct in6_addr *pfx, unsigned int plen)
2888 {
2889         struct inet6_ifaddr *ifp;
2890         struct inet6_dev *idev;
2891         struct net_device *dev;
2892
2893         if (plen > 128)
2894                 return -EINVAL;
2895
2896         dev = __dev_get_by_index(net, ifindex);
2897         if (!dev)
2898                 return -ENODEV;
2899
2900         idev = __in6_dev_get(dev);
2901         if (!idev)
2902                 return -ENXIO;
2903
2904         read_lock_bh(&idev->lock);
2905         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2906                 if (ifp->prefix_len == plen &&
2907                     ipv6_addr_equal(pfx, &ifp->addr)) {
2908                         in6_ifa_hold(ifp);
2909                         read_unlock_bh(&idev->lock);
2910
2911                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2912                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2913                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2914                                                  jiffies);
2915                         ipv6_del_addr(ifp);
2916                         addrconf_verify_rtnl();
2917                         if (ipv6_addr_is_multicast(pfx)) {
2918                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2919                                                false, pfx, dev->ifindex);
2920                         }
2921                         return 0;
2922                 }
2923         }
2924         read_unlock_bh(&idev->lock);
2925         return -EADDRNOTAVAIL;
2926 }
2927
2928
2929 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2930 {
2931         struct in6_ifreq ireq;
2932         int err;
2933
2934         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2935                 return -EPERM;
2936
2937         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2938                 return -EFAULT;
2939
2940         rtnl_lock();
2941         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2942                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2943                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2944         rtnl_unlock();
2945         return err;
2946 }
2947
2948 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2949 {
2950         struct in6_ifreq ireq;
2951         int err;
2952
2953         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2954                 return -EPERM;
2955
2956         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2957                 return -EFAULT;
2958
2959         rtnl_lock();
2960         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2961                              ireq.ifr6_prefixlen);
2962         rtnl_unlock();
2963         return err;
2964 }
2965
2966 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2967                      int plen, int scope)
2968 {
2969         struct inet6_ifaddr *ifp;
2970
2971         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2972                             scope, IFA_F_PERMANENT,
2973                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2974         if (!IS_ERR(ifp)) {
2975                 spin_lock_bh(&ifp->lock);
2976                 ifp->flags &= ~IFA_F_TENTATIVE;
2977                 spin_unlock_bh(&ifp->lock);
2978                 rt_genid_bump_ipv6(dev_net(idev->dev));
2979                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2980                 in6_ifa_put(ifp);
2981         }
2982 }
2983
2984 #if IS_ENABLED(CONFIG_IPV6_SIT)
2985 static void sit_add_v4_addrs(struct inet6_dev *idev)
2986 {
2987         struct in6_addr addr;
2988         struct net_device *dev;
2989         struct net *net = dev_net(idev->dev);
2990         int scope, plen;
2991         u32 pflags = 0;
2992
2993         ASSERT_RTNL();
2994
2995         memset(&addr, 0, sizeof(struct in6_addr));
2996         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2997
2998         if (idev->dev->flags&IFF_POINTOPOINT) {
2999                 addr.s6_addr32[0] = htonl(0xfe800000);
3000                 scope = IFA_LINK;
3001                 plen = 64;
3002         } else {
3003                 scope = IPV6_ADDR_COMPATv4;
3004                 plen = 96;
3005                 pflags |= RTF_NONEXTHOP;
3006         }
3007
3008         if (addr.s6_addr32[3]) {
3009                 add_addr(idev, &addr, plen, scope);
3010                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3011                 return;
3012         }
3013
3014         for_each_netdev(net, dev) {
3015                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3016                 if (in_dev && (dev->flags & IFF_UP)) {
3017                         struct in_ifaddr *ifa;
3018
3019                         int flag = scope;
3020
3021                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3022
3023                                 addr.s6_addr32[3] = ifa->ifa_local;
3024
3025                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3026                                         continue;
3027                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3028                                         if (idev->dev->flags&IFF_POINTOPOINT)
3029                                                 continue;
3030                                         flag |= IFA_HOST;
3031                                 }
3032
3033                                 add_addr(idev, &addr, plen, flag);
3034                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3035                                                       pflags);
3036                         }
3037                 }
3038         }
3039 }
3040 #endif
3041
3042 static void init_loopback(struct net_device *dev)
3043 {
3044         struct inet6_dev  *idev;
3045
3046         /* ::1 */
3047
3048         ASSERT_RTNL();
3049
3050         idev = ipv6_find_idev(dev);
3051         if (!idev) {
3052                 pr_debug("%s: add_dev failed\n", __func__);
3053                 return;
3054         }
3055
3056         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3057 }
3058
3059 void addrconf_add_linklocal(struct inet6_dev *idev,
3060                             const struct in6_addr *addr, u32 flags)
3061 {
3062         struct inet6_ifaddr *ifp;
3063         u32 addr_flags = flags | IFA_F_PERMANENT;
3064
3065 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3066         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3067              idev->cnf.optimistic_dad) &&
3068             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3069                 addr_flags |= IFA_F_OPTIMISTIC;
3070 #endif
3071
3072         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3073                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3074         if (!IS_ERR(ifp)) {
3075                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3076                 addrconf_dad_start(ifp);
3077                 in6_ifa_put(ifp);
3078         }
3079 }
3080 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3081
3082 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3083 {
3084         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3085                 return true;
3086
3087         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3088             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3089                 return true;
3090
3091         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3092             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3093                 return true;
3094
3095         return false;
3096 }
3097
3098 static int ipv6_generate_stable_address(struct in6_addr *address,
3099                                         u8 dad_count,
3100                                         const struct inet6_dev *idev)
3101 {
3102         static DEFINE_SPINLOCK(lock);
3103         static __u32 digest[SHA_DIGEST_WORDS];
3104         static __u32 workspace[SHA_WORKSPACE_WORDS];
3105
3106         static union {
3107                 char __data[SHA_MESSAGE_BYTES];
3108                 struct {
3109                         struct in6_addr secret;
3110                         __be32 prefix[2];
3111                         unsigned char hwaddr[MAX_ADDR_LEN];
3112                         u8 dad_count;
3113                 } __packed;
3114         } data;
3115
3116         struct in6_addr secret;
3117         struct in6_addr temp;
3118         struct net *net = dev_net(idev->dev);
3119
3120         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3121
3122         if (idev->cnf.stable_secret.initialized)
3123                 secret = idev->cnf.stable_secret.secret;
3124         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3125                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3126         else
3127                 return -1;
3128
3129 retry:
3130         spin_lock_bh(&lock);
3131
3132         sha_init(digest);
3133         memset(&data, 0, sizeof(data));
3134         memset(workspace, 0, sizeof(workspace));
3135         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3136         data.prefix[0] = address->s6_addr32[0];
3137         data.prefix[1] = address->s6_addr32[1];
3138         data.secret = secret;
3139         data.dad_count = dad_count;
3140
3141         sha_transform(digest, data.__data, workspace);
3142
3143         temp = *address;
3144         temp.s6_addr32[2] = (__force __be32)digest[0];
3145         temp.s6_addr32[3] = (__force __be32)digest[1];
3146
3147         spin_unlock_bh(&lock);
3148
3149         if (ipv6_reserved_interfaceid(temp)) {
3150                 dad_count++;
3151                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3152                         return -1;
3153                 goto retry;
3154         }
3155
3156         *address = temp;
3157         return 0;
3158 }
3159
3160 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3161 {
3162         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3163
3164         if (s->initialized)
3165                 return;
3166         s = &idev->cnf.stable_secret;
3167         get_random_bytes(&s->secret, sizeof(s->secret));
3168         s->initialized = true;
3169 }
3170
3171 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3172 {
3173         struct in6_addr addr;
3174
3175         /* no link local addresses on L3 master devices */
3176         if (netif_is_l3_master(idev->dev))
3177                 return;
3178
3179         /* no link local addresses on devices flagged as slaves */
3180         if (idev->dev->flags & IFF_SLAVE)
3181                 return;
3182
3183         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3184
3185         switch (idev->cnf.addr_gen_mode) {
3186         case IN6_ADDR_GEN_MODE_RANDOM:
3187                 ipv6_gen_mode_random_init(idev);
3188                 /* fallthrough */
3189         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3190                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3191                         addrconf_add_linklocal(idev, &addr,
3192                                                IFA_F_STABLE_PRIVACY);
3193                 else if (prefix_route)
3194                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3195                 break;
3196         case IN6_ADDR_GEN_MODE_EUI64:
3197                 /* addrconf_add_linklocal also adds a prefix_route and we
3198                  * only need to care about prefix routes if ipv6_generate_eui64
3199                  * couldn't generate one.
3200                  */
3201                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3202                         addrconf_add_linklocal(idev, &addr, 0);
3203                 else if (prefix_route)
3204                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3205                 break;
3206         case IN6_ADDR_GEN_MODE_NONE:
3207         default:
3208                 /* will not add any link local address */
3209                 break;
3210         }
3211 }
3212
3213 static void addrconf_dev_config(struct net_device *dev)
3214 {
3215         struct inet6_dev *idev;
3216
3217         ASSERT_RTNL();
3218
3219         if ((dev->type != ARPHRD_ETHER) &&
3220             (dev->type != ARPHRD_FDDI) &&
3221             (dev->type != ARPHRD_ARCNET) &&
3222             (dev->type != ARPHRD_INFINIBAND) &&
3223             (dev->type != ARPHRD_IEEE1394) &&
3224             (dev->type != ARPHRD_TUNNEL6) &&
3225             (dev->type != ARPHRD_6LOWPAN) &&
3226             (dev->type != ARPHRD_IP6GRE) &&
3227             (dev->type != ARPHRD_IPGRE) &&
3228             (dev->type != ARPHRD_TUNNEL) &&
3229             (dev->type != ARPHRD_NONE)) {
3230                 /* Alas, we support only Ethernet autoconfiguration. */
3231                 idev = __in6_dev_get(dev);
3232                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3233                     dev->flags & IFF_MULTICAST)
3234                         ipv6_mc_up(idev);
3235                 return;
3236         }
3237
3238         idev = addrconf_add_dev(dev);
3239         if (IS_ERR(idev))
3240                 return;
3241
3242         /* this device type has no EUI support */
3243         if (dev->type == ARPHRD_NONE &&
3244             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3245                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3246
3247         addrconf_addr_gen(idev, false);
3248 }
3249
3250 #if IS_ENABLED(CONFIG_IPV6_SIT)
3251 static void addrconf_sit_config(struct net_device *dev)
3252 {
3253         struct inet6_dev *idev;
3254
3255         ASSERT_RTNL();
3256
3257         /*
3258          * Configure the tunnel with one of our IPv4
3259          * addresses... we should configure all of
3260          * our v4 addrs in the tunnel
3261          */
3262
3263         idev = ipv6_find_idev(dev);
3264         if (!idev) {
3265                 pr_debug("%s: add_dev failed\n", __func__);
3266                 return;
3267         }
3268
3269         if (dev->priv_flags & IFF_ISATAP) {
3270                 addrconf_addr_gen(idev, false);
3271                 return;
3272         }
3273
3274         sit_add_v4_addrs(idev);
3275
3276         if (dev->flags&IFF_POINTOPOINT)
3277                 addrconf_add_mroute(dev);
3278 }
3279 #endif
3280
3281 #if IS_ENABLED(CONFIG_NET_IPGRE)
3282 static void addrconf_gre_config(struct net_device *dev)
3283 {
3284         struct inet6_dev *idev;
3285
3286         ASSERT_RTNL();
3287
3288         idev = ipv6_find_idev(dev);
3289         if (!idev) {
3290                 pr_debug("%s: add_dev failed\n", __func__);
3291                 return;
3292         }
3293
3294         addrconf_addr_gen(idev, true);
3295         if (dev->flags & IFF_POINTOPOINT)
3296                 addrconf_add_mroute(dev);
3297 }
3298 #endif
3299
3300 static int fixup_permanent_addr(struct inet6_dev *idev,
3301                                 struct inet6_ifaddr *ifp)
3302 {
3303         /* !rt6i_node means the host route was removed from the
3304          * FIB, for example, if 'lo' device is taken down. In that
3305          * case regenerate the host route.
3306          */
3307         if (!ifp->rt || !ifp->rt->rt6i_node) {
3308                 struct rt6_info *rt, *prev;
3309
3310                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3311                 if (unlikely(IS_ERR(rt)))
3312                         return PTR_ERR(rt);
3313
3314                 /* ifp->rt can be accessed outside of rtnl */
3315                 spin_lock(&ifp->lock);
3316                 prev = ifp->rt;
3317                 ifp->rt = rt;
3318                 spin_unlock(&ifp->lock);
3319
3320                 ip6_rt_put(prev);
3321         }
3322
3323         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3324                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3325                                       idev->dev, 0, 0);
3326         }
3327
3328         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3329                 addrconf_dad_start(ifp);
3330
3331         return 0;
3332 }
3333
3334 static void addrconf_permanent_addr(struct net_device *dev)
3335 {
3336         struct inet6_ifaddr *ifp, *tmp;
3337         struct inet6_dev *idev;
3338
3339         idev = __in6_dev_get(dev);
3340         if (!idev)
3341                 return;
3342
3343         write_lock_bh(&idev->lock);
3344
3345         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3346                 if ((ifp->flags & IFA_F_PERMANENT) &&
3347                     fixup_permanent_addr(idev, ifp) < 0) {
3348                         write_unlock_bh(&idev->lock);
3349                         in6_ifa_hold(ifp);
3350                         ipv6_del_addr(ifp);
3351                         write_lock_bh(&idev->lock);
3352
3353                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3354                                              idev->dev->name, &ifp->addr);
3355                 }
3356         }
3357
3358         write_unlock_bh(&idev->lock);
3359 }
3360
3361 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3362                            void *ptr)
3363 {
3364         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3365         struct netdev_notifier_changeupper_info *info;
3366         struct inet6_dev *idev = __in6_dev_get(dev);
3367         struct net *net = dev_net(dev);
3368         int run_pending = 0;
3369         int err;
3370
3371         switch (event) {
3372         case NETDEV_REGISTER:
3373                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3374                         idev = ipv6_add_dev(dev);
3375                         if (IS_ERR(idev))
3376                                 return notifier_from_errno(PTR_ERR(idev));
3377                 }
3378                 break;
3379
3380         case NETDEV_CHANGEMTU:
3381                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3382                 if (dev->mtu < IPV6_MIN_MTU) {
3383                         addrconf_ifdown(dev, dev != net->loopback_dev);
3384                         break;
3385                 }
3386
3387                 if (idev) {
3388                         rt6_mtu_change(dev, dev->mtu);
3389                         idev->cnf.mtu6 = dev->mtu;
3390                         break;
3391                 }
3392
3393                 /* allocate new idev */
3394                 idev = ipv6_add_dev(dev);
3395                 if (IS_ERR(idev))
3396                         break;
3397
3398                 /* device is still not ready */
3399                 if (!(idev->if_flags & IF_READY))
3400                         break;
3401
3402                 run_pending = 1;
3403
3404                 /* fall through */
3405
3406         case NETDEV_UP:
3407         case NETDEV_CHANGE:
3408                 if (dev->flags & IFF_SLAVE)
3409                         break;
3410
3411                 if (idev && idev->cnf.disable_ipv6)
3412                         break;
3413
3414                 if (event == NETDEV_UP) {
3415                         /* restore routes for permanent addresses */
3416                         addrconf_permanent_addr(dev);
3417
3418                         if (!addrconf_link_ready(dev)) {
3419                                 /* device is not ready yet. */
3420                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3421                                         dev->name);
3422                                 break;
3423                         }
3424
3425                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3426                                 idev = ipv6_add_dev(dev);
3427
3428                         if (!IS_ERR_OR_NULL(idev)) {
3429                                 idev->if_flags |= IF_READY;
3430                                 run_pending = 1;
3431                         }
3432                 } else if (event == NETDEV_CHANGE) {
3433                         if (!addrconf_link_ready(dev)) {
3434                                 /* device is still not ready. */
3435                                 break;
3436                         }
3437
3438                         if (idev) {
3439                                 if (idev->if_flags & IF_READY) {
3440                                         /* device is already configured -
3441                                          * but resend MLD reports, we might
3442                                          * have roamed and need to update
3443                                          * multicast snooping switches
3444                                          */
3445                                         ipv6_mc_up(idev);
3446                                         break;
3447                                 }
3448                                 idev->if_flags |= IF_READY;
3449                         }
3450
3451                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3452                                 dev->name);
3453
3454                         run_pending = 1;
3455                 }
3456
3457                 switch (dev->type) {
3458 #if IS_ENABLED(CONFIG_IPV6_SIT)
3459                 case ARPHRD_SIT:
3460                         addrconf_sit_config(dev);
3461                         break;
3462 #endif
3463 #if IS_ENABLED(CONFIG_NET_IPGRE)
3464                 case ARPHRD_IPGRE:
3465                         addrconf_gre_config(dev);
3466                         break;
3467 #endif
3468                 case ARPHRD_LOOPBACK:
3469                         init_loopback(dev);
3470                         break;
3471
3472                 default:
3473                         addrconf_dev_config(dev);
3474                         break;
3475                 }
3476
3477                 if (!IS_ERR_OR_NULL(idev)) {
3478                         if (run_pending)
3479                                 addrconf_dad_run(idev);
3480
3481                         /*
3482                          * If the MTU changed during the interface down,
3483                          * when the interface up, the changed MTU must be
3484                          * reflected in the idev as well as routers.
3485                          */
3486                         if (idev->cnf.mtu6 != dev->mtu &&
3487                             dev->mtu >= IPV6_MIN_MTU) {
3488                                 rt6_mtu_change(dev, dev->mtu);
3489                                 idev->cnf.mtu6 = dev->mtu;
3490                         }
3491                         idev->tstamp = jiffies;
3492                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3493
3494                         /*
3495                          * If the changed mtu during down is lower than
3496                          * IPV6_MIN_MTU stop IPv6 on this interface.
3497                          */
3498                         if (dev->mtu < IPV6_MIN_MTU)
3499                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3500                 }
3501                 break;
3502
3503         case NETDEV_DOWN:
3504         case NETDEV_UNREGISTER:
3505                 /*
3506                  *      Remove all addresses from this interface.
3507                  */
3508                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3509                 break;
3510
3511         case NETDEV_CHANGENAME:
3512                 if (idev) {
3513                         snmp6_unregister_dev(idev);
3514                         addrconf_sysctl_unregister(idev);
3515                         err = addrconf_sysctl_register(idev);
3516                         if (err)
3517                                 return notifier_from_errno(err);
3518                         err = snmp6_register_dev(idev);
3519                         if (err) {
3520                                 addrconf_sysctl_unregister(idev);
3521                                 return notifier_from_errno(err);
3522                         }
3523                 }
3524                 break;
3525
3526         case NETDEV_PRE_TYPE_CHANGE:
3527         case NETDEV_POST_TYPE_CHANGE:
3528                 if (idev)
3529                         addrconf_type_change(dev, event);
3530                 break;
3531
3532         case NETDEV_CHANGEUPPER:
3533                 info = ptr;
3534
3535                 /* flush all routes if dev is linked to or unlinked from
3536                  * an L3 master device (e.g., VRF)
3537                  */
3538                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3539                         addrconf_ifdown(dev, 0);
3540         }
3541
3542         return NOTIFY_OK;
3543 }
3544
3545 /*
3546  *      addrconf module should be notified of a device going up
3547  */
3548 static struct notifier_block ipv6_dev_notf = {
3549         .notifier_call = addrconf_notify,
3550         .priority = ADDRCONF_NOTIFY_PRIORITY,
3551 };
3552
3553 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3554 {
3555         struct inet6_dev *idev;
3556         ASSERT_RTNL();
3557
3558         idev = __in6_dev_get(dev);
3559
3560         if (event == NETDEV_POST_TYPE_CHANGE)
3561                 ipv6_mc_remap(idev);
3562         else if (event == NETDEV_PRE_TYPE_CHANGE)
3563                 ipv6_mc_unmap(idev);
3564 }
3565
3566 static bool addr_is_local(const struct in6_addr *addr)
3567 {
3568         return ipv6_addr_type(addr) &
3569                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3570 }
3571
3572 static int addrconf_ifdown(struct net_device *dev, int how)
3573 {
3574         struct net *net = dev_net(dev);
3575         struct inet6_dev *idev;
3576         struct inet6_ifaddr *ifa, *tmp;
3577         struct list_head del_list;
3578         int _keep_addr;
3579         bool keep_addr;
3580         int state, i;
3581
3582         ASSERT_RTNL();
3583
3584         rt6_ifdown(net, dev);
3585         neigh_ifdown(&nd_tbl, dev);
3586
3587         idev = __in6_dev_get(dev);
3588         if (!idev)
3589                 return -ENODEV;
3590
3591         /*
3592          * Step 1: remove reference to ipv6 device from parent device.
3593          *         Do not dev_put!
3594          */
3595         if (how) {
3596                 idev->dead = 1;
3597
3598                 /* protected by rtnl_lock */
3599                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3600
3601                 /* Step 1.5: remove snmp6 entry */
3602                 snmp6_unregister_dev(idev);
3603
3604         }
3605
3606         /* aggregate the system setting and interface setting */
3607         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3608         if (!_keep_addr)
3609                 _keep_addr = idev->cnf.keep_addr_on_down;
3610
3611         /* combine the user config with event to determine if permanent
3612          * addresses are to be removed from address hash table
3613          */
3614         keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3615
3616         /* Step 2: clear hash table */
3617         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3618                 struct hlist_head *h = &inet6_addr_lst[i];
3619
3620                 spin_lock_bh(&addrconf_hash_lock);
3621 restart:
3622                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3623                         if (ifa->idev == idev) {
3624                                 addrconf_del_dad_work(ifa);
3625                                 /* combined flag + permanent flag decide if
3626                                  * address is retained on a down event
3627                                  */
3628                                 if (!keep_addr ||
3629                                     !(ifa->flags & IFA_F_PERMANENT) ||
3630                                     addr_is_local(&ifa->addr)) {
3631                                         hlist_del_init_rcu(&ifa->addr_lst);
3632                                         goto restart;
3633                                 }
3634                         }
3635                 }
3636                 spin_unlock_bh(&addrconf_hash_lock);
3637         }
3638
3639         write_lock_bh(&idev->lock);
3640
3641         addrconf_del_rs_timer(idev);
3642
3643         /* Step 2: clear flags for stateless addrconf */
3644         if (!how)
3645                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3646
3647         /* Step 3: clear tempaddr list */
3648         while (!list_empty(&idev->tempaddr_list)) {
3649                 ifa = list_first_entry(&idev->tempaddr_list,
3650                                        struct inet6_ifaddr, tmp_list);
3651                 list_del(&ifa->tmp_list);
3652                 write_unlock_bh(&idev->lock);
3653                 spin_lock_bh(&ifa->lock);
3654
3655                 if (ifa->ifpub) {
3656                         in6_ifa_put(ifa->ifpub);
3657                         ifa->ifpub = NULL;
3658                 }
3659                 spin_unlock_bh(&ifa->lock);
3660                 in6_ifa_put(ifa);
3661                 write_lock_bh(&idev->lock);
3662         }
3663
3664         /* re-combine the user config with event to determine if permanent
3665          * addresses are to be removed from the interface list
3666          */
3667         keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3668
3669         INIT_LIST_HEAD(&del_list);
3670         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3671                 struct rt6_info *rt = NULL;
3672                 bool keep;
3673
3674                 addrconf_del_dad_work(ifa);
3675
3676                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3677                         !addr_is_local(&ifa->addr);
3678                 if (!keep)
3679                         list_move(&ifa->if_list, &del_list);
3680
3681                 write_unlock_bh(&idev->lock);
3682                 spin_lock_bh(&ifa->lock);
3683
3684                 if (keep) {
3685                         /* set state to skip the notifier below */
3686                         state = INET6_IFADDR_STATE_DEAD;
3687                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3688                         if (!(ifa->flags & IFA_F_NODAD))
3689                                 ifa->flags |= IFA_F_TENTATIVE;
3690
3691                         rt = ifa->rt;
3692                         ifa->rt = NULL;
3693                 } else {
3694                         state = ifa->state;
3695                         ifa->state = INET6_IFADDR_STATE_DEAD;
3696                 }
3697
3698                 spin_unlock_bh(&ifa->lock);
3699
3700                 if (rt)
3701                         ip6_del_rt(rt);
3702
3703                 if (state != INET6_IFADDR_STATE_DEAD) {
3704                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3705                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3706                 } else {
3707                         if (idev->cnf.forwarding)
3708                                 addrconf_leave_anycast(ifa);
3709                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3710                 }
3711
3712                 write_lock_bh(&idev->lock);
3713         }
3714
3715         write_unlock_bh(&idev->lock);
3716
3717         /* now clean up addresses to be removed */
3718         while (!list_empty(&del_list)) {
3719                 ifa = list_first_entry(&del_list,
3720                                        struct inet6_ifaddr, if_list);
3721                 list_del(&ifa->if_list);
3722
3723                 in6_ifa_put(ifa);
3724         }
3725
3726         /* Step 5: Discard anycast and multicast list */
3727         if (how) {
3728                 ipv6_ac_destroy_dev(idev);
3729                 ipv6_mc_destroy_dev(idev);
3730         } else {
3731                 ipv6_mc_down(idev);
3732         }
3733
3734         idev->tstamp = jiffies;
3735
3736         /* Last: Shot the device (if unregistered) */
3737         if (how) {
3738                 addrconf_sysctl_unregister(idev);
3739                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3740                 neigh_ifdown(&nd_tbl, dev);
3741                 in6_dev_put(idev);
3742         }
3743         return 0;
3744 }
3745
3746 static void addrconf_rs_timer(unsigned long data)
3747 {
3748         struct inet6_dev *idev = (struct inet6_dev *)data;
3749         struct net_device *dev = idev->dev;
3750         struct in6_addr lladdr;
3751
3752         write_lock(&idev->lock);
3753         if (idev->dead || !(idev->if_flags & IF_READY))
3754                 goto out;
3755
3756         if (!ipv6_accept_ra(idev))
3757                 goto out;
3758
3759         /* Announcement received after solicitation was sent */
3760         if (idev->if_flags & IF_RA_RCVD)
3761                 goto out;
3762
3763         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3764                 write_unlock(&idev->lock);
3765                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3766                         ndisc_send_rs(dev, &lladdr,
3767                                       &in6addr_linklocal_allrouters);
3768                 else
3769                         goto put;
3770
3771                 write_lock(&idev->lock);
3772                 idev->rs_interval = rfc3315_s14_backoff_update(
3773                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3774                 /* The wait after the last probe can be shorter */
3775                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3776                                              idev->cnf.rtr_solicits) ?
3777                                       idev->cnf.rtr_solicit_delay :
3778                                       idev->rs_interval);
3779         } else {
3780                 /*
3781                  * Note: we do not support deprecated "all on-link"
3782                  * assumption any longer.
3783                  */
3784                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3785         }
3786
3787 out:
3788         write_unlock(&idev->lock);
3789 put:
3790         in6_dev_put(idev);
3791 }
3792
3793 /*
3794  *      Duplicate Address Detection
3795  */
3796 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3797 {
3798         unsigned long rand_num;
3799         struct inet6_dev *idev = ifp->idev;
3800         u64 nonce;
3801
3802         if (ifp->flags & IFA_F_OPTIMISTIC)
3803                 rand_num = 0;
3804         else
3805                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3806
3807         nonce = 0;
3808         if (idev->cnf.enhanced_dad ||
3809             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3810                 do
3811                         get_random_bytes(&nonce, 6);
3812                 while (nonce == 0);
3813         }
3814         ifp->dad_nonce = nonce;
3815         ifp->dad_probes = idev->cnf.dad_transmits;
3816         addrconf_mod_dad_work(ifp, rand_num);
3817 }
3818
3819 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3820 {
3821         struct inet6_dev *idev = ifp->idev;
3822         struct net_device *dev = idev->dev;
3823         bool bump_id, notify = false;
3824
3825         addrconf_join_solict(dev, &ifp->addr);
3826
3827         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3828
3829         read_lock_bh(&idev->lock);
3830         spin_lock(&ifp->lock);
3831         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3832                 goto out;
3833
3834         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3835             (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3836              idev->cnf.accept_dad < 1) ||
3837             !(ifp->flags&IFA_F_TENTATIVE) ||
3838             ifp->flags & IFA_F_NODAD) {
3839                 bool send_na = false;
3840
3841                 if (ifp->flags & IFA_F_TENTATIVE &&
3842                     !(ifp->flags & IFA_F_OPTIMISTIC))
3843                         send_na = true;
3844                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3845                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3846                 spin_unlock(&ifp->lock);
3847                 read_unlock_bh(&idev->lock);
3848
3849                 addrconf_dad_completed(ifp, bump_id, send_na);
3850                 return;
3851         }
3852
3853         if (!(idev->if_flags & IF_READY)) {
3854                 spin_unlock(&ifp->lock);
3855                 read_unlock_bh(&idev->lock);
3856                 /*
3857                  * If the device is not ready:
3858                  * - keep it tentative if it is a permanent address.
3859                  * - otherwise, kill it.
3860                  */
3861                 in6_ifa_hold(ifp);
3862                 addrconf_dad_stop(ifp, 0);
3863                 return;
3864         }
3865
3866         /*
3867          * Optimistic nodes can start receiving
3868          * Frames right away
3869          */
3870         if (ifp->flags & IFA_F_OPTIMISTIC) {
3871                 ip6_ins_rt(ifp->rt);
3872                 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3873                         /* Because optimistic nodes can use this address,
3874                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3875                          */
3876                         notify = true;
3877                 }
3878         }
3879
3880         addrconf_dad_kick(ifp);
3881 out:
3882         spin_unlock(&ifp->lock);
3883         read_unlock_bh(&idev->lock);
3884         if (notify)
3885                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3886 }
3887
3888 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3889 {
3890         bool begin_dad = false;
3891
3892         spin_lock_bh(&ifp->lock);
3893         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3894                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3895                 begin_dad = true;
3896         }
3897         spin_unlock_bh(&ifp->lock);
3898
3899         if (begin_dad)
3900                 addrconf_mod_dad_work(ifp, 0);
3901 }
3902
3903 static void addrconf_dad_work(struct work_struct *w)
3904 {
3905         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3906                                                 struct inet6_ifaddr,
3907                                                 dad_work);
3908         struct inet6_dev *idev = ifp->idev;
3909         bool bump_id, disable_ipv6 = false;
3910         struct in6_addr mcaddr;
3911
3912         enum {
3913                 DAD_PROCESS,
3914                 DAD_BEGIN,
3915                 DAD_ABORT,
3916         } action = DAD_PROCESS;
3917
3918         rtnl_lock();
3919
3920         spin_lock_bh(&ifp->lock);
3921         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3922                 action = DAD_BEGIN;
3923                 ifp->state = INET6_IFADDR_STATE_DAD;
3924         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3925                 action = DAD_ABORT;
3926                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3927
3928                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3929                      idev->cnf.accept_dad > 1) &&
3930                     !idev->cnf.disable_ipv6 &&
3931                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3932                         struct in6_addr addr;
3933
3934                         addr.s6_addr32[0] = htonl(0xfe800000);
3935                         addr.s6_addr32[1] = 0;
3936
3937                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3938                             ipv6_addr_equal(&ifp->addr, &addr)) {
3939                                 /* DAD failed for link-local based on MAC */
3940                                 idev->cnf.disable_ipv6 = 1;
3941
3942                                 pr_info("%s: IPv6 being disabled!\n",
3943                                         ifp->idev->dev->name);
3944                                 disable_ipv6 = true;
3945                         }
3946                 }
3947         }
3948         spin_unlock_bh(&ifp->lock);
3949
3950         if (action == DAD_BEGIN) {
3951                 addrconf_dad_begin(ifp);
3952                 goto out;
3953         } else if (action == DAD_ABORT) {
3954                 in6_ifa_hold(ifp);
3955                 addrconf_dad_stop(ifp, 1);
3956                 if (disable_ipv6)
3957                         addrconf_ifdown(idev->dev, 0);
3958                 goto out;
3959         }
3960
3961         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3962                 goto out;
3963
3964         write_lock_bh(&idev->lock);
3965         if (idev->dead || !(idev->if_flags & IF_READY)) {
3966                 write_unlock_bh(&idev->lock);
3967                 goto out;
3968         }
3969
3970         spin_lock(&ifp->lock);
3971         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3972                 spin_unlock(&ifp->lock);
3973                 write_unlock_bh(&idev->lock);
3974                 goto out;
3975         }
3976
3977         if (ifp->dad_probes == 0) {
3978                 bool send_na = false;
3979
3980                 /*
3981                  * DAD was successful
3982                  */
3983
3984                 if (ifp->flags & IFA_F_TENTATIVE &&
3985                     !(ifp->flags & IFA_F_OPTIMISTIC))
3986                         send_na = true;
3987                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3988                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3989                 spin_unlock(&ifp->lock);
3990                 write_unlock_bh(&idev->lock);
3991
3992                 addrconf_dad_completed(ifp, bump_id, send_na);
3993
3994                 goto out;
3995         }
3996
3997         ifp->dad_probes--;
3998         addrconf_mod_dad_work(ifp,
3999                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4000         spin_unlock(&ifp->lock);
4001         write_unlock_bh(&idev->lock);
4002
4003         /* send a neighbour solicitation for our addr */
4004         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4005         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4006                       ifp->dad_nonce);
4007 out:
4008         in6_ifa_put(ifp);
4009         rtnl_unlock();
4010 }
4011
4012 /* ifp->idev must be at least read locked */
4013 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4014 {
4015         struct inet6_ifaddr *ifpiter;
4016         struct inet6_dev *idev = ifp->idev;
4017
4018         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4019                 if (ifpiter->scope > IFA_LINK)
4020                         break;
4021                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4022                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4023                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4024                     IFA_F_PERMANENT)
4025                         return false;
4026         }
4027         return true;
4028 }
4029
4030 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4031                                    bool send_na)
4032 {
4033         struct net_device *dev = ifp->idev->dev;
4034         struct in6_addr lladdr;
4035         bool send_rs, send_mld;
4036
4037         addrconf_del_dad_work(ifp);
4038
4039         /*
4040          *      Configure the address for reception. Now it is valid.
4041          */
4042
4043         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4044
4045         /* If added prefix is link local and we are prepared to process
4046            router advertisements, start sending router solicitations.
4047          */
4048
4049         read_lock_bh(&ifp->idev->lock);
4050         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4051         send_rs = send_mld &&
4052                   ipv6_accept_ra(ifp->idev) &&
4053                   ifp->idev->cnf.rtr_solicits != 0 &&
4054                   (dev->flags&IFF_LOOPBACK) == 0;
4055         read_unlock_bh(&ifp->idev->lock);
4056
4057         /* While dad is in progress mld report's source address is in6_addrany.
4058          * Resend with proper ll now.
4059          */
4060         if (send_mld)
4061                 ipv6_mc_dad_complete(ifp->idev);
4062
4063         /* send unsolicited NA if enabled */
4064         if (send_na &&
4065             (ifp->idev->cnf.ndisc_notify ||
4066              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4067                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4068                               /*router=*/ !!ifp->idev->cnf.forwarding,
4069                               /*solicited=*/ false, /*override=*/ true,
4070                               /*inc_opt=*/ true);
4071         }
4072
4073         if (send_rs) {
4074                 /*
4075                  *      If a host as already performed a random delay
4076                  *      [...] as part of DAD [...] there is no need
4077                  *      to delay again before sending the first RS
4078                  */
4079                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4080                         return;
4081                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4082
4083                 write_lock_bh(&ifp->idev->lock);
4084                 spin_lock(&ifp->lock);
4085                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4086                         ifp->idev->cnf.rtr_solicit_interval);
4087                 ifp->idev->rs_probes = 1;
4088                 ifp->idev->if_flags |= IF_RS_SENT;
4089                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4090                 spin_unlock(&ifp->lock);
4091                 write_unlock_bh(&ifp->idev->lock);
4092         }
4093
4094         if (bump_id)
4095                 rt_genid_bump_ipv6(dev_net(dev));
4096
4097         /* Make sure that a new temporary address will be created
4098          * before this temporary address becomes deprecated.
4099          */
4100         if (ifp->flags & IFA_F_TEMPORARY)
4101                 addrconf_verify_rtnl();
4102 }
4103
4104 static void addrconf_dad_run(struct inet6_dev *idev)
4105 {
4106         struct inet6_ifaddr *ifp;
4107
4108         read_lock_bh(&idev->lock);
4109         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4110                 spin_lock(&ifp->lock);
4111                 if (ifp->flags & IFA_F_TENTATIVE &&
4112                     ifp->state == INET6_IFADDR_STATE_DAD)
4113                         addrconf_dad_kick(ifp);
4114                 spin_unlock(&ifp->lock);
4115         }
4116         read_unlock_bh(&idev->lock);
4117 }
4118
4119 #ifdef CONFIG_PROC_FS
4120 struct if6_iter_state {
4121         struct seq_net_private p;
4122         int bucket;
4123         int offset;
4124 };
4125
4126 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4127 {
4128         struct inet6_ifaddr *ifa = NULL;
4129         struct if6_iter_state *state = seq->private;
4130         struct net *net = seq_file_net(seq);
4131         int p = 0;
4132
4133         /* initial bucket if pos is 0 */
4134         if (pos == 0) {
4135                 state->bucket = 0;
4136                 state->offset = 0;
4137         }
4138
4139         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4140                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4141                                          addr_lst) {
4142                         if (!net_eq(dev_net(ifa->idev->dev), net))
4143                                 continue;
4144                         /* sync with offset */
4145                         if (p < state->offset) {
4146                                 p++;
4147                                 continue;
4148                         }
4149                         return ifa;
4150                 }
4151
4152                 /* prepare for next bucket */
4153                 state->offset = 0;
4154                 p = 0;
4155         }
4156         return NULL;
4157 }
4158
4159 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4160                                          struct inet6_ifaddr *ifa)
4161 {
4162         struct if6_iter_state *state = seq->private;
4163         struct net *net = seq_file_net(seq);
4164
4165         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4166                 if (!net_eq(dev_net(ifa->idev->dev), net))
4167                         continue;
4168                 state->offset++;
4169                 return ifa;
4170         }
4171
4172         state->offset = 0;
4173         while (++state->bucket < IN6_ADDR_HSIZE) {
4174                 hlist_for_each_entry_rcu_bh(ifa,
4175                                      &inet6_addr_lst[state->bucket], addr_lst) {
4176                         if (!net_eq(dev_net(ifa->idev->dev), net))
4177                                 continue;
4178                         return ifa;
4179                 }
4180         }
4181
4182         return NULL;
4183 }
4184
4185 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4186         __acquires(rcu_bh)
4187 {
4188         rcu_read_lock_bh();
4189         return if6_get_first(seq, *pos);
4190 }
4191
4192 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4193 {
4194         struct inet6_ifaddr *ifa;
4195
4196         ifa = if6_get_next(seq, v);
4197         ++*pos;
4198         return ifa;
4199 }
4200
4201 static void if6_seq_stop(struct seq_file *seq, void *v)
4202         __releases(rcu_bh)
4203 {
4204         rcu_read_unlock_bh();
4205 }
4206
4207 static int if6_seq_show(struct seq_file *seq, void *v)
4208 {
4209         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4210         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4211                    &ifp->addr,
4212                    ifp->idev->dev->ifindex,
4213                    ifp->prefix_len,
4214                    ifp->scope,
4215                    (u8) ifp->flags,
4216                    ifp->idev->dev->name);
4217         return 0;
4218 }
4219
4220 static const struct seq_operations if6_seq_ops = {
4221         .start  = if6_seq_start,
4222         .next   = if6_seq_next,
4223         .show   = if6_seq_show,
4224         .stop   = if6_seq_stop,
4225 };
4226
4227 static int if6_seq_open(struct inode *inode, struct file *file)
4228 {
4229         return seq_open_net(inode, file, &if6_seq_ops,
4230                             sizeof(struct if6_iter_state));
4231 }
4232
4233 static const struct file_operations if6_fops = {
4234         .owner          = THIS_MODULE,
4235         .open           = if6_seq_open,
4236         .read           = seq_read,
4237         .llseek         = seq_lseek,
4238         .release        = seq_release_net,
4239 };
4240
4241 static int __net_init if6_proc_net_init(struct net *net)
4242 {
4243         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4244                 return -ENOMEM;
4245         return 0;
4246 }
4247
4248 static void __net_exit if6_proc_net_exit(struct net *net)
4249 {
4250         remove_proc_entry("if_inet6", net->proc_net);
4251 }
4252
4253 static struct pernet_operations if6_proc_net_ops = {
4254         .init = if6_proc_net_init,
4255         .exit = if6_proc_net_exit,
4256 };
4257
4258 int __init if6_proc_init(void)
4259 {
4260         return register_pernet_subsys(&if6_proc_net_ops);
4261 }
4262
4263 void if6_proc_exit(void)
4264 {
4265         unregister_pernet_subsys(&if6_proc_net_ops);
4266 }
4267 #endif  /* CONFIG_PROC_FS */
4268
4269 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4270 /* Check if address is a home address configured on any interface. */
4271 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4272 {
4273         int ret = 0;
4274         struct inet6_ifaddr *ifp = NULL;
4275         unsigned int hash = inet6_addr_hash(addr);
4276
4277         rcu_read_lock_bh();
4278         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4279                 if (!net_eq(dev_net(ifp->idev->dev), net))
4280                         continue;
4281                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4282                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4283                         ret = 1;
4284                         break;
4285                 }
4286         }
4287         rcu_read_unlock_bh();
4288         return ret;
4289 }
4290 #endif
4291
4292 /*
4293  *      Periodic address status verification
4294  */
4295
4296 static void addrconf_verify_rtnl(void)
4297 {
4298         unsigned long now, next, next_sec, next_sched;
4299         struct inet6_ifaddr *ifp;
4300         int i;
4301
4302         ASSERT_RTNL();
4303
4304         rcu_read_lock_bh();
4305         now = jiffies;
4306         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4307
4308         cancel_delayed_work(&addr_chk_work);
4309
4310         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4311 restart:
4312                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4313                         unsigned long age;
4314
4315                         /* When setting preferred_lft to a value not zero or
4316                          * infinity, while valid_lft is infinity
4317                          * IFA_F_PERMANENT has a non-infinity life time.
4318                          */
4319                         if ((ifp->flags & IFA_F_PERMANENT) &&
4320                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4321                                 continue;
4322
4323                         spin_lock(&ifp->lock);
4324                         /* We try to batch several events at once. */
4325                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4326
4327                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4328                             age >= ifp->valid_lft) {
4329                                 spin_unlock(&ifp->lock);
4330                                 in6_ifa_hold(ifp);
4331                                 ipv6_del_addr(ifp);
4332                                 goto restart;
4333                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4334                                 spin_unlock(&ifp->lock);
4335                                 continue;
4336                         } else if (age >= ifp->prefered_lft) {
4337                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4338                                 int deprecate = 0;
4339
4340                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4341                                         deprecate = 1;
4342                                         ifp->flags |= IFA_F_DEPRECATED;
4343                                 }
4344
4345                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4346                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4347                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4348
4349                                 spin_unlock(&ifp->lock);
4350
4351                                 if (deprecate) {
4352                                         in6_ifa_hold(ifp);
4353
4354                                         ipv6_ifa_notify(0, ifp);
4355                                         in6_ifa_put(ifp);
4356                                         goto restart;
4357                                 }
4358                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4359                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4360                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4361                                         ifp->idev->cnf.dad_transmits *
4362                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4363
4364                                 if (age >= ifp->prefered_lft - regen_advance) {
4365                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4366                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4367                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4368                                         if (!ifp->regen_count && ifpub) {
4369                                                 ifp->regen_count++;
4370                                                 in6_ifa_hold(ifp);
4371                                                 in6_ifa_hold(ifpub);
4372                                                 spin_unlock(&ifp->lock);
4373
4374                                                 spin_lock(&ifpub->lock);
4375                                                 ifpub->regen_count = 0;
4376                                                 spin_unlock(&ifpub->lock);
4377                                                 ipv6_create_tempaddr(ifpub, ifp);
4378                                                 in6_ifa_put(ifpub);
4379                                                 in6_ifa_put(ifp);
4380                                                 goto restart;
4381                                         }
4382                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4383                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4384                                 spin_unlock(&ifp->lock);
4385                         } else {
4386                                 /* ifp->prefered_lft <= ifp->valid_lft */
4387                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4388                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4389                                 spin_unlock(&ifp->lock);
4390                         }
4391                 }
4392         }
4393
4394         next_sec = round_jiffies_up(next);
4395         next_sched = next;
4396
4397         /* If rounded timeout is accurate enough, accept it. */
4398         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4399                 next_sched = next_sec;
4400
4401         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4402         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4403                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4404
4405         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4406               now, next, next_sec, next_sched);
4407         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4408         rcu_read_unlock_bh();
4409 }
4410
4411 static void addrconf_verify_work(struct work_struct *w)
4412 {
4413         rtnl_lock();
4414         addrconf_verify_rtnl();
4415         rtnl_unlock();
4416 }
4417
4418 static void addrconf_verify(void)
4419 {
4420         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4421 }
4422
4423 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4424                                      struct in6_addr **peer_pfx)
4425 {
4426         struct in6_addr *pfx = NULL;
4427
4428         *peer_pfx = NULL;
4429
4430         if (addr)
4431                 pfx = nla_data(addr);
4432
4433         if (local) {
4434                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4435                         *peer_pfx = pfx;
4436                 pfx = nla_data(local);
4437         }
4438
4439         return pfx;
4440 }
4441
4442 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4443         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4444         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4445         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4446         [IFA_FLAGS]             = { .len = sizeof(u32) },
4447 };
4448
4449 static int
4450 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4451                   struct netlink_ext_ack *extack)
4452 {
4453         struct net *net = sock_net(skb->sk);
4454         struct ifaddrmsg *ifm;
4455         struct nlattr *tb[IFA_MAX+1];
4456         struct in6_addr *pfx, *peer_pfx;
4457         u32 ifa_flags;
4458         int err;
4459
4460         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4461                           extack);
4462         if (err < 0)
4463                 return err;
4464
4465         ifm = nlmsg_data(nlh);
4466         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4467         if (!pfx)
4468                 return -EINVAL;
4469
4470         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4471
4472         /* We ignore other flags so far. */
4473         ifa_flags &= IFA_F_MANAGETEMPADDR;
4474
4475         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4476                               ifm->ifa_prefixlen);
4477 }
4478
4479 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4480                              u32 prefered_lft, u32 valid_lft)
4481 {
4482         u32 flags;
4483         clock_t expires;
4484         unsigned long timeout;
4485         bool was_managetempaddr;
4486         bool had_prefixroute;
4487
4488         ASSERT_RTNL();
4489
4490         if (!valid_lft || (prefered_lft > valid_lft))
4491                 return -EINVAL;
4492
4493         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4494             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4495                 return -EINVAL;
4496
4497         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4498         if (addrconf_finite_timeout(timeout)) {
4499                 expires = jiffies_to_clock_t(timeout * HZ);
4500                 valid_lft = timeout;
4501                 flags = RTF_EXPIRES;
4502         } else {
4503                 expires = 0;
4504                 flags = 0;
4505                 ifa_flags |= IFA_F_PERMANENT;
4506         }
4507
4508         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4509         if (addrconf_finite_timeout(timeout)) {
4510                 if (timeout == 0)
4511                         ifa_flags |= IFA_F_DEPRECATED;
4512                 prefered_lft = timeout;
4513         }
4514
4515         spin_lock_bh(&ifp->lock);
4516         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4517         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4518                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4519         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4520                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4521                         IFA_F_NOPREFIXROUTE);
4522         ifp->flags |= ifa_flags;
4523         ifp->tstamp = jiffies;
4524         ifp->valid_lft = valid_lft;
4525         ifp->prefered_lft = prefered_lft;
4526
4527         spin_unlock_bh(&ifp->lock);
4528         if (!(ifp->flags&IFA_F_TENTATIVE))
4529                 ipv6_ifa_notify(0, ifp);
4530
4531         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4532                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4533                                       expires, flags);
4534         } else if (had_prefixroute) {
4535                 enum cleanup_prefix_rt_t action;
4536                 unsigned long rt_expires;
4537
4538                 write_lock_bh(&ifp->idev->lock);
4539                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4540                 write_unlock_bh(&ifp->idev->lock);
4541
4542                 if (action != CLEANUP_PREFIX_RT_NOP) {
4543                         cleanup_prefix_route(ifp, rt_expires,
4544                                 action == CLEANUP_PREFIX_RT_DEL);
4545                 }
4546         }
4547
4548         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4549                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4550                         valid_lft = prefered_lft = 0;
4551                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4552                                  !was_managetempaddr, jiffies);
4553         }
4554
4555         addrconf_verify_rtnl();
4556
4557         return 0;
4558 }
4559
4560 static int
4561 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4562                   struct netlink_ext_ack *extack)
4563 {
4564         struct net *net = sock_net(skb->sk);
4565         struct ifaddrmsg *ifm;
4566         struct nlattr *tb[IFA_MAX+1];
4567         struct in6_addr *pfx, *peer_pfx;
4568         struct inet6_ifaddr *ifa;
4569         struct net_device *dev;
4570         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4571         u32 ifa_flags;
4572         int err;
4573
4574         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4575                           extack);
4576         if (err < 0)
4577                 return err;
4578
4579         ifm = nlmsg_data(nlh);
4580         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4581         if (!pfx)
4582                 return -EINVAL;
4583
4584         if (tb[IFA_CACHEINFO]) {
4585                 struct ifa_cacheinfo *ci;
4586
4587                 ci = nla_data(tb[IFA_CACHEINFO]);
4588                 valid_lft = ci->ifa_valid;
4589                 preferred_lft = ci->ifa_prefered;
4590         } else {
4591                 preferred_lft = INFINITY_LIFE_TIME;
4592                 valid_lft = INFINITY_LIFE_TIME;
4593         }
4594
4595         dev =  __dev_get_by_index(net, ifm->ifa_index);
4596         if (!dev)
4597                 return -ENODEV;
4598
4599         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4600
4601         /* We ignore other flags so far. */
4602         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4603                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4604
4605         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4606         if (!ifa) {
4607                 /*
4608                  * It would be best to check for !NLM_F_CREATE here but
4609                  * userspace already relies on not having to provide this.
4610                  */
4611                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4612                                       ifm->ifa_prefixlen, ifa_flags,
4613                                       preferred_lft, valid_lft);
4614         }
4615
4616         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4617             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4618                 err = -EEXIST;
4619         else
4620                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4621
4622         in6_ifa_put(ifa);
4623
4624         return err;
4625 }
4626
4627 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4628                           u8 scope, int ifindex)
4629 {
4630         struct ifaddrmsg *ifm;
4631
4632         ifm = nlmsg_data(nlh);
4633         ifm->ifa_family = AF_INET6;
4634         ifm->ifa_prefixlen = prefixlen;
4635         ifm->ifa_flags = flags;
4636         ifm->ifa_scope = scope;
4637         ifm->ifa_index = ifindex;
4638 }
4639
4640 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4641                          unsigned long tstamp, u32 preferred, u32 valid)
4642 {
4643         struct ifa_cacheinfo ci;
4644
4645         ci.cstamp = cstamp_delta(cstamp);
4646         ci.tstamp = cstamp_delta(tstamp);
4647         ci.ifa_prefered = preferred;
4648         ci.ifa_valid = valid;
4649
4650         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4651 }
4652
4653 static inline int rt_scope(int ifa_scope)
4654 {
4655         if (ifa_scope & IFA_HOST)
4656                 return RT_SCOPE_HOST;
4657         else if (ifa_scope & IFA_LINK)
4658                 return RT_SCOPE_LINK;
4659         else if (ifa_scope & IFA_SITE)
4660                 return RT_SCOPE_SITE;
4661         else
4662                 return RT_SCOPE_UNIVERSE;
4663 }
4664
4665 static inline int inet6_ifaddr_msgsize(void)
4666 {
4667         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4668                + nla_total_size(16) /* IFA_LOCAL */
4669                + nla_total_size(16) /* IFA_ADDRESS */
4670                + nla_total_size(sizeof(struct ifa_cacheinfo))
4671                + nla_total_size(4)  /* IFA_FLAGS */;
4672 }
4673
4674 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4675                              u32 portid, u32 seq, int event, unsigned int flags)
4676 {
4677         struct nlmsghdr  *nlh;
4678         u32 preferred, valid;
4679
4680         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4681         if (!nlh)
4682                 return -EMSGSIZE;
4683
4684         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4685                       ifa->idev->dev->ifindex);
4686
4687         if (!((ifa->flags&IFA_F_PERMANENT) &&
4688               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4689                 preferred = ifa->prefered_lft;
4690                 valid = ifa->valid_lft;
4691                 if (preferred != INFINITY_LIFE_TIME) {
4692                         long tval = (jiffies - ifa->tstamp)/HZ;
4693                         if (preferred > tval)
4694                                 preferred -= tval;
4695                         else
4696                                 preferred = 0;
4697                         if (valid != INFINITY_LIFE_TIME) {
4698                                 if (valid > tval)
4699                                         valid -= tval;
4700                                 else
4701                                         valid = 0;
4702                         }
4703                 }
4704         } else {
4705                 preferred = INFINITY_LIFE_TIME;
4706                 valid = INFINITY_LIFE_TIME;
4707         }
4708
4709         if (!ipv6_addr_any(&ifa->peer_addr)) {
4710                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4711                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4712                         goto error;
4713         } else
4714                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4715                         goto error;
4716
4717         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4718                 goto error;
4719
4720         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4721                 goto error;
4722
4723         nlmsg_end(skb, nlh);
4724         return 0;
4725
4726 error:
4727         nlmsg_cancel(skb, nlh);
4728         return -EMSGSIZE;
4729 }
4730
4731 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4732                                 u32 portid, u32 seq, int event, u16 flags)
4733 {
4734         struct nlmsghdr  *nlh;
4735         u8 scope = RT_SCOPE_UNIVERSE;
4736         int ifindex = ifmca->idev->dev->ifindex;
4737
4738         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4739                 scope = RT_SCOPE_SITE;
4740
4741         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4742         if (!nlh)
4743                 return -EMSGSIZE;
4744
4745         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4746         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4747             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4748                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4749                 nlmsg_cancel(skb, nlh);
4750                 return -EMSGSIZE;
4751         }
4752
4753         nlmsg_end(skb, nlh);
4754         return 0;
4755 }
4756
4757 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4758                                 u32 portid, u32 seq, int event, unsigned int flags)
4759 {
4760         struct nlmsghdr  *nlh;
4761         u8 scope = RT_SCOPE_UNIVERSE;
4762         int ifindex = ifaca->aca_idev->dev->ifindex;
4763
4764         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4765                 scope = RT_SCOPE_SITE;
4766
4767         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4768         if (!nlh)
4769                 return -EMSGSIZE;
4770
4771         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4772         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4773             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4774                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4775                 nlmsg_cancel(skb, nlh);
4776                 return -EMSGSIZE;
4777         }
4778
4779         nlmsg_end(skb, nlh);
4780         return 0;
4781 }
4782
4783 enum addr_type_t {
4784         UNICAST_ADDR,
4785         MULTICAST_ADDR,
4786         ANYCAST_ADDR,
4787 };
4788
4789 /* called with rcu_read_lock() */
4790 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4791                           struct netlink_callback *cb, enum addr_type_t type,
4792                           int s_ip_idx, int *p_ip_idx)
4793 {
4794         struct ifmcaddr6 *ifmca;
4795         struct ifacaddr6 *ifaca;
4796         int err = 1;
4797         int ip_idx = *p_ip_idx;
4798
4799         read_lock_bh(&idev->lock);
4800         switch (type) {
4801         case UNICAST_ADDR: {
4802                 struct inet6_ifaddr *ifa;
4803
4804                 /* unicast address incl. temp addr */
4805                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4806                         if (ip_idx < s_ip_idx)
4807                                 goto next;
4808                         err = inet6_fill_ifaddr(skb, ifa,
4809                                                 NETLINK_CB(cb->skb).portid,
4810                                                 cb->nlh->nlmsg_seq,
4811                                                 RTM_NEWADDR,
4812                                                 NLM_F_MULTI);
4813                         if (err < 0)
4814                                 break;
4815                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4816 next:
4817                         ip_idx++;
4818                 }
4819                 break;
4820         }
4821         case MULTICAST_ADDR:
4822                 /* multicast address */
4823                 for (ifmca = idev->mc_list; ifmca;
4824                      ifmca = ifmca->next, ip_idx++) {
4825                         if (ip_idx < s_ip_idx)
4826                                 continue;
4827                         err = inet6_fill_ifmcaddr(skb, ifmca,
4828                                                   NETLINK_CB(cb->skb).portid,
4829                                                   cb->nlh->nlmsg_seq,
4830                                                   RTM_GETMULTICAST,
4831                                                   NLM_F_MULTI);
4832                         if (err < 0)
4833                                 break;
4834                 }
4835                 break;
4836         case ANYCAST_ADDR:
4837                 /* anycast address */
4838                 for (ifaca = idev->ac_list; ifaca;
4839                      ifaca = ifaca->aca_next, ip_idx++) {
4840                         if (ip_idx < s_ip_idx)
4841                                 continue;
4842                         err = inet6_fill_ifacaddr(skb, ifaca,
4843                                                   NETLINK_CB(cb->skb).portid,
4844                                                   cb->nlh->nlmsg_seq,
4845                                                   RTM_GETANYCAST,
4846                                                   NLM_F_MULTI);
4847                         if (err < 0)
4848                                 break;
4849                 }
4850                 break;
4851         default:
4852                 break;
4853         }
4854         read_unlock_bh(&idev->lock);
4855         *p_ip_idx = ip_idx;
4856         return err;
4857 }
4858
4859 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4860                            enum addr_type_t type)
4861 {
4862         struct net *net = sock_net(skb->sk);
4863         int h, s_h;
4864         int idx, ip_idx;
4865         int s_idx, s_ip_idx;
4866         struct net_device *dev;
4867         struct inet6_dev *idev;
4868         struct hlist_head *head;
4869
4870         s_h = cb->args[0];
4871         s_idx = idx = cb->args[1];
4872         s_ip_idx = ip_idx = cb->args[2];
4873
4874         rcu_read_lock();
4875         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4876         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4877                 idx = 0;
4878                 head = &net->dev_index_head[h];
4879                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4880                         if (idx < s_idx)
4881                                 goto cont;
4882                         if (h > s_h || idx > s_idx)
4883                                 s_ip_idx = 0;
4884                         ip_idx = 0;
4885                         idev = __in6_dev_get(dev);
4886                         if (!idev)
4887                                 goto cont;
4888
4889                         if (in6_dump_addrs(idev, skb, cb, type,
4890                                            s_ip_idx, &ip_idx) < 0)
4891                                 goto done;
4892 cont:
4893                         idx++;
4894                 }
4895         }
4896 done:
4897         rcu_read_unlock();
4898         cb->args[0] = h;
4899         cb->args[1] = idx;
4900         cb->args[2] = ip_idx;
4901
4902         return skb->len;
4903 }
4904
4905 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4906 {
4907         enum addr_type_t type = UNICAST_ADDR;
4908
4909         return inet6_dump_addr(skb, cb, type);
4910 }
4911
4912 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4913 {
4914         enum addr_type_t type = MULTICAST_ADDR;
4915
4916         return inet6_dump_addr(skb, cb, type);
4917 }
4918
4919
4920 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4921 {
4922         enum addr_type_t type = ANYCAST_ADDR;
4923
4924         return inet6_dump_addr(skb, cb, type);
4925 }
4926
4927 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4928                              struct netlink_ext_ack *extack)
4929 {
4930         struct net *net = sock_net(in_skb->sk);
4931         struct ifaddrmsg *ifm;
4932         struct nlattr *tb[IFA_MAX+1];
4933         struct in6_addr *addr = NULL, *peer;
4934         struct net_device *dev = NULL;
4935         struct inet6_ifaddr *ifa;
4936         struct sk_buff *skb;
4937         int err;
4938
4939         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4940                           extack);
4941         if (err < 0)
4942                 goto errout;
4943
4944         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4945         if (!addr) {
4946                 err = -EINVAL;
4947                 goto errout;
4948         }
4949
4950         ifm = nlmsg_data(nlh);
4951         if (ifm->ifa_index)
4952                 dev = __dev_get_by_index(net, ifm->ifa_index);
4953
4954         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4955         if (!ifa) {
4956                 err = -EADDRNOTAVAIL;
4957                 goto errout;
4958         }
4959
4960         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4961         if (!skb) {
4962                 err = -ENOBUFS;
4963                 goto errout_ifa;
4964         }
4965
4966         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4967                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4968         if (err < 0) {
4969                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4970                 WARN_ON(err == -EMSGSIZE);
4971                 kfree_skb(skb);
4972                 goto errout_ifa;
4973         }
4974         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4975 errout_ifa:
4976         in6_ifa_put(ifa);
4977 errout:
4978         return err;
4979 }
4980
4981 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4982 {
4983         struct sk_buff *skb;
4984         struct net *net = dev_net(ifa->idev->dev);
4985         int err = -ENOBUFS;
4986
4987         /* Don't send DELADDR notification for TENTATIVE address,
4988          * since NEWADDR notification is sent only after removing
4989          * TENTATIVE flag, if DAD has not failed.
4990          */
4991         if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
4992             event == RTM_DELADDR)
4993                 return;
4994
4995         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4996         if (!skb)
4997                 goto errout;
4998
4999         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5000         if (err < 0) {
5001                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5002                 WARN_ON(err == -EMSGSIZE);
5003                 kfree_skb(skb);
5004                 goto errout;
5005         }
5006         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5007         return;
5008 errout:
5009         if (err < 0)
5010                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5011 }
5012
5013 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5014                                 __s32 *array, int bytes)
5015 {
5016         BUG_ON(bytes < (DEVCONF_MAX * 4));
5017
5018         memset(array, 0, bytes);
5019         array[DEVCONF_FORWARDING] = cnf->forwarding;
5020         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5021         array[DEVCONF_MTU6] = cnf->mtu6;
5022         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5023         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5024         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5025         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5026         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5027         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5028                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5029         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5030                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5031         array[DEVCONF_RTR_SOLICIT_DELAY] =
5032                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5033         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5034         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5035                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5036         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5037                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5038         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5039         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5040         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5041         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5042         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5043         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5044         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5045         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5046         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5047 #ifdef CONFIG_IPV6_ROUTER_PREF
5048         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5049         array[DEVCONF_RTR_PROBE_INTERVAL] =
5050                 jiffies_to_msecs(cnf->rtr_probe_interval);
5051 #ifdef CONFIG_IPV6_ROUTE_INFO
5052         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5053         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5054 #endif
5055 #endif
5056         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5057         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5058 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5059         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5060         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5061 #endif
5062 #ifdef CONFIG_IPV6_MROUTE
5063         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5064 #endif
5065         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5066         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5067         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5068         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5069         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5070         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5071         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5072         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5073         /* we omit DEVCONF_STABLE_SECRET for now */
5074         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5075         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5076         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5077         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5078         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5079 #ifdef CONFIG_IPV6_SEG6_HMAC
5080         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5081 #endif
5082         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5083         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5084         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5085 }
5086
5087 static inline size_t inet6_ifla6_size(void)
5088 {
5089         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5090              + nla_total_size(sizeof(struct ifla_cacheinfo))
5091              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5092              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5093              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5094              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5095 }
5096
5097 static inline size_t inet6_if_nlmsg_size(void)
5098 {
5099         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5100                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5101                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5102                + nla_total_size(4) /* IFLA_MTU */
5103                + nla_total_size(4) /* IFLA_LINK */
5104                + nla_total_size(1) /* IFLA_OPERSTATE */
5105                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5106 }
5107
5108 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5109                                         int bytes)
5110 {
5111         int i;
5112         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5113         BUG_ON(pad < 0);
5114
5115         /* Use put_unaligned() because stats may not be aligned for u64. */
5116         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5117         for (i = 1; i < ICMP6_MIB_MAX; i++)
5118                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5119
5120         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5121 }
5122
5123 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5124                                         int bytes, size_t syncpoff)
5125 {
5126         int i, c;
5127         u64 buff[IPSTATS_MIB_MAX];
5128         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5129
5130         BUG_ON(pad < 0);
5131
5132         memset(buff, 0, sizeof(buff));
5133         buff[0] = IPSTATS_MIB_MAX;
5134
5135         for_each_possible_cpu(c) {
5136                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5137                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5138         }
5139
5140         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5141         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5142 }
5143
5144 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5145                              int bytes)
5146 {
5147         switch (attrtype) {
5148         case IFLA_INET6_STATS:
5149                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5150                                      offsetof(struct ipstats_mib, syncp));
5151                 break;
5152         case IFLA_INET6_ICMP6STATS:
5153                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5154                 break;
5155         }
5156 }
5157
5158 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5159                                   u32 ext_filter_mask)
5160 {
5161         struct nlattr *nla;
5162         struct ifla_cacheinfo ci;
5163
5164         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5165                 goto nla_put_failure;
5166         ci.max_reasm_len = IPV6_MAXPLEN;
5167         ci.tstamp = cstamp_delta(idev->tstamp);
5168         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5169         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5170         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5171                 goto nla_put_failure;
5172         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5173         if (!nla)
5174                 goto nla_put_failure;
5175         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5176
5177         /* XXX - MC not implemented */
5178
5179         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5180                 return 0;
5181
5182         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5183         if (!nla)
5184                 goto nla_put_failure;
5185         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5186
5187         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5188         if (!nla)
5189                 goto nla_put_failure;
5190         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5191
5192         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5193         if (!nla)
5194                 goto nla_put_failure;
5195
5196         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5197                 goto nla_put_failure;
5198
5199         read_lock_bh(&idev->lock);
5200         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5201         read_unlock_bh(&idev->lock);
5202
5203         return 0;
5204
5205 nla_put_failure:
5206         return -EMSGSIZE;
5207 }
5208
5209 static size_t inet6_get_link_af_size(const struct net_device *dev,
5210                                      u32 ext_filter_mask)
5211 {
5212         if (!__in6_dev_get(dev))
5213                 return 0;
5214
5215         return inet6_ifla6_size();
5216 }
5217
5218 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5219                               u32 ext_filter_mask)
5220 {
5221         struct inet6_dev *idev = __in6_dev_get(dev);
5222
5223         if (!idev)
5224                 return -ENODATA;
5225
5226         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5227                 return -EMSGSIZE;
5228
5229         return 0;
5230 }
5231
5232 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5233 {
5234         struct inet6_ifaddr *ifp;
5235         struct net_device *dev = idev->dev;
5236         bool clear_token, update_rs = false;
5237         struct in6_addr ll_addr;
5238
5239         ASSERT_RTNL();
5240
5241         if (!token)
5242                 return -EINVAL;
5243         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5244                 return -EINVAL;
5245         if (!ipv6_accept_ra(idev))
5246                 return -EINVAL;
5247         if (idev->cnf.rtr_solicits == 0)
5248                 return -EINVAL;
5249
5250         write_lock_bh(&idev->lock);
5251
5252         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5253         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5254
5255         write_unlock_bh(&idev->lock);
5256
5257         clear_token = ipv6_addr_any(token);
5258         if (clear_token)
5259                 goto update_lft;
5260
5261         if (!idev->dead && (idev->if_flags & IF_READY) &&
5262             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5263                              IFA_F_OPTIMISTIC)) {
5264                 /* If we're not ready, then normal ifup will take care
5265                  * of this. Otherwise, we need to request our rs here.
5266                  */
5267                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5268                 update_rs = true;
5269         }
5270
5271 update_lft:
5272         write_lock_bh(&idev->lock);
5273
5274         if (update_rs) {
5275                 idev->if_flags |= IF_RS_SENT;
5276                 idev->rs_interval = rfc3315_s14_backoff_init(
5277                         idev->cnf.rtr_solicit_interval);
5278                 idev->rs_probes = 1;
5279                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5280         }
5281
5282         /* Well, that's kinda nasty ... */
5283         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5284                 spin_lock(&ifp->lock);
5285                 if (ifp->tokenized) {
5286                         ifp->valid_lft = 0;
5287                         ifp->prefered_lft = 0;
5288                 }
5289                 spin_unlock(&ifp->lock);
5290         }
5291
5292         write_unlock_bh(&idev->lock);
5293         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5294         addrconf_verify_rtnl();
5295         return 0;
5296 }
5297
5298 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5299         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5300         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5301 };
5302
5303 static int inet6_validate_link_af(const struct net_device *dev,
5304                                   const struct nlattr *nla)
5305 {
5306         struct nlattr *tb[IFLA_INET6_MAX + 1];
5307
5308         if (dev && !__in6_dev_get(dev))
5309                 return -EAFNOSUPPORT;
5310
5311         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5312                                 NULL);
5313 }
5314
5315 static int check_addr_gen_mode(int mode)
5316 {
5317         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5318             mode != IN6_ADDR_GEN_MODE_NONE &&
5319             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5320             mode != IN6_ADDR_GEN_MODE_RANDOM)
5321                 return -EINVAL;
5322         return 1;
5323 }
5324
5325 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5326                                 int mode)
5327 {
5328         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5329             !idev->cnf.stable_secret.initialized &&
5330             !net->ipv6.devconf_dflt->stable_secret.initialized)
5331                 return -EINVAL;
5332         return 1;
5333 }
5334
5335 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5336 {
5337         int err = -EINVAL;
5338         struct inet6_dev *idev = __in6_dev_get(dev);
5339         struct nlattr *tb[IFLA_INET6_MAX + 1];
5340
5341         if (!idev)
5342                 return -EAFNOSUPPORT;
5343
5344         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5345                 BUG();
5346
5347         if (tb[IFLA_INET6_TOKEN]) {
5348                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5349                 if (err)
5350                         return err;
5351         }
5352
5353         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5354                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5355
5356                 if (check_addr_gen_mode(mode) < 0 ||
5357                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5358                         return -EINVAL;
5359
5360                 idev->cnf.addr_gen_mode = mode;
5361                 err = 0;
5362         }
5363
5364         return err;
5365 }
5366
5367 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5368                              u32 portid, u32 seq, int event, unsigned int flags)
5369 {
5370         struct net_device *dev = idev->dev;
5371         struct ifinfomsg *hdr;
5372         struct nlmsghdr *nlh;
5373         void *protoinfo;
5374
5375         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5376         if (!nlh)
5377                 return -EMSGSIZE;
5378
5379         hdr = nlmsg_data(nlh);
5380         hdr->ifi_family = AF_INET6;
5381         hdr->__ifi_pad = 0;
5382         hdr->ifi_type = dev->type;
5383         hdr->ifi_index = dev->ifindex;
5384         hdr->ifi_flags = dev_get_flags(dev);
5385         hdr->ifi_change = 0;
5386
5387         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5388             (dev->addr_len &&
5389              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5390             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5391             (dev->ifindex != dev_get_iflink(dev) &&
5392              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5393             nla_put_u8(skb, IFLA_OPERSTATE,
5394                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5395                 goto nla_put_failure;
5396         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5397         if (!protoinfo)
5398                 goto nla_put_failure;
5399
5400         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5401                 goto nla_put_failure;
5402
5403         nla_nest_end(skb, protoinfo);
5404         nlmsg_end(skb, nlh);
5405         return 0;
5406
5407 nla_put_failure:
5408         nlmsg_cancel(skb, nlh);
5409         return -EMSGSIZE;
5410 }
5411
5412 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5413 {
5414         struct net *net = sock_net(skb->sk);
5415         int h, s_h;
5416         int idx = 0, s_idx;
5417         struct net_device *dev;
5418         struct inet6_dev *idev;
5419         struct hlist_head *head;
5420
5421         s_h = cb->args[0];
5422         s_idx = cb->args[1];
5423
5424         rcu_read_lock();
5425         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5426                 idx = 0;
5427                 head = &net->dev_index_head[h];
5428                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5429                         if (idx < s_idx)
5430                                 goto cont;
5431                         idev = __in6_dev_get(dev);
5432                         if (!idev)
5433                                 goto cont;
5434                         if (inet6_fill_ifinfo(skb, idev,
5435                                               NETLINK_CB(cb->skb).portid,
5436                                               cb->nlh->nlmsg_seq,
5437                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5438                                 goto out;
5439 cont:
5440                         idx++;
5441                 }
5442         }
5443 out:
5444         rcu_read_unlock();
5445         cb->args[1] = idx;
5446         cb->args[0] = h;
5447
5448         return skb->len;
5449 }
5450
5451 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5452 {
5453         struct sk_buff *skb;
5454         struct net *net = dev_net(idev->dev);
5455         int err = -ENOBUFS;
5456
5457         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5458         if (!skb)
5459                 goto errout;
5460
5461         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5462         if (err < 0) {
5463                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5464                 WARN_ON(err == -EMSGSIZE);
5465                 kfree_skb(skb);
5466                 goto errout;
5467         }
5468         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5469         return;
5470 errout:
5471         if (err < 0)
5472                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5473 }
5474
5475 static inline size_t inet6_prefix_nlmsg_size(void)
5476 {
5477         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5478                + nla_total_size(sizeof(struct in6_addr))
5479                + nla_total_size(sizeof(struct prefix_cacheinfo));
5480 }
5481
5482 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5483                              struct prefix_info *pinfo, u32 portid, u32 seq,
5484                              int event, unsigned int flags)
5485 {
5486         struct prefixmsg *pmsg;
5487         struct nlmsghdr *nlh;
5488         struct prefix_cacheinfo ci;
5489
5490         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5491         if (!nlh)
5492                 return -EMSGSIZE;
5493
5494         pmsg = nlmsg_data(nlh);
5495         pmsg->prefix_family = AF_INET6;
5496         pmsg->prefix_pad1 = 0;
5497         pmsg->prefix_pad2 = 0;
5498         pmsg->prefix_ifindex = idev->dev->ifindex;
5499         pmsg->prefix_len = pinfo->prefix_len;
5500         pmsg->prefix_type = pinfo->type;
5501         pmsg->prefix_pad3 = 0;
5502         pmsg->prefix_flags = 0;
5503         if (pinfo->onlink)
5504                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5505         if (pinfo->autoconf)
5506                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5507
5508         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5509                 goto nla_put_failure;
5510         ci.preferred_time = ntohl(pinfo->prefered);
5511         ci.valid_time = ntohl(pinfo->valid);
5512         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5513                 goto nla_put_failure;
5514         nlmsg_end(skb, nlh);
5515         return 0;
5516
5517 nla_put_failure:
5518         nlmsg_cancel(skb, nlh);
5519         return -EMSGSIZE;
5520 }
5521
5522 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5523                          struct prefix_info *pinfo)
5524 {
5525         struct sk_buff *skb;
5526         struct net *net = dev_net(idev->dev);
5527         int err = -ENOBUFS;
5528
5529         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5530         if (!skb)
5531                 goto errout;
5532
5533         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5534         if (err < 0) {
5535                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5536                 WARN_ON(err == -EMSGSIZE);
5537                 kfree_skb(skb);
5538                 goto errout;
5539         }
5540         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5541         return;
5542 errout:
5543         if (err < 0)
5544                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5545 }
5546
5547 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5548 {
5549         struct net *net = dev_net(ifp->idev->dev);
5550
5551         if (event)
5552                 ASSERT_RTNL();
5553
5554         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5555
5556         switch (event) {
5557         case RTM_NEWADDR:
5558                 /*
5559                  * If the address was optimistic we inserted the route at the
5560                  * start of our DAD process, so we don't need to do it again.
5561                  * If the device was taken down in the middle of the DAD
5562                  * cycle there is a race where we could get here without a
5563                  * host route, so nothing to insert. That will be fixed when
5564                  * the device is brought up.
5565                  */
5566                 if (ifp->rt && !rcu_access_pointer(ifp->rt->rt6i_node)) {
5567                         ip6_ins_rt(ifp->rt);
5568                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
5569                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
5570                                 &ifp->addr, ifp->idev->dev->name);
5571                 }
5572
5573                 if (ifp->idev->cnf.forwarding)
5574                         addrconf_join_anycast(ifp);
5575                 if (!ipv6_addr_any(&ifp->peer_addr))
5576                         addrconf_prefix_route(&ifp->peer_addr, 128,
5577                                               ifp->idev->dev, 0, 0);
5578                 break;
5579         case RTM_DELADDR:
5580                 if (ifp->idev->cnf.forwarding)
5581                         addrconf_leave_anycast(ifp);
5582                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5583                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5584                         struct rt6_info *rt;
5585
5586                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5587                                                        ifp->idev->dev, 0, 0);
5588                         if (rt)
5589                                 ip6_del_rt(rt);
5590                 }
5591                 if (ifp->rt) {
5592                         if (dst_hold_safe(&ifp->rt->dst))
5593                                 ip6_del_rt(ifp->rt);
5594                 }
5595                 rt_genid_bump_ipv6(net);
5596                 break;
5597         }
5598         atomic_inc(&net->ipv6.dev_addr_genid);
5599 }
5600
5601 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5602 {
5603         rcu_read_lock_bh();
5604         if (likely(ifp->idev->dead == 0))
5605                 __ipv6_ifa_notify(event, ifp);
5606         rcu_read_unlock_bh();
5607 }
5608
5609 #ifdef CONFIG_SYSCTL
5610
5611 static
5612 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5613                            void __user *buffer, size_t *lenp, loff_t *ppos)
5614 {
5615         int *valp = ctl->data;
5616         int val = *valp;
5617         loff_t pos = *ppos;
5618         struct ctl_table lctl;
5619         int ret;
5620
5621         /*
5622          * ctl->data points to idev->cnf.forwarding, we should
5623          * not modify it until we get the rtnl lock.
5624          */
5625         lctl = *ctl;
5626         lctl.data = &val;
5627
5628         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5629
5630         if (write)
5631                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5632         if (ret)
5633                 *ppos = pos;
5634         return ret;
5635 }
5636
5637 static
5638 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5639                         void __user *buffer, size_t *lenp, loff_t *ppos)
5640 {
5641         struct inet6_dev *idev = ctl->extra1;
5642         int min_mtu = IPV6_MIN_MTU;
5643         struct ctl_table lctl;
5644
5645         lctl = *ctl;
5646         lctl.extra1 = &min_mtu;
5647         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5648
5649         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5650 }
5651
5652 static void dev_disable_change(struct inet6_dev *idev)
5653 {
5654         struct netdev_notifier_info info;
5655
5656         if (!idev || !idev->dev)
5657                 return;
5658
5659         netdev_notifier_info_init(&info, idev->dev);
5660         if (idev->cnf.disable_ipv6)
5661                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5662         else
5663                 addrconf_notify(NULL, NETDEV_UP, &info);
5664 }
5665
5666 static void addrconf_disable_change(struct net *net, __s32 newf)
5667 {
5668         struct net_device *dev;
5669         struct inet6_dev *idev;
5670
5671         for_each_netdev(net, dev) {
5672                 idev = __in6_dev_get(dev);
5673                 if (idev) {
5674                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5675                         idev->cnf.disable_ipv6 = newf;
5676                         if (changed)
5677                                 dev_disable_change(idev);
5678                 }
5679         }
5680 }
5681
5682 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5683 {
5684         struct net *net;
5685         int old;
5686
5687         if (!rtnl_trylock())
5688                 return restart_syscall();
5689
5690         net = (struct net *)table->extra2;
5691         old = *p;
5692         *p = newf;
5693
5694         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5695                 rtnl_unlock();
5696                 return 0;
5697         }
5698
5699         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5700                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5701                 addrconf_disable_change(net, newf);
5702         } else if ((!newf) ^ (!old))
5703                 dev_disable_change((struct inet6_dev *)table->extra1);
5704
5705         rtnl_unlock();
5706         return 0;
5707 }
5708
5709 static
5710 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5711                             void __user *buffer, size_t *lenp, loff_t *ppos)
5712 {
5713         int *valp = ctl->data;
5714         int val = *valp;
5715         loff_t pos = *ppos;
5716         struct ctl_table lctl;
5717         int ret;
5718
5719         /*
5720          * ctl->data points to idev->cnf.disable_ipv6, we should
5721          * not modify it until we get the rtnl lock.
5722          */
5723         lctl = *ctl;
5724         lctl.data = &val;
5725
5726         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5727
5728         if (write)
5729                 ret = addrconf_disable_ipv6(ctl, valp, val);
5730         if (ret)
5731                 *ppos = pos;
5732         return ret;
5733 }
5734
5735 static
5736 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5737                               void __user *buffer, size_t *lenp, loff_t *ppos)
5738 {
5739         int *valp = ctl->data;
5740         int ret;
5741         int old, new;
5742
5743         old = *valp;
5744         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5745         new = *valp;
5746
5747         if (write && old != new) {
5748                 struct net *net = ctl->extra2;
5749
5750                 if (!rtnl_trylock())
5751                         return restart_syscall();
5752
5753                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5754                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5755                                                      NETCONFA_PROXY_NEIGH,
5756                                                      NETCONFA_IFINDEX_DEFAULT,
5757                                                      net->ipv6.devconf_dflt);
5758                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5759                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5760                                                      NETCONFA_PROXY_NEIGH,
5761                                                      NETCONFA_IFINDEX_ALL,
5762                                                      net->ipv6.devconf_all);
5763                 else {
5764                         struct inet6_dev *idev = ctl->extra1;
5765
5766                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5767                                                      NETCONFA_PROXY_NEIGH,
5768                                                      idev->dev->ifindex,
5769                                                      &idev->cnf);
5770                 }
5771                 rtnl_unlock();
5772         }
5773
5774         return ret;
5775 }
5776
5777 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5778                                          void __user *buffer, size_t *lenp,
5779                                          loff_t *ppos)
5780 {
5781         int ret = 0;
5782         int new_val;
5783         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5784         struct net *net = (struct net *)ctl->extra2;
5785
5786         if (!rtnl_trylock())
5787                 return restart_syscall();
5788
5789         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5790
5791         if (write) {
5792                 new_val = *((int *)ctl->data);
5793
5794                 if (check_addr_gen_mode(new_val) < 0) {
5795                         ret = -EINVAL;
5796                         goto out;
5797                 }
5798
5799                 /* request for default */
5800                 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5801                         ipv6_devconf_dflt.addr_gen_mode = new_val;
5802
5803                 /* request for individual net device */
5804                 } else {
5805                         if (!idev)
5806                                 goto out;
5807
5808                         if (check_stable_privacy(idev, net, new_val) < 0) {
5809                                 ret = -EINVAL;
5810                                 goto out;
5811                         }
5812
5813                         if (idev->cnf.addr_gen_mode != new_val) {
5814                                 idev->cnf.addr_gen_mode = new_val;
5815                                 addrconf_dev_config(idev->dev);
5816                         }
5817                 }
5818         }
5819
5820 out:
5821         rtnl_unlock();
5822
5823         return ret;
5824 }
5825
5826 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5827                                          void __user *buffer, size_t *lenp,
5828                                          loff_t *ppos)
5829 {
5830         int err;
5831         struct in6_addr addr;
5832         char str[IPV6_MAX_STRLEN];
5833         struct ctl_table lctl = *ctl;
5834         struct net *net = ctl->extra2;
5835         struct ipv6_stable_secret *secret = ctl->data;
5836
5837         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5838                 return -EIO;
5839
5840         lctl.maxlen = IPV6_MAX_STRLEN;
5841         lctl.data = str;
5842
5843         if (!rtnl_trylock())
5844                 return restart_syscall();
5845
5846         if (!write && !secret->initialized) {
5847                 err = -EIO;
5848                 goto out;
5849         }
5850
5851         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5852         if (err >= sizeof(str)) {
5853                 err = -EIO;
5854                 goto out;
5855         }
5856
5857         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5858         if (err || !write)
5859                 goto out;
5860
5861         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5862                 err = -EIO;
5863                 goto out;
5864         }
5865
5866         secret->initialized = true;
5867         secret->secret = addr;
5868
5869         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5870                 struct net_device *dev;
5871
5872                 for_each_netdev(net, dev) {
5873                         struct inet6_dev *idev = __in6_dev_get(dev);
5874
5875                         if (idev) {
5876                                 idev->cnf.addr_gen_mode =
5877                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5878                         }
5879                 }
5880         } else {
5881                 struct inet6_dev *idev = ctl->extra1;
5882
5883                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5884         }
5885
5886 out:
5887         rtnl_unlock();
5888
5889         return err;
5890 }
5891
5892 static
5893 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5894                                                 int write,
5895                                                 void __user *buffer,
5896                                                 size_t *lenp,
5897                                                 loff_t *ppos)
5898 {
5899         int *valp = ctl->data;
5900         int val = *valp;
5901         loff_t pos = *ppos;
5902         struct ctl_table lctl;
5903         int ret;
5904
5905         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5906          * we should not modify it until we get the rtnl lock.
5907          */
5908         lctl = *ctl;
5909         lctl.data = &val;
5910
5911         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5912
5913         if (write)
5914                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5915         if (ret)
5916                 *ppos = pos;
5917         return ret;
5918 }
5919
5920 static
5921 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5922 {
5923         if (rt) {
5924                 if (action)
5925                         rt->dst.flags |= DST_NOPOLICY;
5926                 else
5927                         rt->dst.flags &= ~DST_NOPOLICY;
5928         }
5929 }
5930
5931 static
5932 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5933 {
5934         struct inet6_ifaddr *ifa;
5935
5936         read_lock_bh(&idev->lock);
5937         list_for_each_entry(ifa, &idev->addr_list, if_list) {
5938                 spin_lock(&ifa->lock);
5939                 if (ifa->rt) {
5940                         struct rt6_info *rt = ifa->rt;
5941                         struct fib6_table *table = rt->rt6i_table;
5942                         int cpu;
5943
5944                         read_lock(&table->tb6_lock);
5945                         addrconf_set_nopolicy(ifa->rt, val);
5946                         if (rt->rt6i_pcpu) {
5947                                 for_each_possible_cpu(cpu) {
5948                                         struct rt6_info **rtp;
5949
5950                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5951                                         addrconf_set_nopolicy(*rtp, val);
5952                                 }
5953                         }
5954                         read_unlock(&table->tb6_lock);
5955                 }
5956                 spin_unlock(&ifa->lock);
5957         }
5958         read_unlock_bh(&idev->lock);
5959 }
5960
5961 static
5962 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5963 {
5964         struct inet6_dev *idev;
5965         struct net *net;
5966
5967         if (!rtnl_trylock())
5968                 return restart_syscall();
5969
5970         *valp = val;
5971
5972         net = (struct net *)ctl->extra2;
5973         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5974                 rtnl_unlock();
5975                 return 0;
5976         }
5977
5978         if (valp == &net->ipv6.devconf_all->disable_policy)  {
5979                 struct net_device *dev;
5980
5981                 for_each_netdev(net, dev) {
5982                         idev = __in6_dev_get(dev);
5983                         if (idev)
5984                                 addrconf_disable_policy_idev(idev, val);
5985                 }
5986         } else {
5987                 idev = (struct inet6_dev *)ctl->extra1;
5988                 addrconf_disable_policy_idev(idev, val);
5989         }
5990
5991         rtnl_unlock();
5992         return 0;
5993 }
5994
5995 static
5996 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
5997                                    void __user *buffer, size_t *lenp,
5998                                    loff_t *ppos)
5999 {
6000         int *valp = ctl->data;
6001         int val = *valp;
6002         loff_t pos = *ppos;
6003         struct ctl_table lctl;
6004         int ret;
6005
6006         lctl = *ctl;
6007         lctl.data = &val;
6008         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6009
6010         if (write && (*valp != val))
6011                 ret = addrconf_disable_policy(ctl, valp, val);
6012
6013         if (ret)
6014                 *ppos = pos;
6015
6016         return ret;
6017 }
6018
6019 static int minus_one = -1;
6020 static const int one = 1;
6021 static const int two_five_five = 255;
6022
6023 static const struct ctl_table addrconf_sysctl[] = {
6024         {
6025                 .procname       = "forwarding",
6026                 .data           = &ipv6_devconf.forwarding,
6027                 .maxlen         = sizeof(int),
6028                 .mode           = 0644,
6029                 .proc_handler   = addrconf_sysctl_forward,
6030         },
6031         {
6032                 .procname       = "hop_limit",
6033                 .data           = &ipv6_devconf.hop_limit,
6034                 .maxlen         = sizeof(int),
6035                 .mode           = 0644,
6036                 .proc_handler   = proc_dointvec_minmax,
6037                 .extra1         = (void *)&one,
6038                 .extra2         = (void *)&two_five_five,
6039         },
6040         {
6041                 .procname       = "mtu",
6042                 .data           = &ipv6_devconf.mtu6,
6043                 .maxlen         = sizeof(int),
6044                 .mode           = 0644,
6045                 .proc_handler   = addrconf_sysctl_mtu,
6046         },
6047         {
6048                 .procname       = "accept_ra",
6049                 .data           = &ipv6_devconf.accept_ra,
6050                 .maxlen         = sizeof(int),
6051                 .mode           = 0644,
6052                 .proc_handler   = proc_dointvec,
6053         },
6054         {
6055                 .procname       = "accept_redirects",
6056                 .data           = &ipv6_devconf.accept_redirects,
6057                 .maxlen         = sizeof(int),
6058                 .mode           = 0644,
6059                 .proc_handler   = proc_dointvec,
6060         },
6061         {
6062                 .procname       = "autoconf",
6063                 .data           = &ipv6_devconf.autoconf,
6064                 .maxlen         = sizeof(int),
6065                 .mode           = 0644,
6066                 .proc_handler   = proc_dointvec,
6067         },
6068         {
6069                 .procname       = "dad_transmits",
6070                 .data           = &ipv6_devconf.dad_transmits,
6071                 .maxlen         = sizeof(int),
6072                 .mode           = 0644,
6073                 .proc_handler   = proc_dointvec,
6074         },
6075         {
6076                 .procname       = "router_solicitations",
6077                 .data           = &ipv6_devconf.rtr_solicits,
6078                 .maxlen         = sizeof(int),
6079                 .mode           = 0644,
6080                 .proc_handler   = proc_dointvec_minmax,
6081                 .extra1         = &minus_one,
6082         },
6083         {
6084                 .procname       = "router_solicitation_interval",
6085                 .data           = &ipv6_devconf.rtr_solicit_interval,
6086                 .maxlen         = sizeof(int),
6087                 .mode           = 0644,
6088                 .proc_handler   = proc_dointvec_jiffies,
6089         },
6090         {
6091                 .procname       = "router_solicitation_max_interval",
6092                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6093                 .maxlen         = sizeof(int),
6094                 .mode           = 0644,
6095                 .proc_handler   = proc_dointvec_jiffies,
6096         },
6097         {
6098                 .procname       = "router_solicitation_delay",
6099                 .data           = &ipv6_devconf.rtr_solicit_delay,
6100                 .maxlen         = sizeof(int),
6101                 .mode           = 0644,
6102                 .proc_handler   = proc_dointvec_jiffies,
6103         },
6104         {
6105                 .procname       = "force_mld_version",
6106                 .data           = &ipv6_devconf.force_mld_version,
6107                 .maxlen         = sizeof(int),
6108                 .mode           = 0644,
6109                 .proc_handler   = proc_dointvec,
6110         },
6111         {
6112                 .procname       = "mldv1_unsolicited_report_interval",
6113                 .data           =
6114                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6115                 .maxlen         = sizeof(int),
6116                 .mode           = 0644,
6117                 .proc_handler   = proc_dointvec_ms_jiffies,
6118         },
6119         {
6120                 .procname       = "mldv2_unsolicited_report_interval",
6121                 .data           =
6122                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6123                 .maxlen         = sizeof(int),
6124                 .mode           = 0644,
6125                 .proc_handler   = proc_dointvec_ms_jiffies,
6126         },
6127         {
6128                 .procname       = "use_tempaddr",
6129                 .data           = &ipv6_devconf.use_tempaddr,
6130                 .maxlen         = sizeof(int),
6131                 .mode           = 0644,
6132                 .proc_handler   = proc_dointvec,
6133         },
6134         {
6135                 .procname       = "temp_valid_lft",
6136                 .data           = &ipv6_devconf.temp_valid_lft,
6137                 .maxlen         = sizeof(int),
6138                 .mode           = 0644,
6139                 .proc_handler   = proc_dointvec,
6140         },
6141         {
6142                 .procname       = "temp_prefered_lft",
6143                 .data           = &ipv6_devconf.temp_prefered_lft,
6144                 .maxlen         = sizeof(int),
6145                 .mode           = 0644,
6146                 .proc_handler   = proc_dointvec,
6147         },
6148         {
6149                 .procname       = "regen_max_retry",
6150                 .data           = &ipv6_devconf.regen_max_retry,
6151                 .maxlen         = sizeof(int),
6152                 .mode           = 0644,
6153                 .proc_handler   = proc_dointvec,
6154         },
6155         {
6156                 .procname       = "max_desync_factor",
6157                 .data           = &ipv6_devconf.max_desync_factor,
6158                 .maxlen         = sizeof(int),
6159                 .mode           = 0644,
6160                 .proc_handler   = proc_dointvec,
6161         },
6162         {
6163                 .procname       = "max_addresses",
6164                 .data           = &ipv6_devconf.max_addresses,
6165                 .maxlen         = sizeof(int),
6166                 .mode           = 0644,
6167                 .proc_handler   = proc_dointvec,
6168         },
6169         {
6170                 .procname       = "accept_ra_defrtr",
6171                 .data           = &ipv6_devconf.accept_ra_defrtr,
6172                 .maxlen         = sizeof(int),
6173                 .mode           = 0644,
6174                 .proc_handler   = proc_dointvec,
6175         },
6176         {
6177                 .procname       = "accept_ra_min_hop_limit",
6178                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6179                 .maxlen         = sizeof(int),
6180                 .mode           = 0644,
6181                 .proc_handler   = proc_dointvec,
6182         },
6183         {
6184                 .procname       = "accept_ra_pinfo",
6185                 .data           = &ipv6_devconf.accept_ra_pinfo,
6186                 .maxlen         = sizeof(int),
6187                 .mode           = 0644,
6188                 .proc_handler   = proc_dointvec,
6189         },
6190 #ifdef CONFIG_IPV6_ROUTER_PREF
6191         {
6192                 .procname       = "accept_ra_rtr_pref",
6193                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6194                 .maxlen         = sizeof(int),
6195                 .mode           = 0644,
6196                 .proc_handler   = proc_dointvec,
6197         },
6198         {
6199                 .procname       = "router_probe_interval",
6200                 .data           = &ipv6_devconf.rtr_probe_interval,
6201                 .maxlen         = sizeof(int),
6202                 .mode           = 0644,
6203                 .proc_handler   = proc_dointvec_jiffies,
6204         },
6205 #ifdef CONFIG_IPV6_ROUTE_INFO
6206         {
6207                 .procname       = "accept_ra_rt_info_min_plen",
6208                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6209                 .maxlen         = sizeof(int),
6210                 .mode           = 0644,
6211                 .proc_handler   = proc_dointvec,
6212         },
6213         {
6214                 .procname       = "accept_ra_rt_info_max_plen",
6215                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6216                 .maxlen         = sizeof(int),
6217                 .mode           = 0644,
6218                 .proc_handler   = proc_dointvec,
6219         },
6220 #endif
6221 #endif
6222         {
6223                 .procname       = "proxy_ndp",
6224                 .data           = &ipv6_devconf.proxy_ndp,
6225                 .maxlen         = sizeof(int),
6226                 .mode           = 0644,
6227                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6228         },
6229         {
6230                 .procname       = "accept_source_route",
6231                 .data           = &ipv6_devconf.accept_source_route,
6232                 .maxlen         = sizeof(int),
6233                 .mode           = 0644,
6234                 .proc_handler   = proc_dointvec,
6235         },
6236 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6237         {
6238                 .procname       = "optimistic_dad",
6239                 .data           = &ipv6_devconf.optimistic_dad,
6240                 .maxlen         = sizeof(int),
6241                 .mode           = 0644,
6242                 .proc_handler   = proc_dointvec,
6243         },
6244         {
6245                 .procname       = "use_optimistic",
6246                 .data           = &ipv6_devconf.use_optimistic,
6247                 .maxlen         = sizeof(int),
6248                 .mode           = 0644,
6249                 .proc_handler   = proc_dointvec,
6250         },
6251 #endif
6252 #ifdef CONFIG_IPV6_MROUTE
6253         {
6254                 .procname       = "mc_forwarding",
6255                 .data           = &ipv6_devconf.mc_forwarding,
6256                 .maxlen         = sizeof(int),
6257                 .mode           = 0444,
6258                 .proc_handler   = proc_dointvec,
6259         },
6260 #endif
6261         {
6262                 .procname       = "disable_ipv6",
6263                 .data           = &ipv6_devconf.disable_ipv6,
6264                 .maxlen         = sizeof(int),
6265                 .mode           = 0644,
6266                 .proc_handler   = addrconf_sysctl_disable,
6267         },
6268         {
6269                 .procname       = "accept_dad",
6270                 .data           = &ipv6_devconf.accept_dad,
6271                 .maxlen         = sizeof(int),
6272                 .mode           = 0644,
6273                 .proc_handler   = proc_dointvec,
6274         },
6275         {
6276                 .procname       = "force_tllao",
6277                 .data           = &ipv6_devconf.force_tllao,
6278                 .maxlen         = sizeof(int),
6279                 .mode           = 0644,
6280                 .proc_handler   = proc_dointvec
6281         },
6282         {
6283                 .procname       = "ndisc_notify",
6284                 .data           = &ipv6_devconf.ndisc_notify,
6285                 .maxlen         = sizeof(int),
6286                 .mode           = 0644,
6287                 .proc_handler   = proc_dointvec
6288         },
6289         {
6290                 .procname       = "suppress_frag_ndisc",
6291                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6292                 .maxlen         = sizeof(int),
6293                 .mode           = 0644,
6294                 .proc_handler   = proc_dointvec
6295         },
6296         {
6297                 .procname       = "accept_ra_from_local",
6298                 .data           = &ipv6_devconf.accept_ra_from_local,
6299                 .maxlen         = sizeof(int),
6300                 .mode           = 0644,
6301                 .proc_handler   = proc_dointvec,
6302         },
6303         {
6304                 .procname       = "accept_ra_mtu",
6305                 .data           = &ipv6_devconf.accept_ra_mtu,
6306                 .maxlen         = sizeof(int),
6307                 .mode           = 0644,
6308                 .proc_handler   = proc_dointvec,
6309         },
6310         {
6311                 .procname       = "stable_secret",
6312                 .data           = &ipv6_devconf.stable_secret,
6313                 .maxlen         = IPV6_MAX_STRLEN,
6314                 .mode           = 0600,
6315                 .proc_handler   = addrconf_sysctl_stable_secret,
6316         },
6317         {
6318                 .procname       = "use_oif_addrs_only",
6319                 .data           = &ipv6_devconf.use_oif_addrs_only,
6320                 .maxlen         = sizeof(int),
6321                 .mode           = 0644,
6322                 .proc_handler   = proc_dointvec,
6323         },
6324         {
6325                 .procname       = "ignore_routes_with_linkdown",
6326                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6327                 .maxlen         = sizeof(int),
6328                 .mode           = 0644,
6329                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6330         },
6331         {
6332                 .procname       = "drop_unicast_in_l2_multicast",
6333                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6334                 .maxlen         = sizeof(int),
6335                 .mode           = 0644,
6336                 .proc_handler   = proc_dointvec,
6337         },
6338         {
6339                 .procname       = "drop_unsolicited_na",
6340                 .data           = &ipv6_devconf.drop_unsolicited_na,
6341                 .maxlen         = sizeof(int),
6342                 .mode           = 0644,
6343                 .proc_handler   = proc_dointvec,
6344         },
6345         {
6346                 .procname       = "keep_addr_on_down",
6347                 .data           = &ipv6_devconf.keep_addr_on_down,
6348                 .maxlen         = sizeof(int),
6349                 .mode           = 0644,
6350                 .proc_handler   = proc_dointvec,
6351
6352         },
6353         {
6354                 .procname       = "seg6_enabled",
6355                 .data           = &ipv6_devconf.seg6_enabled,
6356                 .maxlen         = sizeof(int),
6357                 .mode           = 0644,
6358                 .proc_handler   = proc_dointvec,
6359         },
6360 #ifdef CONFIG_IPV6_SEG6_HMAC
6361         {
6362                 .procname       = "seg6_require_hmac",
6363                 .data           = &ipv6_devconf.seg6_require_hmac,
6364                 .maxlen         = sizeof(int),
6365                 .mode           = 0644,
6366                 .proc_handler   = proc_dointvec,
6367         },
6368 #endif
6369         {
6370                 .procname       = "enhanced_dad",
6371                 .data           = &ipv6_devconf.enhanced_dad,
6372                 .maxlen         = sizeof(int),
6373                 .mode           = 0644,
6374                 .proc_handler   = proc_dointvec,
6375         },
6376         {
6377                 .procname               = "addr_gen_mode",
6378                 .data                   = &ipv6_devconf.addr_gen_mode,
6379                 .maxlen                 = sizeof(int),
6380                 .mode                   = 0644,
6381                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6382         },
6383         {
6384                 .procname       = "disable_policy",
6385                 .data           = &ipv6_devconf.disable_policy,
6386                 .maxlen         = sizeof(int),
6387                 .mode           = 0644,
6388                 .proc_handler   = addrconf_sysctl_disable_policy,
6389         },
6390         {
6391                 /* sentinel */
6392         }
6393 };
6394
6395 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6396                 struct inet6_dev *idev, struct ipv6_devconf *p)
6397 {
6398         int i, ifindex;
6399         struct ctl_table *table;
6400         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6401
6402         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6403         if (!table)
6404                 goto out;
6405
6406         for (i = 0; table[i].data; i++) {
6407                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6408                 /* If one of these is already set, then it is not safe to
6409                  * overwrite either of them: this makes proc_dointvec_minmax
6410                  * usable.
6411                  */
6412                 if (!table[i].extra1 && !table[i].extra2) {
6413                         table[i].extra1 = idev; /* embedded; no ref */
6414                         table[i].extra2 = net;
6415                 }
6416         }
6417
6418         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6419
6420         p->sysctl_header = register_net_sysctl(net, path, table);
6421         if (!p->sysctl_header)
6422                 goto free;
6423
6424         if (!strcmp(dev_name, "all"))
6425                 ifindex = NETCONFA_IFINDEX_ALL;
6426         else if (!strcmp(dev_name, "default"))
6427                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6428         else
6429                 ifindex = idev->dev->ifindex;
6430         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6431                                      ifindex, p);
6432         return 0;
6433
6434 free:
6435         kfree(table);
6436 out:
6437         return -ENOBUFS;
6438 }
6439
6440 static void __addrconf_sysctl_unregister(struct net *net,
6441                                          struct ipv6_devconf *p, int ifindex)
6442 {
6443         struct ctl_table *table;
6444
6445         if (!p->sysctl_header)
6446                 return;
6447
6448         table = p->sysctl_header->ctl_table_arg;
6449         unregister_net_sysctl_table(p->sysctl_header);
6450         p->sysctl_header = NULL;
6451         kfree(table);
6452
6453         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6454 }
6455
6456 static int addrconf_sysctl_register(struct inet6_dev *idev)
6457 {
6458         int err;
6459
6460         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6461                 return -EINVAL;
6462
6463         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6464                                     &ndisc_ifinfo_sysctl_change);
6465         if (err)
6466                 return err;
6467         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6468                                          idev, &idev->cnf);
6469         if (err)
6470                 neigh_sysctl_unregister(idev->nd_parms);
6471
6472         return err;
6473 }
6474
6475 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6476 {
6477         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6478                                      idev->dev->ifindex);
6479         neigh_sysctl_unregister(idev->nd_parms);
6480 }
6481
6482
6483 #endif
6484
6485 static int __net_init addrconf_init_net(struct net *net)
6486 {
6487         int err = -ENOMEM;
6488         struct ipv6_devconf *all, *dflt;
6489
6490         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6491         if (!all)
6492                 goto err_alloc_all;
6493
6494         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6495         if (!dflt)
6496                 goto err_alloc_dflt;
6497
6498         /* these will be inherited by all namespaces */
6499         dflt->autoconf = ipv6_defaults.autoconf;
6500         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6501
6502         dflt->stable_secret.initialized = false;
6503         all->stable_secret.initialized = false;
6504
6505         net->ipv6.devconf_all = all;
6506         net->ipv6.devconf_dflt = dflt;
6507
6508 #ifdef CONFIG_SYSCTL
6509         err = __addrconf_sysctl_register(net, "all", NULL, all);
6510         if (err < 0)
6511                 goto err_reg_all;
6512
6513         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6514         if (err < 0)
6515                 goto err_reg_dflt;
6516 #endif
6517         return 0;
6518
6519 #ifdef CONFIG_SYSCTL
6520 err_reg_dflt:
6521         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6522 err_reg_all:
6523         kfree(dflt);
6524 #endif
6525 err_alloc_dflt:
6526         kfree(all);
6527 err_alloc_all:
6528         return err;
6529 }
6530
6531 static void __net_exit addrconf_exit_net(struct net *net)
6532 {
6533 #ifdef CONFIG_SYSCTL
6534         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6535                                      NETCONFA_IFINDEX_DEFAULT);
6536         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6537                                      NETCONFA_IFINDEX_ALL);
6538 #endif
6539         kfree(net->ipv6.devconf_dflt);
6540         kfree(net->ipv6.devconf_all);
6541 }
6542
6543 static struct pernet_operations addrconf_ops = {
6544         .init = addrconf_init_net,
6545         .exit = addrconf_exit_net,
6546 };
6547
6548 static struct rtnl_af_ops inet6_ops __read_mostly = {
6549         .family           = AF_INET6,
6550         .fill_link_af     = inet6_fill_link_af,
6551         .get_link_af_size = inet6_get_link_af_size,
6552         .validate_link_af = inet6_validate_link_af,
6553         .set_link_af      = inet6_set_link_af,
6554 };
6555
6556 /*
6557  *      Init / cleanup code
6558  */
6559
6560 int __init addrconf_init(void)
6561 {
6562         struct inet6_dev *idev;
6563         int i, err;
6564
6565         err = ipv6_addr_label_init();
6566         if (err < 0) {
6567                 pr_crit("%s: cannot initialize default policy table: %d\n",
6568                         __func__, err);
6569                 goto out;
6570         }
6571
6572         err = register_pernet_subsys(&addrconf_ops);
6573         if (err < 0)
6574                 goto out_addrlabel;
6575
6576         addrconf_wq = create_workqueue("ipv6_addrconf");
6577         if (!addrconf_wq) {
6578                 err = -ENOMEM;
6579                 goto out_nowq;
6580         }
6581
6582         /* The addrconf netdev notifier requires that loopback_dev
6583          * has it's ipv6 private information allocated and setup
6584          * before it can bring up and give link-local addresses
6585          * to other devices which are up.
6586          *
6587          * Unfortunately, loopback_dev is not necessarily the first
6588          * entry in the global dev_base list of net devices.  In fact,
6589          * it is likely to be the very last entry on that list.
6590          * So this causes the notifier registry below to try and
6591          * give link-local addresses to all devices besides loopback_dev
6592          * first, then loopback_dev, which cases all the non-loopback_dev
6593          * devices to fail to get a link-local address.
6594          *
6595          * So, as a temporary fix, allocate the ipv6 structure for
6596          * loopback_dev first by hand.
6597          * Longer term, all of the dependencies ipv6 has upon the loopback
6598          * device and it being up should be removed.
6599          */
6600         rtnl_lock();
6601         idev = ipv6_add_dev(init_net.loopback_dev);
6602         rtnl_unlock();
6603         if (IS_ERR(idev)) {
6604                 err = PTR_ERR(idev);
6605                 goto errlo;
6606         }
6607
6608         ip6_route_init_special_entries();
6609
6610         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6611                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6612
6613         register_netdevice_notifier(&ipv6_dev_notf);
6614
6615         addrconf_verify();
6616
6617         rtnl_af_register(&inet6_ops);
6618
6619         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6620                               0);
6621         if (err < 0)
6622                 goto errout;
6623
6624         /* Only the first call to __rtnl_register can fail */
6625         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, 0);
6626         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, 0);
6627         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6628                         inet6_dump_ifaddr, 0);
6629         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6630                         inet6_dump_ifmcaddr, 0);
6631         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6632                         inet6_dump_ifacaddr, 0);
6633         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6634                         inet6_netconf_dump_devconf, 0);
6635
6636         ipv6_addr_label_rtnl_register();
6637
6638         return 0;
6639 errout:
6640         rtnl_af_unregister(&inet6_ops);
6641         unregister_netdevice_notifier(&ipv6_dev_notf);
6642 errlo:
6643         destroy_workqueue(addrconf_wq);
6644 out_nowq:
6645         unregister_pernet_subsys(&addrconf_ops);
6646 out_addrlabel:
6647         ipv6_addr_label_cleanup();
6648 out:
6649         return err;
6650 }
6651
6652 void addrconf_cleanup(void)
6653 {
6654         struct net_device *dev;
6655         int i;
6656
6657         unregister_netdevice_notifier(&ipv6_dev_notf);
6658         unregister_pernet_subsys(&addrconf_ops);
6659         ipv6_addr_label_cleanup();
6660
6661         rtnl_lock();
6662
6663         __rtnl_af_unregister(&inet6_ops);
6664
6665         /* clean dev list */
6666         for_each_netdev(&init_net, dev) {
6667                 if (__in6_dev_get(dev) == NULL)
6668                         continue;
6669                 addrconf_ifdown(dev, 1);
6670         }
6671         addrconf_ifdown(init_net.loopback_dev, 2);
6672
6673         /*
6674          *      Check hash table.
6675          */
6676         spin_lock_bh(&addrconf_hash_lock);
6677         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6678                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6679         spin_unlock_bh(&addrconf_hash_lock);
6680         cancel_delayed_work(&addr_chk_work);
6681         rtnl_unlock();
6682
6683         destroy_workqueue(addrconf_wq);
6684 }