net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <[email protected]>
   7  *      Alexey Kuznetsov        <[email protected]>
   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  *      <[email protected]>
  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 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  98
  99 #define IPV6_MAX_STRLEN \
 100         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 108 {
 109         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 110         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 111         do_div(tmp, 1000000);
 112         return (s32)tmp;
 113 }
 114
 115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 116 {
 117         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 118         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 119         do_div(tmp, 1000000);
 120         if ((s32)tmp > mrt) {
 121                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 122                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 123                 do_div(tmp, 1000000);
 124         }
 125         return (s32)tmp;
 126 }
 127
 128 #ifdef CONFIG_SYSCTL
 129 static int addrconf_sysctl_register(struct inet6_dev *idev);
 130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 131 #else
 132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 133 {
 134         return 0;
 135 }
 136
 137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 138 {
 139 }
 140 #endif
 141
 142 static void ipv6_regen_rndid(struct inet6_dev *idev);
 143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 144
 145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 146 static int ipv6_count_addresses(const struct inet6_dev *idev);
 147 static int ipv6_generate_stable_address(struct in6_addr *addr,
 148                                         u8 dad_count,
 149                                         const struct inet6_dev *idev);
 150
 151 #define IN6_ADDR_HSIZE_SHIFT    8
 152 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 153 /*
 154  *      Configured unicast address hash table
 155  */
 156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 157 static DEFINE_SPINLOCK(addrconf_hash_lock);
 158
 159 static void addrconf_verify(void);
 160 static void addrconf_verify_rtnl(void);
 161 static void addrconf_verify_work(struct work_struct *);
 162
 163 static struct workqueue_struct *addrconf_wq;
 164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 165
 166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 168
 169 static void addrconf_type_change(struct net_device *dev,
 170                                  unsigned long event);
 171 static int addrconf_ifdown(struct net_device *dev, int how);
 172
 173 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 174                                                   int plen,
 175                                                   const struct net_device *dev,
 176                                                   u32 flags, u32 noflags);
 177
 178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 179 static void addrconf_dad_work(struct work_struct *w);
 180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 181                                    bool send_na);
 182 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 183 static void addrconf_rs_timer(struct timer_list *t);
 184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 186
 187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 188                                 struct prefix_info *pinfo);
 189
 190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 191         .forwarding             = 0,
 192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 193         .mtu6                   = IPV6_MIN_MTU,
 194         .accept_ra              = 1,
 195         .accept_redirects       = 1,
 196         .autoconf               = 1,
 197         .force_mld_version      = 0,
 198         .mldv1_unsolicited_report_interval = 10 * HZ,
 199         .mldv2_unsolicited_report_interval = HZ,
 200         .dad_transmits          = 1,
 201         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 202         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 203         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205         .use_tempaddr           = 0,
 206         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 207         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 208         .regen_max_retry        = REGEN_MAX_RETRY,
 209         .max_desync_factor      = MAX_DESYNC_FACTOR,
 210         .max_addresses          = IPV6_MAX_ADDRESSES,
 211         .accept_ra_defrtr       = 1,
 212         .accept_ra_from_local   = 0,
 213         .accept_ra_min_hop_limit= 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_min_plen = 0,
 220         .accept_ra_rt_info_max_plen = 0,
 221 #endif
 222 #endif
 223         .proxy_ndp              = 0,
 224         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 225         .disable_ipv6           = 0,
 226         .accept_dad             = 0,
 227         .suppress_frag_ndisc    = 1,
 228         .accept_ra_mtu          = 1,
 229         .stable_secret          = {
 230                 .initialized = false,
 231         },
 232         .use_oif_addrs_only     = 0,
 233         .ignore_routes_with_linkdown = 0,
 234         .keep_addr_on_down      = 0,
 235         .seg6_enabled           = 0,
 236 #ifdef CONFIG_IPV6_SEG6_HMAC
 237         .seg6_require_hmac      = 0,
 238 #endif
 239         .enhanced_dad           = 1,
 240         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 241         .disable_policy         = 0,
 242 };
 243
 244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 245         .forwarding             = 0,
 246         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 247         .mtu6                   = IPV6_MIN_MTU,
 248         .accept_ra              = 1,
 249         .accept_redirects       = 1,
 250         .autoconf               = 1,
 251         .force_mld_version      = 0,
 252         .mldv1_unsolicited_report_interval = 10 * HZ,
 253         .mldv2_unsolicited_report_interval = HZ,
 254         .dad_transmits          = 1,
 255         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 256         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 257         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 258         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 259         .use_tempaddr           = 0,
 260         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 261         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 262         .regen_max_retry        = REGEN_MAX_RETRY,
 263         .max_desync_factor      = MAX_DESYNC_FACTOR,
 264         .max_addresses          = IPV6_MAX_ADDRESSES,
 265         .accept_ra_defrtr       = 1,
 266         .accept_ra_from_local   = 0,
 267         .accept_ra_min_hop_limit= 1,
 268         .accept_ra_pinfo        = 1,
 269 #ifdef CONFIG_IPV6_ROUTER_PREF
 270         .accept_ra_rtr_pref     = 1,
 271         .rtr_probe_interval     = 60 * HZ,
 272 #ifdef CONFIG_IPV6_ROUTE_INFO
 273         .accept_ra_rt_info_min_plen = 0,
 274         .accept_ra_rt_info_max_plen = 0,
 275 #endif
 276 #endif
 277         .proxy_ndp              = 0,
 278         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 279         .disable_ipv6           = 0,
 280         .accept_dad             = 1,
 281         .suppress_frag_ndisc    = 1,
 282         .accept_ra_mtu          = 1,
 283         .stable_secret          = {
 284                 .initialized = false,
 285         },
 286         .use_oif_addrs_only     = 0,
 287         .ignore_routes_with_linkdown = 0,
 288         .keep_addr_on_down      = 0,
 289         .seg6_enabled           = 0,
 290 #ifdef CONFIG_IPV6_SEG6_HMAC
 291         .seg6_require_hmac      = 0,
 292 #endif
 293         .enhanced_dad           = 1,
 294         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 295         .disable_policy         = 0,
 296 };
 297
 298 /* Check if link is ready: is it up and is a valid qdisc available */
 299 static inline bool addrconf_link_ready(const struct net_device *dev)
 300 {
 301         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 302 }
 303
 304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 305 {
 306         if (del_timer(&idev->rs_timer))
 307                 __in6_dev_put(idev);
 308 }
 309
 310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 311 {
 312         if (cancel_delayed_work(&ifp->dad_work))
 313                 __in6_ifa_put(ifp);
 314 }
 315
 316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 317                                   unsigned long when)
 318 {
 319         if (!timer_pending(&idev->rs_timer))
 320                 in6_dev_hold(idev);
 321         mod_timer(&idev->rs_timer, jiffies + when);
 322 }
 323
 324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 325                                    unsigned long delay)
 326 {
 327         in6_ifa_hold(ifp);
 328         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 329                 in6_ifa_put(ifp);
 330 }
 331
 332 static int snmp6_alloc_dev(struct inet6_dev *idev)
 333 {
 334         int i;
 335
 336         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 337         if (!idev->stats.ipv6)
 338                 goto err_ip;
 339
 340         for_each_possible_cpu(i) {
 341                 struct ipstats_mib *addrconf_stats;
 342                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 343                 u64_stats_init(&addrconf_stats->syncp);
 344         }
 345
 346
 347         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 348                                         GFP_KERNEL);
 349         if (!idev->stats.icmpv6dev)
 350                 goto err_icmp;
 351         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 352                                            GFP_KERNEL);
 353         if (!idev->stats.icmpv6msgdev)
 354                 goto err_icmpmsg;
 355
 356         return 0;
 357
 358 err_icmpmsg:
 359         kfree(idev->stats.icmpv6dev);
 360 err_icmp:
 361         free_percpu(idev->stats.ipv6);
 362 err_ip:
 363         return -ENOMEM;
 364 }
 365
 366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 367 {
 368         struct inet6_dev *ndev;
 369         int err = -ENOMEM;
 370
 371         ASSERT_RTNL();
 372
 373         if (dev->mtu < IPV6_MIN_MTU)
 374                 return ERR_PTR(-EINVAL);
 375
 376         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 377         if (!ndev)
 378                 return ERR_PTR(err);
 379
 380         rwlock_init(&ndev->lock);
 381         ndev->dev = dev;
 382         INIT_LIST_HEAD(&ndev->addr_list);
 383         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 384         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 385
 386         if (ndev->cnf.stable_secret.initialized)
 387                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 388
 389         ndev->cnf.mtu6 = dev->mtu;
 390         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 391         if (!ndev->nd_parms) {
 392                 kfree(ndev);
 393                 return ERR_PTR(err);
 394         }
 395         if (ndev->cnf.forwarding)
 396                 dev_disable_lro(dev);
 397         /* We refer to the device */
 398         dev_hold(dev);
 399
 400         if (snmp6_alloc_dev(ndev) < 0) {
 401                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 402                            __func__);
 403                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 404                 dev_put(dev);
 405                 kfree(ndev);
 406                 return ERR_PTR(err);
 407         }
 408
 409         if (snmp6_register_dev(ndev) < 0) {
 410                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 411                            __func__, dev->name);
 412                 goto err_release;
 413         }
 414
 415         /* One reference from device. */
 416         refcount_set(&ndev->refcnt, 1);
 417
 418         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 419                 ndev->cnf.accept_dad = -1;
 420
 421 #if IS_ENABLED(CONFIG_IPV6_SIT)
 422         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 423                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 424                 ndev->cnf.rtr_solicits = 0;
 425         }
 426 #endif
 427
 428         INIT_LIST_HEAD(&ndev->tempaddr_list);
 429         ndev->desync_factor = U32_MAX;
 430         if ((dev->flags&IFF_LOOPBACK) ||
 431             dev->type == ARPHRD_TUNNEL ||
 432             dev->type == ARPHRD_TUNNEL6 ||
 433             dev->type == ARPHRD_SIT ||
 434             dev->type == ARPHRD_NONE) {
 435                 ndev->cnf.use_tempaddr = -1;
 436         } else
 437                 ipv6_regen_rndid(ndev);
 438
 439         ndev->token = in6addr_any;
 440
 441         if (netif_running(dev) && addrconf_link_ready(dev))
 442                 ndev->if_flags |= IF_READY;
 443
 444         ipv6_mc_init_dev(ndev);
 445         ndev->tstamp = jiffies;
 446         err = addrconf_sysctl_register(ndev);
 447         if (err) {
 448                 ipv6_mc_destroy_dev(ndev);
 449                 snmp6_unregister_dev(ndev);
 450                 goto err_release;
 451         }
 452         /* protected by rtnl_lock */
 453         rcu_assign_pointer(dev->ip6_ptr, ndev);
 454
 455         /* Join interface-local all-node multicast group */
 456         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 457
 458         /* Join all-node multicast group */
 459         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 460
 461         /* Join all-router multicast group if forwarding is set */
 462         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 463                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 464
 465         return ndev;
 466
 467 err_release:
 468         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 469         ndev->dead = 1;
 470         in6_dev_finish_destroy(ndev);
 471         return ERR_PTR(err);
 472 }
 473
 474 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 475 {
 476         struct inet6_dev *idev;
 477
 478         ASSERT_RTNL();
 479
 480         idev = __in6_dev_get(dev);
 481         if (!idev) {
 482                 idev = ipv6_add_dev(dev);
 483                 if (IS_ERR(idev))
 484                         return NULL;
 485         }
 486
 487         if (dev->flags&IFF_UP)
 488                 ipv6_mc_up(idev);
 489         return idev;
 490 }
 491
 492 static int inet6_netconf_msgsize_devconf(int type)
 493 {
 494         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 495                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 496         bool all = false;
 497
 498         if (type == NETCONFA_ALL)
 499                 all = true;
 500
 501         if (all || type == NETCONFA_FORWARDING)
 502                 size += nla_total_size(4);
 503 #ifdef CONFIG_IPV6_MROUTE
 504         if (all || type == NETCONFA_MC_FORWARDING)
 505                 size += nla_total_size(4);
 506 #endif
 507         if (all || type == NETCONFA_PROXY_NEIGH)
 508                 size += nla_total_size(4);
 509
 510         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 511                 size += nla_total_size(4);
 512
 513         return size;
 514 }
 515
 516 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 517                                       struct ipv6_devconf *devconf, u32 portid,
 518                                       u32 seq, int event, unsigned int flags,
 519                                       int type)
 520 {
 521         struct nlmsghdr  *nlh;
 522         struct netconfmsg *ncm;
 523         bool all = false;
 524
 525         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 526                         flags);
 527         if (!nlh)
 528                 return -EMSGSIZE;
 529
 530         if (type == NETCONFA_ALL)
 531                 all = true;
 532
 533         ncm = nlmsg_data(nlh);
 534         ncm->ncm_family = AF_INET6;
 535
 536         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 537                 goto nla_put_failure;
 538
 539         if (!devconf)
 540                 goto out;
 541
 542         if ((all || type == NETCONFA_FORWARDING) &&
 543             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 544                 goto nla_put_failure;
 545 #ifdef CONFIG_IPV6_MROUTE
 546         if ((all || type == NETCONFA_MC_FORWARDING) &&
 547             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 548                         devconf->mc_forwarding) < 0)
 549                 goto nla_put_failure;
 550 #endif
 551         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 552             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 553                 goto nla_put_failure;
 554
 555         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 556             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 557                         devconf->ignore_routes_with_linkdown) < 0)
 558                 goto nla_put_failure;
 559
 560 out:
 561         nlmsg_end(skb, nlh);
 562         return 0;
 563
 564 nla_put_failure:
 565         nlmsg_cancel(skb, nlh);
 566         return -EMSGSIZE;
 567 }
 568
 569 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 570                                   int ifindex, struct ipv6_devconf *devconf)
 571 {
 572         struct sk_buff *skb;
 573         int err = -ENOBUFS;
 574
 575         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 576         if (!skb)
 577                 goto errout;
 578
 579         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 580                                          event, 0, type);
 581         if (err < 0) {
 582                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 583                 WARN_ON(err == -EMSGSIZE);
 584                 kfree_skb(skb);
 585                 goto errout;
 586         }
 587         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 588         return;
 589 errout:
 590         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 591 }
 592
 593 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 594         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 595         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 596         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 597         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 598 };
 599
 600 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 601                                      struct nlmsghdr *nlh,
 602                                      struct netlink_ext_ack *extack)
 603 {
 604         struct net *net = sock_net(in_skb->sk);
 605         struct nlattr *tb[NETCONFA_MAX+1];
 606         struct inet6_dev *in6_dev = NULL;
 607         struct net_device *dev = NULL;
 608         struct netconfmsg *ncm;
 609         struct sk_buff *skb;
 610         struct ipv6_devconf *devconf;
 611         int ifindex;
 612         int err;
 613
 614         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 615                           devconf_ipv6_policy, extack);
 616         if (err < 0)
 617                 return err;
 618
 619         if (!tb[NETCONFA_IFINDEX])
 620                 return -EINVAL;
 621
 622         err = -EINVAL;
 623         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 624         switch (ifindex) {
 625         case NETCONFA_IFINDEX_ALL:
 626                 devconf = net->ipv6.devconf_all;
 627                 break;
 628         case NETCONFA_IFINDEX_DEFAULT:
 629                 devconf = net->ipv6.devconf_dflt;
 630                 break;
 631         default:
 632                 dev = dev_get_by_index(net, ifindex);
 633                 if (!dev)
 634                         return -EINVAL;
 635                 in6_dev = in6_dev_get(dev);
 636                 if (!in6_dev)
 637                         goto errout;
 638                 devconf = &in6_dev->cnf;
 639                 break;
 640         }
 641
 642         err = -ENOBUFS;
 643         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 644         if (!skb)
 645                 goto errout;
 646
 647         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 648                                          NETLINK_CB(in_skb).portid,
 649                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 650                                          NETCONFA_ALL);
 651         if (err < 0) {
 652                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 653                 WARN_ON(err == -EMSGSIZE);
 654                 kfree_skb(skb);
 655                 goto errout;
 656         }
 657         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 658 errout:
 659         if (in6_dev)
 660                 in6_dev_put(in6_dev);
 661         if (dev)
 662                 dev_put(dev);
 663         return err;
 664 }
 665
 666 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 667                                       struct netlink_callback *cb)
 668 {
 669         const struct nlmsghdr *nlh = cb->nlh;
 670         struct net *net = sock_net(skb->sk);
 671         int h, s_h;
 672         int idx, s_idx;
 673         struct net_device *dev;
 674         struct inet6_dev *idev;
 675         struct hlist_head *head;
 676
 677         if (cb->strict_check) {
 678                 struct netlink_ext_ack *extack = cb->extack;
 679                 struct netconfmsg *ncm;
 680
 681                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 682                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 683                         return -EINVAL;
 684                 }
 685
 686                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 687                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 688                         return -EINVAL;
 689                 }
 690         }
 691
 692         s_h = cb->args[0];
 693         s_idx = idx = cb->args[1];
 694
 695         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 696                 idx = 0;
 697                 head = &net->dev_index_head[h];
 698                 rcu_read_lock();
 699                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 700                           net->dev_base_seq;
 701                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 702                         if (idx < s_idx)
 703                                 goto cont;
 704                         idev = __in6_dev_get(dev);
 705                         if (!idev)
 706                                 goto cont;
 707
 708                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 709                                                        &idev->cnf,
 710                                                        NETLINK_CB(cb->skb).portid,
 711                                                        nlh->nlmsg_seq,
 712                                                        RTM_NEWNETCONF,
 713                                                        NLM_F_MULTI,
 714                                                        NETCONFA_ALL) < 0) {
 715                                 rcu_read_unlock();
 716                                 goto done;
 717                         }
 718                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 719 cont:
 720                         idx++;
 721                 }
 722                 rcu_read_unlock();
 723         }
 724         if (h == NETDEV_HASHENTRIES) {
 725                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 726                                                net->ipv6.devconf_all,
 727                                                NETLINK_CB(cb->skb).portid,
 728                                                nlh->nlmsg_seq,
 729                                                RTM_NEWNETCONF, NLM_F_MULTI,
 730                                                NETCONFA_ALL) < 0)
 731                         goto done;
 732                 else
 733                         h++;
 734         }
 735         if (h == NETDEV_HASHENTRIES + 1) {
 736                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 737                                                net->ipv6.devconf_dflt,
 738                                                NETLINK_CB(cb->skb).portid,
 739                                                nlh->nlmsg_seq,
 740                                                RTM_NEWNETCONF, NLM_F_MULTI,
 741                                                NETCONFA_ALL) < 0)
 742                         goto done;
 743                 else
 744                         h++;
 745         }
 746 done:
 747         cb->args[0] = h;
 748         cb->args[1] = idx;
 749
 750         return skb->len;
 751 }
 752
 753 #ifdef CONFIG_SYSCTL
 754 static void dev_forward_change(struct inet6_dev *idev)
 755 {
 756         struct net_device *dev;
 757         struct inet6_ifaddr *ifa;
 758
 759         if (!idev)
 760                 return;
 761         dev = idev->dev;
 762         if (idev->cnf.forwarding)
 763                 dev_disable_lro(dev);
 764         if (dev->flags & IFF_MULTICAST) {
 765                 if (idev->cnf.forwarding) {
 766                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 767                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 768                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 769                 } else {
 770                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 771                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 772                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 773                 }
 774         }
 775
 776         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 777                 if (ifa->flags&IFA_F_TENTATIVE)
 778                         continue;
 779                 if (idev->cnf.forwarding)
 780                         addrconf_join_anycast(ifa);
 781                 else
 782                         addrconf_leave_anycast(ifa);
 783         }
 784         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 785                                      NETCONFA_FORWARDING,
 786                                      dev->ifindex, &idev->cnf);
 787 }
 788
 789
 790 static void addrconf_forward_change(struct net *net, __s32 newf)
 791 {
 792         struct net_device *dev;
 793         struct inet6_dev *idev;
 794
 795         for_each_netdev(net, dev) {
 796                 idev = __in6_dev_get(dev);
 797                 if (idev) {
 798                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 799                         idev->cnf.forwarding = newf;
 800                         if (changed)
 801                                 dev_forward_change(idev);
 802                 }
 803         }
 804 }
 805
 806 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 807 {
 808         struct net *net;
 809         int old;
 810
 811         if (!rtnl_trylock())
 812                 return restart_syscall();
 813
 814         net = (struct net *)table->extra2;
 815         old = *p;
 816         *p = newf;
 817
 818         if (p == &net->ipv6.devconf_dflt->forwarding) {
 819                 if ((!newf) ^ (!old))
 820                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 821                                                      NETCONFA_FORWARDING,
 822                                                      NETCONFA_IFINDEX_DEFAULT,
 823                                                      net->ipv6.devconf_dflt);
 824                 rtnl_unlock();
 825                 return 0;
 826         }
 827
 828         if (p == &net->ipv6.devconf_all->forwarding) {
 829                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 830
 831                 net->ipv6.devconf_dflt->forwarding = newf;
 832                 if ((!newf) ^ (!old_dflt))
 833                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 834                                                      NETCONFA_FORWARDING,
 835                                                      NETCONFA_IFINDEX_DEFAULT,
 836                                                      net->ipv6.devconf_dflt);
 837
 838                 addrconf_forward_change(net, newf);
 839                 if ((!newf) ^ (!old))
 840                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 841                                                      NETCONFA_FORWARDING,
 842                                                      NETCONFA_IFINDEX_ALL,
 843                                                      net->ipv6.devconf_all);
 844         } else if ((!newf) ^ (!old))
 845                 dev_forward_change((struct inet6_dev *)table->extra1);
 846         rtnl_unlock();
 847
 848         if (newf)
 849                 rt6_purge_dflt_routers(net);
 850         return 1;
 851 }
 852
 853 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 854 {
 855         struct net_device *dev;
 856         struct inet6_dev *idev;
 857
 858         for_each_netdev(net, dev) {
 859                 idev = __in6_dev_get(dev);
 860                 if (idev) {
 861                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 862
 863                         idev->cnf.ignore_routes_with_linkdown = newf;
 864                         if (changed)
 865                                 inet6_netconf_notify_devconf(dev_net(dev),
 866                                                              RTM_NEWNETCONF,
 867                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 868                                                              dev->ifindex,
 869                                                              &idev->cnf);
 870                 }
 871         }
 872 }
 873
 874 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 875 {
 876         struct net *net;
 877         int old;
 878
 879         if (!rtnl_trylock())
 880                 return restart_syscall();
 881
 882         net = (struct net *)table->extra2;
 883         old = *p;
 884         *p = newf;
 885
 886         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 887                 if ((!newf) ^ (!old))
 888                         inet6_netconf_notify_devconf(net,
 889                                                      RTM_NEWNETCONF,
 890                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 891                                                      NETCONFA_IFINDEX_DEFAULT,
 892                                                      net->ipv6.devconf_dflt);
 893                 rtnl_unlock();
 894                 return 0;
 895         }
 896
 897         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 898                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 899                 addrconf_linkdown_change(net, newf);
 900                 if ((!newf) ^ (!old))
 901                         inet6_netconf_notify_devconf(net,
 902                                                      RTM_NEWNETCONF,
 903                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 904                                                      NETCONFA_IFINDEX_ALL,
 905                                                      net->ipv6.devconf_all);
 906         }
 907         rtnl_unlock();
 908
 909         return 1;
 910 }
 911
 912 #endif
 913
 914 /* Nobody refers to this ifaddr, destroy it */
 915 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 916 {
 917         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 918
 919 #ifdef NET_REFCNT_DEBUG
 920         pr_debug("%s\n", __func__);
 921 #endif
 922
 923         in6_dev_put(ifp->idev);
 924
 925         if (cancel_delayed_work(&ifp->dad_work))
 926                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 927                           ifp);
 928
 929         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 930                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 931                 return;
 932         }
 933
 934         kfree_rcu(ifp, rcu);
 935 }
 936
 937 static void
 938 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 939 {
 940         struct list_head *p;
 941         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 942
 943         /*
 944          * Each device address list is sorted in order of scope -
 945          * global before linklocal.
 946          */
 947         list_for_each(p, &idev->addr_list) {
 948                 struct inet6_ifaddr *ifa
 949                         = list_entry(p, struct inet6_ifaddr, if_list);
 950                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 951                         break;
 952         }
 953
 954         list_add_tail_rcu(&ifp->if_list, p);
 955 }
 956
 957 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 958 {
 959         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 960
 961         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 962 }
 963
 964 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 965                                struct net_device *dev, unsigned int hash)
 966 {
 967         struct inet6_ifaddr *ifp;
 968
 969         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
 970                 if (!net_eq(dev_net(ifp->idev->dev), net))
 971                         continue;
 972                 if (ipv6_addr_equal(&ifp->addr, addr)) {
 973                         if (!dev || ifp->idev->dev == dev)
 974                                 return true;
 975                 }
 976         }
 977         return false;
 978 }
 979
 980 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
 981 {
 982         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
 983         int err = 0;
 984
 985         spin_lock(&addrconf_hash_lock);
 986
 987         /* Ignore adding duplicate addresses on an interface */
 988         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
 989                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
 990                 err = -EEXIST;
 991         } else {
 992                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 993         }
 994
 995         spin_unlock(&addrconf_hash_lock);
 996
 997         return err;
 998 }
 999
1000 /* On success it returns ifp with increased reference count */
1001
1002 static struct inet6_ifaddr *
1003 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1004               bool can_block, struct netlink_ext_ack *extack)
1005 {
1006         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1007         int addr_type = ipv6_addr_type(cfg->pfx);
1008         struct net *net = dev_net(idev->dev);
1009         struct inet6_ifaddr *ifa = NULL;
1010         struct fib6_info *f6i = NULL;
1011         int err = 0;
1012
1013         if (addr_type == IPV6_ADDR_ANY ||
1014             addr_type & IPV6_ADDR_MULTICAST ||
1015             (!(idev->dev->flags & IFF_LOOPBACK) &&
1016              !netif_is_l3_master(idev->dev) &&
1017              addr_type & IPV6_ADDR_LOOPBACK))
1018                 return ERR_PTR(-EADDRNOTAVAIL);
1019
1020         if (idev->dead) {
1021                 err = -ENODEV;                  /*XXX*/
1022                 goto out;
1023         }
1024
1025         if (idev->cnf.disable_ipv6) {
1026                 err = -EACCES;
1027                 goto out;
1028         }
1029
1030         /* validator notifier needs to be blocking;
1031          * do not call in atomic context
1032          */
1033         if (can_block) {
1034                 struct in6_validator_info i6vi = {
1035                         .i6vi_addr = *cfg->pfx,
1036                         .i6vi_dev = idev,
1037                         .extack = extack,
1038                 };
1039
1040                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1041                 err = notifier_to_errno(err);
1042                 if (err < 0)
1043                         goto out;
1044         }
1045
1046         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1047         if (!ifa) {
1048                 err = -ENOBUFS;
1049                 goto out;
1050         }
1051
1052         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1053         if (IS_ERR(f6i)) {
1054                 err = PTR_ERR(f6i);
1055                 f6i = NULL;
1056                 goto out;
1057         }
1058
1059         if (net->ipv6.devconf_all->disable_policy ||
1060             idev->cnf.disable_policy)
1061                 f6i->dst_nopolicy = true;
1062
1063         neigh_parms_data_state_setall(idev->nd_parms);
1064
1065         ifa->addr = *cfg->pfx;
1066         if (cfg->peer_pfx)
1067                 ifa->peer_addr = *cfg->peer_pfx;
1068
1069         spin_lock_init(&ifa->lock);
1070         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1071         INIT_HLIST_NODE(&ifa->addr_lst);
1072         ifa->scope = cfg->scope;
1073         ifa->prefix_len = cfg->plen;
1074         ifa->rt_priority = cfg->rt_priority;
1075         ifa->flags = cfg->ifa_flags;
1076         /* No need to add the TENTATIVE flag for addresses with NODAD */
1077         if (!(cfg->ifa_flags & IFA_F_NODAD))
1078                 ifa->flags |= IFA_F_TENTATIVE;
1079         ifa->valid_lft = cfg->valid_lft;
1080         ifa->prefered_lft = cfg->preferred_lft;
1081         ifa->cstamp = ifa->tstamp = jiffies;
1082         ifa->tokenized = false;
1083
1084         ifa->rt = f6i;
1085
1086         ifa->idev = idev;
1087         in6_dev_hold(idev);
1088
1089         /* For caller */
1090         refcount_set(&ifa->refcnt, 1);
1091
1092         rcu_read_lock_bh();
1093
1094         err = ipv6_add_addr_hash(idev->dev, ifa);
1095         if (err < 0) {
1096                 rcu_read_unlock_bh();
1097                 goto out;
1098         }
1099
1100         write_lock(&idev->lock);
1101
1102         /* Add to inet6_dev unicast addr list. */
1103         ipv6_link_dev_addr(idev, ifa);
1104
1105         if (ifa->flags&IFA_F_TEMPORARY) {
1106                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1107                 in6_ifa_hold(ifa);
1108         }
1109
1110         in6_ifa_hold(ifa);
1111         write_unlock(&idev->lock);
1112
1113         rcu_read_unlock_bh();
1114
1115         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1116 out:
1117         if (unlikely(err < 0)) {
1118                 fib6_info_release(f6i);
1119
1120                 if (ifa) {
1121                         if (ifa->idev)
1122                                 in6_dev_put(ifa->idev);
1123                         kfree(ifa);
1124                 }
1125                 ifa = ERR_PTR(err);
1126         }
1127
1128         return ifa;
1129 }
1130
1131 enum cleanup_prefix_rt_t {
1132         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1133         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1134         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1135 };
1136
1137 /*
1138  * Check, whether the prefix for ifp would still need a prefix route
1139  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1140  * constants.
1141  *
1142  * 1) we don't purge prefix if address was not permanent.
1143  *    prefix is managed by its own lifetime.
1144  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1145  * 3) if there are no addresses, delete prefix.
1146  * 4) if there are still other permanent address(es),
1147  *    corresponding prefix is still permanent.
1148  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1149  *    don't purge the prefix, assume user space is managing it.
1150  * 6) otherwise, update prefix lifetime to the
1151  *    longest valid lifetime among the corresponding
1152  *    addresses on the device.
1153  *    Note: subsequent RA will update lifetime.
1154  **/
1155 static enum cleanup_prefix_rt_t
1156 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1157 {
1158         struct inet6_ifaddr *ifa;
1159         struct inet6_dev *idev = ifp->idev;
1160         unsigned long lifetime;
1161         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1162
1163         *expires = jiffies;
1164
1165         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1166                 if (ifa == ifp)
1167                         continue;
1168                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1169                                        ifp->prefix_len))
1170                         continue;
1171                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1172                         return CLEANUP_PREFIX_RT_NOP;
1173
1174                 action = CLEANUP_PREFIX_RT_EXPIRE;
1175
1176                 spin_lock(&ifa->lock);
1177
1178                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1179                 /*
1180                  * Note: Because this address is
1181                  * not permanent, lifetime <
1182                  * LONG_MAX / HZ here.
1183                  */
1184                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1185                         *expires = ifa->tstamp + lifetime * HZ;
1186                 spin_unlock(&ifa->lock);
1187         }
1188
1189         return action;
1190 }
1191
1192 static void
1193 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1194 {
1195         struct fib6_info *f6i;
1196
1197         f6i = addrconf_get_prefix_route(&ifp->addr,
1198                                        ifp->prefix_len,
1199                                        ifp->idev->dev,
1200                                        0, RTF_GATEWAY | RTF_DEFAULT);
1201         if (f6i) {
1202                 if (del_rt)
1203                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1204                 else {
1205                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1206                                 fib6_set_expires(f6i, expires);
1207                         fib6_info_release(f6i);
1208                 }
1209         }
1210 }
1211
1212
1213 /* This function wants to get referenced ifp and releases it before return */
1214
1215 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1216 {
1217         int state;
1218         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1219         unsigned long expires;
1220
1221         ASSERT_RTNL();
1222
1223         spin_lock_bh(&ifp->lock);
1224         state = ifp->state;
1225         ifp->state = INET6_IFADDR_STATE_DEAD;
1226         spin_unlock_bh(&ifp->lock);
1227
1228         if (state == INET6_IFADDR_STATE_DEAD)
1229                 goto out;
1230
1231         spin_lock_bh(&addrconf_hash_lock);
1232         hlist_del_init_rcu(&ifp->addr_lst);
1233         spin_unlock_bh(&addrconf_hash_lock);
1234
1235         write_lock_bh(&ifp->idev->lock);
1236
1237         if (ifp->flags&IFA_F_TEMPORARY) {
1238                 list_del(&ifp->tmp_list);
1239                 if (ifp->ifpub) {
1240                         in6_ifa_put(ifp->ifpub);
1241                         ifp->ifpub = NULL;
1242                 }
1243                 __in6_ifa_put(ifp);
1244         }
1245
1246         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1247                 action = check_cleanup_prefix_route(ifp, &expires);
1248
1249         list_del_rcu(&ifp->if_list);
1250         __in6_ifa_put(ifp);
1251
1252         write_unlock_bh(&ifp->idev->lock);
1253
1254         addrconf_del_dad_work(ifp);
1255
1256         ipv6_ifa_notify(RTM_DELADDR, ifp);
1257
1258         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1259
1260         if (action != CLEANUP_PREFIX_RT_NOP) {
1261                 cleanup_prefix_route(ifp, expires,
1262                         action == CLEANUP_PREFIX_RT_DEL);
1263         }
1264
1265         /* clean up prefsrc entries */
1266         rt6_remove_prefsrc(ifp);
1267 out:
1268         in6_ifa_put(ifp);
1269 }
1270
1271 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1272                                 struct inet6_ifaddr *ift,
1273                                 bool block)
1274 {
1275         struct inet6_dev *idev = ifp->idev;
1276         struct in6_addr addr, *tmpaddr;
1277         unsigned long tmp_tstamp, age;
1278         unsigned long regen_advance;
1279         struct ifa6_config cfg;
1280         int ret = 0;
1281         unsigned long now = jiffies;
1282         long max_desync_factor;
1283         s32 cnf_temp_preferred_lft;
1284
1285         write_lock_bh(&idev->lock);
1286         if (ift) {
1287                 spin_lock_bh(&ift->lock);
1288                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1289                 spin_unlock_bh(&ift->lock);
1290                 tmpaddr = &addr;
1291         } else {
1292                 tmpaddr = NULL;
1293         }
1294 retry:
1295         in6_dev_hold(idev);
1296         if (idev->cnf.use_tempaddr <= 0) {
1297                 write_unlock_bh(&idev->lock);
1298                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1299                 in6_dev_put(idev);
1300                 ret = -1;
1301                 goto out;
1302         }
1303         spin_lock_bh(&ifp->lock);
1304         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1305                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1306                 spin_unlock_bh(&ifp->lock);
1307                 write_unlock_bh(&idev->lock);
1308                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1309                         __func__);
1310                 in6_dev_put(idev);
1311                 ret = -1;
1312                 goto out;
1313         }
1314         in6_ifa_hold(ifp);
1315         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1316         ipv6_try_regen_rndid(idev, tmpaddr);
1317         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1318         age = (now - ifp->tstamp) / HZ;
1319
1320         regen_advance = idev->cnf.regen_max_retry *
1321                         idev->cnf.dad_transmits *
1322                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1323
1324         /* recalculate max_desync_factor each time and update
1325          * idev->desync_factor if it's larger
1326          */
1327         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1328         max_desync_factor = min_t(__u32,
1329                                   idev->cnf.max_desync_factor,
1330                                   cnf_temp_preferred_lft - regen_advance);
1331
1332         if (unlikely(idev->desync_factor > max_desync_factor)) {
1333                 if (max_desync_factor > 0) {
1334                         get_random_bytes(&idev->desync_factor,
1335                                          sizeof(idev->desync_factor));
1336                         idev->desync_factor %= max_desync_factor;
1337                 } else {
1338                         idev->desync_factor = 0;
1339                 }
1340         }
1341
1342         memset(&cfg, 0, sizeof(cfg));
1343         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1344                               idev->cnf.temp_valid_lft + age);
1345         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1346         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1347
1348         cfg.plen = ifp->prefix_len;
1349         tmp_tstamp = ifp->tstamp;
1350         spin_unlock_bh(&ifp->lock);
1351
1352         write_unlock_bh(&idev->lock);
1353
1354         /* A temporary address is created only if this calculated Preferred
1355          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1356          * an implementation must not create a temporary address with a zero
1357          * Preferred Lifetime.
1358          * Use age calculation as in addrconf_verify to avoid unnecessary
1359          * temporary addresses being generated.
1360          */
1361         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1362         if (cfg.preferred_lft <= regen_advance + age) {
1363                 in6_ifa_put(ifp);
1364                 in6_dev_put(idev);
1365                 ret = -1;
1366                 goto out;
1367         }
1368
1369         cfg.ifa_flags = IFA_F_TEMPORARY;
1370         /* set in addrconf_prefix_rcv() */
1371         if (ifp->flags & IFA_F_OPTIMISTIC)
1372                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1373
1374         cfg.pfx = &addr;
1375         cfg.scope = ipv6_addr_scope(cfg.pfx);
1376
1377         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1378         if (IS_ERR(ift)) {
1379                 in6_ifa_put(ifp);
1380                 in6_dev_put(idev);
1381                 pr_info("%s: retry temporary address regeneration\n", __func__);
1382                 tmpaddr = &addr;
1383                 write_lock_bh(&idev->lock);
1384                 goto retry;
1385         }
1386
1387         spin_lock_bh(&ift->lock);
1388         ift->ifpub = ifp;
1389         ift->cstamp = now;
1390         ift->tstamp = tmp_tstamp;
1391         spin_unlock_bh(&ift->lock);
1392
1393         addrconf_dad_start(ift);
1394         in6_ifa_put(ift);
1395         in6_dev_put(idev);
1396 out:
1397         return ret;
1398 }
1399
1400 /*
1401  *      Choose an appropriate source address (RFC3484)
1402  */
1403 enum {
1404         IPV6_SADDR_RULE_INIT = 0,
1405         IPV6_SADDR_RULE_LOCAL,
1406         IPV6_SADDR_RULE_SCOPE,
1407         IPV6_SADDR_RULE_PREFERRED,
1408 #ifdef CONFIG_IPV6_MIP6
1409         IPV6_SADDR_RULE_HOA,
1410 #endif
1411         IPV6_SADDR_RULE_OIF,
1412         IPV6_SADDR_RULE_LABEL,
1413         IPV6_SADDR_RULE_PRIVACY,
1414         IPV6_SADDR_RULE_ORCHID,
1415         IPV6_SADDR_RULE_PREFIX,
1416 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1417         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1418 #endif
1419         IPV6_SADDR_RULE_MAX
1420 };
1421
1422 struct ipv6_saddr_score {
1423         int                     rule;
1424         int                     addr_type;
1425         struct inet6_ifaddr     *ifa;
1426         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1427         int                     scopedist;
1428         int                     matchlen;
1429 };
1430
1431 struct ipv6_saddr_dst {
1432         const struct in6_addr *addr;
1433         int ifindex;
1434         int scope;
1435         int label;
1436         unsigned int prefs;
1437 };
1438
1439 static inline int ipv6_saddr_preferred(int type)
1440 {
1441         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1442                 return 1;
1443         return 0;
1444 }
1445
1446 static bool ipv6_use_optimistic_addr(struct net *net,
1447                                      struct inet6_dev *idev)
1448 {
1449 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1450         if (!idev)
1451                 return false;
1452         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1453                 return false;
1454         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1455                 return false;
1456
1457         return true;
1458 #else
1459         return false;
1460 #endif
1461 }
1462
1463 static bool ipv6_allow_optimistic_dad(struct net *net,
1464                                       struct inet6_dev *idev)
1465 {
1466 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1467         if (!idev)
1468                 return false;
1469         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1470                 return false;
1471
1472         return true;
1473 #else
1474         return false;
1475 #endif
1476 }
1477
1478 static int ipv6_get_saddr_eval(struct net *net,
1479                                struct ipv6_saddr_score *score,
1480                                struct ipv6_saddr_dst *dst,
1481                                int i)
1482 {
1483         int ret;
1484
1485         if (i <= score->rule) {
1486                 switch (i) {
1487                 case IPV6_SADDR_RULE_SCOPE:
1488                         ret = score->scopedist;
1489                         break;
1490                 case IPV6_SADDR_RULE_PREFIX:
1491                         ret = score->matchlen;
1492                         break;
1493                 default:
1494                         ret = !!test_bit(i, score->scorebits);
1495                 }
1496                 goto out;
1497         }
1498
1499         switch (i) {
1500         case IPV6_SADDR_RULE_INIT:
1501                 /* Rule 0: remember if hiscore is not ready yet */
1502                 ret = !!score->ifa;
1503                 break;
1504         case IPV6_SADDR_RULE_LOCAL:
1505                 /* Rule 1: Prefer same address */
1506                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1507                 break;
1508         case IPV6_SADDR_RULE_SCOPE:
1509                 /* Rule 2: Prefer appropriate scope
1510                  *
1511                  *      ret
1512                  *       ^
1513                  *    -1 |  d 15
1514                  *    ---+--+-+---> scope
1515                  *       |
1516                  *       |             d is scope of the destination.
1517                  *  B-d  |  \
1518                  *       |   \      <- smaller scope is better if
1519                  *  B-15 |    \        if scope is enough for destination.
1520                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1521                  * d-C-1 | /
1522                  *       |/         <- greater is better
1523                  *   -C  /             if scope is not enough for destination.
1524                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1525                  *
1526                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1527                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1528                  * Assume B = 0 and we get C > 29.
1529                  */
1530                 ret = __ipv6_addr_src_scope(score->addr_type);
1531                 if (ret >= dst->scope)
1532                         ret = -ret;
1533                 else
1534                         ret -= 128;     /* 30 is enough */
1535                 score->scopedist = ret;
1536                 break;
1537         case IPV6_SADDR_RULE_PREFERRED:
1538             {
1539                 /* Rule 3: Avoid deprecated and optimistic addresses */
1540                 u8 avoid = IFA_F_DEPRECATED;
1541
1542                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1543                         avoid |= IFA_F_OPTIMISTIC;
1544                 ret = ipv6_saddr_preferred(score->addr_type) ||
1545                       !(score->ifa->flags & avoid);
1546                 break;
1547             }
1548 #ifdef CONFIG_IPV6_MIP6
1549         case IPV6_SADDR_RULE_HOA:
1550             {
1551                 /* Rule 4: Prefer home address */
1552                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1553                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1554                 break;
1555             }
1556 #endif
1557         case IPV6_SADDR_RULE_OIF:
1558                 /* Rule 5: Prefer outgoing interface */
1559                 ret = (!dst->ifindex ||
1560                        dst->ifindex == score->ifa->idev->dev->ifindex);
1561                 break;
1562         case IPV6_SADDR_RULE_LABEL:
1563                 /* Rule 6: Prefer matching label */
1564                 ret = ipv6_addr_label(net,
1565                                       &score->ifa->addr, score->addr_type,
1566                                       score->ifa->idev->dev->ifindex) == dst->label;
1567                 break;
1568         case IPV6_SADDR_RULE_PRIVACY:
1569             {
1570                 /* Rule 7: Prefer public address
1571                  * Note: prefer temporary address if use_tempaddr >= 2
1572                  */
1573                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1574                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1575                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1576                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1577                 break;
1578             }
1579         case IPV6_SADDR_RULE_ORCHID:
1580                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1581                  *          non-ORCHID vs non-ORCHID
1582                  */
1583                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1584                         ipv6_addr_orchid(dst->addr));
1585                 break;
1586         case IPV6_SADDR_RULE_PREFIX:
1587                 /* Rule 8: Use longest matching prefix */
1588                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1589                 if (ret > score->ifa->prefix_len)
1590                         ret = score->ifa->prefix_len;
1591                 score->matchlen = ret;
1592                 break;
1593 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1594         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1595                 /* Optimistic addresses still have lower precedence than other
1596                  * preferred addresses.
1597                  */
1598                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1599                 break;
1600 #endif
1601         default:
1602                 ret = 0;
1603         }
1604
1605         if (ret)
1606                 __set_bit(i, score->scorebits);
1607         score->rule = i;
1608 out:
1609         return ret;
1610 }
1611
1612 static int __ipv6_dev_get_saddr(struct net *net,
1613                                 struct ipv6_saddr_dst *dst,
1614                                 struct inet6_dev *idev,
1615                                 struct ipv6_saddr_score *scores,
1616                                 int hiscore_idx)
1617 {
1618         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1619
1620         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1621                 int i;
1622
1623                 /*
1624                  * - Tentative Address (RFC2462 section 5.4)
1625                  *  - A tentative address is not considered
1626                  *    "assigned to an interface" in the traditional
1627                  *    sense, unless it is also flagged as optimistic.
1628                  * - Candidate Source Address (section 4)
1629                  *  - In any case, anycast addresses, multicast
1630                  *    addresses, and the unspecified address MUST
1631                  *    NOT be included in a candidate set.
1632                  */
1633                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1634                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1635                         continue;
1636
1637                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1638
1639                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1640                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1641                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1642                                             idev->dev->name);
1643                         continue;
1644                 }
1645
1646                 score->rule = -1;
1647                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1648
1649                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1650                         int minihiscore, miniscore;
1651
1652                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1653                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1654
1655                         if (minihiscore > miniscore) {
1656                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1657                                     score->scopedist > 0) {
1658                                         /*
1659                                          * special case:
1660                                          * each remaining entry
1661                                          * has too small (not enough)
1662                                          * scope, because ifa entries
1663                                          * are sorted by their scope
1664                                          * values.
1665                                          */
1666                                         goto out;
1667                                 }
1668                                 break;
1669                         } else if (minihiscore < miniscore) {
1670                                 swap(hiscore, score);
1671                                 hiscore_idx = 1 - hiscore_idx;
1672
1673                                 /* restore our iterator */
1674                                 score->ifa = hiscore->ifa;
1675
1676                                 break;
1677                         }
1678                 }
1679         }
1680 out:
1681         return hiscore_idx;
1682 }
1683
1684 static int ipv6_get_saddr_master(struct net *net,
1685                                  const struct net_device *dst_dev,
1686                                  const struct net_device *master,
1687                                  struct ipv6_saddr_dst *dst,
1688                                  struct ipv6_saddr_score *scores,
1689                                  int hiscore_idx)
1690 {
1691         struct inet6_dev *idev;
1692
1693         idev = __in6_dev_get(dst_dev);
1694         if (idev)
1695                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1696                                                    scores, hiscore_idx);
1697
1698         idev = __in6_dev_get(master);
1699         if (idev)
1700                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1701                                                    scores, hiscore_idx);
1702
1703         return hiscore_idx;
1704 }
1705
1706 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1707                        const struct in6_addr *daddr, unsigned int prefs,
1708                        struct in6_addr *saddr)
1709 {
1710         struct ipv6_saddr_score scores[2], *hiscore;
1711         struct ipv6_saddr_dst dst;
1712         struct inet6_dev *idev;
1713         struct net_device *dev;
1714         int dst_type;
1715         bool use_oif_addr = false;
1716         int hiscore_idx = 0;
1717         int ret = 0;
1718
1719         dst_type = __ipv6_addr_type(daddr);
1720         dst.addr = daddr;
1721         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1722         dst.scope = __ipv6_addr_src_scope(dst_type);
1723         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1724         dst.prefs = prefs;
1725
1726         scores[hiscore_idx].rule = -1;
1727         scores[hiscore_idx].ifa = NULL;
1728
1729         rcu_read_lock();
1730
1731         /* Candidate Source Address (section 4)
1732          *  - multicast and link-local destination address,
1733          *    the set of candidate source address MUST only
1734          *    include addresses assigned to interfaces
1735          *    belonging to the same link as the outgoing
1736          *    interface.
1737          * (- For site-local destination addresses, the
1738          *    set of candidate source addresses MUST only
1739          *    include addresses assigned to interfaces
1740          *    belonging to the same site as the outgoing
1741          *    interface.)
1742          *  - "It is RECOMMENDED that the candidate source addresses
1743          *    be the set of unicast addresses assigned to the
1744          *    interface that will be used to send to the destination
1745          *    (the 'outgoing' interface)." (RFC 6724)
1746          */
1747         if (dst_dev) {
1748                 idev = __in6_dev_get(dst_dev);
1749                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1750                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1751                     (idev && idev->cnf.use_oif_addrs_only)) {
1752                         use_oif_addr = true;
1753                 }
1754         }
1755
1756         if (use_oif_addr) {
1757                 if (idev)
1758                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1759         } else {
1760                 const struct net_device *master;
1761                 int master_idx = 0;
1762
1763                 /* if dst_dev exists and is enslaved to an L3 device, then
1764                  * prefer addresses from dst_dev and then the master over
1765                  * any other enslaved devices in the L3 domain.
1766                  */
1767                 master = l3mdev_master_dev_rcu(dst_dev);
1768                 if (master) {
1769                         master_idx = master->ifindex;
1770
1771                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1772                                                             master, &dst,
1773                                                             scores, hiscore_idx);
1774
1775                         if (scores[hiscore_idx].ifa)
1776                                 goto out;
1777                 }
1778
1779                 for_each_netdev_rcu(net, dev) {
1780                         /* only consider addresses on devices in the
1781                          * same L3 domain
1782                          */
1783                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1784                                 continue;
1785                         idev = __in6_dev_get(dev);
1786                         if (!idev)
1787                                 continue;
1788                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1789                 }
1790         }
1791
1792 out:
1793         hiscore = &scores[hiscore_idx];
1794         if (!hiscore->ifa)
1795                 ret = -EADDRNOTAVAIL;
1796         else
1797                 *saddr = hiscore->ifa->addr;
1798
1799         rcu_read_unlock();
1800         return ret;
1801 }
1802 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1803
1804 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1805                       u32 banned_flags)
1806 {
1807         struct inet6_ifaddr *ifp;
1808         int err = -EADDRNOTAVAIL;
1809
1810         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1811                 if (ifp->scope > IFA_LINK)
1812                         break;
1813                 if (ifp->scope == IFA_LINK &&
1814                     !(ifp->flags & banned_flags)) {
1815                         *addr = ifp->addr;
1816                         err = 0;
1817                         break;
1818                 }
1819         }
1820         return err;
1821 }
1822
1823 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1824                     u32 banned_flags)
1825 {
1826         struct inet6_dev *idev;
1827         int err = -EADDRNOTAVAIL;
1828
1829         rcu_read_lock();
1830         idev = __in6_dev_get(dev);
1831         if (idev) {
1832                 read_lock_bh(&idev->lock);
1833                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1834                 read_unlock_bh(&idev->lock);
1835         }
1836         rcu_read_unlock();
1837         return err;
1838 }
1839
1840 static int ipv6_count_addresses(const struct inet6_dev *idev)
1841 {
1842         const struct inet6_ifaddr *ifp;
1843         int cnt = 0;
1844
1845         rcu_read_lock();
1846         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1847                 cnt++;
1848         rcu_read_unlock();
1849         return cnt;
1850 }
1851
1852 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1853                   const struct net_device *dev, int strict)
1854 {
1855         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1856                                        strict, IFA_F_TENTATIVE);
1857 }
1858 EXPORT_SYMBOL(ipv6_chk_addr);
1859
1860 /* device argument is used to find the L3 domain of interest. If
1861  * skip_dev_check is set, then the ifp device is not checked against
1862  * the passed in dev argument. So the 2 cases for addresses checks are:
1863  *   1. does the address exist in the L3 domain that dev is part of
1864  *      (skip_dev_check = true), or
1865  *
1866  *   2. does the address exist on the specific device
1867  *      (skip_dev_check = false)
1868  */
1869 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1870                             const struct net_device *dev, bool skip_dev_check,
1871                             int strict, u32 banned_flags)
1872 {
1873         unsigned int hash = inet6_addr_hash(net, addr);
1874         const struct net_device *l3mdev;
1875         struct inet6_ifaddr *ifp;
1876         u32 ifp_flags;
1877
1878         rcu_read_lock();
1879
1880         l3mdev = l3mdev_master_dev_rcu(dev);
1881         if (skip_dev_check)
1882                 dev = NULL;
1883
1884         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1885                 if (!net_eq(dev_net(ifp->idev->dev), net))
1886                         continue;
1887
1888                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1889                         continue;
1890
1891                 /* Decouple optimistic from tentative for evaluation here.
1892                  * Ban optimistic addresses explicitly, when required.
1893                  */
1894                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1895                             ? (ifp->flags&~IFA_F_TENTATIVE)
1896                             : ifp->flags;
1897                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1898                     !(ifp_flags&banned_flags) &&
1899                     (!dev || ifp->idev->dev == dev ||
1900                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1901                         rcu_read_unlock();
1902                         return 1;
1903                 }
1904         }
1905
1906         rcu_read_unlock();
1907         return 0;
1908 }
1909 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1910
1911
1912 /* Compares an address/prefix_len with addresses on device @dev.
1913  * If one is found it returns true.
1914  */
1915 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1916         const unsigned int prefix_len, struct net_device *dev)
1917 {
1918         const struct inet6_ifaddr *ifa;
1919         const struct inet6_dev *idev;
1920         bool ret = false;
1921
1922         rcu_read_lock();
1923         idev = __in6_dev_get(dev);
1924         if (idev) {
1925                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1926                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1927                         if (ret)
1928                                 break;
1929                 }
1930         }
1931         rcu_read_unlock();
1932
1933         return ret;
1934 }
1935 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1936
1937 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1938 {
1939         const struct inet6_ifaddr *ifa;
1940         const struct inet6_dev *idev;
1941         int     onlink;
1942
1943         onlink = 0;
1944         rcu_read_lock();
1945         idev = __in6_dev_get(dev);
1946         if (idev) {
1947                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1948                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1949                                                    ifa->prefix_len);
1950                         if (onlink)
1951                                 break;
1952                 }
1953         }
1954         rcu_read_unlock();
1955         return onlink;
1956 }
1957 EXPORT_SYMBOL(ipv6_chk_prefix);
1958
1959 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1960                                      struct net_device *dev, int strict)
1961 {
1962         unsigned int hash = inet6_addr_hash(net, addr);
1963         struct inet6_ifaddr *ifp, *result = NULL;
1964
1965         rcu_read_lock();
1966         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1967                 if (!net_eq(dev_net(ifp->idev->dev), net))
1968                         continue;
1969                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1970                         if (!dev || ifp->idev->dev == dev ||
1971                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1972                                 result = ifp;
1973                                 in6_ifa_hold(ifp);
1974                                 break;
1975                         }
1976                 }
1977         }
1978         rcu_read_unlock();
1979
1980         return result;
1981 }
1982
1983 /* Gets referenced address, destroys ifaddr */
1984
1985 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1986 {
1987         if (dad_failed)
1988                 ifp->flags |= IFA_F_DADFAILED;
1989
1990         if (ifp->flags&IFA_F_TEMPORARY) {
1991                 struct inet6_ifaddr *ifpub;
1992                 spin_lock_bh(&ifp->lock);
1993                 ifpub = ifp->ifpub;
1994                 if (ifpub) {
1995                         in6_ifa_hold(ifpub);
1996                         spin_unlock_bh(&ifp->lock);
1997                         ipv6_create_tempaddr(ifpub, ifp, true);
1998                         in6_ifa_put(ifpub);
1999                 } else {
2000                         spin_unlock_bh(&ifp->lock);
2001                 }
2002                 ipv6_del_addr(ifp);
2003         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2004                 spin_lock_bh(&ifp->lock);
2005                 addrconf_del_dad_work(ifp);
2006                 ifp->flags |= IFA_F_TENTATIVE;
2007                 if (dad_failed)
2008                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2009                 spin_unlock_bh(&ifp->lock);
2010                 if (dad_failed)
2011                         ipv6_ifa_notify(0, ifp);
2012                 in6_ifa_put(ifp);
2013         } else {
2014                 ipv6_del_addr(ifp);
2015         }
2016 }
2017
2018 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2019 {
2020         int err = -ENOENT;
2021
2022         spin_lock_bh(&ifp->lock);
2023         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2024                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2025                 err = 0;
2026         }
2027         spin_unlock_bh(&ifp->lock);
2028
2029         return err;
2030 }
2031
2032 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2033 {
2034         struct inet6_dev *idev = ifp->idev;
2035         struct net *net = dev_net(ifp->idev->dev);
2036
2037         if (addrconf_dad_end(ifp)) {
2038                 in6_ifa_put(ifp);
2039                 return;
2040         }
2041
2042         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2043                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2044
2045         spin_lock_bh(&ifp->lock);
2046
2047         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2048                 struct in6_addr new_addr;
2049                 struct inet6_ifaddr *ifp2;
2050                 int retries = ifp->stable_privacy_retry + 1;
2051                 struct ifa6_config cfg = {
2052                         .pfx = &new_addr,
2053                         .plen = ifp->prefix_len,
2054                         .ifa_flags = ifp->flags,
2055                         .valid_lft = ifp->valid_lft,
2056                         .preferred_lft = ifp->prefered_lft,
2057                         .scope = ifp->scope,
2058                 };
2059
2060                 if (retries > net->ipv6.sysctl.idgen_retries) {
2061                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2062                                              ifp->idev->dev->name);
2063                         goto errdad;
2064                 }
2065
2066                 new_addr = ifp->addr;
2067                 if (ipv6_generate_stable_address(&new_addr, retries,
2068                                                  idev))
2069                         goto errdad;
2070
2071                 spin_unlock_bh(&ifp->lock);
2072
2073                 if (idev->cnf.max_addresses &&
2074                     ipv6_count_addresses(idev) >=
2075                     idev->cnf.max_addresses)
2076                         goto lock_errdad;
2077
2078                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2079                                      ifp->idev->dev->name);
2080
2081                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2082                 if (IS_ERR(ifp2))
2083                         goto lock_errdad;
2084
2085                 spin_lock_bh(&ifp2->lock);
2086                 ifp2->stable_privacy_retry = retries;
2087                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2088                 spin_unlock_bh(&ifp2->lock);
2089
2090                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2091                 in6_ifa_put(ifp2);
2092 lock_errdad:
2093                 spin_lock_bh(&ifp->lock);
2094         }
2095
2096 errdad:
2097         /* transition from _POSTDAD to _ERRDAD */
2098         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2099         spin_unlock_bh(&ifp->lock);
2100
2101         addrconf_mod_dad_work(ifp, 0);
2102         in6_ifa_put(ifp);
2103 }
2104
2105 /* Join to solicited addr multicast group.
2106  * caller must hold RTNL */
2107 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2108 {
2109         struct in6_addr maddr;
2110
2111         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2112                 return;
2113
2114         addrconf_addr_solict_mult(addr, &maddr);
2115         ipv6_dev_mc_inc(dev, &maddr);
2116 }
2117
2118 /* caller must hold RTNL */
2119 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2120 {
2121         struct in6_addr maddr;
2122
2123         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2124                 return;
2125
2126         addrconf_addr_solict_mult(addr, &maddr);
2127         __ipv6_dev_mc_dec(idev, &maddr);
2128 }
2129
2130 /* caller must hold RTNL */
2131 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2132 {
2133         struct in6_addr addr;
2134
2135         if (ifp->prefix_len >= 127) /* RFC 6164 */
2136                 return;
2137         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2138         if (ipv6_addr_any(&addr))
2139                 return;
2140         __ipv6_dev_ac_inc(ifp->idev, &addr);
2141 }
2142
2143 /* caller must hold RTNL */
2144 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2145 {
2146         struct in6_addr addr;
2147
2148         if (ifp->prefix_len >= 127) /* RFC 6164 */
2149                 return;
2150         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2151         if (ipv6_addr_any(&addr))
2152                 return;
2153         __ipv6_dev_ac_dec(ifp->idev, &addr);
2154 }
2155
2156 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2157 {
2158         switch (dev->addr_len) {
2159         case ETH_ALEN:
2160                 memcpy(eui, dev->dev_addr, 3);
2161                 eui[3] = 0xFF;
2162                 eui[4] = 0xFE;
2163                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2164                 break;
2165         case EUI64_ADDR_LEN:
2166                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2167                 eui[0] ^= 2;
2168                 break;
2169         default:
2170                 return -1;
2171         }
2172
2173         return 0;
2174 }
2175
2176 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2177 {
2178         union fwnet_hwaddr *ha;
2179
2180         if (dev->addr_len != FWNET_ALEN)
2181                 return -1;
2182
2183         ha = (union fwnet_hwaddr *)dev->dev_addr;
2184
2185         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2186         eui[0] ^= 2;
2187         return 0;
2188 }
2189
2190 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2191 {
2192         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2193         if (dev->addr_len != ARCNET_ALEN)
2194                 return -1;
2195         memset(eui, 0, 7);
2196         eui[7] = *(u8 *)dev->dev_addr;
2197         return 0;
2198 }
2199
2200 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2201 {
2202         if (dev->addr_len != INFINIBAND_ALEN)
2203                 return -1;
2204         memcpy(eui, dev->dev_addr + 12, 8);
2205         eui[0] |= 2;
2206         return 0;
2207 }
2208
2209 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2210 {
2211         if (addr == 0)
2212                 return -1;
2213         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2214                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2215                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2216                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2217                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2218                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2219         eui[1] = 0;
2220         eui[2] = 0x5E;
2221         eui[3] = 0xFE;
2222         memcpy(eui + 4, &addr, 4);
2223         return 0;
2224 }
2225
2226 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2227 {
2228         if (dev->priv_flags & IFF_ISATAP)
2229                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2230         return -1;
2231 }
2232
2233 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2234 {
2235         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2236 }
2237
2238 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2239 {
2240         memcpy(eui, dev->perm_addr, 3);
2241         memcpy(eui + 5, dev->perm_addr + 3, 3);
2242         eui[3] = 0xFF;
2243         eui[4] = 0xFE;
2244         eui[0] ^= 2;
2245         return 0;
2246 }
2247
2248 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2249 {
2250         switch (dev->type) {
2251         case ARPHRD_ETHER:
2252         case ARPHRD_FDDI:
2253                 return addrconf_ifid_eui48(eui, dev);
2254         case ARPHRD_ARCNET:
2255                 return addrconf_ifid_arcnet(eui, dev);
2256         case ARPHRD_INFINIBAND:
2257                 return addrconf_ifid_infiniband(eui, dev);
2258         case ARPHRD_SIT:
2259                 return addrconf_ifid_sit(eui, dev);
2260         case ARPHRD_IPGRE:
2261         case ARPHRD_TUNNEL:
2262                 return addrconf_ifid_gre(eui, dev);
2263         case ARPHRD_6LOWPAN:
2264                 return addrconf_ifid_6lowpan(eui, dev);
2265         case ARPHRD_IEEE1394:
2266                 return addrconf_ifid_ieee1394(eui, dev);
2267         case ARPHRD_TUNNEL6:
2268         case ARPHRD_IP6GRE:
2269         case ARPHRD_RAWIP:
2270                 return addrconf_ifid_ip6tnl(eui, dev);
2271         }
2272         return -1;
2273 }
2274
2275 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2276 {
2277         int err = -1;
2278         struct inet6_ifaddr *ifp;
2279
2280         read_lock_bh(&idev->lock);
2281         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2282                 if (ifp->scope > IFA_LINK)
2283                         break;
2284                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2285                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2286                         err = 0;
2287                         break;
2288                 }
2289         }
2290         read_unlock_bh(&idev->lock);
2291         return err;
2292 }
2293
2294 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2295 static void ipv6_regen_rndid(struct inet6_dev *idev)
2296 {
2297 regen:
2298         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2299         idev->rndid[0] &= ~0x02;
2300
2301         /*
2302          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2303          * check if generated address is not inappropriate
2304          *
2305          *  - Reserved subnet anycast (RFC 2526)
2306          *      11111101 11....11 1xxxxxxx
2307          *  - ISATAP (RFC4214) 6.1
2308          *      00-00-5E-FE-xx-xx-xx-xx
2309          *  - value 0
2310          *  - XXX: already assigned to an address on the device
2311          */
2312         if (idev->rndid[0] == 0xfd &&
2313             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2314             (idev->rndid[7]&0x80))
2315                 goto regen;
2316         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2317                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2318                         goto regen;
2319                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2320                         goto regen;
2321         }
2322 }
2323
2324 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2325 {
2326         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2327                 ipv6_regen_rndid(idev);
2328 }
2329
2330 /*
2331  *      Add prefix route.
2332  */
2333
2334 static void
2335 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2336                       struct net_device *dev, unsigned long expires,
2337                       u32 flags, gfp_t gfp_flags)
2338 {
2339         struct fib6_config cfg = {
2340                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2341                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2342                 .fc_ifindex = dev->ifindex,
2343                 .fc_expires = expires,
2344                 .fc_dst_len = plen,
2345                 .fc_flags = RTF_UP | flags,
2346                 .fc_nlinfo.nl_net = dev_net(dev),
2347                 .fc_protocol = RTPROT_KERNEL,
2348                 .fc_type = RTN_UNICAST,
2349         };
2350
2351         cfg.fc_dst = *pfx;
2352
2353         /* Prevent useless cloning on PtP SIT.
2354            This thing is done here expecting that the whole
2355            class of non-broadcast devices need not cloning.
2356          */
2357 #if IS_ENABLED(CONFIG_IPV6_SIT)
2358         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2359                 cfg.fc_flags |= RTF_NONEXTHOP;
2360 #endif
2361
2362         ip6_route_add(&cfg, gfp_flags, NULL);
2363 }
2364
2365
2366 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2367                                                   int plen,
2368                                                   const struct net_device *dev,
2369                                                   u32 flags, u32 noflags)
2370 {
2371         struct fib6_node *fn;
2372         struct fib6_info *rt = NULL;
2373         struct fib6_table *table;
2374         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2375
2376         table = fib6_get_table(dev_net(dev), tb_id);
2377         if (!table)
2378                 return NULL;
2379
2380         rcu_read_lock();
2381         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2382         if (!fn)
2383                 goto out;
2384
2385         for_each_fib6_node_rt_rcu(fn) {
2386                 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2387                         continue;
2388                 if ((rt->fib6_flags & flags) != flags)
2389                         continue;
2390                 if ((rt->fib6_flags & noflags) != 0)
2391                         continue;
2392                 if (!fib6_info_hold_safe(rt))
2393                         continue;
2394                 break;
2395         }
2396 out:
2397         rcu_read_unlock();
2398         return rt;
2399 }
2400
2401
2402 /* Create "default" multicast route to the interface */
2403
2404 static void addrconf_add_mroute(struct net_device *dev)
2405 {
2406         struct fib6_config cfg = {
2407                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2408                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2409                 .fc_ifindex = dev->ifindex,
2410                 .fc_dst_len = 8,
2411                 .fc_flags = RTF_UP,
2412                 .fc_type = RTN_UNICAST,
2413                 .fc_nlinfo.nl_net = dev_net(dev),
2414         };
2415
2416         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2417
2418         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2419 }
2420
2421 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2422 {
2423         struct inet6_dev *idev;
2424
2425         ASSERT_RTNL();
2426
2427         idev = ipv6_find_idev(dev);
2428         if (!idev)
2429                 return ERR_PTR(-ENOBUFS);
2430
2431         if (idev->cnf.disable_ipv6)
2432                 return ERR_PTR(-EACCES);
2433
2434         /* Add default multicast route */
2435         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2436                 addrconf_add_mroute(dev);
2437
2438         return idev;
2439 }
2440
2441 static void manage_tempaddrs(struct inet6_dev *idev,
2442                              struct inet6_ifaddr *ifp,
2443                              __u32 valid_lft, __u32 prefered_lft,
2444                              bool create, unsigned long now)
2445 {
2446         u32 flags;
2447         struct inet6_ifaddr *ift;
2448
2449         read_lock_bh(&idev->lock);
2450         /* update all temporary addresses in the list */
2451         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2452                 int age, max_valid, max_prefered;
2453
2454                 if (ifp != ift->ifpub)
2455                         continue;
2456
2457                 /* RFC 4941 section 3.3:
2458                  * If a received option will extend the lifetime of a public
2459                  * address, the lifetimes of temporary addresses should
2460                  * be extended, subject to the overall constraint that no
2461                  * temporary addresses should ever remain "valid" or "preferred"
2462                  * for a time longer than (TEMP_VALID_LIFETIME) or
2463                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2464                  */
2465                 age = (now - ift->cstamp) / HZ;
2466                 max_valid = idev->cnf.temp_valid_lft - age;
2467                 if (max_valid < 0)
2468                         max_valid = 0;
2469
2470                 max_prefered = idev->cnf.temp_prefered_lft -
2471                                idev->desync_factor - age;
2472                 if (max_prefered < 0)
2473                         max_prefered = 0;
2474
2475                 if (valid_lft > max_valid)
2476                         valid_lft = max_valid;
2477
2478                 if (prefered_lft > max_prefered)
2479                         prefered_lft = max_prefered;
2480
2481                 spin_lock(&ift->lock);
2482                 flags = ift->flags;
2483                 ift->valid_lft = valid_lft;
2484                 ift->prefered_lft = prefered_lft;
2485                 ift->tstamp = now;
2486                 if (prefered_lft > 0)
2487                         ift->flags &= ~IFA_F_DEPRECATED;
2488
2489                 spin_unlock(&ift->lock);
2490                 if (!(flags&IFA_F_TENTATIVE))
2491                         ipv6_ifa_notify(0, ift);
2492         }
2493
2494         if ((create || list_empty(&idev->tempaddr_list)) &&
2495             idev->cnf.use_tempaddr > 0) {
2496                 /* When a new public address is created as described
2497                  * in [ADDRCONF], also create a new temporary address.
2498                  * Also create a temporary address if it's enabled but
2499                  * no temporary address currently exists.
2500                  */
2501                 read_unlock_bh(&idev->lock);
2502                 ipv6_create_tempaddr(ifp, NULL, false);
2503         } else {
2504                 read_unlock_bh(&idev->lock);
2505         }
2506 }
2507
2508 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2509 {
2510         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2511                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2512 }
2513
2514 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2515                                  const struct prefix_info *pinfo,
2516                                  struct inet6_dev *in6_dev,
2517                                  const struct in6_addr *addr, int addr_type,
2518                                  u32 addr_flags, bool sllao, bool tokenized,
2519                                  __u32 valid_lft, u32 prefered_lft)
2520 {
2521         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2522         int create = 0, update_lft = 0;
2523
2524         if (!ifp && valid_lft) {
2525                 int max_addresses = in6_dev->cnf.max_addresses;
2526                 struct ifa6_config cfg = {
2527                         .pfx = addr,
2528                         .plen = pinfo->prefix_len,
2529                         .ifa_flags = addr_flags,
2530                         .valid_lft = valid_lft,
2531                         .preferred_lft = prefered_lft,
2532                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2533                 };
2534
2535 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2536                 if ((net->ipv6.devconf_all->optimistic_dad ||
2537                      in6_dev->cnf.optimistic_dad) &&
2538                     !net->ipv6.devconf_all->forwarding && sllao)
2539                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2540 #endif
2541
2542                 /* Do not allow to create too much of autoconfigured
2543                  * addresses; this would be too easy way to crash kernel.
2544                  */
2545                 if (!max_addresses ||
2546                     ipv6_count_addresses(in6_dev) < max_addresses)
2547                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2548
2549                 if (IS_ERR_OR_NULL(ifp))
2550                         return -1;
2551
2552                 create = 1;
2553                 spin_lock_bh(&ifp->lock);
2554                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2555                 ifp->cstamp = jiffies;
2556                 ifp->tokenized = tokenized;
2557                 spin_unlock_bh(&ifp->lock);
2558                 addrconf_dad_start(ifp);
2559         }
2560
2561         if (ifp) {
2562                 u32 flags;
2563                 unsigned long now;
2564                 u32 stored_lft;
2565
2566                 /* update lifetime (RFC2462 5.5.3 e) */
2567                 spin_lock_bh(&ifp->lock);
2568                 now = jiffies;
2569                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2570                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2571                 else
2572                         stored_lft = 0;
2573                 if (!create && stored_lft) {
2574                         const u32 minimum_lft = min_t(u32,
2575                                 stored_lft, MIN_VALID_LIFETIME);
2576                         valid_lft = max(valid_lft, minimum_lft);
2577
2578                         /* RFC4862 Section 5.5.3e:
2579                          * "Note that the preferred lifetime of the
2580                          *  corresponding address is always reset to
2581                          *  the Preferred Lifetime in the received
2582                          *  Prefix Information option, regardless of
2583                          *  whether the valid lifetime is also reset or
2584                          *  ignored."
2585                          *
2586                          * So we should always update prefered_lft here.
2587                          */
2588                         update_lft = 1;
2589                 }
2590
2591                 if (update_lft) {
2592                         ifp->valid_lft = valid_lft;
2593                         ifp->prefered_lft = prefered_lft;
2594                         ifp->tstamp = now;
2595                         flags = ifp->flags;
2596                         ifp->flags &= ~IFA_F_DEPRECATED;
2597                         spin_unlock_bh(&ifp->lock);
2598
2599                         if (!(flags&IFA_F_TENTATIVE))
2600                                 ipv6_ifa_notify(0, ifp);
2601                 } else
2602                         spin_unlock_bh(&ifp->lock);
2603
2604                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2605                                  create, now);
2606
2607                 in6_ifa_put(ifp);
2608                 addrconf_verify();
2609         }
2610
2611         return 0;
2612 }
2613 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2614
2615 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2616 {
2617         struct prefix_info *pinfo;
2618         __u32 valid_lft;
2619         __u32 prefered_lft;
2620         int addr_type, err;
2621         u32 addr_flags = 0;
2622         struct inet6_dev *in6_dev;
2623         struct net *net = dev_net(dev);
2624
2625         pinfo = (struct prefix_info *) opt;
2626
2627         if (len < sizeof(struct prefix_info)) {
2628                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2629                 return;
2630         }
2631
2632         /*
2633          *      Validation checks ([ADDRCONF], page 19)
2634          */
2635
2636         addr_type = ipv6_addr_type(&pinfo->prefix);
2637
2638         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2639                 return;
2640
2641         valid_lft = ntohl(pinfo->valid);
2642         prefered_lft = ntohl(pinfo->prefered);
2643
2644         if (prefered_lft > valid_lft) {
2645                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2646                 return;
2647         }
2648
2649         in6_dev = in6_dev_get(dev);
2650
2651         if (!in6_dev) {
2652                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2653                                     dev->name);
2654                 return;
2655         }
2656
2657         /*
2658          *      Two things going on here:
2659          *      1) Add routes for on-link prefixes
2660          *      2) Configure prefixes with the auto flag set
2661          */
2662
2663         if (pinfo->onlink) {
2664                 struct fib6_info *rt;
2665                 unsigned long rt_expires;
2666
2667                 /* Avoid arithmetic overflow. Really, we could
2668                  * save rt_expires in seconds, likely valid_lft,
2669                  * but it would require division in fib gc, that it
2670                  * not good.
2671                  */
2672                 if (HZ > USER_HZ)
2673                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2674                 else
2675                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2676
2677                 if (addrconf_finite_timeout(rt_expires))
2678                         rt_expires *= HZ;
2679
2680                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2681                                                pinfo->prefix_len,
2682                                                dev,
2683                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2684                                                RTF_GATEWAY | RTF_DEFAULT);
2685
2686                 if (rt) {
2687                         /* Autoconf prefix route */
2688                         if (valid_lft == 0) {
2689                                 ip6_del_rt(net, rt);
2690                                 rt = NULL;
2691                         } else if (addrconf_finite_timeout(rt_expires)) {
2692                                 /* not infinity */
2693                                 fib6_set_expires(rt, jiffies + rt_expires);
2694                         } else {
2695                                 fib6_clean_expires(rt);
2696                         }
2697                 } else if (valid_lft) {
2698                         clock_t expires = 0;
2699                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2700                         if (addrconf_finite_timeout(rt_expires)) {
2701                                 /* not infinity */
2702                                 flags |= RTF_EXPIRES;
2703                                 expires = jiffies_to_clock_t(rt_expires);
2704                         }
2705                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2706                                               0, dev, expires, flags,
2707                                               GFP_ATOMIC);
2708                 }
2709                 fib6_info_release(rt);
2710         }
2711
2712         /* Try to figure out our local address for this prefix */
2713
2714         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2715                 struct in6_addr addr;
2716                 bool tokenized = false, dev_addr_generated = false;
2717
2718                 if (pinfo->prefix_len == 64) {
2719                         memcpy(&addr, &pinfo->prefix, 8);
2720
2721                         if (!ipv6_addr_any(&in6_dev->token)) {
2722                                 read_lock_bh(&in6_dev->lock);
2723                                 memcpy(addr.s6_addr + 8,
2724                                        in6_dev->token.s6_addr + 8, 8);
2725                                 read_unlock_bh(&in6_dev->lock);
2726                                 tokenized = true;
2727                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2728                                    !ipv6_generate_stable_address(&addr, 0,
2729                                                                  in6_dev)) {
2730                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2731                                 goto ok;
2732                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2733                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2734                                 goto put;
2735                         } else {
2736                                 dev_addr_generated = true;
2737                         }
2738                         goto ok;
2739                 }
2740                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2741                                     pinfo->prefix_len);
2742                 goto put;
2743
2744 ok:
2745                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2746                                                    &addr, addr_type,
2747                                                    addr_flags, sllao,
2748                                                    tokenized, valid_lft,
2749                                                    prefered_lft);
2750                 if (err)
2751                         goto put;
2752
2753                 /* Ignore error case here because previous prefix add addr was
2754                  * successful which will be notified.
2755                  */
2756                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2757                                               addr_type, addr_flags, sllao,
2758                                               tokenized, valid_lft,
2759                                               prefered_lft,
2760                                               dev_addr_generated);
2761         }
2762         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2763 put:
2764         in6_dev_put(in6_dev);
2765 }
2766
2767 /*
2768  *      Set destination address.
2769  *      Special case for SIT interfaces where we create a new "virtual"
2770  *      device.
2771  */
2772 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2773 {
2774         struct in6_ifreq ireq;
2775         struct net_device *dev;
2776         int err = -EINVAL;
2777
2778         rtnl_lock();
2779
2780         err = -EFAULT;
2781         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2782                 goto err_exit;
2783
2784         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2785
2786         err = -ENODEV;
2787         if (!dev)
2788                 goto err_exit;
2789
2790 #if IS_ENABLED(CONFIG_IPV6_SIT)
2791         if (dev->type == ARPHRD_SIT) {
2792                 const struct net_device_ops *ops = dev->netdev_ops;
2793                 struct ifreq ifr;
2794                 struct ip_tunnel_parm p;
2795
2796                 err = -EADDRNOTAVAIL;
2797                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2798                         goto err_exit;
2799
2800                 memset(&p, 0, sizeof(p));
2801                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2802                 p.iph.saddr = 0;
2803                 p.iph.version = 4;
2804                 p.iph.ihl = 5;
2805                 p.iph.protocol = IPPROTO_IPV6;
2806                 p.iph.ttl = 64;
2807                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2808
2809                 if (ops->ndo_do_ioctl) {
2810                         mm_segment_t oldfs = get_fs();
2811
2812                         set_fs(KERNEL_DS);
2813                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2814                         set_fs(oldfs);
2815                 } else
2816                         err = -EOPNOTSUPP;
2817
2818                 if (err == 0) {
2819                         err = -ENOBUFS;
2820                         dev = __dev_get_by_name(net, p.name);
2821                         if (!dev)
2822                                 goto err_exit;
2823                         err = dev_open(dev);
2824                 }
2825         }
2826 #endif
2827
2828 err_exit:
2829         rtnl_unlock();
2830         return err;
2831 }
2832
2833 static int ipv6_mc_config(struct sock *sk, bool join,
2834                           const struct in6_addr *addr, int ifindex)
2835 {
2836         int ret;
2837
2838         ASSERT_RTNL();
2839
2840         lock_sock(sk);
2841         if (join)
2842                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2843         else
2844                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2845         release_sock(sk);
2846
2847         return ret;
2848 }
2849
2850 /*
2851  *      Manual configuration of address on an interface
2852  */
2853 static int inet6_addr_add(struct net *net, int ifindex,
2854                           struct ifa6_config *cfg,
2855                           struct netlink_ext_ack *extack)
2856 {
2857         struct inet6_ifaddr *ifp;
2858         struct inet6_dev *idev;
2859         struct net_device *dev;
2860         unsigned long timeout;
2861         clock_t expires;
2862         u32 flags;
2863
2864         ASSERT_RTNL();
2865
2866         if (cfg->plen > 128)
2867                 return -EINVAL;
2868
2869         /* check the lifetime */
2870         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2871                 return -EINVAL;
2872
2873         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2874                 return -EINVAL;
2875
2876         dev = __dev_get_by_index(net, ifindex);
2877         if (!dev)
2878                 return -ENODEV;
2879
2880         idev = addrconf_add_dev(dev);
2881         if (IS_ERR(idev))
2882                 return PTR_ERR(idev);
2883
2884         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2885                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2886                                          true, cfg->pfx, ifindex);
2887
2888                 if (ret < 0)
2889                         return ret;
2890         }
2891
2892         cfg->scope = ipv6_addr_scope(cfg->pfx);
2893
2894         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2895         if (addrconf_finite_timeout(timeout)) {
2896                 expires = jiffies_to_clock_t(timeout * HZ);
2897                 cfg->valid_lft = timeout;
2898                 flags = RTF_EXPIRES;
2899         } else {
2900                 expires = 0;
2901                 flags = 0;
2902                 cfg->ifa_flags |= IFA_F_PERMANENT;
2903         }
2904
2905         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2906         if (addrconf_finite_timeout(timeout)) {
2907                 if (timeout == 0)
2908                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2909                 cfg->preferred_lft = timeout;
2910         }
2911
2912         ifp = ipv6_add_addr(idev, cfg, true, extack);
2913         if (!IS_ERR(ifp)) {
2914                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2915                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2916                                               ifp->rt_priority, dev, expires,
2917                                               flags, GFP_KERNEL);
2918                 }
2919
2920                 /* Send a netlink notification if DAD is enabled and
2921                  * optimistic flag is not set
2922                  */
2923                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2924                         ipv6_ifa_notify(0, ifp);
2925                 /*
2926                  * Note that section 3.1 of RFC 4429 indicates
2927                  * that the Optimistic flag should not be set for
2928                  * manually configured addresses
2929                  */
2930                 addrconf_dad_start(ifp);
2931                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2932                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2933                                          cfg->preferred_lft, true, jiffies);
2934                 in6_ifa_put(ifp);
2935                 addrconf_verify_rtnl();
2936                 return 0;
2937         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2938                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2939                                cfg->pfx, ifindex);
2940         }
2941
2942         return PTR_ERR(ifp);
2943 }
2944
2945 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2946                           const struct in6_addr *pfx, unsigned int plen)
2947 {
2948         struct inet6_ifaddr *ifp;
2949         struct inet6_dev *idev;
2950         struct net_device *dev;
2951
2952         if (plen > 128)
2953                 return -EINVAL;
2954
2955         dev = __dev_get_by_index(net, ifindex);
2956         if (!dev)
2957                 return -ENODEV;
2958
2959         idev = __in6_dev_get(dev);
2960         if (!idev)
2961                 return -ENXIO;
2962
2963         read_lock_bh(&idev->lock);
2964         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2965                 if (ifp->prefix_len == plen &&
2966                     ipv6_addr_equal(pfx, &ifp->addr)) {
2967                         in6_ifa_hold(ifp);
2968                         read_unlock_bh(&idev->lock);
2969
2970                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2971                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2972                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2973                                                  jiffies);
2974                         ipv6_del_addr(ifp);
2975                         addrconf_verify_rtnl();
2976                         if (ipv6_addr_is_multicast(pfx)) {
2977                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2978                                                false, pfx, dev->ifindex);
2979                         }
2980                         return 0;
2981                 }
2982         }
2983         read_unlock_bh(&idev->lock);
2984         return -EADDRNOTAVAIL;
2985 }
2986
2987
2988 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2989 {
2990         struct ifa6_config cfg = {
2991                 .ifa_flags = IFA_F_PERMANENT,
2992                 .preferred_lft = INFINITY_LIFE_TIME,
2993                 .valid_lft = INFINITY_LIFE_TIME,
2994         };
2995         struct in6_ifreq ireq;
2996         int err;
2997
2998         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2999                 return -EPERM;
3000
3001         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3002                 return -EFAULT;
3003
3004         cfg.pfx = &ireq.ifr6_addr;
3005         cfg.plen = ireq.ifr6_prefixlen;
3006
3007         rtnl_lock();
3008         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3009         rtnl_unlock();
3010         return err;
3011 }
3012
3013 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3014 {
3015         struct in6_ifreq ireq;
3016         int err;
3017
3018         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3019                 return -EPERM;
3020
3021         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3022                 return -EFAULT;
3023
3024         rtnl_lock();
3025         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3026                              ireq.ifr6_prefixlen);
3027         rtnl_unlock();
3028         return err;
3029 }
3030
3031 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3032                      int plen, int scope)
3033 {
3034         struct inet6_ifaddr *ifp;
3035         struct ifa6_config cfg = {
3036                 .pfx = addr,
3037                 .plen = plen,
3038                 .ifa_flags = IFA_F_PERMANENT,
3039                 .valid_lft = INFINITY_LIFE_TIME,
3040                 .preferred_lft = INFINITY_LIFE_TIME,
3041                 .scope = scope
3042         };
3043
3044         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3045         if (!IS_ERR(ifp)) {
3046                 spin_lock_bh(&ifp->lock);
3047                 ifp->flags &= ~IFA_F_TENTATIVE;
3048                 spin_unlock_bh(&ifp->lock);
3049                 rt_genid_bump_ipv6(dev_net(idev->dev));
3050                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3051                 in6_ifa_put(ifp);
3052         }
3053 }
3054
3055 #if IS_ENABLED(CONFIG_IPV6_SIT)
3056 static void sit_add_v4_addrs(struct inet6_dev *idev)
3057 {
3058         struct in6_addr addr;
3059         struct net_device *dev;
3060         struct net *net = dev_net(idev->dev);
3061         int scope, plen;
3062         u32 pflags = 0;
3063
3064         ASSERT_RTNL();
3065
3066         memset(&addr, 0, sizeof(struct in6_addr));
3067         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3068
3069         if (idev->dev->flags&IFF_POINTOPOINT) {
3070                 addr.s6_addr32[0] = htonl(0xfe800000);
3071                 scope = IFA_LINK;
3072                 plen = 64;
3073         } else {
3074                 scope = IPV6_ADDR_COMPATv4;
3075                 plen = 96;
3076                 pflags |= RTF_NONEXTHOP;
3077         }
3078
3079         if (addr.s6_addr32[3]) {
3080                 add_addr(idev, &addr, plen, scope);
3081                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3082                                       GFP_KERNEL);
3083                 return;
3084         }
3085
3086         for_each_netdev(net, dev) {
3087                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3088                 if (in_dev && (dev->flags & IFF_UP)) {
3089                         struct in_ifaddr *ifa;
3090
3091                         int flag = scope;
3092
3093                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3094
3095                                 addr.s6_addr32[3] = ifa->ifa_local;
3096
3097                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3098                                         continue;
3099                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3100                                         if (idev->dev->flags&IFF_POINTOPOINT)
3101                                                 continue;
3102                                         flag |= IFA_HOST;
3103                                 }
3104
3105                                 add_addr(idev, &addr, plen, flag);
3106                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3107                                                       0, pflags, GFP_KERNEL);
3108                         }
3109                 }
3110         }
3111 }
3112 #endif
3113
3114 static void init_loopback(struct net_device *dev)
3115 {
3116         struct inet6_dev  *idev;
3117
3118         /* ::1 */
3119
3120         ASSERT_RTNL();
3121
3122         idev = ipv6_find_idev(dev);
3123         if (!idev) {
3124                 pr_debug("%s: add_dev failed\n", __func__);
3125                 return;
3126         }
3127
3128         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3129 }
3130
3131 void addrconf_add_linklocal(struct inet6_dev *idev,
3132                             const struct in6_addr *addr, u32 flags)
3133 {
3134         struct ifa6_config cfg = {
3135                 .pfx = addr,
3136                 .plen = 64,
3137                 .ifa_flags = flags | IFA_F_PERMANENT,
3138                 .valid_lft = INFINITY_LIFE_TIME,
3139                 .preferred_lft = INFINITY_LIFE_TIME,
3140                 .scope = IFA_LINK
3141         };
3142         struct inet6_ifaddr *ifp;
3143
3144 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3145         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3146              idev->cnf.optimistic_dad) &&
3147             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3148                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3149 #endif
3150
3151         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3152         if (!IS_ERR(ifp)) {
3153                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3154                                       0, 0, GFP_ATOMIC);
3155                 addrconf_dad_start(ifp);
3156                 in6_ifa_put(ifp);
3157         }
3158 }
3159 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3160
3161 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3162 {
3163         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3164                 return true;
3165
3166         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3167             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3168                 return true;
3169
3170         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3171             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3172                 return true;
3173
3174         return false;
3175 }
3176
3177 static int ipv6_generate_stable_address(struct in6_addr *address,
3178                                         u8 dad_count,
3179                                         const struct inet6_dev *idev)
3180 {
3181         static DEFINE_SPINLOCK(lock);
3182         static __u32 digest[SHA_DIGEST_WORDS];
3183         static __u32 workspace[SHA_WORKSPACE_WORDS];
3184
3185         static union {
3186                 char __data[SHA_MESSAGE_BYTES];
3187                 struct {
3188                         struct in6_addr secret;
3189                         __be32 prefix[2];
3190                         unsigned char hwaddr[MAX_ADDR_LEN];
3191                         u8 dad_count;
3192                 } __packed;
3193         } data;
3194
3195         struct in6_addr secret;
3196         struct in6_addr temp;
3197         struct net *net = dev_net(idev->dev);
3198
3199         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3200
3201         if (idev->cnf.stable_secret.initialized)
3202                 secret = idev->cnf.stable_secret.secret;
3203         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3204                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3205         else
3206                 return -1;
3207
3208 retry:
3209         spin_lock_bh(&lock);
3210
3211         sha_init(digest);
3212         memset(&data, 0, sizeof(data));
3213         memset(workspace, 0, sizeof(workspace));
3214         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3215         data.prefix[0] = address->s6_addr32[0];
3216         data.prefix[1] = address->s6_addr32[1];
3217         data.secret = secret;
3218         data.dad_count = dad_count;
3219
3220         sha_transform(digest, data.__data, workspace);
3221
3222         temp = *address;
3223         temp.s6_addr32[2] = (__force __be32)digest[0];
3224         temp.s6_addr32[3] = (__force __be32)digest[1];
3225
3226         spin_unlock_bh(&lock);
3227
3228         if (ipv6_reserved_interfaceid(temp)) {
3229                 dad_count++;
3230                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3231                         return -1;
3232                 goto retry;
3233         }
3234
3235         *address = temp;
3236         return 0;
3237 }
3238
3239 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3240 {
3241         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3242
3243         if (s->initialized)
3244                 return;
3245         s = &idev->cnf.stable_secret;
3246         get_random_bytes(&s->secret, sizeof(s->secret));
3247         s->initialized = true;
3248 }
3249
3250 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3251 {
3252         struct in6_addr addr;
3253
3254         /* no link local addresses on L3 master devices */
3255         if (netif_is_l3_master(idev->dev))
3256                 return;
3257
3258         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3259
3260         switch (idev->cnf.addr_gen_mode) {
3261         case IN6_ADDR_GEN_MODE_RANDOM:
3262                 ipv6_gen_mode_random_init(idev);
3263                 /* fallthrough */
3264         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3265                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3266                         addrconf_add_linklocal(idev, &addr,
3267                                                IFA_F_STABLE_PRIVACY);
3268                 else if (prefix_route)
3269                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3270                                               0, 0, GFP_KERNEL);
3271                 break;
3272         case IN6_ADDR_GEN_MODE_EUI64:
3273                 /* addrconf_add_linklocal also adds a prefix_route and we
3274                  * only need to care about prefix routes if ipv6_generate_eui64
3275                  * couldn't generate one.
3276                  */
3277                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3278                         addrconf_add_linklocal(idev, &addr, 0);
3279                 else if (prefix_route)
3280                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3281                                               0, 0, GFP_KERNEL);
3282                 break;
3283         case IN6_ADDR_GEN_MODE_NONE:
3284         default:
3285                 /* will not add any link local address */
3286                 break;
3287         }
3288 }
3289
3290 static void addrconf_dev_config(struct net_device *dev)
3291 {
3292         struct inet6_dev *idev;
3293
3294         ASSERT_RTNL();
3295
3296         if ((dev->type != ARPHRD_ETHER) &&
3297             (dev->type != ARPHRD_FDDI) &&
3298             (dev->type != ARPHRD_ARCNET) &&
3299             (dev->type != ARPHRD_INFINIBAND) &&
3300             (dev->type != ARPHRD_IEEE1394) &&
3301             (dev->type != ARPHRD_TUNNEL6) &&
3302             (dev->type != ARPHRD_6LOWPAN) &&
3303             (dev->type != ARPHRD_IP6GRE) &&
3304             (dev->type != ARPHRD_IPGRE) &&
3305             (dev->type != ARPHRD_TUNNEL) &&
3306             (dev->type != ARPHRD_NONE) &&
3307             (dev->type != ARPHRD_RAWIP)) {
3308                 /* Alas, we support only Ethernet autoconfiguration. */
3309                 return;
3310         }
3311
3312         idev = addrconf_add_dev(dev);
3313         if (IS_ERR(idev))
3314                 return;
3315
3316         /* this device type has no EUI support */
3317         if (dev->type == ARPHRD_NONE &&
3318             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3319                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3320
3321         addrconf_addr_gen(idev, false);
3322 }
3323
3324 #if IS_ENABLED(CONFIG_IPV6_SIT)
3325 static void addrconf_sit_config(struct net_device *dev)
3326 {
3327         struct inet6_dev *idev;
3328
3329         ASSERT_RTNL();
3330
3331         /*
3332          * Configure the tunnel with one of our IPv4
3333          * addresses... we should configure all of
3334          * our v4 addrs in the tunnel
3335          */
3336
3337         idev = ipv6_find_idev(dev);
3338         if (!idev) {
3339                 pr_debug("%s: add_dev failed\n", __func__);
3340                 return;
3341         }
3342
3343         if (dev->priv_flags & IFF_ISATAP) {
3344                 addrconf_addr_gen(idev, false);
3345                 return;
3346         }
3347
3348         sit_add_v4_addrs(idev);
3349
3350         if (dev->flags&IFF_POINTOPOINT)
3351                 addrconf_add_mroute(dev);
3352 }
3353 #endif
3354
3355 #if IS_ENABLED(CONFIG_NET_IPGRE)
3356 static void addrconf_gre_config(struct net_device *dev)
3357 {
3358         struct inet6_dev *idev;
3359
3360         ASSERT_RTNL();
3361
3362         idev = ipv6_find_idev(dev);
3363         if (!idev) {
3364                 pr_debug("%s: add_dev failed\n", __func__);
3365                 return;
3366         }
3367
3368         addrconf_addr_gen(idev, true);
3369         if (dev->flags & IFF_POINTOPOINT)
3370                 addrconf_add_mroute(dev);
3371 }
3372 #endif
3373
3374 static int fixup_permanent_addr(struct net *net,
3375                                 struct inet6_dev *idev,
3376                                 struct inet6_ifaddr *ifp)
3377 {
3378         /* !fib6_node means the host route was removed from the
3379          * FIB, for example, if 'lo' device is taken down. In that
3380          * case regenerate the host route.
3381          */
3382         if (!ifp->rt || !ifp->rt->fib6_node) {
3383                 struct fib6_info *f6i, *prev;
3384
3385                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3386                                          GFP_ATOMIC);
3387                 if (IS_ERR(f6i))
3388                         return PTR_ERR(f6i);
3389
3390                 /* ifp->rt can be accessed outside of rtnl */
3391                 spin_lock(&ifp->lock);
3392                 prev = ifp->rt;
3393                 ifp->rt = f6i;
3394                 spin_unlock(&ifp->lock);
3395
3396                 fib6_info_release(prev);
3397         }
3398
3399         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3400                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3401                                       ifp->rt_priority, idev->dev, 0, 0,
3402                                       GFP_ATOMIC);
3403         }
3404
3405         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3406                 addrconf_dad_start(ifp);
3407
3408         return 0;
3409 }
3410
3411 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3412 {
3413         struct inet6_ifaddr *ifp, *tmp;
3414         struct inet6_dev *idev;
3415
3416         idev = __in6_dev_get(dev);
3417         if (!idev)
3418                 return;
3419
3420         write_lock_bh(&idev->lock);
3421
3422         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3423                 if ((ifp->flags & IFA_F_PERMANENT) &&
3424                     fixup_permanent_addr(net, idev, ifp) < 0) {
3425                         write_unlock_bh(&idev->lock);
3426                         in6_ifa_hold(ifp);
3427                         ipv6_del_addr(ifp);
3428                         write_lock_bh(&idev->lock);
3429
3430                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3431                                              idev->dev->name, &ifp->addr);
3432                 }
3433         }
3434
3435         write_unlock_bh(&idev->lock);
3436 }
3437
3438 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3439                            void *ptr)
3440 {
3441         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3442         struct netdev_notifier_change_info *change_info;
3443         struct netdev_notifier_changeupper_info *info;
3444         struct inet6_dev *idev = __in6_dev_get(dev);
3445         struct net *net = dev_net(dev);
3446         int run_pending = 0;
3447         int err;
3448
3449         switch (event) {
3450         case NETDEV_REGISTER:
3451                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3452                         idev = ipv6_add_dev(dev);
3453                         if (IS_ERR(idev))
3454                                 return notifier_from_errno(PTR_ERR(idev));
3455                 }
3456                 break;
3457
3458         case NETDEV_CHANGEMTU:
3459                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3460                 if (dev->mtu < IPV6_MIN_MTU) {
3461                         addrconf_ifdown(dev, dev != net->loopback_dev);
3462                         break;
3463                 }
3464
3465                 if (idev) {
3466                         rt6_mtu_change(dev, dev->mtu);
3467                         idev->cnf.mtu6 = dev->mtu;
3468                         break;
3469                 }
3470
3471                 /* allocate new idev */
3472                 idev = ipv6_add_dev(dev);
3473                 if (IS_ERR(idev))
3474                         break;
3475
3476                 /* device is still not ready */
3477                 if (!(idev->if_flags & IF_READY))
3478                         break;
3479
3480                 run_pending = 1;
3481
3482                 /* fall through */
3483
3484         case NETDEV_UP:
3485         case NETDEV_CHANGE:
3486                 if (dev->flags & IFF_SLAVE)
3487                         break;
3488
3489                 if (idev && idev->cnf.disable_ipv6)
3490                         break;
3491
3492                 if (event == NETDEV_UP) {
3493                         /* restore routes for permanent addresses */
3494                         addrconf_permanent_addr(net, dev);
3495
3496                         if (!addrconf_link_ready(dev)) {
3497                                 /* device is not ready yet. */
3498                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3499                                         dev->name);
3500                                 break;
3501                         }
3502
3503                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3504                                 idev = ipv6_add_dev(dev);
3505
3506                         if (!IS_ERR_OR_NULL(idev)) {
3507                                 idev->if_flags |= IF_READY;
3508                                 run_pending = 1;
3509                         }
3510                 } else if (event == NETDEV_CHANGE) {
3511                         if (!addrconf_link_ready(dev)) {
3512                                 /* device is still not ready. */
3513                                 rt6_sync_down_dev(dev, event);
3514                                 break;
3515                         }
3516
3517                         if (!IS_ERR_OR_NULL(idev)) {
3518                                 if (idev->if_flags & IF_READY) {
3519                                         /* device is already configured -
3520                                          * but resend MLD reports, we might
3521                                          * have roamed and need to update
3522                                          * multicast snooping switches
3523                                          */
3524                                         ipv6_mc_up(idev);
3525                                         change_info = ptr;
3526                                         if (change_info->flags_changed & IFF_NOARP)
3527                                                 addrconf_dad_run(idev, true);
3528                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3529                                         break;
3530                                 }
3531                                 idev->if_flags |= IF_READY;
3532                         }
3533
3534                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3535                                 dev->name);
3536
3537                         run_pending = 1;
3538                 }
3539
3540                 switch (dev->type) {
3541 #if IS_ENABLED(CONFIG_IPV6_SIT)
3542                 case ARPHRD_SIT:
3543                         addrconf_sit_config(dev);
3544                         break;
3545 #endif
3546 #if IS_ENABLED(CONFIG_NET_IPGRE)
3547                 case ARPHRD_IPGRE:
3548                         addrconf_gre_config(dev);
3549                         break;
3550 #endif
3551                 case ARPHRD_LOOPBACK:
3552                         init_loopback(dev);
3553                         break;
3554
3555                 default:
3556                         addrconf_dev_config(dev);
3557                         break;
3558                 }
3559
3560                 if (!IS_ERR_OR_NULL(idev)) {
3561                         if (run_pending)
3562                                 addrconf_dad_run(idev, false);
3563
3564                         /* Device has an address by now */
3565                         rt6_sync_up(dev, RTNH_F_DEAD);
3566
3567                         /*
3568                          * If the MTU changed during the interface down,
3569                          * when the interface up, the changed MTU must be
3570                          * reflected in the idev as well as routers.
3571                          */
3572                         if (idev->cnf.mtu6 != dev->mtu &&
3573                             dev->mtu >= IPV6_MIN_MTU) {
3574                                 rt6_mtu_change(dev, dev->mtu);
3575                                 idev->cnf.mtu6 = dev->mtu;
3576                         }
3577                         idev->tstamp = jiffies;
3578                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3579
3580                         /*
3581                          * If the changed mtu during down is lower than
3582                          * IPV6_MIN_MTU stop IPv6 on this interface.
3583                          */
3584                         if (dev->mtu < IPV6_MIN_MTU)
3585                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3586                 }
3587                 break;
3588
3589         case NETDEV_DOWN:
3590         case NETDEV_UNREGISTER:
3591                 /*
3592                  *      Remove all addresses from this interface.
3593                  */
3594                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3595                 break;
3596
3597         case NETDEV_CHANGENAME:
3598                 if (idev) {
3599                         snmp6_unregister_dev(idev);
3600                         addrconf_sysctl_unregister(idev);
3601                         err = addrconf_sysctl_register(idev);
3602                         if (err)
3603                                 return notifier_from_errno(err);
3604                         err = snmp6_register_dev(idev);
3605                         if (err) {
3606                                 addrconf_sysctl_unregister(idev);
3607                                 return notifier_from_errno(err);
3608                         }
3609                 }
3610                 break;
3611
3612         case NETDEV_PRE_TYPE_CHANGE:
3613         case NETDEV_POST_TYPE_CHANGE:
3614                 if (idev)
3615                         addrconf_type_change(dev, event);
3616                 break;
3617
3618         case NETDEV_CHANGEUPPER:
3619                 info = ptr;
3620
3621                 /* flush all routes if dev is linked to or unlinked from
3622                  * an L3 master device (e.g., VRF)
3623                  */
3624                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3625                         addrconf_ifdown(dev, 0);
3626         }
3627
3628         return NOTIFY_OK;
3629 }
3630
3631 /*
3632  *      addrconf module should be notified of a device going up
3633  */
3634 static struct notifier_block ipv6_dev_notf = {
3635         .notifier_call = addrconf_notify,
3636         .priority = ADDRCONF_NOTIFY_PRIORITY,
3637 };
3638
3639 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3640 {
3641         struct inet6_dev *idev;
3642         ASSERT_RTNL();
3643
3644         idev = __in6_dev_get(dev);
3645
3646         if (event == NETDEV_POST_TYPE_CHANGE)
3647                 ipv6_mc_remap(idev);
3648         else if (event == NETDEV_PRE_TYPE_CHANGE)
3649                 ipv6_mc_unmap(idev);
3650 }
3651
3652 static bool addr_is_local(const struct in6_addr *addr)
3653 {
3654         return ipv6_addr_type(addr) &
3655                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3656 }
3657
3658 static int addrconf_ifdown(struct net_device *dev, int how)
3659 {
3660         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3661         struct net *net = dev_net(dev);
3662         struct inet6_dev *idev;
3663         struct inet6_ifaddr *ifa, *tmp;
3664         bool keep_addr = false;
3665         int state, i;
3666
3667         ASSERT_RTNL();
3668
3669         rt6_disable_ip(dev, event);
3670
3671         idev = __in6_dev_get(dev);
3672         if (!idev)
3673                 return -ENODEV;
3674
3675         /*
3676          * Step 1: remove reference to ipv6 device from parent device.
3677          *         Do not dev_put!
3678          */
3679         if (how) {
3680                 idev->dead = 1;
3681
3682                 /* protected by rtnl_lock */
3683                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3684
3685                 /* Step 1.5: remove snmp6 entry */
3686                 snmp6_unregister_dev(idev);
3687
3688         }
3689
3690         /* combine the user config with event to determine if permanent
3691          * addresses are to be removed from address hash table
3692          */
3693         if (!how && !idev->cnf.disable_ipv6) {
3694                 /* aggregate the system setting and interface setting */
3695                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3696
3697                 if (!_keep_addr)
3698                         _keep_addr = idev->cnf.keep_addr_on_down;
3699
3700                 keep_addr = (_keep_addr > 0);
3701         }
3702
3703         /* Step 2: clear hash table */
3704         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3705                 struct hlist_head *h = &inet6_addr_lst[i];
3706
3707                 spin_lock_bh(&addrconf_hash_lock);
3708 restart:
3709                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3710                         if (ifa->idev == idev) {
3711                                 addrconf_del_dad_work(ifa);
3712                                 /* combined flag + permanent flag decide if
3713                                  * address is retained on a down event
3714                                  */
3715                                 if (!keep_addr ||
3716                                     !(ifa->flags & IFA_F_PERMANENT) ||
3717                                     addr_is_local(&ifa->addr)) {
3718                                         hlist_del_init_rcu(&ifa->addr_lst);
3719                                         goto restart;
3720                                 }
3721                         }
3722                 }
3723                 spin_unlock_bh(&addrconf_hash_lock);
3724         }
3725
3726         write_lock_bh(&idev->lock);
3727
3728         addrconf_del_rs_timer(idev);
3729
3730         /* Step 2: clear flags for stateless addrconf */
3731         if (!how)
3732                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3733
3734         /* Step 3: clear tempaddr list */
3735         while (!list_empty(&idev->tempaddr_list)) {
3736                 ifa = list_first_entry(&idev->tempaddr_list,
3737                                        struct inet6_ifaddr, tmp_list);
3738                 list_del(&ifa->tmp_list);
3739                 write_unlock_bh(&idev->lock);
3740                 spin_lock_bh(&ifa->lock);
3741
3742                 if (ifa->ifpub) {
3743                         in6_ifa_put(ifa->ifpub);
3744                         ifa->ifpub = NULL;
3745                 }
3746                 spin_unlock_bh(&ifa->lock);
3747                 in6_ifa_put(ifa);
3748                 write_lock_bh(&idev->lock);
3749         }
3750
3751         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3752                 struct fib6_info *rt = NULL;
3753                 bool keep;
3754
3755                 addrconf_del_dad_work(ifa);
3756
3757                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3758                         !addr_is_local(&ifa->addr);
3759
3760                 write_unlock_bh(&idev->lock);
3761                 spin_lock_bh(&ifa->lock);
3762
3763                 if (keep) {
3764                         /* set state to skip the notifier below */
3765                         state = INET6_IFADDR_STATE_DEAD;
3766                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3767                         if (!(ifa->flags & IFA_F_NODAD))
3768                                 ifa->flags |= IFA_F_TENTATIVE;
3769
3770                         rt = ifa->rt;
3771                         ifa->rt = NULL;
3772                 } else {
3773                         state = ifa->state;
3774                         ifa->state = INET6_IFADDR_STATE_DEAD;
3775                 }
3776
3777                 spin_unlock_bh(&ifa->lock);
3778
3779                 if (rt)
3780                         ip6_del_rt(net, rt);
3781
3782                 if (state != INET6_IFADDR_STATE_DEAD) {
3783                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3784                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3785                 } else {
3786                         if (idev->cnf.forwarding)
3787                                 addrconf_leave_anycast(ifa);
3788                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3789                 }
3790
3791                 write_lock_bh(&idev->lock);
3792                 if (!keep) {
3793                         list_del_rcu(&ifa->if_list);
3794                         in6_ifa_put(ifa);
3795                 }
3796         }
3797
3798         write_unlock_bh(&idev->lock);
3799
3800         /* Step 5: Discard anycast and multicast list */
3801         if (how) {
3802                 ipv6_ac_destroy_dev(idev);
3803                 ipv6_mc_destroy_dev(idev);
3804         } else {
3805                 ipv6_mc_down(idev);
3806         }
3807
3808         idev->tstamp = jiffies;
3809
3810         /* Last: Shot the device (if unregistered) */
3811         if (how) {
3812                 addrconf_sysctl_unregister(idev);
3813                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3814                 neigh_ifdown(&nd_tbl, dev);
3815                 in6_dev_put(idev);
3816         }
3817         return 0;
3818 }
3819
3820 static void addrconf_rs_timer(struct timer_list *t)
3821 {
3822         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3823         struct net_device *dev = idev->dev;
3824         struct in6_addr lladdr;
3825
3826         write_lock(&idev->lock);
3827         if (idev->dead || !(idev->if_flags & IF_READY))
3828                 goto out;
3829
3830         if (!ipv6_accept_ra(idev))
3831                 goto out;
3832
3833         /* Announcement received after solicitation was sent */
3834         if (idev->if_flags & IF_RA_RCVD)
3835                 goto out;
3836
3837         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3838                 write_unlock(&idev->lock);
3839                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3840                         ndisc_send_rs(dev, &lladdr,
3841                                       &in6addr_linklocal_allrouters);
3842                 else
3843                         goto put;
3844
3845                 write_lock(&idev->lock);
3846                 idev->rs_interval = rfc3315_s14_backoff_update(
3847                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3848                 /* The wait after the last probe can be shorter */
3849                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3850                                              idev->cnf.rtr_solicits) ?
3851                                       idev->cnf.rtr_solicit_delay :
3852                                       idev->rs_interval);
3853         } else {
3854                 /*
3855                  * Note: we do not support deprecated "all on-link"
3856                  * assumption any longer.
3857                  */
3858                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3859         }
3860
3861 out:
3862         write_unlock(&idev->lock);
3863 put:
3864         in6_dev_put(idev);
3865 }
3866
3867 /*
3868  *      Duplicate Address Detection
3869  */
3870 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3871 {
3872         unsigned long rand_num;
3873         struct inet6_dev *idev = ifp->idev;
3874         u64 nonce;
3875
3876         if (ifp->flags & IFA_F_OPTIMISTIC)
3877                 rand_num = 0;
3878         else
3879                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3880
3881         nonce = 0;
3882         if (idev->cnf.enhanced_dad ||
3883             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3884                 do
3885                         get_random_bytes(&nonce, 6);
3886                 while (nonce == 0);
3887         }
3888         ifp->dad_nonce = nonce;
3889         ifp->dad_probes = idev->cnf.dad_transmits;
3890         addrconf_mod_dad_work(ifp, rand_num);
3891 }
3892
3893 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3894 {
3895         struct inet6_dev *idev = ifp->idev;
3896         struct net_device *dev = idev->dev;
3897         bool bump_id, notify = false;
3898         struct net *net;
3899
3900         addrconf_join_solict(dev, &ifp->addr);
3901
3902         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3903
3904         read_lock_bh(&idev->lock);
3905         spin_lock(&ifp->lock);
3906         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3907                 goto out;
3908
3909         net = dev_net(dev);
3910         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3911             (net->ipv6.devconf_all->accept_dad < 1 &&
3912              idev->cnf.accept_dad < 1) ||
3913             !(ifp->flags&IFA_F_TENTATIVE) ||
3914             ifp->flags & IFA_F_NODAD) {
3915                 bool send_na = false;
3916
3917                 if (ifp->flags & IFA_F_TENTATIVE &&
3918                     !(ifp->flags & IFA_F_OPTIMISTIC))
3919                         send_na = true;
3920                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3921                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3922                 spin_unlock(&ifp->lock);
3923                 read_unlock_bh(&idev->lock);
3924
3925                 addrconf_dad_completed(ifp, bump_id, send_na);
3926                 return;
3927         }
3928
3929         if (!(idev->if_flags & IF_READY)) {
3930                 spin_unlock(&ifp->lock);
3931                 read_unlock_bh(&idev->lock);
3932                 /*
3933                  * If the device is not ready:
3934                  * - keep it tentative if it is a permanent address.
3935                  * - otherwise, kill it.
3936                  */
3937                 in6_ifa_hold(ifp);
3938                 addrconf_dad_stop(ifp, 0);
3939                 return;
3940         }
3941
3942         /*
3943          * Optimistic nodes can start receiving
3944          * Frames right away
3945          */
3946         if (ifp->flags & IFA_F_OPTIMISTIC) {
3947                 ip6_ins_rt(net, ifp->rt);
3948                 if (ipv6_use_optimistic_addr(net, idev)) {
3949                         /* Because optimistic nodes can use this address,
3950                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3951                          */
3952                         notify = true;
3953                 }
3954         }
3955
3956         addrconf_dad_kick(ifp);
3957 out:
3958         spin_unlock(&ifp->lock);
3959         read_unlock_bh(&idev->lock);
3960         if (notify)
3961                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3962 }
3963
3964 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3965 {
3966         bool begin_dad = false;
3967
3968         spin_lock_bh(&ifp->lock);
3969         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3970                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3971                 begin_dad = true;
3972         }
3973         spin_unlock_bh(&ifp->lock);
3974
3975         if (begin_dad)
3976                 addrconf_mod_dad_work(ifp, 0);
3977 }
3978
3979 static void addrconf_dad_work(struct work_struct *w)
3980 {
3981         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3982                                                 struct inet6_ifaddr,
3983                                                 dad_work);
3984         struct inet6_dev *idev = ifp->idev;
3985         bool bump_id, disable_ipv6 = false;
3986         struct in6_addr mcaddr;
3987
3988         enum {
3989                 DAD_PROCESS,
3990                 DAD_BEGIN,
3991                 DAD_ABORT,
3992         } action = DAD_PROCESS;
3993
3994         rtnl_lock();
3995
3996         spin_lock_bh(&ifp->lock);
3997         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3998                 action = DAD_BEGIN;
3999                 ifp->state = INET6_IFADDR_STATE_DAD;
4000         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4001                 action = DAD_ABORT;
4002                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4003
4004                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4005                      idev->cnf.accept_dad > 1) &&
4006                     !idev->cnf.disable_ipv6 &&
4007                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4008                         struct in6_addr addr;
4009
4010                         addr.s6_addr32[0] = htonl(0xfe800000);
4011                         addr.s6_addr32[1] = 0;
4012
4013                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4014                             ipv6_addr_equal(&ifp->addr, &addr)) {
4015                                 /* DAD failed for link-local based on MAC */
4016                                 idev->cnf.disable_ipv6 = 1;
4017
4018                                 pr_info("%s: IPv6 being disabled!\n",
4019                                         ifp->idev->dev->name);
4020                                 disable_ipv6 = true;
4021                         }
4022                 }
4023         }
4024         spin_unlock_bh(&ifp->lock);
4025
4026         if (action == DAD_BEGIN) {
4027                 addrconf_dad_begin(ifp);
4028                 goto out;
4029         } else if (action == DAD_ABORT) {
4030                 in6_ifa_hold(ifp);
4031                 addrconf_dad_stop(ifp, 1);
4032                 if (disable_ipv6)
4033                         addrconf_ifdown(idev->dev, 0);
4034                 goto out;
4035         }
4036
4037         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4038                 goto out;
4039
4040         write_lock_bh(&idev->lock);
4041         if (idev->dead || !(idev->if_flags & IF_READY)) {
4042                 write_unlock_bh(&idev->lock);
4043                 goto out;
4044         }
4045
4046         spin_lock(&ifp->lock);
4047         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4048                 spin_unlock(&ifp->lock);
4049                 write_unlock_bh(&idev->lock);
4050                 goto out;
4051         }
4052
4053         if (ifp->dad_probes == 0) {
4054                 bool send_na = false;
4055
4056                 /*
4057                  * DAD was successful
4058                  */
4059
4060                 if (ifp->flags & IFA_F_TENTATIVE &&
4061                     !(ifp->flags & IFA_F_OPTIMISTIC))
4062                         send_na = true;
4063                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4064                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4065                 spin_unlock(&ifp->lock);
4066                 write_unlock_bh(&idev->lock);
4067
4068                 addrconf_dad_completed(ifp, bump_id, send_na);
4069
4070                 goto out;
4071         }
4072
4073         ifp->dad_probes--;
4074         addrconf_mod_dad_work(ifp,
4075                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4076         spin_unlock(&ifp->lock);
4077         write_unlock_bh(&idev->lock);
4078
4079         /* send a neighbour solicitation for our addr */
4080         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4081         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4082                       ifp->dad_nonce);
4083 out:
4084         in6_ifa_put(ifp);
4085         rtnl_unlock();
4086 }
4087
4088 /* ifp->idev must be at least read locked */
4089 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4090 {
4091         struct inet6_ifaddr *ifpiter;
4092         struct inet6_dev *idev = ifp->idev;
4093
4094         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4095                 if (ifpiter->scope > IFA_LINK)
4096                         break;
4097                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4098                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4099                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4100                     IFA_F_PERMANENT)
4101                         return false;
4102         }
4103         return true;
4104 }
4105
4106 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4107                                    bool send_na)
4108 {
4109         struct net_device *dev = ifp->idev->dev;
4110         struct in6_addr lladdr;
4111         bool send_rs, send_mld;
4112
4113         addrconf_del_dad_work(ifp);
4114
4115         /*
4116          *      Configure the address for reception. Now it is valid.
4117          */
4118
4119         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4120
4121         /* If added prefix is link local and we are prepared to process
4122            router advertisements, start sending router solicitations.
4123          */
4124
4125         read_lock_bh(&ifp->idev->lock);
4126         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4127         send_rs = send_mld &&
4128                   ipv6_accept_ra(ifp->idev) &&
4129                   ifp->idev->cnf.rtr_solicits != 0 &&
4130                   (dev->flags&IFF_LOOPBACK) == 0;
4131         read_unlock_bh(&ifp->idev->lock);
4132
4133         /* While dad is in progress mld report's source address is in6_addrany.
4134          * Resend with proper ll now.
4135          */
4136         if (send_mld)
4137                 ipv6_mc_dad_complete(ifp->idev);
4138
4139         /* send unsolicited NA if enabled */
4140         if (send_na &&
4141             (ifp->idev->cnf.ndisc_notify ||
4142              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4143                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4144                               /*router=*/ !!ifp->idev->cnf.forwarding,
4145                               /*solicited=*/ false, /*override=*/ true,
4146                               /*inc_opt=*/ true);
4147         }
4148
4149         if (send_rs) {
4150                 /*
4151                  *      If a host as already performed a random delay
4152                  *      [...] as part of DAD [...] there is no need
4153                  *      to delay again before sending the first RS
4154                  */
4155                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4156                         return;
4157                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4158
4159                 write_lock_bh(&ifp->idev->lock);
4160                 spin_lock(&ifp->lock);
4161                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4162                         ifp->idev->cnf.rtr_solicit_interval);
4163                 ifp->idev->rs_probes = 1;
4164                 ifp->idev->if_flags |= IF_RS_SENT;
4165                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4166                 spin_unlock(&ifp->lock);
4167                 write_unlock_bh(&ifp->idev->lock);
4168         }
4169
4170         if (bump_id)
4171                 rt_genid_bump_ipv6(dev_net(dev));
4172
4173         /* Make sure that a new temporary address will be created
4174          * before this temporary address becomes deprecated.
4175          */
4176         if (ifp->flags & IFA_F_TEMPORARY)
4177                 addrconf_verify_rtnl();
4178 }
4179
4180 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4181 {
4182         struct inet6_ifaddr *ifp;
4183
4184         read_lock_bh(&idev->lock);
4185         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4186                 spin_lock(&ifp->lock);
4187                 if ((ifp->flags & IFA_F_TENTATIVE &&
4188                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4189                         if (restart)
4190                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4191                         addrconf_dad_kick(ifp);
4192                 }
4193                 spin_unlock(&ifp->lock);
4194         }
4195         read_unlock_bh(&idev->lock);
4196 }
4197
4198 #ifdef CONFIG_PROC_FS
4199 struct if6_iter_state {
4200         struct seq_net_private p;
4201         int bucket;
4202         int offset;
4203 };
4204
4205 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4206 {
4207         struct if6_iter_state *state = seq->private;
4208         struct net *net = seq_file_net(seq);
4209         struct inet6_ifaddr *ifa = NULL;
4210         int p = 0;
4211
4212         /* initial bucket if pos is 0 */
4213         if (pos == 0) {
4214                 state->bucket = 0;
4215                 state->offset = 0;
4216         }
4217
4218         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4219                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4220                                          addr_lst) {
4221                         if (!net_eq(dev_net(ifa->idev->dev), net))
4222                                 continue;
4223                         /* sync with offset */
4224                         if (p < state->offset) {
4225                                 p++;
4226                                 continue;
4227                         }
4228                         return ifa;
4229                 }
4230
4231                 /* prepare for next bucket */
4232                 state->offset = 0;
4233                 p = 0;
4234         }
4235         return NULL;
4236 }
4237
4238 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4239                                          struct inet6_ifaddr *ifa)
4240 {
4241         struct if6_iter_state *state = seq->private;
4242         struct net *net = seq_file_net(seq);
4243
4244         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4245                 if (!net_eq(dev_net(ifa->idev->dev), net))
4246                         continue;
4247                 state->offset++;
4248                 return ifa;
4249         }
4250
4251         state->offset = 0;
4252         while (++state->bucket < IN6_ADDR_HSIZE) {
4253                 hlist_for_each_entry_rcu(ifa,
4254                                      &inet6_addr_lst[state->bucket], addr_lst) {
4255                         if (!net_eq(dev_net(ifa->idev->dev), net))
4256                                 continue;
4257                         return ifa;
4258                 }
4259         }
4260
4261         return NULL;
4262 }
4263
4264 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4265         __acquires(rcu)
4266 {
4267         rcu_read_lock();
4268         return if6_get_first(seq, *pos);
4269 }
4270
4271 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4272 {
4273         struct inet6_ifaddr *ifa;
4274
4275         ifa = if6_get_next(seq, v);
4276         ++*pos;
4277         return ifa;
4278 }
4279
4280 static void if6_seq_stop(struct seq_file *seq, void *v)
4281         __releases(rcu)
4282 {
4283         rcu_read_unlock();
4284 }
4285
4286 static int if6_seq_show(struct seq_file *seq, void *v)
4287 {
4288         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4289         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4290                    &ifp->addr,
4291                    ifp->idev->dev->ifindex,
4292                    ifp->prefix_len,
4293                    ifp->scope,
4294                    (u8) ifp->flags,
4295                    ifp->idev->dev->name);
4296         return 0;
4297 }
4298
4299 static const struct seq_operations if6_seq_ops = {
4300         .start  = if6_seq_start,
4301         .next   = if6_seq_next,
4302         .show   = if6_seq_show,
4303         .stop   = if6_seq_stop,
4304 };
4305
4306 static int __net_init if6_proc_net_init(struct net *net)
4307 {
4308         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4309                         sizeof(struct if6_iter_state)))
4310                 return -ENOMEM;
4311         return 0;
4312 }
4313
4314 static void __net_exit if6_proc_net_exit(struct net *net)
4315 {
4316         remove_proc_entry("if_inet6", net->proc_net);
4317 }
4318
4319 static struct pernet_operations if6_proc_net_ops = {
4320         .init = if6_proc_net_init,
4321         .exit = if6_proc_net_exit,
4322 };
4323
4324 int __init if6_proc_init(void)
4325 {
4326         return register_pernet_subsys(&if6_proc_net_ops);
4327 }
4328
4329 void if6_proc_exit(void)
4330 {
4331         unregister_pernet_subsys(&if6_proc_net_ops);
4332 }
4333 #endif  /* CONFIG_PROC_FS */
4334
4335 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4336 /* Check if address is a home address configured on any interface. */
4337 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4338 {
4339         unsigned int hash = inet6_addr_hash(net, addr);
4340         struct inet6_ifaddr *ifp = NULL;
4341         int ret = 0;
4342
4343         rcu_read_lock();
4344         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4345                 if (!net_eq(dev_net(ifp->idev->dev), net))
4346                         continue;
4347                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4348                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4349                         ret = 1;
4350                         break;
4351                 }
4352         }
4353         rcu_read_unlock();
4354         return ret;
4355 }
4356 #endif
4357
4358 /*
4359  *      Periodic address status verification
4360  */
4361
4362 static void addrconf_verify_rtnl(void)
4363 {
4364         unsigned long now, next, next_sec, next_sched;
4365         struct inet6_ifaddr *ifp;
4366         int i;
4367
4368         ASSERT_RTNL();
4369
4370         rcu_read_lock_bh();
4371         now = jiffies;
4372         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4373
4374         cancel_delayed_work(&addr_chk_work);
4375
4376         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4377 restart:
4378                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4379                         unsigned long age;
4380
4381                         /* When setting preferred_lft to a value not zero or
4382                          * infinity, while valid_lft is infinity
4383                          * IFA_F_PERMANENT has a non-infinity life time.
4384                          */
4385                         if ((ifp->flags & IFA_F_PERMANENT) &&
4386                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4387                                 continue;
4388
4389                         spin_lock(&ifp->lock);
4390                         /* We try to batch several events at once. */
4391                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4392
4393                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4394                             age >= ifp->valid_lft) {
4395                                 spin_unlock(&ifp->lock);
4396                                 in6_ifa_hold(ifp);
4397                                 ipv6_del_addr(ifp);
4398                                 goto restart;
4399                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4400                                 spin_unlock(&ifp->lock);
4401                                 continue;
4402                         } else if (age >= ifp->prefered_lft) {
4403                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4404                                 int deprecate = 0;
4405
4406                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4407                                         deprecate = 1;
4408                                         ifp->flags |= IFA_F_DEPRECATED;
4409                                 }
4410
4411                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4412                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4413                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4414
4415                                 spin_unlock(&ifp->lock);
4416
4417                                 if (deprecate) {
4418                                         in6_ifa_hold(ifp);
4419
4420                                         ipv6_ifa_notify(0, ifp);
4421                                         in6_ifa_put(ifp);
4422                                         goto restart;
4423                                 }
4424                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4425                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4426                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4427                                         ifp->idev->cnf.dad_transmits *
4428                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4429
4430                                 if (age >= ifp->prefered_lft - regen_advance) {
4431                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4432                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4433                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4434                                         if (!ifp->regen_count && ifpub) {
4435                                                 ifp->regen_count++;
4436                                                 in6_ifa_hold(ifp);
4437                                                 in6_ifa_hold(ifpub);
4438                                                 spin_unlock(&ifp->lock);
4439
4440                                                 spin_lock(&ifpub->lock);
4441                                                 ifpub->regen_count = 0;
4442                                                 spin_unlock(&ifpub->lock);
4443                                                 rcu_read_unlock_bh();
4444                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4445                                                 in6_ifa_put(ifpub);
4446                                                 in6_ifa_put(ifp);
4447                                                 rcu_read_lock_bh();
4448                                                 goto restart;
4449                                         }
4450                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4451                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4452                                 spin_unlock(&ifp->lock);
4453                         } else {
4454                                 /* ifp->prefered_lft <= ifp->valid_lft */
4455                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4456                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4457                                 spin_unlock(&ifp->lock);
4458                         }
4459                 }
4460         }
4461
4462         next_sec = round_jiffies_up(next);
4463         next_sched = next;
4464
4465         /* If rounded timeout is accurate enough, accept it. */
4466         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4467                 next_sched = next_sec;
4468
4469         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4470         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4471                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4472
4473         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4474                  now, next, next_sec, next_sched);
4475         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4476         rcu_read_unlock_bh();
4477 }
4478
4479 static void addrconf_verify_work(struct work_struct *w)
4480 {
4481         rtnl_lock();
4482         addrconf_verify_rtnl();
4483         rtnl_unlock();
4484 }
4485
4486 static void addrconf_verify(void)
4487 {
4488         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4489 }
4490
4491 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4492                                      struct in6_addr **peer_pfx)
4493 {
4494         struct in6_addr *pfx = NULL;
4495
4496         *peer_pfx = NULL;
4497
4498         if (addr)
4499                 pfx = nla_data(addr);
4500
4501         if (local) {
4502                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4503                         *peer_pfx = pfx;
4504                 pfx = nla_data(local);
4505         }
4506
4507         return pfx;
4508 }
4509
4510 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4511         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4512         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4513         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4514         [IFA_FLAGS]             = { .len = sizeof(u32) },
4515         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4516         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4517 };
4518
4519 static int
4520 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4521                   struct netlink_ext_ack *extack)
4522 {
4523         struct net *net = sock_net(skb->sk);
4524         struct ifaddrmsg *ifm;
4525         struct nlattr *tb[IFA_MAX+1];
4526         struct in6_addr *pfx, *peer_pfx;
4527         u32 ifa_flags;
4528         int err;
4529
4530         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4531                           extack);
4532         if (err < 0)
4533                 return err;
4534
4535         ifm = nlmsg_data(nlh);
4536         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4537         if (!pfx)
4538                 return -EINVAL;
4539
4540         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4541
4542         /* We ignore other flags so far. */
4543         ifa_flags &= IFA_F_MANAGETEMPADDR;
4544
4545         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4546                               ifm->ifa_prefixlen);
4547 }
4548
4549 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4550                                unsigned long expires, u32 flags)
4551 {
4552         struct fib6_info *f6i;
4553         u32 prio;
4554
4555         f6i = addrconf_get_prefix_route(&ifp->addr,
4556                                         ifp->prefix_len,
4557                                         ifp->idev->dev,
4558                                         0, RTF_GATEWAY | RTF_DEFAULT);
4559         if (!f6i)
4560                 return -ENOENT;
4561
4562         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4563         if (f6i->fib6_metric != prio) {
4564                 /* delete old one */
4565                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4566
4567                 /* add new one */
4568                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4569                                       ifp->rt_priority, ifp->idev->dev,
4570                                       expires, flags, GFP_KERNEL);
4571         } else {
4572                 if (!expires)
4573                         fib6_clean_expires(f6i);
4574                 else
4575                         fib6_set_expires(f6i, expires);
4576
4577                 fib6_info_release(f6i);
4578         }
4579
4580         return 0;
4581 }
4582
4583 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4584 {
4585         u32 flags;
4586         clock_t expires;
4587         unsigned long timeout;
4588         bool was_managetempaddr;
4589         bool had_prefixroute;
4590
4591         ASSERT_RTNL();
4592
4593         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4594                 return -EINVAL;
4595
4596         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4597             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4598                 return -EINVAL;
4599
4600         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4601                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4602
4603         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4604         if (addrconf_finite_timeout(timeout)) {
4605                 expires = jiffies_to_clock_t(timeout * HZ);
4606                 cfg->valid_lft = timeout;
4607                 flags = RTF_EXPIRES;
4608         } else {
4609                 expires = 0;
4610                 flags = 0;
4611                 cfg->ifa_flags |= IFA_F_PERMANENT;
4612         }
4613
4614         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4615         if (addrconf_finite_timeout(timeout)) {
4616                 if (timeout == 0)
4617                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4618                 cfg->preferred_lft = timeout;
4619         }
4620
4621         spin_lock_bh(&ifp->lock);
4622         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4623         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4624                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4625         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4626                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4627                         IFA_F_NOPREFIXROUTE);
4628         ifp->flags |= cfg->ifa_flags;
4629         ifp->tstamp = jiffies;
4630         ifp->valid_lft = cfg->valid_lft;
4631         ifp->prefered_lft = cfg->preferred_lft;
4632
4633         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4634                 ifp->rt_priority = cfg->rt_priority;
4635
4636         spin_unlock_bh(&ifp->lock);
4637         if (!(ifp->flags&IFA_F_TENTATIVE))
4638                 ipv6_ifa_notify(0, ifp);
4639
4640         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4641                 int rc = -ENOENT;
4642
4643                 if (had_prefixroute)
4644                         rc = modify_prefix_route(ifp, expires, flags);
4645
4646                 /* prefix route could have been deleted; if so restore it */
4647                 if (rc == -ENOENT) {
4648                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4649                                               ifp->rt_priority, ifp->idev->dev,
4650                                               expires, flags, GFP_KERNEL);
4651                 }
4652         } else if (had_prefixroute) {
4653                 enum cleanup_prefix_rt_t action;
4654                 unsigned long rt_expires;
4655
4656                 write_lock_bh(&ifp->idev->lock);
4657                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4658                 write_unlock_bh(&ifp->idev->lock);
4659
4660                 if (action != CLEANUP_PREFIX_RT_NOP) {
4661                         cleanup_prefix_route(ifp, rt_expires,
4662                                 action == CLEANUP_PREFIX_RT_DEL);
4663                 }
4664         }
4665
4666         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4667                 if (was_managetempaddr &&
4668                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4669                         cfg->valid_lft = 0;
4670                         cfg->preferred_lft = 0;
4671                 }
4672                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4673                                  cfg->preferred_lft, !was_managetempaddr,
4674                                  jiffies);
4675         }
4676
4677         addrconf_verify_rtnl();
4678
4679         return 0;
4680 }
4681
4682 static int
4683 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4684                   struct netlink_ext_ack *extack)
4685 {
4686         struct net *net = sock_net(skb->sk);
4687         struct ifaddrmsg *ifm;
4688         struct nlattr *tb[IFA_MAX+1];
4689         struct in6_addr *peer_pfx;
4690         struct inet6_ifaddr *ifa;
4691         struct net_device *dev;
4692         struct inet6_dev *idev;
4693         struct ifa6_config cfg;
4694         int err;
4695
4696         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4697                           extack);
4698         if (err < 0)
4699                 return err;
4700
4701         memset(&cfg, 0, sizeof(cfg));
4702
4703         ifm = nlmsg_data(nlh);
4704         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4705         if (!cfg.pfx)
4706                 return -EINVAL;
4707
4708         cfg.peer_pfx = peer_pfx;
4709         cfg.plen = ifm->ifa_prefixlen;
4710         if (tb[IFA_RT_PRIORITY])
4711                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4712
4713         cfg.valid_lft = INFINITY_LIFE_TIME;
4714         cfg.preferred_lft = INFINITY_LIFE_TIME;
4715
4716         if (tb[IFA_CACHEINFO]) {
4717                 struct ifa_cacheinfo *ci;
4718
4719                 ci = nla_data(tb[IFA_CACHEINFO]);
4720                 cfg.valid_lft = ci->ifa_valid;
4721                 cfg.preferred_lft = ci->ifa_prefered;
4722         }
4723
4724         dev =  __dev_get_by_index(net, ifm->ifa_index);
4725         if (!dev)
4726                 return -ENODEV;
4727
4728         if (tb[IFA_FLAGS])
4729                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4730         else
4731                 cfg.ifa_flags = ifm->ifa_flags;
4732
4733         /* We ignore other flags so far. */
4734         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4735                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4736                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4737
4738         idev = ipv6_find_idev(dev);
4739         if (IS_ERR(idev))
4740                 return PTR_ERR(idev);
4741
4742         if (!ipv6_allow_optimistic_dad(net, idev))
4743                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4744
4745         if (cfg.ifa_flags & IFA_F_NODAD &&
4746             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4747                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4748                 return -EINVAL;
4749         }
4750
4751         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4752         if (!ifa) {
4753                 /*
4754                  * It would be best to check for !NLM_F_CREATE here but
4755                  * userspace already relies on not having to provide this.
4756                  */
4757                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4758         }
4759
4760         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4761             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4762                 err = -EEXIST;
4763         else
4764                 err = inet6_addr_modify(ifa, &cfg);
4765
4766         in6_ifa_put(ifa);
4767
4768         return err;
4769 }
4770
4771 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4772                           u8 scope, int ifindex)
4773 {
4774         struct ifaddrmsg *ifm;
4775
4776         ifm = nlmsg_data(nlh);
4777         ifm->ifa_family = AF_INET6;
4778         ifm->ifa_prefixlen = prefixlen;
4779         ifm->ifa_flags = flags;
4780         ifm->ifa_scope = scope;
4781         ifm->ifa_index = ifindex;
4782 }
4783
4784 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4785                          unsigned long tstamp, u32 preferred, u32 valid)
4786 {
4787         struct ifa_cacheinfo ci;
4788
4789         ci.cstamp = cstamp_delta(cstamp);
4790         ci.tstamp = cstamp_delta(tstamp);
4791         ci.ifa_prefered = preferred;
4792         ci.ifa_valid = valid;
4793
4794         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4795 }
4796
4797 static inline int rt_scope(int ifa_scope)
4798 {
4799         if (ifa_scope & IFA_HOST)
4800                 return RT_SCOPE_HOST;
4801         else if (ifa_scope & IFA_LINK)
4802                 return RT_SCOPE_LINK;
4803         else if (ifa_scope & IFA_SITE)
4804                 return RT_SCOPE_SITE;
4805         else
4806                 return RT_SCOPE_UNIVERSE;
4807 }
4808
4809 static inline int inet6_ifaddr_msgsize(void)
4810 {
4811         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4812                + nla_total_size(16) /* IFA_LOCAL */
4813                + nla_total_size(16) /* IFA_ADDRESS */
4814                + nla_total_size(sizeof(struct ifa_cacheinfo))
4815                + nla_total_size(4)  /* IFA_FLAGS */
4816                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4817 }
4818
4819 enum addr_type_t {
4820         UNICAST_ADDR,
4821         MULTICAST_ADDR,
4822         ANYCAST_ADDR,
4823 };
4824
4825 struct inet6_fill_args {
4826         u32 portid;
4827         u32 seq;
4828         int event;
4829         unsigned int flags;
4830         int netnsid;
4831         int ifindex;
4832         enum addr_type_t type;
4833 };
4834
4835 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4836                              struct inet6_fill_args *args)
4837 {
4838         struct nlmsghdr  *nlh;
4839         u32 preferred, valid;
4840
4841         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4842                         sizeof(struct ifaddrmsg), args->flags);
4843         if (!nlh)
4844                 return -EMSGSIZE;
4845
4846         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4847                       ifa->idev->dev->ifindex);
4848
4849         if (args->netnsid >= 0 &&
4850             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4851                 goto error;
4852
4853         if (!((ifa->flags&IFA_F_PERMANENT) &&
4854               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4855                 preferred = ifa->prefered_lft;
4856                 valid = ifa->valid_lft;
4857                 if (preferred != INFINITY_LIFE_TIME) {
4858                         long tval = (jiffies - ifa->tstamp)/HZ;
4859                         if (preferred > tval)
4860                                 preferred -= tval;
4861                         else
4862                                 preferred = 0;
4863                         if (valid != INFINITY_LIFE_TIME) {
4864                                 if (valid > tval)
4865                                         valid -= tval;
4866                                 else
4867                                         valid = 0;
4868                         }
4869                 }
4870         } else {
4871                 preferred = INFINITY_LIFE_TIME;
4872                 valid = INFINITY_LIFE_TIME;
4873         }
4874
4875         if (!ipv6_addr_any(&ifa->peer_addr)) {
4876                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4877                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4878                         goto error;
4879         } else
4880                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4881                         goto error;
4882
4883         if (ifa->rt_priority &&
4884             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4885                 goto error;
4886
4887         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4888                 goto error;
4889
4890         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4891                 goto error;
4892
4893         nlmsg_end(skb, nlh);
4894         return 0;
4895
4896 error:
4897         nlmsg_cancel(skb, nlh);
4898         return -EMSGSIZE;
4899 }
4900
4901 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4902                                struct inet6_fill_args *args)
4903 {
4904         struct nlmsghdr  *nlh;
4905         u8 scope = RT_SCOPE_UNIVERSE;
4906         int ifindex = ifmca->idev->dev->ifindex;
4907
4908         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4909                 scope = RT_SCOPE_SITE;
4910
4911         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4912                         sizeof(struct ifaddrmsg), args->flags);
4913         if (!nlh)
4914                 return -EMSGSIZE;
4915
4916         if (args->netnsid >= 0 &&
4917             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4918                 return -EMSGSIZE;
4919
4920         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4921         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4922             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4923                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4924                 nlmsg_cancel(skb, nlh);
4925                 return -EMSGSIZE;
4926         }
4927
4928         nlmsg_end(skb, nlh);
4929         return 0;
4930 }
4931
4932 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4933                                struct inet6_fill_args *args)
4934 {
4935         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4936         int ifindex = dev ? dev->ifindex : 1;
4937         struct nlmsghdr  *nlh;
4938         u8 scope = RT_SCOPE_UNIVERSE;
4939
4940         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4941                 scope = RT_SCOPE_SITE;
4942
4943         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4944                         sizeof(struct ifaddrmsg), args->flags);
4945         if (!nlh)
4946                 return -EMSGSIZE;
4947
4948         if (args->netnsid >= 0 &&
4949             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4950                 return -EMSGSIZE;
4951
4952         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4953         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4954             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4955                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4956                 nlmsg_cancel(skb, nlh);
4957                 return -EMSGSIZE;
4958         }
4959
4960         nlmsg_end(skb, nlh);
4961         return 0;
4962 }
4963
4964 /* called with rcu_read_lock() */
4965 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4966                           struct netlink_callback *cb, int s_ip_idx,
4967                           struct inet6_fill_args *fillargs)
4968 {
4969         struct ifmcaddr6 *ifmca;
4970         struct ifacaddr6 *ifaca;
4971         int ip_idx = 0;
4972         int err = 1;
4973
4974         read_lock_bh(&idev->lock);
4975         switch (fillargs->type) {
4976         case UNICAST_ADDR: {
4977                 struct inet6_ifaddr *ifa;
4978                 fillargs->event = RTM_NEWADDR;
4979
4980                 /* unicast address incl. temp addr */
4981                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4982                         if (ip_idx < s_ip_idx)
4983                                 goto next;
4984                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
4985                         if (err < 0)
4986                                 break;
4987                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4988 next:
4989                         ip_idx++;
4990                 }
4991                 break;
4992         }
4993         case MULTICAST_ADDR:
4994                 fillargs->event = RTM_GETMULTICAST;
4995
4996                 /* multicast address */
4997                 for (ifmca = idev->mc_list; ifmca;
4998                      ifmca = ifmca->next, ip_idx++) {
4999                         if (ip_idx < s_ip_idx)
5000                                 continue;
5001                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5002                         if (err < 0)
5003                                 break;
5004                 }
5005                 break;
5006         case ANYCAST_ADDR:
5007                 fillargs->event = RTM_GETANYCAST;
5008                 /* anycast address */
5009                 for (ifaca = idev->ac_list; ifaca;
5010                      ifaca = ifaca->aca_next, ip_idx++) {
5011                         if (ip_idx < s_ip_idx)
5012                                 continue;
5013                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5014                         if (err < 0)
5015                                 break;
5016                 }
5017                 break;
5018         default:
5019                 break;
5020         }
5021         read_unlock_bh(&idev->lock);
5022         cb->args[2] = ip_idx;
5023         return err;
5024 }
5025
5026 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5027                                        struct inet6_fill_args *fillargs,
5028                                        struct net **tgt_net, struct sock *sk,
5029                                        struct netlink_callback *cb)
5030 {
5031         struct netlink_ext_ack *extack = cb->extack;
5032         struct nlattr *tb[IFA_MAX+1];
5033         struct ifaddrmsg *ifm;
5034         int err, i;
5035
5036         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5037                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5038                 return -EINVAL;
5039         }
5040
5041         ifm = nlmsg_data(nlh);
5042         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5043                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5044                 return -EINVAL;
5045         }
5046
5047         fillargs->ifindex = ifm->ifa_index;
5048         if (fillargs->ifindex) {
5049                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5050                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5051         }
5052
5053         err = nlmsg_parse_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5054                                  ifa_ipv6_policy, extack);
5055         if (err < 0)
5056                 return err;
5057
5058         for (i = 0; i <= IFA_MAX; ++i) {
5059                 if (!tb[i])
5060                         continue;
5061
5062                 if (i == IFA_TARGET_NETNSID) {
5063                         struct net *net;
5064
5065                         fillargs->netnsid = nla_get_s32(tb[i]);
5066                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5067                         if (IS_ERR(net)) {
5068                                 fillargs->netnsid = -1;
5069                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5070                                 return PTR_ERR(net);
5071                         }
5072                         *tgt_net = net;
5073                 } else {
5074                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5075                         return -EINVAL;
5076                 }
5077         }
5078
5079         return 0;
5080 }
5081
5082 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5083                            enum addr_type_t type)
5084 {
5085         const struct nlmsghdr *nlh = cb->nlh;
5086         struct inet6_fill_args fillargs = {
5087                 .portid = NETLINK_CB(cb->skb).portid,
5088                 .seq = cb->nlh->nlmsg_seq,
5089                 .flags = NLM_F_MULTI,
5090                 .netnsid = -1,
5091                 .type = type,
5092         };
5093         struct net *net = sock_net(skb->sk);
5094         struct net *tgt_net = net;
5095         int idx, s_idx, s_ip_idx;
5096         int h, s_h;
5097         struct net_device *dev;
5098         struct inet6_dev *idev;
5099         struct hlist_head *head;
5100         int err = 0;
5101
5102         s_h = cb->args[0];
5103         s_idx = idx = cb->args[1];
5104         s_ip_idx = cb->args[2];
5105
5106         if (cb->strict_check) {
5107                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5108                                                   skb->sk, cb);
5109                 if (err < 0)
5110                         goto put_tgt_net;
5111
5112                 err = 0;
5113                 if (fillargs.ifindex) {
5114                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5115                         if (!dev) {
5116                                 err = -ENODEV;
5117                                 goto put_tgt_net;
5118                         }
5119                         idev = __in6_dev_get(dev);
5120                         if (idev) {
5121                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5122                                                      &fillargs);
5123                         }
5124                         goto put_tgt_net;
5125                 }
5126         }
5127
5128         rcu_read_lock();
5129         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5130         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5131                 idx = 0;
5132                 head = &tgt_net->dev_index_head[h];
5133                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5134                         if (idx < s_idx)
5135                                 goto cont;
5136                         if (h > s_h || idx > s_idx)
5137                                 s_ip_idx = 0;
5138                         idev = __in6_dev_get(dev);
5139                         if (!idev)
5140                                 goto cont;
5141
5142                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5143                                            &fillargs) < 0)
5144                                 goto done;
5145 cont:
5146                         idx++;
5147                 }
5148         }
5149 done:
5150         rcu_read_unlock();
5151         cb->args[0] = h;
5152         cb->args[1] = idx;
5153 put_tgt_net:
5154         if (fillargs.netnsid >= 0)
5155                 put_net(tgt_net);
5156
5157         return err < 0 ? err : skb->len;
5158 }
5159
5160 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5161 {
5162         enum addr_type_t type = UNICAST_ADDR;
5163
5164         return inet6_dump_addr(skb, cb, type);
5165 }
5166
5167 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5168 {
5169         enum addr_type_t type = MULTICAST_ADDR;
5170
5171         return inet6_dump_addr(skb, cb, type);
5172 }
5173
5174
5175 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5176 {
5177         enum addr_type_t type = ANYCAST_ADDR;
5178
5179         return inet6_dump_addr(skb, cb, type);
5180 }
5181
5182 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5183                              struct netlink_ext_ack *extack)
5184 {
5185         struct net *net = sock_net(in_skb->sk);
5186         struct inet6_fill_args fillargs = {
5187                 .portid = NETLINK_CB(in_skb).portid,
5188                 .seq = nlh->nlmsg_seq,
5189                 .event = RTM_NEWADDR,
5190                 .flags = 0,
5191                 .netnsid = -1,
5192         };
5193         struct net *tgt_net = net;
5194         struct ifaddrmsg *ifm;
5195         struct nlattr *tb[IFA_MAX+1];
5196         struct in6_addr *addr = NULL, *peer;
5197         struct net_device *dev = NULL;
5198         struct inet6_ifaddr *ifa;
5199         struct sk_buff *skb;
5200         int err;
5201
5202         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5203                           extack);
5204         if (err < 0)
5205                 return err;
5206
5207         if (tb[IFA_TARGET_NETNSID]) {
5208                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5209
5210                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5211                                                   fillargs.netnsid);
5212                 if (IS_ERR(tgt_net))
5213                         return PTR_ERR(tgt_net);
5214         }
5215
5216         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5217         if (!addr)
5218                 return -EINVAL;
5219
5220         ifm = nlmsg_data(nlh);
5221         if (ifm->ifa_index)
5222                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5223
5224         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5225         if (!ifa) {
5226                 err = -EADDRNOTAVAIL;
5227                 goto errout;
5228         }
5229
5230         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5231         if (!skb) {
5232                 err = -ENOBUFS;
5233                 goto errout_ifa;
5234         }
5235
5236         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5237         if (err < 0) {
5238                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5239                 WARN_ON(err == -EMSGSIZE);
5240                 kfree_skb(skb);
5241                 goto errout_ifa;
5242         }
5243         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5244 errout_ifa:
5245         in6_ifa_put(ifa);
5246 errout:
5247         if (dev)
5248                 dev_put(dev);
5249         if (fillargs.netnsid >= 0)
5250                 put_net(tgt_net);
5251
5252         return err;
5253 }
5254
5255 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5256 {
5257         struct sk_buff *skb;
5258         struct net *net = dev_net(ifa->idev->dev);
5259         struct inet6_fill_args fillargs = {
5260                 .portid = 0,
5261                 .seq = 0,
5262                 .event = event,
5263                 .flags = 0,
5264                 .netnsid = -1,
5265         };
5266         int err = -ENOBUFS;
5267
5268         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5269         if (!skb)
5270                 goto errout;
5271
5272         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5273         if (err < 0) {
5274                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5275                 WARN_ON(err == -EMSGSIZE);
5276                 kfree_skb(skb);
5277                 goto errout;
5278         }
5279         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5280         return;
5281 errout:
5282         if (err < 0)
5283                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5284 }
5285
5286 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5287                                 __s32 *array, int bytes)
5288 {
5289         BUG_ON(bytes < (DEVCONF_MAX * 4));
5290
5291         memset(array, 0, bytes);
5292         array[DEVCONF_FORWARDING] = cnf->forwarding;
5293         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5294         array[DEVCONF_MTU6] = cnf->mtu6;
5295         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5296         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5297         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5298         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5299         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5300         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5301                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5302         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5303                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5304         array[DEVCONF_RTR_SOLICIT_DELAY] =
5305                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5306         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5307         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5308                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5309         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5310                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5311         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5312         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5313         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5314         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5315         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5316         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5317         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5318         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5319         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5320 #ifdef CONFIG_IPV6_ROUTER_PREF
5321         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5322         array[DEVCONF_RTR_PROBE_INTERVAL] =
5323                 jiffies_to_msecs(cnf->rtr_probe_interval);
5324 #ifdef CONFIG_IPV6_ROUTE_INFO
5325         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5326         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5327 #endif
5328 #endif
5329         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5330         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5331 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5332         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5333         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5334 #endif
5335 #ifdef CONFIG_IPV6_MROUTE
5336         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5337 #endif
5338         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5339         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5340         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5341         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5342         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5343         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5344         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5345         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5346         /* we omit DEVCONF_STABLE_SECRET for now */
5347         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5348         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5349         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5350         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5351         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5352 #ifdef CONFIG_IPV6_SEG6_HMAC
5353         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5354 #endif
5355         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5356         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5357         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5358         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5359 }
5360
5361 static inline size_t inet6_ifla6_size(void)
5362 {
5363         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5364              + nla_total_size(sizeof(struct ifla_cacheinfo))
5365              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5366              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5367              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5368              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5369              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5370              + 0;
5371 }
5372
5373 static inline size_t inet6_if_nlmsg_size(void)
5374 {
5375         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5376                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5377                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5378                + nla_total_size(4) /* IFLA_MTU */
5379                + nla_total_size(4) /* IFLA_LINK */
5380                + nla_total_size(1) /* IFLA_OPERSTATE */
5381                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5382 }
5383
5384 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5385                                         int bytes)
5386 {
5387         int i;
5388         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5389         BUG_ON(pad < 0);
5390
5391         /* Use put_unaligned() because stats may not be aligned for u64. */
5392         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5393         for (i = 1; i < ICMP6_MIB_MAX; i++)
5394                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5395
5396         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5397 }
5398
5399 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5400                                         int bytes, size_t syncpoff)
5401 {
5402         int i, c;
5403         u64 buff[IPSTATS_MIB_MAX];
5404         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5405
5406         BUG_ON(pad < 0);
5407
5408         memset(buff, 0, sizeof(buff));
5409         buff[0] = IPSTATS_MIB_MAX;
5410
5411         for_each_possible_cpu(c) {
5412                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5413                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5414         }
5415
5416         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5417         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5418 }
5419
5420 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5421                              int bytes)
5422 {
5423         switch (attrtype) {
5424         case IFLA_INET6_STATS:
5425                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5426                                      offsetof(struct ipstats_mib, syncp));
5427                 break;
5428         case IFLA_INET6_ICMP6STATS:
5429                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5430                 break;
5431         }
5432 }
5433
5434 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5435                                   u32 ext_filter_mask)
5436 {
5437         struct nlattr *nla;
5438         struct ifla_cacheinfo ci;
5439
5440         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5441                 goto nla_put_failure;
5442         ci.max_reasm_len = IPV6_MAXPLEN;
5443         ci.tstamp = cstamp_delta(idev->tstamp);
5444         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5445         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5446         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5447                 goto nla_put_failure;
5448         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5449         if (!nla)
5450                 goto nla_put_failure;
5451         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5452
5453         /* XXX - MC not implemented */
5454
5455         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5456                 return 0;
5457
5458         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5459         if (!nla)
5460                 goto nla_put_failure;
5461         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5462
5463         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5464         if (!nla)
5465                 goto nla_put_failure;
5466         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5467
5468         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5469         if (!nla)
5470                 goto nla_put_failure;
5471
5472         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5473                 goto nla_put_failure;
5474
5475         read_lock_bh(&idev->lock);
5476         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5477         read_unlock_bh(&idev->lock);
5478
5479         return 0;
5480
5481 nla_put_failure:
5482         return -EMSGSIZE;
5483 }
5484
5485 static size_t inet6_get_link_af_size(const struct net_device *dev,
5486                                      u32 ext_filter_mask)
5487 {
5488         if (!__in6_dev_get(dev))
5489                 return 0;
5490
5491         return inet6_ifla6_size();
5492 }
5493
5494 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5495                               u32 ext_filter_mask)
5496 {
5497         struct inet6_dev *idev = __in6_dev_get(dev);
5498
5499         if (!idev)
5500                 return -ENODATA;
5501
5502         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5503                 return -EMSGSIZE;
5504
5505         return 0;
5506 }
5507
5508 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5509 {
5510         struct inet6_ifaddr *ifp;
5511         struct net_device *dev = idev->dev;
5512         bool clear_token, update_rs = false;
5513         struct in6_addr ll_addr;
5514
5515         ASSERT_RTNL();
5516
5517         if (!token)
5518                 return -EINVAL;
5519         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5520                 return -EINVAL;
5521         if (!ipv6_accept_ra(idev))
5522                 return -EINVAL;
5523         if (idev->cnf.rtr_solicits == 0)
5524                 return -EINVAL;
5525
5526         write_lock_bh(&idev->lock);
5527
5528         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5529         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5530
5531         write_unlock_bh(&idev->lock);
5532
5533         clear_token = ipv6_addr_any(token);
5534         if (clear_token)
5535                 goto update_lft;
5536
5537         if (!idev->dead && (idev->if_flags & IF_READY) &&
5538             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5539                              IFA_F_OPTIMISTIC)) {
5540                 /* If we're not ready, then normal ifup will take care
5541                  * of this. Otherwise, we need to request our rs here.
5542                  */
5543                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5544                 update_rs = true;
5545         }
5546
5547 update_lft:
5548         write_lock_bh(&idev->lock);
5549
5550         if (update_rs) {
5551                 idev->if_flags |= IF_RS_SENT;
5552                 idev->rs_interval = rfc3315_s14_backoff_init(
5553                         idev->cnf.rtr_solicit_interval);
5554                 idev->rs_probes = 1;
5555                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5556         }
5557
5558         /* Well, that's kinda nasty ... */
5559         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5560                 spin_lock(&ifp->lock);
5561                 if (ifp->tokenized) {
5562                         ifp->valid_lft = 0;
5563                         ifp->prefered_lft = 0;
5564                 }
5565                 spin_unlock(&ifp->lock);
5566         }
5567
5568         write_unlock_bh(&idev->lock);
5569         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5570         addrconf_verify_rtnl();
5571         return 0;
5572 }
5573
5574 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5575         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5576         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5577 };
5578
5579 static int inet6_validate_link_af(const struct net_device *dev,
5580                                   const struct nlattr *nla)
5581 {
5582         struct nlattr *tb[IFLA_INET6_MAX + 1];
5583
5584         if (dev && !__in6_dev_get(dev))
5585                 return -EAFNOSUPPORT;
5586
5587         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5588                                 NULL);
5589 }
5590
5591 static int check_addr_gen_mode(int mode)
5592 {
5593         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5594             mode != IN6_ADDR_GEN_MODE_NONE &&
5595             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5596             mode != IN6_ADDR_GEN_MODE_RANDOM)
5597                 return -EINVAL;
5598         return 1;
5599 }
5600
5601 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5602                                 int mode)
5603 {
5604         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5605             !idev->cnf.stable_secret.initialized &&
5606             !net->ipv6.devconf_dflt->stable_secret.initialized)
5607                 return -EINVAL;
5608         return 1;
5609 }
5610
5611 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5612 {
5613         int err = -EINVAL;
5614         struct inet6_dev *idev = __in6_dev_get(dev);
5615         struct nlattr *tb[IFLA_INET6_MAX + 1];
5616
5617         if (!idev)
5618                 return -EAFNOSUPPORT;
5619
5620         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5621                 BUG();
5622
5623         if (tb[IFLA_INET6_TOKEN]) {
5624                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5625                 if (err)
5626                         return err;
5627         }
5628
5629         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5630                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5631
5632                 if (check_addr_gen_mode(mode) < 0 ||
5633                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5634                         return -EINVAL;
5635
5636                 idev->cnf.addr_gen_mode = mode;
5637                 err = 0;
5638         }
5639
5640         return err;
5641 }
5642
5643 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5644                              u32 portid, u32 seq, int event, unsigned int flags)
5645 {
5646         struct net_device *dev = idev->dev;
5647         struct ifinfomsg *hdr;
5648         struct nlmsghdr *nlh;
5649         void *protoinfo;
5650
5651         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5652         if (!nlh)
5653                 return -EMSGSIZE;
5654
5655         hdr = nlmsg_data(nlh);
5656         hdr->ifi_family = AF_INET6;
5657         hdr->__ifi_pad = 0;
5658         hdr->ifi_type = dev->type;
5659         hdr->ifi_index = dev->ifindex;
5660         hdr->ifi_flags = dev_get_flags(dev);
5661         hdr->ifi_change = 0;
5662
5663         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5664             (dev->addr_len &&
5665              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5666             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5667             (dev->ifindex != dev_get_iflink(dev) &&
5668              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5669             nla_put_u8(skb, IFLA_OPERSTATE,
5670                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5671                 goto nla_put_failure;
5672         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5673         if (!protoinfo)
5674                 goto nla_put_failure;
5675
5676         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5677                 goto nla_put_failure;
5678
5679         nla_nest_end(skb, protoinfo);
5680         nlmsg_end(skb, nlh);
5681         return 0;
5682
5683 nla_put_failure:
5684         nlmsg_cancel(skb, nlh);
5685         return -EMSGSIZE;
5686 }
5687
5688 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5689                                    struct netlink_ext_ack *extack)
5690 {
5691         struct ifinfomsg *ifm;
5692
5693         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5694                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5695                 return -EINVAL;
5696         }
5697
5698         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5699                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5700                 return -EINVAL;
5701         }
5702
5703         ifm = nlmsg_data(nlh);
5704         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5705             ifm->ifi_change || ifm->ifi_index) {
5706                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5707                 return -EINVAL;
5708         }
5709
5710         return 0;
5711 }
5712
5713 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5714 {
5715         struct net *net = sock_net(skb->sk);
5716         int h, s_h;
5717         int idx = 0, s_idx;
5718         struct net_device *dev;
5719         struct inet6_dev *idev;
5720         struct hlist_head *head;
5721
5722         /* only requests using strict checking can pass data to
5723          * influence the dump
5724          */
5725         if (cb->strict_check) {
5726                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5727
5728                 if (err < 0)
5729                         return err;
5730         }
5731
5732         s_h = cb->args[0];
5733         s_idx = cb->args[1];
5734
5735         rcu_read_lock();
5736         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5737                 idx = 0;
5738                 head = &net->dev_index_head[h];
5739                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5740                         if (idx < s_idx)
5741                                 goto cont;
5742                         idev = __in6_dev_get(dev);
5743                         if (!idev)
5744                                 goto cont;
5745                         if (inet6_fill_ifinfo(skb, idev,
5746                                               NETLINK_CB(cb->skb).portid,
5747                                               cb->nlh->nlmsg_seq,
5748                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5749                                 goto out;
5750 cont:
5751                         idx++;
5752                 }
5753         }
5754 out:
5755         rcu_read_unlock();
5756         cb->args[1] = idx;
5757         cb->args[0] = h;
5758
5759         return skb->len;
5760 }
5761
5762 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5763 {
5764         struct sk_buff *skb;
5765         struct net *net = dev_net(idev->dev);
5766         int err = -ENOBUFS;
5767
5768         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5769         if (!skb)
5770                 goto errout;
5771
5772         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5773         if (err < 0) {
5774                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5775                 WARN_ON(err == -EMSGSIZE);
5776                 kfree_skb(skb);
5777                 goto errout;
5778         }
5779         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5780         return;
5781 errout:
5782         if (err < 0)
5783                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5784 }
5785
5786 static inline size_t inet6_prefix_nlmsg_size(void)
5787 {
5788         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5789                + nla_total_size(sizeof(struct in6_addr))
5790                + nla_total_size(sizeof(struct prefix_cacheinfo));
5791 }
5792
5793 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5794                              struct prefix_info *pinfo, u32 portid, u32 seq,
5795                              int event, unsigned int flags)
5796 {
5797         struct prefixmsg *pmsg;
5798         struct nlmsghdr *nlh;
5799         struct prefix_cacheinfo ci;
5800
5801         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5802         if (!nlh)
5803                 return -EMSGSIZE;
5804
5805         pmsg = nlmsg_data(nlh);
5806         pmsg->prefix_family = AF_INET6;
5807         pmsg->prefix_pad1 = 0;
5808         pmsg->prefix_pad2 = 0;
5809         pmsg->prefix_ifindex = idev->dev->ifindex;
5810         pmsg->prefix_len = pinfo->prefix_len;
5811         pmsg->prefix_type = pinfo->type;
5812         pmsg->prefix_pad3 = 0;
5813         pmsg->prefix_flags = 0;
5814         if (pinfo->onlink)
5815                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5816         if (pinfo->autoconf)
5817                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5818
5819         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5820                 goto nla_put_failure;
5821         ci.preferred_time = ntohl(pinfo->prefered);
5822         ci.valid_time = ntohl(pinfo->valid);
5823         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5824                 goto nla_put_failure;
5825         nlmsg_end(skb, nlh);
5826         return 0;
5827
5828 nla_put_failure:
5829         nlmsg_cancel(skb, nlh);
5830         return -EMSGSIZE;
5831 }
5832
5833 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5834                          struct prefix_info *pinfo)
5835 {
5836         struct sk_buff *skb;
5837         struct net *net = dev_net(idev->dev);
5838         int err = -ENOBUFS;
5839
5840         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5841         if (!skb)
5842                 goto errout;
5843
5844         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5845         if (err < 0) {
5846                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5847                 WARN_ON(err == -EMSGSIZE);
5848                 kfree_skb(skb);
5849                 goto errout;
5850         }
5851         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5852         return;
5853 errout:
5854         if (err < 0)
5855                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5856 }
5857
5858 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5859 {
5860         struct net *net = dev_net(ifp->idev->dev);
5861
5862         if (event)
5863                 ASSERT_RTNL();
5864
5865         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5866
5867         switch (event) {
5868         case RTM_NEWADDR:
5869                 /*
5870                  * If the address was optimistic
5871                  * we inserted the route at the start of
5872                  * our DAD process, so we don't need
5873                  * to do it again
5874                  */
5875                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5876                         ip6_ins_rt(net, ifp->rt);
5877                 if (ifp->idev->cnf.forwarding)
5878                         addrconf_join_anycast(ifp);
5879                 if (!ipv6_addr_any(&ifp->peer_addr))
5880                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5881                                               ifp->idev->dev, 0, 0,
5882                                               GFP_ATOMIC);
5883                 break;
5884         case RTM_DELADDR:
5885                 if (ifp->idev->cnf.forwarding)
5886                         addrconf_leave_anycast(ifp);
5887                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5888                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5889                         struct fib6_info *rt;
5890
5891                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5892                                                        ifp->idev->dev, 0, 0);
5893                         if (rt)
5894                                 ip6_del_rt(net, rt);
5895                 }
5896                 if (ifp->rt) {
5897                         ip6_del_rt(net, ifp->rt);
5898                         ifp->rt = NULL;
5899                 }
5900                 rt_genid_bump_ipv6(net);
5901                 break;
5902         }
5903         atomic_inc(&net->ipv6.dev_addr_genid);
5904 }
5905
5906 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5907 {
5908         rcu_read_lock_bh();
5909         if (likely(ifp->idev->dead == 0))
5910                 __ipv6_ifa_notify(event, ifp);
5911         rcu_read_unlock_bh();
5912 }
5913
5914 #ifdef CONFIG_SYSCTL
5915
5916 static
5917 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5918                            void __user *buffer, size_t *lenp, loff_t *ppos)
5919 {
5920         int *valp = ctl->data;
5921         int val = *valp;
5922         loff_t pos = *ppos;
5923         struct ctl_table lctl;
5924         int ret;
5925
5926         /*
5927          * ctl->data points to idev->cnf.forwarding, we should
5928          * not modify it until we get the rtnl lock.
5929          */
5930         lctl = *ctl;
5931         lctl.data = &val;
5932
5933         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5934
5935         if (write)
5936                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5937         if (ret)
5938                 *ppos = pos;
5939         return ret;
5940 }
5941
5942 static
5943 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5944                         void __user *buffer, size_t *lenp, loff_t *ppos)
5945 {
5946         struct inet6_dev *idev = ctl->extra1;
5947         int min_mtu = IPV6_MIN_MTU;
5948         struct ctl_table lctl;
5949
5950         lctl = *ctl;
5951         lctl.extra1 = &min_mtu;
5952         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5953
5954         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5955 }
5956
5957 static void dev_disable_change(struct inet6_dev *idev)
5958 {
5959         struct netdev_notifier_info info;
5960
5961         if (!idev || !idev->dev)
5962                 return;
5963
5964         netdev_notifier_info_init(&info, idev->dev);
5965         if (idev->cnf.disable_ipv6)
5966                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5967         else
5968                 addrconf_notify(NULL, NETDEV_UP, &info);
5969 }
5970
5971 static void addrconf_disable_change(struct net *net, __s32 newf)
5972 {
5973         struct net_device *dev;
5974         struct inet6_dev *idev;
5975
5976         for_each_netdev(net, dev) {
5977                 idev = __in6_dev_get(dev);
5978                 if (idev) {
5979                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5980                         idev->cnf.disable_ipv6 = newf;
5981                         if (changed)
5982                                 dev_disable_change(idev);
5983                 }
5984         }
5985 }
5986
5987 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5988 {
5989         struct net *net;
5990         int old;
5991
5992         if (!rtnl_trylock())
5993                 return restart_syscall();
5994
5995         net = (struct net *)table->extra2;
5996         old = *p;
5997         *p = newf;
5998
5999         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6000                 rtnl_unlock();
6001                 return 0;
6002         }
6003
6004         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6005                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6006                 addrconf_disable_change(net, newf);
6007         } else if ((!newf) ^ (!old))
6008                 dev_disable_change((struct inet6_dev *)table->extra1);
6009
6010         rtnl_unlock();
6011         return 0;
6012 }
6013
6014 static
6015 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6016                             void __user *buffer, size_t *lenp, loff_t *ppos)
6017 {
6018         int *valp = ctl->data;
6019         int val = *valp;
6020         loff_t pos = *ppos;
6021         struct ctl_table lctl;
6022         int ret;
6023
6024         /*
6025          * ctl->data points to idev->cnf.disable_ipv6, we should
6026          * not modify it until we get the rtnl lock.
6027          */
6028         lctl = *ctl;
6029         lctl.data = &val;
6030
6031         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6032
6033         if (write)
6034                 ret = addrconf_disable_ipv6(ctl, valp, val);
6035         if (ret)
6036                 *ppos = pos;
6037         return ret;
6038 }
6039
6040 static
6041 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6042                               void __user *buffer, size_t *lenp, loff_t *ppos)
6043 {
6044         int *valp = ctl->data;
6045         int ret;
6046         int old, new;
6047
6048         old = *valp;
6049         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6050         new = *valp;
6051
6052         if (write && old != new) {
6053                 struct net *net = ctl->extra2;
6054
6055                 if (!rtnl_trylock())
6056                         return restart_syscall();
6057
6058                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6059                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6060                                                      NETCONFA_PROXY_NEIGH,
6061                                                      NETCONFA_IFINDEX_DEFAULT,
6062                                                      net->ipv6.devconf_dflt);
6063                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6064                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6065                                                      NETCONFA_PROXY_NEIGH,
6066                                                      NETCONFA_IFINDEX_ALL,
6067                                                      net->ipv6.devconf_all);
6068                 else {
6069                         struct inet6_dev *idev = ctl->extra1;
6070
6071                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6072                                                      NETCONFA_PROXY_NEIGH,
6073                                                      idev->dev->ifindex,
6074                                                      &idev->cnf);
6075                 }
6076                 rtnl_unlock();
6077         }
6078
6079         return ret;
6080 }
6081
6082 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6083                                          void __user *buffer, size_t *lenp,
6084                                          loff_t *ppos)
6085 {
6086         int ret = 0;
6087         u32 new_val;
6088         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6089         struct net *net = (struct net *)ctl->extra2;
6090         struct ctl_table tmp = {
6091                 .data = &new_val,
6092                 .maxlen = sizeof(new_val),
6093                 .mode = ctl->mode,
6094         };
6095
6096         if (!rtnl_trylock())
6097                 return restart_syscall();
6098
6099         new_val = *((u32 *)ctl->data);
6100
6101         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6102         if (ret != 0)
6103                 goto out;
6104
6105         if (write) {
6106                 if (check_addr_gen_mode(new_val) < 0) {
6107                         ret = -EINVAL;
6108                         goto out;
6109                 }
6110
6111                 if (idev) {
6112                         if (check_stable_privacy(idev, net, new_val) < 0) {
6113                                 ret = -EINVAL;
6114                                 goto out;
6115                         }
6116
6117                         if (idev->cnf.addr_gen_mode != new_val) {
6118                                 idev->cnf.addr_gen_mode = new_val;
6119                                 addrconf_dev_config(idev->dev);
6120                         }
6121                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6122                         struct net_device *dev;
6123
6124                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6125                         for_each_netdev(net, dev) {
6126                                 idev = __in6_dev_get(dev);
6127                                 if (idev &&
6128                                     idev->cnf.addr_gen_mode != new_val) {
6129                                         idev->cnf.addr_gen_mode = new_val;
6130                                         addrconf_dev_config(idev->dev);
6131                                 }
6132                         }
6133                 }
6134
6135                 *((u32 *)ctl->data) = new_val;
6136         }
6137
6138 out:
6139         rtnl_unlock();
6140
6141         return ret;
6142 }
6143
6144 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6145                                          void __user *buffer, size_t *lenp,
6146                                          loff_t *ppos)
6147 {
6148         int err;
6149         struct in6_addr addr;
6150         char str[IPV6_MAX_STRLEN];
6151         struct ctl_table lctl = *ctl;
6152         struct net *net = ctl->extra2;
6153         struct ipv6_stable_secret *secret = ctl->data;
6154
6155         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6156                 return -EIO;
6157
6158         lctl.maxlen = IPV6_MAX_STRLEN;
6159         lctl.data = str;
6160
6161         if (!rtnl_trylock())
6162                 return restart_syscall();
6163
6164         if (!write && !secret->initialized) {
6165                 err = -EIO;
6166                 goto out;
6167         }
6168
6169         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6170         if (err >= sizeof(str)) {
6171                 err = -EIO;
6172                 goto out;
6173         }
6174
6175         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6176         if (err || !write)
6177                 goto out;
6178
6179         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6180                 err = -EIO;
6181                 goto out;
6182         }
6183
6184         secret->initialized = true;
6185         secret->secret = addr;
6186
6187         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6188                 struct net_device *dev;
6189
6190                 for_each_netdev(net, dev) {
6191                         struct inet6_dev *idev = __in6_dev_get(dev);
6192
6193                         if (idev) {
6194                                 idev->cnf.addr_gen_mode =
6195                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6196                         }
6197                 }
6198         } else {
6199                 struct inet6_dev *idev = ctl->extra1;
6200
6201                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6202         }
6203
6204 out:
6205         rtnl_unlock();
6206
6207         return err;
6208 }
6209
6210 static
6211 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6212                                                 int write,
6213                                                 void __user *buffer,
6214                                                 size_t *lenp,
6215                                                 loff_t *ppos)
6216 {
6217         int *valp = ctl->data;
6218         int val = *valp;
6219         loff_t pos = *ppos;
6220         struct ctl_table lctl;
6221         int ret;
6222
6223         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6224          * we should not modify it until we get the rtnl lock.
6225          */
6226         lctl = *ctl;
6227         lctl.data = &val;
6228
6229         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6230
6231         if (write)
6232                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6233         if (ret)
6234                 *ppos = pos;
6235         return ret;
6236 }
6237
6238 static
6239 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6240 {
6241         if (rt) {
6242                 if (action)
6243                         rt->dst.flags |= DST_NOPOLICY;
6244                 else
6245                         rt->dst.flags &= ~DST_NOPOLICY;
6246         }
6247 }
6248
6249 static
6250 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6251 {
6252         struct inet6_ifaddr *ifa;
6253
6254         read_lock_bh(&idev->lock);
6255         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6256                 spin_lock(&ifa->lock);
6257                 if (ifa->rt) {
6258                         struct fib6_info *rt = ifa->rt;
6259                         int cpu;
6260
6261                         rcu_read_lock();
6262                         ifa->rt->dst_nopolicy = val ? true : false;
6263                         if (rt->rt6i_pcpu) {
6264                                 for_each_possible_cpu(cpu) {
6265                                         struct rt6_info **rtp;
6266
6267                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6268                                         addrconf_set_nopolicy(*rtp, val);
6269                                 }
6270                         }
6271                         rcu_read_unlock();
6272                 }
6273                 spin_unlock(&ifa->lock);
6274         }
6275         read_unlock_bh(&idev->lock);
6276 }
6277
6278 static
6279 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6280 {
6281         struct inet6_dev *idev;
6282         struct net *net;
6283
6284         if (!rtnl_trylock())
6285                 return restart_syscall();
6286
6287         *valp = val;
6288
6289         net = (struct net *)ctl->extra2;
6290         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6291                 rtnl_unlock();
6292                 return 0;
6293         }
6294
6295         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6296                 struct net_device *dev;
6297
6298                 for_each_netdev(net, dev) {
6299                         idev = __in6_dev_get(dev);
6300                         if (idev)
6301                                 addrconf_disable_policy_idev(idev, val);
6302                 }
6303         } else {
6304                 idev = (struct inet6_dev *)ctl->extra1;
6305                 addrconf_disable_policy_idev(idev, val);
6306         }
6307
6308         rtnl_unlock();
6309         return 0;
6310 }
6311
6312 static
6313 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6314                                    void __user *buffer, size_t *lenp,
6315                                    loff_t *ppos)
6316 {
6317         int *valp = ctl->data;
6318         int val = *valp;
6319         loff_t pos = *ppos;
6320         struct ctl_table lctl;
6321         int ret;
6322
6323         lctl = *ctl;
6324         lctl.data = &val;
6325         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6326
6327         if (write && (*valp != val))
6328                 ret = addrconf_disable_policy(ctl, valp, val);
6329
6330         if (ret)
6331                 *ppos = pos;
6332
6333         return ret;
6334 }
6335
6336 static int minus_one = -1;
6337 static const int zero = 0;
6338 static const int one = 1;
6339 static const int two_five_five = 255;
6340
6341 static const struct ctl_table addrconf_sysctl[] = {
6342         {
6343                 .procname       = "forwarding",
6344                 .data           = &ipv6_devconf.forwarding,
6345                 .maxlen         = sizeof(int),
6346                 .mode           = 0644,
6347                 .proc_handler   = addrconf_sysctl_forward,
6348         },
6349         {
6350                 .procname       = "hop_limit",
6351                 .data           = &ipv6_devconf.hop_limit,
6352                 .maxlen         = sizeof(int),
6353                 .mode           = 0644,
6354                 .proc_handler   = proc_dointvec_minmax,
6355                 .extra1         = (void *)&one,
6356                 .extra2         = (void *)&two_five_five,
6357         },
6358         {
6359                 .procname       = "mtu",
6360                 .data           = &ipv6_devconf.mtu6,
6361                 .maxlen         = sizeof(int),
6362                 .mode           = 0644,
6363                 .proc_handler   = addrconf_sysctl_mtu,
6364         },
6365         {
6366                 .procname       = "accept_ra",
6367                 .data           = &ipv6_devconf.accept_ra,
6368                 .maxlen         = sizeof(int),
6369                 .mode           = 0644,
6370                 .proc_handler   = proc_dointvec,
6371         },
6372         {
6373                 .procname       = "accept_redirects",
6374                 .data           = &ipv6_devconf.accept_redirects,
6375                 .maxlen         = sizeof(int),
6376                 .mode           = 0644,
6377                 .proc_handler   = proc_dointvec,
6378         },
6379         {
6380                 .procname       = "autoconf",
6381                 .data           = &ipv6_devconf.autoconf,
6382                 .maxlen         = sizeof(int),
6383                 .mode           = 0644,
6384                 .proc_handler   = proc_dointvec,
6385         },
6386         {
6387                 .procname       = "dad_transmits",
6388                 .data           = &ipv6_devconf.dad_transmits,
6389                 .maxlen         = sizeof(int),
6390                 .mode           = 0644,
6391                 .proc_handler   = proc_dointvec,
6392         },
6393         {
6394                 .procname       = "router_solicitations",
6395                 .data           = &ipv6_devconf.rtr_solicits,
6396                 .maxlen         = sizeof(int),
6397                 .mode           = 0644,
6398                 .proc_handler   = proc_dointvec_minmax,
6399                 .extra1         = &minus_one,
6400         },
6401         {
6402                 .procname       = "router_solicitation_interval",
6403                 .data           = &ipv6_devconf.rtr_solicit_interval,
6404                 .maxlen         = sizeof(int),
6405                 .mode           = 0644,
6406                 .proc_handler   = proc_dointvec_jiffies,
6407         },
6408         {
6409                 .procname       = "router_solicitation_max_interval",
6410                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6411                 .maxlen         = sizeof(int),
6412                 .mode           = 0644,
6413                 .proc_handler   = proc_dointvec_jiffies,
6414         },
6415         {
6416                 .procname       = "router_solicitation_delay",
6417                 .data           = &ipv6_devconf.rtr_solicit_delay,
6418                 .maxlen         = sizeof(int),
6419                 .mode           = 0644,
6420                 .proc_handler   = proc_dointvec_jiffies,
6421         },
6422         {
6423                 .procname       = "force_mld_version",
6424                 .data           = &ipv6_devconf.force_mld_version,
6425                 .maxlen         = sizeof(int),
6426                 .mode           = 0644,
6427                 .proc_handler   = proc_dointvec,
6428         },
6429         {
6430                 .procname       = "mldv1_unsolicited_report_interval",
6431                 .data           =
6432                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6433                 .maxlen         = sizeof(int),
6434                 .mode           = 0644,
6435                 .proc_handler   = proc_dointvec_ms_jiffies,
6436         },
6437         {
6438                 .procname       = "mldv2_unsolicited_report_interval",
6439                 .data           =
6440                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6441                 .maxlen         = sizeof(int),
6442                 .mode           = 0644,
6443                 .proc_handler   = proc_dointvec_ms_jiffies,
6444         },
6445         {
6446                 .procname       = "use_tempaddr",
6447                 .data           = &ipv6_devconf.use_tempaddr,
6448                 .maxlen         = sizeof(int),
6449                 .mode           = 0644,
6450                 .proc_handler   = proc_dointvec,
6451         },
6452         {
6453                 .procname       = "temp_valid_lft",
6454                 .data           = &ipv6_devconf.temp_valid_lft,
6455                 .maxlen         = sizeof(int),
6456                 .mode           = 0644,
6457                 .proc_handler   = proc_dointvec,
6458         },
6459         {
6460                 .procname       = "temp_prefered_lft",
6461                 .data           = &ipv6_devconf.temp_prefered_lft,
6462                 .maxlen         = sizeof(int),
6463                 .mode           = 0644,
6464                 .proc_handler   = proc_dointvec,
6465         },
6466         {
6467                 .procname       = "regen_max_retry",
6468                 .data           = &ipv6_devconf.regen_max_retry,
6469                 .maxlen         = sizeof(int),
6470                 .mode           = 0644,
6471                 .proc_handler   = proc_dointvec,
6472         },
6473         {
6474                 .procname       = "max_desync_factor",
6475                 .data           = &ipv6_devconf.max_desync_factor,
6476                 .maxlen         = sizeof(int),
6477                 .mode           = 0644,
6478                 .proc_handler   = proc_dointvec,
6479         },
6480         {
6481                 .procname       = "max_addresses",
6482                 .data           = &ipv6_devconf.max_addresses,
6483                 .maxlen         = sizeof(int),
6484                 .mode           = 0644,
6485                 .proc_handler   = proc_dointvec,
6486         },
6487         {
6488                 .procname       = "accept_ra_defrtr",
6489                 .data           = &ipv6_devconf.accept_ra_defrtr,
6490                 .maxlen         = sizeof(int),
6491                 .mode           = 0644,
6492                 .proc_handler   = proc_dointvec,
6493         },
6494         {
6495                 .procname       = "accept_ra_min_hop_limit",
6496                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6497                 .maxlen         = sizeof(int),
6498                 .mode           = 0644,
6499                 .proc_handler   = proc_dointvec,
6500         },
6501         {
6502                 .procname       = "accept_ra_pinfo",
6503                 .data           = &ipv6_devconf.accept_ra_pinfo,
6504                 .maxlen         = sizeof(int),
6505                 .mode           = 0644,
6506                 .proc_handler   = proc_dointvec,
6507         },
6508 #ifdef CONFIG_IPV6_ROUTER_PREF
6509         {
6510                 .procname       = "accept_ra_rtr_pref",
6511                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6512                 .maxlen         = sizeof(int),
6513                 .mode           = 0644,
6514                 .proc_handler   = proc_dointvec,
6515         },
6516         {
6517                 .procname       = "router_probe_interval",
6518                 .data           = &ipv6_devconf.rtr_probe_interval,
6519                 .maxlen         = sizeof(int),
6520                 .mode           = 0644,
6521                 .proc_handler   = proc_dointvec_jiffies,
6522         },
6523 #ifdef CONFIG_IPV6_ROUTE_INFO
6524         {
6525                 .procname       = "accept_ra_rt_info_min_plen",
6526                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6527                 .maxlen         = sizeof(int),
6528                 .mode           = 0644,
6529                 .proc_handler   = proc_dointvec,
6530         },
6531         {
6532                 .procname       = "accept_ra_rt_info_max_plen",
6533                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6534                 .maxlen         = sizeof(int),
6535                 .mode           = 0644,
6536                 .proc_handler   = proc_dointvec,
6537         },
6538 #endif
6539 #endif
6540         {
6541                 .procname       = "proxy_ndp",
6542                 .data           = &ipv6_devconf.proxy_ndp,
6543                 .maxlen         = sizeof(int),
6544                 .mode           = 0644,
6545                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6546         },
6547         {
6548                 .procname       = "accept_source_route",
6549                 .data           = &ipv6_devconf.accept_source_route,
6550                 .maxlen         = sizeof(int),
6551                 .mode           = 0644,
6552                 .proc_handler   = proc_dointvec,
6553         },
6554 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6555         {
6556                 .procname       = "optimistic_dad",
6557                 .data           = &ipv6_devconf.optimistic_dad,
6558                 .maxlen         = sizeof(int),
6559                 .mode           = 0644,
6560                 .proc_handler   = proc_dointvec,
6561         },
6562         {
6563                 .procname       = "use_optimistic",
6564                 .data           = &ipv6_devconf.use_optimistic,
6565                 .maxlen         = sizeof(int),
6566                 .mode           = 0644,
6567                 .proc_handler   = proc_dointvec,
6568         },
6569 #endif
6570 #ifdef CONFIG_IPV6_MROUTE
6571         {
6572                 .procname       = "mc_forwarding",
6573                 .data           = &ipv6_devconf.mc_forwarding,
6574                 .maxlen         = sizeof(int),
6575                 .mode           = 0444,
6576                 .proc_handler   = proc_dointvec,
6577         },
6578 #endif
6579         {
6580                 .procname       = "disable_ipv6",
6581                 .data           = &ipv6_devconf.disable_ipv6,
6582                 .maxlen         = sizeof(int),
6583                 .mode           = 0644,
6584                 .proc_handler   = addrconf_sysctl_disable,
6585         },
6586         {
6587                 .procname       = "accept_dad",
6588                 .data           = &ipv6_devconf.accept_dad,
6589                 .maxlen         = sizeof(int),
6590                 .mode           = 0644,
6591                 .proc_handler   = proc_dointvec,
6592         },
6593         {
6594                 .procname       = "force_tllao",
6595                 .data           = &ipv6_devconf.force_tllao,
6596                 .maxlen         = sizeof(int),
6597                 .mode           = 0644,
6598                 .proc_handler   = proc_dointvec
6599         },
6600         {
6601                 .procname       = "ndisc_notify",
6602                 .data           = &ipv6_devconf.ndisc_notify,
6603                 .maxlen         = sizeof(int),
6604                 .mode           = 0644,
6605                 .proc_handler   = proc_dointvec
6606         },
6607         {
6608                 .procname       = "suppress_frag_ndisc",
6609                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6610                 .maxlen         = sizeof(int),
6611                 .mode           = 0644,
6612                 .proc_handler   = proc_dointvec
6613         },
6614         {
6615                 .procname       = "accept_ra_from_local",
6616                 .data           = &ipv6_devconf.accept_ra_from_local,
6617                 .maxlen         = sizeof(int),
6618                 .mode           = 0644,
6619                 .proc_handler   = proc_dointvec,
6620         },
6621         {
6622                 .procname       = "accept_ra_mtu",
6623                 .data           = &ipv6_devconf.accept_ra_mtu,
6624                 .maxlen         = sizeof(int),
6625                 .mode           = 0644,
6626                 .proc_handler   = proc_dointvec,
6627         },
6628         {
6629                 .procname       = "stable_secret",
6630                 .data           = &ipv6_devconf.stable_secret,
6631                 .maxlen         = IPV6_MAX_STRLEN,
6632                 .mode           = 0600,
6633                 .proc_handler   = addrconf_sysctl_stable_secret,
6634         },
6635         {
6636                 .procname       = "use_oif_addrs_only",
6637                 .data           = &ipv6_devconf.use_oif_addrs_only,
6638                 .maxlen         = sizeof(int),
6639                 .mode           = 0644,
6640                 .proc_handler   = proc_dointvec,
6641         },
6642         {
6643                 .procname       = "ignore_routes_with_linkdown",
6644                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6645                 .maxlen         = sizeof(int),
6646                 .mode           = 0644,
6647                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6648         },
6649         {
6650                 .procname       = "drop_unicast_in_l2_multicast",
6651                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6652                 .maxlen         = sizeof(int),
6653                 .mode           = 0644,
6654                 .proc_handler   = proc_dointvec,
6655         },
6656         {
6657                 .procname       = "drop_unsolicited_na",
6658                 .data           = &ipv6_devconf.drop_unsolicited_na,
6659                 .maxlen         = sizeof(int),
6660                 .mode           = 0644,
6661                 .proc_handler   = proc_dointvec,
6662         },
6663         {
6664                 .procname       = "keep_addr_on_down",
6665                 .data           = &ipv6_devconf.keep_addr_on_down,
6666                 .maxlen         = sizeof(int),
6667                 .mode           = 0644,
6668                 .proc_handler   = proc_dointvec,
6669
6670         },
6671         {
6672                 .procname       = "seg6_enabled",
6673                 .data           = &ipv6_devconf.seg6_enabled,
6674                 .maxlen         = sizeof(int),
6675                 .mode           = 0644,
6676                 .proc_handler   = proc_dointvec,
6677         },
6678 #ifdef CONFIG_IPV6_SEG6_HMAC
6679         {
6680                 .procname       = "seg6_require_hmac",
6681                 .data           = &ipv6_devconf.seg6_require_hmac,
6682                 .maxlen         = sizeof(int),
6683                 .mode           = 0644,
6684                 .proc_handler   = proc_dointvec,
6685         },
6686 #endif
6687         {
6688                 .procname       = "enhanced_dad",
6689                 .data           = &ipv6_devconf.enhanced_dad,
6690                 .maxlen         = sizeof(int),
6691                 .mode           = 0644,
6692                 .proc_handler   = proc_dointvec,
6693         },
6694         {
6695                 .procname               = "addr_gen_mode",
6696                 .data                   = &ipv6_devconf.addr_gen_mode,
6697                 .maxlen                 = sizeof(int),
6698                 .mode                   = 0644,
6699                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6700         },
6701         {
6702                 .procname       = "disable_policy",
6703                 .data           = &ipv6_devconf.disable_policy,
6704                 .maxlen         = sizeof(int),
6705                 .mode           = 0644,
6706                 .proc_handler   = addrconf_sysctl_disable_policy,
6707         },
6708         {
6709                 .procname       = "ndisc_tclass",
6710                 .data           = &ipv6_devconf.ndisc_tclass,
6711                 .maxlen         = sizeof(int),
6712                 .mode           = 0644,
6713                 .proc_handler   = proc_dointvec_minmax,
6714                 .extra1         = (void *)&zero,
6715                 .extra2         = (void *)&two_five_five,
6716         },
6717         {
6718                 /* sentinel */
6719         }
6720 };
6721
6722 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6723                 struct inet6_dev *idev, struct ipv6_devconf *p)
6724 {
6725         int i, ifindex;
6726         struct ctl_table *table;
6727         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6728
6729         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6730         if (!table)
6731                 goto out;
6732
6733         for (i = 0; table[i].data; i++) {
6734                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6735                 /* If one of these is already set, then it is not safe to
6736                  * overwrite either of them: this makes proc_dointvec_minmax
6737                  * usable.
6738                  */
6739                 if (!table[i].extra1 && !table[i].extra2) {
6740                         table[i].extra1 = idev; /* embedded; no ref */
6741                         table[i].extra2 = net;
6742                 }
6743         }
6744
6745         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6746
6747         p->sysctl_header = register_net_sysctl(net, path, table);
6748         if (!p->sysctl_header)
6749                 goto free;
6750
6751         if (!strcmp(dev_name, "all"))
6752                 ifindex = NETCONFA_IFINDEX_ALL;
6753         else if (!strcmp(dev_name, "default"))
6754                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6755         else
6756                 ifindex = idev->dev->ifindex;
6757         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6758                                      ifindex, p);
6759         return 0;
6760
6761 free:
6762         kfree(table);
6763 out:
6764         return -ENOBUFS;
6765 }
6766
6767 static void __addrconf_sysctl_unregister(struct net *net,
6768                                          struct ipv6_devconf *p, int ifindex)
6769 {
6770         struct ctl_table *table;
6771
6772         if (!p->sysctl_header)
6773                 return;
6774
6775         table = p->sysctl_header->ctl_table_arg;
6776         unregister_net_sysctl_table(p->sysctl_header);
6777         p->sysctl_header = NULL;
6778         kfree(table);
6779
6780         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6781 }
6782
6783 static int addrconf_sysctl_register(struct inet6_dev *idev)
6784 {
6785         int err;
6786
6787         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6788                 return -EINVAL;
6789
6790         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6791                                     &ndisc_ifinfo_sysctl_change);
6792         if (err)
6793                 return err;
6794         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6795                                          idev, &idev->cnf);
6796         if (err)
6797                 neigh_sysctl_unregister(idev->nd_parms);
6798
6799         return err;
6800 }
6801
6802 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6803 {
6804         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6805                                      idev->dev->ifindex);
6806         neigh_sysctl_unregister(idev->nd_parms);
6807 }
6808
6809
6810 #endif
6811
6812 static int __net_init addrconf_init_net(struct net *net)
6813 {
6814         int err = -ENOMEM;
6815         struct ipv6_devconf *all, *dflt;
6816
6817         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6818         if (!all)
6819                 goto err_alloc_all;
6820
6821         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6822         if (!dflt)
6823                 goto err_alloc_dflt;
6824
6825         /* these will be inherited by all namespaces */
6826         dflt->autoconf = ipv6_defaults.autoconf;
6827         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6828
6829         dflt->stable_secret.initialized = false;
6830         all->stable_secret.initialized = false;
6831
6832         net->ipv6.devconf_all = all;
6833         net->ipv6.devconf_dflt = dflt;
6834
6835 #ifdef CONFIG_SYSCTL
6836         err = __addrconf_sysctl_register(net, "all", NULL, all);
6837         if (err < 0)
6838                 goto err_reg_all;
6839
6840         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6841         if (err < 0)
6842                 goto err_reg_dflt;
6843 #endif
6844         return 0;
6845
6846 #ifdef CONFIG_SYSCTL
6847 err_reg_dflt:
6848         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6849 err_reg_all:
6850         kfree(dflt);
6851 #endif
6852 err_alloc_dflt:
6853         kfree(all);
6854 err_alloc_all:
6855         return err;
6856 }
6857
6858 static void __net_exit addrconf_exit_net(struct net *net)
6859 {
6860 #ifdef CONFIG_SYSCTL
6861         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6862                                      NETCONFA_IFINDEX_DEFAULT);
6863         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6864                                      NETCONFA_IFINDEX_ALL);
6865 #endif
6866         kfree(net->ipv6.devconf_dflt);
6867         kfree(net->ipv6.devconf_all);
6868 }
6869
6870 static struct pernet_operations addrconf_ops = {
6871         .init = addrconf_init_net,
6872         .exit = addrconf_exit_net,
6873 };
6874
6875 static struct rtnl_af_ops inet6_ops __read_mostly = {
6876         .family           = AF_INET6,
6877         .fill_link_af     = inet6_fill_link_af,
6878         .get_link_af_size = inet6_get_link_af_size,
6879         .validate_link_af = inet6_validate_link_af,
6880         .set_link_af      = inet6_set_link_af,
6881 };
6882
6883 /*
6884  *      Init / cleanup code
6885  */
6886
6887 int __init addrconf_init(void)
6888 {
6889         struct inet6_dev *idev;
6890         int i, err;
6891
6892         err = ipv6_addr_label_init();
6893         if (err < 0) {
6894                 pr_crit("%s: cannot initialize default policy table: %d\n",
6895                         __func__, err);
6896                 goto out;
6897         }
6898
6899         err = register_pernet_subsys(&addrconf_ops);
6900         if (err < 0)
6901                 goto out_addrlabel;
6902
6903         addrconf_wq = create_workqueue("ipv6_addrconf");
6904         if (!addrconf_wq) {
6905                 err = -ENOMEM;
6906                 goto out_nowq;
6907         }
6908
6909         /* The addrconf netdev notifier requires that loopback_dev
6910          * has it's ipv6 private information allocated and setup
6911          * before it can bring up and give link-local addresses
6912          * to other devices which are up.
6913          *
6914          * Unfortunately, loopback_dev is not necessarily the first
6915          * entry in the global dev_base list of net devices.  In fact,
6916          * it is likely to be the very last entry on that list.
6917          * So this causes the notifier registry below to try and
6918          * give link-local addresses to all devices besides loopback_dev
6919          * first, then loopback_dev, which cases all the non-loopback_dev
6920          * devices to fail to get a link-local address.
6921          *
6922          * So, as a temporary fix, allocate the ipv6 structure for
6923          * loopback_dev first by hand.
6924          * Longer term, all of the dependencies ipv6 has upon the loopback
6925          * device and it being up should be removed.
6926          */
6927         rtnl_lock();
6928         idev = ipv6_add_dev(init_net.loopback_dev);
6929         rtnl_unlock();
6930         if (IS_ERR(idev)) {
6931                 err = PTR_ERR(idev);
6932                 goto errlo;
6933         }
6934
6935         ip6_route_init_special_entries();
6936
6937         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6938                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6939
6940         register_netdevice_notifier(&ipv6_dev_notf);
6941
6942         addrconf_verify();
6943
6944         rtnl_af_register(&inet6_ops);
6945
6946         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6947                                    NULL, inet6_dump_ifinfo, 0);
6948         if (err < 0)
6949                 goto errout;
6950
6951         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6952                                    inet6_rtm_newaddr, NULL, 0);
6953         if (err < 0)
6954                 goto errout;
6955         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6956                                    inet6_rtm_deladdr, NULL, 0);
6957         if (err < 0)
6958                 goto errout;
6959         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6960                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
6961                                    RTNL_FLAG_DOIT_UNLOCKED);
6962         if (err < 0)
6963                 goto errout;
6964         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6965                                    NULL, inet6_dump_ifmcaddr, 0);
6966         if (err < 0)
6967                 goto errout;
6968         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6969                                    NULL, inet6_dump_ifacaddr, 0);
6970         if (err < 0)
6971                 goto errout;
6972         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6973                                    inet6_netconf_get_devconf,
6974                                    inet6_netconf_dump_devconf,
6975                                    RTNL_FLAG_DOIT_UNLOCKED);
6976         if (err < 0)
6977                 goto errout;
6978         err = ipv6_addr_label_rtnl_register();
6979         if (err < 0)
6980                 goto errout;
6981
6982         return 0;
6983 errout:
6984         rtnl_unregister_all(PF_INET6);
6985         rtnl_af_unregister(&inet6_ops);
6986         unregister_netdevice_notifier(&ipv6_dev_notf);
6987 errlo:
6988         destroy_workqueue(addrconf_wq);
6989 out_nowq:
6990         unregister_pernet_subsys(&addrconf_ops);
6991 out_addrlabel:
6992         ipv6_addr_label_cleanup();
6993 out:
6994         return err;
6995 }
6996
6997 void addrconf_cleanup(void)
6998 {
6999         struct net_device *dev;
7000         int i;
7001
7002         unregister_netdevice_notifier(&ipv6_dev_notf);
7003         unregister_pernet_subsys(&addrconf_ops);
7004         ipv6_addr_label_cleanup();
7005
7006         rtnl_af_unregister(&inet6_ops);
7007
7008         rtnl_lock();
7009
7010         /* clean dev list */
7011         for_each_netdev(&init_net, dev) {
7012                 if (__in6_dev_get(dev) == NULL)
7013                         continue;
7014                 addrconf_ifdown(dev, 1);
7015         }
7016         addrconf_ifdown(init_net.loopback_dev, 2);
7017
7018         /*
7019          *      Check hash table.
7020          */
7021         spin_lock_bh(&addrconf_hash_lock);
7022         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7023                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7024         spin_unlock_bh(&addrconf_hash_lock);
7025         cancel_delayed_work(&addr_chk_work);
7026         rtnl_unlock();
7027
7028         destroy_workqueue(addrconf_wq);
7029 }