net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92
  93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  94
  95 #define IPV6_MAX_STRLEN \
  96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  97
  98 static inline u32 cstamp_delta(unsigned long cstamp)
  99 {
 100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 101 }
 102
 103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 104 {
 105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 107         do_div(tmp, 1000000);
 108         return (s32)tmp;
 109 }
 110
 111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 112 {
 113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 119                 do_div(tmp, 1000000);
 120         }
 121         return (s32)tmp;
 122 }
 123
 124 #ifdef CONFIG_SYSCTL
 125 static int addrconf_sysctl_register(struct inet6_dev *idev);
 126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 127 #else
 128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 129 {
 130         return 0;
 131 }
 132
 133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 134 {
 135 }
 136 #endif
 137
 138 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 139
 140 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 141 static int ipv6_count_addresses(const struct inet6_dev *idev);
 142 static int ipv6_generate_stable_address(struct in6_addr *addr,
 143                                         u8 dad_count,
 144                                         const struct inet6_dev *idev);
 145
 146 #define IN6_ADDR_HSIZE_SHIFT    8
 147 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 148 /*
 149  *      Configured unicast address hash table
 150  */
 151 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 152 static DEFINE_SPINLOCK(addrconf_hash_lock);
 153
 154 static void addrconf_verify(void);
 155 static void addrconf_verify_rtnl(void);
 156 static void addrconf_verify_work(struct work_struct *);
 157
 158 static struct workqueue_struct *addrconf_wq;
 159 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 160
 161 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 162 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 163
 164 static void addrconf_type_change(struct net_device *dev,
 165                                  unsigned long event);
 166 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 167
 168 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 169                                                   int plen,
 170                                                   const struct net_device *dev,
 171                                                   u32 flags, u32 noflags,
 172                                                   bool no_gw);
 173
 174 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 175 static void addrconf_dad_work(struct work_struct *w);
 176 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 177                                    bool send_na);
 178 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 179 static void addrconf_rs_timer(struct timer_list *t);
 180 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 181 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182
 183 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 184                                 struct prefix_info *pinfo);
 185
 186 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 187         .forwarding             = 0,
 188         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 189         .mtu6                   = IPV6_MIN_MTU,
 190         .accept_ra              = 1,
 191         .accept_redirects       = 1,
 192         .autoconf               = 1,
 193         .force_mld_version      = 0,
 194         .mldv1_unsolicited_report_interval = 10 * HZ,
 195         .mldv2_unsolicited_report_interval = HZ,
 196         .dad_transmits          = 1,
 197         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 198         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 199         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 201         .use_tempaddr           = 0,
 202         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 203         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 204         .regen_max_retry        = REGEN_MAX_RETRY,
 205         .max_desync_factor      = MAX_DESYNC_FACTOR,
 206         .max_addresses          = IPV6_MAX_ADDRESSES,
 207         .accept_ra_defrtr       = 1,
 208         .accept_ra_from_local   = 0,
 209         .accept_ra_min_hop_limit= 1,
 210         .accept_ra_min_lft      = 0,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_min_plen = 0,
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 0,
 224         .suppress_frag_ndisc    = 1,
 225         .accept_ra_mtu          = 1,
 226         .stable_secret          = {
 227                 .initialized = false,
 228         },
 229         .use_oif_addrs_only     = 0,
 230         .ignore_routes_with_linkdown = 0,
 231         .keep_addr_on_down      = 0,
 232         .seg6_enabled           = 0,
 233 #ifdef CONFIG_IPV6_SEG6_HMAC
 234         .seg6_require_hmac      = 0,
 235 #endif
 236         .enhanced_dad           = 1,
 237         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 238         .disable_policy         = 0,
 239         .rpl_seg_enabled        = 0,
 240 };
 241
 242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 243         .forwarding             = 0,
 244         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 245         .mtu6                   = IPV6_MIN_MTU,
 246         .accept_ra              = 1,
 247         .accept_redirects       = 1,
 248         .autoconf               = 1,
 249         .force_mld_version      = 0,
 250         .mldv1_unsolicited_report_interval = 10 * HZ,
 251         .mldv2_unsolicited_report_interval = HZ,
 252         .dad_transmits          = 1,
 253         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 254         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 255         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 256         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 257         .use_tempaddr           = 0,
 258         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 259         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 260         .regen_max_retry        = REGEN_MAX_RETRY,
 261         .max_desync_factor      = MAX_DESYNC_FACTOR,
 262         .max_addresses          = IPV6_MAX_ADDRESSES,
 263         .accept_ra_defrtr       = 1,
 264         .accept_ra_from_local   = 0,
 265         .accept_ra_min_hop_limit= 1,
 266         .accept_ra_min_lft      = 0,
 267         .accept_ra_pinfo        = 1,
 268 #ifdef CONFIG_IPV6_ROUTER_PREF
 269         .accept_ra_rtr_pref     = 1,
 270         .rtr_probe_interval     = 60 * HZ,
 271 #ifdef CONFIG_IPV6_ROUTE_INFO
 272         .accept_ra_rt_info_min_plen = 0,
 273         .accept_ra_rt_info_max_plen = 0,
 274 #endif
 275 #endif
 276         .proxy_ndp              = 0,
 277         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 278         .disable_ipv6           = 0,
 279         .accept_dad             = 1,
 280         .suppress_frag_ndisc    = 1,
 281         .accept_ra_mtu          = 1,
 282         .stable_secret          = {
 283                 .initialized = false,
 284         },
 285         .use_oif_addrs_only     = 0,
 286         .ignore_routes_with_linkdown = 0,
 287         .keep_addr_on_down      = 0,
 288         .seg6_enabled           = 0,
 289 #ifdef CONFIG_IPV6_SEG6_HMAC
 290         .seg6_require_hmac      = 0,
 291 #endif
 292         .enhanced_dad           = 1,
 293         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 294         .disable_policy         = 0,
 295         .rpl_seg_enabled        = 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 (!mod_timer(&idev->rs_timer, jiffies + when))
 320                 in6_dev_hold(idev);
 321 }
 322
 323 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 324                                    unsigned long delay)
 325 {
 326         in6_ifa_hold(ifp);
 327         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 328                 in6_ifa_put(ifp);
 329 }
 330
 331 static int snmp6_alloc_dev(struct inet6_dev *idev)
 332 {
 333         int i;
 334
 335         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 336         if (!idev->stats.ipv6)
 337                 goto err_ip;
 338
 339         for_each_possible_cpu(i) {
 340                 struct ipstats_mib *addrconf_stats;
 341                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 342                 u64_stats_init(&addrconf_stats->syncp);
 343         }
 344
 345
 346         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 347                                         GFP_KERNEL);
 348         if (!idev->stats.icmpv6dev)
 349                 goto err_icmp;
 350         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 351                                            GFP_KERNEL);
 352         if (!idev->stats.icmpv6msgdev)
 353                 goto err_icmpmsg;
 354
 355         return 0;
 356
 357 err_icmpmsg:
 358         kfree(idev->stats.icmpv6dev);
 359 err_icmp:
 360         free_percpu(idev->stats.ipv6);
 361 err_ip:
 362         return -ENOMEM;
 363 }
 364
 365 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 366 {
 367         struct inet6_dev *ndev;
 368         int err = -ENOMEM;
 369
 370         ASSERT_RTNL();
 371
 372         if (dev->mtu < IPV6_MIN_MTU)
 373                 return ERR_PTR(-EINVAL);
 374
 375         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 376         if (!ndev)
 377                 return ERR_PTR(err);
 378
 379         rwlock_init(&ndev->lock);
 380         ndev->dev = dev;
 381         INIT_LIST_HEAD(&ndev->addr_list);
 382         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 383         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 384
 385         if (ndev->cnf.stable_secret.initialized)
 386                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 387
 388         ndev->cnf.mtu6 = dev->mtu;
 389         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 390         if (!ndev->nd_parms) {
 391                 kfree(ndev);
 392                 return ERR_PTR(err);
 393         }
 394         if (ndev->cnf.forwarding)
 395                 dev_disable_lro(dev);
 396         /* We refer to the device */
 397         dev_hold(dev);
 398
 399         if (snmp6_alloc_dev(ndev) < 0) {
 400                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 401                            __func__);
 402                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 403                 dev_put(dev);
 404                 kfree(ndev);
 405                 return ERR_PTR(err);
 406         }
 407
 408         if (snmp6_register_dev(ndev) < 0) {
 409                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 410                            __func__, dev->name);
 411                 goto err_release;
 412         }
 413
 414         /* One reference from device. */
 415         refcount_set(&ndev->refcnt, 1);
 416
 417         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 418                 ndev->cnf.accept_dad = -1;
 419
 420 #if IS_ENABLED(CONFIG_IPV6_SIT)
 421         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 422                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 423                 ndev->cnf.rtr_solicits = 0;
 424         }
 425 #endif
 426
 427         INIT_LIST_HEAD(&ndev->tempaddr_list);
 428         ndev->desync_factor = U32_MAX;
 429         if ((dev->flags&IFF_LOOPBACK) ||
 430             dev->type == ARPHRD_TUNNEL ||
 431             dev->type == ARPHRD_TUNNEL6 ||
 432             dev->type == ARPHRD_SIT ||
 433             dev->type == ARPHRD_NONE) {
 434                 ndev->cnf.use_tempaddr = -1;
 435         }
 436
 437         ndev->token = in6addr_any;
 438
 439         if (netif_running(dev) && addrconf_link_ready(dev))
 440                 ndev->if_flags |= IF_READY;
 441
 442         ipv6_mc_init_dev(ndev);
 443         ndev->tstamp = jiffies;
 444         err = addrconf_sysctl_register(ndev);
 445         if (err) {
 446                 ipv6_mc_destroy_dev(ndev);
 447                 snmp6_unregister_dev(ndev);
 448                 goto err_release;
 449         }
 450         /* protected by rtnl_lock */
 451         rcu_assign_pointer(dev->ip6_ptr, ndev);
 452
 453         /* Join interface-local all-node multicast group */
 454         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 455
 456         /* Join all-node multicast group */
 457         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 458
 459         /* Join all-router multicast group if forwarding is set */
 460         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 461                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 462
 463         return ndev;
 464
 465 err_release:
 466         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 467         ndev->dead = 1;
 468         in6_dev_finish_destroy(ndev);
 469         return ERR_PTR(err);
 470 }
 471
 472 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 473 {
 474         struct inet6_dev *idev;
 475
 476         ASSERT_RTNL();
 477
 478         idev = __in6_dev_get(dev);
 479         if (!idev) {
 480                 idev = ipv6_add_dev(dev);
 481                 if (IS_ERR(idev))
 482                         return idev;
 483         }
 484
 485         if (dev->flags&IFF_UP)
 486                 ipv6_mc_up(idev);
 487         return idev;
 488 }
 489
 490 static int inet6_netconf_msgsize_devconf(int type)
 491 {
 492         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 493                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 494         bool all = false;
 495
 496         if (type == NETCONFA_ALL)
 497                 all = true;
 498
 499         if (all || type == NETCONFA_FORWARDING)
 500                 size += nla_total_size(4);
 501 #ifdef CONFIG_IPV6_MROUTE
 502         if (all || type == NETCONFA_MC_FORWARDING)
 503                 size += nla_total_size(4);
 504 #endif
 505         if (all || type == NETCONFA_PROXY_NEIGH)
 506                 size += nla_total_size(4);
 507
 508         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 509                 size += nla_total_size(4);
 510
 511         return size;
 512 }
 513
 514 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 515                                       struct ipv6_devconf *devconf, u32 portid,
 516                                       u32 seq, int event, unsigned int flags,
 517                                       int type)
 518 {
 519         struct nlmsghdr  *nlh;
 520         struct netconfmsg *ncm;
 521         bool all = false;
 522
 523         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 524                         flags);
 525         if (!nlh)
 526                 return -EMSGSIZE;
 527
 528         if (type == NETCONFA_ALL)
 529                 all = true;
 530
 531         ncm = nlmsg_data(nlh);
 532         ncm->ncm_family = AF_INET6;
 533
 534         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 535                 goto nla_put_failure;
 536
 537         if (!devconf)
 538                 goto out;
 539
 540         if ((all || type == NETCONFA_FORWARDING) &&
 541             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 542                 goto nla_put_failure;
 543 #ifdef CONFIG_IPV6_MROUTE
 544         if ((all || type == NETCONFA_MC_FORWARDING) &&
 545             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 546                         atomic_read(&devconf->mc_forwarding)) < 0)
 547                 goto nla_put_failure;
 548 #endif
 549         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 550             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 551                 goto nla_put_failure;
 552
 553         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 554             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 555                         devconf->ignore_routes_with_linkdown) < 0)
 556                 goto nla_put_failure;
 557
 558 out:
 559         nlmsg_end(skb, nlh);
 560         return 0;
 561
 562 nla_put_failure:
 563         nlmsg_cancel(skb, nlh);
 564         return -EMSGSIZE;
 565 }
 566
 567 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 568                                   int ifindex, struct ipv6_devconf *devconf)
 569 {
 570         struct sk_buff *skb;
 571         int err = -ENOBUFS;
 572
 573         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 574         if (!skb)
 575                 goto errout;
 576
 577         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 578                                          event, 0, type);
 579         if (err < 0) {
 580                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 581                 WARN_ON(err == -EMSGSIZE);
 582                 kfree_skb(skb);
 583                 goto errout;
 584         }
 585         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 586         return;
 587 errout:
 588         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 589 }
 590
 591 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 592         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 593         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 594         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 595         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 596 };
 597
 598 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 599                                        const struct nlmsghdr *nlh,
 600                                        struct nlattr **tb,
 601                                        struct netlink_ext_ack *extack)
 602 {
 603         int i, err;
 604
 605         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 606                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 607                 return -EINVAL;
 608         }
 609
 610         if (!netlink_strict_get_check(skb))
 611                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 612                                               tb, NETCONFA_MAX,
 613                                               devconf_ipv6_policy, extack);
 614
 615         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 616                                             tb, NETCONFA_MAX,
 617                                             devconf_ipv6_policy, extack);
 618         if (err)
 619                 return err;
 620
 621         for (i = 0; i <= NETCONFA_MAX; i++) {
 622                 if (!tb[i])
 623                         continue;
 624
 625                 switch (i) {
 626                 case NETCONFA_IFINDEX:
 627                         break;
 628                 default:
 629                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 630                         return -EINVAL;
 631                 }
 632         }
 633
 634         return 0;
 635 }
 636
 637 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 638                                      struct nlmsghdr *nlh,
 639                                      struct netlink_ext_ack *extack)
 640 {
 641         struct net *net = sock_net(in_skb->sk);
 642         struct nlattr *tb[NETCONFA_MAX+1];
 643         struct inet6_dev *in6_dev = NULL;
 644         struct net_device *dev = NULL;
 645         struct sk_buff *skb;
 646         struct ipv6_devconf *devconf;
 647         int ifindex;
 648         int err;
 649
 650         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 651         if (err < 0)
 652                 return err;
 653
 654         if (!tb[NETCONFA_IFINDEX])
 655                 return -EINVAL;
 656
 657         err = -EINVAL;
 658         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 659         switch (ifindex) {
 660         case NETCONFA_IFINDEX_ALL:
 661                 devconf = net->ipv6.devconf_all;
 662                 break;
 663         case NETCONFA_IFINDEX_DEFAULT:
 664                 devconf = net->ipv6.devconf_dflt;
 665                 break;
 666         default:
 667                 dev = dev_get_by_index(net, ifindex);
 668                 if (!dev)
 669                         return -EINVAL;
 670                 in6_dev = in6_dev_get(dev);
 671                 if (!in6_dev)
 672                         goto errout;
 673                 devconf = &in6_dev->cnf;
 674                 break;
 675         }
 676
 677         err = -ENOBUFS;
 678         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 679         if (!skb)
 680                 goto errout;
 681
 682         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 683                                          NETLINK_CB(in_skb).portid,
 684                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 685                                          NETCONFA_ALL);
 686         if (err < 0) {
 687                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 688                 WARN_ON(err == -EMSGSIZE);
 689                 kfree_skb(skb);
 690                 goto errout;
 691         }
 692         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 693 errout:
 694         if (in6_dev)
 695                 in6_dev_put(in6_dev);
 696         if (dev)
 697                 dev_put(dev);
 698         return err;
 699 }
 700
 701 /* Combine dev_addr_genid and dev_base_seq to detect changes.
 702  */
 703 static u32 inet6_base_seq(const struct net *net)
 704 {
 705         u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
 706                   net->dev_base_seq;
 707
 708         /* Must not return 0 (see nl_dump_check_consistent()).
 709          * Chose a value far away from 0.
 710          */
 711         if (!res)
 712                 res = 0x80000000;
 713         return res;
 714 }
 715
 716
 717 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 718                                       struct netlink_callback *cb)
 719 {
 720         const struct nlmsghdr *nlh = cb->nlh;
 721         struct net *net = sock_net(skb->sk);
 722         int h, s_h;
 723         int idx, s_idx;
 724         struct net_device *dev;
 725         struct inet6_dev *idev;
 726         struct hlist_head *head;
 727
 728         if (cb->strict_check) {
 729                 struct netlink_ext_ack *extack = cb->extack;
 730                 struct netconfmsg *ncm;
 731
 732                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 733                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 734                         return -EINVAL;
 735                 }
 736
 737                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 738                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 739                         return -EINVAL;
 740                 }
 741         }
 742
 743         s_h = cb->args[0];
 744         s_idx = idx = cb->args[1];
 745
 746         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 747                 idx = 0;
 748                 head = &net->dev_index_head[h];
 749                 rcu_read_lock();
 750                 cb->seq = inet6_base_seq(net);
 751                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 752                         if (idx < s_idx)
 753                                 goto cont;
 754                         idev = __in6_dev_get(dev);
 755                         if (!idev)
 756                                 goto cont;
 757
 758                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 759                                                        &idev->cnf,
 760                                                        NETLINK_CB(cb->skb).portid,
 761                                                        nlh->nlmsg_seq,
 762                                                        RTM_NEWNETCONF,
 763                                                        NLM_F_MULTI,
 764                                                        NETCONFA_ALL) < 0) {
 765                                 rcu_read_unlock();
 766                                 goto done;
 767                         }
 768                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 769 cont:
 770                         idx++;
 771                 }
 772                 rcu_read_unlock();
 773         }
 774         if (h == NETDEV_HASHENTRIES) {
 775                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 776                                                net->ipv6.devconf_all,
 777                                                NETLINK_CB(cb->skb).portid,
 778                                                nlh->nlmsg_seq,
 779                                                RTM_NEWNETCONF, NLM_F_MULTI,
 780                                                NETCONFA_ALL) < 0)
 781                         goto done;
 782                 else
 783                         h++;
 784         }
 785         if (h == NETDEV_HASHENTRIES + 1) {
 786                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 787                                                net->ipv6.devconf_dflt,
 788                                                NETLINK_CB(cb->skb).portid,
 789                                                nlh->nlmsg_seq,
 790                                                RTM_NEWNETCONF, NLM_F_MULTI,
 791                                                NETCONFA_ALL) < 0)
 792                         goto done;
 793                 else
 794                         h++;
 795         }
 796 done:
 797         cb->args[0] = h;
 798         cb->args[1] = idx;
 799
 800         return skb->len;
 801 }
 802
 803 #ifdef CONFIG_SYSCTL
 804 static void dev_forward_change(struct inet6_dev *idev)
 805 {
 806         struct net_device *dev;
 807         struct inet6_ifaddr *ifa;
 808         LIST_HEAD(tmp_addr_list);
 809
 810         if (!idev)
 811                 return;
 812         dev = idev->dev;
 813         if (idev->cnf.forwarding)
 814                 dev_disable_lro(dev);
 815         if (dev->flags & IFF_MULTICAST) {
 816                 if (idev->cnf.forwarding) {
 817                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 818                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 819                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 820                 } else {
 821                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 822                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 823                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 824                 }
 825         }
 826
 827         read_lock_bh(&idev->lock);
 828         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 829                 if (ifa->flags&IFA_F_TENTATIVE)
 830                         continue;
 831                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
 832         }
 833         read_unlock_bh(&idev->lock);
 834
 835         while (!list_empty(&tmp_addr_list)) {
 836                 ifa = list_first_entry(&tmp_addr_list,
 837                                        struct inet6_ifaddr, if_list_aux);
 838                 list_del(&ifa->if_list_aux);
 839                 if (idev->cnf.forwarding)
 840                         addrconf_join_anycast(ifa);
 841                 else
 842                         addrconf_leave_anycast(ifa);
 843         }
 844
 845         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 846                                      NETCONFA_FORWARDING,
 847                                      dev->ifindex, &idev->cnf);
 848 }
 849
 850
 851 static void addrconf_forward_change(struct net *net, __s32 newf)
 852 {
 853         struct net_device *dev;
 854         struct inet6_dev *idev;
 855
 856         for_each_netdev(net, dev) {
 857                 idev = __in6_dev_get(dev);
 858                 if (idev) {
 859                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 860                         idev->cnf.forwarding = newf;
 861                         if (changed)
 862                                 dev_forward_change(idev);
 863                 }
 864         }
 865 }
 866
 867 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 868 {
 869         struct net *net;
 870         int old;
 871
 872         if (!rtnl_trylock())
 873                 return restart_syscall();
 874
 875         net = (struct net *)table->extra2;
 876         old = *p;
 877         *p = newf;
 878
 879         if (p == &net->ipv6.devconf_dflt->forwarding) {
 880                 if ((!newf) ^ (!old))
 881                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 882                                                      NETCONFA_FORWARDING,
 883                                                      NETCONFA_IFINDEX_DEFAULT,
 884                                                      net->ipv6.devconf_dflt);
 885                 rtnl_unlock();
 886                 return 0;
 887         }
 888
 889         if (p == &net->ipv6.devconf_all->forwarding) {
 890                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 891
 892                 net->ipv6.devconf_dflt->forwarding = newf;
 893                 if ((!newf) ^ (!old_dflt))
 894                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 895                                                      NETCONFA_FORWARDING,
 896                                                      NETCONFA_IFINDEX_DEFAULT,
 897                                                      net->ipv6.devconf_dflt);
 898
 899                 addrconf_forward_change(net, newf);
 900                 if ((!newf) ^ (!old))
 901                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 902                                                      NETCONFA_FORWARDING,
 903                                                      NETCONFA_IFINDEX_ALL,
 904                                                      net->ipv6.devconf_all);
 905         } else if ((!newf) ^ (!old))
 906                 dev_forward_change((struct inet6_dev *)table->extra1);
 907         rtnl_unlock();
 908
 909         if (newf)
 910                 rt6_purge_dflt_routers(net);
 911         return 1;
 912 }
 913
 914 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 915 {
 916         struct net_device *dev;
 917         struct inet6_dev *idev;
 918
 919         for_each_netdev(net, dev) {
 920                 idev = __in6_dev_get(dev);
 921                 if (idev) {
 922                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 923
 924                         idev->cnf.ignore_routes_with_linkdown = newf;
 925                         if (changed)
 926                                 inet6_netconf_notify_devconf(dev_net(dev),
 927                                                              RTM_NEWNETCONF,
 928                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 929                                                              dev->ifindex,
 930                                                              &idev->cnf);
 931                 }
 932         }
 933 }
 934
 935 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 936 {
 937         struct net *net;
 938         int old;
 939
 940         if (!rtnl_trylock())
 941                 return restart_syscall();
 942
 943         net = (struct net *)table->extra2;
 944         old = *p;
 945         *p = newf;
 946
 947         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 948                 if ((!newf) ^ (!old))
 949                         inet6_netconf_notify_devconf(net,
 950                                                      RTM_NEWNETCONF,
 951                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 952                                                      NETCONFA_IFINDEX_DEFAULT,
 953                                                      net->ipv6.devconf_dflt);
 954                 rtnl_unlock();
 955                 return 0;
 956         }
 957
 958         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 959                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 960                 addrconf_linkdown_change(net, newf);
 961                 if ((!newf) ^ (!old))
 962                         inet6_netconf_notify_devconf(net,
 963                                                      RTM_NEWNETCONF,
 964                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 965                                                      NETCONFA_IFINDEX_ALL,
 966                                                      net->ipv6.devconf_all);
 967         }
 968         rtnl_unlock();
 969
 970         return 1;
 971 }
 972
 973 #endif
 974
 975 /* Nobody refers to this ifaddr, destroy it */
 976 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 977 {
 978         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 979
 980 #ifdef NET_REFCNT_DEBUG
 981         pr_debug("%s\n", __func__);
 982 #endif
 983
 984         in6_dev_put(ifp->idev);
 985
 986         if (cancel_delayed_work(&ifp->dad_work))
 987                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 988                           ifp);
 989
 990         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 991                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 992                 return;
 993         }
 994
 995         kfree_rcu(ifp, rcu);
 996 }
 997
 998 static void
 999 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1000 {
1001         struct list_head *p;
1002         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1003
1004         /*
1005          * Each device address list is sorted in order of scope -
1006          * global before linklocal.
1007          */
1008         list_for_each(p, &idev->addr_list) {
1009                 struct inet6_ifaddr *ifa
1010                         = list_entry(p, struct inet6_ifaddr, if_list);
1011                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1012                         break;
1013         }
1014
1015         list_add_tail_rcu(&ifp->if_list, p);
1016 }
1017
1018 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1019 {
1020         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1021
1022         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1023 }
1024
1025 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1026                                struct net_device *dev, unsigned int hash)
1027 {
1028         struct inet6_ifaddr *ifp;
1029
1030         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1031                 if (!net_eq(dev_net(ifp->idev->dev), net))
1032                         continue;
1033                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1034                         if (!dev || ifp->idev->dev == dev)
1035                                 return true;
1036                 }
1037         }
1038         return false;
1039 }
1040
1041 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1042 {
1043         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1044         int err = 0;
1045
1046         spin_lock(&addrconf_hash_lock);
1047
1048         /* Ignore adding duplicate addresses on an interface */
1049         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1050                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1051                 err = -EEXIST;
1052         } else {
1053                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1054         }
1055
1056         spin_unlock(&addrconf_hash_lock);
1057
1058         return err;
1059 }
1060
1061 /* On success it returns ifp with increased reference count */
1062
1063 static struct inet6_ifaddr *
1064 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1065               bool can_block, struct netlink_ext_ack *extack)
1066 {
1067         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1068         int addr_type = ipv6_addr_type(cfg->pfx);
1069         struct net *net = dev_net(idev->dev);
1070         struct inet6_ifaddr *ifa = NULL;
1071         struct fib6_info *f6i = NULL;
1072         int err = 0;
1073
1074         if (addr_type == IPV6_ADDR_ANY ||
1075             (addr_type & IPV6_ADDR_MULTICAST &&
1076              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1077             (!(idev->dev->flags & IFF_LOOPBACK) &&
1078              !netif_is_l3_master(idev->dev) &&
1079              addr_type & IPV6_ADDR_LOOPBACK))
1080                 return ERR_PTR(-EADDRNOTAVAIL);
1081
1082         if (idev->dead) {
1083                 err = -ENODEV;                  /*XXX*/
1084                 goto out;
1085         }
1086
1087         if (idev->cnf.disable_ipv6) {
1088                 err = -EACCES;
1089                 goto out;
1090         }
1091
1092         /* validator notifier needs to be blocking;
1093          * do not call in atomic context
1094          */
1095         if (can_block) {
1096                 struct in6_validator_info i6vi = {
1097                         .i6vi_addr = *cfg->pfx,
1098                         .i6vi_dev = idev,
1099                         .extack = extack,
1100                 };
1101
1102                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1103                 err = notifier_to_errno(err);
1104                 if (err < 0)
1105                         goto out;
1106         }
1107
1108         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1109         if (!ifa) {
1110                 err = -ENOBUFS;
1111                 goto out;
1112         }
1113
1114         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1115         if (IS_ERR(f6i)) {
1116                 err = PTR_ERR(f6i);
1117                 f6i = NULL;
1118                 goto out;
1119         }
1120
1121         neigh_parms_data_state_setall(idev->nd_parms);
1122
1123         ifa->addr = *cfg->pfx;
1124         if (cfg->peer_pfx)
1125                 ifa->peer_addr = *cfg->peer_pfx;
1126
1127         spin_lock_init(&ifa->lock);
1128         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1129         INIT_HLIST_NODE(&ifa->addr_lst);
1130         ifa->scope = cfg->scope;
1131         ifa->prefix_len = cfg->plen;
1132         ifa->rt_priority = cfg->rt_priority;
1133         ifa->flags = cfg->ifa_flags;
1134         /* No need to add the TENTATIVE flag for addresses with NODAD */
1135         if (!(cfg->ifa_flags & IFA_F_NODAD))
1136                 ifa->flags |= IFA_F_TENTATIVE;
1137         ifa->valid_lft = cfg->valid_lft;
1138         ifa->prefered_lft = cfg->preferred_lft;
1139         ifa->cstamp = ifa->tstamp = jiffies;
1140         ifa->tokenized = false;
1141
1142         ifa->rt = f6i;
1143
1144         ifa->idev = idev;
1145         in6_dev_hold(idev);
1146
1147         /* For caller */
1148         refcount_set(&ifa->refcnt, 1);
1149
1150         rcu_read_lock_bh();
1151
1152         err = ipv6_add_addr_hash(idev->dev, ifa);
1153         if (err < 0) {
1154                 rcu_read_unlock_bh();
1155                 goto out;
1156         }
1157
1158         write_lock(&idev->lock);
1159
1160         /* Add to inet6_dev unicast addr list. */
1161         ipv6_link_dev_addr(idev, ifa);
1162
1163         if (ifa->flags&IFA_F_TEMPORARY) {
1164                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1165                 in6_ifa_hold(ifa);
1166         }
1167
1168         in6_ifa_hold(ifa);
1169         write_unlock(&idev->lock);
1170
1171         rcu_read_unlock_bh();
1172
1173         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1174 out:
1175         if (unlikely(err < 0)) {
1176                 fib6_info_release(f6i);
1177
1178                 if (ifa) {
1179                         if (ifa->idev)
1180                                 in6_dev_put(ifa->idev);
1181                         kfree(ifa);
1182                 }
1183                 ifa = ERR_PTR(err);
1184         }
1185
1186         return ifa;
1187 }
1188
1189 enum cleanup_prefix_rt_t {
1190         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1191         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1192         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1193 };
1194
1195 /*
1196  * Check, whether the prefix for ifp would still need a prefix route
1197  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1198  * constants.
1199  *
1200  * 1) we don't purge prefix if address was not permanent.
1201  *    prefix is managed by its own lifetime.
1202  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1203  * 3) if there are no addresses, delete prefix.
1204  * 4) if there are still other permanent address(es),
1205  *    corresponding prefix is still permanent.
1206  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1207  *    don't purge the prefix, assume user space is managing it.
1208  * 6) otherwise, update prefix lifetime to the
1209  *    longest valid lifetime among the corresponding
1210  *    addresses on the device.
1211  *    Note: subsequent RA will update lifetime.
1212  **/
1213 static enum cleanup_prefix_rt_t
1214 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1215 {
1216         struct inet6_ifaddr *ifa;
1217         struct inet6_dev *idev = ifp->idev;
1218         unsigned long lifetime;
1219         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1220
1221         *expires = jiffies;
1222
1223         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1224                 if (ifa == ifp)
1225                         continue;
1226                 if (ifa->prefix_len != ifp->prefix_len ||
1227                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1228                                        ifp->prefix_len))
1229                         continue;
1230                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1231                         return CLEANUP_PREFIX_RT_NOP;
1232
1233                 action = CLEANUP_PREFIX_RT_EXPIRE;
1234
1235                 spin_lock(&ifa->lock);
1236
1237                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1238                 /*
1239                  * Note: Because this address is
1240                  * not permanent, lifetime <
1241                  * LONG_MAX / HZ here.
1242                  */
1243                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1244                         *expires = ifa->tstamp + lifetime * HZ;
1245                 spin_unlock(&ifa->lock);
1246         }
1247
1248         return action;
1249 }
1250
1251 static void
1252 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1253                      bool del_rt, bool del_peer)
1254 {
1255         struct fib6_info *f6i;
1256
1257         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1258                                         ifp->prefix_len,
1259                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1260         if (f6i) {
1261                 if (del_rt)
1262                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1263                 else {
1264                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1265                                 fib6_set_expires(f6i, expires);
1266                         fib6_info_release(f6i);
1267                 }
1268         }
1269 }
1270
1271
1272 /* This function wants to get referenced ifp and releases it before return */
1273
1274 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1275 {
1276         int state;
1277         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1278         unsigned long expires;
1279
1280         ASSERT_RTNL();
1281
1282         spin_lock_bh(&ifp->lock);
1283         state = ifp->state;
1284         ifp->state = INET6_IFADDR_STATE_DEAD;
1285         spin_unlock_bh(&ifp->lock);
1286
1287         if (state == INET6_IFADDR_STATE_DEAD)
1288                 goto out;
1289
1290         spin_lock_bh(&addrconf_hash_lock);
1291         hlist_del_init_rcu(&ifp->addr_lst);
1292         spin_unlock_bh(&addrconf_hash_lock);
1293
1294         write_lock_bh(&ifp->idev->lock);
1295
1296         if (ifp->flags&IFA_F_TEMPORARY) {
1297                 list_del(&ifp->tmp_list);
1298                 if (ifp->ifpub) {
1299                         in6_ifa_put(ifp->ifpub);
1300                         ifp->ifpub = NULL;
1301                 }
1302                 __in6_ifa_put(ifp);
1303         }
1304
1305         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1306                 action = check_cleanup_prefix_route(ifp, &expires);
1307
1308         list_del_rcu(&ifp->if_list);
1309         __in6_ifa_put(ifp);
1310
1311         write_unlock_bh(&ifp->idev->lock);
1312
1313         addrconf_del_dad_work(ifp);
1314
1315         ipv6_ifa_notify(RTM_DELADDR, ifp);
1316
1317         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1318
1319         if (action != CLEANUP_PREFIX_RT_NOP) {
1320                 cleanup_prefix_route(ifp, expires,
1321                         action == CLEANUP_PREFIX_RT_DEL, false);
1322         }
1323
1324         /* clean up prefsrc entries */
1325         rt6_remove_prefsrc(ifp);
1326 out:
1327         in6_ifa_put(ifp);
1328 }
1329
1330 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1331 {
1332         struct inet6_dev *idev = ifp->idev;
1333         unsigned long tmp_tstamp, age;
1334         unsigned long regen_advance;
1335         unsigned long now = jiffies;
1336         s32 cnf_temp_preferred_lft;
1337         struct inet6_ifaddr *ift;
1338         struct ifa6_config cfg;
1339         long max_desync_factor;
1340         struct in6_addr addr;
1341         int ret = 0;
1342
1343         write_lock_bh(&idev->lock);
1344
1345 retry:
1346         in6_dev_hold(idev);
1347         if (idev->cnf.use_tempaddr <= 0) {
1348                 write_unlock_bh(&idev->lock);
1349                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1350                 in6_dev_put(idev);
1351                 ret = -1;
1352                 goto out;
1353         }
1354         spin_lock_bh(&ifp->lock);
1355         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1356                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1357                 spin_unlock_bh(&ifp->lock);
1358                 write_unlock_bh(&idev->lock);
1359                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1360                         __func__);
1361                 in6_dev_put(idev);
1362                 ret = -1;
1363                 goto out;
1364         }
1365         in6_ifa_hold(ifp);
1366         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1367         ipv6_gen_rnd_iid(&addr);
1368
1369         age = (now - ifp->tstamp) / HZ;
1370
1371         regen_advance = idev->cnf.regen_max_retry *
1372                         idev->cnf.dad_transmits *
1373                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1374
1375         /* recalculate max_desync_factor each time and update
1376          * idev->desync_factor if it's larger
1377          */
1378         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1379         max_desync_factor = min_t(long,
1380                                   idev->cnf.max_desync_factor,
1381                                   cnf_temp_preferred_lft - regen_advance);
1382
1383         if (unlikely(idev->desync_factor > max_desync_factor)) {
1384                 if (max_desync_factor > 0) {
1385                         get_random_bytes(&idev->desync_factor,
1386                                          sizeof(idev->desync_factor));
1387                         idev->desync_factor %= max_desync_factor;
1388                 } else {
1389                         idev->desync_factor = 0;
1390                 }
1391         }
1392
1393         memset(&cfg, 0, sizeof(cfg));
1394         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1395                               idev->cnf.temp_valid_lft + age);
1396         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1397         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1398
1399         cfg.plen = ifp->prefix_len;
1400         tmp_tstamp = ifp->tstamp;
1401         spin_unlock_bh(&ifp->lock);
1402
1403         write_unlock_bh(&idev->lock);
1404
1405         /* A temporary address is created only if this calculated Preferred
1406          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1407          * an implementation must not create a temporary address with a zero
1408          * Preferred Lifetime.
1409          * Use age calculation as in addrconf_verify to avoid unnecessary
1410          * temporary addresses being generated.
1411          */
1412         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1413         if (cfg.preferred_lft <= regen_advance + age) {
1414                 in6_ifa_put(ifp);
1415                 in6_dev_put(idev);
1416                 ret = -1;
1417                 goto out;
1418         }
1419
1420         cfg.ifa_flags = IFA_F_TEMPORARY;
1421         /* set in addrconf_prefix_rcv() */
1422         if (ifp->flags & IFA_F_OPTIMISTIC)
1423                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1424
1425         cfg.pfx = &addr;
1426         cfg.scope = ipv6_addr_scope(cfg.pfx);
1427
1428         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1429         if (IS_ERR(ift)) {
1430                 in6_ifa_put(ifp);
1431                 in6_dev_put(idev);
1432                 pr_info("%s: retry temporary address regeneration\n", __func__);
1433                 write_lock_bh(&idev->lock);
1434                 goto retry;
1435         }
1436
1437         spin_lock_bh(&ift->lock);
1438         ift->ifpub = ifp;
1439         ift->cstamp = now;
1440         ift->tstamp = tmp_tstamp;
1441         spin_unlock_bh(&ift->lock);
1442
1443         addrconf_dad_start(ift);
1444         in6_ifa_put(ift);
1445         in6_dev_put(idev);
1446 out:
1447         return ret;
1448 }
1449
1450 /*
1451  *      Choose an appropriate source address (RFC3484)
1452  */
1453 enum {
1454         IPV6_SADDR_RULE_INIT = 0,
1455         IPV6_SADDR_RULE_LOCAL,
1456         IPV6_SADDR_RULE_SCOPE,
1457         IPV6_SADDR_RULE_PREFERRED,
1458 #ifdef CONFIG_IPV6_MIP6
1459         IPV6_SADDR_RULE_HOA,
1460 #endif
1461         IPV6_SADDR_RULE_OIF,
1462         IPV6_SADDR_RULE_LABEL,
1463         IPV6_SADDR_RULE_PRIVACY,
1464         IPV6_SADDR_RULE_ORCHID,
1465         IPV6_SADDR_RULE_PREFIX,
1466 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1467         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1468 #endif
1469         IPV6_SADDR_RULE_MAX
1470 };
1471
1472 struct ipv6_saddr_score {
1473         int                     rule;
1474         int                     addr_type;
1475         struct inet6_ifaddr     *ifa;
1476         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1477         int                     scopedist;
1478         int                     matchlen;
1479 };
1480
1481 struct ipv6_saddr_dst {
1482         const struct in6_addr *addr;
1483         int ifindex;
1484         int scope;
1485         int label;
1486         unsigned int prefs;
1487 };
1488
1489 static inline int ipv6_saddr_preferred(int type)
1490 {
1491         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1492                 return 1;
1493         return 0;
1494 }
1495
1496 static bool ipv6_use_optimistic_addr(struct net *net,
1497                                      struct inet6_dev *idev)
1498 {
1499 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1500         if (!idev)
1501                 return false;
1502         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1503                 return false;
1504         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1505                 return false;
1506
1507         return true;
1508 #else
1509         return false;
1510 #endif
1511 }
1512
1513 static bool ipv6_allow_optimistic_dad(struct net *net,
1514                                       struct inet6_dev *idev)
1515 {
1516 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1517         if (!idev)
1518                 return false;
1519         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1520                 return false;
1521
1522         return true;
1523 #else
1524         return false;
1525 #endif
1526 }
1527
1528 static int ipv6_get_saddr_eval(struct net *net,
1529                                struct ipv6_saddr_score *score,
1530                                struct ipv6_saddr_dst *dst,
1531                                int i)
1532 {
1533         int ret;
1534
1535         if (i <= score->rule) {
1536                 switch (i) {
1537                 case IPV6_SADDR_RULE_SCOPE:
1538                         ret = score->scopedist;
1539                         break;
1540                 case IPV6_SADDR_RULE_PREFIX:
1541                         ret = score->matchlen;
1542                         break;
1543                 default:
1544                         ret = !!test_bit(i, score->scorebits);
1545                 }
1546                 goto out;
1547         }
1548
1549         switch (i) {
1550         case IPV6_SADDR_RULE_INIT:
1551                 /* Rule 0: remember if hiscore is not ready yet */
1552                 ret = !!score->ifa;
1553                 break;
1554         case IPV6_SADDR_RULE_LOCAL:
1555                 /* Rule 1: Prefer same address */
1556                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1557                 break;
1558         case IPV6_SADDR_RULE_SCOPE:
1559                 /* Rule 2: Prefer appropriate scope
1560                  *
1561                  *      ret
1562                  *       ^
1563                  *    -1 |  d 15
1564                  *    ---+--+-+---> scope
1565                  *       |
1566                  *       |             d is scope of the destination.
1567                  *  B-d  |  \
1568                  *       |   \      <- smaller scope is better if
1569                  *  B-15 |    \        if scope is enough for destination.
1570                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1571                  * d-C-1 | /
1572                  *       |/         <- greater is better
1573                  *   -C  /             if scope is not enough for destination.
1574                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1575                  *
1576                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1577                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1578                  * Assume B = 0 and we get C > 29.
1579                  */
1580                 ret = __ipv6_addr_src_scope(score->addr_type);
1581                 if (ret >= dst->scope)
1582                         ret = -ret;
1583                 else
1584                         ret -= 128;     /* 30 is enough */
1585                 score->scopedist = ret;
1586                 break;
1587         case IPV6_SADDR_RULE_PREFERRED:
1588             {
1589                 /* Rule 3: Avoid deprecated and optimistic addresses */
1590                 u8 avoid = IFA_F_DEPRECATED;
1591
1592                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1593                         avoid |= IFA_F_OPTIMISTIC;
1594                 ret = ipv6_saddr_preferred(score->addr_type) ||
1595                       !(score->ifa->flags & avoid);
1596                 break;
1597             }
1598 #ifdef CONFIG_IPV6_MIP6
1599         case IPV6_SADDR_RULE_HOA:
1600             {
1601                 /* Rule 4: Prefer home address */
1602                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1603                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1604                 break;
1605             }
1606 #endif
1607         case IPV6_SADDR_RULE_OIF:
1608                 /* Rule 5: Prefer outgoing interface */
1609                 ret = (!dst->ifindex ||
1610                        dst->ifindex == score->ifa->idev->dev->ifindex);
1611                 break;
1612         case IPV6_SADDR_RULE_LABEL:
1613                 /* Rule 6: Prefer matching label */
1614                 ret = ipv6_addr_label(net,
1615                                       &score->ifa->addr, score->addr_type,
1616                                       score->ifa->idev->dev->ifindex) == dst->label;
1617                 break;
1618         case IPV6_SADDR_RULE_PRIVACY:
1619             {
1620                 /* Rule 7: Prefer public address
1621                  * Note: prefer temporary address if use_tempaddr >= 2
1622                  */
1623                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1624                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1625                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1626                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1627                 break;
1628             }
1629         case IPV6_SADDR_RULE_ORCHID:
1630                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1631                  *          non-ORCHID vs non-ORCHID
1632                  */
1633                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1634                         ipv6_addr_orchid(dst->addr));
1635                 break;
1636         case IPV6_SADDR_RULE_PREFIX:
1637                 /* Rule 8: Use longest matching prefix */
1638                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1639                 if (ret > score->ifa->prefix_len)
1640                         ret = score->ifa->prefix_len;
1641                 score->matchlen = ret;
1642                 break;
1643 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1644         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1645                 /* Optimistic addresses still have lower precedence than other
1646                  * preferred addresses.
1647                  */
1648                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1649                 break;
1650 #endif
1651         default:
1652                 ret = 0;
1653         }
1654
1655         if (ret)
1656                 __set_bit(i, score->scorebits);
1657         score->rule = i;
1658 out:
1659         return ret;
1660 }
1661
1662 static int __ipv6_dev_get_saddr(struct net *net,
1663                                 struct ipv6_saddr_dst *dst,
1664                                 struct inet6_dev *idev,
1665                                 struct ipv6_saddr_score *scores,
1666                                 int hiscore_idx)
1667 {
1668         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1669
1670         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1671                 int i;
1672
1673                 /*
1674                  * - Tentative Address (RFC2462 section 5.4)
1675                  *  - A tentative address is not considered
1676                  *    "assigned to an interface" in the traditional
1677                  *    sense, unless it is also flagged as optimistic.
1678                  * - Candidate Source Address (section 4)
1679                  *  - In any case, anycast addresses, multicast
1680                  *    addresses, and the unspecified address MUST
1681                  *    NOT be included in a candidate set.
1682                  */
1683                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1684                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1685                         continue;
1686
1687                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1688
1689                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1690                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1691                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1692                                             idev->dev->name);
1693                         continue;
1694                 }
1695
1696                 score->rule = -1;
1697                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1698
1699                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1700                         int minihiscore, miniscore;
1701
1702                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1703                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1704
1705                         if (minihiscore > miniscore) {
1706                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1707                                     score->scopedist > 0) {
1708                                         /*
1709                                          * special case:
1710                                          * each remaining entry
1711                                          * has too small (not enough)
1712                                          * scope, because ifa entries
1713                                          * are sorted by their scope
1714                                          * values.
1715                                          */
1716                                         goto out;
1717                                 }
1718                                 break;
1719                         } else if (minihiscore < miniscore) {
1720                                 swap(hiscore, score);
1721                                 hiscore_idx = 1 - hiscore_idx;
1722
1723                                 /* restore our iterator */
1724                                 score->ifa = hiscore->ifa;
1725
1726                                 break;
1727                         }
1728                 }
1729         }
1730 out:
1731         return hiscore_idx;
1732 }
1733
1734 static int ipv6_get_saddr_master(struct net *net,
1735                                  const struct net_device *dst_dev,
1736                                  const struct net_device *master,
1737                                  struct ipv6_saddr_dst *dst,
1738                                  struct ipv6_saddr_score *scores,
1739                                  int hiscore_idx)
1740 {
1741         struct inet6_dev *idev;
1742
1743         idev = __in6_dev_get(dst_dev);
1744         if (idev)
1745                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1746                                                    scores, hiscore_idx);
1747
1748         idev = __in6_dev_get(master);
1749         if (idev)
1750                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1751                                                    scores, hiscore_idx);
1752
1753         return hiscore_idx;
1754 }
1755
1756 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1757                        const struct in6_addr *daddr, unsigned int prefs,
1758                        struct in6_addr *saddr)
1759 {
1760         struct ipv6_saddr_score scores[2], *hiscore;
1761         struct ipv6_saddr_dst dst;
1762         struct inet6_dev *idev;
1763         struct net_device *dev;
1764         int dst_type;
1765         bool use_oif_addr = false;
1766         int hiscore_idx = 0;
1767         int ret = 0;
1768
1769         dst_type = __ipv6_addr_type(daddr);
1770         dst.addr = daddr;
1771         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1772         dst.scope = __ipv6_addr_src_scope(dst_type);
1773         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1774         dst.prefs = prefs;
1775
1776         scores[hiscore_idx].rule = -1;
1777         scores[hiscore_idx].ifa = NULL;
1778
1779         rcu_read_lock();
1780
1781         /* Candidate Source Address (section 4)
1782          *  - multicast and link-local destination address,
1783          *    the set of candidate source address MUST only
1784          *    include addresses assigned to interfaces
1785          *    belonging to the same link as the outgoing
1786          *    interface.
1787          * (- For site-local destination addresses, the
1788          *    set of candidate source addresses MUST only
1789          *    include addresses assigned to interfaces
1790          *    belonging to the same site as the outgoing
1791          *    interface.)
1792          *  - "It is RECOMMENDED that the candidate source addresses
1793          *    be the set of unicast addresses assigned to the
1794          *    interface that will be used to send to the destination
1795          *    (the 'outgoing' interface)." (RFC 6724)
1796          */
1797         if (dst_dev) {
1798                 idev = __in6_dev_get(dst_dev);
1799                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1800                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1801                     (idev && idev->cnf.use_oif_addrs_only)) {
1802                         use_oif_addr = true;
1803                 }
1804         }
1805
1806         if (use_oif_addr) {
1807                 if (idev)
1808                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1809         } else {
1810                 const struct net_device *master;
1811                 int master_idx = 0;
1812
1813                 /* if dst_dev exists and is enslaved to an L3 device, then
1814                  * prefer addresses from dst_dev and then the master over
1815                  * any other enslaved devices in the L3 domain.
1816                  */
1817                 master = l3mdev_master_dev_rcu(dst_dev);
1818                 if (master) {
1819                         master_idx = master->ifindex;
1820
1821                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1822                                                             master, &dst,
1823                                                             scores, hiscore_idx);
1824
1825                         if (scores[hiscore_idx].ifa)
1826                                 goto out;
1827                 }
1828
1829                 for_each_netdev_rcu(net, dev) {
1830                         /* only consider addresses on devices in the
1831                          * same L3 domain
1832                          */
1833                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1834                                 continue;
1835                         idev = __in6_dev_get(dev);
1836                         if (!idev)
1837                                 continue;
1838                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1839                 }
1840         }
1841
1842 out:
1843         hiscore = &scores[hiscore_idx];
1844         if (!hiscore->ifa)
1845                 ret = -EADDRNOTAVAIL;
1846         else
1847                 *saddr = hiscore->ifa->addr;
1848
1849         rcu_read_unlock();
1850         return ret;
1851 }
1852 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1853
1854 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1855                       u32 banned_flags)
1856 {
1857         struct inet6_ifaddr *ifp;
1858         int err = -EADDRNOTAVAIL;
1859
1860         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1861                 if (ifp->scope > IFA_LINK)
1862                         break;
1863                 if (ifp->scope == IFA_LINK &&
1864                     !(ifp->flags & banned_flags)) {
1865                         *addr = ifp->addr;
1866                         err = 0;
1867                         break;
1868                 }
1869         }
1870         return err;
1871 }
1872
1873 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1874                     u32 banned_flags)
1875 {
1876         struct inet6_dev *idev;
1877         int err = -EADDRNOTAVAIL;
1878
1879         rcu_read_lock();
1880         idev = __in6_dev_get(dev);
1881         if (idev) {
1882                 read_lock_bh(&idev->lock);
1883                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1884                 read_unlock_bh(&idev->lock);
1885         }
1886         rcu_read_unlock();
1887         return err;
1888 }
1889
1890 static int ipv6_count_addresses(const struct inet6_dev *idev)
1891 {
1892         const struct inet6_ifaddr *ifp;
1893         int cnt = 0;
1894
1895         rcu_read_lock();
1896         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1897                 cnt++;
1898         rcu_read_unlock();
1899         return cnt;
1900 }
1901
1902 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1903                   const struct net_device *dev, int strict)
1904 {
1905         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1906                                        strict, IFA_F_TENTATIVE);
1907 }
1908 EXPORT_SYMBOL(ipv6_chk_addr);
1909
1910 /* device argument is used to find the L3 domain of interest. If
1911  * skip_dev_check is set, then the ifp device is not checked against
1912  * the passed in dev argument. So the 2 cases for addresses checks are:
1913  *   1. does the address exist in the L3 domain that dev is part of
1914  *      (skip_dev_check = true), or
1915  *
1916  *   2. does the address exist on the specific device
1917  *      (skip_dev_check = false)
1918  */
1919 static struct net_device *
1920 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1921                           const struct net_device *dev, bool skip_dev_check,
1922                           int strict, u32 banned_flags)
1923 {
1924         unsigned int hash = inet6_addr_hash(net, addr);
1925         struct net_device *l3mdev, *ndev;
1926         struct inet6_ifaddr *ifp;
1927         u32 ifp_flags;
1928
1929         rcu_read_lock();
1930
1931         l3mdev = l3mdev_master_dev_rcu(dev);
1932         if (skip_dev_check)
1933                 dev = NULL;
1934
1935         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1936                 ndev = ifp->idev->dev;
1937                 if (!net_eq(dev_net(ndev), net))
1938                         continue;
1939
1940                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1941                         continue;
1942
1943                 /* Decouple optimistic from tentative for evaluation here.
1944                  * Ban optimistic addresses explicitly, when required.
1945                  */
1946                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1947                             ? (ifp->flags&~IFA_F_TENTATIVE)
1948                             : ifp->flags;
1949                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1950                     !(ifp_flags&banned_flags) &&
1951                     (!dev || ndev == dev ||
1952                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1953                         rcu_read_unlock();
1954                         return ndev;
1955                 }
1956         }
1957
1958         rcu_read_unlock();
1959         return NULL;
1960 }
1961
1962 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1963                             const struct net_device *dev, bool skip_dev_check,
1964                             int strict, u32 banned_flags)
1965 {
1966         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1967                                          strict, banned_flags) ? 1 : 0;
1968 }
1969 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1970
1971
1972 /* Compares an address/prefix_len with addresses on device @dev.
1973  * If one is found it returns true.
1974  */
1975 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1976         const unsigned int prefix_len, struct net_device *dev)
1977 {
1978         const struct inet6_ifaddr *ifa;
1979         const struct inet6_dev *idev;
1980         bool ret = false;
1981
1982         rcu_read_lock();
1983         idev = __in6_dev_get(dev);
1984         if (idev) {
1985                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1986                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1987                         if (ret)
1988                                 break;
1989                 }
1990         }
1991         rcu_read_unlock();
1992
1993         return ret;
1994 }
1995 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1996
1997 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1998 {
1999         const struct inet6_ifaddr *ifa;
2000         const struct inet6_dev *idev;
2001         int     onlink;
2002
2003         onlink = 0;
2004         rcu_read_lock();
2005         idev = __in6_dev_get(dev);
2006         if (idev) {
2007                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2008                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
2009                                                    ifa->prefix_len);
2010                         if (onlink)
2011                                 break;
2012                 }
2013         }
2014         rcu_read_unlock();
2015         return onlink;
2016 }
2017 EXPORT_SYMBOL(ipv6_chk_prefix);
2018
2019 /**
2020  * ipv6_dev_find - find the first device with a given source address.
2021  * @net: the net namespace
2022  * @addr: the source address
2023  *
2024  * The caller should be protected by RCU, or RTNL.
2025  */
2026 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2027                                  struct net_device *dev)
2028 {
2029         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2030                                          IFA_F_TENTATIVE);
2031 }
2032 EXPORT_SYMBOL(ipv6_dev_find);
2033
2034 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2035                                      struct net_device *dev, int strict)
2036 {
2037         unsigned int hash = inet6_addr_hash(net, addr);
2038         struct inet6_ifaddr *ifp, *result = NULL;
2039
2040         rcu_read_lock();
2041         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2042                 if (!net_eq(dev_net(ifp->idev->dev), net))
2043                         continue;
2044                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2045                         if (!dev || ifp->idev->dev == dev ||
2046                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2047                                 if (in6_ifa_hold_safe(ifp)) {
2048                                         result = ifp;
2049                                         break;
2050                                 }
2051                         }
2052                 }
2053         }
2054         rcu_read_unlock();
2055
2056         return result;
2057 }
2058
2059 /* Gets referenced address, destroys ifaddr */
2060
2061 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2062 {
2063         if (dad_failed)
2064                 ifp->flags |= IFA_F_DADFAILED;
2065
2066         if (ifp->flags&IFA_F_TEMPORARY) {
2067                 struct inet6_ifaddr *ifpub;
2068                 spin_lock_bh(&ifp->lock);
2069                 ifpub = ifp->ifpub;
2070                 if (ifpub) {
2071                         in6_ifa_hold(ifpub);
2072                         spin_unlock_bh(&ifp->lock);
2073                         ipv6_create_tempaddr(ifpub, true);
2074                         in6_ifa_put(ifpub);
2075                 } else {
2076                         spin_unlock_bh(&ifp->lock);
2077                 }
2078                 ipv6_del_addr(ifp);
2079         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2080                 spin_lock_bh(&ifp->lock);
2081                 addrconf_del_dad_work(ifp);
2082                 ifp->flags |= IFA_F_TENTATIVE;
2083                 if (dad_failed)
2084                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2085                 spin_unlock_bh(&ifp->lock);
2086                 if (dad_failed)
2087                         ipv6_ifa_notify(0, ifp);
2088                 in6_ifa_put(ifp);
2089         } else {
2090                 ipv6_del_addr(ifp);
2091         }
2092 }
2093
2094 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2095 {
2096         int err = -ENOENT;
2097
2098         spin_lock_bh(&ifp->lock);
2099         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2100                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2101                 err = 0;
2102         }
2103         spin_unlock_bh(&ifp->lock);
2104
2105         return err;
2106 }
2107
2108 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2109 {
2110         struct inet6_dev *idev = ifp->idev;
2111         struct net *net = dev_net(ifp->idev->dev);
2112
2113         if (addrconf_dad_end(ifp)) {
2114                 in6_ifa_put(ifp);
2115                 return;
2116         }
2117
2118         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2119                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2120
2121         spin_lock_bh(&ifp->lock);
2122
2123         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2124                 struct in6_addr new_addr;
2125                 struct inet6_ifaddr *ifp2;
2126                 int retries = ifp->stable_privacy_retry + 1;
2127                 struct ifa6_config cfg = {
2128                         .pfx = &new_addr,
2129                         .plen = ifp->prefix_len,
2130                         .ifa_flags = ifp->flags,
2131                         .valid_lft = ifp->valid_lft,
2132                         .preferred_lft = ifp->prefered_lft,
2133                         .scope = ifp->scope,
2134                 };
2135
2136                 if (retries > net->ipv6.sysctl.idgen_retries) {
2137                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2138                                              ifp->idev->dev->name);
2139                         goto errdad;
2140                 }
2141
2142                 new_addr = ifp->addr;
2143                 if (ipv6_generate_stable_address(&new_addr, retries,
2144                                                  idev))
2145                         goto errdad;
2146
2147                 spin_unlock_bh(&ifp->lock);
2148
2149                 if (idev->cnf.max_addresses &&
2150                     ipv6_count_addresses(idev) >=
2151                     idev->cnf.max_addresses)
2152                         goto lock_errdad;
2153
2154                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2155                                      ifp->idev->dev->name);
2156
2157                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2158                 if (IS_ERR(ifp2))
2159                         goto lock_errdad;
2160
2161                 spin_lock_bh(&ifp2->lock);
2162                 ifp2->stable_privacy_retry = retries;
2163                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2164                 spin_unlock_bh(&ifp2->lock);
2165
2166                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2167                 in6_ifa_put(ifp2);
2168 lock_errdad:
2169                 spin_lock_bh(&ifp->lock);
2170         }
2171
2172 errdad:
2173         /* transition from _POSTDAD to _ERRDAD */
2174         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2175         spin_unlock_bh(&ifp->lock);
2176
2177         addrconf_mod_dad_work(ifp, 0);
2178         in6_ifa_put(ifp);
2179 }
2180
2181 /* Join to solicited addr multicast group.
2182  * caller must hold RTNL */
2183 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2184 {
2185         struct in6_addr maddr;
2186
2187         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2188                 return;
2189
2190         addrconf_addr_solict_mult(addr, &maddr);
2191         ipv6_dev_mc_inc(dev, &maddr);
2192 }
2193
2194 /* caller must hold RTNL */
2195 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2196 {
2197         struct in6_addr maddr;
2198
2199         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2200                 return;
2201
2202         addrconf_addr_solict_mult(addr, &maddr);
2203         __ipv6_dev_mc_dec(idev, &maddr);
2204 }
2205
2206 /* caller must hold RTNL */
2207 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2208 {
2209         struct in6_addr addr;
2210
2211         if (ifp->prefix_len >= 127) /* RFC 6164 */
2212                 return;
2213         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2214         if (ipv6_addr_any(&addr))
2215                 return;
2216         __ipv6_dev_ac_inc(ifp->idev, &addr);
2217 }
2218
2219 /* caller must hold RTNL */
2220 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2221 {
2222         struct in6_addr addr;
2223
2224         if (ifp->prefix_len >= 127) /* RFC 6164 */
2225                 return;
2226         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2227         if (ipv6_addr_any(&addr))
2228                 return;
2229         __ipv6_dev_ac_dec(ifp->idev, &addr);
2230 }
2231
2232 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2233 {
2234         switch (dev->addr_len) {
2235         case ETH_ALEN:
2236                 memcpy(eui, dev->dev_addr, 3);
2237                 eui[3] = 0xFF;
2238                 eui[4] = 0xFE;
2239                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2240                 break;
2241         case EUI64_ADDR_LEN:
2242                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2243                 eui[0] ^= 2;
2244                 break;
2245         default:
2246                 return -1;
2247         }
2248
2249         return 0;
2250 }
2251
2252 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2253 {
2254         union fwnet_hwaddr *ha;
2255
2256         if (dev->addr_len != FWNET_ALEN)
2257                 return -1;
2258
2259         ha = (union fwnet_hwaddr *)dev->dev_addr;
2260
2261         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2262         eui[0] ^= 2;
2263         return 0;
2264 }
2265
2266 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2267 {
2268         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2269         if (dev->addr_len != ARCNET_ALEN)
2270                 return -1;
2271         memset(eui, 0, 7);
2272         eui[7] = *(u8 *)dev->dev_addr;
2273         return 0;
2274 }
2275
2276 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2277 {
2278         if (dev->addr_len != INFINIBAND_ALEN)
2279                 return -1;
2280         memcpy(eui, dev->dev_addr + 12, 8);
2281         eui[0] |= 2;
2282         return 0;
2283 }
2284
2285 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2286 {
2287         if (addr == 0)
2288                 return -1;
2289         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2290                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2291                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2292                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2293                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2294                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2295         eui[1] = 0;
2296         eui[2] = 0x5E;
2297         eui[3] = 0xFE;
2298         memcpy(eui + 4, &addr, 4);
2299         return 0;
2300 }
2301
2302 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2303 {
2304         if (dev->priv_flags & IFF_ISATAP)
2305                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2306         return -1;
2307 }
2308
2309 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2310 {
2311         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2312 }
2313
2314 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2315 {
2316         memcpy(eui, dev->perm_addr, 3);
2317         memcpy(eui + 5, dev->perm_addr + 3, 3);
2318         eui[3] = 0xFF;
2319         eui[4] = 0xFE;
2320         eui[0] ^= 2;
2321         return 0;
2322 }
2323
2324 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2325 {
2326         switch (dev->type) {
2327         case ARPHRD_ETHER:
2328         case ARPHRD_FDDI:
2329                 return addrconf_ifid_eui48(eui, dev);
2330         case ARPHRD_ARCNET:
2331                 return addrconf_ifid_arcnet(eui, dev);
2332         case ARPHRD_INFINIBAND:
2333                 return addrconf_ifid_infiniband(eui, dev);
2334         case ARPHRD_SIT:
2335                 return addrconf_ifid_sit(eui, dev);
2336         case ARPHRD_IPGRE:
2337         case ARPHRD_TUNNEL:
2338                 return addrconf_ifid_gre(eui, dev);
2339         case ARPHRD_6LOWPAN:
2340                 return addrconf_ifid_6lowpan(eui, dev);
2341         case ARPHRD_IEEE1394:
2342                 return addrconf_ifid_ieee1394(eui, dev);
2343         case ARPHRD_TUNNEL6:
2344         case ARPHRD_IP6GRE:
2345         case ARPHRD_RAWIP:
2346                 return addrconf_ifid_ip6tnl(eui, dev);
2347         }
2348         return -1;
2349 }
2350
2351 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2352 {
2353         int err = -1;
2354         struct inet6_ifaddr *ifp;
2355
2356         read_lock_bh(&idev->lock);
2357         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2358                 if (ifp->scope > IFA_LINK)
2359                         break;
2360                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2361                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2362                         err = 0;
2363                         break;
2364                 }
2365         }
2366         read_unlock_bh(&idev->lock);
2367         return err;
2368 }
2369
2370 /* Generation of a randomized Interface Identifier
2371  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2372  */
2373
2374 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2375 {
2376 regen:
2377         get_random_bytes(&addr->s6_addr[8], 8);
2378
2379         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2380          * check if generated address is not inappropriate:
2381          *
2382          * - Reserved IPv6 Interface Identifers
2383          * - XXX: already assigned to an address on the device
2384          */
2385
2386         /* Subnet-router anycast: 0000:0000:0000:0000 */
2387         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2388                 goto regen;
2389
2390         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2391          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2392          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2393          */
2394         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2395             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2396                 goto regen;
2397
2398         /* Reserved subnet anycast addresses */
2399         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2400             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2401                 goto regen;
2402 }
2403
2404 /*
2405  *      Add prefix route.
2406  */
2407
2408 static void
2409 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2410                       struct net_device *dev, unsigned long expires,
2411                       u32 flags, gfp_t gfp_flags)
2412 {
2413         struct fib6_config cfg = {
2414                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2415                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2416                 .fc_ifindex = dev->ifindex,
2417                 .fc_expires = expires,
2418                 .fc_dst_len = plen,
2419                 .fc_flags = RTF_UP | flags,
2420                 .fc_nlinfo.nl_net = dev_net(dev),
2421                 .fc_protocol = RTPROT_KERNEL,
2422                 .fc_type = RTN_UNICAST,
2423         };
2424
2425         cfg.fc_dst = *pfx;
2426
2427         /* Prevent useless cloning on PtP SIT.
2428            This thing is done here expecting that the whole
2429            class of non-broadcast devices need not cloning.
2430          */
2431 #if IS_ENABLED(CONFIG_IPV6_SIT)
2432         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2433                 cfg.fc_flags |= RTF_NONEXTHOP;
2434 #endif
2435
2436         ip6_route_add(&cfg, gfp_flags, NULL);
2437 }
2438
2439
2440 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2441                                                   int plen,
2442                                                   const struct net_device *dev,
2443                                                   u32 flags, u32 noflags,
2444                                                   bool no_gw)
2445 {
2446         struct fib6_node *fn;
2447         struct fib6_info *rt = NULL;
2448         struct fib6_table *table;
2449         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2450
2451         table = fib6_get_table(dev_net(dev), tb_id);
2452         if (!table)
2453                 return NULL;
2454
2455         rcu_read_lock();
2456         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2457         if (!fn)
2458                 goto out;
2459
2460         for_each_fib6_node_rt_rcu(fn) {
2461                 /* prefix routes only use builtin fib6_nh */
2462                 if (rt->nh)
2463                         continue;
2464
2465                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2466                         continue;
2467                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2468                         continue;
2469                 if ((rt->fib6_flags & flags) != flags)
2470                         continue;
2471                 if ((rt->fib6_flags & noflags) != 0)
2472                         continue;
2473                 if (!fib6_info_hold_safe(rt))
2474                         continue;
2475                 break;
2476         }
2477 out:
2478         rcu_read_unlock();
2479         return rt;
2480 }
2481
2482
2483 /* Create "default" multicast route to the interface */
2484
2485 static void addrconf_add_mroute(struct net_device *dev)
2486 {
2487         struct fib6_config cfg = {
2488                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2489                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2490                 .fc_ifindex = dev->ifindex,
2491                 .fc_dst_len = 8,
2492                 .fc_flags = RTF_UP,
2493                 .fc_type = RTN_MULTICAST,
2494                 .fc_nlinfo.nl_net = dev_net(dev),
2495                 .fc_protocol = RTPROT_KERNEL,
2496         };
2497
2498         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2499
2500         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2501 }
2502
2503 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2504 {
2505         struct inet6_dev *idev;
2506
2507         ASSERT_RTNL();
2508
2509         idev = ipv6_find_idev(dev);
2510         if (IS_ERR(idev))
2511                 return idev;
2512
2513         if (idev->cnf.disable_ipv6)
2514                 return ERR_PTR(-EACCES);
2515
2516         /* Add default multicast route */
2517         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2518                 addrconf_add_mroute(dev);
2519
2520         return idev;
2521 }
2522
2523 static void manage_tempaddrs(struct inet6_dev *idev,
2524                              struct inet6_ifaddr *ifp,
2525                              __u32 valid_lft, __u32 prefered_lft,
2526                              bool create, unsigned long now)
2527 {
2528         u32 flags;
2529         struct inet6_ifaddr *ift;
2530
2531         read_lock_bh(&idev->lock);
2532         /* update all temporary addresses in the list */
2533         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2534                 int age, max_valid, max_prefered;
2535
2536                 if (ifp != ift->ifpub)
2537                         continue;
2538
2539                 /* RFC 4941 section 3.3:
2540                  * If a received option will extend the lifetime of a public
2541                  * address, the lifetimes of temporary addresses should
2542                  * be extended, subject to the overall constraint that no
2543                  * temporary addresses should ever remain "valid" or "preferred"
2544                  * for a time longer than (TEMP_VALID_LIFETIME) or
2545                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2546                  */
2547                 age = (now - ift->cstamp) / HZ;
2548                 max_valid = idev->cnf.temp_valid_lft - age;
2549                 if (max_valid < 0)
2550                         max_valid = 0;
2551
2552                 max_prefered = idev->cnf.temp_prefered_lft -
2553                                idev->desync_factor - age;
2554                 if (max_prefered < 0)
2555                         max_prefered = 0;
2556
2557                 if (valid_lft > max_valid)
2558                         valid_lft = max_valid;
2559
2560                 if (prefered_lft > max_prefered)
2561                         prefered_lft = max_prefered;
2562
2563                 spin_lock(&ift->lock);
2564                 flags = ift->flags;
2565                 ift->valid_lft = valid_lft;
2566                 ift->prefered_lft = prefered_lft;
2567                 ift->tstamp = now;
2568                 if (prefered_lft > 0)
2569                         ift->flags &= ~IFA_F_DEPRECATED;
2570
2571                 spin_unlock(&ift->lock);
2572                 if (!(flags&IFA_F_TENTATIVE))
2573                         ipv6_ifa_notify(0, ift);
2574         }
2575
2576         /* Also create a temporary address if it's enabled but no temporary
2577          * address currently exists.
2578          * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2579          * as part of cleanup (ie. deleting the mngtmpaddr).
2580          * We don't want that to result in creating a new temporary ip address.
2581          */
2582         if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2583                 create = true;
2584
2585         if (create && idev->cnf.use_tempaddr > 0) {
2586                 /* When a new public address is created as described
2587                  * in [ADDRCONF], also create a new temporary address.
2588                  */
2589                 read_unlock_bh(&idev->lock);
2590                 ipv6_create_tempaddr(ifp, false);
2591         } else {
2592                 read_unlock_bh(&idev->lock);
2593         }
2594 }
2595
2596 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2597 {
2598         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2599                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2600 }
2601
2602 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2603                                  const struct prefix_info *pinfo,
2604                                  struct inet6_dev *in6_dev,
2605                                  const struct in6_addr *addr, int addr_type,
2606                                  u32 addr_flags, bool sllao, bool tokenized,
2607                                  __u32 valid_lft, u32 prefered_lft)
2608 {
2609         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2610         int create = 0, update_lft = 0;
2611
2612         if (!ifp && valid_lft) {
2613                 int max_addresses = in6_dev->cnf.max_addresses;
2614                 struct ifa6_config cfg = {
2615                         .pfx = addr,
2616                         .plen = pinfo->prefix_len,
2617                         .ifa_flags = addr_flags,
2618                         .valid_lft = valid_lft,
2619                         .preferred_lft = prefered_lft,
2620                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2621                 };
2622
2623 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2624                 if ((net->ipv6.devconf_all->optimistic_dad ||
2625                      in6_dev->cnf.optimistic_dad) &&
2626                     !net->ipv6.devconf_all->forwarding && sllao)
2627                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2628 #endif
2629
2630                 /* Do not allow to create too much of autoconfigured
2631                  * addresses; this would be too easy way to crash kernel.
2632                  */
2633                 if (!max_addresses ||
2634                     ipv6_count_addresses(in6_dev) < max_addresses)
2635                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2636
2637                 if (IS_ERR_OR_NULL(ifp))
2638                         return -1;
2639
2640                 create = 1;
2641                 spin_lock_bh(&ifp->lock);
2642                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2643                 ifp->cstamp = jiffies;
2644                 ifp->tokenized = tokenized;
2645                 spin_unlock_bh(&ifp->lock);
2646                 addrconf_dad_start(ifp);
2647         }
2648
2649         if (ifp) {
2650                 u32 flags;
2651                 unsigned long now;
2652                 u32 stored_lft;
2653
2654                 /* update lifetime (RFC2462 5.5.3 e) */
2655                 spin_lock_bh(&ifp->lock);
2656                 now = jiffies;
2657                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2658                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2659                 else
2660                         stored_lft = 0;
2661                 if (!create && stored_lft) {
2662                         const u32 minimum_lft = min_t(u32,
2663                                 stored_lft, MIN_VALID_LIFETIME);
2664                         valid_lft = max(valid_lft, minimum_lft);
2665
2666                         /* RFC4862 Section 5.5.3e:
2667                          * "Note that the preferred lifetime of the
2668                          *  corresponding address is always reset to
2669                          *  the Preferred Lifetime in the received
2670                          *  Prefix Information option, regardless of
2671                          *  whether the valid lifetime is also reset or
2672                          *  ignored."
2673                          *
2674                          * So we should always update prefered_lft here.
2675                          */
2676                         update_lft = 1;
2677                 }
2678
2679                 if (update_lft) {
2680                         ifp->valid_lft = valid_lft;
2681                         ifp->prefered_lft = prefered_lft;
2682                         ifp->tstamp = now;
2683                         flags = ifp->flags;
2684                         ifp->flags &= ~IFA_F_DEPRECATED;
2685                         spin_unlock_bh(&ifp->lock);
2686
2687                         if (!(flags&IFA_F_TENTATIVE))
2688                                 ipv6_ifa_notify(0, ifp);
2689                 } else
2690                         spin_unlock_bh(&ifp->lock);
2691
2692                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2693                                  create, now);
2694
2695                 in6_ifa_put(ifp);
2696                 addrconf_verify();
2697         }
2698
2699         return 0;
2700 }
2701 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2702
2703 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2704 {
2705         struct prefix_info *pinfo;
2706         __u32 valid_lft;
2707         __u32 prefered_lft;
2708         int addr_type, err;
2709         u32 addr_flags = 0;
2710         struct inet6_dev *in6_dev;
2711         struct net *net = dev_net(dev);
2712
2713         pinfo = (struct prefix_info *) opt;
2714
2715         if (len < sizeof(struct prefix_info)) {
2716                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2717                 return;
2718         }
2719
2720         /*
2721          *      Validation checks ([ADDRCONF], page 19)
2722          */
2723
2724         addr_type = ipv6_addr_type(&pinfo->prefix);
2725
2726         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2727                 return;
2728
2729         valid_lft = ntohl(pinfo->valid);
2730         prefered_lft = ntohl(pinfo->prefered);
2731
2732         if (prefered_lft > valid_lft) {
2733                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2734                 return;
2735         }
2736
2737         in6_dev = in6_dev_get(dev);
2738
2739         if (!in6_dev) {
2740                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2741                                     dev->name);
2742                 return;
2743         }
2744
2745         if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2746                 goto put;
2747
2748         /*
2749          *      Two things going on here:
2750          *      1) Add routes for on-link prefixes
2751          *      2) Configure prefixes with the auto flag set
2752          */
2753
2754         if (pinfo->onlink) {
2755                 struct fib6_info *rt;
2756                 unsigned long rt_expires;
2757
2758                 /* Avoid arithmetic overflow. Really, we could
2759                  * save rt_expires in seconds, likely valid_lft,
2760                  * but it would require division in fib gc, that it
2761                  * not good.
2762                  */
2763                 if (HZ > USER_HZ)
2764                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2765                 else
2766                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2767
2768                 if (addrconf_finite_timeout(rt_expires))
2769                         rt_expires *= HZ;
2770
2771                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2772                                                pinfo->prefix_len,
2773                                                dev,
2774                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2775                                                RTF_DEFAULT, true);
2776
2777                 if (rt) {
2778                         /* Autoconf prefix route */
2779                         if (valid_lft == 0) {
2780                                 ip6_del_rt(net, rt, false);
2781                                 rt = NULL;
2782                         } else if (addrconf_finite_timeout(rt_expires)) {
2783                                 /* not infinity */
2784                                 fib6_set_expires(rt, jiffies + rt_expires);
2785                         } else {
2786                                 fib6_clean_expires(rt);
2787                         }
2788                 } else if (valid_lft) {
2789                         clock_t expires = 0;
2790                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2791                         if (addrconf_finite_timeout(rt_expires)) {
2792                                 /* not infinity */
2793                                 flags |= RTF_EXPIRES;
2794                                 expires = jiffies_to_clock_t(rt_expires);
2795                         }
2796                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2797                                               0, dev, expires, flags,
2798                                               GFP_ATOMIC);
2799                 }
2800                 fib6_info_release(rt);
2801         }
2802
2803         /* Try to figure out our local address for this prefix */
2804
2805         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2806                 struct in6_addr addr;
2807                 bool tokenized = false, dev_addr_generated = false;
2808
2809                 if (pinfo->prefix_len == 64) {
2810                         memcpy(&addr, &pinfo->prefix, 8);
2811
2812                         if (!ipv6_addr_any(&in6_dev->token)) {
2813                                 read_lock_bh(&in6_dev->lock);
2814                                 memcpy(addr.s6_addr + 8,
2815                                        in6_dev->token.s6_addr + 8, 8);
2816                                 read_unlock_bh(&in6_dev->lock);
2817                                 tokenized = true;
2818                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2819                                    !ipv6_generate_stable_address(&addr, 0,
2820                                                                  in6_dev)) {
2821                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2822                                 goto ok;
2823                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2824                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2825                                 goto put;
2826                         } else {
2827                                 dev_addr_generated = true;
2828                         }
2829                         goto ok;
2830                 }
2831                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2832                                     pinfo->prefix_len);
2833                 goto put;
2834
2835 ok:
2836                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2837                                                    &addr, addr_type,
2838                                                    addr_flags, sllao,
2839                                                    tokenized, valid_lft,
2840                                                    prefered_lft);
2841                 if (err)
2842                         goto put;
2843
2844                 /* Ignore error case here because previous prefix add addr was
2845                  * successful which will be notified.
2846                  */
2847                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2848                                               addr_type, addr_flags, sllao,
2849                                               tokenized, valid_lft,
2850                                               prefered_lft,
2851                                               dev_addr_generated);
2852         }
2853         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2854 put:
2855         in6_dev_put(in6_dev);
2856 }
2857
2858 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2859                 struct in6_ifreq *ireq)
2860 {
2861         struct ip_tunnel_parm p = { };
2862         int err;
2863
2864         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2865                 return -EADDRNOTAVAIL;
2866
2867         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2868         p.iph.version = 4;
2869         p.iph.ihl = 5;
2870         p.iph.protocol = IPPROTO_IPV6;
2871         p.iph.ttl = 64;
2872
2873         if (!dev->netdev_ops->ndo_tunnel_ctl)
2874                 return -EOPNOTSUPP;
2875         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2876         if (err)
2877                 return err;
2878
2879         dev = __dev_get_by_name(net, p.name);
2880         if (!dev)
2881                 return -ENOBUFS;
2882         return dev_open(dev, NULL);
2883 }
2884
2885 /*
2886  *      Set destination address.
2887  *      Special case for SIT interfaces where we create a new "virtual"
2888  *      device.
2889  */
2890 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2891 {
2892         struct net_device *dev;
2893         struct in6_ifreq ireq;
2894         int err = -ENODEV;
2895
2896         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2897                 return -ENODEV;
2898         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2899                 return -EFAULT;
2900
2901         rtnl_lock();
2902         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2903         if (dev && dev->type == ARPHRD_SIT)
2904                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2905         rtnl_unlock();
2906         return err;
2907 }
2908
2909 static int ipv6_mc_config(struct sock *sk, bool join,
2910                           const struct in6_addr *addr, int ifindex)
2911 {
2912         int ret;
2913
2914         ASSERT_RTNL();
2915
2916         lock_sock(sk);
2917         if (join)
2918                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2919         else
2920                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2921         release_sock(sk);
2922
2923         return ret;
2924 }
2925
2926 /*
2927  *      Manual configuration of address on an interface
2928  */
2929 static int inet6_addr_add(struct net *net, int ifindex,
2930                           struct ifa6_config *cfg,
2931                           struct netlink_ext_ack *extack)
2932 {
2933         struct inet6_ifaddr *ifp;
2934         struct inet6_dev *idev;
2935         struct net_device *dev;
2936         unsigned long timeout;
2937         clock_t expires;
2938         u32 flags;
2939
2940         ASSERT_RTNL();
2941
2942         if (cfg->plen > 128)
2943                 return -EINVAL;
2944
2945         /* check the lifetime */
2946         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2947                 return -EINVAL;
2948
2949         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2950                 return -EINVAL;
2951
2952         dev = __dev_get_by_index(net, ifindex);
2953         if (!dev)
2954                 return -ENODEV;
2955
2956         idev = addrconf_add_dev(dev);
2957         if (IS_ERR(idev))
2958                 return PTR_ERR(idev);
2959
2960         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2961                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2962                                          true, cfg->pfx, ifindex);
2963
2964                 if (ret < 0)
2965                         return ret;
2966         }
2967
2968         cfg->scope = ipv6_addr_scope(cfg->pfx);
2969
2970         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2971         if (addrconf_finite_timeout(timeout)) {
2972                 expires = jiffies_to_clock_t(timeout * HZ);
2973                 cfg->valid_lft = timeout;
2974                 flags = RTF_EXPIRES;
2975         } else {
2976                 expires = 0;
2977                 flags = 0;
2978                 cfg->ifa_flags |= IFA_F_PERMANENT;
2979         }
2980
2981         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2982         if (addrconf_finite_timeout(timeout)) {
2983                 if (timeout == 0)
2984                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2985                 cfg->preferred_lft = timeout;
2986         }
2987
2988         ifp = ipv6_add_addr(idev, cfg, true, extack);
2989         if (!IS_ERR(ifp)) {
2990                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2991                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2992                                               ifp->rt_priority, dev, expires,
2993                                               flags, GFP_KERNEL);
2994                 }
2995
2996                 /* Send a netlink notification if DAD is enabled and
2997                  * optimistic flag is not set
2998                  */
2999                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3000                         ipv6_ifa_notify(0, ifp);
3001                 /*
3002                  * Note that section 3.1 of RFC 4429 indicates
3003                  * that the Optimistic flag should not be set for
3004                  * manually configured addresses
3005                  */
3006                 addrconf_dad_start(ifp);
3007                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3008                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
3009                                          cfg->preferred_lft, true, jiffies);
3010                 in6_ifa_put(ifp);
3011                 addrconf_verify_rtnl();
3012                 return 0;
3013         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3014                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3015                                cfg->pfx, ifindex);
3016         }
3017
3018         return PTR_ERR(ifp);
3019 }
3020
3021 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3022                           const struct in6_addr *pfx, unsigned int plen)
3023 {
3024         struct inet6_ifaddr *ifp;
3025         struct inet6_dev *idev;
3026         struct net_device *dev;
3027
3028         if (plen > 128)
3029                 return -EINVAL;
3030
3031         dev = __dev_get_by_index(net, ifindex);
3032         if (!dev)
3033                 return -ENODEV;
3034
3035         idev = __in6_dev_get(dev);
3036         if (!idev)
3037                 return -ENXIO;
3038
3039         read_lock_bh(&idev->lock);
3040         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3041                 if (ifp->prefix_len == plen &&
3042                     ipv6_addr_equal(pfx, &ifp->addr)) {
3043                         in6_ifa_hold(ifp);
3044                         read_unlock_bh(&idev->lock);
3045
3046                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3047                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3048                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3049                                                  jiffies);
3050                         ipv6_del_addr(ifp);
3051                         addrconf_verify_rtnl();
3052                         if (ipv6_addr_is_multicast(pfx)) {
3053                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3054                                                false, pfx, dev->ifindex);
3055                         }
3056                         return 0;
3057                 }
3058         }
3059         read_unlock_bh(&idev->lock);
3060         return -EADDRNOTAVAIL;
3061 }
3062
3063
3064 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3065 {
3066         struct ifa6_config cfg = {
3067                 .ifa_flags = IFA_F_PERMANENT,
3068                 .preferred_lft = INFINITY_LIFE_TIME,
3069                 .valid_lft = INFINITY_LIFE_TIME,
3070         };
3071         struct in6_ifreq ireq;
3072         int err;
3073
3074         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3075                 return -EPERM;
3076
3077         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3078                 return -EFAULT;
3079
3080         cfg.pfx = &ireq.ifr6_addr;
3081         cfg.plen = ireq.ifr6_prefixlen;
3082
3083         rtnl_lock();
3084         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3085         rtnl_unlock();
3086         return err;
3087 }
3088
3089 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3090 {
3091         struct in6_ifreq ireq;
3092         int err;
3093
3094         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3095                 return -EPERM;
3096
3097         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3098                 return -EFAULT;
3099
3100         rtnl_lock();
3101         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3102                              ireq.ifr6_prefixlen);
3103         rtnl_unlock();
3104         return err;
3105 }
3106
3107 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3108                      int plen, int scope)
3109 {
3110         struct inet6_ifaddr *ifp;
3111         struct ifa6_config cfg = {
3112                 .pfx = addr,
3113                 .plen = plen,
3114                 .ifa_flags = IFA_F_PERMANENT,
3115                 .valid_lft = INFINITY_LIFE_TIME,
3116                 .preferred_lft = INFINITY_LIFE_TIME,
3117                 .scope = scope
3118         };
3119
3120         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3121         if (!IS_ERR(ifp)) {
3122                 spin_lock_bh(&ifp->lock);
3123                 ifp->flags &= ~IFA_F_TENTATIVE;
3124                 spin_unlock_bh(&ifp->lock);
3125                 rt_genid_bump_ipv6(dev_net(idev->dev));
3126                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3127                 in6_ifa_put(ifp);
3128         }
3129 }
3130
3131 #if IS_ENABLED(CONFIG_IPV6_SIT)
3132 static void sit_add_v4_addrs(struct inet6_dev *idev)
3133 {
3134         struct in6_addr addr;
3135         struct net_device *dev;
3136         struct net *net = dev_net(idev->dev);
3137         int scope, plen;
3138         u32 pflags = 0;
3139
3140         ASSERT_RTNL();
3141
3142         memset(&addr, 0, sizeof(struct in6_addr));
3143         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3144
3145         if (idev->dev->flags&IFF_POINTOPOINT) {
3146                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3147                         return;
3148
3149                 addr.s6_addr32[0] = htonl(0xfe800000);
3150                 scope = IFA_LINK;
3151                 plen = 64;
3152         } else {
3153                 scope = IPV6_ADDR_COMPATv4;
3154                 plen = 96;
3155                 pflags |= RTF_NONEXTHOP;
3156         }
3157
3158         if (addr.s6_addr32[3]) {
3159                 add_addr(idev, &addr, plen, scope);
3160                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3161                                       GFP_KERNEL);
3162                 return;
3163         }
3164
3165         for_each_netdev(net, dev) {
3166                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3167                 if (in_dev && (dev->flags & IFF_UP)) {
3168                         struct in_ifaddr *ifa;
3169                         int flag = scope;
3170
3171                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3172                                 addr.s6_addr32[3] = ifa->ifa_local;
3173
3174                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3175                                         continue;
3176                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3177                                         if (idev->dev->flags&IFF_POINTOPOINT)
3178                                                 continue;
3179                                         flag |= IFA_HOST;
3180                                 }
3181
3182                                 add_addr(idev, &addr, plen, flag);
3183                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3184                                                       0, pflags, GFP_KERNEL);
3185                         }
3186                 }
3187         }
3188 }
3189 #endif
3190
3191 static void init_loopback(struct net_device *dev)
3192 {
3193         struct inet6_dev  *idev;
3194
3195         /* ::1 */
3196
3197         ASSERT_RTNL();
3198
3199         idev = ipv6_find_idev(dev);
3200         if (IS_ERR(idev)) {
3201                 pr_debug("%s: add_dev failed\n", __func__);
3202                 return;
3203         }
3204
3205         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3206 }
3207
3208 void addrconf_add_linklocal(struct inet6_dev *idev,
3209                             const struct in6_addr *addr, u32 flags)
3210 {
3211         struct ifa6_config cfg = {
3212                 .pfx = addr,
3213                 .plen = 64,
3214                 .ifa_flags = flags | IFA_F_PERMANENT,
3215                 .valid_lft = INFINITY_LIFE_TIME,
3216                 .preferred_lft = INFINITY_LIFE_TIME,
3217                 .scope = IFA_LINK
3218         };
3219         struct inet6_ifaddr *ifp;
3220
3221 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3222         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3223              idev->cnf.optimistic_dad) &&
3224             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3225                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3226 #endif
3227
3228         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3229         if (!IS_ERR(ifp)) {
3230                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3231                                       0, 0, GFP_ATOMIC);
3232                 addrconf_dad_start(ifp);
3233                 in6_ifa_put(ifp);
3234         }
3235 }
3236 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3237
3238 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3239 {
3240         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3241                 return true;
3242
3243         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3244             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3245                 return true;
3246
3247         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3248             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3249                 return true;
3250
3251         return false;
3252 }
3253
3254 static int ipv6_generate_stable_address(struct in6_addr *address,
3255                                         u8 dad_count,
3256                                         const struct inet6_dev *idev)
3257 {
3258         static DEFINE_SPINLOCK(lock);
3259         static __u32 digest[SHA1_DIGEST_WORDS];
3260         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3261
3262         static union {
3263                 char __data[SHA1_BLOCK_SIZE];
3264                 struct {
3265                         struct in6_addr secret;
3266                         __be32 prefix[2];
3267                         unsigned char hwaddr[MAX_ADDR_LEN];
3268                         u8 dad_count;
3269                 } __packed;
3270         } data;
3271
3272         struct in6_addr secret;
3273         struct in6_addr temp;
3274         struct net *net = dev_net(idev->dev);
3275
3276         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3277
3278         if (idev->cnf.stable_secret.initialized)
3279                 secret = idev->cnf.stable_secret.secret;
3280         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3281                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3282         else
3283                 return -1;
3284
3285 retry:
3286         spin_lock_bh(&lock);
3287
3288         sha1_init(digest);
3289         memset(&data, 0, sizeof(data));
3290         memset(workspace, 0, sizeof(workspace));
3291         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3292         data.prefix[0] = address->s6_addr32[0];
3293         data.prefix[1] = address->s6_addr32[1];
3294         data.secret = secret;
3295         data.dad_count = dad_count;
3296
3297         sha1_transform(digest, data.__data, workspace);
3298
3299         temp = *address;
3300         temp.s6_addr32[2] = (__force __be32)digest[0];
3301         temp.s6_addr32[3] = (__force __be32)digest[1];
3302
3303         spin_unlock_bh(&lock);
3304
3305         if (ipv6_reserved_interfaceid(temp)) {
3306                 dad_count++;
3307                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3308                         return -1;
3309                 goto retry;
3310         }
3311
3312         *address = temp;
3313         return 0;
3314 }
3315
3316 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3317 {
3318         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3319
3320         if (s->initialized)
3321                 return;
3322         s = &idev->cnf.stable_secret;
3323         get_random_bytes(&s->secret, sizeof(s->secret));
3324         s->initialized = true;
3325 }
3326
3327 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3328 {
3329         struct in6_addr addr;
3330
3331         /* no link local addresses on L3 master devices */
3332         if (netif_is_l3_master(idev->dev))
3333                 return;
3334
3335         /* no link local addresses on devices flagged as slaves */
3336         if (idev->dev->flags & IFF_SLAVE)
3337                 return;
3338
3339         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3340
3341         switch (idev->cnf.addr_gen_mode) {
3342         case IN6_ADDR_GEN_MODE_RANDOM:
3343                 ipv6_gen_mode_random_init(idev);
3344                 fallthrough;
3345         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3346                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3347                         addrconf_add_linklocal(idev, &addr,
3348                                                IFA_F_STABLE_PRIVACY);
3349                 else if (prefix_route)
3350                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3351                                               0, 0, GFP_KERNEL);
3352                 break;
3353         case IN6_ADDR_GEN_MODE_EUI64:
3354                 /* addrconf_add_linklocal also adds a prefix_route and we
3355                  * only need to care about prefix routes if ipv6_generate_eui64
3356                  * couldn't generate one.
3357                  */
3358                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3359                         addrconf_add_linklocal(idev, &addr, 0);
3360                 else if (prefix_route)
3361                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3362                                               0, 0, GFP_KERNEL);
3363                 break;
3364         case IN6_ADDR_GEN_MODE_NONE:
3365         default:
3366                 /* will not add any link local address */
3367                 break;
3368         }
3369 }
3370
3371 static void addrconf_dev_config(struct net_device *dev)
3372 {
3373         struct inet6_dev *idev;
3374
3375         ASSERT_RTNL();
3376
3377         if ((dev->type != ARPHRD_ETHER) &&
3378             (dev->type != ARPHRD_FDDI) &&
3379             (dev->type != ARPHRD_ARCNET) &&
3380             (dev->type != ARPHRD_INFINIBAND) &&
3381             (dev->type != ARPHRD_IEEE1394) &&
3382             (dev->type != ARPHRD_TUNNEL6) &&
3383             (dev->type != ARPHRD_6LOWPAN) &&
3384             (dev->type != ARPHRD_IP6GRE) &&
3385             (dev->type != ARPHRD_IPGRE) &&
3386             (dev->type != ARPHRD_TUNNEL) &&
3387             (dev->type != ARPHRD_NONE) &&
3388             (dev->type != ARPHRD_RAWIP)) {
3389                 /* Alas, we support only Ethernet autoconfiguration. */
3390                 idev = __in6_dev_get(dev);
3391                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3392                     dev->flags & IFF_MULTICAST)
3393                         ipv6_mc_up(idev);
3394                 return;
3395         }
3396
3397         idev = addrconf_add_dev(dev);
3398         if (IS_ERR(idev))
3399                 return;
3400
3401         /* this device type has no EUI support */
3402         if (dev->type == ARPHRD_NONE &&
3403             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3404                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3405
3406         addrconf_addr_gen(idev, false);
3407 }
3408
3409 #if IS_ENABLED(CONFIG_IPV6_SIT)
3410 static void addrconf_sit_config(struct net_device *dev)
3411 {
3412         struct inet6_dev *idev;
3413
3414         ASSERT_RTNL();
3415
3416         /*
3417          * Configure the tunnel with one of our IPv4
3418          * addresses... we should configure all of
3419          * our v4 addrs in the tunnel
3420          */
3421
3422         idev = ipv6_find_idev(dev);
3423         if (IS_ERR(idev)) {
3424                 pr_debug("%s: add_dev failed\n", __func__);
3425                 return;
3426         }
3427
3428         if (dev->priv_flags & IFF_ISATAP) {
3429                 addrconf_addr_gen(idev, false);
3430                 return;
3431         }
3432
3433         sit_add_v4_addrs(idev);
3434
3435         if (dev->flags&IFF_POINTOPOINT)
3436                 addrconf_add_mroute(dev);
3437 }
3438 #endif
3439
3440 #if IS_ENABLED(CONFIG_NET_IPGRE)
3441 static void addrconf_gre_config(struct net_device *dev)
3442 {
3443         struct inet6_dev *idev;
3444
3445         ASSERT_RTNL();
3446
3447         idev = ipv6_find_idev(dev);
3448         if (IS_ERR(idev)) {
3449                 pr_debug("%s: add_dev failed\n", __func__);
3450                 return;
3451         }
3452
3453         addrconf_addr_gen(idev, true);
3454         if (dev->flags & IFF_POINTOPOINT)
3455                 addrconf_add_mroute(dev);
3456 }
3457 #endif
3458
3459 static int fixup_permanent_addr(struct net *net,
3460                                 struct inet6_dev *idev,
3461                                 struct inet6_ifaddr *ifp)
3462 {
3463         /* !fib6_node means the host route was removed from the
3464          * FIB, for example, if 'lo' device is taken down. In that
3465          * case regenerate the host route.
3466          */
3467         if (!ifp->rt || !ifp->rt->fib6_node) {
3468                 struct fib6_info *f6i, *prev;
3469
3470                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3471                                          GFP_ATOMIC);
3472                 if (IS_ERR(f6i))
3473                         return PTR_ERR(f6i);
3474
3475                 /* ifp->rt can be accessed outside of rtnl */
3476                 spin_lock(&ifp->lock);
3477                 prev = ifp->rt;
3478                 ifp->rt = f6i;
3479                 spin_unlock(&ifp->lock);
3480
3481                 fib6_info_release(prev);
3482         }
3483
3484         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3485                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3486                                       ifp->rt_priority, idev->dev, 0, 0,
3487                                       GFP_ATOMIC);
3488         }
3489
3490         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3491                 addrconf_dad_start(ifp);
3492
3493         return 0;
3494 }
3495
3496 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3497 {
3498         struct inet6_ifaddr *ifp, *tmp;
3499         struct inet6_dev *idev;
3500
3501         idev = __in6_dev_get(dev);
3502         if (!idev)
3503                 return;
3504
3505         write_lock_bh(&idev->lock);
3506
3507         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3508                 if ((ifp->flags & IFA_F_PERMANENT) &&
3509                     fixup_permanent_addr(net, idev, ifp) < 0) {
3510                         write_unlock_bh(&idev->lock);
3511                         in6_ifa_hold(ifp);
3512                         ipv6_del_addr(ifp);
3513                         write_lock_bh(&idev->lock);
3514
3515                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3516                                              idev->dev->name, &ifp->addr);
3517                 }
3518         }
3519
3520         write_unlock_bh(&idev->lock);
3521 }
3522
3523 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3524                            void *ptr)
3525 {
3526         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3527         struct netdev_notifier_change_info *change_info;
3528         struct netdev_notifier_changeupper_info *info;
3529         struct inet6_dev *idev = __in6_dev_get(dev);
3530         struct net *net = dev_net(dev);
3531         int run_pending = 0;
3532         int err;
3533
3534         switch (event) {
3535         case NETDEV_REGISTER:
3536                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3537                         idev = ipv6_add_dev(dev);
3538                         if (IS_ERR(idev))
3539                                 return notifier_from_errno(PTR_ERR(idev));
3540                 }
3541                 break;
3542
3543         case NETDEV_CHANGEMTU:
3544                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3545                 if (dev->mtu < IPV6_MIN_MTU) {
3546                         addrconf_ifdown(dev, dev != net->loopback_dev);
3547                         break;
3548                 }
3549
3550                 if (idev) {
3551                         rt6_mtu_change(dev, dev->mtu);
3552                         idev->cnf.mtu6 = dev->mtu;
3553                         break;
3554                 }
3555
3556                 /* allocate new idev */
3557                 idev = ipv6_add_dev(dev);
3558                 if (IS_ERR(idev))
3559                         break;
3560
3561                 /* device is still not ready */
3562                 if (!(idev->if_flags & IF_READY))
3563                         break;
3564
3565                 run_pending = 1;
3566                 fallthrough;
3567         case NETDEV_UP:
3568         case NETDEV_CHANGE:
3569                 if (dev->flags & IFF_SLAVE)
3570                         break;
3571
3572                 if (idev && idev->cnf.disable_ipv6)
3573                         break;
3574
3575                 if (event == NETDEV_UP) {
3576                         /* restore routes for permanent addresses */
3577                         addrconf_permanent_addr(net, dev);
3578
3579                         if (!addrconf_link_ready(dev)) {
3580                                 /* device is not ready yet. */
3581                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3582                                          dev->name);
3583                                 break;
3584                         }
3585
3586                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3587                                 idev = ipv6_add_dev(dev);
3588
3589                         if (!IS_ERR_OR_NULL(idev)) {
3590                                 idev->if_flags |= IF_READY;
3591                                 run_pending = 1;
3592                         }
3593                 } else if (event == NETDEV_CHANGE) {
3594                         if (!addrconf_link_ready(dev)) {
3595                                 /* device is still not ready. */
3596                                 rt6_sync_down_dev(dev, event);
3597                                 break;
3598                         }
3599
3600                         if (!IS_ERR_OR_NULL(idev)) {
3601                                 if (idev->if_flags & IF_READY) {
3602                                         /* device is already configured -
3603                                          * but resend MLD reports, we might
3604                                          * have roamed and need to update
3605                                          * multicast snooping switches
3606                                          */
3607                                         ipv6_mc_up(idev);
3608                                         change_info = ptr;
3609                                         if (change_info->flags_changed & IFF_NOARP)
3610                                                 addrconf_dad_run(idev, true);
3611                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3612                                         break;
3613                                 }
3614                                 idev->if_flags |= IF_READY;
3615                         }
3616
3617                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3618                                 dev->name);
3619
3620                         run_pending = 1;
3621                 }
3622
3623                 switch (dev->type) {
3624 #if IS_ENABLED(CONFIG_IPV6_SIT)
3625                 case ARPHRD_SIT:
3626                         addrconf_sit_config(dev);
3627                         break;
3628 #endif
3629 #if IS_ENABLED(CONFIG_NET_IPGRE)
3630                 case ARPHRD_IPGRE:
3631                         addrconf_gre_config(dev);
3632                         break;
3633 #endif
3634                 case ARPHRD_LOOPBACK:
3635                         init_loopback(dev);
3636                         break;
3637
3638                 default:
3639                         addrconf_dev_config(dev);
3640                         break;
3641                 }
3642
3643                 if (!IS_ERR_OR_NULL(idev)) {
3644                         if (run_pending)
3645                                 addrconf_dad_run(idev, false);
3646
3647                         /* Device has an address by now */
3648                         rt6_sync_up(dev, RTNH_F_DEAD);
3649
3650                         /*
3651                          * If the MTU changed during the interface down,
3652                          * when the interface up, the changed MTU must be
3653                          * reflected in the idev as well as routers.
3654                          */
3655                         if (idev->cnf.mtu6 != dev->mtu &&
3656                             dev->mtu >= IPV6_MIN_MTU) {
3657                                 rt6_mtu_change(dev, dev->mtu);
3658                                 idev->cnf.mtu6 = dev->mtu;
3659                         }
3660                         idev->tstamp = jiffies;
3661                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3662
3663                         /*
3664                          * If the changed mtu during down is lower than
3665                          * IPV6_MIN_MTU stop IPv6 on this interface.
3666                          */
3667                         if (dev->mtu < IPV6_MIN_MTU)
3668                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3669                 }
3670                 break;
3671
3672         case NETDEV_DOWN:
3673         case NETDEV_UNREGISTER:
3674                 /*
3675                  *      Remove all addresses from this interface.
3676                  */
3677                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3678                 break;
3679
3680         case NETDEV_CHANGENAME:
3681                 if (idev) {
3682                         snmp6_unregister_dev(idev);
3683                         addrconf_sysctl_unregister(idev);
3684                         err = addrconf_sysctl_register(idev);
3685                         if (err)
3686                                 return notifier_from_errno(err);
3687                         err = snmp6_register_dev(idev);
3688                         if (err) {
3689                                 addrconf_sysctl_unregister(idev);
3690                                 return notifier_from_errno(err);
3691                         }
3692                 }
3693                 break;
3694
3695         case NETDEV_PRE_TYPE_CHANGE:
3696         case NETDEV_POST_TYPE_CHANGE:
3697                 if (idev)
3698                         addrconf_type_change(dev, event);
3699                 break;
3700
3701         case NETDEV_CHANGEUPPER:
3702                 info = ptr;
3703
3704                 /* flush all routes if dev is linked to or unlinked from
3705                  * an L3 master device (e.g., VRF)
3706                  */
3707                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3708                         addrconf_ifdown(dev, false);
3709         }
3710
3711         return NOTIFY_OK;
3712 }
3713
3714 /*
3715  *      addrconf module should be notified of a device going up
3716  */
3717 static struct notifier_block ipv6_dev_notf = {
3718         .notifier_call = addrconf_notify,
3719         .priority = ADDRCONF_NOTIFY_PRIORITY,
3720 };
3721
3722 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3723 {
3724         struct inet6_dev *idev;
3725         ASSERT_RTNL();
3726
3727         idev = __in6_dev_get(dev);
3728
3729         if (event == NETDEV_POST_TYPE_CHANGE)
3730                 ipv6_mc_remap(idev);
3731         else if (event == NETDEV_PRE_TYPE_CHANGE)
3732                 ipv6_mc_unmap(idev);
3733 }
3734
3735 static bool addr_is_local(const struct in6_addr *addr)
3736 {
3737         return ipv6_addr_type(addr) &
3738                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3739 }
3740
3741 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3742 {
3743         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3744         struct net *net = dev_net(dev);
3745         struct inet6_dev *idev;
3746         struct inet6_ifaddr *ifa;
3747         LIST_HEAD(tmp_addr_list);
3748         bool keep_addr = false;
3749         bool was_ready;
3750         int state, i;
3751
3752         ASSERT_RTNL();
3753
3754         rt6_disable_ip(dev, event);
3755
3756         idev = __in6_dev_get(dev);
3757         if (!idev)
3758                 return -ENODEV;
3759
3760         /*
3761          * Step 1: remove reference to ipv6 device from parent device.
3762          *         Do not dev_put!
3763          */
3764         if (unregister) {
3765                 idev->dead = 1;
3766
3767                 /* protected by rtnl_lock */
3768                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3769
3770                 /* Step 1.5: remove snmp6 entry */
3771                 snmp6_unregister_dev(idev);
3772
3773         }
3774
3775         /* combine the user config with event to determine if permanent
3776          * addresses are to be removed from address hash table
3777          */
3778         if (!unregister && !idev->cnf.disable_ipv6) {
3779                 /* aggregate the system setting and interface setting */
3780                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3781
3782                 if (!_keep_addr)
3783                         _keep_addr = idev->cnf.keep_addr_on_down;
3784
3785                 keep_addr = (_keep_addr > 0);
3786         }
3787
3788         /* Step 2: clear hash table */
3789         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3790                 struct hlist_head *h = &inet6_addr_lst[i];
3791
3792                 spin_lock_bh(&addrconf_hash_lock);
3793 restart:
3794                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3795                         if (ifa->idev == idev) {
3796                                 addrconf_del_dad_work(ifa);
3797                                 /* combined flag + permanent flag decide if
3798                                  * address is retained on a down event
3799                                  */
3800                                 if (!keep_addr ||
3801                                     !(ifa->flags & IFA_F_PERMANENT) ||
3802                                     addr_is_local(&ifa->addr)) {
3803                                         hlist_del_init_rcu(&ifa->addr_lst);
3804                                         goto restart;
3805                                 }
3806                         }
3807                 }
3808                 spin_unlock_bh(&addrconf_hash_lock);
3809         }
3810
3811         write_lock_bh(&idev->lock);
3812
3813         addrconf_del_rs_timer(idev);
3814
3815         /* Step 2: clear flags for stateless addrconf, repeated down
3816          *         detection
3817          */
3818         was_ready = idev->if_flags & IF_READY;
3819         if (!unregister)
3820                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3821
3822         /* Step 3: clear tempaddr list */
3823         while (!list_empty(&idev->tempaddr_list)) {
3824                 ifa = list_first_entry(&idev->tempaddr_list,
3825                                        struct inet6_ifaddr, tmp_list);
3826                 list_del(&ifa->tmp_list);
3827                 write_unlock_bh(&idev->lock);
3828                 spin_lock_bh(&ifa->lock);
3829
3830                 if (ifa->ifpub) {
3831                         in6_ifa_put(ifa->ifpub);
3832                         ifa->ifpub = NULL;
3833                 }
3834                 spin_unlock_bh(&ifa->lock);
3835                 in6_ifa_put(ifa);
3836                 write_lock_bh(&idev->lock);
3837         }
3838
3839         list_for_each_entry(ifa, &idev->addr_list, if_list)
3840                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3841         write_unlock_bh(&idev->lock);
3842
3843         while (!list_empty(&tmp_addr_list)) {
3844                 struct fib6_info *rt = NULL;
3845                 bool keep;
3846
3847                 ifa = list_first_entry(&tmp_addr_list,
3848                                        struct inet6_ifaddr, if_list_aux);
3849                 list_del(&ifa->if_list_aux);
3850
3851                 addrconf_del_dad_work(ifa);
3852
3853                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3854                         !addr_is_local(&ifa->addr);
3855
3856                 spin_lock_bh(&ifa->lock);
3857
3858                 if (keep) {
3859                         /* set state to skip the notifier below */
3860                         state = INET6_IFADDR_STATE_DEAD;
3861                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3862                         if (!(ifa->flags & IFA_F_NODAD))
3863                                 ifa->flags |= IFA_F_TENTATIVE;
3864
3865                         rt = ifa->rt;
3866                         ifa->rt = NULL;
3867                 } else {
3868                         state = ifa->state;
3869                         ifa->state = INET6_IFADDR_STATE_DEAD;
3870                 }
3871
3872                 spin_unlock_bh(&ifa->lock);
3873
3874                 if (rt)
3875                         ip6_del_rt(net, rt, false);
3876
3877                 if (state != INET6_IFADDR_STATE_DEAD) {
3878                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3879                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3880                 } else {
3881                         if (idev->cnf.forwarding)
3882                                 addrconf_leave_anycast(ifa);
3883                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3884                 }
3885
3886                 if (!keep) {
3887                         write_lock_bh(&idev->lock);
3888                         list_del_rcu(&ifa->if_list);
3889                         write_unlock_bh(&idev->lock);
3890                         in6_ifa_put(ifa);
3891                 }
3892         }
3893
3894         /* Step 5: Discard anycast and multicast list */
3895         if (unregister) {
3896                 ipv6_ac_destroy_dev(idev);
3897                 ipv6_mc_destroy_dev(idev);
3898         } else if (was_ready) {
3899                 ipv6_mc_down(idev);
3900         }
3901
3902         idev->tstamp = jiffies;
3903
3904         /* Last: Shot the device (if unregistered) */
3905         if (unregister) {
3906                 addrconf_sysctl_unregister(idev);
3907                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3908                 neigh_ifdown(&nd_tbl, dev);
3909                 in6_dev_put(idev);
3910         }
3911         return 0;
3912 }
3913
3914 static void addrconf_rs_timer(struct timer_list *t)
3915 {
3916         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3917         struct net_device *dev = idev->dev;
3918         struct in6_addr lladdr;
3919
3920         write_lock(&idev->lock);
3921         if (idev->dead || !(idev->if_flags & IF_READY))
3922                 goto out;
3923
3924         if (!ipv6_accept_ra(idev))
3925                 goto out;
3926
3927         /* Announcement received after solicitation was sent */
3928         if (idev->if_flags & IF_RA_RCVD)
3929                 goto out;
3930
3931         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3932                 write_unlock(&idev->lock);
3933                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3934                         ndisc_send_rs(dev, &lladdr,
3935                                       &in6addr_linklocal_allrouters);
3936                 else
3937                         goto put;
3938
3939                 write_lock(&idev->lock);
3940                 idev->rs_interval = rfc3315_s14_backoff_update(
3941                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3942                 /* The wait after the last probe can be shorter */
3943                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3944                                              idev->cnf.rtr_solicits) ?
3945                                       idev->cnf.rtr_solicit_delay :
3946                                       idev->rs_interval);
3947         } else {
3948                 /*
3949                  * Note: we do not support deprecated "all on-link"
3950                  * assumption any longer.
3951                  */
3952                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3953         }
3954
3955 out:
3956         write_unlock(&idev->lock);
3957 put:
3958         in6_dev_put(idev);
3959 }
3960
3961 /*
3962  *      Duplicate Address Detection
3963  */
3964 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3965 {
3966         unsigned long rand_num;
3967         struct inet6_dev *idev = ifp->idev;
3968         u64 nonce;
3969
3970         if (ifp->flags & IFA_F_OPTIMISTIC)
3971                 rand_num = 0;
3972         else
3973                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3974
3975         nonce = 0;
3976         if (idev->cnf.enhanced_dad ||
3977             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3978                 do
3979                         get_random_bytes(&nonce, 6);
3980                 while (nonce == 0);
3981         }
3982         ifp->dad_nonce = nonce;
3983         ifp->dad_probes = idev->cnf.dad_transmits;
3984         addrconf_mod_dad_work(ifp, rand_num);
3985 }
3986
3987 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3988 {
3989         struct inet6_dev *idev = ifp->idev;
3990         struct net_device *dev = idev->dev;
3991         bool bump_id, notify = false;
3992         struct net *net;
3993
3994         addrconf_join_solict(dev, &ifp->addr);
3995
3996         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3997
3998         read_lock_bh(&idev->lock);
3999         spin_lock(&ifp->lock);
4000         if (ifp->state == INET6_IFADDR_STATE_DEAD)
4001                 goto out;
4002
4003         net = dev_net(dev);
4004         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4005             (net->ipv6.devconf_all->accept_dad < 1 &&
4006              idev->cnf.accept_dad < 1) ||
4007             !(ifp->flags&IFA_F_TENTATIVE) ||
4008             ifp->flags & IFA_F_NODAD) {
4009                 bool send_na = false;
4010
4011                 if (ifp->flags & IFA_F_TENTATIVE &&
4012                     !(ifp->flags & IFA_F_OPTIMISTIC))
4013                         send_na = true;
4014                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4015                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4016                 spin_unlock(&ifp->lock);
4017                 read_unlock_bh(&idev->lock);
4018
4019                 addrconf_dad_completed(ifp, bump_id, send_na);
4020                 return;
4021         }
4022
4023         if (!(idev->if_flags & IF_READY)) {
4024                 spin_unlock(&ifp->lock);
4025                 read_unlock_bh(&idev->lock);
4026                 /*
4027                  * If the device is not ready:
4028                  * - keep it tentative if it is a permanent address.
4029                  * - otherwise, kill it.
4030                  */
4031                 in6_ifa_hold(ifp);
4032                 addrconf_dad_stop(ifp, 0);
4033                 return;
4034         }
4035
4036         /*
4037          * Optimistic nodes can start receiving
4038          * Frames right away
4039          */
4040         if (ifp->flags & IFA_F_OPTIMISTIC) {
4041                 ip6_ins_rt(net, ifp->rt);
4042                 if (ipv6_use_optimistic_addr(net, idev)) {
4043                         /* Because optimistic nodes can use this address,
4044                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4045                          */
4046                         notify = true;
4047                 }
4048         }
4049
4050         addrconf_dad_kick(ifp);
4051 out:
4052         spin_unlock(&ifp->lock);
4053         read_unlock_bh(&idev->lock);
4054         if (notify)
4055                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4056 }
4057
4058 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4059 {
4060         bool begin_dad = false;
4061
4062         spin_lock_bh(&ifp->lock);
4063         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4064                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4065                 begin_dad = true;
4066         }
4067         spin_unlock_bh(&ifp->lock);
4068
4069         if (begin_dad)
4070                 addrconf_mod_dad_work(ifp, 0);
4071 }
4072
4073 static void addrconf_dad_work(struct work_struct *w)
4074 {
4075         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4076                                                 struct inet6_ifaddr,
4077                                                 dad_work);
4078         struct inet6_dev *idev = ifp->idev;
4079         bool bump_id, disable_ipv6 = false;
4080         struct in6_addr mcaddr;
4081
4082         enum {
4083                 DAD_PROCESS,
4084                 DAD_BEGIN,
4085                 DAD_ABORT,
4086         } action = DAD_PROCESS;
4087
4088         rtnl_lock();
4089
4090         spin_lock_bh(&ifp->lock);
4091         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4092                 action = DAD_BEGIN;
4093                 ifp->state = INET6_IFADDR_STATE_DAD;
4094         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4095                 action = DAD_ABORT;
4096                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4097
4098                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4099                      idev->cnf.accept_dad > 1) &&
4100                     !idev->cnf.disable_ipv6 &&
4101                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4102                         struct in6_addr addr;
4103
4104                         addr.s6_addr32[0] = htonl(0xfe800000);
4105                         addr.s6_addr32[1] = 0;
4106
4107                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4108                             ipv6_addr_equal(&ifp->addr, &addr)) {
4109                                 /* DAD failed for link-local based on MAC */
4110                                 idev->cnf.disable_ipv6 = 1;
4111
4112                                 pr_info("%s: IPv6 being disabled!\n",
4113                                         ifp->idev->dev->name);
4114                                 disable_ipv6 = true;
4115                         }
4116                 }
4117         }
4118         spin_unlock_bh(&ifp->lock);
4119
4120         if (action == DAD_BEGIN) {
4121                 addrconf_dad_begin(ifp);
4122                 goto out;
4123         } else if (action == DAD_ABORT) {
4124                 in6_ifa_hold(ifp);
4125                 addrconf_dad_stop(ifp, 1);
4126                 if (disable_ipv6)
4127                         addrconf_ifdown(idev->dev, false);
4128                 goto out;
4129         }
4130
4131         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4132                 goto out;
4133
4134         write_lock_bh(&idev->lock);
4135         if (idev->dead || !(idev->if_flags & IF_READY)) {
4136                 write_unlock_bh(&idev->lock);
4137                 goto out;
4138         }
4139
4140         spin_lock(&ifp->lock);
4141         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4142                 spin_unlock(&ifp->lock);
4143                 write_unlock_bh(&idev->lock);
4144                 goto out;
4145         }
4146
4147         if (ifp->dad_probes == 0) {
4148                 bool send_na = false;
4149
4150                 /*
4151                  * DAD was successful
4152                  */
4153
4154                 if (ifp->flags & IFA_F_TENTATIVE &&
4155                     !(ifp->flags & IFA_F_OPTIMISTIC))
4156                         send_na = true;
4157                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4158                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4159                 spin_unlock(&ifp->lock);
4160                 write_unlock_bh(&idev->lock);
4161
4162                 addrconf_dad_completed(ifp, bump_id, send_na);
4163
4164                 goto out;
4165         }
4166
4167         ifp->dad_probes--;
4168         addrconf_mod_dad_work(ifp,
4169                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4170                                   HZ/100));
4171         spin_unlock(&ifp->lock);
4172         write_unlock_bh(&idev->lock);
4173
4174         /* send a neighbour solicitation for our addr */
4175         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4176         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4177                       ifp->dad_nonce);
4178 out:
4179         in6_ifa_put(ifp);
4180         rtnl_unlock();
4181 }
4182
4183 /* ifp->idev must be at least read locked */
4184 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4185 {
4186         struct inet6_ifaddr *ifpiter;
4187         struct inet6_dev *idev = ifp->idev;
4188
4189         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4190                 if (ifpiter->scope > IFA_LINK)
4191                         break;
4192                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4193                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4194                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4195                     IFA_F_PERMANENT)
4196                         return false;
4197         }
4198         return true;
4199 }
4200
4201 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4202                                    bool send_na)
4203 {
4204         struct net_device *dev = ifp->idev->dev;
4205         struct in6_addr lladdr;
4206         bool send_rs, send_mld;
4207
4208         addrconf_del_dad_work(ifp);
4209
4210         /*
4211          *      Configure the address for reception. Now it is valid.
4212          */
4213
4214         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4215
4216         /* If added prefix is link local and we are prepared to process
4217            router advertisements, start sending router solicitations.
4218          */
4219
4220         read_lock_bh(&ifp->idev->lock);
4221         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4222         send_rs = send_mld &&
4223                   ipv6_accept_ra(ifp->idev) &&
4224                   ifp->idev->cnf.rtr_solicits != 0 &&
4225                   (dev->flags & IFF_LOOPBACK) == 0 &&
4226                   (dev->type != ARPHRD_TUNNEL);
4227         read_unlock_bh(&ifp->idev->lock);
4228
4229         /* While dad is in progress mld report's source address is in6_addrany.
4230          * Resend with proper ll now.
4231          */
4232         if (send_mld)
4233                 ipv6_mc_dad_complete(ifp->idev);
4234
4235         /* send unsolicited NA if enabled */
4236         if (send_na &&
4237             (ifp->idev->cnf.ndisc_notify ||
4238              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4239                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4240                               /*router=*/ !!ifp->idev->cnf.forwarding,
4241                               /*solicited=*/ false, /*override=*/ true,
4242                               /*inc_opt=*/ true);
4243         }
4244
4245         if (send_rs) {
4246                 /*
4247                  *      If a host as already performed a random delay
4248                  *      [...] as part of DAD [...] there is no need
4249                  *      to delay again before sending the first RS
4250                  */
4251                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4252                         return;
4253                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4254
4255                 write_lock_bh(&ifp->idev->lock);
4256                 spin_lock(&ifp->lock);
4257                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4258                         ifp->idev->cnf.rtr_solicit_interval);
4259                 ifp->idev->rs_probes = 1;
4260                 ifp->idev->if_flags |= IF_RS_SENT;
4261                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4262                 spin_unlock(&ifp->lock);
4263                 write_unlock_bh(&ifp->idev->lock);
4264         }
4265
4266         if (bump_id)
4267                 rt_genid_bump_ipv6(dev_net(dev));
4268
4269         /* Make sure that a new temporary address will be created
4270          * before this temporary address becomes deprecated.
4271          */
4272         if (ifp->flags & IFA_F_TEMPORARY)
4273                 addrconf_verify_rtnl();
4274 }
4275
4276 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4277 {
4278         struct inet6_ifaddr *ifp;
4279
4280         read_lock_bh(&idev->lock);
4281         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4282                 spin_lock(&ifp->lock);
4283                 if ((ifp->flags & IFA_F_TENTATIVE &&
4284                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4285                         if (restart)
4286                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4287                         addrconf_dad_kick(ifp);
4288                 }
4289                 spin_unlock(&ifp->lock);
4290         }
4291         read_unlock_bh(&idev->lock);
4292 }
4293
4294 #ifdef CONFIG_PROC_FS
4295 struct if6_iter_state {
4296         struct seq_net_private p;
4297         int bucket;
4298         int offset;
4299 };
4300
4301 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4302 {
4303         struct if6_iter_state *state = seq->private;
4304         struct net *net = seq_file_net(seq);
4305         struct inet6_ifaddr *ifa = NULL;
4306         int p = 0;
4307
4308         /* initial bucket if pos is 0 */
4309         if (pos == 0) {
4310                 state->bucket = 0;
4311                 state->offset = 0;
4312         }
4313
4314         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4315                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4316                                          addr_lst) {
4317                         if (!net_eq(dev_net(ifa->idev->dev), net))
4318                                 continue;
4319                         /* sync with offset */
4320                         if (p < state->offset) {
4321                                 p++;
4322                                 continue;
4323                         }
4324                         return ifa;
4325                 }
4326
4327                 /* prepare for next bucket */
4328                 state->offset = 0;
4329                 p = 0;
4330         }
4331         return NULL;
4332 }
4333
4334 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4335                                          struct inet6_ifaddr *ifa)
4336 {
4337         struct if6_iter_state *state = seq->private;
4338         struct net *net = seq_file_net(seq);
4339
4340         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4341                 if (!net_eq(dev_net(ifa->idev->dev), net))
4342                         continue;
4343                 state->offset++;
4344                 return ifa;
4345         }
4346
4347         state->offset = 0;
4348         while (++state->bucket < IN6_ADDR_HSIZE) {
4349                 hlist_for_each_entry_rcu(ifa,
4350                                      &inet6_addr_lst[state->bucket], addr_lst) {
4351                         if (!net_eq(dev_net(ifa->idev->dev), net))
4352                                 continue;
4353                         return ifa;
4354                 }
4355         }
4356
4357         return NULL;
4358 }
4359
4360 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4361         __acquires(rcu)
4362 {
4363         rcu_read_lock();
4364         return if6_get_first(seq, *pos);
4365 }
4366
4367 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4368 {
4369         struct inet6_ifaddr *ifa;
4370
4371         ifa = if6_get_next(seq, v);
4372         ++*pos;
4373         return ifa;
4374 }
4375
4376 static void if6_seq_stop(struct seq_file *seq, void *v)
4377         __releases(rcu)
4378 {
4379         rcu_read_unlock();
4380 }
4381
4382 static int if6_seq_show(struct seq_file *seq, void *v)
4383 {
4384         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4385         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4386                    &ifp->addr,
4387                    ifp->idev->dev->ifindex,
4388                    ifp->prefix_len,
4389                    ifp->scope,
4390                    (u8) ifp->flags,
4391                    ifp->idev->dev->name);
4392         return 0;
4393 }
4394
4395 static const struct seq_operations if6_seq_ops = {
4396         .start  = if6_seq_start,
4397         .next   = if6_seq_next,
4398         .show   = if6_seq_show,
4399         .stop   = if6_seq_stop,
4400 };
4401
4402 static int __net_init if6_proc_net_init(struct net *net)
4403 {
4404         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4405                         sizeof(struct if6_iter_state)))
4406                 return -ENOMEM;
4407         return 0;
4408 }
4409
4410 static void __net_exit if6_proc_net_exit(struct net *net)
4411 {
4412         remove_proc_entry("if_inet6", net->proc_net);
4413 }
4414
4415 static struct pernet_operations if6_proc_net_ops = {
4416         .init = if6_proc_net_init,
4417         .exit = if6_proc_net_exit,
4418 };
4419
4420 int __init if6_proc_init(void)
4421 {
4422         return register_pernet_subsys(&if6_proc_net_ops);
4423 }
4424
4425 void if6_proc_exit(void)
4426 {
4427         unregister_pernet_subsys(&if6_proc_net_ops);
4428 }
4429 #endif  /* CONFIG_PROC_FS */
4430
4431 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4432 /* Check if address is a home address configured on any interface. */
4433 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4434 {
4435         unsigned int hash = inet6_addr_hash(net, addr);
4436         struct inet6_ifaddr *ifp = NULL;
4437         int ret = 0;
4438
4439         rcu_read_lock();
4440         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4441                 if (!net_eq(dev_net(ifp->idev->dev), net))
4442                         continue;
4443                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4444                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4445                         ret = 1;
4446                         break;
4447                 }
4448         }
4449         rcu_read_unlock();
4450         return ret;
4451 }
4452 #endif
4453
4454 /* RFC6554 has some algorithm to avoid loops in segment routing by
4455  * checking if the segments contains any of a local interface address.
4456  *
4457  * Quote:
4458  *
4459  * To detect loops in the SRH, a router MUST determine if the SRH
4460  * includes multiple addresses assigned to any interface on that router.
4461  * If such addresses appear more than once and are separated by at least
4462  * one address not assigned to that router.
4463  */
4464 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4465                           unsigned char nsegs)
4466 {
4467         const struct in6_addr *addr;
4468         int i, ret = 0, found = 0;
4469         struct inet6_ifaddr *ifp;
4470         bool separated = false;
4471         unsigned int hash;
4472         bool hash_found;
4473
4474         rcu_read_lock();
4475         for (i = 0; i < nsegs; i++) {
4476                 addr = &segs[i];
4477                 hash = inet6_addr_hash(net, addr);
4478
4479                 hash_found = false;
4480                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4481                         if (!net_eq(dev_net(ifp->idev->dev), net))
4482                                 continue;
4483
4484                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4485                                 hash_found = true;
4486                                 break;
4487                         }
4488                 }
4489
4490                 if (hash_found) {
4491                         if (found > 1 && separated) {
4492                                 ret = 1;
4493                                 break;
4494                         }
4495
4496                         separated = false;
4497                         found++;
4498                 } else {
4499                         separated = true;
4500                 }
4501         }
4502         rcu_read_unlock();
4503
4504         return ret;
4505 }
4506
4507 /*
4508  *      Periodic address status verification
4509  */
4510
4511 static void addrconf_verify_rtnl(void)
4512 {
4513         unsigned long now, next, next_sec, next_sched;
4514         struct inet6_ifaddr *ifp;
4515         int i;
4516
4517         ASSERT_RTNL();
4518
4519         rcu_read_lock_bh();
4520         now = jiffies;
4521         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4522
4523         cancel_delayed_work(&addr_chk_work);
4524
4525         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4526 restart:
4527                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4528                         unsigned long age;
4529
4530                         /* When setting preferred_lft to a value not zero or
4531                          * infinity, while valid_lft is infinity
4532                          * IFA_F_PERMANENT has a non-infinity life time.
4533                          */
4534                         if ((ifp->flags & IFA_F_PERMANENT) &&
4535                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4536                                 continue;
4537
4538                         spin_lock(&ifp->lock);
4539                         /* We try to batch several events at once. */
4540                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4541
4542                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4543                             age >= ifp->valid_lft) {
4544                                 spin_unlock(&ifp->lock);
4545                                 in6_ifa_hold(ifp);
4546                                 ipv6_del_addr(ifp);
4547                                 goto restart;
4548                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4549                                 spin_unlock(&ifp->lock);
4550                                 continue;
4551                         } else if (age >= ifp->prefered_lft) {
4552                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4553                                 int deprecate = 0;
4554
4555                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4556                                         deprecate = 1;
4557                                         ifp->flags |= IFA_F_DEPRECATED;
4558                                 }
4559
4560                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4561                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4562                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4563
4564                                 spin_unlock(&ifp->lock);
4565
4566                                 if (deprecate) {
4567                                         in6_ifa_hold(ifp);
4568
4569                                         ipv6_ifa_notify(0, ifp);
4570                                         in6_ifa_put(ifp);
4571                                         goto restart;
4572                                 }
4573                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4574                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4575                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4576                                         ifp->idev->cnf.dad_transmits *
4577                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4578
4579                                 if (age >= ifp->prefered_lft - regen_advance) {
4580                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4581                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4582                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4583                                         if (!ifp->regen_count && ifpub) {
4584                                                 ifp->regen_count++;
4585                                                 in6_ifa_hold(ifp);
4586                                                 in6_ifa_hold(ifpub);
4587                                                 spin_unlock(&ifp->lock);
4588
4589                                                 spin_lock(&ifpub->lock);
4590                                                 ifpub->regen_count = 0;
4591                                                 spin_unlock(&ifpub->lock);
4592                                                 rcu_read_unlock_bh();
4593                                                 ipv6_create_tempaddr(ifpub, true);
4594                                                 in6_ifa_put(ifpub);
4595                                                 in6_ifa_put(ifp);
4596                                                 rcu_read_lock_bh();
4597                                                 goto restart;
4598                                         }
4599                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4600                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4601                                 spin_unlock(&ifp->lock);
4602                         } else {
4603                                 /* ifp->prefered_lft <= ifp->valid_lft */
4604                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4605                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4606                                 spin_unlock(&ifp->lock);
4607                         }
4608                 }
4609         }
4610
4611         next_sec = round_jiffies_up(next);
4612         next_sched = next;
4613
4614         /* If rounded timeout is accurate enough, accept it. */
4615         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4616                 next_sched = next_sec;
4617
4618         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4619         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4620                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4621
4622         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4623                  now, next, next_sec, next_sched);
4624         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4625         rcu_read_unlock_bh();
4626 }
4627
4628 static void addrconf_verify_work(struct work_struct *w)
4629 {
4630         rtnl_lock();
4631         addrconf_verify_rtnl();
4632         rtnl_unlock();
4633 }
4634
4635 static void addrconf_verify(void)
4636 {
4637         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4638 }
4639
4640 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4641                                      struct in6_addr **peer_pfx)
4642 {
4643         struct in6_addr *pfx = NULL;
4644
4645         *peer_pfx = NULL;
4646
4647         if (addr)
4648                 pfx = nla_data(addr);
4649
4650         if (local) {
4651                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4652                         *peer_pfx = pfx;
4653                 pfx = nla_data(local);
4654         }
4655
4656         return pfx;
4657 }
4658
4659 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4660         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4661         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4662         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4663         [IFA_FLAGS]             = { .len = sizeof(u32) },
4664         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4665         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4666 };
4667
4668 static int
4669 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4670                   struct netlink_ext_ack *extack)
4671 {
4672         struct net *net = sock_net(skb->sk);
4673         struct ifaddrmsg *ifm;
4674         struct nlattr *tb[IFA_MAX+1];
4675         struct in6_addr *pfx, *peer_pfx;
4676         u32 ifa_flags;
4677         int err;
4678
4679         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4680                                      ifa_ipv6_policy, extack);
4681         if (err < 0)
4682                 return err;
4683
4684         ifm = nlmsg_data(nlh);
4685         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4686         if (!pfx)
4687                 return -EINVAL;
4688
4689         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4690
4691         /* We ignore other flags so far. */
4692         ifa_flags &= IFA_F_MANAGETEMPADDR;
4693
4694         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4695                               ifm->ifa_prefixlen);
4696 }
4697
4698 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4699                                unsigned long expires, u32 flags,
4700                                bool modify_peer)
4701 {
4702         struct fib6_info *f6i;
4703         u32 prio;
4704
4705         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4706                                         ifp->prefix_len,
4707                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4708         if (!f6i)
4709                 return -ENOENT;
4710
4711         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4712         if (f6i->fib6_metric != prio) {
4713                 /* delete old one */
4714                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4715
4716                 /* add new one */
4717                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4718                                       ifp->prefix_len,
4719                                       ifp->rt_priority, ifp->idev->dev,
4720                                       expires, flags, GFP_KERNEL);
4721         } else {
4722                 if (!expires)
4723                         fib6_clean_expires(f6i);
4724                 else
4725                         fib6_set_expires(f6i, expires);
4726
4727                 fib6_info_release(f6i);
4728         }
4729
4730         return 0;
4731 }
4732
4733 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4734 {
4735         u32 flags;
4736         clock_t expires;
4737         unsigned long timeout;
4738         bool was_managetempaddr;
4739         bool had_prefixroute;
4740         bool new_peer = false;
4741
4742         ASSERT_RTNL();
4743
4744         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4745                 return -EINVAL;
4746
4747         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4748             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4749                 return -EINVAL;
4750
4751         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4752                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4753
4754         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4755         if (addrconf_finite_timeout(timeout)) {
4756                 expires = jiffies_to_clock_t(timeout * HZ);
4757                 cfg->valid_lft = timeout;
4758                 flags = RTF_EXPIRES;
4759         } else {
4760                 expires = 0;
4761                 flags = 0;
4762                 cfg->ifa_flags |= IFA_F_PERMANENT;
4763         }
4764
4765         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4766         if (addrconf_finite_timeout(timeout)) {
4767                 if (timeout == 0)
4768                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4769                 cfg->preferred_lft = timeout;
4770         }
4771
4772         if (cfg->peer_pfx &&
4773             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4774                 if (!ipv6_addr_any(&ifp->peer_addr))
4775                         cleanup_prefix_route(ifp, expires, true, true);
4776                 new_peer = true;
4777         }
4778
4779         spin_lock_bh(&ifp->lock);
4780         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4781         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4782                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4783         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4784                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4785                         IFA_F_NOPREFIXROUTE);
4786         ifp->flags |= cfg->ifa_flags;
4787         ifp->tstamp = jiffies;
4788         ifp->valid_lft = cfg->valid_lft;
4789         ifp->prefered_lft = cfg->preferred_lft;
4790
4791         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4792                 ifp->rt_priority = cfg->rt_priority;
4793
4794         if (new_peer)
4795                 ifp->peer_addr = *cfg->peer_pfx;
4796
4797         spin_unlock_bh(&ifp->lock);
4798         if (!(ifp->flags&IFA_F_TENTATIVE))
4799                 ipv6_ifa_notify(0, ifp);
4800
4801         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4802                 int rc = -ENOENT;
4803
4804                 if (had_prefixroute)
4805                         rc = modify_prefix_route(ifp, expires, flags, false);
4806
4807                 /* prefix route could have been deleted; if so restore it */
4808                 if (rc == -ENOENT) {
4809                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4810                                               ifp->rt_priority, ifp->idev->dev,
4811                                               expires, flags, GFP_KERNEL);
4812                 }
4813
4814                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4815                         rc = modify_prefix_route(ifp, expires, flags, true);
4816
4817                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4818                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4819                                               ifp->rt_priority, ifp->idev->dev,
4820                                               expires, flags, GFP_KERNEL);
4821                 }
4822         } else if (had_prefixroute) {
4823                 enum cleanup_prefix_rt_t action;
4824                 unsigned long rt_expires;
4825
4826                 write_lock_bh(&ifp->idev->lock);
4827                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4828                 write_unlock_bh(&ifp->idev->lock);
4829
4830                 if (action != CLEANUP_PREFIX_RT_NOP) {
4831                         cleanup_prefix_route(ifp, rt_expires,
4832                                 action == CLEANUP_PREFIX_RT_DEL, false);
4833                 }
4834         }
4835
4836         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4837                 if (was_managetempaddr &&
4838                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4839                         cfg->valid_lft = 0;
4840                         cfg->preferred_lft = 0;
4841                 }
4842                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4843                                  cfg->preferred_lft, !was_managetempaddr,
4844                                  jiffies);
4845         }
4846
4847         addrconf_verify_rtnl();
4848
4849         return 0;
4850 }
4851
4852 static int
4853 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4854                   struct netlink_ext_ack *extack)
4855 {
4856         struct net *net = sock_net(skb->sk);
4857         struct ifaddrmsg *ifm;
4858         struct nlattr *tb[IFA_MAX+1];
4859         struct in6_addr *peer_pfx;
4860         struct inet6_ifaddr *ifa;
4861         struct net_device *dev;
4862         struct inet6_dev *idev;
4863         struct ifa6_config cfg;
4864         int err;
4865
4866         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4867                                      ifa_ipv6_policy, extack);
4868         if (err < 0)
4869                 return err;
4870
4871         memset(&cfg, 0, sizeof(cfg));
4872
4873         ifm = nlmsg_data(nlh);
4874         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4875         if (!cfg.pfx)
4876                 return -EINVAL;
4877
4878         cfg.peer_pfx = peer_pfx;
4879         cfg.plen = ifm->ifa_prefixlen;
4880         if (tb[IFA_RT_PRIORITY])
4881                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4882
4883         cfg.valid_lft = INFINITY_LIFE_TIME;
4884         cfg.preferred_lft = INFINITY_LIFE_TIME;
4885
4886         if (tb[IFA_CACHEINFO]) {
4887                 struct ifa_cacheinfo *ci;
4888
4889                 ci = nla_data(tb[IFA_CACHEINFO]);
4890                 cfg.valid_lft = ci->ifa_valid;
4891                 cfg.preferred_lft = ci->ifa_prefered;
4892         }
4893
4894         dev =  __dev_get_by_index(net, ifm->ifa_index);
4895         if (!dev)
4896                 return -ENODEV;
4897
4898         if (tb[IFA_FLAGS])
4899                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4900         else
4901                 cfg.ifa_flags = ifm->ifa_flags;
4902
4903         /* We ignore other flags so far. */
4904         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4905                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4906                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4907
4908         idev = ipv6_find_idev(dev);
4909         if (IS_ERR(idev))
4910                 return PTR_ERR(idev);
4911
4912         if (!ipv6_allow_optimistic_dad(net, idev))
4913                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4914
4915         if (cfg.ifa_flags & IFA_F_NODAD &&
4916             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4917                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4918                 return -EINVAL;
4919         }
4920
4921         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4922         if (!ifa) {
4923                 /*
4924                  * It would be best to check for !NLM_F_CREATE here but
4925                  * userspace already relies on not having to provide this.
4926                  */
4927                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4928         }
4929
4930         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4931             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4932                 err = -EEXIST;
4933         else
4934                 err = inet6_addr_modify(ifa, &cfg);
4935
4936         in6_ifa_put(ifa);
4937
4938         return err;
4939 }
4940
4941 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4942                           u8 scope, int ifindex)
4943 {
4944         struct ifaddrmsg *ifm;
4945
4946         ifm = nlmsg_data(nlh);
4947         ifm->ifa_family = AF_INET6;
4948         ifm->ifa_prefixlen = prefixlen;
4949         ifm->ifa_flags = flags;
4950         ifm->ifa_scope = scope;
4951         ifm->ifa_index = ifindex;
4952 }
4953
4954 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4955                          unsigned long tstamp, u32 preferred, u32 valid)
4956 {
4957         struct ifa_cacheinfo ci;
4958
4959         ci.cstamp = cstamp_delta(cstamp);
4960         ci.tstamp = cstamp_delta(tstamp);
4961         ci.ifa_prefered = preferred;
4962         ci.ifa_valid = valid;
4963
4964         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4965 }
4966
4967 static inline int rt_scope(int ifa_scope)
4968 {
4969         if (ifa_scope & IFA_HOST)
4970                 return RT_SCOPE_HOST;
4971         else if (ifa_scope & IFA_LINK)
4972                 return RT_SCOPE_LINK;
4973         else if (ifa_scope & IFA_SITE)
4974                 return RT_SCOPE_SITE;
4975         else
4976                 return RT_SCOPE_UNIVERSE;
4977 }
4978
4979 static inline int inet6_ifaddr_msgsize(void)
4980 {
4981         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4982                + nla_total_size(16) /* IFA_LOCAL */
4983                + nla_total_size(16) /* IFA_ADDRESS */
4984                + nla_total_size(sizeof(struct ifa_cacheinfo))
4985                + nla_total_size(4)  /* IFA_FLAGS */
4986                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4987 }
4988
4989 enum addr_type_t {
4990         UNICAST_ADDR,
4991         MULTICAST_ADDR,
4992         ANYCAST_ADDR,
4993 };
4994
4995 struct inet6_fill_args {
4996         u32 portid;
4997         u32 seq;
4998         int event;
4999         unsigned int flags;
5000         int netnsid;
5001         int ifindex;
5002         enum addr_type_t type;
5003 };
5004
5005 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5006                              struct inet6_fill_args *args)
5007 {
5008         struct nlmsghdr  *nlh;
5009         u32 preferred, valid;
5010
5011         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5012                         sizeof(struct ifaddrmsg), args->flags);
5013         if (!nlh)
5014                 return -EMSGSIZE;
5015
5016         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5017                       ifa->idev->dev->ifindex);
5018
5019         if (args->netnsid >= 0 &&
5020             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5021                 goto error;
5022
5023         spin_lock_bh(&ifa->lock);
5024         if (!((ifa->flags&IFA_F_PERMANENT) &&
5025               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5026                 preferred = ifa->prefered_lft;
5027                 valid = ifa->valid_lft;
5028                 if (preferred != INFINITY_LIFE_TIME) {
5029                         long tval = (jiffies - ifa->tstamp)/HZ;
5030                         if (preferred > tval)
5031                                 preferred -= tval;
5032                         else
5033                                 preferred = 0;
5034                         if (valid != INFINITY_LIFE_TIME) {
5035                                 if (valid > tval)
5036                                         valid -= tval;
5037                                 else
5038                                         valid = 0;
5039                         }
5040                 }
5041         } else {
5042                 preferred = INFINITY_LIFE_TIME;
5043                 valid = INFINITY_LIFE_TIME;
5044         }
5045         spin_unlock_bh(&ifa->lock);
5046
5047         if (!ipv6_addr_any(&ifa->peer_addr)) {
5048                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5049                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5050                         goto error;
5051         } else
5052                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5053                         goto error;
5054
5055         if (ifa->rt_priority &&
5056             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5057                 goto error;
5058
5059         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5060                 goto error;
5061
5062         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5063                 goto error;
5064
5065         nlmsg_end(skb, nlh);
5066         return 0;
5067
5068 error:
5069         nlmsg_cancel(skb, nlh);
5070         return -EMSGSIZE;
5071 }
5072
5073 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5074                                struct inet6_fill_args *args)
5075 {
5076         struct nlmsghdr  *nlh;
5077         u8 scope = RT_SCOPE_UNIVERSE;
5078         int ifindex = ifmca->idev->dev->ifindex;
5079
5080         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5081                 scope = RT_SCOPE_SITE;
5082
5083         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5084                         sizeof(struct ifaddrmsg), args->flags);
5085         if (!nlh)
5086                 return -EMSGSIZE;
5087
5088         if (args->netnsid >= 0 &&
5089             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5090                 nlmsg_cancel(skb, nlh);
5091                 return -EMSGSIZE;
5092         }
5093
5094         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5095         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5096             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5097                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5098                 nlmsg_cancel(skb, nlh);
5099                 return -EMSGSIZE;
5100         }
5101
5102         nlmsg_end(skb, nlh);
5103         return 0;
5104 }
5105
5106 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5107                                struct inet6_fill_args *args)
5108 {
5109         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5110         int ifindex = dev ? dev->ifindex : 1;
5111         struct nlmsghdr  *nlh;
5112         u8 scope = RT_SCOPE_UNIVERSE;
5113
5114         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5115                 scope = RT_SCOPE_SITE;
5116
5117         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5118                         sizeof(struct ifaddrmsg), args->flags);
5119         if (!nlh)
5120                 return -EMSGSIZE;
5121
5122         if (args->netnsid >= 0 &&
5123             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5124                 nlmsg_cancel(skb, nlh);
5125                 return -EMSGSIZE;
5126         }
5127
5128         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5129         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5130             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5131                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5132                 nlmsg_cancel(skb, nlh);
5133                 return -EMSGSIZE;
5134         }
5135
5136         nlmsg_end(skb, nlh);
5137         return 0;
5138 }
5139
5140 /* called with rcu_read_lock() */
5141 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5142                           struct netlink_callback *cb, int s_ip_idx,
5143                           struct inet6_fill_args *fillargs)
5144 {
5145         struct ifmcaddr6 *ifmca;
5146         struct ifacaddr6 *ifaca;
5147         int ip_idx = 0;
5148         int err = 1;
5149
5150         read_lock_bh(&idev->lock);
5151         switch (fillargs->type) {
5152         case UNICAST_ADDR: {
5153                 struct inet6_ifaddr *ifa;
5154                 fillargs->event = RTM_NEWADDR;
5155
5156                 /* unicast address incl. temp addr */
5157                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5158                         if (ip_idx < s_ip_idx)
5159                                 goto next;
5160                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5161                         if (err < 0)
5162                                 break;
5163                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5164 next:
5165                         ip_idx++;
5166                 }
5167                 break;
5168         }
5169         case MULTICAST_ADDR:
5170                 fillargs->event = RTM_GETMULTICAST;
5171
5172                 /* multicast address */
5173                 for (ifmca = idev->mc_list; ifmca;
5174                      ifmca = ifmca->next, ip_idx++) {
5175                         if (ip_idx < s_ip_idx)
5176                                 continue;
5177                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5178                         if (err < 0)
5179                                 break;
5180                 }
5181                 break;
5182         case ANYCAST_ADDR:
5183                 fillargs->event = RTM_GETANYCAST;
5184                 /* anycast address */
5185                 for (ifaca = idev->ac_list; ifaca;
5186                      ifaca = ifaca->aca_next, ip_idx++) {
5187                         if (ip_idx < s_ip_idx)
5188                                 continue;
5189                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5190                         if (err < 0)
5191                                 break;
5192                 }
5193                 break;
5194         default:
5195                 break;
5196         }
5197         read_unlock_bh(&idev->lock);
5198         cb->args[2] = ip_idx;
5199         return err;
5200 }
5201
5202 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5203                                        struct inet6_fill_args *fillargs,
5204                                        struct net **tgt_net, struct sock *sk,
5205                                        struct netlink_callback *cb)
5206 {
5207         struct netlink_ext_ack *extack = cb->extack;
5208         struct nlattr *tb[IFA_MAX+1];
5209         struct ifaddrmsg *ifm;
5210         int err, i;
5211
5212         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5213                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5214                 return -EINVAL;
5215         }
5216
5217         ifm = nlmsg_data(nlh);
5218         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5219                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5220                 return -EINVAL;
5221         }
5222
5223         fillargs->ifindex = ifm->ifa_index;
5224         if (fillargs->ifindex) {
5225                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5226                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5227         }
5228
5229         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5230                                             ifa_ipv6_policy, extack);
5231         if (err < 0)
5232                 return err;
5233
5234         for (i = 0; i <= IFA_MAX; ++i) {
5235                 if (!tb[i])
5236                         continue;
5237
5238                 if (i == IFA_TARGET_NETNSID) {
5239                         struct net *net;
5240
5241                         fillargs->netnsid = nla_get_s32(tb[i]);
5242                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5243                         if (IS_ERR(net)) {
5244                                 fillargs->netnsid = -1;
5245                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5246                                 return PTR_ERR(net);
5247                         }
5248                         *tgt_net = net;
5249                 } else {
5250                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5251                         return -EINVAL;
5252                 }
5253         }
5254
5255         return 0;
5256 }
5257
5258 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5259                            enum addr_type_t type)
5260 {
5261         const struct nlmsghdr *nlh = cb->nlh;
5262         struct inet6_fill_args fillargs = {
5263                 .portid = NETLINK_CB(cb->skb).portid,
5264                 .seq = cb->nlh->nlmsg_seq,
5265                 .flags = NLM_F_MULTI,
5266                 .netnsid = -1,
5267                 .type = type,
5268         };
5269         struct net *net = sock_net(skb->sk);
5270         struct net *tgt_net = net;
5271         int idx, s_idx, s_ip_idx;
5272         int h, s_h;
5273         struct net_device *dev;
5274         struct inet6_dev *idev;
5275         struct hlist_head *head;
5276         int err = 0;
5277
5278         s_h = cb->args[0];
5279         s_idx = idx = cb->args[1];
5280         s_ip_idx = cb->args[2];
5281
5282         if (cb->strict_check) {
5283                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5284                                                   skb->sk, cb);
5285                 if (err < 0)
5286                         goto put_tgt_net;
5287
5288                 err = 0;
5289                 if (fillargs.ifindex) {
5290                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5291                         if (!dev) {
5292                                 err = -ENODEV;
5293                                 goto put_tgt_net;
5294                         }
5295                         idev = __in6_dev_get(dev);
5296                         if (idev) {
5297                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5298                                                      &fillargs);
5299                                 if (err > 0)
5300                                         err = 0;
5301                         }
5302                         goto put_tgt_net;
5303                 }
5304         }
5305
5306         rcu_read_lock();
5307         cb->seq = inet6_base_seq(tgt_net);
5308         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5309                 idx = 0;
5310                 head = &tgt_net->dev_index_head[h];
5311                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5312                         if (idx < s_idx)
5313                                 goto cont;
5314                         if (h > s_h || idx > s_idx)
5315                                 s_ip_idx = 0;
5316                         idev = __in6_dev_get(dev);
5317                         if (!idev)
5318                                 goto cont;
5319
5320                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5321                                            &fillargs) < 0)
5322                                 goto done;
5323 cont:
5324                         idx++;
5325                 }
5326         }
5327 done:
5328         rcu_read_unlock();
5329         cb->args[0] = h;
5330         cb->args[1] = idx;
5331 put_tgt_net:
5332         if (fillargs.netnsid >= 0)
5333                 put_net(tgt_net);
5334
5335         return skb->len ? : err;
5336 }
5337
5338 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5339 {
5340         enum addr_type_t type = UNICAST_ADDR;
5341
5342         return inet6_dump_addr(skb, cb, type);
5343 }
5344
5345 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5346 {
5347         enum addr_type_t type = MULTICAST_ADDR;
5348
5349         return inet6_dump_addr(skb, cb, type);
5350 }
5351
5352
5353 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5354 {
5355         enum addr_type_t type = ANYCAST_ADDR;
5356
5357         return inet6_dump_addr(skb, cb, type);
5358 }
5359
5360 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5361                                        const struct nlmsghdr *nlh,
5362                                        struct nlattr **tb,
5363                                        struct netlink_ext_ack *extack)
5364 {
5365         struct ifaddrmsg *ifm;
5366         int i, err;
5367
5368         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5369                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5370                 return -EINVAL;
5371         }
5372
5373         if (!netlink_strict_get_check(skb))
5374                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5375                                               ifa_ipv6_policy, extack);
5376
5377         ifm = nlmsg_data(nlh);
5378         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5379                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5380                 return -EINVAL;
5381         }
5382
5383         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5384                                             ifa_ipv6_policy, extack);
5385         if (err)
5386                 return err;
5387
5388         for (i = 0; i <= IFA_MAX; i++) {
5389                 if (!tb[i])
5390                         continue;
5391
5392                 switch (i) {
5393                 case IFA_TARGET_NETNSID:
5394                 case IFA_ADDRESS:
5395                 case IFA_LOCAL:
5396                         break;
5397                 default:
5398                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5399                         return -EINVAL;
5400                 }
5401         }
5402
5403         return 0;
5404 }
5405
5406 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5407                              struct netlink_ext_ack *extack)
5408 {
5409         struct net *net = sock_net(in_skb->sk);
5410         struct inet6_fill_args fillargs = {
5411                 .portid = NETLINK_CB(in_skb).portid,
5412                 .seq = nlh->nlmsg_seq,
5413                 .event = RTM_NEWADDR,
5414                 .flags = 0,
5415                 .netnsid = -1,
5416         };
5417         struct net *tgt_net = net;
5418         struct ifaddrmsg *ifm;
5419         struct nlattr *tb[IFA_MAX+1];
5420         struct in6_addr *addr = NULL, *peer;
5421         struct net_device *dev = NULL;
5422         struct inet6_ifaddr *ifa;
5423         struct sk_buff *skb;
5424         int err;
5425
5426         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5427         if (err < 0)
5428                 return err;
5429
5430         if (tb[IFA_TARGET_NETNSID]) {
5431                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5432
5433                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5434                                                   fillargs.netnsid);
5435                 if (IS_ERR(tgt_net))
5436                         return PTR_ERR(tgt_net);
5437         }
5438
5439         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5440         if (!addr) {
5441                 err = -EINVAL;
5442                 goto errout;
5443         }
5444         ifm = nlmsg_data(nlh);
5445         if (ifm->ifa_index)
5446                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5447
5448         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5449         if (!ifa) {
5450                 err = -EADDRNOTAVAIL;
5451                 goto errout;
5452         }
5453
5454         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5455         if (!skb) {
5456                 err = -ENOBUFS;
5457                 goto errout_ifa;
5458         }
5459
5460         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5461         if (err < 0) {
5462                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5463                 WARN_ON(err == -EMSGSIZE);
5464                 kfree_skb(skb);
5465                 goto errout_ifa;
5466         }
5467         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5468 errout_ifa:
5469         in6_ifa_put(ifa);
5470 errout:
5471         if (dev)
5472                 dev_put(dev);
5473         if (fillargs.netnsid >= 0)
5474                 put_net(tgt_net);
5475
5476         return err;
5477 }
5478
5479 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5480 {
5481         struct sk_buff *skb;
5482         struct net *net = dev_net(ifa->idev->dev);
5483         struct inet6_fill_args fillargs = {
5484                 .portid = 0,
5485                 .seq = 0,
5486                 .event = event,
5487                 .flags = 0,
5488                 .netnsid = -1,
5489         };
5490         int err = -ENOBUFS;
5491
5492         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5493         if (!skb)
5494                 goto errout;
5495
5496         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5497         if (err < 0) {
5498                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5499                 WARN_ON(err == -EMSGSIZE);
5500                 kfree_skb(skb);
5501                 goto errout;
5502         }
5503         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5504         return;
5505 errout:
5506         if (err < 0)
5507                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5508 }
5509
5510 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5511                                 __s32 *array, int bytes)
5512 {
5513         BUG_ON(bytes < (DEVCONF_MAX * 4));
5514
5515         memset(array, 0, bytes);
5516         array[DEVCONF_FORWARDING] = cnf->forwarding;
5517         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5518         array[DEVCONF_MTU6] = cnf->mtu6;
5519         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5520         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5521         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5522         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5523         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5524         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5525                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5526         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5527                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5528         array[DEVCONF_RTR_SOLICIT_DELAY] =
5529                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5530         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5531         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5532                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5533         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5534                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5535         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5536         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5537         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5538         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5539         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5540         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5541         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5542         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5543         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5544 #ifdef CONFIG_IPV6_ROUTER_PREF
5545         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5546         array[DEVCONF_RTR_PROBE_INTERVAL] =
5547                 jiffies_to_msecs(cnf->rtr_probe_interval);
5548 #ifdef CONFIG_IPV6_ROUTE_INFO
5549         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5550         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5551 #endif
5552 #endif
5553         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5554         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5555 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5556         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5557         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5558 #endif
5559 #ifdef CONFIG_IPV6_MROUTE
5560         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5561 #endif
5562         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5563         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5564         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5565         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5566         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5567         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5568         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5569         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5570         /* we omit DEVCONF_STABLE_SECRET for now */
5571         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5572         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5573         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5574         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5575         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5576 #ifdef CONFIG_IPV6_SEG6_HMAC
5577         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5578 #endif
5579         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5580         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5581         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5582         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5583         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5584         array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft;
5585 }
5586
5587 static inline size_t inet6_ifla6_size(void)
5588 {
5589         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5590              + nla_total_size(sizeof(struct ifla_cacheinfo))
5591              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5592              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5593              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5594              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5595              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5596              + 0;
5597 }
5598
5599 static inline size_t inet6_if_nlmsg_size(void)
5600 {
5601         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5602                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5603                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5604                + nla_total_size(4) /* IFLA_MTU */
5605                + nla_total_size(4) /* IFLA_LINK */
5606                + nla_total_size(1) /* IFLA_OPERSTATE */
5607                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5608 }
5609
5610 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5611                                         int bytes)
5612 {
5613         int i;
5614         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5615         BUG_ON(pad < 0);
5616
5617         /* Use put_unaligned() because stats may not be aligned for u64. */
5618         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5619         for (i = 1; i < ICMP6_MIB_MAX; i++)
5620                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5621
5622         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5623 }
5624
5625 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5626                                         int bytes, size_t syncpoff)
5627 {
5628         int i, c;
5629         u64 buff[IPSTATS_MIB_MAX];
5630         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5631
5632         BUG_ON(pad < 0);
5633
5634         memset(buff, 0, sizeof(buff));
5635         buff[0] = IPSTATS_MIB_MAX;
5636
5637         for_each_possible_cpu(c) {
5638                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5639                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5640         }
5641
5642         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5643         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5644 }
5645
5646 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5647                              int bytes)
5648 {
5649         switch (attrtype) {
5650         case IFLA_INET6_STATS:
5651                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5652                                      offsetof(struct ipstats_mib, syncp));
5653                 break;
5654         case IFLA_INET6_ICMP6STATS:
5655                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5656                 break;
5657         }
5658 }
5659
5660 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5661                                   u32 ext_filter_mask)
5662 {
5663         struct nlattr *nla;
5664         struct ifla_cacheinfo ci;
5665
5666         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5667                 goto nla_put_failure;
5668         ci.max_reasm_len = IPV6_MAXPLEN;
5669         ci.tstamp = cstamp_delta(idev->tstamp);
5670         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5671         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5672         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5673                 goto nla_put_failure;
5674         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5675         if (!nla)
5676                 goto nla_put_failure;
5677         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5678
5679         /* XXX - MC not implemented */
5680
5681         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5682                 return 0;
5683
5684         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5685         if (!nla)
5686                 goto nla_put_failure;
5687         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5688
5689         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5690         if (!nla)
5691                 goto nla_put_failure;
5692         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5693
5694         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5695         if (!nla)
5696                 goto nla_put_failure;
5697         read_lock_bh(&idev->lock);
5698         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5699         read_unlock_bh(&idev->lock);
5700
5701         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5702                 goto nla_put_failure;
5703
5704         return 0;
5705
5706 nla_put_failure:
5707         return -EMSGSIZE;
5708 }
5709
5710 static size_t inet6_get_link_af_size(const struct net_device *dev,
5711                                      u32 ext_filter_mask)
5712 {
5713         if (!__in6_dev_get(dev))
5714                 return 0;
5715
5716         return inet6_ifla6_size();
5717 }
5718
5719 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5720                               u32 ext_filter_mask)
5721 {
5722         struct inet6_dev *idev = __in6_dev_get(dev);
5723
5724         if (!idev)
5725                 return -ENODATA;
5726
5727         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5728                 return -EMSGSIZE;
5729
5730         return 0;
5731 }
5732
5733 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5734 {
5735         struct inet6_ifaddr *ifp;
5736         struct net_device *dev = idev->dev;
5737         bool clear_token, update_rs = false;
5738         struct in6_addr ll_addr;
5739
5740         ASSERT_RTNL();
5741
5742         if (!token)
5743                 return -EINVAL;
5744         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5745                 return -EINVAL;
5746         if (!ipv6_accept_ra(idev))
5747                 return -EINVAL;
5748         if (idev->cnf.rtr_solicits == 0)
5749                 return -EINVAL;
5750
5751         write_lock_bh(&idev->lock);
5752
5753         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5754         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5755
5756         write_unlock_bh(&idev->lock);
5757
5758         clear_token = ipv6_addr_any(token);
5759         if (clear_token)
5760                 goto update_lft;
5761
5762         if (!idev->dead && (idev->if_flags & IF_READY) &&
5763             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5764                              IFA_F_OPTIMISTIC)) {
5765                 /* If we're not ready, then normal ifup will take care
5766                  * of this. Otherwise, we need to request our rs here.
5767                  */
5768                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5769                 update_rs = true;
5770         }
5771
5772 update_lft:
5773         write_lock_bh(&idev->lock);
5774
5775         if (update_rs) {
5776                 idev->if_flags |= IF_RS_SENT;
5777                 idev->rs_interval = rfc3315_s14_backoff_init(
5778                         idev->cnf.rtr_solicit_interval);
5779                 idev->rs_probes = 1;
5780                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5781         }
5782
5783         /* Well, that's kinda nasty ... */
5784         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5785                 spin_lock(&ifp->lock);
5786                 if (ifp->tokenized) {
5787                         ifp->valid_lft = 0;
5788                         ifp->prefered_lft = 0;
5789                 }
5790                 spin_unlock(&ifp->lock);
5791         }
5792
5793         write_unlock_bh(&idev->lock);
5794         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5795         addrconf_verify_rtnl();
5796         return 0;
5797 }
5798
5799 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5800         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5801         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5802 };
5803
5804 static int check_addr_gen_mode(int mode)
5805 {
5806         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5807             mode != IN6_ADDR_GEN_MODE_NONE &&
5808             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5809             mode != IN6_ADDR_GEN_MODE_RANDOM)
5810                 return -EINVAL;
5811         return 1;
5812 }
5813
5814 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5815                                 int mode)
5816 {
5817         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5818             !idev->cnf.stable_secret.initialized &&
5819             !net->ipv6.devconf_dflt->stable_secret.initialized)
5820                 return -EINVAL;
5821         return 1;
5822 }
5823
5824 static int inet6_validate_link_af(const struct net_device *dev,
5825                                   const struct nlattr *nla)
5826 {
5827         struct nlattr *tb[IFLA_INET6_MAX + 1];
5828         struct inet6_dev *idev = NULL;
5829         int err;
5830
5831         if (dev) {
5832                 idev = __in6_dev_get(dev);
5833                 if (!idev)
5834                         return -EAFNOSUPPORT;
5835         }
5836
5837         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5838                                           inet6_af_policy, NULL);
5839         if (err)
5840                 return err;
5841
5842         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5843                 return -EINVAL;
5844
5845         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5846                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5847
5848                 if (check_addr_gen_mode(mode) < 0)
5849                         return -EINVAL;
5850                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5851                         return -EINVAL;
5852         }
5853
5854         return 0;
5855 }
5856
5857 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5858 {
5859         struct inet6_dev *idev = __in6_dev_get(dev);
5860         struct nlattr *tb[IFLA_INET6_MAX + 1];
5861         int err;
5862
5863         if (!idev)
5864                 return -EAFNOSUPPORT;
5865
5866         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5867                 return -EINVAL;
5868
5869         if (tb[IFLA_INET6_TOKEN]) {
5870                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5871                 if (err)
5872                         return err;
5873         }
5874
5875         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5876                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5877
5878                 idev->cnf.addr_gen_mode = mode;
5879         }
5880
5881         return 0;
5882 }
5883
5884 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5885                              u32 portid, u32 seq, int event, unsigned int flags)
5886 {
5887         struct net_device *dev = idev->dev;
5888         struct ifinfomsg *hdr;
5889         struct nlmsghdr *nlh;
5890         void *protoinfo;
5891
5892         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5893         if (!nlh)
5894                 return -EMSGSIZE;
5895
5896         hdr = nlmsg_data(nlh);
5897         hdr->ifi_family = AF_INET6;
5898         hdr->__ifi_pad = 0;
5899         hdr->ifi_type = dev->type;
5900         hdr->ifi_index = dev->ifindex;
5901         hdr->ifi_flags = dev_get_flags(dev);
5902         hdr->ifi_change = 0;
5903
5904         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5905             (dev->addr_len &&
5906              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5907             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5908             (dev->ifindex != dev_get_iflink(dev) &&
5909              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5910             nla_put_u8(skb, IFLA_OPERSTATE,
5911                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5912                 goto nla_put_failure;
5913         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5914         if (!protoinfo)
5915                 goto nla_put_failure;
5916
5917         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5918                 goto nla_put_failure;
5919
5920         nla_nest_end(skb, protoinfo);
5921         nlmsg_end(skb, nlh);
5922         return 0;
5923
5924 nla_put_failure:
5925         nlmsg_cancel(skb, nlh);
5926         return -EMSGSIZE;
5927 }
5928
5929 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5930                                    struct netlink_ext_ack *extack)
5931 {
5932         struct ifinfomsg *ifm;
5933
5934         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5935                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5936                 return -EINVAL;
5937         }
5938
5939         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5940                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5941                 return -EINVAL;
5942         }
5943
5944         ifm = nlmsg_data(nlh);
5945         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5946             ifm->ifi_change || ifm->ifi_index) {
5947                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5948                 return -EINVAL;
5949         }
5950
5951         return 0;
5952 }
5953
5954 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5955 {
5956         struct net *net = sock_net(skb->sk);
5957         int h, s_h;
5958         int idx = 0, s_idx;
5959         struct net_device *dev;
5960         struct inet6_dev *idev;
5961         struct hlist_head *head;
5962
5963         /* only requests using strict checking can pass data to
5964          * influence the dump
5965          */
5966         if (cb->strict_check) {
5967                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5968
5969                 if (err < 0)
5970                         return err;
5971         }
5972
5973         s_h = cb->args[0];
5974         s_idx = cb->args[1];
5975
5976         rcu_read_lock();
5977         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5978                 idx = 0;
5979                 head = &net->dev_index_head[h];
5980                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5981                         if (idx < s_idx)
5982                                 goto cont;
5983                         idev = __in6_dev_get(dev);
5984                         if (!idev)
5985                                 goto cont;
5986                         if (inet6_fill_ifinfo(skb, idev,
5987                                               NETLINK_CB(cb->skb).portid,
5988                                               cb->nlh->nlmsg_seq,
5989                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5990                                 goto out;
5991 cont:
5992                         idx++;
5993                 }
5994         }
5995 out:
5996         rcu_read_unlock();
5997         cb->args[1] = idx;
5998         cb->args[0] = h;
5999
6000         return skb->len;
6001 }
6002
6003 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6004 {
6005         struct sk_buff *skb;
6006         struct net *net = dev_net(idev->dev);
6007         int err = -ENOBUFS;
6008
6009         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6010         if (!skb)
6011                 goto errout;
6012
6013         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6014         if (err < 0) {
6015                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6016                 WARN_ON(err == -EMSGSIZE);
6017                 kfree_skb(skb);
6018                 goto errout;
6019         }
6020         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6021         return;
6022 errout:
6023         if (err < 0)
6024                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6025 }
6026
6027 static inline size_t inet6_prefix_nlmsg_size(void)
6028 {
6029         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6030                + nla_total_size(sizeof(struct in6_addr))
6031                + nla_total_size(sizeof(struct prefix_cacheinfo));
6032 }
6033
6034 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6035                              struct prefix_info *pinfo, u32 portid, u32 seq,
6036                              int event, unsigned int flags)
6037 {
6038         struct prefixmsg *pmsg;
6039         struct nlmsghdr *nlh;
6040         struct prefix_cacheinfo ci;
6041
6042         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6043         if (!nlh)
6044                 return -EMSGSIZE;
6045
6046         pmsg = nlmsg_data(nlh);
6047         pmsg->prefix_family = AF_INET6;
6048         pmsg->prefix_pad1 = 0;
6049         pmsg->prefix_pad2 = 0;
6050         pmsg->prefix_ifindex = idev->dev->ifindex;
6051         pmsg->prefix_len = pinfo->prefix_len;
6052         pmsg->prefix_type = pinfo->type;
6053         pmsg->prefix_pad3 = 0;
6054         pmsg->prefix_flags = pinfo->flags;
6055
6056         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6057                 goto nla_put_failure;
6058         ci.preferred_time = ntohl(pinfo->prefered);
6059         ci.valid_time = ntohl(pinfo->valid);
6060         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6061                 goto nla_put_failure;
6062         nlmsg_end(skb, nlh);
6063         return 0;
6064
6065 nla_put_failure:
6066         nlmsg_cancel(skb, nlh);
6067         return -EMSGSIZE;
6068 }
6069
6070 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6071                          struct prefix_info *pinfo)
6072 {
6073         struct sk_buff *skb;
6074         struct net *net = dev_net(idev->dev);
6075         int err = -ENOBUFS;
6076
6077         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6078         if (!skb)
6079                 goto errout;
6080
6081         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6082         if (err < 0) {
6083                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6084                 WARN_ON(err == -EMSGSIZE);
6085                 kfree_skb(skb);
6086                 goto errout;
6087         }
6088         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6089         return;
6090 errout:
6091         if (err < 0)
6092                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6093 }
6094
6095 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6096 {
6097         struct net *net = dev_net(ifp->idev->dev);
6098
6099         if (event)
6100                 ASSERT_RTNL();
6101
6102         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6103
6104         switch (event) {
6105         case RTM_NEWADDR:
6106                 /*
6107                  * If the address was optimistic we inserted the route at the
6108                  * start of our DAD process, so we don't need to do it again.
6109                  * If the device was taken down in the middle of the DAD
6110                  * cycle there is a race where we could get here without a
6111                  * host route, so nothing to insert. That will be fixed when
6112                  * the device is brought up.
6113                  */
6114                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6115                         ip6_ins_rt(net, ifp->rt);
6116                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6117                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6118                                 &ifp->addr, ifp->idev->dev->name);
6119                 }
6120
6121                 if (ifp->idev->cnf.forwarding)
6122                         addrconf_join_anycast(ifp);
6123                 if (!ipv6_addr_any(&ifp->peer_addr))
6124                         addrconf_prefix_route(&ifp->peer_addr, 128,
6125                                               ifp->rt_priority, ifp->idev->dev,
6126                                               0, 0, GFP_ATOMIC);
6127                 break;
6128         case RTM_DELADDR:
6129                 if (ifp->idev->cnf.forwarding)
6130                         addrconf_leave_anycast(ifp);
6131                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6132                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6133                         struct fib6_info *rt;
6134
6135                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6136                                                        ifp->idev->dev, 0, 0,
6137                                                        false);
6138                         if (rt)
6139                                 ip6_del_rt(net, rt, false);
6140                 }
6141                 if (ifp->rt) {
6142                         ip6_del_rt(net, ifp->rt, false);
6143                         ifp->rt = NULL;
6144                 }
6145                 rt_genid_bump_ipv6(net);
6146                 break;
6147         }
6148         atomic_inc(&net->ipv6.dev_addr_genid);
6149 }
6150
6151 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6152 {
6153         rcu_read_lock_bh();
6154         if (likely(ifp->idev->dead == 0))
6155                 __ipv6_ifa_notify(event, ifp);
6156         rcu_read_unlock_bh();
6157 }
6158
6159 #ifdef CONFIG_SYSCTL
6160
6161 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6162                 void *buffer, size_t *lenp, loff_t *ppos)
6163 {
6164         int *valp = ctl->data;
6165         int val = *valp;
6166         loff_t pos = *ppos;
6167         struct ctl_table lctl;
6168         int ret;
6169
6170         /*
6171          * ctl->data points to idev->cnf.forwarding, we should
6172          * not modify it until we get the rtnl lock.
6173          */
6174         lctl = *ctl;
6175         lctl.data = &val;
6176
6177         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6178
6179         if (write)
6180                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6181         if (ret)
6182                 *ppos = pos;
6183         return ret;
6184 }
6185
6186 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6187                 void *buffer, size_t *lenp, loff_t *ppos)
6188 {
6189         struct inet6_dev *idev = ctl->extra1;
6190         int min_mtu = IPV6_MIN_MTU;
6191         struct ctl_table lctl;
6192
6193         lctl = *ctl;
6194         lctl.extra1 = &min_mtu;
6195         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6196
6197         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6198 }
6199
6200 static void dev_disable_change(struct inet6_dev *idev)
6201 {
6202         struct netdev_notifier_info info;
6203
6204         if (!idev || !idev->dev)
6205                 return;
6206
6207         netdev_notifier_info_init(&info, idev->dev);
6208         if (idev->cnf.disable_ipv6)
6209                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6210         else
6211                 addrconf_notify(NULL, NETDEV_UP, &info);
6212 }
6213
6214 static void addrconf_disable_change(struct net *net, __s32 newf)
6215 {
6216         struct net_device *dev;
6217         struct inet6_dev *idev;
6218
6219         for_each_netdev(net, dev) {
6220                 idev = __in6_dev_get(dev);
6221                 if (idev) {
6222                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6223                         idev->cnf.disable_ipv6 = newf;
6224                         if (changed)
6225                                 dev_disable_change(idev);
6226                 }
6227         }
6228 }
6229
6230 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6231 {
6232         struct net *net;
6233         int old;
6234
6235         if (!rtnl_trylock())
6236                 return restart_syscall();
6237
6238         net = (struct net *)table->extra2;
6239         old = *p;
6240         *p = newf;
6241
6242         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6243                 rtnl_unlock();
6244                 return 0;
6245         }
6246
6247         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6248                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6249                 addrconf_disable_change(net, newf);
6250         } else if ((!newf) ^ (!old))
6251                 dev_disable_change((struct inet6_dev *)table->extra1);
6252
6253         rtnl_unlock();
6254         return 0;
6255 }
6256
6257 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6258                 void *buffer, size_t *lenp, loff_t *ppos)
6259 {
6260         int *valp = ctl->data;
6261         int val = *valp;
6262         loff_t pos = *ppos;
6263         struct ctl_table lctl;
6264         int ret;
6265
6266         /*
6267          * ctl->data points to idev->cnf.disable_ipv6, we should
6268          * not modify it until we get the rtnl lock.
6269          */
6270         lctl = *ctl;
6271         lctl.data = &val;
6272
6273         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6274
6275         if (write)
6276                 ret = addrconf_disable_ipv6(ctl, valp, val);
6277         if (ret)
6278                 *ppos = pos;
6279         return ret;
6280 }
6281
6282 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6283                 void *buffer, size_t *lenp, loff_t *ppos)
6284 {
6285         int *valp = ctl->data;
6286         int ret;
6287         int old, new;
6288
6289         old = *valp;
6290         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6291         new = *valp;
6292
6293         if (write && old != new) {
6294                 struct net *net = ctl->extra2;
6295
6296                 if (!rtnl_trylock())
6297                         return restart_syscall();
6298
6299                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6300                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6301                                                      NETCONFA_PROXY_NEIGH,
6302                                                      NETCONFA_IFINDEX_DEFAULT,
6303                                                      net->ipv6.devconf_dflt);
6304                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6305                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6306                                                      NETCONFA_PROXY_NEIGH,
6307                                                      NETCONFA_IFINDEX_ALL,
6308                                                      net->ipv6.devconf_all);
6309                 else {
6310                         struct inet6_dev *idev = ctl->extra1;
6311
6312                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6313                                                      NETCONFA_PROXY_NEIGH,
6314                                                      idev->dev->ifindex,
6315                                                      &idev->cnf);
6316                 }
6317                 rtnl_unlock();
6318         }
6319
6320         return ret;
6321 }
6322
6323 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6324                                          void *buffer, size_t *lenp,
6325                                          loff_t *ppos)
6326 {
6327         int ret = 0;
6328         u32 new_val;
6329         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6330         struct net *net = (struct net *)ctl->extra2;
6331         struct ctl_table tmp = {
6332                 .data = &new_val,
6333                 .maxlen = sizeof(new_val),
6334                 .mode = ctl->mode,
6335         };
6336
6337         if (!rtnl_trylock())
6338                 return restart_syscall();
6339
6340         new_val = *((u32 *)ctl->data);
6341
6342         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6343         if (ret != 0)
6344                 goto out;
6345
6346         if (write) {
6347                 if (check_addr_gen_mode(new_val) < 0) {
6348                         ret = -EINVAL;
6349                         goto out;
6350                 }
6351
6352                 if (idev) {
6353                         if (check_stable_privacy(idev, net, new_val) < 0) {
6354                                 ret = -EINVAL;
6355                                 goto out;
6356                         }
6357
6358                         if (idev->cnf.addr_gen_mode != new_val) {
6359                                 idev->cnf.addr_gen_mode = new_val;
6360                                 addrconf_dev_config(idev->dev);
6361                         }
6362                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6363                         struct net_device *dev;
6364
6365                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6366                         for_each_netdev(net, dev) {
6367                                 idev = __in6_dev_get(dev);
6368                                 if (idev &&
6369                                     idev->cnf.addr_gen_mode != new_val) {
6370                                         idev->cnf.addr_gen_mode = new_val;
6371                                         addrconf_dev_config(idev->dev);
6372                                 }
6373                         }
6374                 }
6375
6376                 *((u32 *)ctl->data) = new_val;
6377         }
6378
6379 out:
6380         rtnl_unlock();
6381
6382         return ret;
6383 }
6384
6385 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6386                                          void *buffer, size_t *lenp,
6387                                          loff_t *ppos)
6388 {
6389         int err;
6390         struct in6_addr addr;
6391         char str[IPV6_MAX_STRLEN];
6392         struct ctl_table lctl = *ctl;
6393         struct net *net = ctl->extra2;
6394         struct ipv6_stable_secret *secret = ctl->data;
6395
6396         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6397                 return -EIO;
6398
6399         lctl.maxlen = IPV6_MAX_STRLEN;
6400         lctl.data = str;
6401
6402         if (!rtnl_trylock())
6403                 return restart_syscall();
6404
6405         if (!write && !secret->initialized) {
6406                 err = -EIO;
6407                 goto out;
6408         }
6409
6410         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6411         if (err >= sizeof(str)) {
6412                 err = -EIO;
6413                 goto out;
6414         }
6415
6416         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6417         if (err || !write)
6418                 goto out;
6419
6420         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6421                 err = -EIO;
6422                 goto out;
6423         }
6424
6425         secret->initialized = true;
6426         secret->secret = addr;
6427
6428         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6429                 struct net_device *dev;
6430
6431                 for_each_netdev(net, dev) {
6432                         struct inet6_dev *idev = __in6_dev_get(dev);
6433
6434                         if (idev) {
6435                                 idev->cnf.addr_gen_mode =
6436                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6437                         }
6438                 }
6439         } else {
6440                 struct inet6_dev *idev = ctl->extra1;
6441
6442                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6443         }
6444
6445 out:
6446         rtnl_unlock();
6447
6448         return err;
6449 }
6450
6451 static
6452 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6453                                                 int write, void *buffer,
6454                                                 size_t *lenp,
6455                                                 loff_t *ppos)
6456 {
6457         int *valp = ctl->data;
6458         int val = *valp;
6459         loff_t pos = *ppos;
6460         struct ctl_table lctl;
6461         int ret;
6462
6463         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6464          * we should not modify it until we get the rtnl lock.
6465          */
6466         lctl = *ctl;
6467         lctl.data = &val;
6468
6469         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6470
6471         if (write)
6472                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6473         if (ret)
6474                 *ppos = pos;
6475         return ret;
6476 }
6477
6478 static
6479 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6480 {
6481         if (rt) {
6482                 if (action)
6483                         rt->dst.flags |= DST_NOPOLICY;
6484                 else
6485                         rt->dst.flags &= ~DST_NOPOLICY;
6486         }
6487 }
6488
6489 static
6490 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6491 {
6492         struct inet6_ifaddr *ifa;
6493
6494         read_lock_bh(&idev->lock);
6495         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6496                 spin_lock(&ifa->lock);
6497                 if (ifa->rt) {
6498                         /* host routes only use builtin fib6_nh */
6499                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6500                         int cpu;
6501
6502                         rcu_read_lock();
6503                         ifa->rt->dst_nopolicy = val ? true : false;
6504                         if (nh->rt6i_pcpu) {
6505                                 for_each_possible_cpu(cpu) {
6506                                         struct rt6_info **rtp;
6507
6508                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6509                                         addrconf_set_nopolicy(*rtp, val);
6510                                 }
6511                         }
6512                         rcu_read_unlock();
6513                 }
6514                 spin_unlock(&ifa->lock);
6515         }
6516         read_unlock_bh(&idev->lock);
6517 }
6518
6519 static
6520 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6521 {
6522         struct inet6_dev *idev;
6523         struct net *net;
6524
6525         if (!rtnl_trylock())
6526                 return restart_syscall();
6527
6528         *valp = val;
6529
6530         net = (struct net *)ctl->extra2;
6531         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6532                 rtnl_unlock();
6533                 return 0;
6534         }
6535
6536         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6537                 struct net_device *dev;
6538
6539                 for_each_netdev(net, dev) {
6540                         idev = __in6_dev_get(dev);
6541                         if (idev)
6542                                 addrconf_disable_policy_idev(idev, val);
6543                 }
6544         } else {
6545                 idev = (struct inet6_dev *)ctl->extra1;
6546                 addrconf_disable_policy_idev(idev, val);
6547         }
6548
6549         rtnl_unlock();
6550         return 0;
6551 }
6552
6553 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6554                                    void *buffer, size_t *lenp, loff_t *ppos)
6555 {
6556         int *valp = ctl->data;
6557         int val = *valp;
6558         loff_t pos = *ppos;
6559         struct ctl_table lctl;
6560         int ret;
6561
6562         lctl = *ctl;
6563         lctl.data = &val;
6564         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6565
6566         if (write && (*valp != val))
6567                 ret = addrconf_disable_policy(ctl, valp, val);
6568
6569         if (ret)
6570                 *ppos = pos;
6571
6572         return ret;
6573 }
6574
6575 static int minus_one = -1;
6576 static const int two_five_five = 255;
6577
6578 static const struct ctl_table addrconf_sysctl[] = {
6579         {
6580                 .procname       = "forwarding",
6581                 .data           = &ipv6_devconf.forwarding,
6582                 .maxlen         = sizeof(int),
6583                 .mode           = 0644,
6584                 .proc_handler   = addrconf_sysctl_forward,
6585         },
6586         {
6587                 .procname       = "hop_limit",
6588                 .data           = &ipv6_devconf.hop_limit,
6589                 .maxlen         = sizeof(int),
6590                 .mode           = 0644,
6591                 .proc_handler   = proc_dointvec_minmax,
6592                 .extra1         = (void *)SYSCTL_ONE,
6593                 .extra2         = (void *)&two_five_five,
6594         },
6595         {
6596                 .procname       = "mtu",
6597                 .data           = &ipv6_devconf.mtu6,
6598                 .maxlen         = sizeof(int),
6599                 .mode           = 0644,
6600                 .proc_handler   = addrconf_sysctl_mtu,
6601         },
6602         {
6603                 .procname       = "accept_ra",
6604                 .data           = &ipv6_devconf.accept_ra,
6605                 .maxlen         = sizeof(int),
6606                 .mode           = 0644,
6607                 .proc_handler   = proc_dointvec,
6608         },
6609         {
6610                 .procname       = "accept_redirects",
6611                 .data           = &ipv6_devconf.accept_redirects,
6612                 .maxlen         = sizeof(int),
6613                 .mode           = 0644,
6614                 .proc_handler   = proc_dointvec,
6615         },
6616         {
6617                 .procname       = "autoconf",
6618                 .data           = &ipv6_devconf.autoconf,
6619                 .maxlen         = sizeof(int),
6620                 .mode           = 0644,
6621                 .proc_handler   = proc_dointvec,
6622         },
6623         {
6624                 .procname       = "dad_transmits",
6625                 .data           = &ipv6_devconf.dad_transmits,
6626                 .maxlen         = sizeof(int),
6627                 .mode           = 0644,
6628                 .proc_handler   = proc_dointvec,
6629         },
6630         {
6631                 .procname       = "router_solicitations",
6632                 .data           = &ipv6_devconf.rtr_solicits,
6633                 .maxlen         = sizeof(int),
6634                 .mode           = 0644,
6635                 .proc_handler   = proc_dointvec_minmax,
6636                 .extra1         = &minus_one,
6637         },
6638         {
6639                 .procname       = "router_solicitation_interval",
6640                 .data           = &ipv6_devconf.rtr_solicit_interval,
6641                 .maxlen         = sizeof(int),
6642                 .mode           = 0644,
6643                 .proc_handler   = proc_dointvec_jiffies,
6644         },
6645         {
6646                 .procname       = "router_solicitation_max_interval",
6647                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6648                 .maxlen         = sizeof(int),
6649                 .mode           = 0644,
6650                 .proc_handler   = proc_dointvec_jiffies,
6651         },
6652         {
6653                 .procname       = "router_solicitation_delay",
6654                 .data           = &ipv6_devconf.rtr_solicit_delay,
6655                 .maxlen         = sizeof(int),
6656                 .mode           = 0644,
6657                 .proc_handler   = proc_dointvec_jiffies,
6658         },
6659         {
6660                 .procname       = "force_mld_version",
6661                 .data           = &ipv6_devconf.force_mld_version,
6662                 .maxlen         = sizeof(int),
6663                 .mode           = 0644,
6664                 .proc_handler   = proc_dointvec,
6665         },
6666         {
6667                 .procname       = "mldv1_unsolicited_report_interval",
6668                 .data           =
6669                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6670                 .maxlen         = sizeof(int),
6671                 .mode           = 0644,
6672                 .proc_handler   = proc_dointvec_ms_jiffies,
6673         },
6674         {
6675                 .procname       = "mldv2_unsolicited_report_interval",
6676                 .data           =
6677                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6678                 .maxlen         = sizeof(int),
6679                 .mode           = 0644,
6680                 .proc_handler   = proc_dointvec_ms_jiffies,
6681         },
6682         {
6683                 .procname       = "use_tempaddr",
6684                 .data           = &ipv6_devconf.use_tempaddr,
6685                 .maxlen         = sizeof(int),
6686                 .mode           = 0644,
6687                 .proc_handler   = proc_dointvec,
6688         },
6689         {
6690                 .procname       = "temp_valid_lft",
6691                 .data           = &ipv6_devconf.temp_valid_lft,
6692                 .maxlen         = sizeof(int),
6693                 .mode           = 0644,
6694                 .proc_handler   = proc_dointvec,
6695         },
6696         {
6697                 .procname       = "temp_prefered_lft",
6698                 .data           = &ipv6_devconf.temp_prefered_lft,
6699                 .maxlen         = sizeof(int),
6700                 .mode           = 0644,
6701                 .proc_handler   = proc_dointvec,
6702         },
6703         {
6704                 .procname       = "regen_max_retry",
6705                 .data           = &ipv6_devconf.regen_max_retry,
6706                 .maxlen         = sizeof(int),
6707                 .mode           = 0644,
6708                 .proc_handler   = proc_dointvec,
6709         },
6710         {
6711                 .procname       = "max_desync_factor",
6712                 .data           = &ipv6_devconf.max_desync_factor,
6713                 .maxlen         = sizeof(int),
6714                 .mode           = 0644,
6715                 .proc_handler   = proc_dointvec,
6716         },
6717         {
6718                 .procname       = "max_addresses",
6719                 .data           = &ipv6_devconf.max_addresses,
6720                 .maxlen         = sizeof(int),
6721                 .mode           = 0644,
6722                 .proc_handler   = proc_dointvec,
6723         },
6724         {
6725                 .procname       = "accept_ra_defrtr",
6726                 .data           = &ipv6_devconf.accept_ra_defrtr,
6727                 .maxlen         = sizeof(int),
6728                 .mode           = 0644,
6729                 .proc_handler   = proc_dointvec,
6730         },
6731         {
6732                 .procname       = "accept_ra_min_hop_limit",
6733                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6734                 .maxlen         = sizeof(int),
6735                 .mode           = 0644,
6736                 .proc_handler   = proc_dointvec,
6737         },
6738         {
6739                 .procname       = "accept_ra_min_lft",
6740                 .data           = &ipv6_devconf.accept_ra_min_lft,
6741                 .maxlen         = sizeof(int),
6742                 .mode           = 0644,
6743                 .proc_handler   = proc_dointvec,
6744         },
6745         {
6746                 .procname       = "accept_ra_pinfo",
6747                 .data           = &ipv6_devconf.accept_ra_pinfo,
6748                 .maxlen         = sizeof(int),
6749                 .mode           = 0644,
6750                 .proc_handler   = proc_dointvec,
6751         },
6752 #ifdef CONFIG_IPV6_ROUTER_PREF
6753         {
6754                 .procname       = "accept_ra_rtr_pref",
6755                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6756                 .maxlen         = sizeof(int),
6757                 .mode           = 0644,
6758                 .proc_handler   = proc_dointvec,
6759         },
6760         {
6761                 .procname       = "router_probe_interval",
6762                 .data           = &ipv6_devconf.rtr_probe_interval,
6763                 .maxlen         = sizeof(int),
6764                 .mode           = 0644,
6765                 .proc_handler   = proc_dointvec_jiffies,
6766         },
6767 #ifdef CONFIG_IPV6_ROUTE_INFO
6768         {
6769                 .procname       = "accept_ra_rt_info_min_plen",
6770                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6771                 .maxlen         = sizeof(int),
6772                 .mode           = 0644,
6773                 .proc_handler   = proc_dointvec,
6774         },
6775         {
6776                 .procname       = "accept_ra_rt_info_max_plen",
6777                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6778                 .maxlen         = sizeof(int),
6779                 .mode           = 0644,
6780                 .proc_handler   = proc_dointvec,
6781         },
6782 #endif
6783 #endif
6784         {
6785                 .procname       = "proxy_ndp",
6786                 .data           = &ipv6_devconf.proxy_ndp,
6787                 .maxlen         = sizeof(int),
6788                 .mode           = 0644,
6789                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6790         },
6791         {
6792                 .procname       = "accept_source_route",
6793                 .data           = &ipv6_devconf.accept_source_route,
6794                 .maxlen         = sizeof(int),
6795                 .mode           = 0644,
6796                 .proc_handler   = proc_dointvec,
6797         },
6798 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6799         {
6800                 .procname       = "optimistic_dad",
6801                 .data           = &ipv6_devconf.optimistic_dad,
6802                 .maxlen         = sizeof(int),
6803                 .mode           = 0644,
6804                 .proc_handler   = proc_dointvec,
6805         },
6806         {
6807                 .procname       = "use_optimistic",
6808                 .data           = &ipv6_devconf.use_optimistic,
6809                 .maxlen         = sizeof(int),
6810                 .mode           = 0644,
6811                 .proc_handler   = proc_dointvec,
6812         },
6813 #endif
6814 #ifdef CONFIG_IPV6_MROUTE
6815         {
6816                 .procname       = "mc_forwarding",
6817                 .data           = &ipv6_devconf.mc_forwarding,
6818                 .maxlen         = sizeof(int),
6819                 .mode           = 0444,
6820                 .proc_handler   = proc_dointvec,
6821         },
6822 #endif
6823         {
6824                 .procname       = "disable_ipv6",
6825                 .data           = &ipv6_devconf.disable_ipv6,
6826                 .maxlen         = sizeof(int),
6827                 .mode           = 0644,
6828                 .proc_handler   = addrconf_sysctl_disable,
6829         },
6830         {
6831                 .procname       = "accept_dad",
6832                 .data           = &ipv6_devconf.accept_dad,
6833                 .maxlen         = sizeof(int),
6834                 .mode           = 0644,
6835                 .proc_handler   = proc_dointvec,
6836         },
6837         {
6838                 .procname       = "force_tllao",
6839                 .data           = &ipv6_devconf.force_tllao,
6840                 .maxlen         = sizeof(int),
6841                 .mode           = 0644,
6842                 .proc_handler   = proc_dointvec
6843         },
6844         {
6845                 .procname       = "ndisc_notify",
6846                 .data           = &ipv6_devconf.ndisc_notify,
6847                 .maxlen         = sizeof(int),
6848                 .mode           = 0644,
6849                 .proc_handler   = proc_dointvec
6850         },
6851         {
6852                 .procname       = "suppress_frag_ndisc",
6853                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6854                 .maxlen         = sizeof(int),
6855                 .mode           = 0644,
6856                 .proc_handler   = proc_dointvec
6857         },
6858         {
6859                 .procname       = "accept_ra_from_local",
6860                 .data           = &ipv6_devconf.accept_ra_from_local,
6861                 .maxlen         = sizeof(int),
6862                 .mode           = 0644,
6863                 .proc_handler   = proc_dointvec,
6864         },
6865         {
6866                 .procname       = "accept_ra_mtu",
6867                 .data           = &ipv6_devconf.accept_ra_mtu,
6868                 .maxlen         = sizeof(int),
6869                 .mode           = 0644,
6870                 .proc_handler   = proc_dointvec,
6871         },
6872         {
6873                 .procname       = "stable_secret",
6874                 .data           = &ipv6_devconf.stable_secret,
6875                 .maxlen         = IPV6_MAX_STRLEN,
6876                 .mode           = 0600,
6877                 .proc_handler   = addrconf_sysctl_stable_secret,
6878         },
6879         {
6880                 .procname       = "use_oif_addrs_only",
6881                 .data           = &ipv6_devconf.use_oif_addrs_only,
6882                 .maxlen         = sizeof(int),
6883                 .mode           = 0644,
6884                 .proc_handler   = proc_dointvec,
6885         },
6886         {
6887                 .procname       = "ignore_routes_with_linkdown",
6888                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6889                 .maxlen         = sizeof(int),
6890                 .mode           = 0644,
6891                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6892         },
6893         {
6894                 .procname       = "drop_unicast_in_l2_multicast",
6895                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6896                 .maxlen         = sizeof(int),
6897                 .mode           = 0644,
6898                 .proc_handler   = proc_dointvec,
6899         },
6900         {
6901                 .procname       = "drop_unsolicited_na",
6902                 .data           = &ipv6_devconf.drop_unsolicited_na,
6903                 .maxlen         = sizeof(int),
6904                 .mode           = 0644,
6905                 .proc_handler   = proc_dointvec,
6906         },
6907         {
6908                 .procname       = "keep_addr_on_down",
6909                 .data           = &ipv6_devconf.keep_addr_on_down,
6910                 .maxlen         = sizeof(int),
6911                 .mode           = 0644,
6912                 .proc_handler   = proc_dointvec,
6913
6914         },
6915         {
6916                 .procname       = "seg6_enabled",
6917                 .data           = &ipv6_devconf.seg6_enabled,
6918                 .maxlen         = sizeof(int),
6919                 .mode           = 0644,
6920                 .proc_handler   = proc_dointvec,
6921         },
6922 #ifdef CONFIG_IPV6_SEG6_HMAC
6923         {
6924                 .procname       = "seg6_require_hmac",
6925                 .data           = &ipv6_devconf.seg6_require_hmac,
6926                 .maxlen         = sizeof(int),
6927                 .mode           = 0644,
6928                 .proc_handler   = proc_dointvec,
6929         },
6930 #endif
6931         {
6932                 .procname       = "enhanced_dad",
6933                 .data           = &ipv6_devconf.enhanced_dad,
6934                 .maxlen         = sizeof(int),
6935                 .mode           = 0644,
6936                 .proc_handler   = proc_dointvec,
6937         },
6938         {
6939                 .procname               = "addr_gen_mode",
6940                 .data                   = &ipv6_devconf.addr_gen_mode,
6941                 .maxlen                 = sizeof(int),
6942                 .mode                   = 0644,
6943                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6944         },
6945         {
6946                 .procname       = "disable_policy",
6947                 .data           = &ipv6_devconf.disable_policy,
6948                 .maxlen         = sizeof(int),
6949                 .mode           = 0644,
6950                 .proc_handler   = addrconf_sysctl_disable_policy,
6951         },
6952         {
6953                 .procname       = "ndisc_tclass",
6954                 .data           = &ipv6_devconf.ndisc_tclass,
6955                 .maxlen         = sizeof(int),
6956                 .mode           = 0644,
6957                 .proc_handler   = proc_dointvec_minmax,
6958                 .extra1         = (void *)SYSCTL_ZERO,
6959                 .extra2         = (void *)&two_five_five,
6960         },
6961         {
6962                 .procname       = "rpl_seg_enabled",
6963                 .data           = &ipv6_devconf.rpl_seg_enabled,
6964                 .maxlen         = sizeof(int),
6965                 .mode           = 0644,
6966                 .proc_handler   = proc_dointvec,
6967         },
6968         {
6969                 /* sentinel */
6970         }
6971 };
6972
6973 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6974                 struct inet6_dev *idev, struct ipv6_devconf *p)
6975 {
6976         int i, ifindex;
6977         struct ctl_table *table;
6978         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6979
6980         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6981         if (!table)
6982                 goto out;
6983
6984         for (i = 0; table[i].data; i++) {
6985                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6986                 /* If one of these is already set, then it is not safe to
6987                  * overwrite either of them: this makes proc_dointvec_minmax
6988                  * usable.
6989                  */
6990                 if (!table[i].extra1 && !table[i].extra2) {
6991                         table[i].extra1 = idev; /* embedded; no ref */
6992                         table[i].extra2 = net;
6993                 }
6994         }
6995
6996         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6997
6998         p->sysctl_header = register_net_sysctl(net, path, table);
6999         if (!p->sysctl_header)
7000                 goto free;
7001
7002         if (!strcmp(dev_name, "all"))
7003                 ifindex = NETCONFA_IFINDEX_ALL;
7004         else if (!strcmp(dev_name, "default"))
7005                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7006         else
7007                 ifindex = idev->dev->ifindex;
7008         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7009                                      ifindex, p);
7010         return 0;
7011
7012 free:
7013         kfree(table);
7014 out:
7015         return -ENOBUFS;
7016 }
7017
7018 static void __addrconf_sysctl_unregister(struct net *net,
7019                                          struct ipv6_devconf *p, int ifindex)
7020 {
7021         struct ctl_table *table;
7022
7023         if (!p->sysctl_header)
7024                 return;
7025
7026         table = p->sysctl_header->ctl_table_arg;
7027         unregister_net_sysctl_table(p->sysctl_header);
7028         p->sysctl_header = NULL;
7029         kfree(table);
7030
7031         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7032 }
7033
7034 static int addrconf_sysctl_register(struct inet6_dev *idev)
7035 {
7036         int err;
7037
7038         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7039                 return -EINVAL;
7040
7041         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7042                                     &ndisc_ifinfo_sysctl_change);
7043         if (err)
7044                 return err;
7045         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7046                                          idev, &idev->cnf);
7047         if (err)
7048                 neigh_sysctl_unregister(idev->nd_parms);
7049
7050         return err;
7051 }
7052
7053 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7054 {
7055         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7056                                      idev->dev->ifindex);
7057         neigh_sysctl_unregister(idev->nd_parms);
7058 }
7059
7060
7061 #endif
7062
7063 static int __net_init addrconf_init_net(struct net *net)
7064 {
7065         int err = -ENOMEM;
7066         struct ipv6_devconf *all, *dflt;
7067
7068         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7069         if (!all)
7070                 goto err_alloc_all;
7071
7072         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7073         if (!dflt)
7074                 goto err_alloc_dflt;
7075
7076         if (!net_eq(net, &init_net)) {
7077                 switch (net_inherit_devconf()) {
7078                 case 1:  /* copy from init_net */
7079                         memcpy(all, init_net.ipv6.devconf_all,
7080                                sizeof(ipv6_devconf));
7081                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7082                                sizeof(ipv6_devconf_dflt));
7083                         break;
7084                 case 3: /* copy from the current netns */
7085                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7086                                sizeof(ipv6_devconf));
7087                         memcpy(dflt,
7088                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7089                                sizeof(ipv6_devconf_dflt));
7090                         break;
7091                 case 0:
7092                 case 2:
7093                         /* use compiled values */
7094                         break;
7095                 }
7096         }
7097
7098         /* these will be inherited by all namespaces */
7099         dflt->autoconf = ipv6_defaults.autoconf;
7100         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7101
7102         dflt->stable_secret.initialized = false;
7103         all->stable_secret.initialized = false;
7104
7105         net->ipv6.devconf_all = all;
7106         net->ipv6.devconf_dflt = dflt;
7107
7108 #ifdef CONFIG_SYSCTL
7109         err = __addrconf_sysctl_register(net, "all", NULL, all);
7110         if (err < 0)
7111                 goto err_reg_all;
7112
7113         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7114         if (err < 0)
7115                 goto err_reg_dflt;
7116 #endif
7117         return 0;
7118
7119 #ifdef CONFIG_SYSCTL
7120 err_reg_dflt:
7121         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7122 err_reg_all:
7123         kfree(dflt);
7124 #endif
7125 err_alloc_dflt:
7126         kfree(all);
7127 err_alloc_all:
7128         return err;
7129 }
7130
7131 static void __net_exit addrconf_exit_net(struct net *net)
7132 {
7133 #ifdef CONFIG_SYSCTL
7134         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7135                                      NETCONFA_IFINDEX_DEFAULT);
7136         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7137                                      NETCONFA_IFINDEX_ALL);
7138 #endif
7139         kfree(net->ipv6.devconf_dflt);
7140         kfree(net->ipv6.devconf_all);
7141 }
7142
7143 static struct pernet_operations addrconf_ops = {
7144         .init = addrconf_init_net,
7145         .exit = addrconf_exit_net,
7146 };
7147
7148 static struct rtnl_af_ops inet6_ops __read_mostly = {
7149         .family           = AF_INET6,
7150         .fill_link_af     = inet6_fill_link_af,
7151         .get_link_af_size = inet6_get_link_af_size,
7152         .validate_link_af = inet6_validate_link_af,
7153         .set_link_af      = inet6_set_link_af,
7154 };
7155
7156 /*
7157  *      Init / cleanup code
7158  */
7159
7160 int __init addrconf_init(void)
7161 {
7162         struct inet6_dev *idev;
7163         int i, err;
7164
7165         err = ipv6_addr_label_init();
7166         if (err < 0) {
7167                 pr_crit("%s: cannot initialize default policy table: %d\n",
7168                         __func__, err);
7169                 goto out;
7170         }
7171
7172         err = register_pernet_subsys(&addrconf_ops);
7173         if (err < 0)
7174                 goto out_addrlabel;
7175
7176         addrconf_wq = create_workqueue("ipv6_addrconf");
7177         if (!addrconf_wq) {
7178                 err = -ENOMEM;
7179                 goto out_nowq;
7180         }
7181
7182         /* The addrconf netdev notifier requires that loopback_dev
7183          * has it's ipv6 private information allocated and setup
7184          * before it can bring up and give link-local addresses
7185          * to other devices which are up.
7186          *
7187          * Unfortunately, loopback_dev is not necessarily the first
7188          * entry in the global dev_base list of net devices.  In fact,
7189          * it is likely to be the very last entry on that list.
7190          * So this causes the notifier registry below to try and
7191          * give link-local addresses to all devices besides loopback_dev
7192          * first, then loopback_dev, which cases all the non-loopback_dev
7193          * devices to fail to get a link-local address.
7194          *
7195          * So, as a temporary fix, allocate the ipv6 structure for
7196          * loopback_dev first by hand.
7197          * Longer term, all of the dependencies ipv6 has upon the loopback
7198          * device and it being up should be removed.
7199          */
7200         rtnl_lock();
7201         idev = ipv6_add_dev(init_net.loopback_dev);
7202         rtnl_unlock();
7203         if (IS_ERR(idev)) {
7204                 err = PTR_ERR(idev);
7205                 goto errlo;
7206         }
7207
7208         ip6_route_init_special_entries();
7209
7210         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7211                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7212
7213         register_netdevice_notifier(&ipv6_dev_notf);
7214
7215         addrconf_verify();
7216
7217         rtnl_af_register(&inet6_ops);
7218
7219         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7220                                    NULL, inet6_dump_ifinfo, 0);
7221         if (err < 0)
7222                 goto errout;
7223
7224         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7225                                    inet6_rtm_newaddr, NULL, 0);
7226         if (err < 0)
7227                 goto errout;
7228         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7229                                    inet6_rtm_deladdr, NULL, 0);
7230         if (err < 0)
7231                 goto errout;
7232         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7233                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7234                                    RTNL_FLAG_DOIT_UNLOCKED);
7235         if (err < 0)
7236                 goto errout;
7237         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7238                                    NULL, inet6_dump_ifmcaddr, 0);
7239         if (err < 0)
7240                 goto errout;
7241         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7242                                    NULL, inet6_dump_ifacaddr, 0);
7243         if (err < 0)
7244                 goto errout;
7245         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7246                                    inet6_netconf_get_devconf,
7247                                    inet6_netconf_dump_devconf,
7248                                    RTNL_FLAG_DOIT_UNLOCKED);
7249         if (err < 0)
7250                 goto errout;
7251         err = ipv6_addr_label_rtnl_register();
7252         if (err < 0)
7253                 goto errout;
7254
7255         return 0;
7256 errout:
7257         rtnl_unregister_all(PF_INET6);
7258         rtnl_af_unregister(&inet6_ops);
7259         unregister_netdevice_notifier(&ipv6_dev_notf);
7260 errlo:
7261         destroy_workqueue(addrconf_wq);
7262 out_nowq:
7263         unregister_pernet_subsys(&addrconf_ops);
7264 out_addrlabel:
7265         ipv6_addr_label_cleanup();
7266 out:
7267         return err;
7268 }
7269
7270 void addrconf_cleanup(void)
7271 {
7272         struct net_device *dev;
7273         int i;
7274
7275         unregister_netdevice_notifier(&ipv6_dev_notf);
7276         unregister_pernet_subsys(&addrconf_ops);
7277         ipv6_addr_label_cleanup();
7278
7279         rtnl_af_unregister(&inet6_ops);
7280
7281         rtnl_lock();
7282
7283         /* clean dev list */
7284         for_each_netdev(&init_net, dev) {
7285                 if (__in6_dev_get(dev) == NULL)
7286                         continue;
7287                 addrconf_ifdown(dev, true);
7288         }
7289         addrconf_ifdown(init_net.loopback_dev, true);
7290
7291         /*
7292          *      Check hash table.
7293          */
7294         spin_lock_bh(&addrconf_hash_lock);
7295         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7296                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7297         spin_unlock_bh(&addrconf_hash_lock);
7298         cancel_delayed_work(&addr_chk_work);
7299         rtnl_unlock();
7300
7301         destroy_workqueue(addrconf_wq);
7302 }