net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/sched/signal.h>
  47 #include <linux/socket.h>
  48 #include <linux/sockios.h>
  49 #include <linux/net.h>
  50 #include <linux/inet.h>
  51 #include <linux/in6.h>
  52 #include <linux/netdevice.h>
  53 #include <linux/if_addr.h>
  54 #include <linux/if_arp.h>
  55 #include <linux/if_arcnet.h>
  56 #include <linux/if_infiniband.h>
  57 #include <linux/route.h>
  58 #include <linux/inetdevice.h>
  59 #include <linux/init.h>
  60 #include <linux/slab.h>
  61 #ifdef CONFIG_SYSCTL
  62 #include <linux/sysctl.h>
  63 #endif
  64 #include <linux/capability.h>
  65 #include <linux/delay.h>
  66 #include <linux/notifier.h>
  67 #include <linux/string.h>
  68 #include <linux/hash.h>
  69
  70 #include <net/net_namespace.h>
  71 #include <net/sock.h>
  72 #include <net/snmp.h>
  73
  74 #include <net/6lowpan.h>
  75 #include <net/firewire.h>
  76 #include <net/ipv6.h>
  77 #include <net/protocol.h>
  78 #include <net/ndisc.h>
  79 #include <net/ip6_route.h>
  80 #include <net/addrconf.h>
  81 #include <net/tcp.h>
  82 #include <net/ip.h>
  83 #include <net/netlink.h>
  84 #include <net/pkt_sched.h>
  85 #include <net/l3mdev.h>
  86 #include <linux/if_tunnel.h>
  87 #include <linux/rtnetlink.h>
  88 #include <linux/netconf.h>
  89 #include <linux/random.h>
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 #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 (ifa->prefix_len != ifp->prefix_len ||
1169                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1170                                        ifp->prefix_len))
1171                         continue;
1172                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1173                         return CLEANUP_PREFIX_RT_NOP;
1174
1175                 action = CLEANUP_PREFIX_RT_EXPIRE;
1176
1177                 spin_lock(&ifa->lock);
1178
1179                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1180                 /*
1181                  * Note: Because this address is
1182                  * not permanent, lifetime <
1183                  * LONG_MAX / HZ here.
1184                  */
1185                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1186                         *expires = ifa->tstamp + lifetime * HZ;
1187                 spin_unlock(&ifa->lock);
1188         }
1189
1190         return action;
1191 }
1192
1193 static void
1194 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1195 {
1196         struct fib6_info *f6i;
1197
1198         f6i = addrconf_get_prefix_route(&ifp->addr,
1199                                        ifp->prefix_len,
1200                                        ifp->idev->dev,
1201                                        0, RTF_GATEWAY | RTF_DEFAULT);
1202         if (f6i) {
1203                 if (del_rt)
1204                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1205                 else {
1206                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1207                                 fib6_set_expires(f6i, expires);
1208                         fib6_info_release(f6i);
1209                 }
1210         }
1211 }
1212
1213
1214 /* This function wants to get referenced ifp and releases it before return */
1215
1216 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1217 {
1218         int state;
1219         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1220         unsigned long expires;
1221
1222         ASSERT_RTNL();
1223
1224         spin_lock_bh(&ifp->lock);
1225         state = ifp->state;
1226         ifp->state = INET6_IFADDR_STATE_DEAD;
1227         spin_unlock_bh(&ifp->lock);
1228
1229         if (state == INET6_IFADDR_STATE_DEAD)
1230                 goto out;
1231
1232         spin_lock_bh(&addrconf_hash_lock);
1233         hlist_del_init_rcu(&ifp->addr_lst);
1234         spin_unlock_bh(&addrconf_hash_lock);
1235
1236         write_lock_bh(&ifp->idev->lock);
1237
1238         if (ifp->flags&IFA_F_TEMPORARY) {
1239                 list_del(&ifp->tmp_list);
1240                 if (ifp->ifpub) {
1241                         in6_ifa_put(ifp->ifpub);
1242                         ifp->ifpub = NULL;
1243                 }
1244                 __in6_ifa_put(ifp);
1245         }
1246
1247         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1248                 action = check_cleanup_prefix_route(ifp, &expires);
1249
1250         list_del_rcu(&ifp->if_list);
1251         __in6_ifa_put(ifp);
1252
1253         write_unlock_bh(&ifp->idev->lock);
1254
1255         addrconf_del_dad_work(ifp);
1256
1257         ipv6_ifa_notify(RTM_DELADDR, ifp);
1258
1259         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1260
1261         if (action != CLEANUP_PREFIX_RT_NOP) {
1262                 cleanup_prefix_route(ifp, expires,
1263                         action == CLEANUP_PREFIX_RT_DEL);
1264         }
1265
1266         /* clean up prefsrc entries */
1267         rt6_remove_prefsrc(ifp);
1268 out:
1269         in6_ifa_put(ifp);
1270 }
1271
1272 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1273                                 struct inet6_ifaddr *ift,
1274                                 bool block)
1275 {
1276         struct inet6_dev *idev = ifp->idev;
1277         struct in6_addr addr, *tmpaddr;
1278         unsigned long tmp_tstamp, age;
1279         unsigned long regen_advance;
1280         struct ifa6_config cfg;
1281         int ret = 0;
1282         unsigned long now = jiffies;
1283         long max_desync_factor;
1284         s32 cnf_temp_preferred_lft;
1285
1286         write_lock_bh(&idev->lock);
1287         if (ift) {
1288                 spin_lock_bh(&ift->lock);
1289                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1290                 spin_unlock_bh(&ift->lock);
1291                 tmpaddr = &addr;
1292         } else {
1293                 tmpaddr = NULL;
1294         }
1295 retry:
1296         in6_dev_hold(idev);
1297         if (idev->cnf.use_tempaddr <= 0) {
1298                 write_unlock_bh(&idev->lock);
1299                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1300                 in6_dev_put(idev);
1301                 ret = -1;
1302                 goto out;
1303         }
1304         spin_lock_bh(&ifp->lock);
1305         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1306                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1307                 spin_unlock_bh(&ifp->lock);
1308                 write_unlock_bh(&idev->lock);
1309                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1310                         __func__);
1311                 in6_dev_put(idev);
1312                 ret = -1;
1313                 goto out;
1314         }
1315         in6_ifa_hold(ifp);
1316         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1317         ipv6_try_regen_rndid(idev, tmpaddr);
1318         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1319         age = (now - ifp->tstamp) / HZ;
1320
1321         regen_advance = idev->cnf.regen_max_retry *
1322                         idev->cnf.dad_transmits *
1323                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1324
1325         /* recalculate max_desync_factor each time and update
1326          * idev->desync_factor if it's larger
1327          */
1328         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1329         max_desync_factor = min_t(__u32,
1330                                   idev->cnf.max_desync_factor,
1331                                   cnf_temp_preferred_lft - regen_advance);
1332
1333         if (unlikely(idev->desync_factor > max_desync_factor)) {
1334                 if (max_desync_factor > 0) {
1335                         get_random_bytes(&idev->desync_factor,
1336                                          sizeof(idev->desync_factor));
1337                         idev->desync_factor %= max_desync_factor;
1338                 } else {
1339                         idev->desync_factor = 0;
1340                 }
1341         }
1342
1343         memset(&cfg, 0, sizeof(cfg));
1344         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1345                               idev->cnf.temp_valid_lft + age);
1346         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1347         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1348
1349         cfg.plen = ifp->prefix_len;
1350         tmp_tstamp = ifp->tstamp;
1351         spin_unlock_bh(&ifp->lock);
1352
1353         write_unlock_bh(&idev->lock);
1354
1355         /* A temporary address is created only if this calculated Preferred
1356          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1357          * an implementation must not create a temporary address with a zero
1358          * Preferred Lifetime.
1359          * Use age calculation as in addrconf_verify to avoid unnecessary
1360          * temporary addresses being generated.
1361          */
1362         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1363         if (cfg.preferred_lft <= regen_advance + age) {
1364                 in6_ifa_put(ifp);
1365                 in6_dev_put(idev);
1366                 ret = -1;
1367                 goto out;
1368         }
1369
1370         cfg.ifa_flags = IFA_F_TEMPORARY;
1371         /* set in addrconf_prefix_rcv() */
1372         if (ifp->flags & IFA_F_OPTIMISTIC)
1373                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1374
1375         cfg.pfx = &addr;
1376         cfg.scope = ipv6_addr_scope(cfg.pfx);
1377
1378         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1379         if (IS_ERR(ift)) {
1380                 in6_ifa_put(ifp);
1381                 in6_dev_put(idev);
1382                 pr_info("%s: retry temporary address regeneration\n", __func__);
1383                 tmpaddr = &addr;
1384                 write_lock_bh(&idev->lock);
1385                 goto retry;
1386         }
1387
1388         spin_lock_bh(&ift->lock);
1389         ift->ifpub = ifp;
1390         ift->cstamp = now;
1391         ift->tstamp = tmp_tstamp;
1392         spin_unlock_bh(&ift->lock);
1393
1394         addrconf_dad_start(ift);
1395         in6_ifa_put(ift);
1396         in6_dev_put(idev);
1397 out:
1398         return ret;
1399 }
1400
1401 /*
1402  *      Choose an appropriate source address (RFC3484)
1403  */
1404 enum {
1405         IPV6_SADDR_RULE_INIT = 0,
1406         IPV6_SADDR_RULE_LOCAL,
1407         IPV6_SADDR_RULE_SCOPE,
1408         IPV6_SADDR_RULE_PREFERRED,
1409 #ifdef CONFIG_IPV6_MIP6
1410         IPV6_SADDR_RULE_HOA,
1411 #endif
1412         IPV6_SADDR_RULE_OIF,
1413         IPV6_SADDR_RULE_LABEL,
1414         IPV6_SADDR_RULE_PRIVACY,
1415         IPV6_SADDR_RULE_ORCHID,
1416         IPV6_SADDR_RULE_PREFIX,
1417 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1418         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1419 #endif
1420         IPV6_SADDR_RULE_MAX
1421 };
1422
1423 struct ipv6_saddr_score {
1424         int                     rule;
1425         int                     addr_type;
1426         struct inet6_ifaddr     *ifa;
1427         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1428         int                     scopedist;
1429         int                     matchlen;
1430 };
1431
1432 struct ipv6_saddr_dst {
1433         const struct in6_addr *addr;
1434         int ifindex;
1435         int scope;
1436         int label;
1437         unsigned int prefs;
1438 };
1439
1440 static inline int ipv6_saddr_preferred(int type)
1441 {
1442         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1443                 return 1;
1444         return 0;
1445 }
1446
1447 static bool ipv6_use_optimistic_addr(struct net *net,
1448                                      struct inet6_dev *idev)
1449 {
1450 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1451         if (!idev)
1452                 return false;
1453         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1454                 return false;
1455         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1456                 return false;
1457
1458         return true;
1459 #else
1460         return false;
1461 #endif
1462 }
1463
1464 static bool ipv6_allow_optimistic_dad(struct net *net,
1465                                       struct inet6_dev *idev)
1466 {
1467 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1468         if (!idev)
1469                 return false;
1470         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1471                 return false;
1472
1473         return true;
1474 #else
1475         return false;
1476 #endif
1477 }
1478
1479 static int ipv6_get_saddr_eval(struct net *net,
1480                                struct ipv6_saddr_score *score,
1481                                struct ipv6_saddr_dst *dst,
1482                                int i)
1483 {
1484         int ret;
1485
1486         if (i <= score->rule) {
1487                 switch (i) {
1488                 case IPV6_SADDR_RULE_SCOPE:
1489                         ret = score->scopedist;
1490                         break;
1491                 case IPV6_SADDR_RULE_PREFIX:
1492                         ret = score->matchlen;
1493                         break;
1494                 default:
1495                         ret = !!test_bit(i, score->scorebits);
1496                 }
1497                 goto out;
1498         }
1499
1500         switch (i) {
1501         case IPV6_SADDR_RULE_INIT:
1502                 /* Rule 0: remember if hiscore is not ready yet */
1503                 ret = !!score->ifa;
1504                 break;
1505         case IPV6_SADDR_RULE_LOCAL:
1506                 /* Rule 1: Prefer same address */
1507                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1508                 break;
1509         case IPV6_SADDR_RULE_SCOPE:
1510                 /* Rule 2: Prefer appropriate scope
1511                  *
1512                  *      ret
1513                  *       ^
1514                  *    -1 |  d 15
1515                  *    ---+--+-+---> scope
1516                  *       |
1517                  *       |             d is scope of the destination.
1518                  *  B-d  |  \
1519                  *       |   \      <- smaller scope is better if
1520                  *  B-15 |    \        if scope is enough for destination.
1521                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1522                  * d-C-1 | /
1523                  *       |/         <- greater is better
1524                  *   -C  /             if scope is not enough for destination.
1525                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1526                  *
1527                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1528                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1529                  * Assume B = 0 and we get C > 29.
1530                  */
1531                 ret = __ipv6_addr_src_scope(score->addr_type);
1532                 if (ret >= dst->scope)
1533                         ret = -ret;
1534                 else
1535                         ret -= 128;     /* 30 is enough */
1536                 score->scopedist = ret;
1537                 break;
1538         case IPV6_SADDR_RULE_PREFERRED:
1539             {
1540                 /* Rule 3: Avoid deprecated and optimistic addresses */
1541                 u8 avoid = IFA_F_DEPRECATED;
1542
1543                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1544                         avoid |= IFA_F_OPTIMISTIC;
1545                 ret = ipv6_saddr_preferred(score->addr_type) ||
1546                       !(score->ifa->flags & avoid);
1547                 break;
1548             }
1549 #ifdef CONFIG_IPV6_MIP6
1550         case IPV6_SADDR_RULE_HOA:
1551             {
1552                 /* Rule 4: Prefer home address */
1553                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1554                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1555                 break;
1556             }
1557 #endif
1558         case IPV6_SADDR_RULE_OIF:
1559                 /* Rule 5: Prefer outgoing interface */
1560                 ret = (!dst->ifindex ||
1561                        dst->ifindex == score->ifa->idev->dev->ifindex);
1562                 break;
1563         case IPV6_SADDR_RULE_LABEL:
1564                 /* Rule 6: Prefer matching label */
1565                 ret = ipv6_addr_label(net,
1566                                       &score->ifa->addr, score->addr_type,
1567                                       score->ifa->idev->dev->ifindex) == dst->label;
1568                 break;
1569         case IPV6_SADDR_RULE_PRIVACY:
1570             {
1571                 /* Rule 7: Prefer public address
1572                  * Note: prefer temporary address if use_tempaddr >= 2
1573                  */
1574                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1575                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1576                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1577                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1578                 break;
1579             }
1580         case IPV6_SADDR_RULE_ORCHID:
1581                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1582                  *          non-ORCHID vs non-ORCHID
1583                  */
1584                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1585                         ipv6_addr_orchid(dst->addr));
1586                 break;
1587         case IPV6_SADDR_RULE_PREFIX:
1588                 /* Rule 8: Use longest matching prefix */
1589                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1590                 if (ret > score->ifa->prefix_len)
1591                         ret = score->ifa->prefix_len;
1592                 score->matchlen = ret;
1593                 break;
1594 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1595         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1596                 /* Optimistic addresses still have lower precedence than other
1597                  * preferred addresses.
1598                  */
1599                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1600                 break;
1601 #endif
1602         default:
1603                 ret = 0;
1604         }
1605
1606         if (ret)
1607                 __set_bit(i, score->scorebits);
1608         score->rule = i;
1609 out:
1610         return ret;
1611 }
1612
1613 static int __ipv6_dev_get_saddr(struct net *net,
1614                                 struct ipv6_saddr_dst *dst,
1615                                 struct inet6_dev *idev,
1616                                 struct ipv6_saddr_score *scores,
1617                                 int hiscore_idx)
1618 {
1619         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1620
1621         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1622                 int i;
1623
1624                 /*
1625                  * - Tentative Address (RFC2462 section 5.4)
1626                  *  - A tentative address is not considered
1627                  *    "assigned to an interface" in the traditional
1628                  *    sense, unless it is also flagged as optimistic.
1629                  * - Candidate Source Address (section 4)
1630                  *  - In any case, anycast addresses, multicast
1631                  *    addresses, and the unspecified address MUST
1632                  *    NOT be included in a candidate set.
1633                  */
1634                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1635                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1636                         continue;
1637
1638                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1639
1640                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1641                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1642                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1643                                             idev->dev->name);
1644                         continue;
1645                 }
1646
1647                 score->rule = -1;
1648                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1649
1650                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1651                         int minihiscore, miniscore;
1652
1653                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1654                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1655
1656                         if (minihiscore > miniscore) {
1657                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1658                                     score->scopedist > 0) {
1659                                         /*
1660                                          * special case:
1661                                          * each remaining entry
1662                                          * has too small (not enough)
1663                                          * scope, because ifa entries
1664                                          * are sorted by their scope
1665                                          * values.
1666                                          */
1667                                         goto out;
1668                                 }
1669                                 break;
1670                         } else if (minihiscore < miniscore) {
1671                                 swap(hiscore, score);
1672                                 hiscore_idx = 1 - hiscore_idx;
1673
1674                                 /* restore our iterator */
1675                                 score->ifa = hiscore->ifa;
1676
1677                                 break;
1678                         }
1679                 }
1680         }
1681 out:
1682         return hiscore_idx;
1683 }
1684
1685 static int ipv6_get_saddr_master(struct net *net,
1686                                  const struct net_device *dst_dev,
1687                                  const struct net_device *master,
1688                                  struct ipv6_saddr_dst *dst,
1689                                  struct ipv6_saddr_score *scores,
1690                                  int hiscore_idx)
1691 {
1692         struct inet6_dev *idev;
1693
1694         idev = __in6_dev_get(dst_dev);
1695         if (idev)
1696                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1697                                                    scores, hiscore_idx);
1698
1699         idev = __in6_dev_get(master);
1700         if (idev)
1701                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1702                                                    scores, hiscore_idx);
1703
1704         return hiscore_idx;
1705 }
1706
1707 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1708                        const struct in6_addr *daddr, unsigned int prefs,
1709                        struct in6_addr *saddr)
1710 {
1711         struct ipv6_saddr_score scores[2], *hiscore;
1712         struct ipv6_saddr_dst dst;
1713         struct inet6_dev *idev;
1714         struct net_device *dev;
1715         int dst_type;
1716         bool use_oif_addr = false;
1717         int hiscore_idx = 0;
1718         int ret = 0;
1719
1720         dst_type = __ipv6_addr_type(daddr);
1721         dst.addr = daddr;
1722         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1723         dst.scope = __ipv6_addr_src_scope(dst_type);
1724         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1725         dst.prefs = prefs;
1726
1727         scores[hiscore_idx].rule = -1;
1728         scores[hiscore_idx].ifa = NULL;
1729
1730         rcu_read_lock();
1731
1732         /* Candidate Source Address (section 4)
1733          *  - multicast and link-local destination address,
1734          *    the set of candidate source address MUST only
1735          *    include addresses assigned to interfaces
1736          *    belonging to the same link as the outgoing
1737          *    interface.
1738          * (- For site-local destination addresses, the
1739          *    set of candidate source addresses MUST only
1740          *    include addresses assigned to interfaces
1741          *    belonging to the same site as the outgoing
1742          *    interface.)
1743          *  - "It is RECOMMENDED that the candidate source addresses
1744          *    be the set of unicast addresses assigned to the
1745          *    interface that will be used to send to the destination
1746          *    (the 'outgoing' interface)." (RFC 6724)
1747          */
1748         if (dst_dev) {
1749                 idev = __in6_dev_get(dst_dev);
1750                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1751                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1752                     (idev && idev->cnf.use_oif_addrs_only)) {
1753                         use_oif_addr = true;
1754                 }
1755         }
1756
1757         if (use_oif_addr) {
1758                 if (idev)
1759                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1760         } else {
1761                 const struct net_device *master;
1762                 int master_idx = 0;
1763
1764                 /* if dst_dev exists and is enslaved to an L3 device, then
1765                  * prefer addresses from dst_dev and then the master over
1766                  * any other enslaved devices in the L3 domain.
1767                  */
1768                 master = l3mdev_master_dev_rcu(dst_dev);
1769                 if (master) {
1770                         master_idx = master->ifindex;
1771
1772                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1773                                                             master, &dst,
1774                                                             scores, hiscore_idx);
1775
1776                         if (scores[hiscore_idx].ifa)
1777                                 goto out;
1778                 }
1779
1780                 for_each_netdev_rcu(net, dev) {
1781                         /* only consider addresses on devices in the
1782                          * same L3 domain
1783                          */
1784                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1785                                 continue;
1786                         idev = __in6_dev_get(dev);
1787                         if (!idev)
1788                                 continue;
1789                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1790                 }
1791         }
1792
1793 out:
1794         hiscore = &scores[hiscore_idx];
1795         if (!hiscore->ifa)
1796                 ret = -EADDRNOTAVAIL;
1797         else
1798                 *saddr = hiscore->ifa->addr;
1799
1800         rcu_read_unlock();
1801         return ret;
1802 }
1803 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1804
1805 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1806                       u32 banned_flags)
1807 {
1808         struct inet6_ifaddr *ifp;
1809         int err = -EADDRNOTAVAIL;
1810
1811         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1812                 if (ifp->scope > IFA_LINK)
1813                         break;
1814                 if (ifp->scope == IFA_LINK &&
1815                     !(ifp->flags & banned_flags)) {
1816                         *addr = ifp->addr;
1817                         err = 0;
1818                         break;
1819                 }
1820         }
1821         return err;
1822 }
1823
1824 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1825                     u32 banned_flags)
1826 {
1827         struct inet6_dev *idev;
1828         int err = -EADDRNOTAVAIL;
1829
1830         rcu_read_lock();
1831         idev = __in6_dev_get(dev);
1832         if (idev) {
1833                 read_lock_bh(&idev->lock);
1834                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1835                 read_unlock_bh(&idev->lock);
1836         }
1837         rcu_read_unlock();
1838         return err;
1839 }
1840
1841 static int ipv6_count_addresses(const struct inet6_dev *idev)
1842 {
1843         const struct inet6_ifaddr *ifp;
1844         int cnt = 0;
1845
1846         rcu_read_lock();
1847         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1848                 cnt++;
1849         rcu_read_unlock();
1850         return cnt;
1851 }
1852
1853 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1854                   const struct net_device *dev, int strict)
1855 {
1856         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1857                                        strict, IFA_F_TENTATIVE);
1858 }
1859 EXPORT_SYMBOL(ipv6_chk_addr);
1860
1861 /* device argument is used to find the L3 domain of interest. If
1862  * skip_dev_check is set, then the ifp device is not checked against
1863  * the passed in dev argument. So the 2 cases for addresses checks are:
1864  *   1. does the address exist in the L3 domain that dev is part of
1865  *      (skip_dev_check = true), or
1866  *
1867  *   2. does the address exist on the specific device
1868  *      (skip_dev_check = false)
1869  */
1870 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1871                             const struct net_device *dev, bool skip_dev_check,
1872                             int strict, u32 banned_flags)
1873 {
1874         unsigned int hash = inet6_addr_hash(net, addr);
1875         const struct net_device *l3mdev;
1876         struct inet6_ifaddr *ifp;
1877         u32 ifp_flags;
1878
1879         rcu_read_lock();
1880
1881         l3mdev = l3mdev_master_dev_rcu(dev);
1882         if (skip_dev_check)
1883                 dev = NULL;
1884
1885         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1886                 if (!net_eq(dev_net(ifp->idev->dev), net))
1887                         continue;
1888
1889                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1890                         continue;
1891
1892                 /* Decouple optimistic from tentative for evaluation here.
1893                  * Ban optimistic addresses explicitly, when required.
1894                  */
1895                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1896                             ? (ifp->flags&~IFA_F_TENTATIVE)
1897                             : ifp->flags;
1898                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1899                     !(ifp_flags&banned_flags) &&
1900                     (!dev || ifp->idev->dev == dev ||
1901                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1902                         rcu_read_unlock();
1903                         return 1;
1904                 }
1905         }
1906
1907         rcu_read_unlock();
1908         return 0;
1909 }
1910 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1911
1912
1913 /* Compares an address/prefix_len with addresses on device @dev.
1914  * If one is found it returns true.
1915  */
1916 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1917         const unsigned int prefix_len, struct net_device *dev)
1918 {
1919         const struct inet6_ifaddr *ifa;
1920         const struct inet6_dev *idev;
1921         bool ret = false;
1922
1923         rcu_read_lock();
1924         idev = __in6_dev_get(dev);
1925         if (idev) {
1926                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1927                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1928                         if (ret)
1929                                 break;
1930                 }
1931         }
1932         rcu_read_unlock();
1933
1934         return ret;
1935 }
1936 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1937
1938 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1939 {
1940         const struct inet6_ifaddr *ifa;
1941         const struct inet6_dev *idev;
1942         int     onlink;
1943
1944         onlink = 0;
1945         rcu_read_lock();
1946         idev = __in6_dev_get(dev);
1947         if (idev) {
1948                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1949                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1950                                                    ifa->prefix_len);
1951                         if (onlink)
1952                                 break;
1953                 }
1954         }
1955         rcu_read_unlock();
1956         return onlink;
1957 }
1958 EXPORT_SYMBOL(ipv6_chk_prefix);
1959
1960 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1961                                      struct net_device *dev, int strict)
1962 {
1963         unsigned int hash = inet6_addr_hash(net, addr);
1964         struct inet6_ifaddr *ifp, *result = NULL;
1965
1966         rcu_read_lock();
1967         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1968                 if (!net_eq(dev_net(ifp->idev->dev), net))
1969                         continue;
1970                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1971                         if (!dev || ifp->idev->dev == dev ||
1972                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1973                                 result = ifp;
1974                                 in6_ifa_hold(ifp);
1975                                 break;
1976                         }
1977                 }
1978         }
1979         rcu_read_unlock();
1980
1981         return result;
1982 }
1983
1984 /* Gets referenced address, destroys ifaddr */
1985
1986 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1987 {
1988         if (dad_failed)
1989                 ifp->flags |= IFA_F_DADFAILED;
1990
1991         if (ifp->flags&IFA_F_TEMPORARY) {
1992                 struct inet6_ifaddr *ifpub;
1993                 spin_lock_bh(&ifp->lock);
1994                 ifpub = ifp->ifpub;
1995                 if (ifpub) {
1996                         in6_ifa_hold(ifpub);
1997                         spin_unlock_bh(&ifp->lock);
1998                         ipv6_create_tempaddr(ifpub, ifp, true);
1999                         in6_ifa_put(ifpub);
2000                 } else {
2001                         spin_unlock_bh(&ifp->lock);
2002                 }
2003                 ipv6_del_addr(ifp);
2004         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2005                 spin_lock_bh(&ifp->lock);
2006                 addrconf_del_dad_work(ifp);
2007                 ifp->flags |= IFA_F_TENTATIVE;
2008                 if (dad_failed)
2009                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2010                 spin_unlock_bh(&ifp->lock);
2011                 if (dad_failed)
2012                         ipv6_ifa_notify(0, ifp);
2013                 in6_ifa_put(ifp);
2014         } else {
2015                 ipv6_del_addr(ifp);
2016         }
2017 }
2018
2019 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2020 {
2021         int err = -ENOENT;
2022
2023         spin_lock_bh(&ifp->lock);
2024         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2025                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2026                 err = 0;
2027         }
2028         spin_unlock_bh(&ifp->lock);
2029
2030         return err;
2031 }
2032
2033 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2034 {
2035         struct inet6_dev *idev = ifp->idev;
2036         struct net *net = dev_net(ifp->idev->dev);
2037
2038         if (addrconf_dad_end(ifp)) {
2039                 in6_ifa_put(ifp);
2040                 return;
2041         }
2042
2043         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2044                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2045
2046         spin_lock_bh(&ifp->lock);
2047
2048         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2049                 struct in6_addr new_addr;
2050                 struct inet6_ifaddr *ifp2;
2051                 int retries = ifp->stable_privacy_retry + 1;
2052                 struct ifa6_config cfg = {
2053                         .pfx = &new_addr,
2054                         .plen = ifp->prefix_len,
2055                         .ifa_flags = ifp->flags,
2056                         .valid_lft = ifp->valid_lft,
2057                         .preferred_lft = ifp->prefered_lft,
2058                         .scope = ifp->scope,
2059                 };
2060
2061                 if (retries > net->ipv6.sysctl.idgen_retries) {
2062                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2063                                              ifp->idev->dev->name);
2064                         goto errdad;
2065                 }
2066
2067                 new_addr = ifp->addr;
2068                 if (ipv6_generate_stable_address(&new_addr, retries,
2069                                                  idev))
2070                         goto errdad;
2071
2072                 spin_unlock_bh(&ifp->lock);
2073
2074                 if (idev->cnf.max_addresses &&
2075                     ipv6_count_addresses(idev) >=
2076                     idev->cnf.max_addresses)
2077                         goto lock_errdad;
2078
2079                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2080                                      ifp->idev->dev->name);
2081
2082                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2083                 if (IS_ERR(ifp2))
2084                         goto lock_errdad;
2085
2086                 spin_lock_bh(&ifp2->lock);
2087                 ifp2->stable_privacy_retry = retries;
2088                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2089                 spin_unlock_bh(&ifp2->lock);
2090
2091                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2092                 in6_ifa_put(ifp2);
2093 lock_errdad:
2094                 spin_lock_bh(&ifp->lock);
2095         }
2096
2097 errdad:
2098         /* transition from _POSTDAD to _ERRDAD */
2099         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2100         spin_unlock_bh(&ifp->lock);
2101
2102         addrconf_mod_dad_work(ifp, 0);
2103         in6_ifa_put(ifp);
2104 }
2105
2106 /* Join to solicited addr multicast group.
2107  * caller must hold RTNL */
2108 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2109 {
2110         struct in6_addr maddr;
2111
2112         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2113                 return;
2114
2115         addrconf_addr_solict_mult(addr, &maddr);
2116         ipv6_dev_mc_inc(dev, &maddr);
2117 }
2118
2119 /* caller must hold RTNL */
2120 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2121 {
2122         struct in6_addr maddr;
2123
2124         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2125                 return;
2126
2127         addrconf_addr_solict_mult(addr, &maddr);
2128         __ipv6_dev_mc_dec(idev, &maddr);
2129 }
2130
2131 /* caller must hold RTNL */
2132 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2133 {
2134         struct in6_addr addr;
2135
2136         if (ifp->prefix_len >= 127) /* RFC 6164 */
2137                 return;
2138         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2139         if (ipv6_addr_any(&addr))
2140                 return;
2141         __ipv6_dev_ac_inc(ifp->idev, &addr);
2142 }
2143
2144 /* caller must hold RTNL */
2145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2146 {
2147         struct in6_addr addr;
2148
2149         if (ifp->prefix_len >= 127) /* RFC 6164 */
2150                 return;
2151         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2152         if (ipv6_addr_any(&addr))
2153                 return;
2154         __ipv6_dev_ac_dec(ifp->idev, &addr);
2155 }
2156
2157 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2158 {
2159         switch (dev->addr_len) {
2160         case ETH_ALEN:
2161                 memcpy(eui, dev->dev_addr, 3);
2162                 eui[3] = 0xFF;
2163                 eui[4] = 0xFE;
2164                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2165                 break;
2166         case EUI64_ADDR_LEN:
2167                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2168                 eui[0] ^= 2;
2169                 break;
2170         default:
2171                 return -1;
2172         }
2173
2174         return 0;
2175 }
2176
2177 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2178 {
2179         union fwnet_hwaddr *ha;
2180
2181         if (dev->addr_len != FWNET_ALEN)
2182                 return -1;
2183
2184         ha = (union fwnet_hwaddr *)dev->dev_addr;
2185
2186         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2187         eui[0] ^= 2;
2188         return 0;
2189 }
2190
2191 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2192 {
2193         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2194         if (dev->addr_len != ARCNET_ALEN)
2195                 return -1;
2196         memset(eui, 0, 7);
2197         eui[7] = *(u8 *)dev->dev_addr;
2198         return 0;
2199 }
2200
2201 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2202 {
2203         if (dev->addr_len != INFINIBAND_ALEN)
2204                 return -1;
2205         memcpy(eui, dev->dev_addr + 12, 8);
2206         eui[0] |= 2;
2207         return 0;
2208 }
2209
2210 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2211 {
2212         if (addr == 0)
2213                 return -1;
2214         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2215                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2216                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2217                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2218                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2219                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2220         eui[1] = 0;
2221         eui[2] = 0x5E;
2222         eui[3] = 0xFE;
2223         memcpy(eui + 4, &addr, 4);
2224         return 0;
2225 }
2226
2227 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2228 {
2229         if (dev->priv_flags & IFF_ISATAP)
2230                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2231         return -1;
2232 }
2233
2234 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2235 {
2236         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2237 }
2238
2239 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2240 {
2241         memcpy(eui, dev->perm_addr, 3);
2242         memcpy(eui + 5, dev->perm_addr + 3, 3);
2243         eui[3] = 0xFF;
2244         eui[4] = 0xFE;
2245         eui[0] ^= 2;
2246         return 0;
2247 }
2248
2249 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2250 {
2251         switch (dev->type) {
2252         case ARPHRD_ETHER:
2253         case ARPHRD_FDDI:
2254                 return addrconf_ifid_eui48(eui, dev);
2255         case ARPHRD_ARCNET:
2256                 return addrconf_ifid_arcnet(eui, dev);
2257         case ARPHRD_INFINIBAND:
2258                 return addrconf_ifid_infiniband(eui, dev);
2259         case ARPHRD_SIT:
2260                 return addrconf_ifid_sit(eui, dev);
2261         case ARPHRD_IPGRE:
2262         case ARPHRD_TUNNEL:
2263                 return addrconf_ifid_gre(eui, dev);
2264         case ARPHRD_6LOWPAN:
2265                 return addrconf_ifid_6lowpan(eui, dev);
2266         case ARPHRD_IEEE1394:
2267                 return addrconf_ifid_ieee1394(eui, dev);
2268         case ARPHRD_TUNNEL6:
2269         case ARPHRD_IP6GRE:
2270         case ARPHRD_RAWIP:
2271                 return addrconf_ifid_ip6tnl(eui, dev);
2272         }
2273         return -1;
2274 }
2275
2276 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2277 {
2278         int err = -1;
2279         struct inet6_ifaddr *ifp;
2280
2281         read_lock_bh(&idev->lock);
2282         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2283                 if (ifp->scope > IFA_LINK)
2284                         break;
2285                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2286                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2287                         err = 0;
2288                         break;
2289                 }
2290         }
2291         read_unlock_bh(&idev->lock);
2292         return err;
2293 }
2294
2295 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2296 static void ipv6_regen_rndid(struct inet6_dev *idev)
2297 {
2298 regen:
2299         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2300         idev->rndid[0] &= ~0x02;
2301
2302         /*
2303          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2304          * check if generated address is not inappropriate
2305          *
2306          *  - Reserved subnet anycast (RFC 2526)
2307          *      11111101 11....11 1xxxxxxx
2308          *  - ISATAP (RFC4214) 6.1
2309          *      00-00-5E-FE-xx-xx-xx-xx
2310          *  - value 0
2311          *  - XXX: already assigned to an address on the device
2312          */
2313         if (idev->rndid[0] == 0xfd &&
2314             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2315             (idev->rndid[7]&0x80))
2316                 goto regen;
2317         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2318                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2319                         goto regen;
2320                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2321                         goto regen;
2322         }
2323 }
2324
2325 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2326 {
2327         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2328                 ipv6_regen_rndid(idev);
2329 }
2330
2331 /*
2332  *      Add prefix route.
2333  */
2334
2335 static void
2336 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2337                       struct net_device *dev, unsigned long expires,
2338                       u32 flags, gfp_t gfp_flags)
2339 {
2340         struct fib6_config cfg = {
2341                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2342                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2343                 .fc_ifindex = dev->ifindex,
2344                 .fc_expires = expires,
2345                 .fc_dst_len = plen,
2346                 .fc_flags = RTF_UP | flags,
2347                 .fc_nlinfo.nl_net = dev_net(dev),
2348                 .fc_protocol = RTPROT_KERNEL,
2349                 .fc_type = RTN_UNICAST,
2350         };
2351
2352         cfg.fc_dst = *pfx;
2353
2354         /* Prevent useless cloning on PtP SIT.
2355            This thing is done here expecting that the whole
2356            class of non-broadcast devices need not cloning.
2357          */
2358 #if IS_ENABLED(CONFIG_IPV6_SIT)
2359         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2360                 cfg.fc_flags |= RTF_NONEXTHOP;
2361 #endif
2362
2363         ip6_route_add(&cfg, gfp_flags, NULL);
2364 }
2365
2366
2367 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2368                                                   int plen,
2369                                                   const struct net_device *dev,
2370                                                   u32 flags, u32 noflags)
2371 {
2372         struct fib6_node *fn;
2373         struct fib6_info *rt = NULL;
2374         struct fib6_table *table;
2375         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2376
2377         table = fib6_get_table(dev_net(dev), tb_id);
2378         if (!table)
2379                 return NULL;
2380
2381         rcu_read_lock();
2382         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2383         if (!fn)
2384                 goto out;
2385
2386         for_each_fib6_node_rt_rcu(fn) {
2387                 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2388                         continue;
2389                 if ((rt->fib6_flags & flags) != flags)
2390                         continue;
2391                 if ((rt->fib6_flags & noflags) != 0)
2392                         continue;
2393                 if (!fib6_info_hold_safe(rt))
2394                         continue;
2395                 break;
2396         }
2397 out:
2398         rcu_read_unlock();
2399         return rt;
2400 }
2401
2402
2403 /* Create "default" multicast route to the interface */
2404
2405 static void addrconf_add_mroute(struct net_device *dev)
2406 {
2407         struct fib6_config cfg = {
2408                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2409                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2410                 .fc_ifindex = dev->ifindex,
2411                 .fc_dst_len = 8,
2412                 .fc_flags = RTF_UP,
2413                 .fc_type = RTN_UNICAST,
2414                 .fc_nlinfo.nl_net = dev_net(dev),
2415         };
2416
2417         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2418
2419         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2420 }
2421
2422 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2423 {
2424         struct inet6_dev *idev;
2425
2426         ASSERT_RTNL();
2427
2428         idev = ipv6_find_idev(dev);
2429         if (!idev)
2430                 return ERR_PTR(-ENOBUFS);
2431
2432         if (idev->cnf.disable_ipv6)
2433                 return ERR_PTR(-EACCES);
2434
2435         /* Add default multicast route */
2436         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2437                 addrconf_add_mroute(dev);
2438
2439         return idev;
2440 }
2441
2442 static void manage_tempaddrs(struct inet6_dev *idev,
2443                              struct inet6_ifaddr *ifp,
2444                              __u32 valid_lft, __u32 prefered_lft,
2445                              bool create, unsigned long now)
2446 {
2447         u32 flags;
2448         struct inet6_ifaddr *ift;
2449
2450         read_lock_bh(&idev->lock);
2451         /* update all temporary addresses in the list */
2452         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2453                 int age, max_valid, max_prefered;
2454
2455                 if (ifp != ift->ifpub)
2456                         continue;
2457
2458                 /* RFC 4941 section 3.3:
2459                  * If a received option will extend the lifetime of a public
2460                  * address, the lifetimes of temporary addresses should
2461                  * be extended, subject to the overall constraint that no
2462                  * temporary addresses should ever remain "valid" or "preferred"
2463                  * for a time longer than (TEMP_VALID_LIFETIME) or
2464                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2465                  */
2466                 age = (now - ift->cstamp) / HZ;
2467                 max_valid = idev->cnf.temp_valid_lft - age;
2468                 if (max_valid < 0)
2469                         max_valid = 0;
2470
2471                 max_prefered = idev->cnf.temp_prefered_lft -
2472                                idev->desync_factor - age;
2473                 if (max_prefered < 0)
2474                         max_prefered = 0;
2475
2476                 if (valid_lft > max_valid)
2477                         valid_lft = max_valid;
2478
2479                 if (prefered_lft > max_prefered)
2480                         prefered_lft = max_prefered;
2481
2482                 spin_lock(&ift->lock);
2483                 flags = ift->flags;
2484                 ift->valid_lft = valid_lft;
2485                 ift->prefered_lft = prefered_lft;
2486                 ift->tstamp = now;
2487                 if (prefered_lft > 0)
2488                         ift->flags &= ~IFA_F_DEPRECATED;
2489
2490                 spin_unlock(&ift->lock);
2491                 if (!(flags&IFA_F_TENTATIVE))
2492                         ipv6_ifa_notify(0, ift);
2493         }
2494
2495         if ((create || list_empty(&idev->tempaddr_list)) &&
2496             idev->cnf.use_tempaddr > 0) {
2497                 /* When a new public address is created as described
2498                  * in [ADDRCONF], also create a new temporary address.
2499                  * Also create a temporary address if it's enabled but
2500                  * no temporary address currently exists.
2501                  */
2502                 read_unlock_bh(&idev->lock);
2503                 ipv6_create_tempaddr(ifp, NULL, false);
2504         } else {
2505                 read_unlock_bh(&idev->lock);
2506         }
2507 }
2508
2509 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2510 {
2511         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2512                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2513 }
2514
2515 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2516                                  const struct prefix_info *pinfo,
2517                                  struct inet6_dev *in6_dev,
2518                                  const struct in6_addr *addr, int addr_type,
2519                                  u32 addr_flags, bool sllao, bool tokenized,
2520                                  __u32 valid_lft, u32 prefered_lft)
2521 {
2522         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2523         int create = 0, update_lft = 0;
2524
2525         if (!ifp && valid_lft) {
2526                 int max_addresses = in6_dev->cnf.max_addresses;
2527                 struct ifa6_config cfg = {
2528                         .pfx = addr,
2529                         .plen = pinfo->prefix_len,
2530                         .ifa_flags = addr_flags,
2531                         .valid_lft = valid_lft,
2532                         .preferred_lft = prefered_lft,
2533                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2534                 };
2535
2536 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2537                 if ((net->ipv6.devconf_all->optimistic_dad ||
2538                      in6_dev->cnf.optimistic_dad) &&
2539                     !net->ipv6.devconf_all->forwarding && sllao)
2540                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2541 #endif
2542
2543                 /* Do not allow to create too much of autoconfigured
2544                  * addresses; this would be too easy way to crash kernel.
2545                  */
2546                 if (!max_addresses ||
2547                     ipv6_count_addresses(in6_dev) < max_addresses)
2548                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2549
2550                 if (IS_ERR_OR_NULL(ifp))
2551                         return -1;
2552
2553                 create = 1;
2554                 spin_lock_bh(&ifp->lock);
2555                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2556                 ifp->cstamp = jiffies;
2557                 ifp->tokenized = tokenized;
2558                 spin_unlock_bh(&ifp->lock);
2559                 addrconf_dad_start(ifp);
2560         }
2561
2562         if (ifp) {
2563                 u32 flags;
2564                 unsigned long now;
2565                 u32 stored_lft;
2566
2567                 /* update lifetime (RFC2462 5.5.3 e) */
2568                 spin_lock_bh(&ifp->lock);
2569                 now = jiffies;
2570                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2571                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2572                 else
2573                         stored_lft = 0;
2574                 if (!create && stored_lft) {
2575                         const u32 minimum_lft = min_t(u32,
2576                                 stored_lft, MIN_VALID_LIFETIME);
2577                         valid_lft = max(valid_lft, minimum_lft);
2578
2579                         /* RFC4862 Section 5.5.3e:
2580                          * "Note that the preferred lifetime of the
2581                          *  corresponding address is always reset to
2582                          *  the Preferred Lifetime in the received
2583                          *  Prefix Information option, regardless of
2584                          *  whether the valid lifetime is also reset or
2585                          *  ignored."
2586                          *
2587                          * So we should always update prefered_lft here.
2588                          */
2589                         update_lft = 1;
2590                 }
2591
2592                 if (update_lft) {
2593                         ifp->valid_lft = valid_lft;
2594                         ifp->prefered_lft = prefered_lft;
2595                         ifp->tstamp = now;
2596                         flags = ifp->flags;
2597                         ifp->flags &= ~IFA_F_DEPRECATED;
2598                         spin_unlock_bh(&ifp->lock);
2599
2600                         if (!(flags&IFA_F_TENTATIVE))
2601                                 ipv6_ifa_notify(0, ifp);
2602                 } else
2603                         spin_unlock_bh(&ifp->lock);
2604
2605                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2606                                  create, now);
2607
2608                 in6_ifa_put(ifp);
2609                 addrconf_verify();
2610         }
2611
2612         return 0;
2613 }
2614 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2615
2616 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2617 {
2618         struct prefix_info *pinfo;
2619         __u32 valid_lft;
2620         __u32 prefered_lft;
2621         int addr_type, err;
2622         u32 addr_flags = 0;
2623         struct inet6_dev *in6_dev;
2624         struct net *net = dev_net(dev);
2625
2626         pinfo = (struct prefix_info *) opt;
2627
2628         if (len < sizeof(struct prefix_info)) {
2629                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2630                 return;
2631         }
2632
2633         /*
2634          *      Validation checks ([ADDRCONF], page 19)
2635          */
2636
2637         addr_type = ipv6_addr_type(&pinfo->prefix);
2638
2639         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2640                 return;
2641
2642         valid_lft = ntohl(pinfo->valid);
2643         prefered_lft = ntohl(pinfo->prefered);
2644
2645         if (prefered_lft > valid_lft) {
2646                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2647                 return;
2648         }
2649
2650         in6_dev = in6_dev_get(dev);
2651
2652         if (!in6_dev) {
2653                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2654                                     dev->name);
2655                 return;
2656         }
2657
2658         /*
2659          *      Two things going on here:
2660          *      1) Add routes for on-link prefixes
2661          *      2) Configure prefixes with the auto flag set
2662          */
2663
2664         if (pinfo->onlink) {
2665                 struct fib6_info *rt;
2666                 unsigned long rt_expires;
2667
2668                 /* Avoid arithmetic overflow. Really, we could
2669                  * save rt_expires in seconds, likely valid_lft,
2670                  * but it would require division in fib gc, that it
2671                  * not good.
2672                  */
2673                 if (HZ > USER_HZ)
2674                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2675                 else
2676                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2677
2678                 if (addrconf_finite_timeout(rt_expires))
2679                         rt_expires *= HZ;
2680
2681                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2682                                                pinfo->prefix_len,
2683                                                dev,
2684                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2685                                                RTF_GATEWAY | RTF_DEFAULT);
2686
2687                 if (rt) {
2688                         /* Autoconf prefix route */
2689                         if (valid_lft == 0) {
2690                                 ip6_del_rt(net, rt);
2691                                 rt = NULL;
2692                         } else if (addrconf_finite_timeout(rt_expires)) {
2693                                 /* not infinity */
2694                                 fib6_set_expires(rt, jiffies + rt_expires);
2695                         } else {
2696                                 fib6_clean_expires(rt);
2697                         }
2698                 } else if (valid_lft) {
2699                         clock_t expires = 0;
2700                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2701                         if (addrconf_finite_timeout(rt_expires)) {
2702                                 /* not infinity */
2703                                 flags |= RTF_EXPIRES;
2704                                 expires = jiffies_to_clock_t(rt_expires);
2705                         }
2706                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2707                                               0, dev, expires, flags,
2708                                               GFP_ATOMIC);
2709                 }
2710                 fib6_info_release(rt);
2711         }
2712
2713         /* Try to figure out our local address for this prefix */
2714
2715         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2716                 struct in6_addr addr;
2717                 bool tokenized = false, dev_addr_generated = false;
2718
2719                 if (pinfo->prefix_len == 64) {
2720                         memcpy(&addr, &pinfo->prefix, 8);
2721
2722                         if (!ipv6_addr_any(&in6_dev->token)) {
2723                                 read_lock_bh(&in6_dev->lock);
2724                                 memcpy(addr.s6_addr + 8,
2725                                        in6_dev->token.s6_addr + 8, 8);
2726                                 read_unlock_bh(&in6_dev->lock);
2727                                 tokenized = true;
2728                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2729                                    !ipv6_generate_stable_address(&addr, 0,
2730                                                                  in6_dev)) {
2731                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2732                                 goto ok;
2733                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2734                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2735                                 goto put;
2736                         } else {
2737                                 dev_addr_generated = true;
2738                         }
2739                         goto ok;
2740                 }
2741                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2742                                     pinfo->prefix_len);
2743                 goto put;
2744
2745 ok:
2746                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2747                                                    &addr, addr_type,
2748                                                    addr_flags, sllao,
2749                                                    tokenized, valid_lft,
2750                                                    prefered_lft);
2751                 if (err)
2752                         goto put;
2753
2754                 /* Ignore error case here because previous prefix add addr was
2755                  * successful which will be notified.
2756                  */
2757                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2758                                               addr_type, addr_flags, sllao,
2759                                               tokenized, valid_lft,
2760                                               prefered_lft,
2761                                               dev_addr_generated);
2762         }
2763         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2764 put:
2765         in6_dev_put(in6_dev);
2766 }
2767
2768 /*
2769  *      Set destination address.
2770  *      Special case for SIT interfaces where we create a new "virtual"
2771  *      device.
2772  */
2773 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2774 {
2775         struct in6_ifreq ireq;
2776         struct net_device *dev;
2777         int err = -EINVAL;
2778
2779         rtnl_lock();
2780
2781         err = -EFAULT;
2782         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2783                 goto err_exit;
2784
2785         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2786
2787         err = -ENODEV;
2788         if (!dev)
2789                 goto err_exit;
2790
2791 #if IS_ENABLED(CONFIG_IPV6_SIT)
2792         if (dev->type == ARPHRD_SIT) {
2793                 const struct net_device_ops *ops = dev->netdev_ops;
2794                 struct ifreq ifr;
2795                 struct ip_tunnel_parm p;
2796
2797                 err = -EADDRNOTAVAIL;
2798                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2799                         goto err_exit;
2800
2801                 memset(&p, 0, sizeof(p));
2802                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2803                 p.iph.saddr = 0;
2804                 p.iph.version = 4;
2805                 p.iph.ihl = 5;
2806                 p.iph.protocol = IPPROTO_IPV6;
2807                 p.iph.ttl = 64;
2808                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2809
2810                 if (ops->ndo_do_ioctl) {
2811                         mm_segment_t oldfs = get_fs();
2812
2813                         set_fs(KERNEL_DS);
2814                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2815                         set_fs(oldfs);
2816                 } else
2817                         err = -EOPNOTSUPP;
2818
2819                 if (err == 0) {
2820                         err = -ENOBUFS;
2821                         dev = __dev_get_by_name(net, p.name);
2822                         if (!dev)
2823                                 goto err_exit;
2824                         err = dev_open(dev, NULL);
2825                 }
2826         }
2827 #endif
2828
2829 err_exit:
2830         rtnl_unlock();
2831         return err;
2832 }
2833
2834 static int ipv6_mc_config(struct sock *sk, bool join,
2835                           const struct in6_addr *addr, int ifindex)
2836 {
2837         int ret;
2838
2839         ASSERT_RTNL();
2840
2841         lock_sock(sk);
2842         if (join)
2843                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2844         else
2845                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2846         release_sock(sk);
2847
2848         return ret;
2849 }
2850
2851 /*
2852  *      Manual configuration of address on an interface
2853  */
2854 static int inet6_addr_add(struct net *net, int ifindex,
2855                           struct ifa6_config *cfg,
2856                           struct netlink_ext_ack *extack)
2857 {
2858         struct inet6_ifaddr *ifp;
2859         struct inet6_dev *idev;
2860         struct net_device *dev;
2861         unsigned long timeout;
2862         clock_t expires;
2863         u32 flags;
2864
2865         ASSERT_RTNL();
2866
2867         if (cfg->plen > 128)
2868                 return -EINVAL;
2869
2870         /* check the lifetime */
2871         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2872                 return -EINVAL;
2873
2874         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2875                 return -EINVAL;
2876
2877         dev = __dev_get_by_index(net, ifindex);
2878         if (!dev)
2879                 return -ENODEV;
2880
2881         idev = addrconf_add_dev(dev);
2882         if (IS_ERR(idev))
2883                 return PTR_ERR(idev);
2884
2885         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2886                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2887                                          true, cfg->pfx, ifindex);
2888
2889                 if (ret < 0)
2890                         return ret;
2891         }
2892
2893         cfg->scope = ipv6_addr_scope(cfg->pfx);
2894
2895         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2896         if (addrconf_finite_timeout(timeout)) {
2897                 expires = jiffies_to_clock_t(timeout * HZ);
2898                 cfg->valid_lft = timeout;
2899                 flags = RTF_EXPIRES;
2900         } else {
2901                 expires = 0;
2902                 flags = 0;
2903                 cfg->ifa_flags |= IFA_F_PERMANENT;
2904         }
2905
2906         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2907         if (addrconf_finite_timeout(timeout)) {
2908                 if (timeout == 0)
2909                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2910                 cfg->preferred_lft = timeout;
2911         }
2912
2913         ifp = ipv6_add_addr(idev, cfg, true, extack);
2914         if (!IS_ERR(ifp)) {
2915                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2916                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2917                                               ifp->rt_priority, dev, expires,
2918                                               flags, GFP_KERNEL);
2919                 }
2920
2921                 /* Send a netlink notification if DAD is enabled and
2922                  * optimistic flag is not set
2923                  */
2924                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2925                         ipv6_ifa_notify(0, ifp);
2926                 /*
2927                  * Note that section 3.1 of RFC 4429 indicates
2928                  * that the Optimistic flag should not be set for
2929                  * manually configured addresses
2930                  */
2931                 addrconf_dad_start(ifp);
2932                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2933                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2934                                          cfg->preferred_lft, true, jiffies);
2935                 in6_ifa_put(ifp);
2936                 addrconf_verify_rtnl();
2937                 return 0;
2938         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2939                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2940                                cfg->pfx, ifindex);
2941         }
2942
2943         return PTR_ERR(ifp);
2944 }
2945
2946 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2947                           const struct in6_addr *pfx, unsigned int plen)
2948 {
2949         struct inet6_ifaddr *ifp;
2950         struct inet6_dev *idev;
2951         struct net_device *dev;
2952
2953         if (plen > 128)
2954                 return -EINVAL;
2955
2956         dev = __dev_get_by_index(net, ifindex);
2957         if (!dev)
2958                 return -ENODEV;
2959
2960         idev = __in6_dev_get(dev);
2961         if (!idev)
2962                 return -ENXIO;
2963
2964         read_lock_bh(&idev->lock);
2965         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2966                 if (ifp->prefix_len == plen &&
2967                     ipv6_addr_equal(pfx, &ifp->addr)) {
2968                         in6_ifa_hold(ifp);
2969                         read_unlock_bh(&idev->lock);
2970
2971                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2972                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2973                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2974                                                  jiffies);
2975                         ipv6_del_addr(ifp);
2976                         addrconf_verify_rtnl();
2977                         if (ipv6_addr_is_multicast(pfx)) {
2978                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2979                                                false, pfx, dev->ifindex);
2980                         }
2981                         return 0;
2982                 }
2983         }
2984         read_unlock_bh(&idev->lock);
2985         return -EADDRNOTAVAIL;
2986 }
2987
2988
2989 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2990 {
2991         struct ifa6_config cfg = {
2992                 .ifa_flags = IFA_F_PERMANENT,
2993                 .preferred_lft = INFINITY_LIFE_TIME,
2994                 .valid_lft = INFINITY_LIFE_TIME,
2995         };
2996         struct in6_ifreq ireq;
2997         int err;
2998
2999         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3000                 return -EPERM;
3001
3002         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3003                 return -EFAULT;
3004
3005         cfg.pfx = &ireq.ifr6_addr;
3006         cfg.plen = ireq.ifr6_prefixlen;
3007
3008         rtnl_lock();
3009         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3010         rtnl_unlock();
3011         return err;
3012 }
3013
3014 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3015 {
3016         struct in6_ifreq ireq;
3017         int err;
3018
3019         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3020                 return -EPERM;
3021
3022         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3023                 return -EFAULT;
3024
3025         rtnl_lock();
3026         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3027                              ireq.ifr6_prefixlen);
3028         rtnl_unlock();
3029         return err;
3030 }
3031
3032 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3033                      int plen, int scope)
3034 {
3035         struct inet6_ifaddr *ifp;
3036         struct ifa6_config cfg = {
3037                 .pfx = addr,
3038                 .plen = plen,
3039                 .ifa_flags = IFA_F_PERMANENT,
3040                 .valid_lft = INFINITY_LIFE_TIME,
3041                 .preferred_lft = INFINITY_LIFE_TIME,
3042                 .scope = scope
3043         };
3044
3045         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3046         if (!IS_ERR(ifp)) {
3047                 spin_lock_bh(&ifp->lock);
3048                 ifp->flags &= ~IFA_F_TENTATIVE;
3049                 spin_unlock_bh(&ifp->lock);
3050                 rt_genid_bump_ipv6(dev_net(idev->dev));
3051                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3052                 in6_ifa_put(ifp);
3053         }
3054 }
3055
3056 #if IS_ENABLED(CONFIG_IPV6_SIT)
3057 static void sit_add_v4_addrs(struct inet6_dev *idev)
3058 {
3059         struct in6_addr addr;
3060         struct net_device *dev;
3061         struct net *net = dev_net(idev->dev);
3062         int scope, plen;
3063         u32 pflags = 0;
3064
3065         ASSERT_RTNL();
3066
3067         memset(&addr, 0, sizeof(struct in6_addr));
3068         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3069
3070         if (idev->dev->flags&IFF_POINTOPOINT) {
3071                 addr.s6_addr32[0] = htonl(0xfe800000);
3072                 scope = IFA_LINK;
3073                 plen = 64;
3074         } else {
3075                 scope = IPV6_ADDR_COMPATv4;
3076                 plen = 96;
3077                 pflags |= RTF_NONEXTHOP;
3078         }
3079
3080         if (addr.s6_addr32[3]) {
3081                 add_addr(idev, &addr, plen, scope);
3082                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3083                                       GFP_KERNEL);
3084                 return;
3085         }
3086
3087         for_each_netdev(net, dev) {
3088                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3089                 if (in_dev && (dev->flags & IFF_UP)) {
3090                         struct in_ifaddr *ifa;
3091
3092                         int flag = scope;
3093
3094                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3095
3096                                 addr.s6_addr32[3] = ifa->ifa_local;
3097
3098                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3099                                         continue;
3100                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3101                                         if (idev->dev->flags&IFF_POINTOPOINT)
3102                                                 continue;
3103                                         flag |= IFA_HOST;
3104                                 }
3105
3106                                 add_addr(idev, &addr, plen, flag);
3107                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3108                                                       0, pflags, GFP_KERNEL);
3109                         }
3110                 }
3111         }
3112 }
3113 #endif
3114
3115 static void init_loopback(struct net_device *dev)
3116 {
3117         struct inet6_dev  *idev;
3118
3119         /* ::1 */
3120
3121         ASSERT_RTNL();
3122
3123         idev = ipv6_find_idev(dev);
3124         if (!idev) {
3125                 pr_debug("%s: add_dev failed\n", __func__);
3126                 return;
3127         }
3128
3129         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3130 }
3131
3132 void addrconf_add_linklocal(struct inet6_dev *idev,
3133                             const struct in6_addr *addr, u32 flags)
3134 {
3135         struct ifa6_config cfg = {
3136                 .pfx = addr,
3137                 .plen = 64,
3138                 .ifa_flags = flags | IFA_F_PERMANENT,
3139                 .valid_lft = INFINITY_LIFE_TIME,
3140                 .preferred_lft = INFINITY_LIFE_TIME,
3141                 .scope = IFA_LINK
3142         };
3143         struct inet6_ifaddr *ifp;
3144
3145 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3146         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3147              idev->cnf.optimistic_dad) &&
3148             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3149                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3150 #endif
3151
3152         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3153         if (!IS_ERR(ifp)) {
3154                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3155                                       0, 0, GFP_ATOMIC);
3156                 addrconf_dad_start(ifp);
3157                 in6_ifa_put(ifp);
3158         }
3159 }
3160 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3161
3162 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3163 {
3164         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3165                 return true;
3166
3167         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3168             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3169                 return true;
3170
3171         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3172             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3173                 return true;
3174
3175         return false;
3176 }
3177
3178 static int ipv6_generate_stable_address(struct in6_addr *address,
3179                                         u8 dad_count,
3180                                         const struct inet6_dev *idev)
3181 {
3182         static DEFINE_SPINLOCK(lock);
3183         static __u32 digest[SHA_DIGEST_WORDS];
3184         static __u32 workspace[SHA_WORKSPACE_WORDS];
3185
3186         static union {
3187                 char __data[SHA_MESSAGE_BYTES];
3188                 struct {
3189                         struct in6_addr secret;
3190                         __be32 prefix[2];
3191                         unsigned char hwaddr[MAX_ADDR_LEN];
3192                         u8 dad_count;
3193                 } __packed;
3194         } data;
3195
3196         struct in6_addr secret;
3197         struct in6_addr temp;
3198         struct net *net = dev_net(idev->dev);
3199
3200         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3201
3202         if (idev->cnf.stable_secret.initialized)
3203                 secret = idev->cnf.stable_secret.secret;
3204         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3205                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3206         else
3207                 return -1;
3208
3209 retry:
3210         spin_lock_bh(&lock);
3211
3212         sha_init(digest);
3213         memset(&data, 0, sizeof(data));
3214         memset(workspace, 0, sizeof(workspace));
3215         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3216         data.prefix[0] = address->s6_addr32[0];
3217         data.prefix[1] = address->s6_addr32[1];
3218         data.secret = secret;
3219         data.dad_count = dad_count;
3220
3221         sha_transform(digest, data.__data, workspace);
3222
3223         temp = *address;
3224         temp.s6_addr32[2] = (__force __be32)digest[0];
3225         temp.s6_addr32[3] = (__force __be32)digest[1];
3226
3227         spin_unlock_bh(&lock);
3228
3229         if (ipv6_reserved_interfaceid(temp)) {
3230                 dad_count++;
3231                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3232                         return -1;
3233                 goto retry;
3234         }
3235
3236         *address = temp;
3237         return 0;
3238 }
3239
3240 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3241 {
3242         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3243
3244         if (s->initialized)
3245                 return;
3246         s = &idev->cnf.stable_secret;
3247         get_random_bytes(&s->secret, sizeof(s->secret));
3248         s->initialized = true;
3249 }
3250
3251 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3252 {
3253         struct in6_addr addr;
3254
3255         /* no link local addresses on L3 master devices */
3256         if (netif_is_l3_master(idev->dev))
3257                 return;
3258
3259         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3260
3261         switch (idev->cnf.addr_gen_mode) {
3262         case IN6_ADDR_GEN_MODE_RANDOM:
3263                 ipv6_gen_mode_random_init(idev);
3264                 /* fallthrough */
3265         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3266                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3267                         addrconf_add_linklocal(idev, &addr,
3268                                                IFA_F_STABLE_PRIVACY);
3269                 else if (prefix_route)
3270                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3271                                               0, 0, GFP_KERNEL);
3272                 break;
3273         case IN6_ADDR_GEN_MODE_EUI64:
3274                 /* addrconf_add_linklocal also adds a prefix_route and we
3275                  * only need to care about prefix routes if ipv6_generate_eui64
3276                  * couldn't generate one.
3277                  */
3278                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3279                         addrconf_add_linklocal(idev, &addr, 0);
3280                 else if (prefix_route)
3281                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3282                                               0, 0, GFP_KERNEL);
3283                 break;
3284         case IN6_ADDR_GEN_MODE_NONE:
3285         default:
3286                 /* will not add any link local address */
3287                 break;
3288         }
3289 }
3290
3291 static void addrconf_dev_config(struct net_device *dev)
3292 {
3293         struct inet6_dev *idev;
3294
3295         ASSERT_RTNL();
3296
3297         if ((dev->type != ARPHRD_ETHER) &&
3298             (dev->type != ARPHRD_FDDI) &&
3299             (dev->type != ARPHRD_ARCNET) &&
3300             (dev->type != ARPHRD_INFINIBAND) &&
3301             (dev->type != ARPHRD_IEEE1394) &&
3302             (dev->type != ARPHRD_TUNNEL6) &&
3303             (dev->type != ARPHRD_6LOWPAN) &&
3304             (dev->type != ARPHRD_IP6GRE) &&
3305             (dev->type != ARPHRD_IPGRE) &&
3306             (dev->type != ARPHRD_TUNNEL) &&
3307             (dev->type != ARPHRD_NONE) &&
3308             (dev->type != ARPHRD_RAWIP)) {
3309                 /* Alas, we support only Ethernet autoconfiguration. */
3310                 return;
3311         }
3312
3313         idev = addrconf_add_dev(dev);
3314         if (IS_ERR(idev))
3315                 return;
3316
3317         /* this device type has no EUI support */
3318         if (dev->type == ARPHRD_NONE &&
3319             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3320                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3321
3322         addrconf_addr_gen(idev, false);
3323 }
3324
3325 #if IS_ENABLED(CONFIG_IPV6_SIT)
3326 static void addrconf_sit_config(struct net_device *dev)
3327 {
3328         struct inet6_dev *idev;
3329
3330         ASSERT_RTNL();
3331
3332         /*
3333          * Configure the tunnel with one of our IPv4
3334          * addresses... we should configure all of
3335          * our v4 addrs in the tunnel
3336          */
3337
3338         idev = ipv6_find_idev(dev);
3339         if (!idev) {
3340                 pr_debug("%s: add_dev failed\n", __func__);
3341                 return;
3342         }
3343
3344         if (dev->priv_flags & IFF_ISATAP) {
3345                 addrconf_addr_gen(idev, false);
3346                 return;
3347         }
3348
3349         sit_add_v4_addrs(idev);
3350
3351         if (dev->flags&IFF_POINTOPOINT)
3352                 addrconf_add_mroute(dev);
3353 }
3354 #endif
3355
3356 #if IS_ENABLED(CONFIG_NET_IPGRE)
3357 static void addrconf_gre_config(struct net_device *dev)
3358 {
3359         struct inet6_dev *idev;
3360
3361         ASSERT_RTNL();
3362
3363         idev = ipv6_find_idev(dev);
3364         if (!idev) {
3365                 pr_debug("%s: add_dev failed\n", __func__);
3366                 return;
3367         }
3368
3369         addrconf_addr_gen(idev, true);
3370         if (dev->flags & IFF_POINTOPOINT)
3371                 addrconf_add_mroute(dev);
3372 }
3373 #endif
3374
3375 static int fixup_permanent_addr(struct net *net,
3376                                 struct inet6_dev *idev,
3377                                 struct inet6_ifaddr *ifp)
3378 {
3379         /* !fib6_node means the host route was removed from the
3380          * FIB, for example, if 'lo' device is taken down. In that
3381          * case regenerate the host route.
3382          */
3383         if (!ifp->rt || !ifp->rt->fib6_node) {
3384                 struct fib6_info *f6i, *prev;
3385
3386                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3387                                          GFP_ATOMIC);
3388                 if (IS_ERR(f6i))
3389                         return PTR_ERR(f6i);
3390
3391                 /* ifp->rt can be accessed outside of rtnl */
3392                 spin_lock(&ifp->lock);
3393                 prev = ifp->rt;
3394                 ifp->rt = f6i;
3395                 spin_unlock(&ifp->lock);
3396
3397                 fib6_info_release(prev);
3398         }
3399
3400         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3401                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3402                                       ifp->rt_priority, idev->dev, 0, 0,
3403                                       GFP_ATOMIC);
3404         }
3405
3406         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3407                 addrconf_dad_start(ifp);
3408
3409         return 0;
3410 }
3411
3412 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3413 {
3414         struct inet6_ifaddr *ifp, *tmp;
3415         struct inet6_dev *idev;
3416
3417         idev = __in6_dev_get(dev);
3418         if (!idev)
3419                 return;
3420
3421         write_lock_bh(&idev->lock);
3422
3423         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3424                 if ((ifp->flags & IFA_F_PERMANENT) &&
3425                     fixup_permanent_addr(net, idev, ifp) < 0) {
3426                         write_unlock_bh(&idev->lock);
3427                         in6_ifa_hold(ifp);
3428                         ipv6_del_addr(ifp);
3429                         write_lock_bh(&idev->lock);
3430
3431                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3432                                              idev->dev->name, &ifp->addr);
3433                 }
3434         }
3435
3436         write_unlock_bh(&idev->lock);
3437 }
3438
3439 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3440                            void *ptr)
3441 {
3442         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3443         struct netdev_notifier_change_info *change_info;
3444         struct netdev_notifier_changeupper_info *info;
3445         struct inet6_dev *idev = __in6_dev_get(dev);
3446         struct net *net = dev_net(dev);
3447         int run_pending = 0;
3448         int err;
3449
3450         switch (event) {
3451         case NETDEV_REGISTER:
3452                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3453                         idev = ipv6_add_dev(dev);
3454                         if (IS_ERR(idev))
3455                                 return notifier_from_errno(PTR_ERR(idev));
3456                 }
3457                 break;
3458
3459         case NETDEV_CHANGEMTU:
3460                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3461                 if (dev->mtu < IPV6_MIN_MTU) {
3462                         addrconf_ifdown(dev, dev != net->loopback_dev);
3463                         break;
3464                 }
3465
3466                 if (idev) {
3467                         rt6_mtu_change(dev, dev->mtu);
3468                         idev->cnf.mtu6 = dev->mtu;
3469                         break;
3470                 }
3471
3472                 /* allocate new idev */
3473                 idev = ipv6_add_dev(dev);
3474                 if (IS_ERR(idev))
3475                         break;
3476
3477                 /* device is still not ready */
3478                 if (!(idev->if_flags & IF_READY))
3479                         break;
3480
3481                 run_pending = 1;
3482
3483                 /* fall through */
3484
3485         case NETDEV_UP:
3486         case NETDEV_CHANGE:
3487                 if (dev->flags & IFF_SLAVE)
3488                         break;
3489
3490                 if (idev && idev->cnf.disable_ipv6)
3491                         break;
3492
3493                 if (event == NETDEV_UP) {
3494                         /* restore routes for permanent addresses */
3495                         addrconf_permanent_addr(net, dev);
3496
3497                         if (!addrconf_link_ready(dev)) {
3498                                 /* device is not ready yet. */
3499                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3500                                          dev->name);
3501                                 break;
3502                         }
3503
3504                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3505                                 idev = ipv6_add_dev(dev);
3506
3507                         if (!IS_ERR_OR_NULL(idev)) {
3508                                 idev->if_flags |= IF_READY;
3509                                 run_pending = 1;
3510                         }
3511                 } else if (event == NETDEV_CHANGE) {
3512                         if (!addrconf_link_ready(dev)) {
3513                                 /* device is still not ready. */
3514                                 rt6_sync_down_dev(dev, event);
3515                                 break;
3516                         }
3517
3518                         if (!IS_ERR_OR_NULL(idev)) {
3519                                 if (idev->if_flags & IF_READY) {
3520                                         /* device is already configured -
3521                                          * but resend MLD reports, we might
3522                                          * have roamed and need to update
3523                                          * multicast snooping switches
3524                                          */
3525                                         ipv6_mc_up(idev);
3526                                         change_info = ptr;
3527                                         if (change_info->flags_changed & IFF_NOARP)
3528                                                 addrconf_dad_run(idev, true);
3529                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3530                                         break;
3531                                 }
3532                                 idev->if_flags |= IF_READY;
3533                         }
3534
3535                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3536                                 dev->name);
3537
3538                         run_pending = 1;
3539                 }
3540
3541                 switch (dev->type) {
3542 #if IS_ENABLED(CONFIG_IPV6_SIT)
3543                 case ARPHRD_SIT:
3544                         addrconf_sit_config(dev);
3545                         break;
3546 #endif
3547 #if IS_ENABLED(CONFIG_NET_IPGRE)
3548                 case ARPHRD_IPGRE:
3549                         addrconf_gre_config(dev);
3550                         break;
3551 #endif
3552                 case ARPHRD_LOOPBACK:
3553                         init_loopback(dev);
3554                         break;
3555
3556                 default:
3557                         addrconf_dev_config(dev);
3558                         break;
3559                 }
3560
3561                 if (!IS_ERR_OR_NULL(idev)) {
3562                         if (run_pending)
3563                                 addrconf_dad_run(idev, false);
3564
3565                         /* Device has an address by now */
3566                         rt6_sync_up(dev, RTNH_F_DEAD);
3567
3568                         /*
3569                          * If the MTU changed during the interface down,
3570                          * when the interface up, the changed MTU must be
3571                          * reflected in the idev as well as routers.
3572                          */
3573                         if (idev->cnf.mtu6 != dev->mtu &&
3574                             dev->mtu >= IPV6_MIN_MTU) {
3575                                 rt6_mtu_change(dev, dev->mtu);
3576                                 idev->cnf.mtu6 = dev->mtu;
3577                         }
3578                         idev->tstamp = jiffies;
3579                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3580
3581                         /*
3582                          * If the changed mtu during down is lower than
3583                          * IPV6_MIN_MTU stop IPv6 on this interface.
3584                          */
3585                         if (dev->mtu < IPV6_MIN_MTU)
3586                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3587                 }
3588                 break;
3589
3590         case NETDEV_DOWN:
3591         case NETDEV_UNREGISTER:
3592                 /*
3593                  *      Remove all addresses from this interface.
3594                  */
3595                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3596                 break;
3597
3598         case NETDEV_CHANGENAME:
3599                 if (idev) {
3600                         snmp6_unregister_dev(idev);
3601                         addrconf_sysctl_unregister(idev);
3602                         err = addrconf_sysctl_register(idev);
3603                         if (err)
3604                                 return notifier_from_errno(err);
3605                         err = snmp6_register_dev(idev);
3606                         if (err) {
3607                                 addrconf_sysctl_unregister(idev);
3608                                 return notifier_from_errno(err);
3609                         }
3610                 }
3611                 break;
3612
3613         case NETDEV_PRE_TYPE_CHANGE:
3614         case NETDEV_POST_TYPE_CHANGE:
3615                 if (idev)
3616                         addrconf_type_change(dev, event);
3617                 break;
3618
3619         case NETDEV_CHANGEUPPER:
3620                 info = ptr;
3621
3622                 /* flush all routes if dev is linked to or unlinked from
3623                  * an L3 master device (e.g., VRF)
3624                  */
3625                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3626                         addrconf_ifdown(dev, 0);
3627         }
3628
3629         return NOTIFY_OK;
3630 }
3631
3632 /*
3633  *      addrconf module should be notified of a device going up
3634  */
3635 static struct notifier_block ipv6_dev_notf = {
3636         .notifier_call = addrconf_notify,
3637         .priority = ADDRCONF_NOTIFY_PRIORITY,
3638 };
3639
3640 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3641 {
3642         struct inet6_dev *idev;
3643         ASSERT_RTNL();
3644
3645         idev = __in6_dev_get(dev);
3646
3647         if (event == NETDEV_POST_TYPE_CHANGE)
3648                 ipv6_mc_remap(idev);
3649         else if (event == NETDEV_PRE_TYPE_CHANGE)
3650                 ipv6_mc_unmap(idev);
3651 }
3652
3653 static bool addr_is_local(const struct in6_addr *addr)
3654 {
3655         return ipv6_addr_type(addr) &
3656                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3657 }
3658
3659 static int addrconf_ifdown(struct net_device *dev, int how)
3660 {
3661         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3662         struct net *net = dev_net(dev);
3663         struct inet6_dev *idev;
3664         struct inet6_ifaddr *ifa, *tmp;
3665         bool keep_addr = false;
3666         int state, i;
3667
3668         ASSERT_RTNL();
3669
3670         rt6_disable_ip(dev, event);
3671
3672         idev = __in6_dev_get(dev);
3673         if (!idev)
3674                 return -ENODEV;
3675
3676         /*
3677          * Step 1: remove reference to ipv6 device from parent device.
3678          *         Do not dev_put!
3679          */
3680         if (how) {
3681                 idev->dead = 1;
3682
3683                 /* protected by rtnl_lock */
3684                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3685
3686                 /* Step 1.5: remove snmp6 entry */
3687                 snmp6_unregister_dev(idev);
3688
3689         }
3690
3691         /* combine the user config with event to determine if permanent
3692          * addresses are to be removed from address hash table
3693          */
3694         if (!how && !idev->cnf.disable_ipv6) {
3695                 /* aggregate the system setting and interface setting */
3696                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3697
3698                 if (!_keep_addr)
3699                         _keep_addr = idev->cnf.keep_addr_on_down;
3700
3701                 keep_addr = (_keep_addr > 0);
3702         }
3703
3704         /* Step 2: clear hash table */
3705         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3706                 struct hlist_head *h = &inet6_addr_lst[i];
3707
3708                 spin_lock_bh(&addrconf_hash_lock);
3709 restart:
3710                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3711                         if (ifa->idev == idev) {
3712                                 addrconf_del_dad_work(ifa);
3713                                 /* combined flag + permanent flag decide if
3714                                  * address is retained on a down event
3715                                  */
3716                                 if (!keep_addr ||
3717                                     !(ifa->flags & IFA_F_PERMANENT) ||
3718                                     addr_is_local(&ifa->addr)) {
3719                                         hlist_del_init_rcu(&ifa->addr_lst);
3720                                         goto restart;
3721                                 }
3722                         }
3723                 }
3724                 spin_unlock_bh(&addrconf_hash_lock);
3725         }
3726
3727         write_lock_bh(&idev->lock);
3728
3729         addrconf_del_rs_timer(idev);
3730
3731         /* Step 2: clear flags for stateless addrconf */
3732         if (!how)
3733                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3734
3735         /* Step 3: clear tempaddr list */
3736         while (!list_empty(&idev->tempaddr_list)) {
3737                 ifa = list_first_entry(&idev->tempaddr_list,
3738                                        struct inet6_ifaddr, tmp_list);
3739                 list_del(&ifa->tmp_list);
3740                 write_unlock_bh(&idev->lock);
3741                 spin_lock_bh(&ifa->lock);
3742
3743                 if (ifa->ifpub) {
3744                         in6_ifa_put(ifa->ifpub);
3745                         ifa->ifpub = NULL;
3746                 }
3747                 spin_unlock_bh(&ifa->lock);
3748                 in6_ifa_put(ifa);
3749                 write_lock_bh(&idev->lock);
3750         }
3751
3752         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3753                 struct fib6_info *rt = NULL;
3754                 bool keep;
3755
3756                 addrconf_del_dad_work(ifa);
3757
3758                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3759                         !addr_is_local(&ifa->addr);
3760
3761                 write_unlock_bh(&idev->lock);
3762                 spin_lock_bh(&ifa->lock);
3763
3764                 if (keep) {
3765                         /* set state to skip the notifier below */
3766                         state = INET6_IFADDR_STATE_DEAD;
3767                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3768                         if (!(ifa->flags & IFA_F_NODAD))
3769                                 ifa->flags |= IFA_F_TENTATIVE;
3770
3771                         rt = ifa->rt;
3772                         ifa->rt = NULL;
3773                 } else {
3774                         state = ifa->state;
3775                         ifa->state = INET6_IFADDR_STATE_DEAD;
3776                 }
3777
3778                 spin_unlock_bh(&ifa->lock);
3779
3780                 if (rt)
3781                         ip6_del_rt(net, rt);
3782
3783                 if (state != INET6_IFADDR_STATE_DEAD) {
3784                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3785                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3786                 } else {
3787                         if (idev->cnf.forwarding)
3788                                 addrconf_leave_anycast(ifa);
3789                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3790                 }
3791
3792                 write_lock_bh(&idev->lock);
3793                 if (!keep) {
3794                         list_del_rcu(&ifa->if_list);
3795                         in6_ifa_put(ifa);
3796                 }
3797         }
3798
3799         write_unlock_bh(&idev->lock);
3800
3801         /* Step 5: Discard anycast and multicast list */
3802         if (how) {
3803                 ipv6_ac_destroy_dev(idev);
3804                 ipv6_mc_destroy_dev(idev);
3805         } else {
3806                 ipv6_mc_down(idev);
3807         }
3808
3809         idev->tstamp = jiffies;
3810
3811         /* Last: Shot the device (if unregistered) */
3812         if (how) {
3813                 addrconf_sysctl_unregister(idev);
3814                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3815                 neigh_ifdown(&nd_tbl, dev);
3816                 in6_dev_put(idev);
3817         }
3818         return 0;
3819 }
3820
3821 static void addrconf_rs_timer(struct timer_list *t)
3822 {
3823         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3824         struct net_device *dev = idev->dev;
3825         struct in6_addr lladdr;
3826
3827         write_lock(&idev->lock);
3828         if (idev->dead || !(idev->if_flags & IF_READY))
3829                 goto out;
3830
3831         if (!ipv6_accept_ra(idev))
3832                 goto out;
3833
3834         /* Announcement received after solicitation was sent */
3835         if (idev->if_flags & IF_RA_RCVD)
3836                 goto out;
3837
3838         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3839                 write_unlock(&idev->lock);
3840                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3841                         ndisc_send_rs(dev, &lladdr,
3842                                       &in6addr_linklocal_allrouters);
3843                 else
3844                         goto put;
3845
3846                 write_lock(&idev->lock);
3847                 idev->rs_interval = rfc3315_s14_backoff_update(
3848                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3849                 /* The wait after the last probe can be shorter */
3850                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3851                                              idev->cnf.rtr_solicits) ?
3852                                       idev->cnf.rtr_solicit_delay :
3853                                       idev->rs_interval);
3854         } else {
3855                 /*
3856                  * Note: we do not support deprecated "all on-link"
3857                  * assumption any longer.
3858                  */
3859                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3860         }
3861
3862 out:
3863         write_unlock(&idev->lock);
3864 put:
3865         in6_dev_put(idev);
3866 }
3867
3868 /*
3869  *      Duplicate Address Detection
3870  */
3871 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3872 {
3873         unsigned long rand_num;
3874         struct inet6_dev *idev = ifp->idev;
3875         u64 nonce;
3876
3877         if (ifp->flags & IFA_F_OPTIMISTIC)
3878                 rand_num = 0;
3879         else
3880                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3881
3882         nonce = 0;
3883         if (idev->cnf.enhanced_dad ||
3884             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3885                 do
3886                         get_random_bytes(&nonce, 6);
3887                 while (nonce == 0);
3888         }
3889         ifp->dad_nonce = nonce;
3890         ifp->dad_probes = idev->cnf.dad_transmits;
3891         addrconf_mod_dad_work(ifp, rand_num);
3892 }
3893
3894 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3895 {
3896         struct inet6_dev *idev = ifp->idev;
3897         struct net_device *dev = idev->dev;
3898         bool bump_id, notify = false;
3899         struct net *net;
3900
3901         addrconf_join_solict(dev, &ifp->addr);
3902
3903         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3904
3905         read_lock_bh(&idev->lock);
3906         spin_lock(&ifp->lock);
3907         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3908                 goto out;
3909
3910         net = dev_net(dev);
3911         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3912             (net->ipv6.devconf_all->accept_dad < 1 &&
3913              idev->cnf.accept_dad < 1) ||
3914             !(ifp->flags&IFA_F_TENTATIVE) ||
3915             ifp->flags & IFA_F_NODAD) {
3916                 bool send_na = false;
3917
3918                 if (ifp->flags & IFA_F_TENTATIVE &&
3919                     !(ifp->flags & IFA_F_OPTIMISTIC))
3920                         send_na = true;
3921                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3922                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3923                 spin_unlock(&ifp->lock);
3924                 read_unlock_bh(&idev->lock);
3925
3926                 addrconf_dad_completed(ifp, bump_id, send_na);
3927                 return;
3928         }
3929
3930         if (!(idev->if_flags & IF_READY)) {
3931                 spin_unlock(&ifp->lock);
3932                 read_unlock_bh(&idev->lock);
3933                 /*
3934                  * If the device is not ready:
3935                  * - keep it tentative if it is a permanent address.
3936                  * - otherwise, kill it.
3937                  */
3938                 in6_ifa_hold(ifp);
3939                 addrconf_dad_stop(ifp, 0);
3940                 return;
3941         }
3942
3943         /*
3944          * Optimistic nodes can start receiving
3945          * Frames right away
3946          */
3947         if (ifp->flags & IFA_F_OPTIMISTIC) {
3948                 ip6_ins_rt(net, ifp->rt);
3949                 if (ipv6_use_optimistic_addr(net, idev)) {
3950                         /* Because optimistic nodes can use this address,
3951                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3952                          */
3953                         notify = true;
3954                 }
3955         }
3956
3957         addrconf_dad_kick(ifp);
3958 out:
3959         spin_unlock(&ifp->lock);
3960         read_unlock_bh(&idev->lock);
3961         if (notify)
3962                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3963 }
3964
3965 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3966 {
3967         bool begin_dad = false;
3968
3969         spin_lock_bh(&ifp->lock);
3970         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3971                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3972                 begin_dad = true;
3973         }
3974         spin_unlock_bh(&ifp->lock);
3975
3976         if (begin_dad)
3977                 addrconf_mod_dad_work(ifp, 0);
3978 }
3979
3980 static void addrconf_dad_work(struct work_struct *w)
3981 {
3982         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3983                                                 struct inet6_ifaddr,
3984                                                 dad_work);
3985         struct inet6_dev *idev = ifp->idev;
3986         bool bump_id, disable_ipv6 = false;
3987         struct in6_addr mcaddr;
3988
3989         enum {
3990                 DAD_PROCESS,
3991                 DAD_BEGIN,
3992                 DAD_ABORT,
3993         } action = DAD_PROCESS;
3994
3995         rtnl_lock();
3996
3997         spin_lock_bh(&ifp->lock);
3998         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3999                 action = DAD_BEGIN;
4000                 ifp->state = INET6_IFADDR_STATE_DAD;
4001         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4002                 action = DAD_ABORT;
4003                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4004
4005                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4006                      idev->cnf.accept_dad > 1) &&
4007                     !idev->cnf.disable_ipv6 &&
4008                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4009                         struct in6_addr addr;
4010
4011                         addr.s6_addr32[0] = htonl(0xfe800000);
4012                         addr.s6_addr32[1] = 0;
4013
4014                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4015                             ipv6_addr_equal(&ifp->addr, &addr)) {
4016                                 /* DAD failed for link-local based on MAC */
4017                                 idev->cnf.disable_ipv6 = 1;
4018
4019                                 pr_info("%s: IPv6 being disabled!\n",
4020                                         ifp->idev->dev->name);
4021                                 disable_ipv6 = true;
4022                         }
4023                 }
4024         }
4025         spin_unlock_bh(&ifp->lock);
4026
4027         if (action == DAD_BEGIN) {
4028                 addrconf_dad_begin(ifp);
4029                 goto out;
4030         } else if (action == DAD_ABORT) {
4031                 in6_ifa_hold(ifp);
4032                 addrconf_dad_stop(ifp, 1);
4033                 if (disable_ipv6)
4034                         addrconf_ifdown(idev->dev, 0);
4035                 goto out;
4036         }
4037
4038         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4039                 goto out;
4040
4041         write_lock_bh(&idev->lock);
4042         if (idev->dead || !(idev->if_flags & IF_READY)) {
4043                 write_unlock_bh(&idev->lock);
4044                 goto out;
4045         }
4046
4047         spin_lock(&ifp->lock);
4048         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4049                 spin_unlock(&ifp->lock);
4050                 write_unlock_bh(&idev->lock);
4051                 goto out;
4052         }
4053
4054         if (ifp->dad_probes == 0) {
4055                 bool send_na = false;
4056
4057                 /*
4058                  * DAD was successful
4059                  */
4060
4061                 if (ifp->flags & IFA_F_TENTATIVE &&
4062                     !(ifp->flags & IFA_F_OPTIMISTIC))
4063                         send_na = true;
4064                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4065                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4066                 spin_unlock(&ifp->lock);
4067                 write_unlock_bh(&idev->lock);
4068
4069                 addrconf_dad_completed(ifp, bump_id, send_na);
4070
4071                 goto out;
4072         }
4073
4074         ifp->dad_probes--;
4075         addrconf_mod_dad_work(ifp,
4076                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4077         spin_unlock(&ifp->lock);
4078         write_unlock_bh(&idev->lock);
4079
4080         /* send a neighbour solicitation for our addr */
4081         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4082         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4083                       ifp->dad_nonce);
4084 out:
4085         in6_ifa_put(ifp);
4086         rtnl_unlock();
4087 }
4088
4089 /* ifp->idev must be at least read locked */
4090 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4091 {
4092         struct inet6_ifaddr *ifpiter;
4093         struct inet6_dev *idev = ifp->idev;
4094
4095         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4096                 if (ifpiter->scope > IFA_LINK)
4097                         break;
4098                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4099                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4100                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4101                     IFA_F_PERMANENT)
4102                         return false;
4103         }
4104         return true;
4105 }
4106
4107 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4108                                    bool send_na)
4109 {
4110         struct net_device *dev = ifp->idev->dev;
4111         struct in6_addr lladdr;
4112         bool send_rs, send_mld;
4113
4114         addrconf_del_dad_work(ifp);
4115
4116         /*
4117          *      Configure the address for reception. Now it is valid.
4118          */
4119
4120         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4121
4122         /* If added prefix is link local and we are prepared to process
4123            router advertisements, start sending router solicitations.
4124          */
4125
4126         read_lock_bh(&ifp->idev->lock);
4127         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4128         send_rs = send_mld &&
4129                   ipv6_accept_ra(ifp->idev) &&
4130                   ifp->idev->cnf.rtr_solicits != 0 &&
4131                   (dev->flags&IFF_LOOPBACK) == 0;
4132         read_unlock_bh(&ifp->idev->lock);
4133
4134         /* While dad is in progress mld report's source address is in6_addrany.
4135          * Resend with proper ll now.
4136          */
4137         if (send_mld)
4138                 ipv6_mc_dad_complete(ifp->idev);
4139
4140         /* send unsolicited NA if enabled */
4141         if (send_na &&
4142             (ifp->idev->cnf.ndisc_notify ||
4143              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4144                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4145                               /*router=*/ !!ifp->idev->cnf.forwarding,
4146                               /*solicited=*/ false, /*override=*/ true,
4147                               /*inc_opt=*/ true);
4148         }
4149
4150         if (send_rs) {
4151                 /*
4152                  *      If a host as already performed a random delay
4153                  *      [...] as part of DAD [...] there is no need
4154                  *      to delay again before sending the first RS
4155                  */
4156                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4157                         return;
4158                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4159
4160                 write_lock_bh(&ifp->idev->lock);
4161                 spin_lock(&ifp->lock);
4162                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4163                         ifp->idev->cnf.rtr_solicit_interval);
4164                 ifp->idev->rs_probes = 1;
4165                 ifp->idev->if_flags |= IF_RS_SENT;
4166                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4167                 spin_unlock(&ifp->lock);
4168                 write_unlock_bh(&ifp->idev->lock);
4169         }
4170
4171         if (bump_id)
4172                 rt_genid_bump_ipv6(dev_net(dev));
4173
4174         /* Make sure that a new temporary address will be created
4175          * before this temporary address becomes deprecated.
4176          */
4177         if (ifp->flags & IFA_F_TEMPORARY)
4178                 addrconf_verify_rtnl();
4179 }
4180
4181 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4182 {
4183         struct inet6_ifaddr *ifp;
4184
4185         read_lock_bh(&idev->lock);
4186         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4187                 spin_lock(&ifp->lock);
4188                 if ((ifp->flags & IFA_F_TENTATIVE &&
4189                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4190                         if (restart)
4191                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4192                         addrconf_dad_kick(ifp);
4193                 }
4194                 spin_unlock(&ifp->lock);
4195         }
4196         read_unlock_bh(&idev->lock);
4197 }
4198
4199 #ifdef CONFIG_PROC_FS
4200 struct if6_iter_state {
4201         struct seq_net_private p;
4202         int bucket;
4203         int offset;
4204 };
4205
4206 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4207 {
4208         struct if6_iter_state *state = seq->private;
4209         struct net *net = seq_file_net(seq);
4210         struct inet6_ifaddr *ifa = NULL;
4211         int p = 0;
4212
4213         /* initial bucket if pos is 0 */
4214         if (pos == 0) {
4215                 state->bucket = 0;
4216                 state->offset = 0;
4217         }
4218
4219         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4220                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4221                                          addr_lst) {
4222                         if (!net_eq(dev_net(ifa->idev->dev), net))
4223                                 continue;
4224                         /* sync with offset */
4225                         if (p < state->offset) {
4226                                 p++;
4227                                 continue;
4228                         }
4229                         return ifa;
4230                 }
4231
4232                 /* prepare for next bucket */
4233                 state->offset = 0;
4234                 p = 0;
4235         }
4236         return NULL;
4237 }
4238
4239 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4240                                          struct inet6_ifaddr *ifa)
4241 {
4242         struct if6_iter_state *state = seq->private;
4243         struct net *net = seq_file_net(seq);
4244
4245         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4246                 if (!net_eq(dev_net(ifa->idev->dev), net))
4247                         continue;
4248                 state->offset++;
4249                 return ifa;
4250         }
4251
4252         state->offset = 0;
4253         while (++state->bucket < IN6_ADDR_HSIZE) {
4254                 hlist_for_each_entry_rcu(ifa,
4255                                      &inet6_addr_lst[state->bucket], addr_lst) {
4256                         if (!net_eq(dev_net(ifa->idev->dev), net))
4257                                 continue;
4258                         return ifa;
4259                 }
4260         }
4261
4262         return NULL;
4263 }
4264
4265 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4266         __acquires(rcu)
4267 {
4268         rcu_read_lock();
4269         return if6_get_first(seq, *pos);
4270 }
4271
4272 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4273 {
4274         struct inet6_ifaddr *ifa;
4275
4276         ifa = if6_get_next(seq, v);
4277         ++*pos;
4278         return ifa;
4279 }
4280
4281 static void if6_seq_stop(struct seq_file *seq, void *v)
4282         __releases(rcu)
4283 {
4284         rcu_read_unlock();
4285 }
4286
4287 static int if6_seq_show(struct seq_file *seq, void *v)
4288 {
4289         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4290         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4291                    &ifp->addr,
4292                    ifp->idev->dev->ifindex,
4293                    ifp->prefix_len,
4294                    ifp->scope,
4295                    (u8) ifp->flags,
4296                    ifp->idev->dev->name);
4297         return 0;
4298 }
4299
4300 static const struct seq_operations if6_seq_ops = {
4301         .start  = if6_seq_start,
4302         .next   = if6_seq_next,
4303         .show   = if6_seq_show,
4304         .stop   = if6_seq_stop,
4305 };
4306
4307 static int __net_init if6_proc_net_init(struct net *net)
4308 {
4309         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4310                         sizeof(struct if6_iter_state)))
4311                 return -ENOMEM;
4312         return 0;
4313 }
4314
4315 static void __net_exit if6_proc_net_exit(struct net *net)
4316 {
4317         remove_proc_entry("if_inet6", net->proc_net);
4318 }
4319
4320 static struct pernet_operations if6_proc_net_ops = {
4321         .init = if6_proc_net_init,
4322         .exit = if6_proc_net_exit,
4323 };
4324
4325 int __init if6_proc_init(void)
4326 {
4327         return register_pernet_subsys(&if6_proc_net_ops);
4328 }
4329
4330 void if6_proc_exit(void)
4331 {
4332         unregister_pernet_subsys(&if6_proc_net_ops);
4333 }
4334 #endif  /* CONFIG_PROC_FS */
4335
4336 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4337 /* Check if address is a home address configured on any interface. */
4338 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4339 {
4340         unsigned int hash = inet6_addr_hash(net, addr);
4341         struct inet6_ifaddr *ifp = NULL;
4342         int ret = 0;
4343
4344         rcu_read_lock();
4345         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4346                 if (!net_eq(dev_net(ifp->idev->dev), net))
4347                         continue;
4348                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4349                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4350                         ret = 1;
4351                         break;
4352                 }
4353         }
4354         rcu_read_unlock();
4355         return ret;
4356 }
4357 #endif
4358
4359 /*
4360  *      Periodic address status verification
4361  */
4362
4363 static void addrconf_verify_rtnl(void)
4364 {
4365         unsigned long now, next, next_sec, next_sched;
4366         struct inet6_ifaddr *ifp;
4367         int i;
4368
4369         ASSERT_RTNL();
4370
4371         rcu_read_lock_bh();
4372         now = jiffies;
4373         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4374
4375         cancel_delayed_work(&addr_chk_work);
4376
4377         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4378 restart:
4379                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4380                         unsigned long age;
4381
4382                         /* When setting preferred_lft to a value not zero or
4383                          * infinity, while valid_lft is infinity
4384                          * IFA_F_PERMANENT has a non-infinity life time.
4385                          */
4386                         if ((ifp->flags & IFA_F_PERMANENT) &&
4387                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4388                                 continue;
4389
4390                         spin_lock(&ifp->lock);
4391                         /* We try to batch several events at once. */
4392                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4393
4394                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4395                             age >= ifp->valid_lft) {
4396                                 spin_unlock(&ifp->lock);
4397                                 in6_ifa_hold(ifp);
4398                                 ipv6_del_addr(ifp);
4399                                 goto restart;
4400                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4401                                 spin_unlock(&ifp->lock);
4402                                 continue;
4403                         } else if (age >= ifp->prefered_lft) {
4404                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4405                                 int deprecate = 0;
4406
4407                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4408                                         deprecate = 1;
4409                                         ifp->flags |= IFA_F_DEPRECATED;
4410                                 }
4411
4412                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4413                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4414                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4415
4416                                 spin_unlock(&ifp->lock);
4417
4418                                 if (deprecate) {
4419                                         in6_ifa_hold(ifp);
4420
4421                                         ipv6_ifa_notify(0, ifp);
4422                                         in6_ifa_put(ifp);
4423                                         goto restart;
4424                                 }
4425                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4426                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4427                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4428                                         ifp->idev->cnf.dad_transmits *
4429                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4430
4431                                 if (age >= ifp->prefered_lft - regen_advance) {
4432                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4433                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4434                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4435                                         if (!ifp->regen_count && ifpub) {
4436                                                 ifp->regen_count++;
4437                                                 in6_ifa_hold(ifp);
4438                                                 in6_ifa_hold(ifpub);
4439                                                 spin_unlock(&ifp->lock);
4440
4441                                                 spin_lock(&ifpub->lock);
4442                                                 ifpub->regen_count = 0;
4443                                                 spin_unlock(&ifpub->lock);
4444                                                 rcu_read_unlock_bh();
4445                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4446                                                 in6_ifa_put(ifpub);
4447                                                 in6_ifa_put(ifp);
4448                                                 rcu_read_lock_bh();
4449                                                 goto restart;
4450                                         }
4451                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4452                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4453                                 spin_unlock(&ifp->lock);
4454                         } else {
4455                                 /* ifp->prefered_lft <= ifp->valid_lft */
4456                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4457                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4458                                 spin_unlock(&ifp->lock);
4459                         }
4460                 }
4461         }
4462
4463         next_sec = round_jiffies_up(next);
4464         next_sched = next;
4465
4466         /* If rounded timeout is accurate enough, accept it. */
4467         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4468                 next_sched = next_sec;
4469
4470         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4471         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4472                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4473
4474         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4475                  now, next, next_sec, next_sched);
4476         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4477         rcu_read_unlock_bh();
4478 }
4479
4480 static void addrconf_verify_work(struct work_struct *w)
4481 {
4482         rtnl_lock();
4483         addrconf_verify_rtnl();
4484         rtnl_unlock();
4485 }
4486
4487 static void addrconf_verify(void)
4488 {
4489         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4490 }
4491
4492 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4493                                      struct in6_addr **peer_pfx)
4494 {
4495         struct in6_addr *pfx = NULL;
4496
4497         *peer_pfx = NULL;
4498
4499         if (addr)
4500                 pfx = nla_data(addr);
4501
4502         if (local) {
4503                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4504                         *peer_pfx = pfx;
4505                 pfx = nla_data(local);
4506         }
4507
4508         return pfx;
4509 }
4510
4511 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4512         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4513         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4514         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4515         [IFA_FLAGS]             = { .len = sizeof(u32) },
4516         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4517         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4518 };
4519
4520 static int
4521 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4522                   struct netlink_ext_ack *extack)
4523 {
4524         struct net *net = sock_net(skb->sk);
4525         struct ifaddrmsg *ifm;
4526         struct nlattr *tb[IFA_MAX+1];
4527         struct in6_addr *pfx, *peer_pfx;
4528         u32 ifa_flags;
4529         int err;
4530
4531         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4532                           extack);
4533         if (err < 0)
4534                 return err;
4535
4536         ifm = nlmsg_data(nlh);
4537         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4538         if (!pfx)
4539                 return -EINVAL;
4540
4541         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4542
4543         /* We ignore other flags so far. */
4544         ifa_flags &= IFA_F_MANAGETEMPADDR;
4545
4546         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4547                               ifm->ifa_prefixlen);
4548 }
4549
4550 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4551                                unsigned long expires, u32 flags)
4552 {
4553         struct fib6_info *f6i;
4554         u32 prio;
4555
4556         f6i = addrconf_get_prefix_route(&ifp->addr,
4557                                         ifp->prefix_len,
4558                                         ifp->idev->dev,
4559                                         0, RTF_GATEWAY | RTF_DEFAULT);
4560         if (!f6i)
4561                 return -ENOENT;
4562
4563         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4564         if (f6i->fib6_metric != prio) {
4565                 /* delete old one */
4566                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4567
4568                 /* add new one */
4569                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4570                                       ifp->rt_priority, ifp->idev->dev,
4571                                       expires, flags, GFP_KERNEL);
4572         } else {
4573                 if (!expires)
4574                         fib6_clean_expires(f6i);
4575                 else
4576                         fib6_set_expires(f6i, expires);
4577
4578                 fib6_info_release(f6i);
4579         }
4580
4581         return 0;
4582 }
4583
4584 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4585 {
4586         u32 flags;
4587         clock_t expires;
4588         unsigned long timeout;
4589         bool was_managetempaddr;
4590         bool had_prefixroute;
4591
4592         ASSERT_RTNL();
4593
4594         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4595                 return -EINVAL;
4596
4597         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4598             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4599                 return -EINVAL;
4600
4601         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4602                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4603
4604         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4605         if (addrconf_finite_timeout(timeout)) {
4606                 expires = jiffies_to_clock_t(timeout * HZ);
4607                 cfg->valid_lft = timeout;
4608                 flags = RTF_EXPIRES;
4609         } else {
4610                 expires = 0;
4611                 flags = 0;
4612                 cfg->ifa_flags |= IFA_F_PERMANENT;
4613         }
4614
4615         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4616         if (addrconf_finite_timeout(timeout)) {
4617                 if (timeout == 0)
4618                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4619                 cfg->preferred_lft = timeout;
4620         }
4621
4622         spin_lock_bh(&ifp->lock);
4623         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4624         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4625                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4626         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4627                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4628                         IFA_F_NOPREFIXROUTE);
4629         ifp->flags |= cfg->ifa_flags;
4630         ifp->tstamp = jiffies;
4631         ifp->valid_lft = cfg->valid_lft;
4632         ifp->prefered_lft = cfg->preferred_lft;
4633
4634         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4635                 ifp->rt_priority = cfg->rt_priority;
4636
4637         spin_unlock_bh(&ifp->lock);
4638         if (!(ifp->flags&IFA_F_TENTATIVE))
4639                 ipv6_ifa_notify(0, ifp);
4640
4641         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4642                 int rc = -ENOENT;
4643
4644                 if (had_prefixroute)
4645                         rc = modify_prefix_route(ifp, expires, flags);
4646
4647                 /* prefix route could have been deleted; if so restore it */
4648                 if (rc == -ENOENT) {
4649                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4650                                               ifp->rt_priority, ifp->idev->dev,
4651                                               expires, flags, GFP_KERNEL);
4652                 }
4653         } else if (had_prefixroute) {
4654                 enum cleanup_prefix_rt_t action;
4655                 unsigned long rt_expires;
4656
4657                 write_lock_bh(&ifp->idev->lock);
4658                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4659                 write_unlock_bh(&ifp->idev->lock);
4660
4661                 if (action != CLEANUP_PREFIX_RT_NOP) {
4662                         cleanup_prefix_route(ifp, rt_expires,
4663                                 action == CLEANUP_PREFIX_RT_DEL);
4664                 }
4665         }
4666
4667         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4668                 if (was_managetempaddr &&
4669                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4670                         cfg->valid_lft = 0;
4671                         cfg->preferred_lft = 0;
4672                 }
4673                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4674                                  cfg->preferred_lft, !was_managetempaddr,
4675                                  jiffies);
4676         }
4677
4678         addrconf_verify_rtnl();
4679
4680         return 0;
4681 }
4682
4683 static int
4684 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4685                   struct netlink_ext_ack *extack)
4686 {
4687         struct net *net = sock_net(skb->sk);
4688         struct ifaddrmsg *ifm;
4689         struct nlattr *tb[IFA_MAX+1];
4690         struct in6_addr *peer_pfx;
4691         struct inet6_ifaddr *ifa;
4692         struct net_device *dev;
4693         struct inet6_dev *idev;
4694         struct ifa6_config cfg;
4695         int err;
4696
4697         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4698                           extack);
4699         if (err < 0)
4700                 return err;
4701
4702         memset(&cfg, 0, sizeof(cfg));
4703
4704         ifm = nlmsg_data(nlh);
4705         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4706         if (!cfg.pfx)
4707                 return -EINVAL;
4708
4709         cfg.peer_pfx = peer_pfx;
4710         cfg.plen = ifm->ifa_prefixlen;
4711         if (tb[IFA_RT_PRIORITY])
4712                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4713
4714         cfg.valid_lft = INFINITY_LIFE_TIME;
4715         cfg.preferred_lft = INFINITY_LIFE_TIME;
4716
4717         if (tb[IFA_CACHEINFO]) {
4718                 struct ifa_cacheinfo *ci;
4719
4720                 ci = nla_data(tb[IFA_CACHEINFO]);
4721                 cfg.valid_lft = ci->ifa_valid;
4722                 cfg.preferred_lft = ci->ifa_prefered;
4723         }
4724
4725         dev =  __dev_get_by_index(net, ifm->ifa_index);
4726         if (!dev)
4727                 return -ENODEV;
4728
4729         if (tb[IFA_FLAGS])
4730                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4731         else
4732                 cfg.ifa_flags = ifm->ifa_flags;
4733
4734         /* We ignore other flags so far. */
4735         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4736                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4737                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4738
4739         idev = ipv6_find_idev(dev);
4740         if (!idev)
4741                 return -ENOBUFS;
4742
4743         if (!ipv6_allow_optimistic_dad(net, idev))
4744                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4745
4746         if (cfg.ifa_flags & IFA_F_NODAD &&
4747             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4748                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4749                 return -EINVAL;
4750         }
4751
4752         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4753         if (!ifa) {
4754                 /*
4755                  * It would be best to check for !NLM_F_CREATE here but
4756                  * userspace already relies on not having to provide this.
4757                  */
4758                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4759         }
4760
4761         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4762             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4763                 err = -EEXIST;
4764         else
4765                 err = inet6_addr_modify(ifa, &cfg);
4766
4767         in6_ifa_put(ifa);
4768
4769         return err;
4770 }
4771
4772 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4773                           u8 scope, int ifindex)
4774 {
4775         struct ifaddrmsg *ifm;
4776
4777         ifm = nlmsg_data(nlh);
4778         ifm->ifa_family = AF_INET6;
4779         ifm->ifa_prefixlen = prefixlen;
4780         ifm->ifa_flags = flags;
4781         ifm->ifa_scope = scope;
4782         ifm->ifa_index = ifindex;
4783 }
4784
4785 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4786                          unsigned long tstamp, u32 preferred, u32 valid)
4787 {
4788         struct ifa_cacheinfo ci;
4789
4790         ci.cstamp = cstamp_delta(cstamp);
4791         ci.tstamp = cstamp_delta(tstamp);
4792         ci.ifa_prefered = preferred;
4793         ci.ifa_valid = valid;
4794
4795         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4796 }
4797
4798 static inline int rt_scope(int ifa_scope)
4799 {
4800         if (ifa_scope & IFA_HOST)
4801                 return RT_SCOPE_HOST;
4802         else if (ifa_scope & IFA_LINK)
4803                 return RT_SCOPE_LINK;
4804         else if (ifa_scope & IFA_SITE)
4805                 return RT_SCOPE_SITE;
4806         else
4807                 return RT_SCOPE_UNIVERSE;
4808 }
4809
4810 static inline int inet6_ifaddr_msgsize(void)
4811 {
4812         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4813                + nla_total_size(16) /* IFA_LOCAL */
4814                + nla_total_size(16) /* IFA_ADDRESS */
4815                + nla_total_size(sizeof(struct ifa_cacheinfo))
4816                + nla_total_size(4)  /* IFA_FLAGS */
4817                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4818 }
4819
4820 enum addr_type_t {
4821         UNICAST_ADDR,
4822         MULTICAST_ADDR,
4823         ANYCAST_ADDR,
4824 };
4825
4826 struct inet6_fill_args {
4827         u32 portid;
4828         u32 seq;
4829         int event;
4830         unsigned int flags;
4831         int netnsid;
4832         int ifindex;
4833         enum addr_type_t type;
4834 };
4835
4836 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4837                              struct inet6_fill_args *args)
4838 {
4839         struct nlmsghdr  *nlh;
4840         u32 preferred, valid;
4841
4842         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4843                         sizeof(struct ifaddrmsg), args->flags);
4844         if (!nlh)
4845                 return -EMSGSIZE;
4846
4847         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4848                       ifa->idev->dev->ifindex);
4849
4850         if (args->netnsid >= 0 &&
4851             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4852                 goto error;
4853
4854         if (!((ifa->flags&IFA_F_PERMANENT) &&
4855               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4856                 preferred = ifa->prefered_lft;
4857                 valid = ifa->valid_lft;
4858                 if (preferred != INFINITY_LIFE_TIME) {
4859                         long tval = (jiffies - ifa->tstamp)/HZ;
4860                         if (preferred > tval)
4861                                 preferred -= tval;
4862                         else
4863                                 preferred = 0;
4864                         if (valid != INFINITY_LIFE_TIME) {
4865                                 if (valid > tval)
4866                                         valid -= tval;
4867                                 else
4868                                         valid = 0;
4869                         }
4870                 }
4871         } else {
4872                 preferred = INFINITY_LIFE_TIME;
4873                 valid = INFINITY_LIFE_TIME;
4874         }
4875
4876         if (!ipv6_addr_any(&ifa->peer_addr)) {
4877                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4878                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4879                         goto error;
4880         } else
4881                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4882                         goto error;
4883
4884         if (ifa->rt_priority &&
4885             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4886                 goto error;
4887
4888         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4889                 goto error;
4890
4891         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4892                 goto error;
4893
4894         nlmsg_end(skb, nlh);
4895         return 0;
4896
4897 error:
4898         nlmsg_cancel(skb, nlh);
4899         return -EMSGSIZE;
4900 }
4901
4902 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4903                                struct inet6_fill_args *args)
4904 {
4905         struct nlmsghdr  *nlh;
4906         u8 scope = RT_SCOPE_UNIVERSE;
4907         int ifindex = ifmca->idev->dev->ifindex;
4908
4909         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4910                 scope = RT_SCOPE_SITE;
4911
4912         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4913                         sizeof(struct ifaddrmsg), args->flags);
4914         if (!nlh)
4915                 return -EMSGSIZE;
4916
4917         if (args->netnsid >= 0 &&
4918             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4919                 return -EMSGSIZE;
4920
4921         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4922         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4923             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4924                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4925                 nlmsg_cancel(skb, nlh);
4926                 return -EMSGSIZE;
4927         }
4928
4929         nlmsg_end(skb, nlh);
4930         return 0;
4931 }
4932
4933 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4934                                struct inet6_fill_args *args)
4935 {
4936         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4937         int ifindex = dev ? dev->ifindex : 1;
4938         struct nlmsghdr  *nlh;
4939         u8 scope = RT_SCOPE_UNIVERSE;
4940
4941         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4942                 scope = RT_SCOPE_SITE;
4943
4944         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4945                         sizeof(struct ifaddrmsg), args->flags);
4946         if (!nlh)
4947                 return -EMSGSIZE;
4948
4949         if (args->netnsid >= 0 &&
4950             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4951                 return -EMSGSIZE;
4952
4953         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4954         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4955             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4956                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4957                 nlmsg_cancel(skb, nlh);
4958                 return -EMSGSIZE;
4959         }
4960
4961         nlmsg_end(skb, nlh);
4962         return 0;
4963 }
4964
4965 /* called with rcu_read_lock() */
4966 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4967                           struct netlink_callback *cb, int s_ip_idx,
4968                           struct inet6_fill_args *fillargs)
4969 {
4970         struct ifmcaddr6 *ifmca;
4971         struct ifacaddr6 *ifaca;
4972         int ip_idx = 0;
4973         int err = 1;
4974
4975         read_lock_bh(&idev->lock);
4976         switch (fillargs->type) {
4977         case UNICAST_ADDR: {
4978                 struct inet6_ifaddr *ifa;
4979                 fillargs->event = RTM_NEWADDR;
4980
4981                 /* unicast address incl. temp addr */
4982                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4983                         if (ip_idx < s_ip_idx)
4984                                 goto next;
4985                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
4986                         if (err < 0)
4987                                 break;
4988                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4989 next:
4990                         ip_idx++;
4991                 }
4992                 break;
4993         }
4994         case MULTICAST_ADDR:
4995                 fillargs->event = RTM_GETMULTICAST;
4996
4997                 /* multicast address */
4998                 for (ifmca = idev->mc_list; ifmca;
4999                      ifmca = ifmca->next, ip_idx++) {
5000                         if (ip_idx < s_ip_idx)
5001                                 continue;
5002                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5003                         if (err < 0)
5004                                 break;
5005                 }
5006                 break;
5007         case ANYCAST_ADDR:
5008                 fillargs->event = RTM_GETANYCAST;
5009                 /* anycast address */
5010                 for (ifaca = idev->ac_list; ifaca;
5011                      ifaca = ifaca->aca_next, ip_idx++) {
5012                         if (ip_idx < s_ip_idx)
5013                                 continue;
5014                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5015                         if (err < 0)
5016                                 break;
5017                 }
5018                 break;
5019         default:
5020                 break;
5021         }
5022         read_unlock_bh(&idev->lock);
5023         cb->args[2] = ip_idx;
5024         return err;
5025 }
5026
5027 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5028                                        struct inet6_fill_args *fillargs,
5029                                        struct net **tgt_net, struct sock *sk,
5030                                        struct netlink_callback *cb)
5031 {
5032         struct netlink_ext_ack *extack = cb->extack;
5033         struct nlattr *tb[IFA_MAX+1];
5034         struct ifaddrmsg *ifm;
5035         int err, i;
5036
5037         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5038                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5039                 return -EINVAL;
5040         }
5041
5042         ifm = nlmsg_data(nlh);
5043         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5044                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5045                 return -EINVAL;
5046         }
5047
5048         fillargs->ifindex = ifm->ifa_index;
5049         if (fillargs->ifindex) {
5050                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5051                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5052         }
5053
5054         err = nlmsg_parse_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5055                                  ifa_ipv6_policy, extack);
5056         if (err < 0)
5057                 return err;
5058
5059         for (i = 0; i <= IFA_MAX; ++i) {
5060                 if (!tb[i])
5061                         continue;
5062
5063                 if (i == IFA_TARGET_NETNSID) {
5064                         struct net *net;
5065
5066                         fillargs->netnsid = nla_get_s32(tb[i]);
5067                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5068                         if (IS_ERR(net)) {
5069                                 fillargs->netnsid = -1;
5070                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5071                                 return PTR_ERR(net);
5072                         }
5073                         *tgt_net = net;
5074                 } else {
5075                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5076                         return -EINVAL;
5077                 }
5078         }
5079
5080         return 0;
5081 }
5082
5083 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5084                            enum addr_type_t type)
5085 {
5086         const struct nlmsghdr *nlh = cb->nlh;
5087         struct inet6_fill_args fillargs = {
5088                 .portid = NETLINK_CB(cb->skb).portid,
5089                 .seq = cb->nlh->nlmsg_seq,
5090                 .flags = NLM_F_MULTI,
5091                 .netnsid = -1,
5092                 .type = type,
5093         };
5094         struct net *net = sock_net(skb->sk);
5095         struct net *tgt_net = net;
5096         int idx, s_idx, s_ip_idx;
5097         int h, s_h;
5098         struct net_device *dev;
5099         struct inet6_dev *idev;
5100         struct hlist_head *head;
5101         int err = 0;
5102
5103         s_h = cb->args[0];
5104         s_idx = idx = cb->args[1];
5105         s_ip_idx = cb->args[2];
5106
5107         if (cb->strict_check) {
5108                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5109                                                   skb->sk, cb);
5110                 if (err < 0)
5111                         goto put_tgt_net;
5112
5113                 err = 0;
5114                 if (fillargs.ifindex) {
5115                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5116                         if (!dev) {
5117                                 err = -ENODEV;
5118                                 goto put_tgt_net;
5119                         }
5120                         idev = __in6_dev_get(dev);
5121                         if (idev) {
5122                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5123                                                      &fillargs);
5124                                 if (err > 0)
5125                                         err = 0;
5126                         }
5127                         goto put_tgt_net;
5128                 }
5129         }
5130
5131         rcu_read_lock();
5132         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5133         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5134                 idx = 0;
5135                 head = &tgt_net->dev_index_head[h];
5136                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5137                         if (idx < s_idx)
5138                                 goto cont;
5139                         if (h > s_h || idx > s_idx)
5140                                 s_ip_idx = 0;
5141                         idev = __in6_dev_get(dev);
5142                         if (!idev)
5143                                 goto cont;
5144
5145                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5146                                            &fillargs) < 0)
5147                                 goto done;
5148 cont:
5149                         idx++;
5150                 }
5151         }
5152 done:
5153         rcu_read_unlock();
5154         cb->args[0] = h;
5155         cb->args[1] = idx;
5156 put_tgt_net:
5157         if (fillargs.netnsid >= 0)
5158                 put_net(tgt_net);
5159
5160         return skb->len ? : err;
5161 }
5162
5163 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5164 {
5165         enum addr_type_t type = UNICAST_ADDR;
5166
5167         return inet6_dump_addr(skb, cb, type);
5168 }
5169
5170 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5171 {
5172         enum addr_type_t type = MULTICAST_ADDR;
5173
5174         return inet6_dump_addr(skb, cb, type);
5175 }
5176
5177
5178 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5179 {
5180         enum addr_type_t type = ANYCAST_ADDR;
5181
5182         return inet6_dump_addr(skb, cb, type);
5183 }
5184
5185 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5186                              struct netlink_ext_ack *extack)
5187 {
5188         struct net *net = sock_net(in_skb->sk);
5189         struct inet6_fill_args fillargs = {
5190                 .portid = NETLINK_CB(in_skb).portid,
5191                 .seq = nlh->nlmsg_seq,
5192                 .event = RTM_NEWADDR,
5193                 .flags = 0,
5194                 .netnsid = -1,
5195         };
5196         struct net *tgt_net = net;
5197         struct ifaddrmsg *ifm;
5198         struct nlattr *tb[IFA_MAX+1];
5199         struct in6_addr *addr = NULL, *peer;
5200         struct net_device *dev = NULL;
5201         struct inet6_ifaddr *ifa;
5202         struct sk_buff *skb;
5203         int err;
5204
5205         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5206                           extack);
5207         if (err < 0)
5208                 return err;
5209
5210         if (tb[IFA_TARGET_NETNSID]) {
5211                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5212
5213                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5214                                                   fillargs.netnsid);
5215                 if (IS_ERR(tgt_net))
5216                         return PTR_ERR(tgt_net);
5217         }
5218
5219         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5220         if (!addr)
5221                 return -EINVAL;
5222
5223         ifm = nlmsg_data(nlh);
5224         if (ifm->ifa_index)
5225                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5226
5227         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5228         if (!ifa) {
5229                 err = -EADDRNOTAVAIL;
5230                 goto errout;
5231         }
5232
5233         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5234         if (!skb) {
5235                 err = -ENOBUFS;
5236                 goto errout_ifa;
5237         }
5238
5239         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5240         if (err < 0) {
5241                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5242                 WARN_ON(err == -EMSGSIZE);
5243                 kfree_skb(skb);
5244                 goto errout_ifa;
5245         }
5246         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5247 errout_ifa:
5248         in6_ifa_put(ifa);
5249 errout:
5250         if (dev)
5251                 dev_put(dev);
5252         if (fillargs.netnsid >= 0)
5253                 put_net(tgt_net);
5254
5255         return err;
5256 }
5257
5258 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5259 {
5260         struct sk_buff *skb;
5261         struct net *net = dev_net(ifa->idev->dev);
5262         struct inet6_fill_args fillargs = {
5263                 .portid = 0,
5264                 .seq = 0,
5265                 .event = event,
5266                 .flags = 0,
5267                 .netnsid = -1,
5268         };
5269         int err = -ENOBUFS;
5270
5271         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5272         if (!skb)
5273                 goto errout;
5274
5275         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5276         if (err < 0) {
5277                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5278                 WARN_ON(err == -EMSGSIZE);
5279                 kfree_skb(skb);
5280                 goto errout;
5281         }
5282         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5283         return;
5284 errout:
5285         if (err < 0)
5286                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5287 }
5288
5289 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5290                                 __s32 *array, int bytes)
5291 {
5292         BUG_ON(bytes < (DEVCONF_MAX * 4));
5293
5294         memset(array, 0, bytes);
5295         array[DEVCONF_FORWARDING] = cnf->forwarding;
5296         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5297         array[DEVCONF_MTU6] = cnf->mtu6;
5298         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5299         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5300         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5301         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5302         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5303         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5304                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5305         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5306                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5307         array[DEVCONF_RTR_SOLICIT_DELAY] =
5308                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5309         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5310         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5311                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5312         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5313                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5314         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5315         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5316         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5317         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5318         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5319         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5320         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5321         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5322         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5323 #ifdef CONFIG_IPV6_ROUTER_PREF
5324         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5325         array[DEVCONF_RTR_PROBE_INTERVAL] =
5326                 jiffies_to_msecs(cnf->rtr_probe_interval);
5327 #ifdef CONFIG_IPV6_ROUTE_INFO
5328         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5329         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5330 #endif
5331 #endif
5332         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5333         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5334 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5335         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5336         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5337 #endif
5338 #ifdef CONFIG_IPV6_MROUTE
5339         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5340 #endif
5341         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5342         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5343         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5344         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5345         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5346         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5347         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5348         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5349         /* we omit DEVCONF_STABLE_SECRET for now */
5350         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5351         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5352         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5353         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5354         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5355 #ifdef CONFIG_IPV6_SEG6_HMAC
5356         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5357 #endif
5358         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5359         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5360         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5361         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5362 }
5363
5364 static inline size_t inet6_ifla6_size(void)
5365 {
5366         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5367              + nla_total_size(sizeof(struct ifla_cacheinfo))
5368              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5369              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5370              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5371              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5372              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5373              + 0;
5374 }
5375
5376 static inline size_t inet6_if_nlmsg_size(void)
5377 {
5378         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5379                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5380                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5381                + nla_total_size(4) /* IFLA_MTU */
5382                + nla_total_size(4) /* IFLA_LINK */
5383                + nla_total_size(1) /* IFLA_OPERSTATE */
5384                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5385 }
5386
5387 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5388                                         int bytes)
5389 {
5390         int i;
5391         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5392         BUG_ON(pad < 0);
5393
5394         /* Use put_unaligned() because stats may not be aligned for u64. */
5395         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5396         for (i = 1; i < ICMP6_MIB_MAX; i++)
5397                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5398
5399         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5400 }
5401
5402 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5403                                         int bytes, size_t syncpoff)
5404 {
5405         int i, c;
5406         u64 buff[IPSTATS_MIB_MAX];
5407         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5408
5409         BUG_ON(pad < 0);
5410
5411         memset(buff, 0, sizeof(buff));
5412         buff[0] = IPSTATS_MIB_MAX;
5413
5414         for_each_possible_cpu(c) {
5415                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5416                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5417         }
5418
5419         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5420         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5421 }
5422
5423 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5424                              int bytes)
5425 {
5426         switch (attrtype) {
5427         case IFLA_INET6_STATS:
5428                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5429                                      offsetof(struct ipstats_mib, syncp));
5430                 break;
5431         case IFLA_INET6_ICMP6STATS:
5432                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5433                 break;
5434         }
5435 }
5436
5437 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5438                                   u32 ext_filter_mask)
5439 {
5440         struct nlattr *nla;
5441         struct ifla_cacheinfo ci;
5442
5443         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5444                 goto nla_put_failure;
5445         ci.max_reasm_len = IPV6_MAXPLEN;
5446         ci.tstamp = cstamp_delta(idev->tstamp);
5447         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5448         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5449         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5450                 goto nla_put_failure;
5451         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5452         if (!nla)
5453                 goto nla_put_failure;
5454         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5455
5456         /* XXX - MC not implemented */
5457
5458         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5459                 return 0;
5460
5461         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5462         if (!nla)
5463                 goto nla_put_failure;
5464         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5465
5466         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5467         if (!nla)
5468                 goto nla_put_failure;
5469         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5470
5471         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5472         if (!nla)
5473                 goto nla_put_failure;
5474
5475         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5476                 goto nla_put_failure;
5477
5478         read_lock_bh(&idev->lock);
5479         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5480         read_unlock_bh(&idev->lock);
5481
5482         return 0;
5483
5484 nla_put_failure:
5485         return -EMSGSIZE;
5486 }
5487
5488 static size_t inet6_get_link_af_size(const struct net_device *dev,
5489                                      u32 ext_filter_mask)
5490 {
5491         if (!__in6_dev_get(dev))
5492                 return 0;
5493
5494         return inet6_ifla6_size();
5495 }
5496
5497 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5498                               u32 ext_filter_mask)
5499 {
5500         struct inet6_dev *idev = __in6_dev_get(dev);
5501
5502         if (!idev)
5503                 return -ENODATA;
5504
5505         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5506                 return -EMSGSIZE;
5507
5508         return 0;
5509 }
5510
5511 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5512 {
5513         struct inet6_ifaddr *ifp;
5514         struct net_device *dev = idev->dev;
5515         bool clear_token, update_rs = false;
5516         struct in6_addr ll_addr;
5517
5518         ASSERT_RTNL();
5519
5520         if (!token)
5521                 return -EINVAL;
5522         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5523                 return -EINVAL;
5524         if (!ipv6_accept_ra(idev))
5525                 return -EINVAL;
5526         if (idev->cnf.rtr_solicits == 0)
5527                 return -EINVAL;
5528
5529         write_lock_bh(&idev->lock);
5530
5531         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5532         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5533
5534         write_unlock_bh(&idev->lock);
5535
5536         clear_token = ipv6_addr_any(token);
5537         if (clear_token)
5538                 goto update_lft;
5539
5540         if (!idev->dead && (idev->if_flags & IF_READY) &&
5541             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5542                              IFA_F_OPTIMISTIC)) {
5543                 /* If we're not ready, then normal ifup will take care
5544                  * of this. Otherwise, we need to request our rs here.
5545                  */
5546                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5547                 update_rs = true;
5548         }
5549
5550 update_lft:
5551         write_lock_bh(&idev->lock);
5552
5553         if (update_rs) {
5554                 idev->if_flags |= IF_RS_SENT;
5555                 idev->rs_interval = rfc3315_s14_backoff_init(
5556                         idev->cnf.rtr_solicit_interval);
5557                 idev->rs_probes = 1;
5558                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5559         }
5560
5561         /* Well, that's kinda nasty ... */
5562         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5563                 spin_lock(&ifp->lock);
5564                 if (ifp->tokenized) {
5565                         ifp->valid_lft = 0;
5566                         ifp->prefered_lft = 0;
5567                 }
5568                 spin_unlock(&ifp->lock);
5569         }
5570
5571         write_unlock_bh(&idev->lock);
5572         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5573         addrconf_verify_rtnl();
5574         return 0;
5575 }
5576
5577 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5578         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5579         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5580 };
5581
5582 static int inet6_validate_link_af(const struct net_device *dev,
5583                                   const struct nlattr *nla)
5584 {
5585         struct nlattr *tb[IFLA_INET6_MAX + 1];
5586
5587         if (dev && !__in6_dev_get(dev))
5588                 return -EAFNOSUPPORT;
5589
5590         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5591                                 NULL);
5592 }
5593
5594 static int check_addr_gen_mode(int mode)
5595 {
5596         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5597             mode != IN6_ADDR_GEN_MODE_NONE &&
5598             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5599             mode != IN6_ADDR_GEN_MODE_RANDOM)
5600                 return -EINVAL;
5601         return 1;
5602 }
5603
5604 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5605                                 int mode)
5606 {
5607         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5608             !idev->cnf.stable_secret.initialized &&
5609             !net->ipv6.devconf_dflt->stable_secret.initialized)
5610                 return -EINVAL;
5611         return 1;
5612 }
5613
5614 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5615 {
5616         int err = -EINVAL;
5617         struct inet6_dev *idev = __in6_dev_get(dev);
5618         struct nlattr *tb[IFLA_INET6_MAX + 1];
5619
5620         if (!idev)
5621                 return -EAFNOSUPPORT;
5622
5623         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5624                 BUG();
5625
5626         if (tb[IFLA_INET6_TOKEN]) {
5627                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5628                 if (err)
5629                         return err;
5630         }
5631
5632         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5633                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5634
5635                 if (check_addr_gen_mode(mode) < 0 ||
5636                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5637                         return -EINVAL;
5638
5639                 idev->cnf.addr_gen_mode = mode;
5640                 err = 0;
5641         }
5642
5643         return err;
5644 }
5645
5646 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5647                              u32 portid, u32 seq, int event, unsigned int flags)
5648 {
5649         struct net_device *dev = idev->dev;
5650         struct ifinfomsg *hdr;
5651         struct nlmsghdr *nlh;
5652         void *protoinfo;
5653
5654         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5655         if (!nlh)
5656                 return -EMSGSIZE;
5657
5658         hdr = nlmsg_data(nlh);
5659         hdr->ifi_family = AF_INET6;
5660         hdr->__ifi_pad = 0;
5661         hdr->ifi_type = dev->type;
5662         hdr->ifi_index = dev->ifindex;
5663         hdr->ifi_flags = dev_get_flags(dev);
5664         hdr->ifi_change = 0;
5665
5666         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5667             (dev->addr_len &&
5668              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5669             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5670             (dev->ifindex != dev_get_iflink(dev) &&
5671              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5672             nla_put_u8(skb, IFLA_OPERSTATE,
5673                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5674                 goto nla_put_failure;
5675         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5676         if (!protoinfo)
5677                 goto nla_put_failure;
5678
5679         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5680                 goto nla_put_failure;
5681
5682         nla_nest_end(skb, protoinfo);
5683         nlmsg_end(skb, nlh);
5684         return 0;
5685
5686 nla_put_failure:
5687         nlmsg_cancel(skb, nlh);
5688         return -EMSGSIZE;
5689 }
5690
5691 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5692                                    struct netlink_ext_ack *extack)
5693 {
5694         struct ifinfomsg *ifm;
5695
5696         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5697                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5698                 return -EINVAL;
5699         }
5700
5701         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5702                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5703                 return -EINVAL;
5704         }
5705
5706         ifm = nlmsg_data(nlh);
5707         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5708             ifm->ifi_change || ifm->ifi_index) {
5709                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5710                 return -EINVAL;
5711         }
5712
5713         return 0;
5714 }
5715
5716 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5717 {
5718         struct net *net = sock_net(skb->sk);
5719         int h, s_h;
5720         int idx = 0, s_idx;
5721         struct net_device *dev;
5722         struct inet6_dev *idev;
5723         struct hlist_head *head;
5724
5725         /* only requests using strict checking can pass data to
5726          * influence the dump
5727          */
5728         if (cb->strict_check) {
5729                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5730
5731                 if (err < 0)
5732                         return err;
5733         }
5734
5735         s_h = cb->args[0];
5736         s_idx = cb->args[1];
5737
5738         rcu_read_lock();
5739         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5740                 idx = 0;
5741                 head = &net->dev_index_head[h];
5742                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5743                         if (idx < s_idx)
5744                                 goto cont;
5745                         idev = __in6_dev_get(dev);
5746                         if (!idev)
5747                                 goto cont;
5748                         if (inet6_fill_ifinfo(skb, idev,
5749                                               NETLINK_CB(cb->skb).portid,
5750                                               cb->nlh->nlmsg_seq,
5751                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5752                                 goto out;
5753 cont:
5754                         idx++;
5755                 }
5756         }
5757 out:
5758         rcu_read_unlock();
5759         cb->args[1] = idx;
5760         cb->args[0] = h;
5761
5762         return skb->len;
5763 }
5764
5765 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5766 {
5767         struct sk_buff *skb;
5768         struct net *net = dev_net(idev->dev);
5769         int err = -ENOBUFS;
5770
5771         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5772         if (!skb)
5773                 goto errout;
5774
5775         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5776         if (err < 0) {
5777                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5778                 WARN_ON(err == -EMSGSIZE);
5779                 kfree_skb(skb);
5780                 goto errout;
5781         }
5782         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5783         return;
5784 errout:
5785         if (err < 0)
5786                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5787 }
5788
5789 static inline size_t inet6_prefix_nlmsg_size(void)
5790 {
5791         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5792                + nla_total_size(sizeof(struct in6_addr))
5793                + nla_total_size(sizeof(struct prefix_cacheinfo));
5794 }
5795
5796 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5797                              struct prefix_info *pinfo, u32 portid, u32 seq,
5798                              int event, unsigned int flags)
5799 {
5800         struct prefixmsg *pmsg;
5801         struct nlmsghdr *nlh;
5802         struct prefix_cacheinfo ci;
5803
5804         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5805         if (!nlh)
5806                 return -EMSGSIZE;
5807
5808         pmsg = nlmsg_data(nlh);
5809         pmsg->prefix_family = AF_INET6;
5810         pmsg->prefix_pad1 = 0;
5811         pmsg->prefix_pad2 = 0;
5812         pmsg->prefix_ifindex = idev->dev->ifindex;
5813         pmsg->prefix_len = pinfo->prefix_len;
5814         pmsg->prefix_type = pinfo->type;
5815         pmsg->prefix_pad3 = 0;
5816         pmsg->prefix_flags = 0;
5817         if (pinfo->onlink)
5818                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5819         if (pinfo->autoconf)
5820                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5821
5822         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5823                 goto nla_put_failure;
5824         ci.preferred_time = ntohl(pinfo->prefered);
5825         ci.valid_time = ntohl(pinfo->valid);
5826         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5827                 goto nla_put_failure;
5828         nlmsg_end(skb, nlh);
5829         return 0;
5830
5831 nla_put_failure:
5832         nlmsg_cancel(skb, nlh);
5833         return -EMSGSIZE;
5834 }
5835
5836 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5837                          struct prefix_info *pinfo)
5838 {
5839         struct sk_buff *skb;
5840         struct net *net = dev_net(idev->dev);
5841         int err = -ENOBUFS;
5842
5843         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5844         if (!skb)
5845                 goto errout;
5846
5847         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5848         if (err < 0) {
5849                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5850                 WARN_ON(err == -EMSGSIZE);
5851                 kfree_skb(skb);
5852                 goto errout;
5853         }
5854         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5855         return;
5856 errout:
5857         if (err < 0)
5858                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5859 }
5860
5861 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5862 {
5863         struct net *net = dev_net(ifp->idev->dev);
5864
5865         if (event)
5866                 ASSERT_RTNL();
5867
5868         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5869
5870         switch (event) {
5871         case RTM_NEWADDR:
5872                 /*
5873                  * If the address was optimistic
5874                  * we inserted the route at the start of
5875                  * our DAD process, so we don't need
5876                  * to do it again
5877                  */
5878                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5879                         ip6_ins_rt(net, ifp->rt);
5880                 if (ifp->idev->cnf.forwarding)
5881                         addrconf_join_anycast(ifp);
5882                 if (!ipv6_addr_any(&ifp->peer_addr))
5883                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5884                                               ifp->idev->dev, 0, 0,
5885                                               GFP_ATOMIC);
5886                 break;
5887         case RTM_DELADDR:
5888                 if (ifp->idev->cnf.forwarding)
5889                         addrconf_leave_anycast(ifp);
5890                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5891                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5892                         struct fib6_info *rt;
5893
5894                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5895                                                        ifp->idev->dev, 0, 0);
5896                         if (rt)
5897                                 ip6_del_rt(net, rt);
5898                 }
5899                 if (ifp->rt) {
5900                         ip6_del_rt(net, ifp->rt);
5901                         ifp->rt = NULL;
5902                 }
5903                 rt_genid_bump_ipv6(net);
5904                 break;
5905         }
5906         atomic_inc(&net->ipv6.dev_addr_genid);
5907 }
5908
5909 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5910 {
5911         rcu_read_lock_bh();
5912         if (likely(ifp->idev->dead == 0))
5913                 __ipv6_ifa_notify(event, ifp);
5914         rcu_read_unlock_bh();
5915 }
5916
5917 #ifdef CONFIG_SYSCTL
5918
5919 static
5920 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5921                            void __user *buffer, size_t *lenp, loff_t *ppos)
5922 {
5923         int *valp = ctl->data;
5924         int val = *valp;
5925         loff_t pos = *ppos;
5926         struct ctl_table lctl;
5927         int ret;
5928
5929         /*
5930          * ctl->data points to idev->cnf.forwarding, we should
5931          * not modify it until we get the rtnl lock.
5932          */
5933         lctl = *ctl;
5934         lctl.data = &val;
5935
5936         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5937
5938         if (write)
5939                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5940         if (ret)
5941                 *ppos = pos;
5942         return ret;
5943 }
5944
5945 static
5946 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5947                         void __user *buffer, size_t *lenp, loff_t *ppos)
5948 {
5949         struct inet6_dev *idev = ctl->extra1;
5950         int min_mtu = IPV6_MIN_MTU;
5951         struct ctl_table lctl;
5952
5953         lctl = *ctl;
5954         lctl.extra1 = &min_mtu;
5955         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5956
5957         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5958 }
5959
5960 static void dev_disable_change(struct inet6_dev *idev)
5961 {
5962         struct netdev_notifier_info info;
5963
5964         if (!idev || !idev->dev)
5965                 return;
5966
5967         netdev_notifier_info_init(&info, idev->dev);
5968         if (idev->cnf.disable_ipv6)
5969                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5970         else
5971                 addrconf_notify(NULL, NETDEV_UP, &info);
5972 }
5973
5974 static void addrconf_disable_change(struct net *net, __s32 newf)
5975 {
5976         struct net_device *dev;
5977         struct inet6_dev *idev;
5978
5979         for_each_netdev(net, dev) {
5980                 idev = __in6_dev_get(dev);
5981                 if (idev) {
5982                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5983                         idev->cnf.disable_ipv6 = newf;
5984                         if (changed)
5985                                 dev_disable_change(idev);
5986                 }
5987         }
5988 }
5989
5990 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5991 {
5992         struct net *net;
5993         int old;
5994
5995         if (!rtnl_trylock())
5996                 return restart_syscall();
5997
5998         net = (struct net *)table->extra2;
5999         old = *p;
6000         *p = newf;
6001
6002         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6003                 rtnl_unlock();
6004                 return 0;
6005         }
6006
6007         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6008                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6009                 addrconf_disable_change(net, newf);
6010         } else if ((!newf) ^ (!old))
6011                 dev_disable_change((struct inet6_dev *)table->extra1);
6012
6013         rtnl_unlock();
6014         return 0;
6015 }
6016
6017 static
6018 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6019                             void __user *buffer, size_t *lenp, loff_t *ppos)
6020 {
6021         int *valp = ctl->data;
6022         int val = *valp;
6023         loff_t pos = *ppos;
6024         struct ctl_table lctl;
6025         int ret;
6026
6027         /*
6028          * ctl->data points to idev->cnf.disable_ipv6, we should
6029          * not modify it until we get the rtnl lock.
6030          */
6031         lctl = *ctl;
6032         lctl.data = &val;
6033
6034         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6035
6036         if (write)
6037                 ret = addrconf_disable_ipv6(ctl, valp, val);
6038         if (ret)
6039                 *ppos = pos;
6040         return ret;
6041 }
6042
6043 static
6044 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6045                               void __user *buffer, size_t *lenp, loff_t *ppos)
6046 {
6047         int *valp = ctl->data;
6048         int ret;
6049         int old, new;
6050
6051         old = *valp;
6052         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6053         new = *valp;
6054
6055         if (write && old != new) {
6056                 struct net *net = ctl->extra2;
6057
6058                 if (!rtnl_trylock())
6059                         return restart_syscall();
6060
6061                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6062                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6063                                                      NETCONFA_PROXY_NEIGH,
6064                                                      NETCONFA_IFINDEX_DEFAULT,
6065                                                      net->ipv6.devconf_dflt);
6066                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6067                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6068                                                      NETCONFA_PROXY_NEIGH,
6069                                                      NETCONFA_IFINDEX_ALL,
6070                                                      net->ipv6.devconf_all);
6071                 else {
6072                         struct inet6_dev *idev = ctl->extra1;
6073
6074                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6075                                                      NETCONFA_PROXY_NEIGH,
6076                                                      idev->dev->ifindex,
6077                                                      &idev->cnf);
6078                 }
6079                 rtnl_unlock();
6080         }
6081
6082         return ret;
6083 }
6084
6085 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6086                                          void __user *buffer, size_t *lenp,
6087                                          loff_t *ppos)
6088 {
6089         int ret = 0;
6090         u32 new_val;
6091         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6092         struct net *net = (struct net *)ctl->extra2;
6093         struct ctl_table tmp = {
6094                 .data = &new_val,
6095                 .maxlen = sizeof(new_val),
6096                 .mode = ctl->mode,
6097         };
6098
6099         if (!rtnl_trylock())
6100                 return restart_syscall();
6101
6102         new_val = *((u32 *)ctl->data);
6103
6104         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6105         if (ret != 0)
6106                 goto out;
6107
6108         if (write) {
6109                 if (check_addr_gen_mode(new_val) < 0) {
6110                         ret = -EINVAL;
6111                         goto out;
6112                 }
6113
6114                 if (idev) {
6115                         if (check_stable_privacy(idev, net, new_val) < 0) {
6116                                 ret = -EINVAL;
6117                                 goto out;
6118                         }
6119
6120                         if (idev->cnf.addr_gen_mode != new_val) {
6121                                 idev->cnf.addr_gen_mode = new_val;
6122                                 addrconf_dev_config(idev->dev);
6123                         }
6124                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6125                         struct net_device *dev;
6126
6127                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6128                         for_each_netdev(net, dev) {
6129                                 idev = __in6_dev_get(dev);
6130                                 if (idev &&
6131                                     idev->cnf.addr_gen_mode != new_val) {
6132                                         idev->cnf.addr_gen_mode = new_val;
6133                                         addrconf_dev_config(idev->dev);
6134                                 }
6135                         }
6136                 }
6137
6138                 *((u32 *)ctl->data) = new_val;
6139         }
6140
6141 out:
6142         rtnl_unlock();
6143
6144         return ret;
6145 }
6146
6147 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6148                                          void __user *buffer, size_t *lenp,
6149                                          loff_t *ppos)
6150 {
6151         int err;
6152         struct in6_addr addr;
6153         char str[IPV6_MAX_STRLEN];
6154         struct ctl_table lctl = *ctl;
6155         struct net *net = ctl->extra2;
6156         struct ipv6_stable_secret *secret = ctl->data;
6157
6158         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6159                 return -EIO;
6160
6161         lctl.maxlen = IPV6_MAX_STRLEN;
6162         lctl.data = str;
6163
6164         if (!rtnl_trylock())
6165                 return restart_syscall();
6166
6167         if (!write && !secret->initialized) {
6168                 err = -EIO;
6169                 goto out;
6170         }
6171
6172         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6173         if (err >= sizeof(str)) {
6174                 err = -EIO;
6175                 goto out;
6176         }
6177
6178         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6179         if (err || !write)
6180                 goto out;
6181
6182         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6183                 err = -EIO;
6184                 goto out;
6185         }
6186
6187         secret->initialized = true;
6188         secret->secret = addr;
6189
6190         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6191                 struct net_device *dev;
6192
6193                 for_each_netdev(net, dev) {
6194                         struct inet6_dev *idev = __in6_dev_get(dev);
6195
6196                         if (idev) {
6197                                 idev->cnf.addr_gen_mode =
6198                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6199                         }
6200                 }
6201         } else {
6202                 struct inet6_dev *idev = ctl->extra1;
6203
6204                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6205         }
6206
6207 out:
6208         rtnl_unlock();
6209
6210         return err;
6211 }
6212
6213 static
6214 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6215                                                 int write,
6216                                                 void __user *buffer,
6217                                                 size_t *lenp,
6218                                                 loff_t *ppos)
6219 {
6220         int *valp = ctl->data;
6221         int val = *valp;
6222         loff_t pos = *ppos;
6223         struct ctl_table lctl;
6224         int ret;
6225
6226         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6227          * we should not modify it until we get the rtnl lock.
6228          */
6229         lctl = *ctl;
6230         lctl.data = &val;
6231
6232         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6233
6234         if (write)
6235                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6236         if (ret)
6237                 *ppos = pos;
6238         return ret;
6239 }
6240
6241 static
6242 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6243 {
6244         if (rt) {
6245                 if (action)
6246                         rt->dst.flags |= DST_NOPOLICY;
6247                 else
6248                         rt->dst.flags &= ~DST_NOPOLICY;
6249         }
6250 }
6251
6252 static
6253 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6254 {
6255         struct inet6_ifaddr *ifa;
6256
6257         read_lock_bh(&idev->lock);
6258         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6259                 spin_lock(&ifa->lock);
6260                 if (ifa->rt) {
6261                         struct fib6_info *rt = ifa->rt;
6262                         int cpu;
6263
6264                         rcu_read_lock();
6265                         ifa->rt->dst_nopolicy = val ? true : false;
6266                         if (rt->rt6i_pcpu) {
6267                                 for_each_possible_cpu(cpu) {
6268                                         struct rt6_info **rtp;
6269
6270                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6271                                         addrconf_set_nopolicy(*rtp, val);
6272                                 }
6273                         }
6274                         rcu_read_unlock();
6275                 }
6276                 spin_unlock(&ifa->lock);
6277         }
6278         read_unlock_bh(&idev->lock);
6279 }
6280
6281 static
6282 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6283 {
6284         struct inet6_dev *idev;
6285         struct net *net;
6286
6287         if (!rtnl_trylock())
6288                 return restart_syscall();
6289
6290         *valp = val;
6291
6292         net = (struct net *)ctl->extra2;
6293         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6294                 rtnl_unlock();
6295                 return 0;
6296         }
6297
6298         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6299                 struct net_device *dev;
6300
6301                 for_each_netdev(net, dev) {
6302                         idev = __in6_dev_get(dev);
6303                         if (idev)
6304                                 addrconf_disable_policy_idev(idev, val);
6305                 }
6306         } else {
6307                 idev = (struct inet6_dev *)ctl->extra1;
6308                 addrconf_disable_policy_idev(idev, val);
6309         }
6310
6311         rtnl_unlock();
6312         return 0;
6313 }
6314
6315 static
6316 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6317                                    void __user *buffer, size_t *lenp,
6318                                    loff_t *ppos)
6319 {
6320         int *valp = ctl->data;
6321         int val = *valp;
6322         loff_t pos = *ppos;
6323         struct ctl_table lctl;
6324         int ret;
6325
6326         lctl = *ctl;
6327         lctl.data = &val;
6328         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6329
6330         if (write && (*valp != val))
6331                 ret = addrconf_disable_policy(ctl, valp, val);
6332
6333         if (ret)
6334                 *ppos = pos;
6335
6336         return ret;
6337 }
6338
6339 static int minus_one = -1;
6340 static const int zero = 0;
6341 static const int one = 1;
6342 static const int two_five_five = 255;
6343
6344 static const struct ctl_table addrconf_sysctl[] = {
6345         {
6346                 .procname       = "forwarding",
6347                 .data           = &ipv6_devconf.forwarding,
6348                 .maxlen         = sizeof(int),
6349                 .mode           = 0644,
6350                 .proc_handler   = addrconf_sysctl_forward,
6351         },
6352         {
6353                 .procname       = "hop_limit",
6354                 .data           = &ipv6_devconf.hop_limit,
6355                 .maxlen         = sizeof(int),
6356                 .mode           = 0644,
6357                 .proc_handler   = proc_dointvec_minmax,
6358                 .extra1         = (void *)&one,
6359                 .extra2         = (void *)&two_five_five,
6360         },
6361         {
6362                 .procname       = "mtu",
6363                 .data           = &ipv6_devconf.mtu6,
6364                 .maxlen         = sizeof(int),
6365                 .mode           = 0644,
6366                 .proc_handler   = addrconf_sysctl_mtu,
6367         },
6368         {
6369                 .procname       = "accept_ra",
6370                 .data           = &ipv6_devconf.accept_ra,
6371                 .maxlen         = sizeof(int),
6372                 .mode           = 0644,
6373                 .proc_handler   = proc_dointvec,
6374         },
6375         {
6376                 .procname       = "accept_redirects",
6377                 .data           = &ipv6_devconf.accept_redirects,
6378                 .maxlen         = sizeof(int),
6379                 .mode           = 0644,
6380                 .proc_handler   = proc_dointvec,
6381         },
6382         {
6383                 .procname       = "autoconf",
6384                 .data           = &ipv6_devconf.autoconf,
6385                 .maxlen         = sizeof(int),
6386                 .mode           = 0644,
6387                 .proc_handler   = proc_dointvec,
6388         },
6389         {
6390                 .procname       = "dad_transmits",
6391                 .data           = &ipv6_devconf.dad_transmits,
6392                 .maxlen         = sizeof(int),
6393                 .mode           = 0644,
6394                 .proc_handler   = proc_dointvec,
6395         },
6396         {
6397                 .procname       = "router_solicitations",
6398                 .data           = &ipv6_devconf.rtr_solicits,
6399                 .maxlen         = sizeof(int),
6400                 .mode           = 0644,
6401                 .proc_handler   = proc_dointvec_minmax,
6402                 .extra1         = &minus_one,
6403         },
6404         {
6405                 .procname       = "router_solicitation_interval",
6406                 .data           = &ipv6_devconf.rtr_solicit_interval,
6407                 .maxlen         = sizeof(int),
6408                 .mode           = 0644,
6409                 .proc_handler   = proc_dointvec_jiffies,
6410         },
6411         {
6412                 .procname       = "router_solicitation_max_interval",
6413                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6414                 .maxlen         = sizeof(int),
6415                 .mode           = 0644,
6416                 .proc_handler   = proc_dointvec_jiffies,
6417         },
6418         {
6419                 .procname       = "router_solicitation_delay",
6420                 .data           = &ipv6_devconf.rtr_solicit_delay,
6421                 .maxlen         = sizeof(int),
6422                 .mode           = 0644,
6423                 .proc_handler   = proc_dointvec_jiffies,
6424         },
6425         {
6426                 .procname       = "force_mld_version",
6427                 .data           = &ipv6_devconf.force_mld_version,
6428                 .maxlen         = sizeof(int),
6429                 .mode           = 0644,
6430                 .proc_handler   = proc_dointvec,
6431         },
6432         {
6433                 .procname       = "mldv1_unsolicited_report_interval",
6434                 .data           =
6435                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6436                 .maxlen         = sizeof(int),
6437                 .mode           = 0644,
6438                 .proc_handler   = proc_dointvec_ms_jiffies,
6439         },
6440         {
6441                 .procname       = "mldv2_unsolicited_report_interval",
6442                 .data           =
6443                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6444                 .maxlen         = sizeof(int),
6445                 .mode           = 0644,
6446                 .proc_handler   = proc_dointvec_ms_jiffies,
6447         },
6448         {
6449                 .procname       = "use_tempaddr",
6450                 .data           = &ipv6_devconf.use_tempaddr,
6451                 .maxlen         = sizeof(int),
6452                 .mode           = 0644,
6453                 .proc_handler   = proc_dointvec,
6454         },
6455         {
6456                 .procname       = "temp_valid_lft",
6457                 .data           = &ipv6_devconf.temp_valid_lft,
6458                 .maxlen         = sizeof(int),
6459                 .mode           = 0644,
6460                 .proc_handler   = proc_dointvec,
6461         },
6462         {
6463                 .procname       = "temp_prefered_lft",
6464                 .data           = &ipv6_devconf.temp_prefered_lft,
6465                 .maxlen         = sizeof(int),
6466                 .mode           = 0644,
6467                 .proc_handler   = proc_dointvec,
6468         },
6469         {
6470                 .procname       = "regen_max_retry",
6471                 .data           = &ipv6_devconf.regen_max_retry,
6472                 .maxlen         = sizeof(int),
6473                 .mode           = 0644,
6474                 .proc_handler   = proc_dointvec,
6475         },
6476         {
6477                 .procname       = "max_desync_factor",
6478                 .data           = &ipv6_devconf.max_desync_factor,
6479                 .maxlen         = sizeof(int),
6480                 .mode           = 0644,
6481                 .proc_handler   = proc_dointvec,
6482         },
6483         {
6484                 .procname       = "max_addresses",
6485                 .data           = &ipv6_devconf.max_addresses,
6486                 .maxlen         = sizeof(int),
6487                 .mode           = 0644,
6488                 .proc_handler   = proc_dointvec,
6489         },
6490         {
6491                 .procname       = "accept_ra_defrtr",
6492                 .data           = &ipv6_devconf.accept_ra_defrtr,
6493                 .maxlen         = sizeof(int),
6494                 .mode           = 0644,
6495                 .proc_handler   = proc_dointvec,
6496         },
6497         {
6498                 .procname       = "accept_ra_min_hop_limit",
6499                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6500                 .maxlen         = sizeof(int),
6501                 .mode           = 0644,
6502                 .proc_handler   = proc_dointvec,
6503         },
6504         {
6505                 .procname       = "accept_ra_pinfo",
6506                 .data           = &ipv6_devconf.accept_ra_pinfo,
6507                 .maxlen         = sizeof(int),
6508                 .mode           = 0644,
6509                 .proc_handler   = proc_dointvec,
6510         },
6511 #ifdef CONFIG_IPV6_ROUTER_PREF
6512         {
6513                 .procname       = "accept_ra_rtr_pref",
6514                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6515                 .maxlen         = sizeof(int),
6516                 .mode           = 0644,
6517                 .proc_handler   = proc_dointvec,
6518         },
6519         {
6520                 .procname       = "router_probe_interval",
6521                 .data           = &ipv6_devconf.rtr_probe_interval,
6522                 .maxlen         = sizeof(int),
6523                 .mode           = 0644,
6524                 .proc_handler   = proc_dointvec_jiffies,
6525         },
6526 #ifdef CONFIG_IPV6_ROUTE_INFO
6527         {
6528                 .procname       = "accept_ra_rt_info_min_plen",
6529                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6530                 .maxlen         = sizeof(int),
6531                 .mode           = 0644,
6532                 .proc_handler   = proc_dointvec,
6533         },
6534         {
6535                 .procname       = "accept_ra_rt_info_max_plen",
6536                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6537                 .maxlen         = sizeof(int),
6538                 .mode           = 0644,
6539                 .proc_handler   = proc_dointvec,
6540         },
6541 #endif
6542 #endif
6543         {
6544                 .procname       = "proxy_ndp",
6545                 .data           = &ipv6_devconf.proxy_ndp,
6546                 .maxlen         = sizeof(int),
6547                 .mode           = 0644,
6548                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6549         },
6550         {
6551                 .procname       = "accept_source_route",
6552                 .data           = &ipv6_devconf.accept_source_route,
6553                 .maxlen         = sizeof(int),
6554                 .mode           = 0644,
6555                 .proc_handler   = proc_dointvec,
6556         },
6557 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6558         {
6559                 .procname       = "optimistic_dad",
6560                 .data           = &ipv6_devconf.optimistic_dad,
6561                 .maxlen         = sizeof(int),
6562                 .mode           = 0644,
6563                 .proc_handler   = proc_dointvec,
6564         },
6565         {
6566                 .procname       = "use_optimistic",
6567                 .data           = &ipv6_devconf.use_optimistic,
6568                 .maxlen         = sizeof(int),
6569                 .mode           = 0644,
6570                 .proc_handler   = proc_dointvec,
6571         },
6572 #endif
6573 #ifdef CONFIG_IPV6_MROUTE
6574         {
6575                 .procname       = "mc_forwarding",
6576                 .data           = &ipv6_devconf.mc_forwarding,
6577                 .maxlen         = sizeof(int),
6578                 .mode           = 0444,
6579                 .proc_handler   = proc_dointvec,
6580         },
6581 #endif
6582         {
6583                 .procname       = "disable_ipv6",
6584                 .data           = &ipv6_devconf.disable_ipv6,
6585                 .maxlen         = sizeof(int),
6586                 .mode           = 0644,
6587                 .proc_handler   = addrconf_sysctl_disable,
6588         },
6589         {
6590                 .procname       = "accept_dad",
6591                 .data           = &ipv6_devconf.accept_dad,
6592                 .maxlen         = sizeof(int),
6593                 .mode           = 0644,
6594                 .proc_handler   = proc_dointvec,
6595         },
6596         {
6597                 .procname       = "force_tllao",
6598                 .data           = &ipv6_devconf.force_tllao,
6599                 .maxlen         = sizeof(int),
6600                 .mode           = 0644,
6601                 .proc_handler   = proc_dointvec
6602         },
6603         {
6604                 .procname       = "ndisc_notify",
6605                 .data           = &ipv6_devconf.ndisc_notify,
6606                 .maxlen         = sizeof(int),
6607                 .mode           = 0644,
6608                 .proc_handler   = proc_dointvec
6609         },
6610         {
6611                 .procname       = "suppress_frag_ndisc",
6612                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6613                 .maxlen         = sizeof(int),
6614                 .mode           = 0644,
6615                 .proc_handler   = proc_dointvec
6616         },
6617         {
6618                 .procname       = "accept_ra_from_local",
6619                 .data           = &ipv6_devconf.accept_ra_from_local,
6620                 .maxlen         = sizeof(int),
6621                 .mode           = 0644,
6622                 .proc_handler   = proc_dointvec,
6623         },
6624         {
6625                 .procname       = "accept_ra_mtu",
6626                 .data           = &ipv6_devconf.accept_ra_mtu,
6627                 .maxlen         = sizeof(int),
6628                 .mode           = 0644,
6629                 .proc_handler   = proc_dointvec,
6630         },
6631         {
6632                 .procname       = "stable_secret",
6633                 .data           = &ipv6_devconf.stable_secret,
6634                 .maxlen         = IPV6_MAX_STRLEN,
6635                 .mode           = 0600,
6636                 .proc_handler   = addrconf_sysctl_stable_secret,
6637         },
6638         {
6639                 .procname       = "use_oif_addrs_only",
6640                 .data           = &ipv6_devconf.use_oif_addrs_only,
6641                 .maxlen         = sizeof(int),
6642                 .mode           = 0644,
6643                 .proc_handler   = proc_dointvec,
6644         },
6645         {
6646                 .procname       = "ignore_routes_with_linkdown",
6647                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6648                 .maxlen         = sizeof(int),
6649                 .mode           = 0644,
6650                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6651         },
6652         {
6653                 .procname       = "drop_unicast_in_l2_multicast",
6654                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6655                 .maxlen         = sizeof(int),
6656                 .mode           = 0644,
6657                 .proc_handler   = proc_dointvec,
6658         },
6659         {
6660                 .procname       = "drop_unsolicited_na",
6661                 .data           = &ipv6_devconf.drop_unsolicited_na,
6662                 .maxlen         = sizeof(int),
6663                 .mode           = 0644,
6664                 .proc_handler   = proc_dointvec,
6665         },
6666         {
6667                 .procname       = "keep_addr_on_down",
6668                 .data           = &ipv6_devconf.keep_addr_on_down,
6669                 .maxlen         = sizeof(int),
6670                 .mode           = 0644,
6671                 .proc_handler   = proc_dointvec,
6672
6673         },
6674         {
6675                 .procname       = "seg6_enabled",
6676                 .data           = &ipv6_devconf.seg6_enabled,
6677                 .maxlen         = sizeof(int),
6678                 .mode           = 0644,
6679                 .proc_handler   = proc_dointvec,
6680         },
6681 #ifdef CONFIG_IPV6_SEG6_HMAC
6682         {
6683                 .procname       = "seg6_require_hmac",
6684                 .data           = &ipv6_devconf.seg6_require_hmac,
6685                 .maxlen         = sizeof(int),
6686                 .mode           = 0644,
6687                 .proc_handler   = proc_dointvec,
6688         },
6689 #endif
6690         {
6691                 .procname       = "enhanced_dad",
6692                 .data           = &ipv6_devconf.enhanced_dad,
6693                 .maxlen         = sizeof(int),
6694                 .mode           = 0644,
6695                 .proc_handler   = proc_dointvec,
6696         },
6697         {
6698                 .procname               = "addr_gen_mode",
6699                 .data                   = &ipv6_devconf.addr_gen_mode,
6700                 .maxlen                 = sizeof(int),
6701                 .mode                   = 0644,
6702                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6703         },
6704         {
6705                 .procname       = "disable_policy",
6706                 .data           = &ipv6_devconf.disable_policy,
6707                 .maxlen         = sizeof(int),
6708                 .mode           = 0644,
6709                 .proc_handler   = addrconf_sysctl_disable_policy,
6710         },
6711         {
6712                 .procname       = "ndisc_tclass",
6713                 .data           = &ipv6_devconf.ndisc_tclass,
6714                 .maxlen         = sizeof(int),
6715                 .mode           = 0644,
6716                 .proc_handler   = proc_dointvec_minmax,
6717                 .extra1         = (void *)&zero,
6718                 .extra2         = (void *)&two_five_five,
6719         },
6720         {
6721                 /* sentinel */
6722         }
6723 };
6724
6725 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6726                 struct inet6_dev *idev, struct ipv6_devconf *p)
6727 {
6728         int i, ifindex;
6729         struct ctl_table *table;
6730         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6731
6732         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6733         if (!table)
6734                 goto out;
6735
6736         for (i = 0; table[i].data; i++) {
6737                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6738                 /* If one of these is already set, then it is not safe to
6739                  * overwrite either of them: this makes proc_dointvec_minmax
6740                  * usable.
6741                  */
6742                 if (!table[i].extra1 && !table[i].extra2) {
6743                         table[i].extra1 = idev; /* embedded; no ref */
6744                         table[i].extra2 = net;
6745                 }
6746         }
6747
6748         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6749
6750         p->sysctl_header = register_net_sysctl(net, path, table);
6751         if (!p->sysctl_header)
6752                 goto free;
6753
6754         if (!strcmp(dev_name, "all"))
6755                 ifindex = NETCONFA_IFINDEX_ALL;
6756         else if (!strcmp(dev_name, "default"))
6757                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6758         else
6759                 ifindex = idev->dev->ifindex;
6760         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6761                                      ifindex, p);
6762         return 0;
6763
6764 free:
6765         kfree(table);
6766 out:
6767         return -ENOBUFS;
6768 }
6769
6770 static void __addrconf_sysctl_unregister(struct net *net,
6771                                          struct ipv6_devconf *p, int ifindex)
6772 {
6773         struct ctl_table *table;
6774
6775         if (!p->sysctl_header)
6776                 return;
6777
6778         table = p->sysctl_header->ctl_table_arg;
6779         unregister_net_sysctl_table(p->sysctl_header);
6780         p->sysctl_header = NULL;
6781         kfree(table);
6782
6783         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6784 }
6785
6786 static int addrconf_sysctl_register(struct inet6_dev *idev)
6787 {
6788         int err;
6789
6790         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6791                 return -EINVAL;
6792
6793         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6794                                     &ndisc_ifinfo_sysctl_change);
6795         if (err)
6796                 return err;
6797         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6798                                          idev, &idev->cnf);
6799         if (err)
6800                 neigh_sysctl_unregister(idev->nd_parms);
6801
6802         return err;
6803 }
6804
6805 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6806 {
6807         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6808                                      idev->dev->ifindex);
6809         neigh_sysctl_unregister(idev->nd_parms);
6810 }
6811
6812
6813 #endif
6814
6815 static int __net_init addrconf_init_net(struct net *net)
6816 {
6817         int err = -ENOMEM;
6818         struct ipv6_devconf *all, *dflt;
6819
6820         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6821         if (!all)
6822                 goto err_alloc_all;
6823
6824         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6825         if (!dflt)
6826                 goto err_alloc_dflt;
6827
6828         /* these will be inherited by all namespaces */
6829         dflt->autoconf = ipv6_defaults.autoconf;
6830         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6831
6832         dflt->stable_secret.initialized = false;
6833         all->stable_secret.initialized = false;
6834
6835         net->ipv6.devconf_all = all;
6836         net->ipv6.devconf_dflt = dflt;
6837
6838 #ifdef CONFIG_SYSCTL
6839         err = __addrconf_sysctl_register(net, "all", NULL, all);
6840         if (err < 0)
6841                 goto err_reg_all;
6842
6843         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6844         if (err < 0)
6845                 goto err_reg_dflt;
6846 #endif
6847         return 0;
6848
6849 #ifdef CONFIG_SYSCTL
6850 err_reg_dflt:
6851         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6852 err_reg_all:
6853         kfree(dflt);
6854 #endif
6855 err_alloc_dflt:
6856         kfree(all);
6857 err_alloc_all:
6858         return err;
6859 }
6860
6861 static void __net_exit addrconf_exit_net(struct net *net)
6862 {
6863 #ifdef CONFIG_SYSCTL
6864         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6865                                      NETCONFA_IFINDEX_DEFAULT);
6866         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6867                                      NETCONFA_IFINDEX_ALL);
6868 #endif
6869         kfree(net->ipv6.devconf_dflt);
6870         kfree(net->ipv6.devconf_all);
6871 }
6872
6873 static struct pernet_operations addrconf_ops = {
6874         .init = addrconf_init_net,
6875         .exit = addrconf_exit_net,
6876 };
6877
6878 static struct rtnl_af_ops inet6_ops __read_mostly = {
6879         .family           = AF_INET6,
6880         .fill_link_af     = inet6_fill_link_af,
6881         .get_link_af_size = inet6_get_link_af_size,
6882         .validate_link_af = inet6_validate_link_af,
6883         .set_link_af      = inet6_set_link_af,
6884 };
6885
6886 /*
6887  *      Init / cleanup code
6888  */
6889
6890 int __init addrconf_init(void)
6891 {
6892         struct inet6_dev *idev;
6893         int i, err;
6894
6895         err = ipv6_addr_label_init();
6896         if (err < 0) {
6897                 pr_crit("%s: cannot initialize default policy table: %d\n",
6898                         __func__, err);
6899                 goto out;
6900         }
6901
6902         err = register_pernet_subsys(&addrconf_ops);
6903         if (err < 0)
6904                 goto out_addrlabel;
6905
6906         addrconf_wq = create_workqueue("ipv6_addrconf");
6907         if (!addrconf_wq) {
6908                 err = -ENOMEM;
6909                 goto out_nowq;
6910         }
6911
6912         /* The addrconf netdev notifier requires that loopback_dev
6913          * has it's ipv6 private information allocated and setup
6914          * before it can bring up and give link-local addresses
6915          * to other devices which are up.
6916          *
6917          * Unfortunately, loopback_dev is not necessarily the first
6918          * entry in the global dev_base list of net devices.  In fact,
6919          * it is likely to be the very last entry on that list.
6920          * So this causes the notifier registry below to try and
6921          * give link-local addresses to all devices besides loopback_dev
6922          * first, then loopback_dev, which cases all the non-loopback_dev
6923          * devices to fail to get a link-local address.
6924          *
6925          * So, as a temporary fix, allocate the ipv6 structure for
6926          * loopback_dev first by hand.
6927          * Longer term, all of the dependencies ipv6 has upon the loopback
6928          * device and it being up should be removed.
6929          */
6930         rtnl_lock();
6931         idev = ipv6_add_dev(init_net.loopback_dev);
6932         rtnl_unlock();
6933         if (IS_ERR(idev)) {
6934                 err = PTR_ERR(idev);
6935                 goto errlo;
6936         }
6937
6938         ip6_route_init_special_entries();
6939
6940         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6941                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6942
6943         register_netdevice_notifier(&ipv6_dev_notf);
6944
6945         addrconf_verify();
6946
6947         rtnl_af_register(&inet6_ops);
6948
6949         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6950                                    NULL, inet6_dump_ifinfo, 0);
6951         if (err < 0)
6952                 goto errout;
6953
6954         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6955                                    inet6_rtm_newaddr, NULL, 0);
6956         if (err < 0)
6957                 goto errout;
6958         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6959                                    inet6_rtm_deladdr, NULL, 0);
6960         if (err < 0)
6961                 goto errout;
6962         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6963                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
6964                                    RTNL_FLAG_DOIT_UNLOCKED);
6965         if (err < 0)
6966                 goto errout;
6967         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6968                                    NULL, inet6_dump_ifmcaddr, 0);
6969         if (err < 0)
6970                 goto errout;
6971         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6972                                    NULL, inet6_dump_ifacaddr, 0);
6973         if (err < 0)
6974                 goto errout;
6975         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6976                                    inet6_netconf_get_devconf,
6977                                    inet6_netconf_dump_devconf,
6978                                    RTNL_FLAG_DOIT_UNLOCKED);
6979         if (err < 0)
6980                 goto errout;
6981         err = ipv6_addr_label_rtnl_register();
6982         if (err < 0)
6983                 goto errout;
6984
6985         return 0;
6986 errout:
6987         rtnl_unregister_all(PF_INET6);
6988         rtnl_af_unregister(&inet6_ops);
6989         unregister_netdevice_notifier(&ipv6_dev_notf);
6990 errlo:
6991         destroy_workqueue(addrconf_wq);
6992 out_nowq:
6993         unregister_pernet_subsys(&addrconf_ops);
6994 out_addrlabel:
6995         ipv6_addr_label_cleanup();
6996 out:
6997         return err;
6998 }
6999
7000 void addrconf_cleanup(void)
7001 {
7002         struct net_device *dev;
7003         int i;
7004
7005         unregister_netdevice_notifier(&ipv6_dev_notf);
7006         unregister_pernet_subsys(&addrconf_ops);
7007         ipv6_addr_label_cleanup();
7008
7009         rtnl_af_unregister(&inet6_ops);
7010
7011         rtnl_lock();
7012
7013         /* clean dev list */
7014         for_each_netdev(&init_net, dev) {
7015                 if (__in6_dev_get(dev) == NULL)
7016                         continue;
7017                 addrconf_ifdown(dev, 1);
7018         }
7019         addrconf_ifdown(init_net.loopback_dev, 2);
7020
7021         /*
7022          *      Check hash table.
7023          */
7024         spin_lock_bh(&addrconf_hash_lock);
7025         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7026                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7027         spin_unlock_bh(&addrconf_hash_lock);
7028         cancel_delayed_work(&addr_chk_work);
7029         rtnl_unlock();
7030
7031         destroy_workqueue(addrconf_wq);
7032 }