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 #include <linux/ioam6.h>
  93
  94 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  95
  96 #define IPV6_MAX_STRLEN \
  97         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  98
  99 static inline u32 cstamp_delta(unsigned long cstamp)
 100 {
 101         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 102 }
 103
 104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 105 {
 106         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 107         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 108         do_div(tmp, 1000000);
 109         return (s32)tmp;
 110 }
 111
 112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 113 {
 114         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 115         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 116         do_div(tmp, 1000000);
 117         if ((s32)tmp > mrt) {
 118                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 119                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 120                 do_div(tmp, 1000000);
 121         }
 122         return (s32)tmp;
 123 }
 124
 125 #ifdef CONFIG_SYSCTL
 126 static int addrconf_sysctl_register(struct inet6_dev *idev);
 127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 128 #else
 129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 130 {
 131         return 0;
 132 }
 133
 134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 135 {
 136 }
 137 #endif
 138
 139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 140
 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 142 static int ipv6_count_addresses(const struct inet6_dev *idev);
 143 static int ipv6_generate_stable_address(struct in6_addr *addr,
 144                                         u8 dad_count,
 145                                         const struct inet6_dev *idev);
 146
 147 #define IN6_ADDR_HSIZE_SHIFT    8
 148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 149 /*
 150  *      Configured unicast address hash table
 151  */
 152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 153 static DEFINE_SPINLOCK(addrconf_hash_lock);
 154
 155 static void addrconf_verify(void);
 156 static void addrconf_verify_rtnl(void);
 157 static void addrconf_verify_work(struct work_struct *);
 158
 159 static struct workqueue_struct *addrconf_wq;
 160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 161
 162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 164
 165 static void addrconf_type_change(struct net_device *dev,
 166                                  unsigned long event);
 167 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 168
 169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 170                                                   int plen,
 171                                                   const struct net_device *dev,
 172                                                   u32 flags, u32 noflags,
 173                                                   bool no_gw);
 174
 175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 176 static void addrconf_dad_work(struct work_struct *w);
 177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 178                                    bool send_na);
 179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 180 static void addrconf_rs_timer(struct timer_list *t);
 181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 183
 184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 185                                 struct prefix_info *pinfo);
 186
 187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 188         .forwarding             = 0,
 189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 190         .mtu6                   = IPV6_MIN_MTU,
 191         .accept_ra              = 1,
 192         .accept_redirects       = 1,
 193         .autoconf               = 1,
 194         .force_mld_version      = 0,
 195         .mldv1_unsolicited_report_interval = 10 * HZ,
 196         .mldv2_unsolicited_report_interval = HZ,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 210         .accept_ra_from_local   = 0,
 211         .accept_ra_min_hop_limit= 1,
 212         .accept_ra_pinfo        = 1,
 213 #ifdef CONFIG_IPV6_ROUTER_PREF
 214         .accept_ra_rtr_pref     = 1,
 215         .rtr_probe_interval     = 60 * HZ,
 216 #ifdef CONFIG_IPV6_ROUTE_INFO
 217         .accept_ra_rt_info_min_plen = 0,
 218         .accept_ra_rt_info_max_plen = 0,
 219 #endif
 220 #endif
 221         .proxy_ndp              = 0,
 222         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 223         .disable_ipv6           = 0,
 224         .accept_dad             = 0,
 225         .suppress_frag_ndisc    = 1,
 226         .accept_ra_mtu          = 1,
 227         .stable_secret          = {
 228                 .initialized = false,
 229         },
 230         .use_oif_addrs_only     = 0,
 231         .ignore_routes_with_linkdown = 0,
 232         .keep_addr_on_down      = 0,
 233         .seg6_enabled           = 0,
 234 #ifdef CONFIG_IPV6_SEG6_HMAC
 235         .seg6_require_hmac      = 0,
 236 #endif
 237         .enhanced_dad           = 1,
 238         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 239         .disable_policy         = 0,
 240         .rpl_seg_enabled        = 0,
 241         .ioam6_enabled          = 0,
 242         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 243         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 244 };
 245
 246 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 247         .forwarding             = 0,
 248         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 249         .mtu6                   = IPV6_MIN_MTU,
 250         .accept_ra              = 1,
 251         .accept_redirects       = 1,
 252         .autoconf               = 1,
 253         .force_mld_version      = 0,
 254         .mldv1_unsolicited_report_interval = 10 * HZ,
 255         .mldv2_unsolicited_report_interval = HZ,
 256         .dad_transmits          = 1,
 257         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 258         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 259         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 260         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 261         .use_tempaddr           = 0,
 262         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 263         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 264         .regen_max_retry        = REGEN_MAX_RETRY,
 265         .max_desync_factor      = MAX_DESYNC_FACTOR,
 266         .max_addresses          = IPV6_MAX_ADDRESSES,
 267         .accept_ra_defrtr       = 1,
 268         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 269         .accept_ra_from_local   = 0,
 270         .accept_ra_min_hop_limit= 1,
 271         .accept_ra_pinfo        = 1,
 272 #ifdef CONFIG_IPV6_ROUTER_PREF
 273         .accept_ra_rtr_pref     = 1,
 274         .rtr_probe_interval     = 60 * HZ,
 275 #ifdef CONFIG_IPV6_ROUTE_INFO
 276         .accept_ra_rt_info_min_plen = 0,
 277         .accept_ra_rt_info_max_plen = 0,
 278 #endif
 279 #endif
 280         .proxy_ndp              = 0,
 281         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 282         .disable_ipv6           = 0,
 283         .accept_dad             = 1,
 284         .suppress_frag_ndisc    = 1,
 285         .accept_ra_mtu          = 1,
 286         .stable_secret          = {
 287                 .initialized = false,
 288         },
 289         .use_oif_addrs_only     = 0,
 290         .ignore_routes_with_linkdown = 0,
 291         .keep_addr_on_down      = 0,
 292         .seg6_enabled           = 0,
 293 #ifdef CONFIG_IPV6_SEG6_HMAC
 294         .seg6_require_hmac      = 0,
 295 #endif
 296         .enhanced_dad           = 1,
 297         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 298         .disable_policy         = 0,
 299         .rpl_seg_enabled        = 0,
 300         .ioam6_enabled          = 0,
 301         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 302         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 303 };
 304
 305 /* Check if link is ready: is it up and is a valid qdisc available */
 306 static inline bool addrconf_link_ready(const struct net_device *dev)
 307 {
 308         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 309 }
 310
 311 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 312 {
 313         if (del_timer(&idev->rs_timer))
 314                 __in6_dev_put(idev);
 315 }
 316
 317 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 318 {
 319         if (cancel_delayed_work(&ifp->dad_work))
 320                 __in6_ifa_put(ifp);
 321 }
 322
 323 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 324                                   unsigned long when)
 325 {
 326         if (!timer_pending(&idev->rs_timer))
 327                 in6_dev_hold(idev);
 328         mod_timer(&idev->rs_timer, jiffies + when);
 329 }
 330
 331 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 332                                    unsigned long delay)
 333 {
 334         in6_ifa_hold(ifp);
 335         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 336                 in6_ifa_put(ifp);
 337 }
 338
 339 static int snmp6_alloc_dev(struct inet6_dev *idev)
 340 {
 341         int i;
 342
 343         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 344         if (!idev->stats.ipv6)
 345                 goto err_ip;
 346
 347         for_each_possible_cpu(i) {
 348                 struct ipstats_mib *addrconf_stats;
 349                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 350                 u64_stats_init(&addrconf_stats->syncp);
 351         }
 352
 353
 354         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 355                                         GFP_KERNEL);
 356         if (!idev->stats.icmpv6dev)
 357                 goto err_icmp;
 358         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 359                                            GFP_KERNEL);
 360         if (!idev->stats.icmpv6msgdev)
 361                 goto err_icmpmsg;
 362
 363         return 0;
 364
 365 err_icmpmsg:
 366         kfree(idev->stats.icmpv6dev);
 367 err_icmp:
 368         free_percpu(idev->stats.ipv6);
 369 err_ip:
 370         return -ENOMEM;
 371 }
 372
 373 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 374 {
 375         struct inet6_dev *ndev;
 376         int err = -ENOMEM;
 377
 378         ASSERT_RTNL();
 379
 380         if (dev->mtu < IPV6_MIN_MTU)
 381                 return ERR_PTR(-EINVAL);
 382
 383         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 384         if (!ndev)
 385                 return ERR_PTR(err);
 386
 387         rwlock_init(&ndev->lock);
 388         ndev->dev = dev;
 389         INIT_LIST_HEAD(&ndev->addr_list);
 390         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 391         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 392
 393         if (ndev->cnf.stable_secret.initialized)
 394                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 395
 396         ndev->cnf.mtu6 = dev->mtu;
 397         ndev->ra_mtu = 0;
 398         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 399         if (!ndev->nd_parms) {
 400                 kfree(ndev);
 401                 return ERR_PTR(err);
 402         }
 403         if (ndev->cnf.forwarding)
 404                 dev_disable_lro(dev);
 405         /* We refer to the device */
 406         dev_hold(dev);
 407
 408         if (snmp6_alloc_dev(ndev) < 0) {
 409                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 410                            __func__);
 411                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 412                 dev_put(dev);
 413                 kfree(ndev);
 414                 return ERR_PTR(err);
 415         }
 416
 417         if (snmp6_register_dev(ndev) < 0) {
 418                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 419                            __func__, dev->name);
 420                 goto err_release;
 421         }
 422
 423         /* One reference from device. */
 424         refcount_set(&ndev->refcnt, 1);
 425
 426         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 427                 ndev->cnf.accept_dad = -1;
 428
 429 #if IS_ENABLED(CONFIG_IPV6_SIT)
 430         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 431                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 432                 ndev->cnf.rtr_solicits = 0;
 433         }
 434 #endif
 435
 436         INIT_LIST_HEAD(&ndev->tempaddr_list);
 437         ndev->desync_factor = U32_MAX;
 438         if ((dev->flags&IFF_LOOPBACK) ||
 439             dev->type == ARPHRD_TUNNEL ||
 440             dev->type == ARPHRD_TUNNEL6 ||
 441             dev->type == ARPHRD_SIT ||
 442             dev->type == ARPHRD_NONE) {
 443                 ndev->cnf.use_tempaddr = -1;
 444         }
 445
 446         ndev->token = in6addr_any;
 447
 448         if (netif_running(dev) && addrconf_link_ready(dev))
 449                 ndev->if_flags |= IF_READY;
 450
 451         ipv6_mc_init_dev(ndev);
 452         ndev->tstamp = jiffies;
 453         err = addrconf_sysctl_register(ndev);
 454         if (err) {
 455                 ipv6_mc_destroy_dev(ndev);
 456                 snmp6_unregister_dev(ndev);
 457                 goto err_release;
 458         }
 459         /* protected by rtnl_lock */
 460         rcu_assign_pointer(dev->ip6_ptr, ndev);
 461
 462         /* Join interface-local all-node multicast group */
 463         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 464
 465         /* Join all-node multicast group */
 466         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 467
 468         /* Join all-router multicast group if forwarding is set */
 469         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 470                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 471
 472         return ndev;
 473
 474 err_release:
 475         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 476         ndev->dead = 1;
 477         in6_dev_finish_destroy(ndev);
 478         return ERR_PTR(err);
 479 }
 480
 481 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 482 {
 483         struct inet6_dev *idev;
 484
 485         ASSERT_RTNL();
 486
 487         idev = __in6_dev_get(dev);
 488         if (!idev) {
 489                 idev = ipv6_add_dev(dev);
 490                 if (IS_ERR(idev))
 491                         return idev;
 492         }
 493
 494         if (dev->flags&IFF_UP)
 495                 ipv6_mc_up(idev);
 496         return idev;
 497 }
 498
 499 static int inet6_netconf_msgsize_devconf(int type)
 500 {
 501         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 502                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 503         bool all = false;
 504
 505         if (type == NETCONFA_ALL)
 506                 all = true;
 507
 508         if (all || type == NETCONFA_FORWARDING)
 509                 size += nla_total_size(4);
 510 #ifdef CONFIG_IPV6_MROUTE
 511         if (all || type == NETCONFA_MC_FORWARDING)
 512                 size += nla_total_size(4);
 513 #endif
 514         if (all || type == NETCONFA_PROXY_NEIGH)
 515                 size += nla_total_size(4);
 516
 517         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 518                 size += nla_total_size(4);
 519
 520         return size;
 521 }
 522
 523 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 524                                       struct ipv6_devconf *devconf, u32 portid,
 525                                       u32 seq, int event, unsigned int flags,
 526                                       int type)
 527 {
 528         struct nlmsghdr  *nlh;
 529         struct netconfmsg *ncm;
 530         bool all = false;
 531
 532         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 533                         flags);
 534         if (!nlh)
 535                 return -EMSGSIZE;
 536
 537         if (type == NETCONFA_ALL)
 538                 all = true;
 539
 540         ncm = nlmsg_data(nlh);
 541         ncm->ncm_family = AF_INET6;
 542
 543         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 544                 goto nla_put_failure;
 545
 546         if (!devconf)
 547                 goto out;
 548
 549         if ((all || type == NETCONFA_FORWARDING) &&
 550             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 551                 goto nla_put_failure;
 552 #ifdef CONFIG_IPV6_MROUTE
 553         if ((all || type == NETCONFA_MC_FORWARDING) &&
 554             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 555                         atomic_read(&devconf->mc_forwarding)) < 0)
 556                 goto nla_put_failure;
 557 #endif
 558         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 559             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 560                 goto nla_put_failure;
 561
 562         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 563             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 564                         devconf->ignore_routes_with_linkdown) < 0)
 565                 goto nla_put_failure;
 566
 567 out:
 568         nlmsg_end(skb, nlh);
 569         return 0;
 570
 571 nla_put_failure:
 572         nlmsg_cancel(skb, nlh);
 573         return -EMSGSIZE;
 574 }
 575
 576 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 577                                   int ifindex, struct ipv6_devconf *devconf)
 578 {
 579         struct sk_buff *skb;
 580         int err = -ENOBUFS;
 581
 582         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 583         if (!skb)
 584                 goto errout;
 585
 586         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 587                                          event, 0, type);
 588         if (err < 0) {
 589                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 590                 WARN_ON(err == -EMSGSIZE);
 591                 kfree_skb(skb);
 592                 goto errout;
 593         }
 594         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 595         return;
 596 errout:
 597         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 598 }
 599
 600 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 601         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 602         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 603         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 604         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 605 };
 606
 607 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 608                                        const struct nlmsghdr *nlh,
 609                                        struct nlattr **tb,
 610                                        struct netlink_ext_ack *extack)
 611 {
 612         int i, err;
 613
 614         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 615                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 616                 return -EINVAL;
 617         }
 618
 619         if (!netlink_strict_get_check(skb))
 620                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 621                                               tb, NETCONFA_MAX,
 622                                               devconf_ipv6_policy, extack);
 623
 624         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 625                                             tb, NETCONFA_MAX,
 626                                             devconf_ipv6_policy, extack);
 627         if (err)
 628                 return err;
 629
 630         for (i = 0; i <= NETCONFA_MAX; i++) {
 631                 if (!tb[i])
 632                         continue;
 633
 634                 switch (i) {
 635                 case NETCONFA_IFINDEX:
 636                         break;
 637                 default:
 638                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 639                         return -EINVAL;
 640                 }
 641         }
 642
 643         return 0;
 644 }
 645
 646 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 647                                      struct nlmsghdr *nlh,
 648                                      struct netlink_ext_ack *extack)
 649 {
 650         struct net *net = sock_net(in_skb->sk);
 651         struct nlattr *tb[NETCONFA_MAX+1];
 652         struct inet6_dev *in6_dev = NULL;
 653         struct net_device *dev = NULL;
 654         struct sk_buff *skb;
 655         struct ipv6_devconf *devconf;
 656         int ifindex;
 657         int err;
 658
 659         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 660         if (err < 0)
 661                 return err;
 662
 663         if (!tb[NETCONFA_IFINDEX])
 664                 return -EINVAL;
 665
 666         err = -EINVAL;
 667         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 668         switch (ifindex) {
 669         case NETCONFA_IFINDEX_ALL:
 670                 devconf = net->ipv6.devconf_all;
 671                 break;
 672         case NETCONFA_IFINDEX_DEFAULT:
 673                 devconf = net->ipv6.devconf_dflt;
 674                 break;
 675         default:
 676                 dev = dev_get_by_index(net, ifindex);
 677                 if (!dev)
 678                         return -EINVAL;
 679                 in6_dev = in6_dev_get(dev);
 680                 if (!in6_dev)
 681                         goto errout;
 682                 devconf = &in6_dev->cnf;
 683                 break;
 684         }
 685
 686         err = -ENOBUFS;
 687         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 688         if (!skb)
 689                 goto errout;
 690
 691         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 692                                          NETLINK_CB(in_skb).portid,
 693                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 694                                          NETCONFA_ALL);
 695         if (err < 0) {
 696                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 697                 WARN_ON(err == -EMSGSIZE);
 698                 kfree_skb(skb);
 699                 goto errout;
 700         }
 701         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 702 errout:
 703         if (in6_dev)
 704                 in6_dev_put(in6_dev);
 705         dev_put(dev);
 706         return err;
 707 }
 708
 709 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 710                                       struct netlink_callback *cb)
 711 {
 712         const struct nlmsghdr *nlh = cb->nlh;
 713         struct net *net = sock_net(skb->sk);
 714         int h, s_h;
 715         int idx, s_idx;
 716         struct net_device *dev;
 717         struct inet6_dev *idev;
 718         struct hlist_head *head;
 719
 720         if (cb->strict_check) {
 721                 struct netlink_ext_ack *extack = cb->extack;
 722                 struct netconfmsg *ncm;
 723
 724                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 725                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 726                         return -EINVAL;
 727                 }
 728
 729                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 730                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 731                         return -EINVAL;
 732                 }
 733         }
 734
 735         s_h = cb->args[0];
 736         s_idx = idx = cb->args[1];
 737
 738         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 739                 idx = 0;
 740                 head = &net->dev_index_head[h];
 741                 rcu_read_lock();
 742                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 743                           net->dev_base_seq;
 744                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 745                         if (idx < s_idx)
 746                                 goto cont;
 747                         idev = __in6_dev_get(dev);
 748                         if (!idev)
 749                                 goto cont;
 750
 751                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 752                                                        &idev->cnf,
 753                                                        NETLINK_CB(cb->skb).portid,
 754                                                        nlh->nlmsg_seq,
 755                                                        RTM_NEWNETCONF,
 756                                                        NLM_F_MULTI,
 757                                                        NETCONFA_ALL) < 0) {
 758                                 rcu_read_unlock();
 759                                 goto done;
 760                         }
 761                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 762 cont:
 763                         idx++;
 764                 }
 765                 rcu_read_unlock();
 766         }
 767         if (h == NETDEV_HASHENTRIES) {
 768                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 769                                                net->ipv6.devconf_all,
 770                                                NETLINK_CB(cb->skb).portid,
 771                                                nlh->nlmsg_seq,
 772                                                RTM_NEWNETCONF, NLM_F_MULTI,
 773                                                NETCONFA_ALL) < 0)
 774                         goto done;
 775                 else
 776                         h++;
 777         }
 778         if (h == NETDEV_HASHENTRIES + 1) {
 779                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 780                                                net->ipv6.devconf_dflt,
 781                                                NETLINK_CB(cb->skb).portid,
 782                                                nlh->nlmsg_seq,
 783                                                RTM_NEWNETCONF, NLM_F_MULTI,
 784                                                NETCONFA_ALL) < 0)
 785                         goto done;
 786                 else
 787                         h++;
 788         }
 789 done:
 790         cb->args[0] = h;
 791         cb->args[1] = idx;
 792
 793         return skb->len;
 794 }
 795
 796 #ifdef CONFIG_SYSCTL
 797 static void dev_forward_change(struct inet6_dev *idev)
 798 {
 799         struct net_device *dev;
 800         struct inet6_ifaddr *ifa;
 801         LIST_HEAD(tmp_addr_list);
 802
 803         if (!idev)
 804                 return;
 805         dev = idev->dev;
 806         if (idev->cnf.forwarding)
 807                 dev_disable_lro(dev);
 808         if (dev->flags & IFF_MULTICAST) {
 809                 if (idev->cnf.forwarding) {
 810                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 811                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 812                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 813                 } else {
 814                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 815                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 816                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 817                 }
 818         }
 819
 820         read_lock_bh(&idev->lock);
 821         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 822                 if (ifa->flags&IFA_F_TENTATIVE)
 823                         continue;
 824                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
 825         }
 826         read_unlock_bh(&idev->lock);
 827
 828         while (!list_empty(&tmp_addr_list)) {
 829                 ifa = list_first_entry(&tmp_addr_list,
 830                                        struct inet6_ifaddr, if_list_aux);
 831                 list_del(&ifa->if_list_aux);
 832                 if (idev->cnf.forwarding)
 833                         addrconf_join_anycast(ifa);
 834                 else
 835                         addrconf_leave_anycast(ifa);
 836         }
 837
 838         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 839                                      NETCONFA_FORWARDING,
 840                                      dev->ifindex, &idev->cnf);
 841 }
 842
 843
 844 static void addrconf_forward_change(struct net *net, __s32 newf)
 845 {
 846         struct net_device *dev;
 847         struct inet6_dev *idev;
 848
 849         for_each_netdev(net, dev) {
 850                 idev = __in6_dev_get(dev);
 851                 if (idev) {
 852                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 853                         idev->cnf.forwarding = newf;
 854                         if (changed)
 855                                 dev_forward_change(idev);
 856                 }
 857         }
 858 }
 859
 860 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 861 {
 862         struct net *net;
 863         int old;
 864
 865         if (!rtnl_trylock())
 866                 return restart_syscall();
 867
 868         net = (struct net *)table->extra2;
 869         old = *p;
 870         *p = newf;
 871
 872         if (p == &net->ipv6.devconf_dflt->forwarding) {
 873                 if ((!newf) ^ (!old))
 874                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 875                                                      NETCONFA_FORWARDING,
 876                                                      NETCONFA_IFINDEX_DEFAULT,
 877                                                      net->ipv6.devconf_dflt);
 878                 rtnl_unlock();
 879                 return 0;
 880         }
 881
 882         if (p == &net->ipv6.devconf_all->forwarding) {
 883                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 884
 885                 net->ipv6.devconf_dflt->forwarding = newf;
 886                 if ((!newf) ^ (!old_dflt))
 887                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 888                                                      NETCONFA_FORWARDING,
 889                                                      NETCONFA_IFINDEX_DEFAULT,
 890                                                      net->ipv6.devconf_dflt);
 891
 892                 addrconf_forward_change(net, newf);
 893                 if ((!newf) ^ (!old))
 894                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 895                                                      NETCONFA_FORWARDING,
 896                                                      NETCONFA_IFINDEX_ALL,
 897                                                      net->ipv6.devconf_all);
 898         } else if ((!newf) ^ (!old))
 899                 dev_forward_change((struct inet6_dev *)table->extra1);
 900         rtnl_unlock();
 901
 902         if (newf)
 903                 rt6_purge_dflt_routers(net);
 904         return 1;
 905 }
 906
 907 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 908 {
 909         struct net_device *dev;
 910         struct inet6_dev *idev;
 911
 912         for_each_netdev(net, dev) {
 913                 idev = __in6_dev_get(dev);
 914                 if (idev) {
 915                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 916
 917                         idev->cnf.ignore_routes_with_linkdown = newf;
 918                         if (changed)
 919                                 inet6_netconf_notify_devconf(dev_net(dev),
 920                                                              RTM_NEWNETCONF,
 921                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 922                                                              dev->ifindex,
 923                                                              &idev->cnf);
 924                 }
 925         }
 926 }
 927
 928 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 929 {
 930         struct net *net;
 931         int old;
 932
 933         if (!rtnl_trylock())
 934                 return restart_syscall();
 935
 936         net = (struct net *)table->extra2;
 937         old = *p;
 938         *p = newf;
 939
 940         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 941                 if ((!newf) ^ (!old))
 942                         inet6_netconf_notify_devconf(net,
 943                                                      RTM_NEWNETCONF,
 944                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 945                                                      NETCONFA_IFINDEX_DEFAULT,
 946                                                      net->ipv6.devconf_dflt);
 947                 rtnl_unlock();
 948                 return 0;
 949         }
 950
 951         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 952                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 953                 addrconf_linkdown_change(net, newf);
 954                 if ((!newf) ^ (!old))
 955                         inet6_netconf_notify_devconf(net,
 956                                                      RTM_NEWNETCONF,
 957                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 958                                                      NETCONFA_IFINDEX_ALL,
 959                                                      net->ipv6.devconf_all);
 960         }
 961         rtnl_unlock();
 962
 963         return 1;
 964 }
 965
 966 #endif
 967
 968 /* Nobody refers to this ifaddr, destroy it */
 969 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 970 {
 971         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 972
 973 #ifdef NET_REFCNT_DEBUG
 974         pr_debug("%s\n", __func__);
 975 #endif
 976
 977         in6_dev_put(ifp->idev);
 978
 979         if (cancel_delayed_work(&ifp->dad_work))
 980                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 981                           ifp);
 982
 983         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 984                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 985                 return;
 986         }
 987
 988         kfree_rcu(ifp, rcu);
 989 }
 990
 991 static void
 992 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 993 {
 994         struct list_head *p;
 995         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 996
 997         /*
 998          * Each device address list is sorted in order of scope -
 999          * global before linklocal.
1000          */
1001         list_for_each(p, &idev->addr_list) {
1002                 struct inet6_ifaddr *ifa
1003                         = list_entry(p, struct inet6_ifaddr, if_list);
1004                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1005                         break;
1006         }
1007
1008         list_add_tail_rcu(&ifp->if_list, p);
1009 }
1010
1011 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1012 {
1013         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1014
1015         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1016 }
1017
1018 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1019                                struct net_device *dev, unsigned int hash)
1020 {
1021         struct inet6_ifaddr *ifp;
1022
1023         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1024                 if (!net_eq(dev_net(ifp->idev->dev), net))
1025                         continue;
1026                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1027                         if (!dev || ifp->idev->dev == dev)
1028                                 return true;
1029                 }
1030         }
1031         return false;
1032 }
1033
1034 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1035 {
1036         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1037         int err = 0;
1038
1039         spin_lock(&addrconf_hash_lock);
1040
1041         /* Ignore adding duplicate addresses on an interface */
1042         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1043                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1044                 err = -EEXIST;
1045         } else {
1046                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1047         }
1048
1049         spin_unlock(&addrconf_hash_lock);
1050
1051         return err;
1052 }
1053
1054 /* On success it returns ifp with increased reference count */
1055
1056 static struct inet6_ifaddr *
1057 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1058               bool can_block, struct netlink_ext_ack *extack)
1059 {
1060         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1061         int addr_type = ipv6_addr_type(cfg->pfx);
1062         struct net *net = dev_net(idev->dev);
1063         struct inet6_ifaddr *ifa = NULL;
1064         struct fib6_info *f6i = NULL;
1065         int err = 0;
1066
1067         if (addr_type == IPV6_ADDR_ANY ||
1068             (addr_type & IPV6_ADDR_MULTICAST &&
1069              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1070             (!(idev->dev->flags & IFF_LOOPBACK) &&
1071              !netif_is_l3_master(idev->dev) &&
1072              addr_type & IPV6_ADDR_LOOPBACK))
1073                 return ERR_PTR(-EADDRNOTAVAIL);
1074
1075         if (idev->dead) {
1076                 err = -ENODEV;                  /*XXX*/
1077                 goto out;
1078         }
1079
1080         if (idev->cnf.disable_ipv6) {
1081                 err = -EACCES;
1082                 goto out;
1083         }
1084
1085         /* validator notifier needs to be blocking;
1086          * do not call in atomic context
1087          */
1088         if (can_block) {
1089                 struct in6_validator_info i6vi = {
1090                         .i6vi_addr = *cfg->pfx,
1091                         .i6vi_dev = idev,
1092                         .extack = extack,
1093                 };
1094
1095                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1096                 err = notifier_to_errno(err);
1097                 if (err < 0)
1098                         goto out;
1099         }
1100
1101         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1102         if (!ifa) {
1103                 err = -ENOBUFS;
1104                 goto out;
1105         }
1106
1107         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1108         if (IS_ERR(f6i)) {
1109                 err = PTR_ERR(f6i);
1110                 f6i = NULL;
1111                 goto out;
1112         }
1113
1114         neigh_parms_data_state_setall(idev->nd_parms);
1115
1116         ifa->addr = *cfg->pfx;
1117         if (cfg->peer_pfx)
1118                 ifa->peer_addr = *cfg->peer_pfx;
1119
1120         spin_lock_init(&ifa->lock);
1121         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1122         INIT_HLIST_NODE(&ifa->addr_lst);
1123         ifa->scope = cfg->scope;
1124         ifa->prefix_len = cfg->plen;
1125         ifa->rt_priority = cfg->rt_priority;
1126         ifa->flags = cfg->ifa_flags;
1127         /* No need to add the TENTATIVE flag for addresses with NODAD */
1128         if (!(cfg->ifa_flags & IFA_F_NODAD))
1129                 ifa->flags |= IFA_F_TENTATIVE;
1130         ifa->valid_lft = cfg->valid_lft;
1131         ifa->prefered_lft = cfg->preferred_lft;
1132         ifa->cstamp = ifa->tstamp = jiffies;
1133         ifa->tokenized = false;
1134
1135         ifa->rt = f6i;
1136
1137         ifa->idev = idev;
1138         in6_dev_hold(idev);
1139
1140         /* For caller */
1141         refcount_set(&ifa->refcnt, 1);
1142
1143         rcu_read_lock_bh();
1144
1145         err = ipv6_add_addr_hash(idev->dev, ifa);
1146         if (err < 0) {
1147                 rcu_read_unlock_bh();
1148                 goto out;
1149         }
1150
1151         write_lock(&idev->lock);
1152
1153         /* Add to inet6_dev unicast addr list. */
1154         ipv6_link_dev_addr(idev, ifa);
1155
1156         if (ifa->flags&IFA_F_TEMPORARY) {
1157                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1158                 in6_ifa_hold(ifa);
1159         }
1160
1161         in6_ifa_hold(ifa);
1162         write_unlock(&idev->lock);
1163
1164         rcu_read_unlock_bh();
1165
1166         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1167 out:
1168         if (unlikely(err < 0)) {
1169                 fib6_info_release(f6i);
1170
1171                 if (ifa) {
1172                         if (ifa->idev)
1173                                 in6_dev_put(ifa->idev);
1174                         kfree(ifa);
1175                 }
1176                 ifa = ERR_PTR(err);
1177         }
1178
1179         return ifa;
1180 }
1181
1182 enum cleanup_prefix_rt_t {
1183         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1184         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1185         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1186 };
1187
1188 /*
1189  * Check, whether the prefix for ifp would still need a prefix route
1190  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1191  * constants.
1192  *
1193  * 1) we don't purge prefix if address was not permanent.
1194  *    prefix is managed by its own lifetime.
1195  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1196  * 3) if there are no addresses, delete prefix.
1197  * 4) if there are still other permanent address(es),
1198  *    corresponding prefix is still permanent.
1199  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1200  *    don't purge the prefix, assume user space is managing it.
1201  * 6) otherwise, update prefix lifetime to the
1202  *    longest valid lifetime among the corresponding
1203  *    addresses on the device.
1204  *    Note: subsequent RA will update lifetime.
1205  **/
1206 static enum cleanup_prefix_rt_t
1207 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1208 {
1209         struct inet6_ifaddr *ifa;
1210         struct inet6_dev *idev = ifp->idev;
1211         unsigned long lifetime;
1212         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1213
1214         *expires = jiffies;
1215
1216         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1217                 if (ifa == ifp)
1218                         continue;
1219                 if (ifa->prefix_len != ifp->prefix_len ||
1220                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1221                                        ifp->prefix_len))
1222                         continue;
1223                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1224                         return CLEANUP_PREFIX_RT_NOP;
1225
1226                 action = CLEANUP_PREFIX_RT_EXPIRE;
1227
1228                 spin_lock(&ifa->lock);
1229
1230                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1231                 /*
1232                  * Note: Because this address is
1233                  * not permanent, lifetime <
1234                  * LONG_MAX / HZ here.
1235                  */
1236                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1237                         *expires = ifa->tstamp + lifetime * HZ;
1238                 spin_unlock(&ifa->lock);
1239         }
1240
1241         return action;
1242 }
1243
1244 static void
1245 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1246                      bool del_rt, bool del_peer)
1247 {
1248         struct fib6_info *f6i;
1249
1250         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1251                                         ifp->prefix_len,
1252                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1253         if (f6i) {
1254                 if (del_rt)
1255                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1256                 else {
1257                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1258                                 fib6_set_expires(f6i, expires);
1259                         fib6_info_release(f6i);
1260                 }
1261         }
1262 }
1263
1264
1265 /* This function wants to get referenced ifp and releases it before return */
1266
1267 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1268 {
1269         int state;
1270         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1271         unsigned long expires;
1272
1273         ASSERT_RTNL();
1274
1275         spin_lock_bh(&ifp->lock);
1276         state = ifp->state;
1277         ifp->state = INET6_IFADDR_STATE_DEAD;
1278         spin_unlock_bh(&ifp->lock);
1279
1280         if (state == INET6_IFADDR_STATE_DEAD)
1281                 goto out;
1282
1283         spin_lock_bh(&addrconf_hash_lock);
1284         hlist_del_init_rcu(&ifp->addr_lst);
1285         spin_unlock_bh(&addrconf_hash_lock);
1286
1287         write_lock_bh(&ifp->idev->lock);
1288
1289         if (ifp->flags&IFA_F_TEMPORARY) {
1290                 list_del(&ifp->tmp_list);
1291                 if (ifp->ifpub) {
1292                         in6_ifa_put(ifp->ifpub);
1293                         ifp->ifpub = NULL;
1294                 }
1295                 __in6_ifa_put(ifp);
1296         }
1297
1298         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1299                 action = check_cleanup_prefix_route(ifp, &expires);
1300
1301         list_del_rcu(&ifp->if_list);
1302         __in6_ifa_put(ifp);
1303
1304         write_unlock_bh(&ifp->idev->lock);
1305
1306         addrconf_del_dad_work(ifp);
1307
1308         ipv6_ifa_notify(RTM_DELADDR, ifp);
1309
1310         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1311
1312         if (action != CLEANUP_PREFIX_RT_NOP) {
1313                 cleanup_prefix_route(ifp, expires,
1314                         action == CLEANUP_PREFIX_RT_DEL, false);
1315         }
1316
1317         /* clean up prefsrc entries */
1318         rt6_remove_prefsrc(ifp);
1319 out:
1320         in6_ifa_put(ifp);
1321 }
1322
1323 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1324 {
1325         struct inet6_dev *idev = ifp->idev;
1326         unsigned long tmp_tstamp, age;
1327         unsigned long regen_advance;
1328         unsigned long now = jiffies;
1329         s32 cnf_temp_preferred_lft;
1330         struct inet6_ifaddr *ift;
1331         struct ifa6_config cfg;
1332         long max_desync_factor;
1333         struct in6_addr addr;
1334         int ret = 0;
1335
1336         write_lock_bh(&idev->lock);
1337
1338 retry:
1339         in6_dev_hold(idev);
1340         if (idev->cnf.use_tempaddr <= 0) {
1341                 write_unlock_bh(&idev->lock);
1342                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1343                 in6_dev_put(idev);
1344                 ret = -1;
1345                 goto out;
1346         }
1347         spin_lock_bh(&ifp->lock);
1348         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1349                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1350                 spin_unlock_bh(&ifp->lock);
1351                 write_unlock_bh(&idev->lock);
1352                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1353                         __func__);
1354                 in6_dev_put(idev);
1355                 ret = -1;
1356                 goto out;
1357         }
1358         in6_ifa_hold(ifp);
1359         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1360         ipv6_gen_rnd_iid(&addr);
1361
1362         age = (now - ifp->tstamp) / HZ;
1363
1364         regen_advance = idev->cnf.regen_max_retry *
1365                         idev->cnf.dad_transmits *
1366                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1367
1368         /* recalculate max_desync_factor each time and update
1369          * idev->desync_factor if it's larger
1370          */
1371         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1372         max_desync_factor = min_t(__u32,
1373                                   idev->cnf.max_desync_factor,
1374                                   cnf_temp_preferred_lft - regen_advance);
1375
1376         if (unlikely(idev->desync_factor > max_desync_factor)) {
1377                 if (max_desync_factor > 0) {
1378                         get_random_bytes(&idev->desync_factor,
1379                                          sizeof(idev->desync_factor));
1380                         idev->desync_factor %= max_desync_factor;
1381                 } else {
1382                         idev->desync_factor = 0;
1383                 }
1384         }
1385
1386         memset(&cfg, 0, sizeof(cfg));
1387         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1388                               idev->cnf.temp_valid_lft + age);
1389         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1390         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1391
1392         cfg.plen = ifp->prefix_len;
1393         tmp_tstamp = ifp->tstamp;
1394         spin_unlock_bh(&ifp->lock);
1395
1396         write_unlock_bh(&idev->lock);
1397
1398         /* A temporary address is created only if this calculated Preferred
1399          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1400          * an implementation must not create a temporary address with a zero
1401          * Preferred Lifetime.
1402          * Use age calculation as in addrconf_verify to avoid unnecessary
1403          * temporary addresses being generated.
1404          */
1405         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1406         if (cfg.preferred_lft <= regen_advance + age) {
1407                 in6_ifa_put(ifp);
1408                 in6_dev_put(idev);
1409                 ret = -1;
1410                 goto out;
1411         }
1412
1413         cfg.ifa_flags = IFA_F_TEMPORARY;
1414         /* set in addrconf_prefix_rcv() */
1415         if (ifp->flags & IFA_F_OPTIMISTIC)
1416                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1417
1418         cfg.pfx = &addr;
1419         cfg.scope = ipv6_addr_scope(cfg.pfx);
1420
1421         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1422         if (IS_ERR(ift)) {
1423                 in6_ifa_put(ifp);
1424                 in6_dev_put(idev);
1425                 pr_info("%s: retry temporary address regeneration\n", __func__);
1426                 write_lock_bh(&idev->lock);
1427                 goto retry;
1428         }
1429
1430         spin_lock_bh(&ift->lock);
1431         ift->ifpub = ifp;
1432         ift->cstamp = now;
1433         ift->tstamp = tmp_tstamp;
1434         spin_unlock_bh(&ift->lock);
1435
1436         addrconf_dad_start(ift);
1437         in6_ifa_put(ift);
1438         in6_dev_put(idev);
1439 out:
1440         return ret;
1441 }
1442
1443 /*
1444  *      Choose an appropriate source address (RFC3484)
1445  */
1446 enum {
1447         IPV6_SADDR_RULE_INIT = 0,
1448         IPV6_SADDR_RULE_LOCAL,
1449         IPV6_SADDR_RULE_SCOPE,
1450         IPV6_SADDR_RULE_PREFERRED,
1451 #ifdef CONFIG_IPV6_MIP6
1452         IPV6_SADDR_RULE_HOA,
1453 #endif
1454         IPV6_SADDR_RULE_OIF,
1455         IPV6_SADDR_RULE_LABEL,
1456         IPV6_SADDR_RULE_PRIVACY,
1457         IPV6_SADDR_RULE_ORCHID,
1458         IPV6_SADDR_RULE_PREFIX,
1459 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1460         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1461 #endif
1462         IPV6_SADDR_RULE_MAX
1463 };
1464
1465 struct ipv6_saddr_score {
1466         int                     rule;
1467         int                     addr_type;
1468         struct inet6_ifaddr     *ifa;
1469         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1470         int                     scopedist;
1471         int                     matchlen;
1472 };
1473
1474 struct ipv6_saddr_dst {
1475         const struct in6_addr *addr;
1476         int ifindex;
1477         int scope;
1478         int label;
1479         unsigned int prefs;
1480 };
1481
1482 static inline int ipv6_saddr_preferred(int type)
1483 {
1484         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1485                 return 1;
1486         return 0;
1487 }
1488
1489 static bool ipv6_use_optimistic_addr(struct net *net,
1490                                      struct inet6_dev *idev)
1491 {
1492 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1493         if (!idev)
1494                 return false;
1495         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1496                 return false;
1497         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1498                 return false;
1499
1500         return true;
1501 #else
1502         return false;
1503 #endif
1504 }
1505
1506 static bool ipv6_allow_optimistic_dad(struct net *net,
1507                                       struct inet6_dev *idev)
1508 {
1509 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1510         if (!idev)
1511                 return false;
1512         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1513                 return false;
1514
1515         return true;
1516 #else
1517         return false;
1518 #endif
1519 }
1520
1521 static int ipv6_get_saddr_eval(struct net *net,
1522                                struct ipv6_saddr_score *score,
1523                                struct ipv6_saddr_dst *dst,
1524                                int i)
1525 {
1526         int ret;
1527
1528         if (i <= score->rule) {
1529                 switch (i) {
1530                 case IPV6_SADDR_RULE_SCOPE:
1531                         ret = score->scopedist;
1532                         break;
1533                 case IPV6_SADDR_RULE_PREFIX:
1534                         ret = score->matchlen;
1535                         break;
1536                 default:
1537                         ret = !!test_bit(i, score->scorebits);
1538                 }
1539                 goto out;
1540         }
1541
1542         switch (i) {
1543         case IPV6_SADDR_RULE_INIT:
1544                 /* Rule 0: remember if hiscore is not ready yet */
1545                 ret = !!score->ifa;
1546                 break;
1547         case IPV6_SADDR_RULE_LOCAL:
1548                 /* Rule 1: Prefer same address */
1549                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1550                 break;
1551         case IPV6_SADDR_RULE_SCOPE:
1552                 /* Rule 2: Prefer appropriate scope
1553                  *
1554                  *      ret
1555                  *       ^
1556                  *    -1 |  d 15
1557                  *    ---+--+-+---> scope
1558                  *       |
1559                  *       |             d is scope of the destination.
1560                  *  B-d  |  \
1561                  *       |   \      <- smaller scope is better if
1562                  *  B-15 |    \        if scope is enough for destination.
1563                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1564                  * d-C-1 | /
1565                  *       |/         <- greater is better
1566                  *   -C  /             if scope is not enough for destination.
1567                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1568                  *
1569                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1570                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1571                  * Assume B = 0 and we get C > 29.
1572                  */
1573                 ret = __ipv6_addr_src_scope(score->addr_type);
1574                 if (ret >= dst->scope)
1575                         ret = -ret;
1576                 else
1577                         ret -= 128;     /* 30 is enough */
1578                 score->scopedist = ret;
1579                 break;
1580         case IPV6_SADDR_RULE_PREFERRED:
1581             {
1582                 /* Rule 3: Avoid deprecated and optimistic addresses */
1583                 u8 avoid = IFA_F_DEPRECATED;
1584
1585                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1586                         avoid |= IFA_F_OPTIMISTIC;
1587                 ret = ipv6_saddr_preferred(score->addr_type) ||
1588                       !(score->ifa->flags & avoid);
1589                 break;
1590             }
1591 #ifdef CONFIG_IPV6_MIP6
1592         case IPV6_SADDR_RULE_HOA:
1593             {
1594                 /* Rule 4: Prefer home address */
1595                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1596                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1597                 break;
1598             }
1599 #endif
1600         case IPV6_SADDR_RULE_OIF:
1601                 /* Rule 5: Prefer outgoing interface */
1602                 ret = (!dst->ifindex ||
1603                        dst->ifindex == score->ifa->idev->dev->ifindex);
1604                 break;
1605         case IPV6_SADDR_RULE_LABEL:
1606                 /* Rule 6: Prefer matching label */
1607                 ret = ipv6_addr_label(net,
1608                                       &score->ifa->addr, score->addr_type,
1609                                       score->ifa->idev->dev->ifindex) == dst->label;
1610                 break;
1611         case IPV6_SADDR_RULE_PRIVACY:
1612             {
1613                 /* Rule 7: Prefer public address
1614                  * Note: prefer temporary address if use_tempaddr >= 2
1615                  */
1616                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1617                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1618                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1619                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1620                 break;
1621             }
1622         case IPV6_SADDR_RULE_ORCHID:
1623                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1624                  *          non-ORCHID vs non-ORCHID
1625                  */
1626                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1627                         ipv6_addr_orchid(dst->addr));
1628                 break;
1629         case IPV6_SADDR_RULE_PREFIX:
1630                 /* Rule 8: Use longest matching prefix */
1631                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1632                 if (ret > score->ifa->prefix_len)
1633                         ret = score->ifa->prefix_len;
1634                 score->matchlen = ret;
1635                 break;
1636 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1637         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1638                 /* Optimistic addresses still have lower precedence than other
1639                  * preferred addresses.
1640                  */
1641                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1642                 break;
1643 #endif
1644         default:
1645                 ret = 0;
1646         }
1647
1648         if (ret)
1649                 __set_bit(i, score->scorebits);
1650         score->rule = i;
1651 out:
1652         return ret;
1653 }
1654
1655 static int __ipv6_dev_get_saddr(struct net *net,
1656                                 struct ipv6_saddr_dst *dst,
1657                                 struct inet6_dev *idev,
1658                                 struct ipv6_saddr_score *scores,
1659                                 int hiscore_idx)
1660 {
1661         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1662
1663         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1664                 int i;
1665
1666                 /*
1667                  * - Tentative Address (RFC2462 section 5.4)
1668                  *  - A tentative address is not considered
1669                  *    "assigned to an interface" in the traditional
1670                  *    sense, unless it is also flagged as optimistic.
1671                  * - Candidate Source Address (section 4)
1672                  *  - In any case, anycast addresses, multicast
1673                  *    addresses, and the unspecified address MUST
1674                  *    NOT be included in a candidate set.
1675                  */
1676                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1677                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1678                         continue;
1679
1680                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1681
1682                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1683                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1684                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1685                                             idev->dev->name);
1686                         continue;
1687                 }
1688
1689                 score->rule = -1;
1690                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1691
1692                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1693                         int minihiscore, miniscore;
1694
1695                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1696                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1697
1698                         if (minihiscore > miniscore) {
1699                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1700                                     score->scopedist > 0) {
1701                                         /*
1702                                          * special case:
1703                                          * each remaining entry
1704                                          * has too small (not enough)
1705                                          * scope, because ifa entries
1706                                          * are sorted by their scope
1707                                          * values.
1708                                          */
1709                                         goto out;
1710                                 }
1711                                 break;
1712                         } else if (minihiscore < miniscore) {
1713                                 swap(hiscore, score);
1714                                 hiscore_idx = 1 - hiscore_idx;
1715
1716                                 /* restore our iterator */
1717                                 score->ifa = hiscore->ifa;
1718
1719                                 break;
1720                         }
1721                 }
1722         }
1723 out:
1724         return hiscore_idx;
1725 }
1726
1727 static int ipv6_get_saddr_master(struct net *net,
1728                                  const struct net_device *dst_dev,
1729                                  const struct net_device *master,
1730                                  struct ipv6_saddr_dst *dst,
1731                                  struct ipv6_saddr_score *scores,
1732                                  int hiscore_idx)
1733 {
1734         struct inet6_dev *idev;
1735
1736         idev = __in6_dev_get(dst_dev);
1737         if (idev)
1738                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1739                                                    scores, hiscore_idx);
1740
1741         idev = __in6_dev_get(master);
1742         if (idev)
1743                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1744                                                    scores, hiscore_idx);
1745
1746         return hiscore_idx;
1747 }
1748
1749 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1750                        const struct in6_addr *daddr, unsigned int prefs,
1751                        struct in6_addr *saddr)
1752 {
1753         struct ipv6_saddr_score scores[2], *hiscore;
1754         struct ipv6_saddr_dst dst;
1755         struct inet6_dev *idev;
1756         struct net_device *dev;
1757         int dst_type;
1758         bool use_oif_addr = false;
1759         int hiscore_idx = 0;
1760         int ret = 0;
1761
1762         dst_type = __ipv6_addr_type(daddr);
1763         dst.addr = daddr;
1764         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1765         dst.scope = __ipv6_addr_src_scope(dst_type);
1766         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1767         dst.prefs = prefs;
1768
1769         scores[hiscore_idx].rule = -1;
1770         scores[hiscore_idx].ifa = NULL;
1771
1772         rcu_read_lock();
1773
1774         /* Candidate Source Address (section 4)
1775          *  - multicast and link-local destination address,
1776          *    the set of candidate source address MUST only
1777          *    include addresses assigned to interfaces
1778          *    belonging to the same link as the outgoing
1779          *    interface.
1780          * (- For site-local destination addresses, the
1781          *    set of candidate source addresses MUST only
1782          *    include addresses assigned to interfaces
1783          *    belonging to the same site as the outgoing
1784          *    interface.)
1785          *  - "It is RECOMMENDED that the candidate source addresses
1786          *    be the set of unicast addresses assigned to the
1787          *    interface that will be used to send to the destination
1788          *    (the 'outgoing' interface)." (RFC 6724)
1789          */
1790         if (dst_dev) {
1791                 idev = __in6_dev_get(dst_dev);
1792                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1793                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1794                     (idev && idev->cnf.use_oif_addrs_only)) {
1795                         use_oif_addr = true;
1796                 }
1797         }
1798
1799         if (use_oif_addr) {
1800                 if (idev)
1801                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1802         } else {
1803                 const struct net_device *master;
1804                 int master_idx = 0;
1805
1806                 /* if dst_dev exists and is enslaved to an L3 device, then
1807                  * prefer addresses from dst_dev and then the master over
1808                  * any other enslaved devices in the L3 domain.
1809                  */
1810                 master = l3mdev_master_dev_rcu(dst_dev);
1811                 if (master) {
1812                         master_idx = master->ifindex;
1813
1814                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1815                                                             master, &dst,
1816                                                             scores, hiscore_idx);
1817
1818                         if (scores[hiscore_idx].ifa)
1819                                 goto out;
1820                 }
1821
1822                 for_each_netdev_rcu(net, dev) {
1823                         /* only consider addresses on devices in the
1824                          * same L3 domain
1825                          */
1826                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1827                                 continue;
1828                         idev = __in6_dev_get(dev);
1829                         if (!idev)
1830                                 continue;
1831                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1832                 }
1833         }
1834
1835 out:
1836         hiscore = &scores[hiscore_idx];
1837         if (!hiscore->ifa)
1838                 ret = -EADDRNOTAVAIL;
1839         else
1840                 *saddr = hiscore->ifa->addr;
1841
1842         rcu_read_unlock();
1843         return ret;
1844 }
1845 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1846
1847 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1848                               u32 banned_flags)
1849 {
1850         struct inet6_ifaddr *ifp;
1851         int err = -EADDRNOTAVAIL;
1852
1853         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1854                 if (ifp->scope > IFA_LINK)
1855                         break;
1856                 if (ifp->scope == IFA_LINK &&
1857                     !(ifp->flags & banned_flags)) {
1858                         *addr = ifp->addr;
1859                         err = 0;
1860                         break;
1861                 }
1862         }
1863         return err;
1864 }
1865
1866 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1867                     u32 banned_flags)
1868 {
1869         struct inet6_dev *idev;
1870         int err = -EADDRNOTAVAIL;
1871
1872         rcu_read_lock();
1873         idev = __in6_dev_get(dev);
1874         if (idev) {
1875                 read_lock_bh(&idev->lock);
1876                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1877                 read_unlock_bh(&idev->lock);
1878         }
1879         rcu_read_unlock();
1880         return err;
1881 }
1882
1883 static int ipv6_count_addresses(const struct inet6_dev *idev)
1884 {
1885         const struct inet6_ifaddr *ifp;
1886         int cnt = 0;
1887
1888         rcu_read_lock();
1889         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1890                 cnt++;
1891         rcu_read_unlock();
1892         return cnt;
1893 }
1894
1895 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1896                   const struct net_device *dev, int strict)
1897 {
1898         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1899                                        strict, IFA_F_TENTATIVE);
1900 }
1901 EXPORT_SYMBOL(ipv6_chk_addr);
1902
1903 /* device argument is used to find the L3 domain of interest. If
1904  * skip_dev_check is set, then the ifp device is not checked against
1905  * the passed in dev argument. So the 2 cases for addresses checks are:
1906  *   1. does the address exist in the L3 domain that dev is part of
1907  *      (skip_dev_check = true), or
1908  *
1909  *   2. does the address exist on the specific device
1910  *      (skip_dev_check = false)
1911  */
1912 static struct net_device *
1913 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1914                           const struct net_device *dev, bool skip_dev_check,
1915                           int strict, u32 banned_flags)
1916 {
1917         unsigned int hash = inet6_addr_hash(net, addr);
1918         struct net_device *l3mdev, *ndev;
1919         struct inet6_ifaddr *ifp;
1920         u32 ifp_flags;
1921
1922         rcu_read_lock();
1923
1924         l3mdev = l3mdev_master_dev_rcu(dev);
1925         if (skip_dev_check)
1926                 dev = NULL;
1927
1928         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1929                 ndev = ifp->idev->dev;
1930                 if (!net_eq(dev_net(ndev), net))
1931                         continue;
1932
1933                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1934                         continue;
1935
1936                 /* Decouple optimistic from tentative for evaluation here.
1937                  * Ban optimistic addresses explicitly, when required.
1938                  */
1939                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1940                             ? (ifp->flags&~IFA_F_TENTATIVE)
1941                             : ifp->flags;
1942                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1943                     !(ifp_flags&banned_flags) &&
1944                     (!dev || ndev == dev ||
1945                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1946                         rcu_read_unlock();
1947                         return ndev;
1948                 }
1949         }
1950
1951         rcu_read_unlock();
1952         return NULL;
1953 }
1954
1955 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1956                             const struct net_device *dev, bool skip_dev_check,
1957                             int strict, u32 banned_flags)
1958 {
1959         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1960                                          strict, banned_flags) ? 1 : 0;
1961 }
1962 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1963
1964
1965 /* Compares an address/prefix_len with addresses on device @dev.
1966  * If one is found it returns true.
1967  */
1968 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1969         const unsigned int prefix_len, struct net_device *dev)
1970 {
1971         const struct inet6_ifaddr *ifa;
1972         const struct inet6_dev *idev;
1973         bool ret = false;
1974
1975         rcu_read_lock();
1976         idev = __in6_dev_get(dev);
1977         if (idev) {
1978                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1979                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1980                         if (ret)
1981                                 break;
1982                 }
1983         }
1984         rcu_read_unlock();
1985
1986         return ret;
1987 }
1988 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1989
1990 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1991 {
1992         const struct inet6_ifaddr *ifa;
1993         const struct inet6_dev *idev;
1994         int     onlink;
1995
1996         onlink = 0;
1997         rcu_read_lock();
1998         idev = __in6_dev_get(dev);
1999         if (idev) {
2000                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2001                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
2002                                                    ifa->prefix_len);
2003                         if (onlink)
2004                                 break;
2005                 }
2006         }
2007         rcu_read_unlock();
2008         return onlink;
2009 }
2010 EXPORT_SYMBOL(ipv6_chk_prefix);
2011
2012 /**
2013  * ipv6_dev_find - find the first device with a given source address.
2014  * @net: the net namespace
2015  * @addr: the source address
2016  * @dev: used to find the L3 domain of interest
2017  *
2018  * The caller should be protected by RCU, or RTNL.
2019  */
2020 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2021                                  struct net_device *dev)
2022 {
2023         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2024                                          IFA_F_TENTATIVE);
2025 }
2026 EXPORT_SYMBOL(ipv6_dev_find);
2027
2028 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2029                                      struct net_device *dev, int strict)
2030 {
2031         unsigned int hash = inet6_addr_hash(net, addr);
2032         struct inet6_ifaddr *ifp, *result = NULL;
2033
2034         rcu_read_lock();
2035         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2036                 if (!net_eq(dev_net(ifp->idev->dev), net))
2037                         continue;
2038                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2039                         if (!dev || ifp->idev->dev == dev ||
2040                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2041                                 result = ifp;
2042                                 in6_ifa_hold(ifp);
2043                                 break;
2044                         }
2045                 }
2046         }
2047         rcu_read_unlock();
2048
2049         return result;
2050 }
2051
2052 /* Gets referenced address, destroys ifaddr */
2053
2054 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2055 {
2056         if (dad_failed)
2057                 ifp->flags |= IFA_F_DADFAILED;
2058
2059         if (ifp->flags&IFA_F_TEMPORARY) {
2060                 struct inet6_ifaddr *ifpub;
2061                 spin_lock_bh(&ifp->lock);
2062                 ifpub = ifp->ifpub;
2063                 if (ifpub) {
2064                         in6_ifa_hold(ifpub);
2065                         spin_unlock_bh(&ifp->lock);
2066                         ipv6_create_tempaddr(ifpub, true);
2067                         in6_ifa_put(ifpub);
2068                 } else {
2069                         spin_unlock_bh(&ifp->lock);
2070                 }
2071                 ipv6_del_addr(ifp);
2072         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2073                 spin_lock_bh(&ifp->lock);
2074                 addrconf_del_dad_work(ifp);
2075                 ifp->flags |= IFA_F_TENTATIVE;
2076                 if (dad_failed)
2077                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2078                 spin_unlock_bh(&ifp->lock);
2079                 if (dad_failed)
2080                         ipv6_ifa_notify(0, ifp);
2081                 in6_ifa_put(ifp);
2082         } else {
2083                 ipv6_del_addr(ifp);
2084         }
2085 }
2086
2087 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2088 {
2089         int err = -ENOENT;
2090
2091         spin_lock_bh(&ifp->lock);
2092         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2093                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2094                 err = 0;
2095         }
2096         spin_unlock_bh(&ifp->lock);
2097
2098         return err;
2099 }
2100
2101 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2102 {
2103         struct inet6_dev *idev = ifp->idev;
2104         struct net *net = dev_net(ifp->idev->dev);
2105
2106         if (addrconf_dad_end(ifp)) {
2107                 in6_ifa_put(ifp);
2108                 return;
2109         }
2110
2111         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2112                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2113
2114         spin_lock_bh(&ifp->lock);
2115
2116         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2117                 struct in6_addr new_addr;
2118                 struct inet6_ifaddr *ifp2;
2119                 int retries = ifp->stable_privacy_retry + 1;
2120                 struct ifa6_config cfg = {
2121                         .pfx = &new_addr,
2122                         .plen = ifp->prefix_len,
2123                         .ifa_flags = ifp->flags,
2124                         .valid_lft = ifp->valid_lft,
2125                         .preferred_lft = ifp->prefered_lft,
2126                         .scope = ifp->scope,
2127                 };
2128
2129                 if (retries > net->ipv6.sysctl.idgen_retries) {
2130                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2131                                              ifp->idev->dev->name);
2132                         goto errdad;
2133                 }
2134
2135                 new_addr = ifp->addr;
2136                 if (ipv6_generate_stable_address(&new_addr, retries,
2137                                                  idev))
2138                         goto errdad;
2139
2140                 spin_unlock_bh(&ifp->lock);
2141
2142                 if (idev->cnf.max_addresses &&
2143                     ipv6_count_addresses(idev) >=
2144                     idev->cnf.max_addresses)
2145                         goto lock_errdad;
2146
2147                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2148                                      ifp->idev->dev->name);
2149
2150                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2151                 if (IS_ERR(ifp2))
2152                         goto lock_errdad;
2153
2154                 spin_lock_bh(&ifp2->lock);
2155                 ifp2->stable_privacy_retry = retries;
2156                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2157                 spin_unlock_bh(&ifp2->lock);
2158
2159                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2160                 in6_ifa_put(ifp2);
2161 lock_errdad:
2162                 spin_lock_bh(&ifp->lock);
2163         }
2164
2165 errdad:
2166         /* transition from _POSTDAD to _ERRDAD */
2167         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2168         spin_unlock_bh(&ifp->lock);
2169
2170         addrconf_mod_dad_work(ifp, 0);
2171         in6_ifa_put(ifp);
2172 }
2173
2174 /* Join to solicited addr multicast group.
2175  * caller must hold RTNL */
2176 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2177 {
2178         struct in6_addr maddr;
2179
2180         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2181                 return;
2182
2183         addrconf_addr_solict_mult(addr, &maddr);
2184         ipv6_dev_mc_inc(dev, &maddr);
2185 }
2186
2187 /* caller must hold RTNL */
2188 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2189 {
2190         struct in6_addr maddr;
2191
2192         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2193                 return;
2194
2195         addrconf_addr_solict_mult(addr, &maddr);
2196         __ipv6_dev_mc_dec(idev, &maddr);
2197 }
2198
2199 /* caller must hold RTNL */
2200 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2201 {
2202         struct in6_addr addr;
2203
2204         if (ifp->prefix_len >= 127) /* RFC 6164 */
2205                 return;
2206         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2207         if (ipv6_addr_any(&addr))
2208                 return;
2209         __ipv6_dev_ac_inc(ifp->idev, &addr);
2210 }
2211
2212 /* caller must hold RTNL */
2213 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2214 {
2215         struct in6_addr addr;
2216
2217         if (ifp->prefix_len >= 127) /* RFC 6164 */
2218                 return;
2219         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2220         if (ipv6_addr_any(&addr))
2221                 return;
2222         __ipv6_dev_ac_dec(ifp->idev, &addr);
2223 }
2224
2225 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2226 {
2227         switch (dev->addr_len) {
2228         case ETH_ALEN:
2229                 memcpy(eui, dev->dev_addr, 3);
2230                 eui[3] = 0xFF;
2231                 eui[4] = 0xFE;
2232                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2233                 break;
2234         case EUI64_ADDR_LEN:
2235                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2236                 eui[0] ^= 2;
2237                 break;
2238         default:
2239                 return -1;
2240         }
2241
2242         return 0;
2243 }
2244
2245 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2246 {
2247         union fwnet_hwaddr *ha;
2248
2249         if (dev->addr_len != FWNET_ALEN)
2250                 return -1;
2251
2252         ha = (union fwnet_hwaddr *)dev->dev_addr;
2253
2254         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2255         eui[0] ^= 2;
2256         return 0;
2257 }
2258
2259 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2260 {
2261         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2262         if (dev->addr_len != ARCNET_ALEN)
2263                 return -1;
2264         memset(eui, 0, 7);
2265         eui[7] = *(u8 *)dev->dev_addr;
2266         return 0;
2267 }
2268
2269 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2270 {
2271         if (dev->addr_len != INFINIBAND_ALEN)
2272                 return -1;
2273         memcpy(eui, dev->dev_addr + 12, 8);
2274         eui[0] |= 2;
2275         return 0;
2276 }
2277
2278 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2279 {
2280         if (addr == 0)
2281                 return -1;
2282         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2283                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2284                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2285                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2286                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2287                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2288         eui[1] = 0;
2289         eui[2] = 0x5E;
2290         eui[3] = 0xFE;
2291         memcpy(eui + 4, &addr, 4);
2292         return 0;
2293 }
2294
2295 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2296 {
2297         if (dev->priv_flags & IFF_ISATAP)
2298                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2299         return -1;
2300 }
2301
2302 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2303 {
2304         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2305 }
2306
2307 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2308 {
2309         memcpy(eui, dev->perm_addr, 3);
2310         memcpy(eui + 5, dev->perm_addr + 3, 3);
2311         eui[3] = 0xFF;
2312         eui[4] = 0xFE;
2313         eui[0] ^= 2;
2314         return 0;
2315 }
2316
2317 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2318 {
2319         switch (dev->type) {
2320         case ARPHRD_ETHER:
2321         case ARPHRD_FDDI:
2322                 return addrconf_ifid_eui48(eui, dev);
2323         case ARPHRD_ARCNET:
2324                 return addrconf_ifid_arcnet(eui, dev);
2325         case ARPHRD_INFINIBAND:
2326                 return addrconf_ifid_infiniband(eui, dev);
2327         case ARPHRD_SIT:
2328                 return addrconf_ifid_sit(eui, dev);
2329         case ARPHRD_IPGRE:
2330         case ARPHRD_TUNNEL:
2331                 return addrconf_ifid_gre(eui, dev);
2332         case ARPHRD_6LOWPAN:
2333                 return addrconf_ifid_6lowpan(eui, dev);
2334         case ARPHRD_IEEE1394:
2335                 return addrconf_ifid_ieee1394(eui, dev);
2336         case ARPHRD_TUNNEL6:
2337         case ARPHRD_IP6GRE:
2338         case ARPHRD_RAWIP:
2339                 return addrconf_ifid_ip6tnl(eui, dev);
2340         }
2341         return -1;
2342 }
2343
2344 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2345 {
2346         int err = -1;
2347         struct inet6_ifaddr *ifp;
2348
2349         read_lock_bh(&idev->lock);
2350         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2351                 if (ifp->scope > IFA_LINK)
2352                         break;
2353                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2354                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2355                         err = 0;
2356                         break;
2357                 }
2358         }
2359         read_unlock_bh(&idev->lock);
2360         return err;
2361 }
2362
2363 /* Generation of a randomized Interface Identifier
2364  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2365  */
2366
2367 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2368 {
2369 regen:
2370         get_random_bytes(&addr->s6_addr[8], 8);
2371
2372         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2373          * check if generated address is not inappropriate:
2374          *
2375          * - Reserved IPv6 Interface Identifiers
2376          * - XXX: already assigned to an address on the device
2377          */
2378
2379         /* Subnet-router anycast: 0000:0000:0000:0000 */
2380         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2381                 goto regen;
2382
2383         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2384          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2385          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2386          */
2387         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2388             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2389                 goto regen;
2390
2391         /* Reserved subnet anycast addresses */
2392         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2393             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2394                 goto regen;
2395 }
2396
2397 /*
2398  *      Add prefix route.
2399  */
2400
2401 static void
2402 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2403                       struct net_device *dev, unsigned long expires,
2404                       u32 flags, gfp_t gfp_flags)
2405 {
2406         struct fib6_config cfg = {
2407                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2408                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2409                 .fc_ifindex = dev->ifindex,
2410                 .fc_expires = expires,
2411                 .fc_dst_len = plen,
2412                 .fc_flags = RTF_UP | flags,
2413                 .fc_nlinfo.nl_net = dev_net(dev),
2414                 .fc_protocol = RTPROT_KERNEL,
2415                 .fc_type = RTN_UNICAST,
2416         };
2417
2418         cfg.fc_dst = *pfx;
2419
2420         /* Prevent useless cloning on PtP SIT.
2421            This thing is done here expecting that the whole
2422            class of non-broadcast devices need not cloning.
2423          */
2424 #if IS_ENABLED(CONFIG_IPV6_SIT)
2425         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2426                 cfg.fc_flags |= RTF_NONEXTHOP;
2427 #endif
2428
2429         ip6_route_add(&cfg, gfp_flags, NULL);
2430 }
2431
2432
2433 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2434                                                   int plen,
2435                                                   const struct net_device *dev,
2436                                                   u32 flags, u32 noflags,
2437                                                   bool no_gw)
2438 {
2439         struct fib6_node *fn;
2440         struct fib6_info *rt = NULL;
2441         struct fib6_table *table;
2442         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2443
2444         table = fib6_get_table(dev_net(dev), tb_id);
2445         if (!table)
2446                 return NULL;
2447
2448         rcu_read_lock();
2449         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2450         if (!fn)
2451                 goto out;
2452
2453         for_each_fib6_node_rt_rcu(fn) {
2454                 /* prefix routes only use builtin fib6_nh */
2455                 if (rt->nh)
2456                         continue;
2457
2458                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2459                         continue;
2460                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2461                         continue;
2462                 if ((rt->fib6_flags & flags) != flags)
2463                         continue;
2464                 if ((rt->fib6_flags & noflags) != 0)
2465                         continue;
2466                 if (!fib6_info_hold_safe(rt))
2467                         continue;
2468                 break;
2469         }
2470 out:
2471         rcu_read_unlock();
2472         return rt;
2473 }
2474
2475
2476 /* Create "default" multicast route to the interface */
2477
2478 static void addrconf_add_mroute(struct net_device *dev)
2479 {
2480         struct fib6_config cfg = {
2481                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2482                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2483                 .fc_ifindex = dev->ifindex,
2484                 .fc_dst_len = 8,
2485                 .fc_flags = RTF_UP,
2486                 .fc_type = RTN_MULTICAST,
2487                 .fc_nlinfo.nl_net = dev_net(dev),
2488                 .fc_protocol = RTPROT_KERNEL,
2489         };
2490
2491         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2492
2493         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2494 }
2495
2496 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2497 {
2498         struct inet6_dev *idev;
2499
2500         ASSERT_RTNL();
2501
2502         idev = ipv6_find_idev(dev);
2503         if (IS_ERR(idev))
2504                 return idev;
2505
2506         if (idev->cnf.disable_ipv6)
2507                 return ERR_PTR(-EACCES);
2508
2509         /* Add default multicast route */
2510         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2511                 addrconf_add_mroute(dev);
2512
2513         return idev;
2514 }
2515
2516 static void manage_tempaddrs(struct inet6_dev *idev,
2517                              struct inet6_ifaddr *ifp,
2518                              __u32 valid_lft, __u32 prefered_lft,
2519                              bool create, unsigned long now)
2520 {
2521         u32 flags;
2522         struct inet6_ifaddr *ift;
2523
2524         read_lock_bh(&idev->lock);
2525         /* update all temporary addresses in the list */
2526         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2527                 int age, max_valid, max_prefered;
2528
2529                 if (ifp != ift->ifpub)
2530                         continue;
2531
2532                 /* RFC 4941 section 3.3:
2533                  * If a received option will extend the lifetime of a public
2534                  * address, the lifetimes of temporary addresses should
2535                  * be extended, subject to the overall constraint that no
2536                  * temporary addresses should ever remain "valid" or "preferred"
2537                  * for a time longer than (TEMP_VALID_LIFETIME) or
2538                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2539                  */
2540                 age = (now - ift->cstamp) / HZ;
2541                 max_valid = idev->cnf.temp_valid_lft - age;
2542                 if (max_valid < 0)
2543                         max_valid = 0;
2544
2545                 max_prefered = idev->cnf.temp_prefered_lft -
2546                                idev->desync_factor - age;
2547                 if (max_prefered < 0)
2548                         max_prefered = 0;
2549
2550                 if (valid_lft > max_valid)
2551                         valid_lft = max_valid;
2552
2553                 if (prefered_lft > max_prefered)
2554                         prefered_lft = max_prefered;
2555
2556                 spin_lock(&ift->lock);
2557                 flags = ift->flags;
2558                 ift->valid_lft = valid_lft;
2559                 ift->prefered_lft = prefered_lft;
2560                 ift->tstamp = now;
2561                 if (prefered_lft > 0)
2562                         ift->flags &= ~IFA_F_DEPRECATED;
2563
2564                 spin_unlock(&ift->lock);
2565                 if (!(flags&IFA_F_TENTATIVE))
2566                         ipv6_ifa_notify(0, ift);
2567         }
2568
2569         if ((create || list_empty(&idev->tempaddr_list)) &&
2570             idev->cnf.use_tempaddr > 0) {
2571                 /* When a new public address is created as described
2572                  * in [ADDRCONF], also create a new temporary address.
2573                  * Also create a temporary address if it's enabled but
2574                  * no temporary address currently exists.
2575                  */
2576                 read_unlock_bh(&idev->lock);
2577                 ipv6_create_tempaddr(ifp, false);
2578         } else {
2579                 read_unlock_bh(&idev->lock);
2580         }
2581 }
2582
2583 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2584 {
2585         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2586                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2587 }
2588
2589 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2590                                  const struct prefix_info *pinfo,
2591                                  struct inet6_dev *in6_dev,
2592                                  const struct in6_addr *addr, int addr_type,
2593                                  u32 addr_flags, bool sllao, bool tokenized,
2594                                  __u32 valid_lft, u32 prefered_lft)
2595 {
2596         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2597         int create = 0, update_lft = 0;
2598
2599         if (!ifp && valid_lft) {
2600                 int max_addresses = in6_dev->cnf.max_addresses;
2601                 struct ifa6_config cfg = {
2602                         .pfx = addr,
2603                         .plen = pinfo->prefix_len,
2604                         .ifa_flags = addr_flags,
2605                         .valid_lft = valid_lft,
2606                         .preferred_lft = prefered_lft,
2607                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2608                 };
2609
2610 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2611                 if ((net->ipv6.devconf_all->optimistic_dad ||
2612                      in6_dev->cnf.optimistic_dad) &&
2613                     !net->ipv6.devconf_all->forwarding && sllao)
2614                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2615 #endif
2616
2617                 /* Do not allow to create too much of autoconfigured
2618                  * addresses; this would be too easy way to crash kernel.
2619                  */
2620                 if (!max_addresses ||
2621                     ipv6_count_addresses(in6_dev) < max_addresses)
2622                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2623
2624                 if (IS_ERR_OR_NULL(ifp))
2625                         return -1;
2626
2627                 create = 1;
2628                 spin_lock_bh(&ifp->lock);
2629                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2630                 ifp->cstamp = jiffies;
2631                 ifp->tokenized = tokenized;
2632                 spin_unlock_bh(&ifp->lock);
2633                 addrconf_dad_start(ifp);
2634         }
2635
2636         if (ifp) {
2637                 u32 flags;
2638                 unsigned long now;
2639                 u32 stored_lft;
2640
2641                 /* update lifetime (RFC2462 5.5.3 e) */
2642                 spin_lock_bh(&ifp->lock);
2643                 now = jiffies;
2644                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2645                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2646                 else
2647                         stored_lft = 0;
2648                 if (!create && stored_lft) {
2649                         const u32 minimum_lft = min_t(u32,
2650                                 stored_lft, MIN_VALID_LIFETIME);
2651                         valid_lft = max(valid_lft, minimum_lft);
2652
2653                         /* RFC4862 Section 5.5.3e:
2654                          * "Note that the preferred lifetime of the
2655                          *  corresponding address is always reset to
2656                          *  the Preferred Lifetime in the received
2657                          *  Prefix Information option, regardless of
2658                          *  whether the valid lifetime is also reset or
2659                          *  ignored."
2660                          *
2661                          * So we should always update prefered_lft here.
2662                          */
2663                         update_lft = 1;
2664                 }
2665
2666                 if (update_lft) {
2667                         ifp->valid_lft = valid_lft;
2668                         ifp->prefered_lft = prefered_lft;
2669                         ifp->tstamp = now;
2670                         flags = ifp->flags;
2671                         ifp->flags &= ~IFA_F_DEPRECATED;
2672                         spin_unlock_bh(&ifp->lock);
2673
2674                         if (!(flags&IFA_F_TENTATIVE))
2675                                 ipv6_ifa_notify(0, ifp);
2676                 } else
2677                         spin_unlock_bh(&ifp->lock);
2678
2679                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2680                                  create, now);
2681
2682                 in6_ifa_put(ifp);
2683                 addrconf_verify();
2684         }
2685
2686         return 0;
2687 }
2688 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2689
2690 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2691 {
2692         struct prefix_info *pinfo;
2693         __u32 valid_lft;
2694         __u32 prefered_lft;
2695         int addr_type, err;
2696         u32 addr_flags = 0;
2697         struct inet6_dev *in6_dev;
2698         struct net *net = dev_net(dev);
2699
2700         pinfo = (struct prefix_info *) opt;
2701
2702         if (len < sizeof(struct prefix_info)) {
2703                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2704                 return;
2705         }
2706
2707         /*
2708          *      Validation checks ([ADDRCONF], page 19)
2709          */
2710
2711         addr_type = ipv6_addr_type(&pinfo->prefix);
2712
2713         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2714                 return;
2715
2716         valid_lft = ntohl(pinfo->valid);
2717         prefered_lft = ntohl(pinfo->prefered);
2718
2719         if (prefered_lft > valid_lft) {
2720                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2721                 return;
2722         }
2723
2724         in6_dev = in6_dev_get(dev);
2725
2726         if (!in6_dev) {
2727                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2728                                     dev->name);
2729                 return;
2730         }
2731
2732         /*
2733          *      Two things going on here:
2734          *      1) Add routes for on-link prefixes
2735          *      2) Configure prefixes with the auto flag set
2736          */
2737
2738         if (pinfo->onlink) {
2739                 struct fib6_info *rt;
2740                 unsigned long rt_expires;
2741
2742                 /* Avoid arithmetic overflow. Really, we could
2743                  * save rt_expires in seconds, likely valid_lft,
2744                  * but it would require division in fib gc, that it
2745                  * not good.
2746                  */
2747                 if (HZ > USER_HZ)
2748                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2749                 else
2750                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2751
2752                 if (addrconf_finite_timeout(rt_expires))
2753                         rt_expires *= HZ;
2754
2755                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2756                                                pinfo->prefix_len,
2757                                                dev,
2758                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2759                                                RTF_DEFAULT, true);
2760
2761                 if (rt) {
2762                         /* Autoconf prefix route */
2763                         if (valid_lft == 0) {
2764                                 ip6_del_rt(net, rt, false);
2765                                 rt = NULL;
2766                         } else if (addrconf_finite_timeout(rt_expires)) {
2767                                 /* not infinity */
2768                                 fib6_set_expires(rt, jiffies + rt_expires);
2769                         } else {
2770                                 fib6_clean_expires(rt);
2771                         }
2772                 } else if (valid_lft) {
2773                         clock_t expires = 0;
2774                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2775                         if (addrconf_finite_timeout(rt_expires)) {
2776                                 /* not infinity */
2777                                 flags |= RTF_EXPIRES;
2778                                 expires = jiffies_to_clock_t(rt_expires);
2779                         }
2780                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2781                                               0, dev, expires, flags,
2782                                               GFP_ATOMIC);
2783                 }
2784                 fib6_info_release(rt);
2785         }
2786
2787         /* Try to figure out our local address for this prefix */
2788
2789         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2790                 struct in6_addr addr;
2791                 bool tokenized = false, dev_addr_generated = false;
2792
2793                 if (pinfo->prefix_len == 64) {
2794                         memcpy(&addr, &pinfo->prefix, 8);
2795
2796                         if (!ipv6_addr_any(&in6_dev->token)) {
2797                                 read_lock_bh(&in6_dev->lock);
2798                                 memcpy(addr.s6_addr + 8,
2799                                        in6_dev->token.s6_addr + 8, 8);
2800                                 read_unlock_bh(&in6_dev->lock);
2801                                 tokenized = true;
2802                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2803                                    !ipv6_generate_stable_address(&addr, 0,
2804                                                                  in6_dev)) {
2805                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2806                                 goto ok;
2807                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2808                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2809                                 goto put;
2810                         } else {
2811                                 dev_addr_generated = true;
2812                         }
2813                         goto ok;
2814                 }
2815                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2816                                     pinfo->prefix_len);
2817                 goto put;
2818
2819 ok:
2820                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2821                                                    &addr, addr_type,
2822                                                    addr_flags, sllao,
2823                                                    tokenized, valid_lft,
2824                                                    prefered_lft);
2825                 if (err)
2826                         goto put;
2827
2828                 /* Ignore error case here because previous prefix add addr was
2829                  * successful which will be notified.
2830                  */
2831                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2832                                               addr_type, addr_flags, sllao,
2833                                               tokenized, valid_lft,
2834                                               prefered_lft,
2835                                               dev_addr_generated);
2836         }
2837         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2838 put:
2839         in6_dev_put(in6_dev);
2840 }
2841
2842 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2843                 struct in6_ifreq *ireq)
2844 {
2845         struct ip_tunnel_parm p = { };
2846         int err;
2847
2848         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2849                 return -EADDRNOTAVAIL;
2850
2851         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2852         p.iph.version = 4;
2853         p.iph.ihl = 5;
2854         p.iph.protocol = IPPROTO_IPV6;
2855         p.iph.ttl = 64;
2856
2857         if (!dev->netdev_ops->ndo_tunnel_ctl)
2858                 return -EOPNOTSUPP;
2859         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2860         if (err)
2861                 return err;
2862
2863         dev = __dev_get_by_name(net, p.name);
2864         if (!dev)
2865                 return -ENOBUFS;
2866         return dev_open(dev, NULL);
2867 }
2868
2869 /*
2870  *      Set destination address.
2871  *      Special case for SIT interfaces where we create a new "virtual"
2872  *      device.
2873  */
2874 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2875 {
2876         struct net_device *dev;
2877         struct in6_ifreq ireq;
2878         int err = -ENODEV;
2879
2880         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2881                 return -ENODEV;
2882         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2883                 return -EFAULT;
2884
2885         rtnl_lock();
2886         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2887         if (dev && dev->type == ARPHRD_SIT)
2888                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2889         rtnl_unlock();
2890         return err;
2891 }
2892
2893 static int ipv6_mc_config(struct sock *sk, bool join,
2894                           const struct in6_addr *addr, int ifindex)
2895 {
2896         int ret;
2897
2898         ASSERT_RTNL();
2899
2900         lock_sock(sk);
2901         if (join)
2902                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2903         else
2904                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2905         release_sock(sk);
2906
2907         return ret;
2908 }
2909
2910 /*
2911  *      Manual configuration of address on an interface
2912  */
2913 static int inet6_addr_add(struct net *net, int ifindex,
2914                           struct ifa6_config *cfg,
2915                           struct netlink_ext_ack *extack)
2916 {
2917         struct inet6_ifaddr *ifp;
2918         struct inet6_dev *idev;
2919         struct net_device *dev;
2920         unsigned long timeout;
2921         clock_t expires;
2922         u32 flags;
2923
2924         ASSERT_RTNL();
2925
2926         if (cfg->plen > 128)
2927                 return -EINVAL;
2928
2929         /* check the lifetime */
2930         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2931                 return -EINVAL;
2932
2933         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2934                 return -EINVAL;
2935
2936         dev = __dev_get_by_index(net, ifindex);
2937         if (!dev)
2938                 return -ENODEV;
2939
2940         idev = addrconf_add_dev(dev);
2941         if (IS_ERR(idev))
2942                 return PTR_ERR(idev);
2943
2944         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2945                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2946                                          true, cfg->pfx, ifindex);
2947
2948                 if (ret < 0)
2949                         return ret;
2950         }
2951
2952         cfg->scope = ipv6_addr_scope(cfg->pfx);
2953
2954         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2955         if (addrconf_finite_timeout(timeout)) {
2956                 expires = jiffies_to_clock_t(timeout * HZ);
2957                 cfg->valid_lft = timeout;
2958                 flags = RTF_EXPIRES;
2959         } else {
2960                 expires = 0;
2961                 flags = 0;
2962                 cfg->ifa_flags |= IFA_F_PERMANENT;
2963         }
2964
2965         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2966         if (addrconf_finite_timeout(timeout)) {
2967                 if (timeout == 0)
2968                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2969                 cfg->preferred_lft = timeout;
2970         }
2971
2972         ifp = ipv6_add_addr(idev, cfg, true, extack);
2973         if (!IS_ERR(ifp)) {
2974                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2975                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2976                                               ifp->rt_priority, dev, expires,
2977                                               flags, GFP_KERNEL);
2978                 }
2979
2980                 /* Send a netlink notification if DAD is enabled and
2981                  * optimistic flag is not set
2982                  */
2983                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2984                         ipv6_ifa_notify(0, ifp);
2985                 /*
2986                  * Note that section 3.1 of RFC 4429 indicates
2987                  * that the Optimistic flag should not be set for
2988                  * manually configured addresses
2989                  */
2990                 addrconf_dad_start(ifp);
2991                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2992                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2993                                          cfg->preferred_lft, true, jiffies);
2994                 in6_ifa_put(ifp);
2995                 addrconf_verify_rtnl();
2996                 return 0;
2997         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2998                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2999                                cfg->pfx, ifindex);
3000         }
3001
3002         return PTR_ERR(ifp);
3003 }
3004
3005 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3006                           const struct in6_addr *pfx, unsigned int plen)
3007 {
3008         struct inet6_ifaddr *ifp;
3009         struct inet6_dev *idev;
3010         struct net_device *dev;
3011
3012         if (plen > 128)
3013                 return -EINVAL;
3014
3015         dev = __dev_get_by_index(net, ifindex);
3016         if (!dev)
3017                 return -ENODEV;
3018
3019         idev = __in6_dev_get(dev);
3020         if (!idev)
3021                 return -ENXIO;
3022
3023         read_lock_bh(&idev->lock);
3024         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3025                 if (ifp->prefix_len == plen &&
3026                     ipv6_addr_equal(pfx, &ifp->addr)) {
3027                         in6_ifa_hold(ifp);
3028                         read_unlock_bh(&idev->lock);
3029
3030                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3031                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3032                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3033                                                  jiffies);
3034                         ipv6_del_addr(ifp);
3035                         addrconf_verify_rtnl();
3036                         if (ipv6_addr_is_multicast(pfx)) {
3037                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3038                                                false, pfx, dev->ifindex);
3039                         }
3040                         return 0;
3041                 }
3042         }
3043         read_unlock_bh(&idev->lock);
3044         return -EADDRNOTAVAIL;
3045 }
3046
3047
3048 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3049 {
3050         struct ifa6_config cfg = {
3051                 .ifa_flags = IFA_F_PERMANENT,
3052                 .preferred_lft = INFINITY_LIFE_TIME,
3053                 .valid_lft = INFINITY_LIFE_TIME,
3054         };
3055         struct in6_ifreq ireq;
3056         int err;
3057
3058         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3059                 return -EPERM;
3060
3061         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3062                 return -EFAULT;
3063
3064         cfg.pfx = &ireq.ifr6_addr;
3065         cfg.plen = ireq.ifr6_prefixlen;
3066
3067         rtnl_lock();
3068         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3069         rtnl_unlock();
3070         return err;
3071 }
3072
3073 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3074 {
3075         struct in6_ifreq ireq;
3076         int err;
3077
3078         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3079                 return -EPERM;
3080
3081         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3082                 return -EFAULT;
3083
3084         rtnl_lock();
3085         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3086                              ireq.ifr6_prefixlen);
3087         rtnl_unlock();
3088         return err;
3089 }
3090
3091 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3092                      int plen, int scope)
3093 {
3094         struct inet6_ifaddr *ifp;
3095         struct ifa6_config cfg = {
3096                 .pfx = addr,
3097                 .plen = plen,
3098                 .ifa_flags = IFA_F_PERMANENT,
3099                 .valid_lft = INFINITY_LIFE_TIME,
3100                 .preferred_lft = INFINITY_LIFE_TIME,
3101                 .scope = scope
3102         };
3103
3104         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3105         if (!IS_ERR(ifp)) {
3106                 spin_lock_bh(&ifp->lock);
3107                 ifp->flags &= ~IFA_F_TENTATIVE;
3108                 spin_unlock_bh(&ifp->lock);
3109                 rt_genid_bump_ipv6(dev_net(idev->dev));
3110                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3111                 in6_ifa_put(ifp);
3112         }
3113 }
3114
3115 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3116 static void add_v4_addrs(struct inet6_dev *idev)
3117 {
3118         struct in6_addr addr;
3119         struct net_device *dev;
3120         struct net *net = dev_net(idev->dev);
3121         int scope, plen, offset = 0;
3122         u32 pflags = 0;
3123
3124         ASSERT_RTNL();
3125
3126         memset(&addr, 0, sizeof(struct in6_addr));
3127         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3128         if (idev->dev->addr_len == sizeof(struct in6_addr))
3129                 offset = sizeof(struct in6_addr) - 4;
3130         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3131
3132         if (idev->dev->flags&IFF_POINTOPOINT) {
3133                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3134                         return;
3135
3136                 addr.s6_addr32[0] = htonl(0xfe800000);
3137                 scope = IFA_LINK;
3138                 plen = 64;
3139         } else {
3140                 scope = IPV6_ADDR_COMPATv4;
3141                 plen = 96;
3142                 pflags |= RTF_NONEXTHOP;
3143         }
3144
3145         if (addr.s6_addr32[3]) {
3146                 add_addr(idev, &addr, plen, scope);
3147                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3148                                       GFP_KERNEL);
3149                 return;
3150         }
3151
3152         for_each_netdev(net, dev) {
3153                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3154                 if (in_dev && (dev->flags & IFF_UP)) {
3155                         struct in_ifaddr *ifa;
3156                         int flag = scope;
3157
3158                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3159                                 addr.s6_addr32[3] = ifa->ifa_local;
3160
3161                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3162                                         continue;
3163                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3164                                         if (idev->dev->flags&IFF_POINTOPOINT)
3165                                                 continue;
3166                                         flag |= IFA_HOST;
3167                                 }
3168
3169                                 add_addr(idev, &addr, plen, flag);
3170                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3171                                                       0, pflags, GFP_KERNEL);
3172                         }
3173                 }
3174         }
3175 }
3176 #endif
3177
3178 static void init_loopback(struct net_device *dev)
3179 {
3180         struct inet6_dev  *idev;
3181
3182         /* ::1 */
3183
3184         ASSERT_RTNL();
3185
3186         idev = ipv6_find_idev(dev);
3187         if (IS_ERR(idev)) {
3188                 pr_debug("%s: add_dev failed\n", __func__);
3189                 return;
3190         }
3191
3192         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3193 }
3194
3195 void addrconf_add_linklocal(struct inet6_dev *idev,
3196                             const struct in6_addr *addr, u32 flags)
3197 {
3198         struct ifa6_config cfg = {
3199                 .pfx = addr,
3200                 .plen = 64,
3201                 .ifa_flags = flags | IFA_F_PERMANENT,
3202                 .valid_lft = INFINITY_LIFE_TIME,
3203                 .preferred_lft = INFINITY_LIFE_TIME,
3204                 .scope = IFA_LINK
3205         };
3206         struct inet6_ifaddr *ifp;
3207
3208 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3209         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3210              idev->cnf.optimistic_dad) &&
3211             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3212                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3213 #endif
3214
3215         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3216         if (!IS_ERR(ifp)) {
3217                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3218                                       0, 0, GFP_ATOMIC);
3219                 addrconf_dad_start(ifp);
3220                 in6_ifa_put(ifp);
3221         }
3222 }
3223 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3224
3225 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3226 {
3227         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3228                 return true;
3229
3230         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3231             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3232                 return true;
3233
3234         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3235             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3236                 return true;
3237
3238         return false;
3239 }
3240
3241 static int ipv6_generate_stable_address(struct in6_addr *address,
3242                                         u8 dad_count,
3243                                         const struct inet6_dev *idev)
3244 {
3245         static DEFINE_SPINLOCK(lock);
3246         static __u32 digest[SHA1_DIGEST_WORDS];
3247         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3248
3249         static union {
3250                 char __data[SHA1_BLOCK_SIZE];
3251                 struct {
3252                         struct in6_addr secret;
3253                         __be32 prefix[2];
3254                         unsigned char hwaddr[MAX_ADDR_LEN];
3255                         u8 dad_count;
3256                 } __packed;
3257         } data;
3258
3259         struct in6_addr secret;
3260         struct in6_addr temp;
3261         struct net *net = dev_net(idev->dev);
3262
3263         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3264
3265         if (idev->cnf.stable_secret.initialized)
3266                 secret = idev->cnf.stable_secret.secret;
3267         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3268                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3269         else
3270                 return -1;
3271
3272 retry:
3273         spin_lock_bh(&lock);
3274
3275         sha1_init(digest);
3276         memset(&data, 0, sizeof(data));
3277         memset(workspace, 0, sizeof(workspace));
3278         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3279         data.prefix[0] = address->s6_addr32[0];
3280         data.prefix[1] = address->s6_addr32[1];
3281         data.secret = secret;
3282         data.dad_count = dad_count;
3283
3284         sha1_transform(digest, data.__data, workspace);
3285
3286         temp = *address;
3287         temp.s6_addr32[2] = (__force __be32)digest[0];
3288         temp.s6_addr32[3] = (__force __be32)digest[1];
3289
3290         spin_unlock_bh(&lock);
3291
3292         if (ipv6_reserved_interfaceid(temp)) {
3293                 dad_count++;
3294                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3295                         return -1;
3296                 goto retry;
3297         }
3298
3299         *address = temp;
3300         return 0;
3301 }
3302
3303 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3304 {
3305         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3306
3307         if (s->initialized)
3308                 return;
3309         s = &idev->cnf.stable_secret;
3310         get_random_bytes(&s->secret, sizeof(s->secret));
3311         s->initialized = true;
3312 }
3313
3314 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3315 {
3316         struct in6_addr addr;
3317
3318         /* no link local addresses on L3 master devices */
3319         if (netif_is_l3_master(idev->dev))
3320                 return;
3321
3322         /* no link local addresses on devices flagged as slaves */
3323         if (idev->dev->flags & IFF_SLAVE)
3324                 return;
3325
3326         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3327
3328         switch (idev->cnf.addr_gen_mode) {
3329         case IN6_ADDR_GEN_MODE_RANDOM:
3330                 ipv6_gen_mode_random_init(idev);
3331                 fallthrough;
3332         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3333                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3334                         addrconf_add_linklocal(idev, &addr,
3335                                                IFA_F_STABLE_PRIVACY);
3336                 else if (prefix_route)
3337                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3338                                               0, 0, GFP_KERNEL);
3339                 break;
3340         case IN6_ADDR_GEN_MODE_EUI64:
3341                 /* addrconf_add_linklocal also adds a prefix_route and we
3342                  * only need to care about prefix routes if ipv6_generate_eui64
3343                  * couldn't generate one.
3344                  */
3345                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3346                         addrconf_add_linklocal(idev, &addr, 0);
3347                 else if (prefix_route)
3348                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3349                                               0, 0, GFP_KERNEL);
3350                 break;
3351         case IN6_ADDR_GEN_MODE_NONE:
3352         default:
3353                 /* will not add any link local address */
3354                 break;
3355         }
3356 }
3357
3358 static void addrconf_dev_config(struct net_device *dev)
3359 {
3360         struct inet6_dev *idev;
3361
3362         ASSERT_RTNL();
3363
3364         if ((dev->type != ARPHRD_ETHER) &&
3365             (dev->type != ARPHRD_FDDI) &&
3366             (dev->type != ARPHRD_ARCNET) &&
3367             (dev->type != ARPHRD_INFINIBAND) &&
3368             (dev->type != ARPHRD_IEEE1394) &&
3369             (dev->type != ARPHRD_TUNNEL6) &&
3370             (dev->type != ARPHRD_6LOWPAN) &&
3371             (dev->type != ARPHRD_TUNNEL) &&
3372             (dev->type != ARPHRD_NONE) &&
3373             (dev->type != ARPHRD_RAWIP)) {
3374                 /* Alas, we support only Ethernet autoconfiguration. */
3375                 idev = __in6_dev_get(dev);
3376                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3377                     dev->flags & IFF_MULTICAST)
3378                         ipv6_mc_up(idev);
3379                 return;
3380         }
3381
3382         idev = addrconf_add_dev(dev);
3383         if (IS_ERR(idev))
3384                 return;
3385
3386         /* this device type has no EUI support */
3387         if (dev->type == ARPHRD_NONE &&
3388             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3389                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3390
3391         addrconf_addr_gen(idev, false);
3392 }
3393
3394 #if IS_ENABLED(CONFIG_IPV6_SIT)
3395 static void addrconf_sit_config(struct net_device *dev)
3396 {
3397         struct inet6_dev *idev;
3398
3399         ASSERT_RTNL();
3400
3401         /*
3402          * Configure the tunnel with one of our IPv4
3403          * addresses... we should configure all of
3404          * our v4 addrs in the tunnel
3405          */
3406
3407         idev = ipv6_find_idev(dev);
3408         if (IS_ERR(idev)) {
3409                 pr_debug("%s: add_dev failed\n", __func__);
3410                 return;
3411         }
3412
3413         if (dev->priv_flags & IFF_ISATAP) {
3414                 addrconf_addr_gen(idev, false);
3415                 return;
3416         }
3417
3418         add_v4_addrs(idev);
3419
3420         if (dev->flags&IFF_POINTOPOINT)
3421                 addrconf_add_mroute(dev);
3422 }
3423 #endif
3424
3425 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3426 static void addrconf_gre_config(struct net_device *dev)
3427 {
3428         struct inet6_dev *idev;
3429
3430         ASSERT_RTNL();
3431
3432         idev = ipv6_find_idev(dev);
3433         if (IS_ERR(idev)) {
3434                 pr_debug("%s: add_dev failed\n", __func__);
3435                 return;
3436         }
3437
3438         if (dev->type == ARPHRD_ETHER) {
3439                 addrconf_addr_gen(idev, true);
3440                 return;
3441         }
3442
3443         add_v4_addrs(idev);
3444
3445         if (dev->flags & IFF_POINTOPOINT)
3446                 addrconf_add_mroute(dev);
3447 }
3448 #endif
3449
3450 static int fixup_permanent_addr(struct net *net,
3451                                 struct inet6_dev *idev,
3452                                 struct inet6_ifaddr *ifp)
3453 {
3454         /* !fib6_node means the host route was removed from the
3455          * FIB, for example, if 'lo' device is taken down. In that
3456          * case regenerate the host route.
3457          */
3458         if (!ifp->rt || !ifp->rt->fib6_node) {
3459                 struct fib6_info *f6i, *prev;
3460
3461                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3462                                          GFP_ATOMIC);
3463                 if (IS_ERR(f6i))
3464                         return PTR_ERR(f6i);
3465
3466                 /* ifp->rt can be accessed outside of rtnl */
3467                 spin_lock(&ifp->lock);
3468                 prev = ifp->rt;
3469                 ifp->rt = f6i;
3470                 spin_unlock(&ifp->lock);
3471
3472                 fib6_info_release(prev);
3473         }
3474
3475         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3476                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3477                                       ifp->rt_priority, idev->dev, 0, 0,
3478                                       GFP_ATOMIC);
3479         }
3480
3481         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3482                 addrconf_dad_start(ifp);
3483
3484         return 0;
3485 }
3486
3487 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3488 {
3489         struct inet6_ifaddr *ifp, *tmp;
3490         struct inet6_dev *idev;
3491
3492         idev = __in6_dev_get(dev);
3493         if (!idev)
3494                 return;
3495
3496         write_lock_bh(&idev->lock);
3497
3498         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3499                 if ((ifp->flags & IFA_F_PERMANENT) &&
3500                     fixup_permanent_addr(net, idev, ifp) < 0) {
3501                         write_unlock_bh(&idev->lock);
3502                         in6_ifa_hold(ifp);
3503                         ipv6_del_addr(ifp);
3504                         write_lock_bh(&idev->lock);
3505
3506                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3507                                              idev->dev->name, &ifp->addr);
3508                 }
3509         }
3510
3511         write_unlock_bh(&idev->lock);
3512 }
3513
3514 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3515                            void *ptr)
3516 {
3517         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3518         struct netdev_notifier_change_info *change_info;
3519         struct netdev_notifier_changeupper_info *info;
3520         struct inet6_dev *idev = __in6_dev_get(dev);
3521         struct net *net = dev_net(dev);
3522         int run_pending = 0;
3523         int err;
3524
3525         switch (event) {
3526         case NETDEV_REGISTER:
3527                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3528                         idev = ipv6_add_dev(dev);
3529                         if (IS_ERR(idev))
3530                                 return notifier_from_errno(PTR_ERR(idev));
3531                 }
3532                 break;
3533
3534         case NETDEV_CHANGEMTU:
3535                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3536                 if (dev->mtu < IPV6_MIN_MTU) {
3537                         addrconf_ifdown(dev, dev != net->loopback_dev);
3538                         break;
3539                 }
3540
3541                 if (idev) {
3542                         rt6_mtu_change(dev, dev->mtu);
3543                         idev->cnf.mtu6 = dev->mtu;
3544                         break;
3545                 }
3546
3547                 /* allocate new idev */
3548                 idev = ipv6_add_dev(dev);
3549                 if (IS_ERR(idev))
3550                         break;
3551
3552                 /* device is still not ready */
3553                 if (!(idev->if_flags & IF_READY))
3554                         break;
3555
3556                 run_pending = 1;
3557                 fallthrough;
3558         case NETDEV_UP:
3559         case NETDEV_CHANGE:
3560                 if (dev->flags & IFF_SLAVE)
3561                         break;
3562
3563                 if (idev && idev->cnf.disable_ipv6)
3564                         break;
3565
3566                 if (event == NETDEV_UP) {
3567                         /* restore routes for permanent addresses */
3568                         addrconf_permanent_addr(net, dev);
3569
3570                         if (!addrconf_link_ready(dev)) {
3571                                 /* device is not ready yet. */
3572                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3573                                          dev->name);
3574                                 break;
3575                         }
3576
3577                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3578                                 idev = ipv6_add_dev(dev);
3579
3580                         if (!IS_ERR_OR_NULL(idev)) {
3581                                 idev->if_flags |= IF_READY;
3582                                 run_pending = 1;
3583                         }
3584                 } else if (event == NETDEV_CHANGE) {
3585                         if (!addrconf_link_ready(dev)) {
3586                                 /* device is still not ready. */
3587                                 rt6_sync_down_dev(dev, event);
3588                                 break;
3589                         }
3590
3591                         if (!IS_ERR_OR_NULL(idev)) {
3592                                 if (idev->if_flags & IF_READY) {
3593                                         /* device is already configured -
3594                                          * but resend MLD reports, we might
3595                                          * have roamed and need to update
3596                                          * multicast snooping switches
3597                                          */
3598                                         ipv6_mc_up(idev);
3599                                         change_info = ptr;
3600                                         if (change_info->flags_changed & IFF_NOARP)
3601                                                 addrconf_dad_run(idev, true);
3602                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3603                                         break;
3604                                 }
3605                                 idev->if_flags |= IF_READY;
3606                         }
3607
3608                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3609                                 dev->name);
3610
3611                         run_pending = 1;
3612                 }
3613
3614                 switch (dev->type) {
3615 #if IS_ENABLED(CONFIG_IPV6_SIT)
3616                 case ARPHRD_SIT:
3617                         addrconf_sit_config(dev);
3618                         break;
3619 #endif
3620 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3621                 case ARPHRD_IP6GRE:
3622                 case ARPHRD_IPGRE:
3623                         addrconf_gre_config(dev);
3624                         break;
3625 #endif
3626                 case ARPHRD_LOOPBACK:
3627                         init_loopback(dev);
3628                         break;
3629
3630                 default:
3631                         addrconf_dev_config(dev);
3632                         break;
3633                 }
3634
3635                 if (!IS_ERR_OR_NULL(idev)) {
3636                         if (run_pending)
3637                                 addrconf_dad_run(idev, false);
3638
3639                         /* Device has an address by now */
3640                         rt6_sync_up(dev, RTNH_F_DEAD);
3641
3642                         /*
3643                          * If the MTU changed during the interface down,
3644                          * when the interface up, the changed MTU must be
3645                          * reflected in the idev as well as routers.
3646                          */
3647                         if (idev->cnf.mtu6 != dev->mtu &&
3648                             dev->mtu >= IPV6_MIN_MTU) {
3649                                 rt6_mtu_change(dev, dev->mtu);
3650                                 idev->cnf.mtu6 = dev->mtu;
3651                         }
3652                         idev->tstamp = jiffies;
3653                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3654
3655                         /*
3656                          * If the changed mtu during down is lower than
3657                          * IPV6_MIN_MTU stop IPv6 on this interface.
3658                          */
3659                         if (dev->mtu < IPV6_MIN_MTU)
3660                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3661                 }
3662                 break;
3663
3664         case NETDEV_DOWN:
3665         case NETDEV_UNREGISTER:
3666                 /*
3667                  *      Remove all addresses from this interface.
3668                  */
3669                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3670                 break;
3671
3672         case NETDEV_CHANGENAME:
3673                 if (idev) {
3674                         snmp6_unregister_dev(idev);
3675                         addrconf_sysctl_unregister(idev);
3676                         err = addrconf_sysctl_register(idev);
3677                         if (err)
3678                                 return notifier_from_errno(err);
3679                         err = snmp6_register_dev(idev);
3680                         if (err) {
3681                                 addrconf_sysctl_unregister(idev);
3682                                 return notifier_from_errno(err);
3683                         }
3684                 }
3685                 break;
3686
3687         case NETDEV_PRE_TYPE_CHANGE:
3688         case NETDEV_POST_TYPE_CHANGE:
3689                 if (idev)
3690                         addrconf_type_change(dev, event);
3691                 break;
3692
3693         case NETDEV_CHANGEUPPER:
3694                 info = ptr;
3695
3696                 /* flush all routes if dev is linked to or unlinked from
3697                  * an L3 master device (e.g., VRF)
3698                  */
3699                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3700                         addrconf_ifdown(dev, false);
3701         }
3702
3703         return NOTIFY_OK;
3704 }
3705
3706 /*
3707  *      addrconf module should be notified of a device going up
3708  */
3709 static struct notifier_block ipv6_dev_notf = {
3710         .notifier_call = addrconf_notify,
3711         .priority = ADDRCONF_NOTIFY_PRIORITY,
3712 };
3713
3714 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3715 {
3716         struct inet6_dev *idev;
3717         ASSERT_RTNL();
3718
3719         idev = __in6_dev_get(dev);
3720
3721         if (event == NETDEV_POST_TYPE_CHANGE)
3722                 ipv6_mc_remap(idev);
3723         else if (event == NETDEV_PRE_TYPE_CHANGE)
3724                 ipv6_mc_unmap(idev);
3725 }
3726
3727 static bool addr_is_local(const struct in6_addr *addr)
3728 {
3729         return ipv6_addr_type(addr) &
3730                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3731 }
3732
3733 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3734 {
3735         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3736         struct net *net = dev_net(dev);
3737         struct inet6_dev *idev;
3738         struct inet6_ifaddr *ifa;
3739         LIST_HEAD(tmp_addr_list);
3740         bool keep_addr = false;
3741         bool was_ready;
3742         int state, i;
3743
3744         ASSERT_RTNL();
3745
3746         rt6_disable_ip(dev, event);
3747
3748         idev = __in6_dev_get(dev);
3749         if (!idev)
3750                 return -ENODEV;
3751
3752         /*
3753          * Step 1: remove reference to ipv6 device from parent device.
3754          *         Do not dev_put!
3755          */
3756         if (unregister) {
3757                 idev->dead = 1;
3758
3759                 /* protected by rtnl_lock */
3760                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3761
3762                 /* Step 1.5: remove snmp6 entry */
3763                 snmp6_unregister_dev(idev);
3764
3765         }
3766
3767         /* combine the user config with event to determine if permanent
3768          * addresses are to be removed from address hash table
3769          */
3770         if (!unregister && !idev->cnf.disable_ipv6) {
3771                 /* aggregate the system setting and interface setting */
3772                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3773
3774                 if (!_keep_addr)
3775                         _keep_addr = idev->cnf.keep_addr_on_down;
3776
3777                 keep_addr = (_keep_addr > 0);
3778         }
3779
3780         /* Step 2: clear hash table */
3781         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3782                 struct hlist_head *h = &inet6_addr_lst[i];
3783
3784                 spin_lock_bh(&addrconf_hash_lock);
3785 restart:
3786                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3787                         if (ifa->idev == idev) {
3788                                 addrconf_del_dad_work(ifa);
3789                                 /* combined flag + permanent flag decide if
3790                                  * address is retained on a down event
3791                                  */
3792                                 if (!keep_addr ||
3793                                     !(ifa->flags & IFA_F_PERMANENT) ||
3794                                     addr_is_local(&ifa->addr)) {
3795                                         hlist_del_init_rcu(&ifa->addr_lst);
3796                                         goto restart;
3797                                 }
3798                         }
3799                 }
3800                 spin_unlock_bh(&addrconf_hash_lock);
3801         }
3802
3803         write_lock_bh(&idev->lock);
3804
3805         addrconf_del_rs_timer(idev);
3806
3807         /* Step 2: clear flags for stateless addrconf, repeated down
3808          *         detection
3809          */
3810         was_ready = idev->if_flags & IF_READY;
3811         if (!unregister)
3812                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3813
3814         /* Step 3: clear tempaddr list */
3815         while (!list_empty(&idev->tempaddr_list)) {
3816                 ifa = list_first_entry(&idev->tempaddr_list,
3817                                        struct inet6_ifaddr, tmp_list);
3818                 list_del(&ifa->tmp_list);
3819                 write_unlock_bh(&idev->lock);
3820                 spin_lock_bh(&ifa->lock);
3821
3822                 if (ifa->ifpub) {
3823                         in6_ifa_put(ifa->ifpub);
3824                         ifa->ifpub = NULL;
3825                 }
3826                 spin_unlock_bh(&ifa->lock);
3827                 in6_ifa_put(ifa);
3828                 write_lock_bh(&idev->lock);
3829         }
3830
3831         list_for_each_entry(ifa, &idev->addr_list, if_list)
3832                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3833         write_unlock_bh(&idev->lock);
3834
3835         while (!list_empty(&tmp_addr_list)) {
3836                 struct fib6_info *rt = NULL;
3837                 bool keep;
3838
3839                 ifa = list_first_entry(&tmp_addr_list,
3840                                        struct inet6_ifaddr, if_list_aux);
3841                 list_del(&ifa->if_list_aux);
3842
3843                 addrconf_del_dad_work(ifa);
3844
3845                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3846                         !addr_is_local(&ifa->addr);
3847
3848                 spin_lock_bh(&ifa->lock);
3849
3850                 if (keep) {
3851                         /* set state to skip the notifier below */
3852                         state = INET6_IFADDR_STATE_DEAD;
3853                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3854                         if (!(ifa->flags & IFA_F_NODAD))
3855                                 ifa->flags |= IFA_F_TENTATIVE;
3856
3857                         rt = ifa->rt;
3858                         ifa->rt = NULL;
3859                 } else {
3860                         state = ifa->state;
3861                         ifa->state = INET6_IFADDR_STATE_DEAD;
3862                 }
3863
3864                 spin_unlock_bh(&ifa->lock);
3865
3866                 if (rt)
3867                         ip6_del_rt(net, rt, false);
3868
3869                 if (state != INET6_IFADDR_STATE_DEAD) {
3870                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3871                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3872                 } else {
3873                         if (idev->cnf.forwarding)
3874                                 addrconf_leave_anycast(ifa);
3875                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3876                 }
3877
3878                 if (!keep) {
3879                         write_lock_bh(&idev->lock);
3880                         list_del_rcu(&ifa->if_list);
3881                         write_unlock_bh(&idev->lock);
3882                         in6_ifa_put(ifa);
3883                 }
3884         }
3885
3886         /* Step 5: Discard anycast and multicast list */
3887         if (unregister) {
3888                 ipv6_ac_destroy_dev(idev);
3889                 ipv6_mc_destroy_dev(idev);
3890         } else if (was_ready) {
3891                 ipv6_mc_down(idev);
3892         }
3893
3894         idev->tstamp = jiffies;
3895         idev->ra_mtu = 0;
3896
3897         /* Last: Shot the device (if unregistered) */
3898         if (unregister) {
3899                 addrconf_sysctl_unregister(idev);
3900                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3901                 neigh_ifdown(&nd_tbl, dev);
3902                 in6_dev_put(idev);
3903         }
3904         return 0;
3905 }
3906
3907 static void addrconf_rs_timer(struct timer_list *t)
3908 {
3909         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3910         struct net_device *dev = idev->dev;
3911         struct in6_addr lladdr;
3912
3913         write_lock(&idev->lock);
3914         if (idev->dead || !(idev->if_flags & IF_READY))
3915                 goto out;
3916
3917         if (!ipv6_accept_ra(idev))
3918                 goto out;
3919
3920         /* Announcement received after solicitation was sent */
3921         if (idev->if_flags & IF_RA_RCVD)
3922                 goto out;
3923
3924         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3925                 write_unlock(&idev->lock);
3926                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3927                         ndisc_send_rs(dev, &lladdr,
3928                                       &in6addr_linklocal_allrouters);
3929                 else
3930                         goto put;
3931
3932                 write_lock(&idev->lock);
3933                 idev->rs_interval = rfc3315_s14_backoff_update(
3934                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3935                 /* The wait after the last probe can be shorter */
3936                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3937                                              idev->cnf.rtr_solicits) ?
3938                                       idev->cnf.rtr_solicit_delay :
3939                                       idev->rs_interval);
3940         } else {
3941                 /*
3942                  * Note: we do not support deprecated "all on-link"
3943                  * assumption any longer.
3944                  */
3945                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3946         }
3947
3948 out:
3949         write_unlock(&idev->lock);
3950 put:
3951         in6_dev_put(idev);
3952 }
3953
3954 /*
3955  *      Duplicate Address Detection
3956  */
3957 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3958 {
3959         unsigned long rand_num;
3960         struct inet6_dev *idev = ifp->idev;
3961         u64 nonce;
3962
3963         if (ifp->flags & IFA_F_OPTIMISTIC)
3964                 rand_num = 0;
3965         else
3966                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3967
3968         nonce = 0;
3969         if (idev->cnf.enhanced_dad ||
3970             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3971                 do
3972                         get_random_bytes(&nonce, 6);
3973                 while (nonce == 0);
3974         }
3975         ifp->dad_nonce = nonce;
3976         ifp->dad_probes = idev->cnf.dad_transmits;
3977         addrconf_mod_dad_work(ifp, rand_num);
3978 }
3979
3980 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3981 {
3982         struct inet6_dev *idev = ifp->idev;
3983         struct net_device *dev = idev->dev;
3984         bool bump_id, notify = false;
3985         struct net *net;
3986
3987         addrconf_join_solict(dev, &ifp->addr);
3988
3989         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3990
3991         read_lock_bh(&idev->lock);
3992         spin_lock(&ifp->lock);
3993         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3994                 goto out;
3995
3996         net = dev_net(dev);
3997         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3998             (net->ipv6.devconf_all->accept_dad < 1 &&
3999              idev->cnf.accept_dad < 1) ||
4000             !(ifp->flags&IFA_F_TENTATIVE) ||
4001             ifp->flags & IFA_F_NODAD) {
4002                 bool send_na = false;
4003
4004                 if (ifp->flags & IFA_F_TENTATIVE &&
4005                     !(ifp->flags & IFA_F_OPTIMISTIC))
4006                         send_na = true;
4007                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4008                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4009                 spin_unlock(&ifp->lock);
4010                 read_unlock_bh(&idev->lock);
4011
4012                 addrconf_dad_completed(ifp, bump_id, send_na);
4013                 return;
4014         }
4015
4016         if (!(idev->if_flags & IF_READY)) {
4017                 spin_unlock(&ifp->lock);
4018                 read_unlock_bh(&idev->lock);
4019                 /*
4020                  * If the device is not ready:
4021                  * - keep it tentative if it is a permanent address.
4022                  * - otherwise, kill it.
4023                  */
4024                 in6_ifa_hold(ifp);
4025                 addrconf_dad_stop(ifp, 0);
4026                 return;
4027         }
4028
4029         /*
4030          * Optimistic nodes can start receiving
4031          * Frames right away
4032          */
4033         if (ifp->flags & IFA_F_OPTIMISTIC) {
4034                 ip6_ins_rt(net, ifp->rt);
4035                 if (ipv6_use_optimistic_addr(net, idev)) {
4036                         /* Because optimistic nodes can use this address,
4037                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4038                          */
4039                         notify = true;
4040                 }
4041         }
4042
4043         addrconf_dad_kick(ifp);
4044 out:
4045         spin_unlock(&ifp->lock);
4046         read_unlock_bh(&idev->lock);
4047         if (notify)
4048                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4049 }
4050
4051 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4052 {
4053         bool begin_dad = false;
4054
4055         spin_lock_bh(&ifp->lock);
4056         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4057                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4058                 begin_dad = true;
4059         }
4060         spin_unlock_bh(&ifp->lock);
4061
4062         if (begin_dad)
4063                 addrconf_mod_dad_work(ifp, 0);
4064 }
4065
4066 static void addrconf_dad_work(struct work_struct *w)
4067 {
4068         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4069                                                 struct inet6_ifaddr,
4070                                                 dad_work);
4071         struct inet6_dev *idev = ifp->idev;
4072         bool bump_id, disable_ipv6 = false;
4073         struct in6_addr mcaddr;
4074
4075         enum {
4076                 DAD_PROCESS,
4077                 DAD_BEGIN,
4078                 DAD_ABORT,
4079         } action = DAD_PROCESS;
4080
4081         rtnl_lock();
4082
4083         spin_lock_bh(&ifp->lock);
4084         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4085                 action = DAD_BEGIN;
4086                 ifp->state = INET6_IFADDR_STATE_DAD;
4087         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4088                 action = DAD_ABORT;
4089                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4090
4091                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4092                      idev->cnf.accept_dad > 1) &&
4093                     !idev->cnf.disable_ipv6 &&
4094                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4095                         struct in6_addr addr;
4096
4097                         addr.s6_addr32[0] = htonl(0xfe800000);
4098                         addr.s6_addr32[1] = 0;
4099
4100                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4101                             ipv6_addr_equal(&ifp->addr, &addr)) {
4102                                 /* DAD failed for link-local based on MAC */
4103                                 idev->cnf.disable_ipv6 = 1;
4104
4105                                 pr_info("%s: IPv6 being disabled!\n",
4106                                         ifp->idev->dev->name);
4107                                 disable_ipv6 = true;
4108                         }
4109                 }
4110         }
4111         spin_unlock_bh(&ifp->lock);
4112
4113         if (action == DAD_BEGIN) {
4114                 addrconf_dad_begin(ifp);
4115                 goto out;
4116         } else if (action == DAD_ABORT) {
4117                 in6_ifa_hold(ifp);
4118                 addrconf_dad_stop(ifp, 1);
4119                 if (disable_ipv6)
4120                         addrconf_ifdown(idev->dev, false);
4121                 goto out;
4122         }
4123
4124         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4125                 goto out;
4126
4127         write_lock_bh(&idev->lock);
4128         if (idev->dead || !(idev->if_flags & IF_READY)) {
4129                 write_unlock_bh(&idev->lock);
4130                 goto out;
4131         }
4132
4133         spin_lock(&ifp->lock);
4134         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4135                 spin_unlock(&ifp->lock);
4136                 write_unlock_bh(&idev->lock);
4137                 goto out;
4138         }
4139
4140         if (ifp->dad_probes == 0) {
4141                 bool send_na = false;
4142
4143                 /*
4144                  * DAD was successful
4145                  */
4146
4147                 if (ifp->flags & IFA_F_TENTATIVE &&
4148                     !(ifp->flags & IFA_F_OPTIMISTIC))
4149                         send_na = true;
4150                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4151                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4152                 spin_unlock(&ifp->lock);
4153                 write_unlock_bh(&idev->lock);
4154
4155                 addrconf_dad_completed(ifp, bump_id, send_na);
4156
4157                 goto out;
4158         }
4159
4160         ifp->dad_probes--;
4161         addrconf_mod_dad_work(ifp,
4162                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4163                                   HZ/100));
4164         spin_unlock(&ifp->lock);
4165         write_unlock_bh(&idev->lock);
4166
4167         /* send a neighbour solicitation for our addr */
4168         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4169         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4170                       ifp->dad_nonce);
4171 out:
4172         in6_ifa_put(ifp);
4173         rtnl_unlock();
4174 }
4175
4176 /* ifp->idev must be at least read locked */
4177 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4178 {
4179         struct inet6_ifaddr *ifpiter;
4180         struct inet6_dev *idev = ifp->idev;
4181
4182         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4183                 if (ifpiter->scope > IFA_LINK)
4184                         break;
4185                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4186                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4187                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4188                     IFA_F_PERMANENT)
4189                         return false;
4190         }
4191         return true;
4192 }
4193
4194 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4195                                    bool send_na)
4196 {
4197         struct net_device *dev = ifp->idev->dev;
4198         struct in6_addr lladdr;
4199         bool send_rs, send_mld;
4200
4201         addrconf_del_dad_work(ifp);
4202
4203         /*
4204          *      Configure the address for reception. Now it is valid.
4205          */
4206
4207         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4208
4209         /* If added prefix is link local and we are prepared to process
4210            router advertisements, start sending router solicitations.
4211          */
4212
4213         read_lock_bh(&ifp->idev->lock);
4214         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4215         send_rs = send_mld &&
4216                   ipv6_accept_ra(ifp->idev) &&
4217                   ifp->idev->cnf.rtr_solicits != 0 &&
4218                   (dev->flags & IFF_LOOPBACK) == 0 &&
4219                   (dev->type != ARPHRD_TUNNEL);
4220         read_unlock_bh(&ifp->idev->lock);
4221
4222         /* While dad is in progress mld report's source address is in6_addrany.
4223          * Resend with proper ll now.
4224          */
4225         if (send_mld)
4226                 ipv6_mc_dad_complete(ifp->idev);
4227
4228         /* send unsolicited NA if enabled */
4229         if (send_na &&
4230             (ifp->idev->cnf.ndisc_notify ||
4231              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4232                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4233                               /*router=*/ !!ifp->idev->cnf.forwarding,
4234                               /*solicited=*/ false, /*override=*/ true,
4235                               /*inc_opt=*/ true);
4236         }
4237
4238         if (send_rs) {
4239                 /*
4240                  *      If a host as already performed a random delay
4241                  *      [...] as part of DAD [...] there is no need
4242                  *      to delay again before sending the first RS
4243                  */
4244                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4245                         return;
4246                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4247
4248                 write_lock_bh(&ifp->idev->lock);
4249                 spin_lock(&ifp->lock);
4250                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4251                         ifp->idev->cnf.rtr_solicit_interval);
4252                 ifp->idev->rs_probes = 1;
4253                 ifp->idev->if_flags |= IF_RS_SENT;
4254                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4255                 spin_unlock(&ifp->lock);
4256                 write_unlock_bh(&ifp->idev->lock);
4257         }
4258
4259         if (bump_id)
4260                 rt_genid_bump_ipv6(dev_net(dev));
4261
4262         /* Make sure that a new temporary address will be created
4263          * before this temporary address becomes deprecated.
4264          */
4265         if (ifp->flags & IFA_F_TEMPORARY)
4266                 addrconf_verify_rtnl();
4267 }
4268
4269 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4270 {
4271         struct inet6_ifaddr *ifp;
4272
4273         read_lock_bh(&idev->lock);
4274         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4275                 spin_lock(&ifp->lock);
4276                 if ((ifp->flags & IFA_F_TENTATIVE &&
4277                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4278                         if (restart)
4279                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4280                         addrconf_dad_kick(ifp);
4281                 }
4282                 spin_unlock(&ifp->lock);
4283         }
4284         read_unlock_bh(&idev->lock);
4285 }
4286
4287 #ifdef CONFIG_PROC_FS
4288 struct if6_iter_state {
4289         struct seq_net_private p;
4290         int bucket;
4291         int offset;
4292 };
4293
4294 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4295 {
4296         struct if6_iter_state *state = seq->private;
4297         struct net *net = seq_file_net(seq);
4298         struct inet6_ifaddr *ifa = NULL;
4299         int p = 0;
4300
4301         /* initial bucket if pos is 0 */
4302         if (pos == 0) {
4303                 state->bucket = 0;
4304                 state->offset = 0;
4305         }
4306
4307         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4308                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4309                                          addr_lst) {
4310                         if (!net_eq(dev_net(ifa->idev->dev), net))
4311                                 continue;
4312                         /* sync with offset */
4313                         if (p < state->offset) {
4314                                 p++;
4315                                 continue;
4316                         }
4317                         return ifa;
4318                 }
4319
4320                 /* prepare for next bucket */
4321                 state->offset = 0;
4322                 p = 0;
4323         }
4324         return NULL;
4325 }
4326
4327 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4328                                          struct inet6_ifaddr *ifa)
4329 {
4330         struct if6_iter_state *state = seq->private;
4331         struct net *net = seq_file_net(seq);
4332
4333         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4334                 if (!net_eq(dev_net(ifa->idev->dev), net))
4335                         continue;
4336                 state->offset++;
4337                 return ifa;
4338         }
4339
4340         state->offset = 0;
4341         while (++state->bucket < IN6_ADDR_HSIZE) {
4342                 hlist_for_each_entry_rcu(ifa,
4343                                      &inet6_addr_lst[state->bucket], addr_lst) {
4344                         if (!net_eq(dev_net(ifa->idev->dev), net))
4345                                 continue;
4346                         return ifa;
4347                 }
4348         }
4349
4350         return NULL;
4351 }
4352
4353 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4354         __acquires(rcu)
4355 {
4356         rcu_read_lock();
4357         return if6_get_first(seq, *pos);
4358 }
4359
4360 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4361 {
4362         struct inet6_ifaddr *ifa;
4363
4364         ifa = if6_get_next(seq, v);
4365         ++*pos;
4366         return ifa;
4367 }
4368
4369 static void if6_seq_stop(struct seq_file *seq, void *v)
4370         __releases(rcu)
4371 {
4372         rcu_read_unlock();
4373 }
4374
4375 static int if6_seq_show(struct seq_file *seq, void *v)
4376 {
4377         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4378         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4379                    &ifp->addr,
4380                    ifp->idev->dev->ifindex,
4381                    ifp->prefix_len,
4382                    ifp->scope,
4383                    (u8) ifp->flags,
4384                    ifp->idev->dev->name);
4385         return 0;
4386 }
4387
4388 static const struct seq_operations if6_seq_ops = {
4389         .start  = if6_seq_start,
4390         .next   = if6_seq_next,
4391         .show   = if6_seq_show,
4392         .stop   = if6_seq_stop,
4393 };
4394
4395 static int __net_init if6_proc_net_init(struct net *net)
4396 {
4397         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4398                         sizeof(struct if6_iter_state)))
4399                 return -ENOMEM;
4400         return 0;
4401 }
4402
4403 static void __net_exit if6_proc_net_exit(struct net *net)
4404 {
4405         remove_proc_entry("if_inet6", net->proc_net);
4406 }
4407
4408 static struct pernet_operations if6_proc_net_ops = {
4409         .init = if6_proc_net_init,
4410         .exit = if6_proc_net_exit,
4411 };
4412
4413 int __init if6_proc_init(void)
4414 {
4415         return register_pernet_subsys(&if6_proc_net_ops);
4416 }
4417
4418 void if6_proc_exit(void)
4419 {
4420         unregister_pernet_subsys(&if6_proc_net_ops);
4421 }
4422 #endif  /* CONFIG_PROC_FS */
4423
4424 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4425 /* Check if address is a home address configured on any interface. */
4426 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4427 {
4428         unsigned int hash = inet6_addr_hash(net, addr);
4429         struct inet6_ifaddr *ifp = NULL;
4430         int ret = 0;
4431
4432         rcu_read_lock();
4433         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4434                 if (!net_eq(dev_net(ifp->idev->dev), net))
4435                         continue;
4436                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4437                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4438                         ret = 1;
4439                         break;
4440                 }
4441         }
4442         rcu_read_unlock();
4443         return ret;
4444 }
4445 #endif
4446
4447 /* RFC6554 has some algorithm to avoid loops in segment routing by
4448  * checking if the segments contains any of a local interface address.
4449  *
4450  * Quote:
4451  *
4452  * To detect loops in the SRH, a router MUST determine if the SRH
4453  * includes multiple addresses assigned to any interface on that router.
4454  * If such addresses appear more than once and are separated by at least
4455  * one address not assigned to that router.
4456  */
4457 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4458                           unsigned char nsegs)
4459 {
4460         const struct in6_addr *addr;
4461         int i, ret = 0, found = 0;
4462         struct inet6_ifaddr *ifp;
4463         bool separated = false;
4464         unsigned int hash;
4465         bool hash_found;
4466
4467         rcu_read_lock();
4468         for (i = 0; i < nsegs; i++) {
4469                 addr = &segs[i];
4470                 hash = inet6_addr_hash(net, addr);
4471
4472                 hash_found = false;
4473                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4474                         if (!net_eq(dev_net(ifp->idev->dev), net))
4475                                 continue;
4476
4477                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4478                                 hash_found = true;
4479                                 break;
4480                         }
4481                 }
4482
4483                 if (hash_found) {
4484                         if (found > 1 && separated) {
4485                                 ret = 1;
4486                                 break;
4487                         }
4488
4489                         separated = false;
4490                         found++;
4491                 } else {
4492                         separated = true;
4493                 }
4494         }
4495         rcu_read_unlock();
4496
4497         return ret;
4498 }
4499
4500 /*
4501  *      Periodic address status verification
4502  */
4503
4504 static void addrconf_verify_rtnl(void)
4505 {
4506         unsigned long now, next, next_sec, next_sched;
4507         struct inet6_ifaddr *ifp;
4508         int i;
4509
4510         ASSERT_RTNL();
4511
4512         rcu_read_lock_bh();
4513         now = jiffies;
4514         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4515
4516         cancel_delayed_work(&addr_chk_work);
4517
4518         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4519 restart:
4520                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4521                         unsigned long age;
4522
4523                         /* When setting preferred_lft to a value not zero or
4524                          * infinity, while valid_lft is infinity
4525                          * IFA_F_PERMANENT has a non-infinity life time.
4526                          */
4527                         if ((ifp->flags & IFA_F_PERMANENT) &&
4528                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4529                                 continue;
4530
4531                         spin_lock(&ifp->lock);
4532                         /* We try to batch several events at once. */
4533                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4534
4535                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4536                             age >= ifp->valid_lft) {
4537                                 spin_unlock(&ifp->lock);
4538                                 in6_ifa_hold(ifp);
4539                                 rcu_read_unlock_bh();
4540                                 ipv6_del_addr(ifp);
4541                                 rcu_read_lock_bh();
4542                                 goto restart;
4543                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4544                                 spin_unlock(&ifp->lock);
4545                                 continue;
4546                         } else if (age >= ifp->prefered_lft) {
4547                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4548                                 int deprecate = 0;
4549
4550                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4551                                         deprecate = 1;
4552                                         ifp->flags |= IFA_F_DEPRECATED;
4553                                 }
4554
4555                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4556                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4557                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4558
4559                                 spin_unlock(&ifp->lock);
4560
4561                                 if (deprecate) {
4562                                         in6_ifa_hold(ifp);
4563
4564                                         ipv6_ifa_notify(0, ifp);
4565                                         in6_ifa_put(ifp);
4566                                         goto restart;
4567                                 }
4568                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4569                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4570                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4571                                         ifp->idev->cnf.dad_transmits *
4572                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4573
4574                                 if (age >= ifp->prefered_lft - regen_advance) {
4575                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4576                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4577                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4578                                         if (!ifp->regen_count && ifpub) {
4579                                                 ifp->regen_count++;
4580                                                 in6_ifa_hold(ifp);
4581                                                 in6_ifa_hold(ifpub);
4582                                                 spin_unlock(&ifp->lock);
4583
4584                                                 spin_lock(&ifpub->lock);
4585                                                 ifpub->regen_count = 0;
4586                                                 spin_unlock(&ifpub->lock);
4587                                                 rcu_read_unlock_bh();
4588                                                 ipv6_create_tempaddr(ifpub, true);
4589                                                 in6_ifa_put(ifpub);
4590                                                 in6_ifa_put(ifp);
4591                                                 rcu_read_lock_bh();
4592                                                 goto restart;
4593                                         }
4594                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4595                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4596                                 spin_unlock(&ifp->lock);
4597                         } else {
4598                                 /* ifp->prefered_lft <= ifp->valid_lft */
4599                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4600                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4601                                 spin_unlock(&ifp->lock);
4602                         }
4603                 }
4604         }
4605
4606         next_sec = round_jiffies_up(next);
4607         next_sched = next;
4608
4609         /* If rounded timeout is accurate enough, accept it. */
4610         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4611                 next_sched = next_sec;
4612
4613         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4614         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4615                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4616
4617         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4618                  now, next, next_sec, next_sched);
4619         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4620         rcu_read_unlock_bh();
4621 }
4622
4623 static void addrconf_verify_work(struct work_struct *w)
4624 {
4625         rtnl_lock();
4626         addrconf_verify_rtnl();
4627         rtnl_unlock();
4628 }
4629
4630 static void addrconf_verify(void)
4631 {
4632         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4633 }
4634
4635 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4636                                      struct in6_addr **peer_pfx)
4637 {
4638         struct in6_addr *pfx = NULL;
4639
4640         *peer_pfx = NULL;
4641
4642         if (addr)
4643                 pfx = nla_data(addr);
4644
4645         if (local) {
4646                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4647                         *peer_pfx = pfx;
4648                 pfx = nla_data(local);
4649         }
4650
4651         return pfx;
4652 }
4653
4654 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4655         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4656         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4657         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4658         [IFA_FLAGS]             = { .len = sizeof(u32) },
4659         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4660         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4661 };
4662
4663 static int
4664 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4665                   struct netlink_ext_ack *extack)
4666 {
4667         struct net *net = sock_net(skb->sk);
4668         struct ifaddrmsg *ifm;
4669         struct nlattr *tb[IFA_MAX+1];
4670         struct in6_addr *pfx, *peer_pfx;
4671         u32 ifa_flags;
4672         int err;
4673
4674         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4675                                      ifa_ipv6_policy, extack);
4676         if (err < 0)
4677                 return err;
4678
4679         ifm = nlmsg_data(nlh);
4680         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4681         if (!pfx)
4682                 return -EINVAL;
4683
4684         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4685
4686         /* We ignore other flags so far. */
4687         ifa_flags &= IFA_F_MANAGETEMPADDR;
4688
4689         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4690                               ifm->ifa_prefixlen);
4691 }
4692
4693 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4694                                unsigned long expires, u32 flags,
4695                                bool modify_peer)
4696 {
4697         struct fib6_info *f6i;
4698         u32 prio;
4699
4700         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4701                                         ifp->prefix_len,
4702                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4703         if (!f6i)
4704                 return -ENOENT;
4705
4706         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4707         if (f6i->fib6_metric != prio) {
4708                 /* delete old one */
4709                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4710
4711                 /* add new one */
4712                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4713                                       ifp->prefix_len,
4714                                       ifp->rt_priority, ifp->idev->dev,
4715                                       expires, flags, GFP_KERNEL);
4716         } else {
4717                 if (!expires)
4718                         fib6_clean_expires(f6i);
4719                 else
4720                         fib6_set_expires(f6i, expires);
4721
4722                 fib6_info_release(f6i);
4723         }
4724
4725         return 0;
4726 }
4727
4728 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4729 {
4730         u32 flags;
4731         clock_t expires;
4732         unsigned long timeout;
4733         bool was_managetempaddr;
4734         bool had_prefixroute;
4735         bool new_peer = false;
4736
4737         ASSERT_RTNL();
4738
4739         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4740                 return -EINVAL;
4741
4742         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4743             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4744                 return -EINVAL;
4745
4746         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4747                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4748
4749         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4750         if (addrconf_finite_timeout(timeout)) {
4751                 expires = jiffies_to_clock_t(timeout * HZ);
4752                 cfg->valid_lft = timeout;
4753                 flags = RTF_EXPIRES;
4754         } else {
4755                 expires = 0;
4756                 flags = 0;
4757                 cfg->ifa_flags |= IFA_F_PERMANENT;
4758         }
4759
4760         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4761         if (addrconf_finite_timeout(timeout)) {
4762                 if (timeout == 0)
4763                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4764                 cfg->preferred_lft = timeout;
4765         }
4766
4767         if (cfg->peer_pfx &&
4768             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4769                 if (!ipv6_addr_any(&ifp->peer_addr))
4770                         cleanup_prefix_route(ifp, expires, true, true);
4771                 new_peer = true;
4772         }
4773
4774         spin_lock_bh(&ifp->lock);
4775         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4776         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4777                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4778         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4779                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4780                         IFA_F_NOPREFIXROUTE);
4781         ifp->flags |= cfg->ifa_flags;
4782         ifp->tstamp = jiffies;
4783         ifp->valid_lft = cfg->valid_lft;
4784         ifp->prefered_lft = cfg->preferred_lft;
4785
4786         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4787                 ifp->rt_priority = cfg->rt_priority;
4788
4789         if (new_peer)
4790                 ifp->peer_addr = *cfg->peer_pfx;
4791
4792         spin_unlock_bh(&ifp->lock);
4793         if (!(ifp->flags&IFA_F_TENTATIVE))
4794                 ipv6_ifa_notify(0, ifp);
4795
4796         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4797                 int rc = -ENOENT;
4798
4799                 if (had_prefixroute)
4800                         rc = modify_prefix_route(ifp, expires, flags, false);
4801
4802                 /* prefix route could have been deleted; if so restore it */
4803                 if (rc == -ENOENT) {
4804                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4805                                               ifp->rt_priority, ifp->idev->dev,
4806                                               expires, flags, GFP_KERNEL);
4807                 }
4808
4809                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4810                         rc = modify_prefix_route(ifp, expires, flags, true);
4811
4812                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4813                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4814                                               ifp->rt_priority, ifp->idev->dev,
4815                                               expires, flags, GFP_KERNEL);
4816                 }
4817         } else if (had_prefixroute) {
4818                 enum cleanup_prefix_rt_t action;
4819                 unsigned long rt_expires;
4820
4821                 write_lock_bh(&ifp->idev->lock);
4822                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4823                 write_unlock_bh(&ifp->idev->lock);
4824
4825                 if (action != CLEANUP_PREFIX_RT_NOP) {
4826                         cleanup_prefix_route(ifp, rt_expires,
4827                                 action == CLEANUP_PREFIX_RT_DEL, false);
4828                 }
4829         }
4830
4831         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4832                 if (was_managetempaddr &&
4833                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4834                         cfg->valid_lft = 0;
4835                         cfg->preferred_lft = 0;
4836                 }
4837                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4838                                  cfg->preferred_lft, !was_managetempaddr,
4839                                  jiffies);
4840         }
4841
4842         addrconf_verify_rtnl();
4843
4844         return 0;
4845 }
4846
4847 static int
4848 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4849                   struct netlink_ext_ack *extack)
4850 {
4851         struct net *net = sock_net(skb->sk);
4852         struct ifaddrmsg *ifm;
4853         struct nlattr *tb[IFA_MAX+1];
4854         struct in6_addr *peer_pfx;
4855         struct inet6_ifaddr *ifa;
4856         struct net_device *dev;
4857         struct inet6_dev *idev;
4858         struct ifa6_config cfg;
4859         int err;
4860
4861         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4862                                      ifa_ipv6_policy, extack);
4863         if (err < 0)
4864                 return err;
4865
4866         memset(&cfg, 0, sizeof(cfg));
4867
4868         ifm = nlmsg_data(nlh);
4869         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4870         if (!cfg.pfx)
4871                 return -EINVAL;
4872
4873         cfg.peer_pfx = peer_pfx;
4874         cfg.plen = ifm->ifa_prefixlen;
4875         if (tb[IFA_RT_PRIORITY])
4876                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4877
4878         cfg.valid_lft = INFINITY_LIFE_TIME;
4879         cfg.preferred_lft = INFINITY_LIFE_TIME;
4880
4881         if (tb[IFA_CACHEINFO]) {
4882                 struct ifa_cacheinfo *ci;
4883
4884                 ci = nla_data(tb[IFA_CACHEINFO]);
4885                 cfg.valid_lft = ci->ifa_valid;
4886                 cfg.preferred_lft = ci->ifa_prefered;
4887         }
4888
4889         dev =  __dev_get_by_index(net, ifm->ifa_index);
4890         if (!dev)
4891                 return -ENODEV;
4892
4893         if (tb[IFA_FLAGS])
4894                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4895         else
4896                 cfg.ifa_flags = ifm->ifa_flags;
4897
4898         /* We ignore other flags so far. */
4899         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4900                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4901                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4902
4903         idev = ipv6_find_idev(dev);
4904         if (IS_ERR(idev))
4905                 return PTR_ERR(idev);
4906
4907         if (!ipv6_allow_optimistic_dad(net, idev))
4908                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4909
4910         if (cfg.ifa_flags & IFA_F_NODAD &&
4911             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4912                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4913                 return -EINVAL;
4914         }
4915
4916         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4917         if (!ifa) {
4918                 /*
4919                  * It would be best to check for !NLM_F_CREATE here but
4920                  * userspace already relies on not having to provide this.
4921                  */
4922                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4923         }
4924
4925         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4926             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4927                 err = -EEXIST;
4928         else
4929                 err = inet6_addr_modify(ifa, &cfg);
4930
4931         in6_ifa_put(ifa);
4932
4933         return err;
4934 }
4935
4936 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4937                           u8 scope, int ifindex)
4938 {
4939         struct ifaddrmsg *ifm;
4940
4941         ifm = nlmsg_data(nlh);
4942         ifm->ifa_family = AF_INET6;
4943         ifm->ifa_prefixlen = prefixlen;
4944         ifm->ifa_flags = flags;
4945         ifm->ifa_scope = scope;
4946         ifm->ifa_index = ifindex;
4947 }
4948
4949 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4950                          unsigned long tstamp, u32 preferred, u32 valid)
4951 {
4952         struct ifa_cacheinfo ci;
4953
4954         ci.cstamp = cstamp_delta(cstamp);
4955         ci.tstamp = cstamp_delta(tstamp);
4956         ci.ifa_prefered = preferred;
4957         ci.ifa_valid = valid;
4958
4959         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4960 }
4961
4962 static inline int rt_scope(int ifa_scope)
4963 {
4964         if (ifa_scope & IFA_HOST)
4965                 return RT_SCOPE_HOST;
4966         else if (ifa_scope & IFA_LINK)
4967                 return RT_SCOPE_LINK;
4968         else if (ifa_scope & IFA_SITE)
4969                 return RT_SCOPE_SITE;
4970         else
4971                 return RT_SCOPE_UNIVERSE;
4972 }
4973
4974 static inline int inet6_ifaddr_msgsize(void)
4975 {
4976         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4977                + nla_total_size(16) /* IFA_LOCAL */
4978                + nla_total_size(16) /* IFA_ADDRESS */
4979                + nla_total_size(sizeof(struct ifa_cacheinfo))
4980                + nla_total_size(4)  /* IFA_FLAGS */
4981                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4982 }
4983
4984 enum addr_type_t {
4985         UNICAST_ADDR,
4986         MULTICAST_ADDR,
4987         ANYCAST_ADDR,
4988 };
4989
4990 struct inet6_fill_args {
4991         u32 portid;
4992         u32 seq;
4993         int event;
4994         unsigned int flags;
4995         int netnsid;
4996         int ifindex;
4997         enum addr_type_t type;
4998 };
4999
5000 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5001                              struct inet6_fill_args *args)
5002 {
5003         struct nlmsghdr  *nlh;
5004         u32 preferred, valid;
5005
5006         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5007                         sizeof(struct ifaddrmsg), args->flags);
5008         if (!nlh)
5009                 return -EMSGSIZE;
5010
5011         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5012                       ifa->idev->dev->ifindex);
5013
5014         if (args->netnsid >= 0 &&
5015             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5016                 goto error;
5017
5018         spin_lock_bh(&ifa->lock);
5019         if (!((ifa->flags&IFA_F_PERMANENT) &&
5020               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5021                 preferred = ifa->prefered_lft;
5022                 valid = ifa->valid_lft;
5023                 if (preferred != INFINITY_LIFE_TIME) {
5024                         long tval = (jiffies - ifa->tstamp)/HZ;
5025                         if (preferred > tval)
5026                                 preferred -= tval;
5027                         else
5028                                 preferred = 0;
5029                         if (valid != INFINITY_LIFE_TIME) {
5030                                 if (valid > tval)
5031                                         valid -= tval;
5032                                 else
5033                                         valid = 0;
5034                         }
5035                 }
5036         } else {
5037                 preferred = INFINITY_LIFE_TIME;
5038                 valid = INFINITY_LIFE_TIME;
5039         }
5040         spin_unlock_bh(&ifa->lock);
5041
5042         if (!ipv6_addr_any(&ifa->peer_addr)) {
5043                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5044                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5045                         goto error;
5046         } else
5047                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5048                         goto error;
5049
5050         if (ifa->rt_priority &&
5051             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5052                 goto error;
5053
5054         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5055                 goto error;
5056
5057         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5058                 goto error;
5059
5060         nlmsg_end(skb, nlh);
5061         return 0;
5062
5063 error:
5064         nlmsg_cancel(skb, nlh);
5065         return -EMSGSIZE;
5066 }
5067
5068 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5069                                struct inet6_fill_args *args)
5070 {
5071         struct nlmsghdr  *nlh;
5072         u8 scope = RT_SCOPE_UNIVERSE;
5073         int ifindex = ifmca->idev->dev->ifindex;
5074
5075         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5076                 scope = RT_SCOPE_SITE;
5077
5078         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5079                         sizeof(struct ifaddrmsg), args->flags);
5080         if (!nlh)
5081                 return -EMSGSIZE;
5082
5083         if (args->netnsid >= 0 &&
5084             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5085                 nlmsg_cancel(skb, nlh);
5086                 return -EMSGSIZE;
5087         }
5088
5089         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5090         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5091             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5092                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5093                 nlmsg_cancel(skb, nlh);
5094                 return -EMSGSIZE;
5095         }
5096
5097         nlmsg_end(skb, nlh);
5098         return 0;
5099 }
5100
5101 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5102                                struct inet6_fill_args *args)
5103 {
5104         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5105         int ifindex = dev ? dev->ifindex : 1;
5106         struct nlmsghdr  *nlh;
5107         u8 scope = RT_SCOPE_UNIVERSE;
5108
5109         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5110                 scope = RT_SCOPE_SITE;
5111
5112         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5113                         sizeof(struct ifaddrmsg), args->flags);
5114         if (!nlh)
5115                 return -EMSGSIZE;
5116
5117         if (args->netnsid >= 0 &&
5118             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5119                 nlmsg_cancel(skb, nlh);
5120                 return -EMSGSIZE;
5121         }
5122
5123         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5124         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5125             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5126                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5127                 nlmsg_cancel(skb, nlh);
5128                 return -EMSGSIZE;
5129         }
5130
5131         nlmsg_end(skb, nlh);
5132         return 0;
5133 }
5134
5135 /* called with rcu_read_lock() */
5136 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5137                           struct netlink_callback *cb, int s_ip_idx,
5138                           struct inet6_fill_args *fillargs)
5139 {
5140         struct ifmcaddr6 *ifmca;
5141         struct ifacaddr6 *ifaca;
5142         int ip_idx = 0;
5143         int err = 1;
5144
5145         read_lock_bh(&idev->lock);
5146         switch (fillargs->type) {
5147         case UNICAST_ADDR: {
5148                 struct inet6_ifaddr *ifa;
5149                 fillargs->event = RTM_NEWADDR;
5150
5151                 /* unicast address incl. temp addr */
5152                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5153                         if (ip_idx < s_ip_idx)
5154                                 goto next;
5155                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5156                         if (err < 0)
5157                                 break;
5158                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5159 next:
5160                         ip_idx++;
5161                 }
5162                 break;
5163         }
5164         case MULTICAST_ADDR:
5165                 read_unlock_bh(&idev->lock);
5166                 fillargs->event = RTM_GETMULTICAST;
5167
5168                 /* multicast address */
5169                 for (ifmca = rtnl_dereference(idev->mc_list);
5170                      ifmca;
5171                      ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5172                         if (ip_idx < s_ip_idx)
5173                                 continue;
5174                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5175                         if (err < 0)
5176                                 break;
5177                 }
5178                 read_lock_bh(&idev->lock);
5179                 break;
5180         case ANYCAST_ADDR:
5181                 fillargs->event = RTM_GETANYCAST;
5182                 /* anycast address */
5183                 for (ifaca = idev->ac_list; ifaca;
5184                      ifaca = ifaca->aca_next, ip_idx++) {
5185                         if (ip_idx < s_ip_idx)
5186                                 continue;
5187                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5188                         if (err < 0)
5189                                 break;
5190                 }
5191                 break;
5192         default:
5193                 break;
5194         }
5195         read_unlock_bh(&idev->lock);
5196         cb->args[2] = ip_idx;
5197         return err;
5198 }
5199
5200 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5201                                        struct inet6_fill_args *fillargs,
5202                                        struct net **tgt_net, struct sock *sk,
5203                                        struct netlink_callback *cb)
5204 {
5205         struct netlink_ext_ack *extack = cb->extack;
5206         struct nlattr *tb[IFA_MAX+1];
5207         struct ifaddrmsg *ifm;
5208         int err, i;
5209
5210         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5211                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5212                 return -EINVAL;
5213         }
5214
5215         ifm = nlmsg_data(nlh);
5216         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5217                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5218                 return -EINVAL;
5219         }
5220
5221         fillargs->ifindex = ifm->ifa_index;
5222         if (fillargs->ifindex) {
5223                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5224                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5225         }
5226
5227         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5228                                             ifa_ipv6_policy, extack);
5229         if (err < 0)
5230                 return err;
5231
5232         for (i = 0; i <= IFA_MAX; ++i) {
5233                 if (!tb[i])
5234                         continue;
5235
5236                 if (i == IFA_TARGET_NETNSID) {
5237                         struct net *net;
5238
5239                         fillargs->netnsid = nla_get_s32(tb[i]);
5240                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5241                         if (IS_ERR(net)) {
5242                                 fillargs->netnsid = -1;
5243                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5244                                 return PTR_ERR(net);
5245                         }
5246                         *tgt_net = net;
5247                 } else {
5248                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5249                         return -EINVAL;
5250                 }
5251         }
5252
5253         return 0;
5254 }
5255
5256 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5257                            enum addr_type_t type)
5258 {
5259         const struct nlmsghdr *nlh = cb->nlh;
5260         struct inet6_fill_args fillargs = {
5261                 .portid = NETLINK_CB(cb->skb).portid,
5262                 .seq = cb->nlh->nlmsg_seq,
5263                 .flags = NLM_F_MULTI,
5264                 .netnsid = -1,
5265                 .type = type,
5266         };
5267         struct net *tgt_net = sock_net(skb->sk);
5268         int idx, s_idx, s_ip_idx;
5269         int h, s_h;
5270         struct net_device *dev;
5271         struct inet6_dev *idev;
5272         struct hlist_head *head;
5273         int err = 0;
5274
5275         s_h = cb->args[0];
5276         s_idx = idx = cb->args[1];
5277         s_ip_idx = cb->args[2];
5278
5279         if (cb->strict_check) {
5280                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5281                                                   skb->sk, cb);
5282                 if (err < 0)
5283                         goto put_tgt_net;
5284
5285                 err = 0;
5286                 if (fillargs.ifindex) {
5287                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5288                         if (!dev) {
5289                                 err = -ENODEV;
5290                                 goto put_tgt_net;
5291                         }
5292                         idev = __in6_dev_get(dev);
5293                         if (idev) {
5294                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5295                                                      &fillargs);
5296                                 if (err > 0)
5297                                         err = 0;
5298                         }
5299                         goto put_tgt_net;
5300                 }
5301         }
5302
5303         rcu_read_lock();
5304         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5305         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5306                 idx = 0;
5307                 head = &tgt_net->dev_index_head[h];
5308                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5309                         if (idx < s_idx)
5310                                 goto cont;
5311                         if (h > s_h || idx > s_idx)
5312                                 s_ip_idx = 0;
5313                         idev = __in6_dev_get(dev);
5314                         if (!idev)
5315                                 goto cont;
5316
5317                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5318                                            &fillargs) < 0)
5319                                 goto done;
5320 cont:
5321                         idx++;
5322                 }
5323         }
5324 done:
5325         rcu_read_unlock();
5326         cb->args[0] = h;
5327         cb->args[1] = idx;
5328 put_tgt_net:
5329         if (fillargs.netnsid >= 0)
5330                 put_net(tgt_net);
5331
5332         return skb->len ? : err;
5333 }
5334
5335 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5336 {
5337         enum addr_type_t type = UNICAST_ADDR;
5338
5339         return inet6_dump_addr(skb, cb, type);
5340 }
5341
5342 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5343 {
5344         enum addr_type_t type = MULTICAST_ADDR;
5345
5346         return inet6_dump_addr(skb, cb, type);
5347 }
5348
5349
5350 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5351 {
5352         enum addr_type_t type = ANYCAST_ADDR;
5353
5354         return inet6_dump_addr(skb, cb, type);
5355 }
5356
5357 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5358                                        const struct nlmsghdr *nlh,
5359                                        struct nlattr **tb,
5360                                        struct netlink_ext_ack *extack)
5361 {
5362         struct ifaddrmsg *ifm;
5363         int i, err;
5364
5365         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5366                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5367                 return -EINVAL;
5368         }
5369
5370         if (!netlink_strict_get_check(skb))
5371                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5372                                               ifa_ipv6_policy, extack);
5373
5374         ifm = nlmsg_data(nlh);
5375         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5376                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5377                 return -EINVAL;
5378         }
5379
5380         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5381                                             ifa_ipv6_policy, extack);
5382         if (err)
5383                 return err;
5384
5385         for (i = 0; i <= IFA_MAX; i++) {
5386                 if (!tb[i])
5387                         continue;
5388
5389                 switch (i) {
5390                 case IFA_TARGET_NETNSID:
5391                 case IFA_ADDRESS:
5392                 case IFA_LOCAL:
5393                         break;
5394                 default:
5395                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5396                         return -EINVAL;
5397                 }
5398         }
5399
5400         return 0;
5401 }
5402
5403 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5404                              struct netlink_ext_ack *extack)
5405 {
5406         struct net *tgt_net = sock_net(in_skb->sk);
5407         struct inet6_fill_args fillargs = {
5408                 .portid = NETLINK_CB(in_skb).portid,
5409                 .seq = nlh->nlmsg_seq,
5410                 .event = RTM_NEWADDR,
5411                 .flags = 0,
5412                 .netnsid = -1,
5413         };
5414         struct ifaddrmsg *ifm;
5415         struct nlattr *tb[IFA_MAX+1];
5416         struct in6_addr *addr = NULL, *peer;
5417         struct net_device *dev = NULL;
5418         struct inet6_ifaddr *ifa;
5419         struct sk_buff *skb;
5420         int err;
5421
5422         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5423         if (err < 0)
5424                 return err;
5425
5426         if (tb[IFA_TARGET_NETNSID]) {
5427                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5428
5429                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5430                                                   fillargs.netnsid);
5431                 if (IS_ERR(tgt_net))
5432                         return PTR_ERR(tgt_net);
5433         }
5434
5435         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5436         if (!addr)
5437                 return -EINVAL;
5438
5439         ifm = nlmsg_data(nlh);
5440         if (ifm->ifa_index)
5441                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5442
5443         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5444         if (!ifa) {
5445                 err = -EADDRNOTAVAIL;
5446                 goto errout;
5447         }
5448
5449         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5450         if (!skb) {
5451                 err = -ENOBUFS;
5452                 goto errout_ifa;
5453         }
5454
5455         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5456         if (err < 0) {
5457                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5458                 WARN_ON(err == -EMSGSIZE);
5459                 kfree_skb(skb);
5460                 goto errout_ifa;
5461         }
5462         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5463 errout_ifa:
5464         in6_ifa_put(ifa);
5465 errout:
5466         dev_put(dev);
5467         if (fillargs.netnsid >= 0)
5468                 put_net(tgt_net);
5469
5470         return err;
5471 }
5472
5473 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5474 {
5475         struct sk_buff *skb;
5476         struct net *net = dev_net(ifa->idev->dev);
5477         struct inet6_fill_args fillargs = {
5478                 .portid = 0,
5479                 .seq = 0,
5480                 .event = event,
5481                 .flags = 0,
5482                 .netnsid = -1,
5483         };
5484         int err = -ENOBUFS;
5485
5486         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5487         if (!skb)
5488                 goto errout;
5489
5490         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5491         if (err < 0) {
5492                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5493                 WARN_ON(err == -EMSGSIZE);
5494                 kfree_skb(skb);
5495                 goto errout;
5496         }
5497         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5498         return;
5499 errout:
5500         if (err < 0)
5501                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5502 }
5503
5504 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5505                                 __s32 *array, int bytes)
5506 {
5507         BUG_ON(bytes < (DEVCONF_MAX * 4));
5508
5509         memset(array, 0, bytes);
5510         array[DEVCONF_FORWARDING] = cnf->forwarding;
5511         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5512         array[DEVCONF_MTU6] = cnf->mtu6;
5513         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5514         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5515         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5516         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5517         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5518         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5519                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5520         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5521                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5522         array[DEVCONF_RTR_SOLICIT_DELAY] =
5523                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5524         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5525         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5526                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5527         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5528                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5529         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5530         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5531         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5532         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5533         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5534         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5535         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5536         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5537         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5538         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5539 #ifdef CONFIG_IPV6_ROUTER_PREF
5540         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5541         array[DEVCONF_RTR_PROBE_INTERVAL] =
5542                 jiffies_to_msecs(cnf->rtr_probe_interval);
5543 #ifdef CONFIG_IPV6_ROUTE_INFO
5544         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5545         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5546 #endif
5547 #endif
5548         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5549         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5550 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5551         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5552         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5553 #endif
5554 #ifdef CONFIG_IPV6_MROUTE
5555         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5556 #endif
5557         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5558         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5559         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5560         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5561         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5562         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5563         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5564         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5565         /* we omit DEVCONF_STABLE_SECRET for now */
5566         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5567         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5568         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5569         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5570         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5571 #ifdef CONFIG_IPV6_SEG6_HMAC
5572         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5573 #endif
5574         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5575         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5576         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5577         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5578         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5579         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5580         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5581         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5582 }
5583
5584 static inline size_t inet6_ifla6_size(void)
5585 {
5586         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5587              + nla_total_size(sizeof(struct ifla_cacheinfo))
5588              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5589              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5590              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5591              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5592              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5593              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5594              + 0;
5595 }
5596
5597 static inline size_t inet6_if_nlmsg_size(void)
5598 {
5599         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5600                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5601                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5602                + nla_total_size(4) /* IFLA_MTU */
5603                + nla_total_size(4) /* IFLA_LINK */
5604                + nla_total_size(1) /* IFLA_OPERSTATE */
5605                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5606 }
5607
5608 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5609                                         int bytes)
5610 {
5611         int i;
5612         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5613         BUG_ON(pad < 0);
5614
5615         /* Use put_unaligned() because stats may not be aligned for u64. */
5616         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5617         for (i = 1; i < ICMP6_MIB_MAX; i++)
5618                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5619
5620         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5621 }
5622
5623 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5624                                         int bytes, size_t syncpoff)
5625 {
5626         int i, c;
5627         u64 buff[IPSTATS_MIB_MAX];
5628         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5629
5630         BUG_ON(pad < 0);
5631
5632         memset(buff, 0, sizeof(buff));
5633         buff[0] = IPSTATS_MIB_MAX;
5634
5635         for_each_possible_cpu(c) {
5636                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5637                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5638         }
5639
5640         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5641         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5642 }
5643
5644 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5645                              int bytes)
5646 {
5647         switch (attrtype) {
5648         case IFLA_INET6_STATS:
5649                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5650                                      offsetof(struct ipstats_mib, syncp));
5651                 break;
5652         case IFLA_INET6_ICMP6STATS:
5653                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5654                 break;
5655         }
5656 }
5657
5658 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5659                                   u32 ext_filter_mask)
5660 {
5661         struct nlattr *nla;
5662         struct ifla_cacheinfo ci;
5663
5664         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5665                 goto nla_put_failure;
5666         ci.max_reasm_len = IPV6_MAXPLEN;
5667         ci.tstamp = cstamp_delta(idev->tstamp);
5668         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5669         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5670         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5671                 goto nla_put_failure;
5672         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5673         if (!nla)
5674                 goto nla_put_failure;
5675         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5676
5677         /* XXX - MC not implemented */
5678
5679         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5680                 return 0;
5681
5682         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5683         if (!nla)
5684                 goto nla_put_failure;
5685         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5686
5687         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5688         if (!nla)
5689                 goto nla_put_failure;
5690         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5691
5692         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5693         if (!nla)
5694                 goto nla_put_failure;
5695         read_lock_bh(&idev->lock);
5696         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5697         read_unlock_bh(&idev->lock);
5698
5699         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5700                 goto nla_put_failure;
5701
5702         if (idev->ra_mtu &&
5703             nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5704                 goto nla_put_failure;
5705
5706         return 0;
5707
5708 nla_put_failure:
5709         return -EMSGSIZE;
5710 }
5711
5712 static size_t inet6_get_link_af_size(const struct net_device *dev,
5713                                      u32 ext_filter_mask)
5714 {
5715         if (!__in6_dev_get(dev))
5716                 return 0;
5717
5718         return inet6_ifla6_size();
5719 }
5720
5721 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5722                               u32 ext_filter_mask)
5723 {
5724         struct inet6_dev *idev = __in6_dev_get(dev);
5725
5726         if (!idev)
5727                 return -ENODATA;
5728
5729         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5730                 return -EMSGSIZE;
5731
5732         return 0;
5733 }
5734
5735 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5736                              struct netlink_ext_ack *extack)
5737 {
5738         struct inet6_ifaddr *ifp;
5739         struct net_device *dev = idev->dev;
5740         bool clear_token, update_rs = false;
5741         struct in6_addr ll_addr;
5742
5743         ASSERT_RTNL();
5744
5745         if (!token)
5746                 return -EINVAL;
5747
5748         if (dev->flags & IFF_LOOPBACK) {
5749                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5750                 return -EINVAL;
5751         }
5752
5753         if (dev->flags & IFF_NOARP) {
5754                 NL_SET_ERR_MSG_MOD(extack,
5755                                    "Device does not do neighbour discovery");
5756                 return -EINVAL;
5757         }
5758
5759         if (!ipv6_accept_ra(idev)) {
5760                 NL_SET_ERR_MSG_MOD(extack,
5761                                    "Router advertisement is disabled on device");
5762                 return -EINVAL;
5763         }
5764
5765         if (idev->cnf.rtr_solicits == 0) {
5766                 NL_SET_ERR_MSG(extack,
5767                                "Router solicitation is disabled on device");
5768                 return -EINVAL;
5769         }
5770
5771         write_lock_bh(&idev->lock);
5772
5773         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5774         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5775
5776         write_unlock_bh(&idev->lock);
5777
5778         clear_token = ipv6_addr_any(token);
5779         if (clear_token)
5780                 goto update_lft;
5781
5782         if (!idev->dead && (idev->if_flags & IF_READY) &&
5783             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5784                              IFA_F_OPTIMISTIC)) {
5785                 /* If we're not ready, then normal ifup will take care
5786                  * of this. Otherwise, we need to request our rs here.
5787                  */
5788                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5789                 update_rs = true;
5790         }
5791
5792 update_lft:
5793         write_lock_bh(&idev->lock);
5794
5795         if (update_rs) {
5796                 idev->if_flags |= IF_RS_SENT;
5797                 idev->rs_interval = rfc3315_s14_backoff_init(
5798                         idev->cnf.rtr_solicit_interval);
5799                 idev->rs_probes = 1;
5800                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5801         }
5802
5803         /* Well, that's kinda nasty ... */
5804         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5805                 spin_lock(&ifp->lock);
5806                 if (ifp->tokenized) {
5807                         ifp->valid_lft = 0;
5808                         ifp->prefered_lft = 0;
5809                 }
5810                 spin_unlock(&ifp->lock);
5811         }
5812
5813         write_unlock_bh(&idev->lock);
5814         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5815         addrconf_verify_rtnl();
5816         return 0;
5817 }
5818
5819 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5820         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5821         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5822         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5823                                             .reject_message =
5824                                                 "IFLA_INET6_RA_MTU can not be set" },
5825 };
5826
5827 static int check_addr_gen_mode(int mode)
5828 {
5829         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5830             mode != IN6_ADDR_GEN_MODE_NONE &&
5831             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5832             mode != IN6_ADDR_GEN_MODE_RANDOM)
5833                 return -EINVAL;
5834         return 1;
5835 }
5836
5837 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5838                                 int mode)
5839 {
5840         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5841             !idev->cnf.stable_secret.initialized &&
5842             !net->ipv6.devconf_dflt->stable_secret.initialized)
5843                 return -EINVAL;
5844         return 1;
5845 }
5846
5847 static int inet6_validate_link_af(const struct net_device *dev,
5848                                   const struct nlattr *nla,
5849                                   struct netlink_ext_ack *extack)
5850 {
5851         struct nlattr *tb[IFLA_INET6_MAX + 1];
5852         struct inet6_dev *idev = NULL;
5853         int err;
5854
5855         if (dev) {
5856                 idev = __in6_dev_get(dev);
5857                 if (!idev)
5858                         return -EAFNOSUPPORT;
5859         }
5860
5861         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5862                                           inet6_af_policy, extack);
5863         if (err)
5864                 return err;
5865
5866         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5867                 return -EINVAL;
5868
5869         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5870                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5871
5872                 if (check_addr_gen_mode(mode) < 0)
5873                         return -EINVAL;
5874                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5875                         return -EINVAL;
5876         }
5877
5878         return 0;
5879 }
5880
5881 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5882                              struct netlink_ext_ack *extack)
5883 {
5884         struct inet6_dev *idev = __in6_dev_get(dev);
5885         struct nlattr *tb[IFLA_INET6_MAX + 1];
5886         int err;
5887
5888         if (!idev)
5889                 return -EAFNOSUPPORT;
5890
5891         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5892                 return -EINVAL;
5893
5894         if (tb[IFLA_INET6_TOKEN]) {
5895                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5896                                         extack);
5897                 if (err)
5898                         return err;
5899         }
5900
5901         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5902                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5903
5904                 idev->cnf.addr_gen_mode = mode;
5905         }
5906
5907         return 0;
5908 }
5909
5910 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5911                              u32 portid, u32 seq, int event, unsigned int flags)
5912 {
5913         struct net_device *dev = idev->dev;
5914         struct ifinfomsg *hdr;
5915         struct nlmsghdr *nlh;
5916         void *protoinfo;
5917
5918         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5919         if (!nlh)
5920                 return -EMSGSIZE;
5921
5922         hdr = nlmsg_data(nlh);
5923         hdr->ifi_family = AF_INET6;
5924         hdr->__ifi_pad = 0;
5925         hdr->ifi_type = dev->type;
5926         hdr->ifi_index = dev->ifindex;
5927         hdr->ifi_flags = dev_get_flags(dev);
5928         hdr->ifi_change = 0;
5929
5930         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5931             (dev->addr_len &&
5932              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5933             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5934             (dev->ifindex != dev_get_iflink(dev) &&
5935              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5936             nla_put_u8(skb, IFLA_OPERSTATE,
5937                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5938                 goto nla_put_failure;
5939         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5940         if (!protoinfo)
5941                 goto nla_put_failure;
5942
5943         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5944                 goto nla_put_failure;
5945
5946         nla_nest_end(skb, protoinfo);
5947         nlmsg_end(skb, nlh);
5948         return 0;
5949
5950 nla_put_failure:
5951         nlmsg_cancel(skb, nlh);
5952         return -EMSGSIZE;
5953 }
5954
5955 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5956                                    struct netlink_ext_ack *extack)
5957 {
5958         struct ifinfomsg *ifm;
5959
5960         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5961                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5962                 return -EINVAL;
5963         }
5964
5965         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5966                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5967                 return -EINVAL;
5968         }
5969
5970         ifm = nlmsg_data(nlh);
5971         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5972             ifm->ifi_change || ifm->ifi_index) {
5973                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5974                 return -EINVAL;
5975         }
5976
5977         return 0;
5978 }
5979
5980 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5981 {
5982         struct net *net = sock_net(skb->sk);
5983         int h, s_h;
5984         int idx = 0, s_idx;
5985         struct net_device *dev;
5986         struct inet6_dev *idev;
5987         struct hlist_head *head;
5988
5989         /* only requests using strict checking can pass data to
5990          * influence the dump
5991          */
5992         if (cb->strict_check) {
5993                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5994
5995                 if (err < 0)
5996                         return err;
5997         }
5998
5999         s_h = cb->args[0];
6000         s_idx = cb->args[1];
6001
6002         rcu_read_lock();
6003         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6004                 idx = 0;
6005                 head = &net->dev_index_head[h];
6006                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6007                         if (idx < s_idx)
6008                                 goto cont;
6009                         idev = __in6_dev_get(dev);
6010                         if (!idev)
6011                                 goto cont;
6012                         if (inet6_fill_ifinfo(skb, idev,
6013                                               NETLINK_CB(cb->skb).portid,
6014                                               cb->nlh->nlmsg_seq,
6015                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
6016                                 goto out;
6017 cont:
6018                         idx++;
6019                 }
6020         }
6021 out:
6022         rcu_read_unlock();
6023         cb->args[1] = idx;
6024         cb->args[0] = h;
6025
6026         return skb->len;
6027 }
6028
6029 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6030 {
6031         struct sk_buff *skb;
6032         struct net *net = dev_net(idev->dev);
6033         int err = -ENOBUFS;
6034
6035         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6036         if (!skb)
6037                 goto errout;
6038
6039         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6040         if (err < 0) {
6041                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6042                 WARN_ON(err == -EMSGSIZE);
6043                 kfree_skb(skb);
6044                 goto errout;
6045         }
6046         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6047         return;
6048 errout:
6049         if (err < 0)
6050                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6051 }
6052
6053 static inline size_t inet6_prefix_nlmsg_size(void)
6054 {
6055         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6056                + nla_total_size(sizeof(struct in6_addr))
6057                + nla_total_size(sizeof(struct prefix_cacheinfo));
6058 }
6059
6060 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6061                              struct prefix_info *pinfo, u32 portid, u32 seq,
6062                              int event, unsigned int flags)
6063 {
6064         struct prefixmsg *pmsg;
6065         struct nlmsghdr *nlh;
6066         struct prefix_cacheinfo ci;
6067
6068         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6069         if (!nlh)
6070                 return -EMSGSIZE;
6071
6072         pmsg = nlmsg_data(nlh);
6073         pmsg->prefix_family = AF_INET6;
6074         pmsg->prefix_pad1 = 0;
6075         pmsg->prefix_pad2 = 0;
6076         pmsg->prefix_ifindex = idev->dev->ifindex;
6077         pmsg->prefix_len = pinfo->prefix_len;
6078         pmsg->prefix_type = pinfo->type;
6079         pmsg->prefix_pad3 = 0;
6080         pmsg->prefix_flags = 0;
6081         if (pinfo->onlink)
6082                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6083         if (pinfo->autoconf)
6084                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6085
6086         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6087                 goto nla_put_failure;
6088         ci.preferred_time = ntohl(pinfo->prefered);
6089         ci.valid_time = ntohl(pinfo->valid);
6090         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6091                 goto nla_put_failure;
6092         nlmsg_end(skb, nlh);
6093         return 0;
6094
6095 nla_put_failure:
6096         nlmsg_cancel(skb, nlh);
6097         return -EMSGSIZE;
6098 }
6099
6100 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6101                          struct prefix_info *pinfo)
6102 {
6103         struct sk_buff *skb;
6104         struct net *net = dev_net(idev->dev);
6105         int err = -ENOBUFS;
6106
6107         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6108         if (!skb)
6109                 goto errout;
6110
6111         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6112         if (err < 0) {
6113                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6114                 WARN_ON(err == -EMSGSIZE);
6115                 kfree_skb(skb);
6116                 goto errout;
6117         }
6118         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6119         return;
6120 errout:
6121         if (err < 0)
6122                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6123 }
6124
6125 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6126 {
6127         struct net *net = dev_net(ifp->idev->dev);
6128
6129         if (event)
6130                 ASSERT_RTNL();
6131
6132         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6133
6134         switch (event) {
6135         case RTM_NEWADDR:
6136                 /*
6137                  * If the address was optimistic we inserted the route at the
6138                  * start of our DAD process, so we don't need to do it again.
6139                  * If the device was taken down in the middle of the DAD
6140                  * cycle there is a race where we could get here without a
6141                  * host route, so nothing to insert. That will be fixed when
6142                  * the device is brought up.
6143                  */
6144                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6145                         ip6_ins_rt(net, ifp->rt);
6146                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6147                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6148                                 &ifp->addr, ifp->idev->dev->name);
6149                 }
6150
6151                 if (ifp->idev->cnf.forwarding)
6152                         addrconf_join_anycast(ifp);
6153                 if (!ipv6_addr_any(&ifp->peer_addr))
6154                         addrconf_prefix_route(&ifp->peer_addr, 128,
6155                                               ifp->rt_priority, ifp->idev->dev,
6156                                               0, 0, GFP_ATOMIC);
6157                 break;
6158         case RTM_DELADDR:
6159                 if (ifp->idev->cnf.forwarding)
6160                         addrconf_leave_anycast(ifp);
6161                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6162                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6163                         struct fib6_info *rt;
6164
6165                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6166                                                        ifp->idev->dev, 0, 0,
6167                                                        false);
6168                         if (rt)
6169                                 ip6_del_rt(net, rt, false);
6170                 }
6171                 if (ifp->rt) {
6172                         ip6_del_rt(net, ifp->rt, false);
6173                         ifp->rt = NULL;
6174                 }
6175                 rt_genid_bump_ipv6(net);
6176                 break;
6177         }
6178         atomic_inc(&net->ipv6.dev_addr_genid);
6179 }
6180
6181 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6182 {
6183         if (likely(ifp->idev->dead == 0))
6184                 __ipv6_ifa_notify(event, ifp);
6185 }
6186
6187 #ifdef CONFIG_SYSCTL
6188
6189 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6190                 void *buffer, size_t *lenp, loff_t *ppos)
6191 {
6192         int *valp = ctl->data;
6193         int val = *valp;
6194         loff_t pos = *ppos;
6195         struct ctl_table lctl;
6196         int ret;
6197
6198         /*
6199          * ctl->data points to idev->cnf.forwarding, we should
6200          * not modify it until we get the rtnl lock.
6201          */
6202         lctl = *ctl;
6203         lctl.data = &val;
6204
6205         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6206
6207         if (write)
6208                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6209         if (ret)
6210                 *ppos = pos;
6211         return ret;
6212 }
6213
6214 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6215                 void *buffer, size_t *lenp, loff_t *ppos)
6216 {
6217         struct inet6_dev *idev = ctl->extra1;
6218         int min_mtu = IPV6_MIN_MTU;
6219         struct ctl_table lctl;
6220
6221         lctl = *ctl;
6222         lctl.extra1 = &min_mtu;
6223         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6224
6225         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6226 }
6227
6228 static void dev_disable_change(struct inet6_dev *idev)
6229 {
6230         struct netdev_notifier_info info;
6231
6232         if (!idev || !idev->dev)
6233                 return;
6234
6235         netdev_notifier_info_init(&info, idev->dev);
6236         if (idev->cnf.disable_ipv6)
6237                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6238         else
6239                 addrconf_notify(NULL, NETDEV_UP, &info);
6240 }
6241
6242 static void addrconf_disable_change(struct net *net, __s32 newf)
6243 {
6244         struct net_device *dev;
6245         struct inet6_dev *idev;
6246
6247         for_each_netdev(net, dev) {
6248                 idev = __in6_dev_get(dev);
6249                 if (idev) {
6250                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6251                         idev->cnf.disable_ipv6 = newf;
6252                         if (changed)
6253                                 dev_disable_change(idev);
6254                 }
6255         }
6256 }
6257
6258 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6259 {
6260         struct net *net;
6261         int old;
6262
6263         if (!rtnl_trylock())
6264                 return restart_syscall();
6265
6266         net = (struct net *)table->extra2;
6267         old = *p;
6268         *p = newf;
6269
6270         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6271                 rtnl_unlock();
6272                 return 0;
6273         }
6274
6275         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6276                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6277                 addrconf_disable_change(net, newf);
6278         } else if ((!newf) ^ (!old))
6279                 dev_disable_change((struct inet6_dev *)table->extra1);
6280
6281         rtnl_unlock();
6282         return 0;
6283 }
6284
6285 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6286                 void *buffer, size_t *lenp, loff_t *ppos)
6287 {
6288         int *valp = ctl->data;
6289         int val = *valp;
6290         loff_t pos = *ppos;
6291         struct ctl_table lctl;
6292         int ret;
6293
6294         /*
6295          * ctl->data points to idev->cnf.disable_ipv6, we should
6296          * not modify it until we get the rtnl lock.
6297          */
6298         lctl = *ctl;
6299         lctl.data = &val;
6300
6301         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6302
6303         if (write)
6304                 ret = addrconf_disable_ipv6(ctl, valp, val);
6305         if (ret)
6306                 *ppos = pos;
6307         return ret;
6308 }
6309
6310 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6311                 void *buffer, size_t *lenp, loff_t *ppos)
6312 {
6313         int *valp = ctl->data;
6314         int ret;
6315         int old, new;
6316
6317         old = *valp;
6318         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6319         new = *valp;
6320
6321         if (write && old != new) {
6322                 struct net *net = ctl->extra2;
6323
6324                 if (!rtnl_trylock())
6325                         return restart_syscall();
6326
6327                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6328                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6329                                                      NETCONFA_PROXY_NEIGH,
6330                                                      NETCONFA_IFINDEX_DEFAULT,
6331                                                      net->ipv6.devconf_dflt);
6332                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6333                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6334                                                      NETCONFA_PROXY_NEIGH,
6335                                                      NETCONFA_IFINDEX_ALL,
6336                                                      net->ipv6.devconf_all);
6337                 else {
6338                         struct inet6_dev *idev = ctl->extra1;
6339
6340                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6341                                                      NETCONFA_PROXY_NEIGH,
6342                                                      idev->dev->ifindex,
6343                                                      &idev->cnf);
6344                 }
6345                 rtnl_unlock();
6346         }
6347
6348         return ret;
6349 }
6350
6351 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6352                                          void *buffer, size_t *lenp,
6353                                          loff_t *ppos)
6354 {
6355         int ret = 0;
6356         u32 new_val;
6357         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6358         struct net *net = (struct net *)ctl->extra2;
6359         struct ctl_table tmp = {
6360                 .data = &new_val,
6361                 .maxlen = sizeof(new_val),
6362                 .mode = ctl->mode,
6363         };
6364
6365         if (!rtnl_trylock())
6366                 return restart_syscall();
6367
6368         new_val = *((u32 *)ctl->data);
6369
6370         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6371         if (ret != 0)
6372                 goto out;
6373
6374         if (write) {
6375                 if (check_addr_gen_mode(new_val) < 0) {
6376                         ret = -EINVAL;
6377                         goto out;
6378                 }
6379
6380                 if (idev) {
6381                         if (check_stable_privacy(idev, net, new_val) < 0) {
6382                                 ret = -EINVAL;
6383                                 goto out;
6384                         }
6385
6386                         if (idev->cnf.addr_gen_mode != new_val) {
6387                                 idev->cnf.addr_gen_mode = new_val;
6388                                 addrconf_dev_config(idev->dev);
6389                         }
6390                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6391                         struct net_device *dev;
6392
6393                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6394                         for_each_netdev(net, dev) {
6395                                 idev = __in6_dev_get(dev);
6396                                 if (idev &&
6397                                     idev->cnf.addr_gen_mode != new_val) {
6398                                         idev->cnf.addr_gen_mode = new_val;
6399                                         addrconf_dev_config(idev->dev);
6400                                 }
6401                         }
6402                 }
6403
6404                 *((u32 *)ctl->data) = new_val;
6405         }
6406
6407 out:
6408         rtnl_unlock();
6409
6410         return ret;
6411 }
6412
6413 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6414                                          void *buffer, size_t *lenp,
6415                                          loff_t *ppos)
6416 {
6417         int err;
6418         struct in6_addr addr;
6419         char str[IPV6_MAX_STRLEN];
6420         struct ctl_table lctl = *ctl;
6421         struct net *net = ctl->extra2;
6422         struct ipv6_stable_secret *secret = ctl->data;
6423
6424         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6425                 return -EIO;
6426
6427         lctl.maxlen = IPV6_MAX_STRLEN;
6428         lctl.data = str;
6429
6430         if (!rtnl_trylock())
6431                 return restart_syscall();
6432
6433         if (!write && !secret->initialized) {
6434                 err = -EIO;
6435                 goto out;
6436         }
6437
6438         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6439         if (err >= sizeof(str)) {
6440                 err = -EIO;
6441                 goto out;
6442         }
6443
6444         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6445         if (err || !write)
6446                 goto out;
6447
6448         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6449                 err = -EIO;
6450                 goto out;
6451         }
6452
6453         secret->initialized = true;
6454         secret->secret = addr;
6455
6456         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6457                 struct net_device *dev;
6458
6459                 for_each_netdev(net, dev) {
6460                         struct inet6_dev *idev = __in6_dev_get(dev);
6461
6462                         if (idev) {
6463                                 idev->cnf.addr_gen_mode =
6464                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6465                         }
6466                 }
6467         } else {
6468                 struct inet6_dev *idev = ctl->extra1;
6469
6470                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6471         }
6472
6473 out:
6474         rtnl_unlock();
6475
6476         return err;
6477 }
6478
6479 static
6480 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6481                                                 int write, void *buffer,
6482                                                 size_t *lenp,
6483                                                 loff_t *ppos)
6484 {
6485         int *valp = ctl->data;
6486         int val = *valp;
6487         loff_t pos = *ppos;
6488         struct ctl_table lctl;
6489         int ret;
6490
6491         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6492          * we should not modify it until we get the rtnl lock.
6493          */
6494         lctl = *ctl;
6495         lctl.data = &val;
6496
6497         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6498
6499         if (write)
6500                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6501         if (ret)
6502                 *ppos = pos;
6503         return ret;
6504 }
6505
6506 static
6507 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6508 {
6509         if (rt) {
6510                 if (action)
6511                         rt->dst.flags |= DST_NOPOLICY;
6512                 else
6513                         rt->dst.flags &= ~DST_NOPOLICY;
6514         }
6515 }
6516
6517 static
6518 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6519 {
6520         struct inet6_ifaddr *ifa;
6521
6522         read_lock_bh(&idev->lock);
6523         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6524                 spin_lock(&ifa->lock);
6525                 if (ifa->rt) {
6526                         /* host routes only use builtin fib6_nh */
6527                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6528                         int cpu;
6529
6530                         rcu_read_lock();
6531                         ifa->rt->dst_nopolicy = val ? true : false;
6532                         if (nh->rt6i_pcpu) {
6533                                 for_each_possible_cpu(cpu) {
6534                                         struct rt6_info **rtp;
6535
6536                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6537                                         addrconf_set_nopolicy(*rtp, val);
6538                                 }
6539                         }
6540                         rcu_read_unlock();
6541                 }
6542                 spin_unlock(&ifa->lock);
6543         }
6544         read_unlock_bh(&idev->lock);
6545 }
6546
6547 static
6548 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6549 {
6550         struct inet6_dev *idev;
6551         struct net *net;
6552
6553         if (!rtnl_trylock())
6554                 return restart_syscall();
6555
6556         *valp = val;
6557
6558         net = (struct net *)ctl->extra2;
6559         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6560                 rtnl_unlock();
6561                 return 0;
6562         }
6563
6564         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6565                 struct net_device *dev;
6566
6567                 for_each_netdev(net, dev) {
6568                         idev = __in6_dev_get(dev);
6569                         if (idev)
6570                                 addrconf_disable_policy_idev(idev, val);
6571                 }
6572         } else {
6573                 idev = (struct inet6_dev *)ctl->extra1;
6574                 addrconf_disable_policy_idev(idev, val);
6575         }
6576
6577         rtnl_unlock();
6578         return 0;
6579 }
6580
6581 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6582                                    void *buffer, size_t *lenp, loff_t *ppos)
6583 {
6584         int *valp = ctl->data;
6585         int val = *valp;
6586         loff_t pos = *ppos;
6587         struct ctl_table lctl;
6588         int ret;
6589
6590         lctl = *ctl;
6591         lctl.data = &val;
6592         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6593
6594         if (write && (*valp != val))
6595                 ret = addrconf_disable_policy(ctl, valp, val);
6596
6597         if (ret)
6598                 *ppos = pos;
6599
6600         return ret;
6601 }
6602
6603 static int minus_one = -1;
6604 static const int two_five_five = 255;
6605 static u32 ioam6_if_id_max = U16_MAX;
6606
6607 static const struct ctl_table addrconf_sysctl[] = {
6608         {
6609                 .procname       = "forwarding",
6610                 .data           = &ipv6_devconf.forwarding,
6611                 .maxlen         = sizeof(int),
6612                 .mode           = 0644,
6613                 .proc_handler   = addrconf_sysctl_forward,
6614         },
6615         {
6616                 .procname       = "hop_limit",
6617                 .data           = &ipv6_devconf.hop_limit,
6618                 .maxlen         = sizeof(int),
6619                 .mode           = 0644,
6620                 .proc_handler   = proc_dointvec_minmax,
6621                 .extra1         = (void *)SYSCTL_ONE,
6622                 .extra2         = (void *)&two_five_five,
6623         },
6624         {
6625                 .procname       = "mtu",
6626                 .data           = &ipv6_devconf.mtu6,
6627                 .maxlen         = sizeof(int),
6628                 .mode           = 0644,
6629                 .proc_handler   = addrconf_sysctl_mtu,
6630         },
6631         {
6632                 .procname       = "accept_ra",
6633                 .data           = &ipv6_devconf.accept_ra,
6634                 .maxlen         = sizeof(int),
6635                 .mode           = 0644,
6636                 .proc_handler   = proc_dointvec,
6637         },
6638         {
6639                 .procname       = "accept_redirects",
6640                 .data           = &ipv6_devconf.accept_redirects,
6641                 .maxlen         = sizeof(int),
6642                 .mode           = 0644,
6643                 .proc_handler   = proc_dointvec,
6644         },
6645         {
6646                 .procname       = "autoconf",
6647                 .data           = &ipv6_devconf.autoconf,
6648                 .maxlen         = sizeof(int),
6649                 .mode           = 0644,
6650                 .proc_handler   = proc_dointvec,
6651         },
6652         {
6653                 .procname       = "dad_transmits",
6654                 .data           = &ipv6_devconf.dad_transmits,
6655                 .maxlen         = sizeof(int),
6656                 .mode           = 0644,
6657                 .proc_handler   = proc_dointvec,
6658         },
6659         {
6660                 .procname       = "router_solicitations",
6661                 .data           = &ipv6_devconf.rtr_solicits,
6662                 .maxlen         = sizeof(int),
6663                 .mode           = 0644,
6664                 .proc_handler   = proc_dointvec_minmax,
6665                 .extra1         = &minus_one,
6666         },
6667         {
6668                 .procname       = "router_solicitation_interval",
6669                 .data           = &ipv6_devconf.rtr_solicit_interval,
6670                 .maxlen         = sizeof(int),
6671                 .mode           = 0644,
6672                 .proc_handler   = proc_dointvec_jiffies,
6673         },
6674         {
6675                 .procname       = "router_solicitation_max_interval",
6676                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6677                 .maxlen         = sizeof(int),
6678                 .mode           = 0644,
6679                 .proc_handler   = proc_dointvec_jiffies,
6680         },
6681         {
6682                 .procname       = "router_solicitation_delay",
6683                 .data           = &ipv6_devconf.rtr_solicit_delay,
6684                 .maxlen         = sizeof(int),
6685                 .mode           = 0644,
6686                 .proc_handler   = proc_dointvec_jiffies,
6687         },
6688         {
6689                 .procname       = "force_mld_version",
6690                 .data           = &ipv6_devconf.force_mld_version,
6691                 .maxlen         = sizeof(int),
6692                 .mode           = 0644,
6693                 .proc_handler   = proc_dointvec,
6694         },
6695         {
6696                 .procname       = "mldv1_unsolicited_report_interval",
6697                 .data           =
6698                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6699                 .maxlen         = sizeof(int),
6700                 .mode           = 0644,
6701                 .proc_handler   = proc_dointvec_ms_jiffies,
6702         },
6703         {
6704                 .procname       = "mldv2_unsolicited_report_interval",
6705                 .data           =
6706                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6707                 .maxlen         = sizeof(int),
6708                 .mode           = 0644,
6709                 .proc_handler   = proc_dointvec_ms_jiffies,
6710         },
6711         {
6712                 .procname       = "use_tempaddr",
6713                 .data           = &ipv6_devconf.use_tempaddr,
6714                 .maxlen         = sizeof(int),
6715                 .mode           = 0644,
6716                 .proc_handler   = proc_dointvec,
6717         },
6718         {
6719                 .procname       = "temp_valid_lft",
6720                 .data           = &ipv6_devconf.temp_valid_lft,
6721                 .maxlen         = sizeof(int),
6722                 .mode           = 0644,
6723                 .proc_handler   = proc_dointvec,
6724         },
6725         {
6726                 .procname       = "temp_prefered_lft",
6727                 .data           = &ipv6_devconf.temp_prefered_lft,
6728                 .maxlen         = sizeof(int),
6729                 .mode           = 0644,
6730                 .proc_handler   = proc_dointvec,
6731         },
6732         {
6733                 .procname       = "regen_max_retry",
6734                 .data           = &ipv6_devconf.regen_max_retry,
6735                 .maxlen         = sizeof(int),
6736                 .mode           = 0644,
6737                 .proc_handler   = proc_dointvec,
6738         },
6739         {
6740                 .procname       = "max_desync_factor",
6741                 .data           = &ipv6_devconf.max_desync_factor,
6742                 .maxlen         = sizeof(int),
6743                 .mode           = 0644,
6744                 .proc_handler   = proc_dointvec,
6745         },
6746         {
6747                 .procname       = "max_addresses",
6748                 .data           = &ipv6_devconf.max_addresses,
6749                 .maxlen         = sizeof(int),
6750                 .mode           = 0644,
6751                 .proc_handler   = proc_dointvec,
6752         },
6753         {
6754                 .procname       = "accept_ra_defrtr",
6755                 .data           = &ipv6_devconf.accept_ra_defrtr,
6756                 .maxlen         = sizeof(int),
6757                 .mode           = 0644,
6758                 .proc_handler   = proc_dointvec,
6759         },
6760         {
6761                 .procname       = "ra_defrtr_metric",
6762                 .data           = &ipv6_devconf.ra_defrtr_metric,
6763                 .maxlen         = sizeof(u32),
6764                 .mode           = 0644,
6765                 .proc_handler   = proc_douintvec_minmax,
6766                 .extra1         = (void *)SYSCTL_ONE,
6767         },
6768         {
6769                 .procname       = "accept_ra_min_hop_limit",
6770                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6771                 .maxlen         = sizeof(int),
6772                 .mode           = 0644,
6773                 .proc_handler   = proc_dointvec,
6774         },
6775         {
6776                 .procname       = "accept_ra_pinfo",
6777                 .data           = &ipv6_devconf.accept_ra_pinfo,
6778                 .maxlen         = sizeof(int),
6779                 .mode           = 0644,
6780                 .proc_handler   = proc_dointvec,
6781         },
6782 #ifdef CONFIG_IPV6_ROUTER_PREF
6783         {
6784                 .procname       = "accept_ra_rtr_pref",
6785                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6786                 .maxlen         = sizeof(int),
6787                 .mode           = 0644,
6788                 .proc_handler   = proc_dointvec,
6789         },
6790         {
6791                 .procname       = "router_probe_interval",
6792                 .data           = &ipv6_devconf.rtr_probe_interval,
6793                 .maxlen         = sizeof(int),
6794                 .mode           = 0644,
6795                 .proc_handler   = proc_dointvec_jiffies,
6796         },
6797 #ifdef CONFIG_IPV6_ROUTE_INFO
6798         {
6799                 .procname       = "accept_ra_rt_info_min_plen",
6800                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6801                 .maxlen         = sizeof(int),
6802                 .mode           = 0644,
6803                 .proc_handler   = proc_dointvec,
6804         },
6805         {
6806                 .procname       = "accept_ra_rt_info_max_plen",
6807                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6808                 .maxlen         = sizeof(int),
6809                 .mode           = 0644,
6810                 .proc_handler   = proc_dointvec,
6811         },
6812 #endif
6813 #endif
6814         {
6815                 .procname       = "proxy_ndp",
6816                 .data           = &ipv6_devconf.proxy_ndp,
6817                 .maxlen         = sizeof(int),
6818                 .mode           = 0644,
6819                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6820         },
6821         {
6822                 .procname       = "accept_source_route",
6823                 .data           = &ipv6_devconf.accept_source_route,
6824                 .maxlen         = sizeof(int),
6825                 .mode           = 0644,
6826                 .proc_handler   = proc_dointvec,
6827         },
6828 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6829         {
6830                 .procname       = "optimistic_dad",
6831                 .data           = &ipv6_devconf.optimistic_dad,
6832                 .maxlen         = sizeof(int),
6833                 .mode           = 0644,
6834                 .proc_handler   = proc_dointvec,
6835         },
6836         {
6837                 .procname       = "use_optimistic",
6838                 .data           = &ipv6_devconf.use_optimistic,
6839                 .maxlen         = sizeof(int),
6840                 .mode           = 0644,
6841                 .proc_handler   = proc_dointvec,
6842         },
6843 #endif
6844 #ifdef CONFIG_IPV6_MROUTE
6845         {
6846                 .procname       = "mc_forwarding",
6847                 .data           = &ipv6_devconf.mc_forwarding,
6848                 .maxlen         = sizeof(int),
6849                 .mode           = 0444,
6850                 .proc_handler   = proc_dointvec,
6851         },
6852 #endif
6853         {
6854                 .procname       = "disable_ipv6",
6855                 .data           = &ipv6_devconf.disable_ipv6,
6856                 .maxlen         = sizeof(int),
6857                 .mode           = 0644,
6858                 .proc_handler   = addrconf_sysctl_disable,
6859         },
6860         {
6861                 .procname       = "accept_dad",
6862                 .data           = &ipv6_devconf.accept_dad,
6863                 .maxlen         = sizeof(int),
6864                 .mode           = 0644,
6865                 .proc_handler   = proc_dointvec,
6866         },
6867         {
6868                 .procname       = "force_tllao",
6869                 .data           = &ipv6_devconf.force_tllao,
6870                 .maxlen         = sizeof(int),
6871                 .mode           = 0644,
6872                 .proc_handler   = proc_dointvec
6873         },
6874         {
6875                 .procname       = "ndisc_notify",
6876                 .data           = &ipv6_devconf.ndisc_notify,
6877                 .maxlen         = sizeof(int),
6878                 .mode           = 0644,
6879                 .proc_handler   = proc_dointvec
6880         },
6881         {
6882                 .procname       = "suppress_frag_ndisc",
6883                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6884                 .maxlen         = sizeof(int),
6885                 .mode           = 0644,
6886                 .proc_handler   = proc_dointvec
6887         },
6888         {
6889                 .procname       = "accept_ra_from_local",
6890                 .data           = &ipv6_devconf.accept_ra_from_local,
6891                 .maxlen         = sizeof(int),
6892                 .mode           = 0644,
6893                 .proc_handler   = proc_dointvec,
6894         },
6895         {
6896                 .procname       = "accept_ra_mtu",
6897                 .data           = &ipv6_devconf.accept_ra_mtu,
6898                 .maxlen         = sizeof(int),
6899                 .mode           = 0644,
6900                 .proc_handler   = proc_dointvec,
6901         },
6902         {
6903                 .procname       = "stable_secret",
6904                 .data           = &ipv6_devconf.stable_secret,
6905                 .maxlen         = IPV6_MAX_STRLEN,
6906                 .mode           = 0600,
6907                 .proc_handler   = addrconf_sysctl_stable_secret,
6908         },
6909         {
6910                 .procname       = "use_oif_addrs_only",
6911                 .data           = &ipv6_devconf.use_oif_addrs_only,
6912                 .maxlen         = sizeof(int),
6913                 .mode           = 0644,
6914                 .proc_handler   = proc_dointvec,
6915         },
6916         {
6917                 .procname       = "ignore_routes_with_linkdown",
6918                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6919                 .maxlen         = sizeof(int),
6920                 .mode           = 0644,
6921                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6922         },
6923         {
6924                 .procname       = "drop_unicast_in_l2_multicast",
6925                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6926                 .maxlen         = sizeof(int),
6927                 .mode           = 0644,
6928                 .proc_handler   = proc_dointvec,
6929         },
6930         {
6931                 .procname       = "drop_unsolicited_na",
6932                 .data           = &ipv6_devconf.drop_unsolicited_na,
6933                 .maxlen         = sizeof(int),
6934                 .mode           = 0644,
6935                 .proc_handler   = proc_dointvec,
6936         },
6937         {
6938                 .procname       = "keep_addr_on_down",
6939                 .data           = &ipv6_devconf.keep_addr_on_down,
6940                 .maxlen         = sizeof(int),
6941                 .mode           = 0644,
6942                 .proc_handler   = proc_dointvec,
6943
6944         },
6945         {
6946                 .procname       = "seg6_enabled",
6947                 .data           = &ipv6_devconf.seg6_enabled,
6948                 .maxlen         = sizeof(int),
6949                 .mode           = 0644,
6950                 .proc_handler   = proc_dointvec,
6951         },
6952 #ifdef CONFIG_IPV6_SEG6_HMAC
6953         {
6954                 .procname       = "seg6_require_hmac",
6955                 .data           = &ipv6_devconf.seg6_require_hmac,
6956                 .maxlen         = sizeof(int),
6957                 .mode           = 0644,
6958                 .proc_handler   = proc_dointvec,
6959         },
6960 #endif
6961         {
6962                 .procname       = "enhanced_dad",
6963                 .data           = &ipv6_devconf.enhanced_dad,
6964                 .maxlen         = sizeof(int),
6965                 .mode           = 0644,
6966                 .proc_handler   = proc_dointvec,
6967         },
6968         {
6969                 .procname       = "addr_gen_mode",
6970                 .data           = &ipv6_devconf.addr_gen_mode,
6971                 .maxlen         = sizeof(int),
6972                 .mode           = 0644,
6973                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6974         },
6975         {
6976                 .procname       = "disable_policy",
6977                 .data           = &ipv6_devconf.disable_policy,
6978                 .maxlen         = sizeof(int),
6979                 .mode           = 0644,
6980                 .proc_handler   = addrconf_sysctl_disable_policy,
6981         },
6982         {
6983                 .procname       = "ndisc_tclass",
6984                 .data           = &ipv6_devconf.ndisc_tclass,
6985                 .maxlen         = sizeof(int),
6986                 .mode           = 0644,
6987                 .proc_handler   = proc_dointvec_minmax,
6988                 .extra1         = (void *)SYSCTL_ZERO,
6989                 .extra2         = (void *)&two_five_five,
6990         },
6991         {
6992                 .procname       = "rpl_seg_enabled",
6993                 .data           = &ipv6_devconf.rpl_seg_enabled,
6994                 .maxlen         = sizeof(int),
6995                 .mode           = 0644,
6996                 .proc_handler   = proc_dointvec,
6997         },
6998         {
6999                 .procname       = "ioam6_enabled",
7000                 .data           = &ipv6_devconf.ioam6_enabled,
7001                 .maxlen         = sizeof(u8),
7002                 .mode           = 0644,
7003                 .proc_handler   = proc_dou8vec_minmax,
7004                 .extra1         = (void *)SYSCTL_ZERO,
7005                 .extra2         = (void *)SYSCTL_ONE,
7006         },
7007         {
7008                 .procname       = "ioam6_id",
7009                 .data           = &ipv6_devconf.ioam6_id,
7010                 .maxlen         = sizeof(u32),
7011                 .mode           = 0644,
7012                 .proc_handler   = proc_douintvec_minmax,
7013                 .extra1         = (void *)SYSCTL_ZERO,
7014                 .extra2         = (void *)&ioam6_if_id_max,
7015         },
7016         {
7017                 .procname       = "ioam6_id_wide",
7018                 .data           = &ipv6_devconf.ioam6_id_wide,
7019                 .maxlen         = sizeof(u32),
7020                 .mode           = 0644,
7021                 .proc_handler   = proc_douintvec,
7022         },
7023         {
7024                 /* sentinel */
7025         }
7026 };
7027
7028 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7029                 struct inet6_dev *idev, struct ipv6_devconf *p)
7030 {
7031         int i, ifindex;
7032         struct ctl_table *table;
7033         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7034
7035         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
7036         if (!table)
7037                 goto out;
7038
7039         for (i = 0; table[i].data; i++) {
7040                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7041                 /* If one of these is already set, then it is not safe to
7042                  * overwrite either of them: this makes proc_dointvec_minmax
7043                  * usable.
7044                  */
7045                 if (!table[i].extra1 && !table[i].extra2) {
7046                         table[i].extra1 = idev; /* embedded; no ref */
7047                         table[i].extra2 = net;
7048                 }
7049         }
7050
7051         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7052
7053         p->sysctl_header = register_net_sysctl(net, path, table);
7054         if (!p->sysctl_header)
7055                 goto free;
7056
7057         if (!strcmp(dev_name, "all"))
7058                 ifindex = NETCONFA_IFINDEX_ALL;
7059         else if (!strcmp(dev_name, "default"))
7060                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7061         else
7062                 ifindex = idev->dev->ifindex;
7063         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7064                                      ifindex, p);
7065         return 0;
7066
7067 free:
7068         kfree(table);
7069 out:
7070         return -ENOBUFS;
7071 }
7072
7073 static void __addrconf_sysctl_unregister(struct net *net,
7074                                          struct ipv6_devconf *p, int ifindex)
7075 {
7076         struct ctl_table *table;
7077
7078         if (!p->sysctl_header)
7079                 return;
7080
7081         table = p->sysctl_header->ctl_table_arg;
7082         unregister_net_sysctl_table(p->sysctl_header);
7083         p->sysctl_header = NULL;
7084         kfree(table);
7085
7086         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7087 }
7088
7089 static int addrconf_sysctl_register(struct inet6_dev *idev)
7090 {
7091         int err;
7092
7093         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7094                 return -EINVAL;
7095
7096         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7097                                     &ndisc_ifinfo_sysctl_change);
7098         if (err)
7099                 return err;
7100         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7101                                          idev, &idev->cnf);
7102         if (err)
7103                 neigh_sysctl_unregister(idev->nd_parms);
7104
7105         return err;
7106 }
7107
7108 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7109 {
7110         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7111                                      idev->dev->ifindex);
7112         neigh_sysctl_unregister(idev->nd_parms);
7113 }
7114
7115
7116 #endif
7117
7118 static int __net_init addrconf_init_net(struct net *net)
7119 {
7120         int err = -ENOMEM;
7121         struct ipv6_devconf *all, *dflt;
7122
7123         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7124         if (!all)
7125                 goto err_alloc_all;
7126
7127         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7128         if (!dflt)
7129                 goto err_alloc_dflt;
7130
7131         if (!net_eq(net, &init_net)) {
7132                 switch (net_inherit_devconf()) {
7133                 case 1:  /* copy from init_net */
7134                         memcpy(all, init_net.ipv6.devconf_all,
7135                                sizeof(ipv6_devconf));
7136                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7137                                sizeof(ipv6_devconf_dflt));
7138                         break;
7139                 case 3: /* copy from the current netns */
7140                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7141                                sizeof(ipv6_devconf));
7142                         memcpy(dflt,
7143                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7144                                sizeof(ipv6_devconf_dflt));
7145                         break;
7146                 case 0:
7147                 case 2:
7148                         /* use compiled values */
7149                         break;
7150                 }
7151         }
7152
7153         /* these will be inherited by all namespaces */
7154         dflt->autoconf = ipv6_defaults.autoconf;
7155         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7156
7157         dflt->stable_secret.initialized = false;
7158         all->stable_secret.initialized = false;
7159
7160         net->ipv6.devconf_all = all;
7161         net->ipv6.devconf_dflt = dflt;
7162
7163 #ifdef CONFIG_SYSCTL
7164         err = __addrconf_sysctl_register(net, "all", NULL, all);
7165         if (err < 0)
7166                 goto err_reg_all;
7167
7168         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7169         if (err < 0)
7170                 goto err_reg_dflt;
7171 #endif
7172         return 0;
7173
7174 #ifdef CONFIG_SYSCTL
7175 err_reg_dflt:
7176         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7177 err_reg_all:
7178         kfree(dflt);
7179 #endif
7180 err_alloc_dflt:
7181         kfree(all);
7182 err_alloc_all:
7183         return err;
7184 }
7185
7186 static void __net_exit addrconf_exit_net(struct net *net)
7187 {
7188 #ifdef CONFIG_SYSCTL
7189         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7190                                      NETCONFA_IFINDEX_DEFAULT);
7191         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7192                                      NETCONFA_IFINDEX_ALL);
7193 #endif
7194         kfree(net->ipv6.devconf_dflt);
7195         kfree(net->ipv6.devconf_all);
7196 }
7197
7198 static struct pernet_operations addrconf_ops = {
7199         .init = addrconf_init_net,
7200         .exit = addrconf_exit_net,
7201 };
7202
7203 static struct rtnl_af_ops inet6_ops __read_mostly = {
7204         .family           = AF_INET6,
7205         .fill_link_af     = inet6_fill_link_af,
7206         .get_link_af_size = inet6_get_link_af_size,
7207         .validate_link_af = inet6_validate_link_af,
7208         .set_link_af      = inet6_set_link_af,
7209 };
7210
7211 /*
7212  *      Init / cleanup code
7213  */
7214
7215 int __init addrconf_init(void)
7216 {
7217         struct inet6_dev *idev;
7218         int i, err;
7219
7220         err = ipv6_addr_label_init();
7221         if (err < 0) {
7222                 pr_crit("%s: cannot initialize default policy table: %d\n",
7223                         __func__, err);
7224                 goto out;
7225         }
7226
7227         err = register_pernet_subsys(&addrconf_ops);
7228         if (err < 0)
7229                 goto out_addrlabel;
7230
7231         addrconf_wq = create_workqueue("ipv6_addrconf");
7232         if (!addrconf_wq) {
7233                 err = -ENOMEM;
7234                 goto out_nowq;
7235         }
7236
7237         /* The addrconf netdev notifier requires that loopback_dev
7238          * has it's ipv6 private information allocated and setup
7239          * before it can bring up and give link-local addresses
7240          * to other devices which are up.
7241          *
7242          * Unfortunately, loopback_dev is not necessarily the first
7243          * entry in the global dev_base list of net devices.  In fact,
7244          * it is likely to be the very last entry on that list.
7245          * So this causes the notifier registry below to try and
7246          * give link-local addresses to all devices besides loopback_dev
7247          * first, then loopback_dev, which cases all the non-loopback_dev
7248          * devices to fail to get a link-local address.
7249          *
7250          * So, as a temporary fix, allocate the ipv6 structure for
7251          * loopback_dev first by hand.
7252          * Longer term, all of the dependencies ipv6 has upon the loopback
7253          * device and it being up should be removed.
7254          */
7255         rtnl_lock();
7256         idev = ipv6_add_dev(init_net.loopback_dev);
7257         rtnl_unlock();
7258         if (IS_ERR(idev)) {
7259                 err = PTR_ERR(idev);
7260                 goto errlo;
7261         }
7262
7263         ip6_route_init_special_entries();
7264
7265         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7266                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7267
7268         register_netdevice_notifier(&ipv6_dev_notf);
7269
7270         addrconf_verify();
7271
7272         rtnl_af_register(&inet6_ops);
7273
7274         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7275                                    NULL, inet6_dump_ifinfo, 0);
7276         if (err < 0)
7277                 goto errout;
7278
7279         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7280                                    inet6_rtm_newaddr, NULL, 0);
7281         if (err < 0)
7282                 goto errout;
7283         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7284                                    inet6_rtm_deladdr, NULL, 0);
7285         if (err < 0)
7286                 goto errout;
7287         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7288                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7289                                    RTNL_FLAG_DOIT_UNLOCKED);
7290         if (err < 0)
7291                 goto errout;
7292         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7293                                    NULL, inet6_dump_ifmcaddr, 0);
7294         if (err < 0)
7295                 goto errout;
7296         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7297                                    NULL, inet6_dump_ifacaddr, 0);
7298         if (err < 0)
7299                 goto errout;
7300         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7301                                    inet6_netconf_get_devconf,
7302                                    inet6_netconf_dump_devconf,
7303                                    RTNL_FLAG_DOIT_UNLOCKED);
7304         if (err < 0)
7305                 goto errout;
7306         err = ipv6_addr_label_rtnl_register();
7307         if (err < 0)
7308                 goto errout;
7309
7310         return 0;
7311 errout:
7312         rtnl_unregister_all(PF_INET6);
7313         rtnl_af_unregister(&inet6_ops);
7314         unregister_netdevice_notifier(&ipv6_dev_notf);
7315 errlo:
7316         destroy_workqueue(addrconf_wq);
7317 out_nowq:
7318         unregister_pernet_subsys(&addrconf_ops);
7319 out_addrlabel:
7320         ipv6_addr_label_cleanup();
7321 out:
7322         return err;
7323 }
7324
7325 void addrconf_cleanup(void)
7326 {
7327         struct net_device *dev;
7328         int i;
7329
7330         unregister_netdevice_notifier(&ipv6_dev_notf);
7331         unregister_pernet_subsys(&addrconf_ops);
7332         ipv6_addr_label_cleanup();
7333
7334         rtnl_af_unregister(&inet6_ops);
7335
7336         rtnl_lock();
7337
7338         /* clean dev list */
7339         for_each_netdev(&init_net, dev) {
7340                 if (__in6_dev_get(dev) == NULL)
7341                         continue;
7342                 addrconf_ifdown(dev, true);
7343         }
7344         addrconf_ifdown(init_net.loopback_dev, true);
7345
7346         /*
7347          *      Check hash table.
7348          */
7349         spin_lock_bh(&addrconf_hash_lock);
7350         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7351                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7352         spin_unlock_bh(&addrconf_hash_lock);
7353         cancel_delayed_work(&addr_chk_work);
7354         rtnl_unlock();
7355
7356         destroy_workqueue(addrconf_wq);
7357 }