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 static void addrconf_verify(struct net *net);
 151 static void addrconf_verify_rtnl(struct net *net);
 152
 153 static struct workqueue_struct *addrconf_wq;
 154
 155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 157
 158 static void addrconf_type_change(struct net_device *dev,
 159                                  unsigned long event);
 160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 161
 162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 163                                                   int plen,
 164                                                   const struct net_device *dev,
 165                                                   u32 flags, u32 noflags,
 166                                                   bool no_gw);
 167
 168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 169 static void addrconf_dad_work(struct work_struct *w);
 170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 171                                    bool send_na);
 172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 173 static void addrconf_rs_timer(struct timer_list *t);
 174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 176
 177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 178                                 struct prefix_info *pinfo);
 179
 180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 181         .forwarding             = 0,
 182         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 183         .mtu6                   = IPV6_MIN_MTU,
 184         .accept_ra              = 1,
 185         .accept_redirects       = 1,
 186         .autoconf               = 1,
 187         .force_mld_version      = 0,
 188         .mldv1_unsolicited_report_interval = 10 * HZ,
 189         .mldv2_unsolicited_report_interval = HZ,
 190         .dad_transmits          = 1,
 191         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 192         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 193         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 194         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 195         .use_tempaddr           = 0,
 196         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 197         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 198         .regen_max_retry        = REGEN_MAX_RETRY,
 199         .max_desync_factor      = MAX_DESYNC_FACTOR,
 200         .max_addresses          = IPV6_MAX_ADDRESSES,
 201         .accept_ra_defrtr       = 1,
 202         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 203         .accept_ra_from_local   = 0,
 204         .accept_ra_min_hop_limit= 1,
 205         .accept_ra_pinfo        = 1,
 206 #ifdef CONFIG_IPV6_ROUTER_PREF
 207         .accept_ra_rtr_pref     = 1,
 208         .rtr_probe_interval     = 60 * HZ,
 209 #ifdef CONFIG_IPV6_ROUTE_INFO
 210         .accept_ra_rt_info_min_plen = 0,
 211         .accept_ra_rt_info_max_plen = 0,
 212 #endif
 213 #endif
 214         .proxy_ndp              = 0,
 215         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 216         .disable_ipv6           = 0,
 217         .accept_dad             = 0,
 218         .suppress_frag_ndisc    = 1,
 219         .accept_ra_mtu          = 1,
 220         .stable_secret          = {
 221                 .initialized = false,
 222         },
 223         .use_oif_addrs_only     = 0,
 224         .ignore_routes_with_linkdown = 0,
 225         .keep_addr_on_down      = 0,
 226         .seg6_enabled           = 0,
 227 #ifdef CONFIG_IPV6_SEG6_HMAC
 228         .seg6_require_hmac      = 0,
 229 #endif
 230         .enhanced_dad           = 1,
 231         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 232         .disable_policy         = 0,
 233         .rpl_seg_enabled        = 0,
 234         .ioam6_enabled          = 0,
 235         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 236         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 237         .ndisc_evict_nocarrier  = 1,
 238 };
 239
 240 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 241         .forwarding             = 0,
 242         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 243         .mtu6                   = IPV6_MIN_MTU,
 244         .accept_ra              = 1,
 245         .accept_redirects       = 1,
 246         .autoconf               = 1,
 247         .force_mld_version      = 0,
 248         .mldv1_unsolicited_report_interval = 10 * HZ,
 249         .mldv2_unsolicited_report_interval = HZ,
 250         .dad_transmits          = 1,
 251         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 252         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 253         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 254         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 255         .use_tempaddr           = 0,
 256         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 257         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 258         .regen_max_retry        = REGEN_MAX_RETRY,
 259         .max_desync_factor      = MAX_DESYNC_FACTOR,
 260         .max_addresses          = IPV6_MAX_ADDRESSES,
 261         .accept_ra_defrtr       = 1,
 262         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 263         .accept_ra_from_local   = 0,
 264         .accept_ra_min_hop_limit= 1,
 265         .accept_ra_pinfo        = 1,
 266 #ifdef CONFIG_IPV6_ROUTER_PREF
 267         .accept_ra_rtr_pref     = 1,
 268         .rtr_probe_interval     = 60 * HZ,
 269 #ifdef CONFIG_IPV6_ROUTE_INFO
 270         .accept_ra_rt_info_min_plen = 0,
 271         .accept_ra_rt_info_max_plen = 0,
 272 #endif
 273 #endif
 274         .proxy_ndp              = 0,
 275         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 276         .disable_ipv6           = 0,
 277         .accept_dad             = 1,
 278         .suppress_frag_ndisc    = 1,
 279         .accept_ra_mtu          = 1,
 280         .stable_secret          = {
 281                 .initialized = false,
 282         },
 283         .use_oif_addrs_only     = 0,
 284         .ignore_routes_with_linkdown = 0,
 285         .keep_addr_on_down      = 0,
 286         .seg6_enabled           = 0,
 287 #ifdef CONFIG_IPV6_SEG6_HMAC
 288         .seg6_require_hmac      = 0,
 289 #endif
 290         .enhanced_dad           = 1,
 291         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 292         .disable_policy         = 0,
 293         .rpl_seg_enabled        = 0,
 294         .ioam6_enabled          = 0,
 295         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 296         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 297         .ndisc_evict_nocarrier  = 1,
 298 };
 299
 300 /* Check if link is ready: is it up and is a valid qdisc available */
 301 static inline bool addrconf_link_ready(const struct net_device *dev)
 302 {
 303         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 304 }
 305
 306 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 307 {
 308         if (del_timer(&idev->rs_timer))
 309                 __in6_dev_put(idev);
 310 }
 311
 312 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 313 {
 314         if (cancel_delayed_work(&ifp->dad_work))
 315                 __in6_ifa_put(ifp);
 316 }
 317
 318 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 319                                   unsigned long when)
 320 {
 321         if (!timer_pending(&idev->rs_timer))
 322                 in6_dev_hold(idev);
 323         mod_timer(&idev->rs_timer, jiffies + when);
 324 }
 325
 326 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 327                                    unsigned long delay)
 328 {
 329         in6_ifa_hold(ifp);
 330         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 331                 in6_ifa_put(ifp);
 332 }
 333
 334 static int snmp6_alloc_dev(struct inet6_dev *idev)
 335 {
 336         int i;
 337
 338         idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
 339         if (!idev->stats.ipv6)
 340                 goto err_ip;
 341
 342         for_each_possible_cpu(i) {
 343                 struct ipstats_mib *addrconf_stats;
 344                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 345                 u64_stats_init(&addrconf_stats->syncp);
 346         }
 347
 348
 349         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 350                                         GFP_KERNEL);
 351         if (!idev->stats.icmpv6dev)
 352                 goto err_icmp;
 353         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 354                                            GFP_KERNEL_ACCOUNT);
 355         if (!idev->stats.icmpv6msgdev)
 356                 goto err_icmpmsg;
 357
 358         return 0;
 359
 360 err_icmpmsg:
 361         kfree(idev->stats.icmpv6dev);
 362 err_icmp:
 363         free_percpu(idev->stats.ipv6);
 364 err_ip:
 365         return -ENOMEM;
 366 }
 367
 368 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 369 {
 370         struct inet6_dev *ndev;
 371         int err = -ENOMEM;
 372
 373         ASSERT_RTNL();
 374
 375         if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
 376                 return ERR_PTR(-EINVAL);
 377
 378         ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
 379         if (!ndev)
 380                 return ERR_PTR(err);
 381
 382         rwlock_init(&ndev->lock);
 383         ndev->dev = dev;
 384         INIT_LIST_HEAD(&ndev->addr_list);
 385         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 386         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 387
 388         if (ndev->cnf.stable_secret.initialized)
 389                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 390
 391         ndev->cnf.mtu6 = dev->mtu;
 392         ndev->ra_mtu = 0;
 393         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 394         if (!ndev->nd_parms) {
 395                 kfree(ndev);
 396                 return ERR_PTR(err);
 397         }
 398         if (ndev->cnf.forwarding)
 399                 dev_disable_lro(dev);
 400         /* We refer to the device */
 401         netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
 402
 403         if (snmp6_alloc_dev(ndev) < 0) {
 404                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 405                            __func__);
 406                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 407                 netdev_put(dev, &ndev->dev_tracker);
 408                 kfree(ndev);
 409                 return ERR_PTR(err);
 410         }
 411
 412         if (dev != blackhole_netdev) {
 413                 if (snmp6_register_dev(ndev) < 0) {
 414                         netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 415                                    __func__, dev->name);
 416                         goto err_release;
 417                 }
 418         }
 419         /* One reference from device. */
 420         refcount_set(&ndev->refcnt, 1);
 421
 422         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 423                 ndev->cnf.accept_dad = -1;
 424
 425 #if IS_ENABLED(CONFIG_IPV6_SIT)
 426         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 427                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 428                 ndev->cnf.rtr_solicits = 0;
 429         }
 430 #endif
 431
 432         INIT_LIST_HEAD(&ndev->tempaddr_list);
 433         ndev->desync_factor = U32_MAX;
 434         if ((dev->flags&IFF_LOOPBACK) ||
 435             dev->type == ARPHRD_TUNNEL ||
 436             dev->type == ARPHRD_TUNNEL6 ||
 437             dev->type == ARPHRD_SIT ||
 438             dev->type == ARPHRD_NONE) {
 439                 ndev->cnf.use_tempaddr = -1;
 440         }
 441
 442         ndev->token = in6addr_any;
 443
 444         if (netif_running(dev) && addrconf_link_ready(dev))
 445                 ndev->if_flags |= IF_READY;
 446
 447         ipv6_mc_init_dev(ndev);
 448         ndev->tstamp = jiffies;
 449         if (dev != blackhole_netdev) {
 450                 err = addrconf_sysctl_register(ndev);
 451                 if (err) {
 452                         ipv6_mc_destroy_dev(ndev);
 453                         snmp6_unregister_dev(ndev);
 454                         goto err_release;
 455                 }
 456         }
 457         /* protected by rtnl_lock */
 458         rcu_assign_pointer(dev->ip6_ptr, ndev);
 459
 460         if (dev != blackhole_netdev) {
 461                 /* Join interface-local all-node multicast group */
 462                 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 463
 464                 /* Join all-node multicast group */
 465                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 466
 467                 /* Join all-router multicast group if forwarding is set */
 468                 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 469                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 470         }
 471         return ndev;
 472
 473 err_release:
 474         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 475         ndev->dead = 1;
 476         in6_dev_finish_destroy(ndev);
 477         return ERR_PTR(err);
 478 }
 479
 480 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 481 {
 482         struct inet6_dev *idev;
 483
 484         ASSERT_RTNL();
 485
 486         idev = __in6_dev_get(dev);
 487         if (!idev) {
 488                 idev = ipv6_add_dev(dev);
 489                 if (IS_ERR(idev))
 490                         return idev;
 491         }
 492
 493         if (dev->flags&IFF_UP)
 494                 ipv6_mc_up(idev);
 495         return idev;
 496 }
 497
 498 static int inet6_netconf_msgsize_devconf(int type)
 499 {
 500         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 501                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 502         bool all = false;
 503
 504         if (type == NETCONFA_ALL)
 505                 all = true;
 506
 507         if (all || type == NETCONFA_FORWARDING)
 508                 size += nla_total_size(4);
 509 #ifdef CONFIG_IPV6_MROUTE
 510         if (all || type == NETCONFA_MC_FORWARDING)
 511                 size += nla_total_size(4);
 512 #endif
 513         if (all || type == NETCONFA_PROXY_NEIGH)
 514                 size += nla_total_size(4);
 515
 516         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 517                 size += nla_total_size(4);
 518
 519         return size;
 520 }
 521
 522 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 523                                       struct ipv6_devconf *devconf, u32 portid,
 524                                       u32 seq, int event, unsigned int flags,
 525                                       int type)
 526 {
 527         struct nlmsghdr  *nlh;
 528         struct netconfmsg *ncm;
 529         bool all = false;
 530
 531         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 532                         flags);
 533         if (!nlh)
 534                 return -EMSGSIZE;
 535
 536         if (type == NETCONFA_ALL)
 537                 all = true;
 538
 539         ncm = nlmsg_data(nlh);
 540         ncm->ncm_family = AF_INET6;
 541
 542         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 543                 goto nla_put_failure;
 544
 545         if (!devconf)
 546                 goto out;
 547
 548         if ((all || type == NETCONFA_FORWARDING) &&
 549             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 550                 goto nla_put_failure;
 551 #ifdef CONFIG_IPV6_MROUTE
 552         if ((all || type == NETCONFA_MC_FORWARDING) &&
 553             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 554                         atomic_read(&devconf->mc_forwarding)) < 0)
 555                 goto nla_put_failure;
 556 #endif
 557         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 558             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 559                 goto nla_put_failure;
 560
 561         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 562             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 563                         devconf->ignore_routes_with_linkdown) < 0)
 564                 goto nla_put_failure;
 565
 566 out:
 567         nlmsg_end(skb, nlh);
 568         return 0;
 569
 570 nla_put_failure:
 571         nlmsg_cancel(skb, nlh);
 572         return -EMSGSIZE;
 573 }
 574
 575 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 576                                   int ifindex, struct ipv6_devconf *devconf)
 577 {
 578         struct sk_buff *skb;
 579         int err = -ENOBUFS;
 580
 581         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 582         if (!skb)
 583                 goto errout;
 584
 585         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 586                                          event, 0, type);
 587         if (err < 0) {
 588                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 589                 WARN_ON(err == -EMSGSIZE);
 590                 kfree_skb(skb);
 591                 goto errout;
 592         }
 593         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 594         return;
 595 errout:
 596         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 597 }
 598
 599 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 600         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 601         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 602         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 603         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 604 };
 605
 606 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 607                                        const struct nlmsghdr *nlh,
 608                                        struct nlattr **tb,
 609                                        struct netlink_ext_ack *extack)
 610 {
 611         int i, err;
 612
 613         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 614                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 615                 return -EINVAL;
 616         }
 617
 618         if (!netlink_strict_get_check(skb))
 619                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 620                                               tb, NETCONFA_MAX,
 621                                               devconf_ipv6_policy, extack);
 622
 623         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 624                                             tb, NETCONFA_MAX,
 625                                             devconf_ipv6_policy, extack);
 626         if (err)
 627                 return err;
 628
 629         for (i = 0; i <= NETCONFA_MAX; i++) {
 630                 if (!tb[i])
 631                         continue;
 632
 633                 switch (i) {
 634                 case NETCONFA_IFINDEX:
 635                         break;
 636                 default:
 637                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 638                         return -EINVAL;
 639                 }
 640         }
 641
 642         return 0;
 643 }
 644
 645 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 646                                      struct nlmsghdr *nlh,
 647                                      struct netlink_ext_ack *extack)
 648 {
 649         struct net *net = sock_net(in_skb->sk);
 650         struct nlattr *tb[NETCONFA_MAX+1];
 651         struct inet6_dev *in6_dev = NULL;
 652         struct net_device *dev = NULL;
 653         struct sk_buff *skb;
 654         struct ipv6_devconf *devconf;
 655         int ifindex;
 656         int err;
 657
 658         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 659         if (err < 0)
 660                 return err;
 661
 662         if (!tb[NETCONFA_IFINDEX])
 663                 return -EINVAL;
 664
 665         err = -EINVAL;
 666         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 667         switch (ifindex) {
 668         case NETCONFA_IFINDEX_ALL:
 669                 devconf = net->ipv6.devconf_all;
 670                 break;
 671         case NETCONFA_IFINDEX_DEFAULT:
 672                 devconf = net->ipv6.devconf_dflt;
 673                 break;
 674         default:
 675                 dev = dev_get_by_index(net, ifindex);
 676                 if (!dev)
 677                         return -EINVAL;
 678                 in6_dev = in6_dev_get(dev);
 679                 if (!in6_dev)
 680                         goto errout;
 681                 devconf = &in6_dev->cnf;
 682                 break;
 683         }
 684
 685         err = -ENOBUFS;
 686         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 687         if (!skb)
 688                 goto errout;
 689
 690         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 691                                          NETLINK_CB(in_skb).portid,
 692                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 693                                          NETCONFA_ALL);
 694         if (err < 0) {
 695                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 696                 WARN_ON(err == -EMSGSIZE);
 697                 kfree_skb(skb);
 698                 goto errout;
 699         }
 700         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 701 errout:
 702         if (in6_dev)
 703                 in6_dev_put(in6_dev);
 704         dev_put(dev);
 705         return err;
 706 }
 707
 708 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 709                                       struct netlink_callback *cb)
 710 {
 711         const struct nlmsghdr *nlh = cb->nlh;
 712         struct net *net = sock_net(skb->sk);
 713         int h, s_h;
 714         int idx, s_idx;
 715         struct net_device *dev;
 716         struct inet6_dev *idev;
 717         struct hlist_head *head;
 718
 719         if (cb->strict_check) {
 720                 struct netlink_ext_ack *extack = cb->extack;
 721                 struct netconfmsg *ncm;
 722
 723                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 724                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 725                         return -EINVAL;
 726                 }
 727
 728                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 729                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 730                         return -EINVAL;
 731                 }
 732         }
 733
 734         s_h = cb->args[0];
 735         s_idx = idx = cb->args[1];
 736
 737         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 738                 idx = 0;
 739                 head = &net->dev_index_head[h];
 740                 rcu_read_lock();
 741                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 742                           net->dev_base_seq;
 743                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 744                         if (idx < s_idx)
 745                                 goto cont;
 746                         idev = __in6_dev_get(dev);
 747                         if (!idev)
 748                                 goto cont;
 749
 750                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 751                                                        &idev->cnf,
 752                                                        NETLINK_CB(cb->skb).portid,
 753                                                        nlh->nlmsg_seq,
 754                                                        RTM_NEWNETCONF,
 755                                                        NLM_F_MULTI,
 756                                                        NETCONFA_ALL) < 0) {
 757                                 rcu_read_unlock();
 758                                 goto done;
 759                         }
 760                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 761 cont:
 762                         idx++;
 763                 }
 764                 rcu_read_unlock();
 765         }
 766         if (h == NETDEV_HASHENTRIES) {
 767                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 768                                                net->ipv6.devconf_all,
 769                                                NETLINK_CB(cb->skb).portid,
 770                                                nlh->nlmsg_seq,
 771                                                RTM_NEWNETCONF, NLM_F_MULTI,
 772                                                NETCONFA_ALL) < 0)
 773                         goto done;
 774                 else
 775                         h++;
 776         }
 777         if (h == NETDEV_HASHENTRIES + 1) {
 778                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 779                                                net->ipv6.devconf_dflt,
 780                                                NETLINK_CB(cb->skb).portid,
 781                                                nlh->nlmsg_seq,
 782                                                RTM_NEWNETCONF, NLM_F_MULTI,
 783                                                NETCONFA_ALL) < 0)
 784                         goto done;
 785                 else
 786                         h++;
 787         }
 788 done:
 789         cb->args[0] = h;
 790         cb->args[1] = idx;
 791
 792         return skb->len;
 793 }
 794
 795 #ifdef CONFIG_SYSCTL
 796 static void dev_forward_change(struct inet6_dev *idev)
 797 {
 798         struct net_device *dev;
 799         struct inet6_ifaddr *ifa;
 800         LIST_HEAD(tmp_addr_list);
 801
 802         if (!idev)
 803                 return;
 804         dev = idev->dev;
 805         if (idev->cnf.forwarding)
 806                 dev_disable_lro(dev);
 807         if (dev->flags & IFF_MULTICAST) {
 808                 if (idev->cnf.forwarding) {
 809                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 810                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 811                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 812                 } else {
 813                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 814                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 815                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 816                 }
 817         }
 818
 819         read_lock_bh(&idev->lock);
 820         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 821                 if (ifa->flags&IFA_F_TENTATIVE)
 822                         continue;
 823                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
 824         }
 825         read_unlock_bh(&idev->lock);
 826
 827         while (!list_empty(&tmp_addr_list)) {
 828                 ifa = list_first_entry(&tmp_addr_list,
 829                                        struct inet6_ifaddr, if_list_aux);
 830                 list_del(&ifa->if_list_aux);
 831                 if (idev->cnf.forwarding)
 832                         addrconf_join_anycast(ifa);
 833                 else
 834                         addrconf_leave_anycast(ifa);
 835         }
 836
 837         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 838                                      NETCONFA_FORWARDING,
 839                                      dev->ifindex, &idev->cnf);
 840 }
 841
 842
 843 static void addrconf_forward_change(struct net *net, __s32 newf)
 844 {
 845         struct net_device *dev;
 846         struct inet6_dev *idev;
 847
 848         for_each_netdev(net, dev) {
 849                 idev = __in6_dev_get(dev);
 850                 if (idev) {
 851                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 852                         idev->cnf.forwarding = newf;
 853                         if (changed)
 854                                 dev_forward_change(idev);
 855                 }
 856         }
 857 }
 858
 859 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 860 {
 861         struct net *net;
 862         int old;
 863
 864         if (!rtnl_trylock())
 865                 return restart_syscall();
 866
 867         net = (struct net *)table->extra2;
 868         old = *p;
 869         *p = newf;
 870
 871         if (p == &net->ipv6.devconf_dflt->forwarding) {
 872                 if ((!newf) ^ (!old))
 873                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 874                                                      NETCONFA_FORWARDING,
 875                                                      NETCONFA_IFINDEX_DEFAULT,
 876                                                      net->ipv6.devconf_dflt);
 877                 rtnl_unlock();
 878                 return 0;
 879         }
 880
 881         if (p == &net->ipv6.devconf_all->forwarding) {
 882                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 883
 884                 net->ipv6.devconf_dflt->forwarding = newf;
 885                 if ((!newf) ^ (!old_dflt))
 886                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 887                                                      NETCONFA_FORWARDING,
 888                                                      NETCONFA_IFINDEX_DEFAULT,
 889                                                      net->ipv6.devconf_dflt);
 890
 891                 addrconf_forward_change(net, newf);
 892                 if ((!newf) ^ (!old))
 893                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 894                                                      NETCONFA_FORWARDING,
 895                                                      NETCONFA_IFINDEX_ALL,
 896                                                      net->ipv6.devconf_all);
 897         } else if ((!newf) ^ (!old))
 898                 dev_forward_change((struct inet6_dev *)table->extra1);
 899         rtnl_unlock();
 900
 901         if (newf)
 902                 rt6_purge_dflt_routers(net);
 903         return 1;
 904 }
 905
 906 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 907 {
 908         struct net_device *dev;
 909         struct inet6_dev *idev;
 910
 911         for_each_netdev(net, dev) {
 912                 idev = __in6_dev_get(dev);
 913                 if (idev) {
 914                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 915
 916                         idev->cnf.ignore_routes_with_linkdown = newf;
 917                         if (changed)
 918                                 inet6_netconf_notify_devconf(dev_net(dev),
 919                                                              RTM_NEWNETCONF,
 920                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 921                                                              dev->ifindex,
 922                                                              &idev->cnf);
 923                 }
 924         }
 925 }
 926
 927 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 928 {
 929         struct net *net;
 930         int old;
 931
 932         if (!rtnl_trylock())
 933                 return restart_syscall();
 934
 935         net = (struct net *)table->extra2;
 936         old = *p;
 937         *p = newf;
 938
 939         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 940                 if ((!newf) ^ (!old))
 941                         inet6_netconf_notify_devconf(net,
 942                                                      RTM_NEWNETCONF,
 943                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 944                                                      NETCONFA_IFINDEX_DEFAULT,
 945                                                      net->ipv6.devconf_dflt);
 946                 rtnl_unlock();
 947                 return 0;
 948         }
 949
 950         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 951                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 952                 addrconf_linkdown_change(net, newf);
 953                 if ((!newf) ^ (!old))
 954                         inet6_netconf_notify_devconf(net,
 955                                                      RTM_NEWNETCONF,
 956                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 957                                                      NETCONFA_IFINDEX_ALL,
 958                                                      net->ipv6.devconf_all);
 959         }
 960         rtnl_unlock();
 961
 962         return 1;
 963 }
 964
 965 #endif
 966
 967 /* Nobody refers to this ifaddr, destroy it */
 968 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 969 {
 970         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 971
 972 #ifdef NET_REFCNT_DEBUG
 973         pr_debug("%s\n", __func__);
 974 #endif
 975
 976         in6_dev_put(ifp->idev);
 977
 978         if (cancel_delayed_work(&ifp->dad_work))
 979                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 980                           ifp);
 981
 982         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 983                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 984                 return;
 985         }
 986
 987         kfree_rcu(ifp, rcu);
 988 }
 989
 990 static void
 991 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 992 {
 993         struct list_head *p;
 994         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 995
 996         /*
 997          * Each device address list is sorted in order of scope -
 998          * global before linklocal.
 999          */
1000         list_for_each(p, &idev->addr_list) {
1001                 struct inet6_ifaddr *ifa
1002                         = list_entry(p, struct inet6_ifaddr, if_list);
1003                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1004                         break;
1005         }
1006
1007         list_add_tail_rcu(&ifp->if_list, p);
1008 }
1009
1010 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1011 {
1012         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1013
1014         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1015 }
1016
1017 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1018                                struct net_device *dev, unsigned int hash)
1019 {
1020         struct inet6_ifaddr *ifp;
1021
1022         hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1023                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1024                         if (!dev || ifp->idev->dev == dev)
1025                                 return true;
1026                 }
1027         }
1028         return false;
1029 }
1030
1031 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1032 {
1033         struct net *net = dev_net(dev);
1034         unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1035         int err = 0;
1036
1037         spin_lock(&net->ipv6.addrconf_hash_lock);
1038
1039         /* Ignore adding duplicate addresses on an interface */
1040         if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1041                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1042                 err = -EEXIST;
1043         } else {
1044                 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1045         }
1046
1047         spin_unlock(&net->ipv6.addrconf_hash_lock);
1048
1049         return err;
1050 }
1051
1052 /* On success it returns ifp with increased reference count */
1053
1054 static struct inet6_ifaddr *
1055 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1056               bool can_block, struct netlink_ext_ack *extack)
1057 {
1058         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1059         int addr_type = ipv6_addr_type(cfg->pfx);
1060         struct net *net = dev_net(idev->dev);
1061         struct inet6_ifaddr *ifa = NULL;
1062         struct fib6_info *f6i = NULL;
1063         int err = 0;
1064
1065         if (addr_type == IPV6_ADDR_ANY ||
1066             (addr_type & IPV6_ADDR_MULTICAST &&
1067              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1068             (!(idev->dev->flags & IFF_LOOPBACK) &&
1069              !netif_is_l3_master(idev->dev) &&
1070              addr_type & IPV6_ADDR_LOOPBACK))
1071                 return ERR_PTR(-EADDRNOTAVAIL);
1072
1073         if (idev->dead) {
1074                 err = -ENODEV;                  /*XXX*/
1075                 goto out;
1076         }
1077
1078         if (idev->cnf.disable_ipv6) {
1079                 err = -EACCES;
1080                 goto out;
1081         }
1082
1083         /* validator notifier needs to be blocking;
1084          * do not call in atomic context
1085          */
1086         if (can_block) {
1087                 struct in6_validator_info i6vi = {
1088                         .i6vi_addr = *cfg->pfx,
1089                         .i6vi_dev = idev,
1090                         .extack = extack,
1091                 };
1092
1093                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1094                 err = notifier_to_errno(err);
1095                 if (err < 0)
1096                         goto out;
1097         }
1098
1099         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1100         if (!ifa) {
1101                 err = -ENOBUFS;
1102                 goto out;
1103         }
1104
1105         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1106         if (IS_ERR(f6i)) {
1107                 err = PTR_ERR(f6i);
1108                 f6i = NULL;
1109                 goto out;
1110         }
1111
1112         if (net->ipv6.devconf_all->disable_policy ||
1113             idev->cnf.disable_policy)
1114                 f6i->dst_nopolicy = true;
1115
1116         neigh_parms_data_state_setall(idev->nd_parms);
1117
1118         ifa->addr = *cfg->pfx;
1119         if (cfg->peer_pfx)
1120                 ifa->peer_addr = *cfg->peer_pfx;
1121
1122         spin_lock_init(&ifa->lock);
1123         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1124         INIT_HLIST_NODE(&ifa->addr_lst);
1125         ifa->scope = cfg->scope;
1126         ifa->prefix_len = cfg->plen;
1127         ifa->rt_priority = cfg->rt_priority;
1128         ifa->flags = cfg->ifa_flags;
1129         ifa->ifa_proto = cfg->ifa_proto;
1130         /* No need to add the TENTATIVE flag for addresses with NODAD */
1131         if (!(cfg->ifa_flags & IFA_F_NODAD))
1132                 ifa->flags |= IFA_F_TENTATIVE;
1133         ifa->valid_lft = cfg->valid_lft;
1134         ifa->prefered_lft = cfg->preferred_lft;
1135         ifa->cstamp = ifa->tstamp = jiffies;
1136         ifa->tokenized = false;
1137
1138         ifa->rt = f6i;
1139
1140         ifa->idev = idev;
1141         in6_dev_hold(idev);
1142
1143         /* For caller */
1144         refcount_set(&ifa->refcnt, 1);
1145
1146         rcu_read_lock_bh();
1147
1148         err = ipv6_add_addr_hash(idev->dev, ifa);
1149         if (err < 0) {
1150                 rcu_read_unlock_bh();
1151                 goto out;
1152         }
1153
1154         write_lock(&idev->lock);
1155
1156         /* Add to inet6_dev unicast addr list. */
1157         ipv6_link_dev_addr(idev, ifa);
1158
1159         if (ifa->flags&IFA_F_TEMPORARY) {
1160                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1161                 in6_ifa_hold(ifa);
1162         }
1163
1164         in6_ifa_hold(ifa);
1165         write_unlock(&idev->lock);
1166
1167         rcu_read_unlock_bh();
1168
1169         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1170 out:
1171         if (unlikely(err < 0)) {
1172                 fib6_info_release(f6i);
1173
1174                 if (ifa) {
1175                         if (ifa->idev)
1176                                 in6_dev_put(ifa->idev);
1177                         kfree(ifa);
1178                 }
1179                 ifa = ERR_PTR(err);
1180         }
1181
1182         return ifa;
1183 }
1184
1185 enum cleanup_prefix_rt_t {
1186         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1187         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1188         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1189 };
1190
1191 /*
1192  * Check, whether the prefix for ifp would still need a prefix route
1193  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1194  * constants.
1195  *
1196  * 1) we don't purge prefix if address was not permanent.
1197  *    prefix is managed by its own lifetime.
1198  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1199  * 3) if there are no addresses, delete prefix.
1200  * 4) if there are still other permanent address(es),
1201  *    corresponding prefix is still permanent.
1202  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1203  *    don't purge the prefix, assume user space is managing it.
1204  * 6) otherwise, update prefix lifetime to the
1205  *    longest valid lifetime among the corresponding
1206  *    addresses on the device.
1207  *    Note: subsequent RA will update lifetime.
1208  **/
1209 static enum cleanup_prefix_rt_t
1210 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1211 {
1212         struct inet6_ifaddr *ifa;
1213         struct inet6_dev *idev = ifp->idev;
1214         unsigned long lifetime;
1215         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1216
1217         *expires = jiffies;
1218
1219         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1220                 if (ifa == ifp)
1221                         continue;
1222                 if (ifa->prefix_len != ifp->prefix_len ||
1223                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1224                                        ifp->prefix_len))
1225                         continue;
1226                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1227                         return CLEANUP_PREFIX_RT_NOP;
1228
1229                 action = CLEANUP_PREFIX_RT_EXPIRE;
1230
1231                 spin_lock(&ifa->lock);
1232
1233                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1234                 /*
1235                  * Note: Because this address is
1236                  * not permanent, lifetime <
1237                  * LONG_MAX / HZ here.
1238                  */
1239                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1240                         *expires = ifa->tstamp + lifetime * HZ;
1241                 spin_unlock(&ifa->lock);
1242         }
1243
1244         return action;
1245 }
1246
1247 static void
1248 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1249                      bool del_rt, bool del_peer)
1250 {
1251         struct fib6_info *f6i;
1252
1253         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1254                                         ifp->prefix_len,
1255                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1256         if (f6i) {
1257                 if (del_rt)
1258                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1259                 else {
1260                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1261                                 fib6_set_expires(f6i, expires);
1262                         fib6_info_release(f6i);
1263                 }
1264         }
1265 }
1266
1267
1268 /* This function wants to get referenced ifp and releases it before return */
1269
1270 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1271 {
1272         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1273         struct net *net = dev_net(ifp->idev->dev);
1274         unsigned long expires;
1275         int state;
1276
1277         ASSERT_RTNL();
1278
1279         spin_lock_bh(&ifp->lock);
1280         state = ifp->state;
1281         ifp->state = INET6_IFADDR_STATE_DEAD;
1282         spin_unlock_bh(&ifp->lock);
1283
1284         if (state == INET6_IFADDR_STATE_DEAD)
1285                 goto out;
1286
1287         spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1288         hlist_del_init_rcu(&ifp->addr_lst);
1289         spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1290
1291         write_lock_bh(&ifp->idev->lock);
1292
1293         if (ifp->flags&IFA_F_TEMPORARY) {
1294                 list_del(&ifp->tmp_list);
1295                 if (ifp->ifpub) {
1296                         in6_ifa_put(ifp->ifpub);
1297                         ifp->ifpub = NULL;
1298                 }
1299                 __in6_ifa_put(ifp);
1300         }
1301
1302         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1303                 action = check_cleanup_prefix_route(ifp, &expires);
1304
1305         list_del_rcu(&ifp->if_list);
1306         __in6_ifa_put(ifp);
1307
1308         write_unlock_bh(&ifp->idev->lock);
1309
1310         addrconf_del_dad_work(ifp);
1311
1312         ipv6_ifa_notify(RTM_DELADDR, ifp);
1313
1314         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1315
1316         if (action != CLEANUP_PREFIX_RT_NOP) {
1317                 cleanup_prefix_route(ifp, expires,
1318                         action == CLEANUP_PREFIX_RT_DEL, false);
1319         }
1320
1321         /* clean up prefsrc entries */
1322         rt6_remove_prefsrc(ifp);
1323 out:
1324         in6_ifa_put(ifp);
1325 }
1326
1327 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1328 {
1329         struct inet6_dev *idev = ifp->idev;
1330         unsigned long tmp_tstamp, age;
1331         unsigned long regen_advance;
1332         unsigned long now = jiffies;
1333         s32 cnf_temp_preferred_lft;
1334         struct inet6_ifaddr *ift;
1335         struct ifa6_config cfg;
1336         long max_desync_factor;
1337         struct in6_addr addr;
1338         int ret = 0;
1339
1340         write_lock_bh(&idev->lock);
1341
1342 retry:
1343         in6_dev_hold(idev);
1344         if (idev->cnf.use_tempaddr <= 0) {
1345                 write_unlock_bh(&idev->lock);
1346                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1347                 in6_dev_put(idev);
1348                 ret = -1;
1349                 goto out;
1350         }
1351         spin_lock_bh(&ifp->lock);
1352         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1353                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1354                 spin_unlock_bh(&ifp->lock);
1355                 write_unlock_bh(&idev->lock);
1356                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1357                         __func__);
1358                 in6_dev_put(idev);
1359                 ret = -1;
1360                 goto out;
1361         }
1362         in6_ifa_hold(ifp);
1363         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1364         ipv6_gen_rnd_iid(&addr);
1365
1366         age = (now - ifp->tstamp) / HZ;
1367
1368         regen_advance = idev->cnf.regen_max_retry *
1369                         idev->cnf.dad_transmits *
1370                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1371
1372         /* recalculate max_desync_factor each time and update
1373          * idev->desync_factor if it's larger
1374          */
1375         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1376         max_desync_factor = min_t(__u32,
1377                                   idev->cnf.max_desync_factor,
1378                                   cnf_temp_preferred_lft - regen_advance);
1379
1380         if (unlikely(idev->desync_factor > max_desync_factor)) {
1381                 if (max_desync_factor > 0) {
1382                         get_random_bytes(&idev->desync_factor,
1383                                          sizeof(idev->desync_factor));
1384                         idev->desync_factor %= max_desync_factor;
1385                 } else {
1386                         idev->desync_factor = 0;
1387                 }
1388         }
1389
1390         memset(&cfg, 0, sizeof(cfg));
1391         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1392                               idev->cnf.temp_valid_lft + age);
1393         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1394         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1395
1396         cfg.plen = ifp->prefix_len;
1397         tmp_tstamp = ifp->tstamp;
1398         spin_unlock_bh(&ifp->lock);
1399
1400         write_unlock_bh(&idev->lock);
1401
1402         /* A temporary address is created only if this calculated Preferred
1403          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1404          * an implementation must not create a temporary address with a zero
1405          * Preferred Lifetime.
1406          * Use age calculation as in addrconf_verify to avoid unnecessary
1407          * temporary addresses being generated.
1408          */
1409         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1410         if (cfg.preferred_lft <= regen_advance + age) {
1411                 in6_ifa_put(ifp);
1412                 in6_dev_put(idev);
1413                 ret = -1;
1414                 goto out;
1415         }
1416
1417         cfg.ifa_flags = IFA_F_TEMPORARY;
1418         /* set in addrconf_prefix_rcv() */
1419         if (ifp->flags & IFA_F_OPTIMISTIC)
1420                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1421
1422         cfg.pfx = &addr;
1423         cfg.scope = ipv6_addr_scope(cfg.pfx);
1424
1425         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1426         if (IS_ERR(ift)) {
1427                 in6_ifa_put(ifp);
1428                 in6_dev_put(idev);
1429                 pr_info("%s: retry temporary address regeneration\n", __func__);
1430                 write_lock_bh(&idev->lock);
1431                 goto retry;
1432         }
1433
1434         spin_lock_bh(&ift->lock);
1435         ift->ifpub = ifp;
1436         ift->cstamp = now;
1437         ift->tstamp = tmp_tstamp;
1438         spin_unlock_bh(&ift->lock);
1439
1440         addrconf_dad_start(ift);
1441         in6_ifa_put(ift);
1442         in6_dev_put(idev);
1443 out:
1444         return ret;
1445 }
1446
1447 /*
1448  *      Choose an appropriate source address (RFC3484)
1449  */
1450 enum {
1451         IPV6_SADDR_RULE_INIT = 0,
1452         IPV6_SADDR_RULE_LOCAL,
1453         IPV6_SADDR_RULE_SCOPE,
1454         IPV6_SADDR_RULE_PREFERRED,
1455 #ifdef CONFIG_IPV6_MIP6
1456         IPV6_SADDR_RULE_HOA,
1457 #endif
1458         IPV6_SADDR_RULE_OIF,
1459         IPV6_SADDR_RULE_LABEL,
1460         IPV6_SADDR_RULE_PRIVACY,
1461         IPV6_SADDR_RULE_ORCHID,
1462         IPV6_SADDR_RULE_PREFIX,
1463 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1464         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1465 #endif
1466         IPV6_SADDR_RULE_MAX
1467 };
1468
1469 struct ipv6_saddr_score {
1470         int                     rule;
1471         int                     addr_type;
1472         struct inet6_ifaddr     *ifa;
1473         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1474         int                     scopedist;
1475         int                     matchlen;
1476 };
1477
1478 struct ipv6_saddr_dst {
1479         const struct in6_addr *addr;
1480         int ifindex;
1481         int scope;
1482         int label;
1483         unsigned int prefs;
1484 };
1485
1486 static inline int ipv6_saddr_preferred(int type)
1487 {
1488         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1489                 return 1;
1490         return 0;
1491 }
1492
1493 static bool ipv6_use_optimistic_addr(struct net *net,
1494                                      struct inet6_dev *idev)
1495 {
1496 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1497         if (!idev)
1498                 return false;
1499         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1500                 return false;
1501         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1502                 return false;
1503
1504         return true;
1505 #else
1506         return false;
1507 #endif
1508 }
1509
1510 static bool ipv6_allow_optimistic_dad(struct net *net,
1511                                       struct inet6_dev *idev)
1512 {
1513 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1514         if (!idev)
1515                 return false;
1516         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1517                 return false;
1518
1519         return true;
1520 #else
1521         return false;
1522 #endif
1523 }
1524
1525 static int ipv6_get_saddr_eval(struct net *net,
1526                                struct ipv6_saddr_score *score,
1527                                struct ipv6_saddr_dst *dst,
1528                                int i)
1529 {
1530         int ret;
1531
1532         if (i <= score->rule) {
1533                 switch (i) {
1534                 case IPV6_SADDR_RULE_SCOPE:
1535                         ret = score->scopedist;
1536                         break;
1537                 case IPV6_SADDR_RULE_PREFIX:
1538                         ret = score->matchlen;
1539                         break;
1540                 default:
1541                         ret = !!test_bit(i, score->scorebits);
1542                 }
1543                 goto out;
1544         }
1545
1546         switch (i) {
1547         case IPV6_SADDR_RULE_INIT:
1548                 /* Rule 0: remember if hiscore is not ready yet */
1549                 ret = !!score->ifa;
1550                 break;
1551         case IPV6_SADDR_RULE_LOCAL:
1552                 /* Rule 1: Prefer same address */
1553                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1554                 break;
1555         case IPV6_SADDR_RULE_SCOPE:
1556                 /* Rule 2: Prefer appropriate scope
1557                  *
1558                  *      ret
1559                  *       ^
1560                  *    -1 |  d 15
1561                  *    ---+--+-+---> scope
1562                  *       |
1563                  *       |             d is scope of the destination.
1564                  *  B-d  |  \
1565                  *       |   \      <- smaller scope is better if
1566                  *  B-15 |    \        if scope is enough for destination.
1567                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1568                  * d-C-1 | /
1569                  *       |/         <- greater is better
1570                  *   -C  /             if scope is not enough for destination.
1571                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1572                  *
1573                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1574                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1575                  * Assume B = 0 and we get C > 29.
1576                  */
1577                 ret = __ipv6_addr_src_scope(score->addr_type);
1578                 if (ret >= dst->scope)
1579                         ret = -ret;
1580                 else
1581                         ret -= 128;     /* 30 is enough */
1582                 score->scopedist = ret;
1583                 break;
1584         case IPV6_SADDR_RULE_PREFERRED:
1585             {
1586                 /* Rule 3: Avoid deprecated and optimistic addresses */
1587                 u8 avoid = IFA_F_DEPRECATED;
1588
1589                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1590                         avoid |= IFA_F_OPTIMISTIC;
1591                 ret = ipv6_saddr_preferred(score->addr_type) ||
1592                       !(score->ifa->flags & avoid);
1593                 break;
1594             }
1595 #ifdef CONFIG_IPV6_MIP6
1596         case IPV6_SADDR_RULE_HOA:
1597             {
1598                 /* Rule 4: Prefer home address */
1599                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1600                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1601                 break;
1602             }
1603 #endif
1604         case IPV6_SADDR_RULE_OIF:
1605                 /* Rule 5: Prefer outgoing interface */
1606                 ret = (!dst->ifindex ||
1607                        dst->ifindex == score->ifa->idev->dev->ifindex);
1608                 break;
1609         case IPV6_SADDR_RULE_LABEL:
1610                 /* Rule 6: Prefer matching label */
1611                 ret = ipv6_addr_label(net,
1612                                       &score->ifa->addr, score->addr_type,
1613                                       score->ifa->idev->dev->ifindex) == dst->label;
1614                 break;
1615         case IPV6_SADDR_RULE_PRIVACY:
1616             {
1617                 /* Rule 7: Prefer public address
1618                  * Note: prefer temporary address if use_tempaddr >= 2
1619                  */
1620                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1621                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1622                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1623                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1624                 break;
1625             }
1626         case IPV6_SADDR_RULE_ORCHID:
1627                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1628                  *          non-ORCHID vs non-ORCHID
1629                  */
1630                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1631                         ipv6_addr_orchid(dst->addr));
1632                 break;
1633         case IPV6_SADDR_RULE_PREFIX:
1634                 /* Rule 8: Use longest matching prefix */
1635                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1636                 if (ret > score->ifa->prefix_len)
1637                         ret = score->ifa->prefix_len;
1638                 score->matchlen = ret;
1639                 break;
1640 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1641         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1642                 /* Optimistic addresses still have lower precedence than other
1643                  * preferred addresses.
1644                  */
1645                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1646                 break;
1647 #endif
1648         default:
1649                 ret = 0;
1650         }
1651
1652         if (ret)
1653                 __set_bit(i, score->scorebits);
1654         score->rule = i;
1655 out:
1656         return ret;
1657 }
1658
1659 static int __ipv6_dev_get_saddr(struct net *net,
1660                                 struct ipv6_saddr_dst *dst,
1661                                 struct inet6_dev *idev,
1662                                 struct ipv6_saddr_score *scores,
1663                                 int hiscore_idx)
1664 {
1665         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1666
1667         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1668                 int i;
1669
1670                 /*
1671                  * - Tentative Address (RFC2462 section 5.4)
1672                  *  - A tentative address is not considered
1673                  *    "assigned to an interface" in the traditional
1674                  *    sense, unless it is also flagged as optimistic.
1675                  * - Candidate Source Address (section 4)
1676                  *  - In any case, anycast addresses, multicast
1677                  *    addresses, and the unspecified address MUST
1678                  *    NOT be included in a candidate set.
1679                  */
1680                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1681                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1682                         continue;
1683
1684                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1685
1686                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1687                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1688                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1689                                             idev->dev->name);
1690                         continue;
1691                 }
1692
1693                 score->rule = -1;
1694                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1695
1696                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1697                         int minihiscore, miniscore;
1698
1699                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1700                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1701
1702                         if (minihiscore > miniscore) {
1703                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1704                                     score->scopedist > 0) {
1705                                         /*
1706                                          * special case:
1707                                          * each remaining entry
1708                                          * has too small (not enough)
1709                                          * scope, because ifa entries
1710                                          * are sorted by their scope
1711                                          * values.
1712                                          */
1713                                         goto out;
1714                                 }
1715                                 break;
1716                         } else if (minihiscore < miniscore) {
1717                                 swap(hiscore, score);
1718                                 hiscore_idx = 1 - hiscore_idx;
1719
1720                                 /* restore our iterator */
1721                                 score->ifa = hiscore->ifa;
1722
1723                                 break;
1724                         }
1725                 }
1726         }
1727 out:
1728         return hiscore_idx;
1729 }
1730
1731 static int ipv6_get_saddr_master(struct net *net,
1732                                  const struct net_device *dst_dev,
1733                                  const struct net_device *master,
1734                                  struct ipv6_saddr_dst *dst,
1735                                  struct ipv6_saddr_score *scores,
1736                                  int hiscore_idx)
1737 {
1738         struct inet6_dev *idev;
1739
1740         idev = __in6_dev_get(dst_dev);
1741         if (idev)
1742                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1743                                                    scores, hiscore_idx);
1744
1745         idev = __in6_dev_get(master);
1746         if (idev)
1747                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1748                                                    scores, hiscore_idx);
1749
1750         return hiscore_idx;
1751 }
1752
1753 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1754                        const struct in6_addr *daddr, unsigned int prefs,
1755                        struct in6_addr *saddr)
1756 {
1757         struct ipv6_saddr_score scores[2], *hiscore;
1758         struct ipv6_saddr_dst dst;
1759         struct inet6_dev *idev;
1760         struct net_device *dev;
1761         int dst_type;
1762         bool use_oif_addr = false;
1763         int hiscore_idx = 0;
1764         int ret = 0;
1765
1766         dst_type = __ipv6_addr_type(daddr);
1767         dst.addr = daddr;
1768         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1769         dst.scope = __ipv6_addr_src_scope(dst_type);
1770         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1771         dst.prefs = prefs;
1772
1773         scores[hiscore_idx].rule = -1;
1774         scores[hiscore_idx].ifa = NULL;
1775
1776         rcu_read_lock();
1777
1778         /* Candidate Source Address (section 4)
1779          *  - multicast and link-local destination address,
1780          *    the set of candidate source address MUST only
1781          *    include addresses assigned to interfaces
1782          *    belonging to the same link as the outgoing
1783          *    interface.
1784          * (- For site-local destination addresses, the
1785          *    set of candidate source addresses MUST only
1786          *    include addresses assigned to interfaces
1787          *    belonging to the same site as the outgoing
1788          *    interface.)
1789          *  - "It is RECOMMENDED that the candidate source addresses
1790          *    be the set of unicast addresses assigned to the
1791          *    interface that will be used to send to the destination
1792          *    (the 'outgoing' interface)." (RFC 6724)
1793          */
1794         if (dst_dev) {
1795                 idev = __in6_dev_get(dst_dev);
1796                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1797                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1798                     (idev && idev->cnf.use_oif_addrs_only)) {
1799                         use_oif_addr = true;
1800                 }
1801         }
1802
1803         if (use_oif_addr) {
1804                 if (idev)
1805                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1806         } else {
1807                 const struct net_device *master;
1808                 int master_idx = 0;
1809
1810                 /* if dst_dev exists and is enslaved to an L3 device, then
1811                  * prefer addresses from dst_dev and then the master over
1812                  * any other enslaved devices in the L3 domain.
1813                  */
1814                 master = l3mdev_master_dev_rcu(dst_dev);
1815                 if (master) {
1816                         master_idx = master->ifindex;
1817
1818                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1819                                                             master, &dst,
1820                                                             scores, hiscore_idx);
1821
1822                         if (scores[hiscore_idx].ifa)
1823                                 goto out;
1824                 }
1825
1826                 for_each_netdev_rcu(net, dev) {
1827                         /* only consider addresses on devices in the
1828                          * same L3 domain
1829                          */
1830                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1831                                 continue;
1832                         idev = __in6_dev_get(dev);
1833                         if (!idev)
1834                                 continue;
1835                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1836                 }
1837         }
1838
1839 out:
1840         hiscore = &scores[hiscore_idx];
1841         if (!hiscore->ifa)
1842                 ret = -EADDRNOTAVAIL;
1843         else
1844                 *saddr = hiscore->ifa->addr;
1845
1846         rcu_read_unlock();
1847         return ret;
1848 }
1849 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1850
1851 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1852                               u32 banned_flags)
1853 {
1854         struct inet6_ifaddr *ifp;
1855         int err = -EADDRNOTAVAIL;
1856
1857         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1858                 if (ifp->scope > IFA_LINK)
1859                         break;
1860                 if (ifp->scope == IFA_LINK &&
1861                     !(ifp->flags & banned_flags)) {
1862                         *addr = ifp->addr;
1863                         err = 0;
1864                         break;
1865                 }
1866         }
1867         return err;
1868 }
1869
1870 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1871                     u32 banned_flags)
1872 {
1873         struct inet6_dev *idev;
1874         int err = -EADDRNOTAVAIL;
1875
1876         rcu_read_lock();
1877         idev = __in6_dev_get(dev);
1878         if (idev) {
1879                 read_lock_bh(&idev->lock);
1880                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1881                 read_unlock_bh(&idev->lock);
1882         }
1883         rcu_read_unlock();
1884         return err;
1885 }
1886
1887 static int ipv6_count_addresses(const struct inet6_dev *idev)
1888 {
1889         const struct inet6_ifaddr *ifp;
1890         int cnt = 0;
1891
1892         rcu_read_lock();
1893         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1894                 cnt++;
1895         rcu_read_unlock();
1896         return cnt;
1897 }
1898
1899 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1900                   const struct net_device *dev, int strict)
1901 {
1902         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1903                                        strict, IFA_F_TENTATIVE);
1904 }
1905 EXPORT_SYMBOL(ipv6_chk_addr);
1906
1907 /* device argument is used to find the L3 domain of interest. If
1908  * skip_dev_check is set, then the ifp device is not checked against
1909  * the passed in dev argument. So the 2 cases for addresses checks are:
1910  *   1. does the address exist in the L3 domain that dev is part of
1911  *      (skip_dev_check = true), or
1912  *
1913  *   2. does the address exist on the specific device
1914  *      (skip_dev_check = false)
1915  */
1916 static struct net_device *
1917 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1918                           const struct net_device *dev, bool skip_dev_check,
1919                           int strict, u32 banned_flags)
1920 {
1921         unsigned int hash = inet6_addr_hash(net, addr);
1922         struct net_device *l3mdev, *ndev;
1923         struct inet6_ifaddr *ifp;
1924         u32 ifp_flags;
1925
1926         rcu_read_lock();
1927
1928         l3mdev = l3mdev_master_dev_rcu(dev);
1929         if (skip_dev_check)
1930                 dev = NULL;
1931
1932         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1933                 ndev = ifp->idev->dev;
1934
1935                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1936                         continue;
1937
1938                 /* Decouple optimistic from tentative for evaluation here.
1939                  * Ban optimistic addresses explicitly, when required.
1940                  */
1941                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1942                             ? (ifp->flags&~IFA_F_TENTATIVE)
1943                             : ifp->flags;
1944                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1945                     !(ifp_flags&banned_flags) &&
1946                     (!dev || ndev == dev ||
1947                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1948                         rcu_read_unlock();
1949                         return ndev;
1950                 }
1951         }
1952
1953         rcu_read_unlock();
1954         return NULL;
1955 }
1956
1957 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1958                             const struct net_device *dev, bool skip_dev_check,
1959                             int strict, u32 banned_flags)
1960 {
1961         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1962                                          strict, banned_flags) ? 1 : 0;
1963 }
1964 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1965
1966
1967 /* Compares an address/prefix_len with addresses on device @dev.
1968  * If one is found it returns true.
1969  */
1970 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1971         const unsigned int prefix_len, struct net_device *dev)
1972 {
1973         const struct inet6_ifaddr *ifa;
1974         const struct inet6_dev *idev;
1975         bool ret = false;
1976
1977         rcu_read_lock();
1978         idev = __in6_dev_get(dev);
1979         if (idev) {
1980                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1981                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1982                         if (ret)
1983                                 break;
1984                 }
1985         }
1986         rcu_read_unlock();
1987
1988         return ret;
1989 }
1990 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1991
1992 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1993 {
1994         const struct inet6_ifaddr *ifa;
1995         const struct inet6_dev *idev;
1996         int     onlink;
1997
1998         onlink = 0;
1999         rcu_read_lock();
2000         idev = __in6_dev_get(dev);
2001         if (idev) {
2002                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2003                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
2004                                                    ifa->prefix_len);
2005                         if (onlink)
2006                                 break;
2007                 }
2008         }
2009         rcu_read_unlock();
2010         return onlink;
2011 }
2012 EXPORT_SYMBOL(ipv6_chk_prefix);
2013
2014 /**
2015  * ipv6_dev_find - find the first device with a given source address.
2016  * @net: the net namespace
2017  * @addr: the source address
2018  * @dev: used to find the L3 domain of interest
2019  *
2020  * The caller should be protected by RCU, or RTNL.
2021  */
2022 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2023                                  struct net_device *dev)
2024 {
2025         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2026                                          IFA_F_TENTATIVE);
2027 }
2028 EXPORT_SYMBOL(ipv6_dev_find);
2029
2030 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2031                                      struct net_device *dev, int strict)
2032 {
2033         unsigned int hash = inet6_addr_hash(net, addr);
2034         struct inet6_ifaddr *ifp, *result = NULL;
2035
2036         rcu_read_lock();
2037         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
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(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         const union fwnet_hwaddr *ha;
2248
2249         if (dev->addr_len != FWNET_ALEN)
2250                 return -1;
2251
2252         ha = (const 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                         .ifa_proto = IFAPROT_KERNEL_RA
2609                 };
2610
2611 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2612                 if ((net->ipv6.devconf_all->optimistic_dad ||
2613                      in6_dev->cnf.optimistic_dad) &&
2614                     !net->ipv6.devconf_all->forwarding && sllao)
2615                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2616 #endif
2617
2618                 /* Do not allow to create too much of autoconfigured
2619                  * addresses; this would be too easy way to crash kernel.
2620                  */
2621                 if (!max_addresses ||
2622                     ipv6_count_addresses(in6_dev) < max_addresses)
2623                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2624
2625                 if (IS_ERR_OR_NULL(ifp))
2626                         return -1;
2627
2628                 create = 1;
2629                 spin_lock_bh(&ifp->lock);
2630                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2631                 ifp->cstamp = jiffies;
2632                 ifp->tokenized = tokenized;
2633                 spin_unlock_bh(&ifp->lock);
2634                 addrconf_dad_start(ifp);
2635         }
2636
2637         if (ifp) {
2638                 u32 flags;
2639                 unsigned long now;
2640                 u32 stored_lft;
2641
2642                 /* update lifetime (RFC2462 5.5.3 e) */
2643                 spin_lock_bh(&ifp->lock);
2644                 now = jiffies;
2645                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2646                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2647                 else
2648                         stored_lft = 0;
2649                 if (!create && stored_lft) {
2650                         const u32 minimum_lft = min_t(u32,
2651                                 stored_lft, MIN_VALID_LIFETIME);
2652                         valid_lft = max(valid_lft, minimum_lft);
2653
2654                         /* RFC4862 Section 5.5.3e:
2655                          * "Note that the preferred lifetime of the
2656                          *  corresponding address is always reset to
2657                          *  the Preferred Lifetime in the received
2658                          *  Prefix Information option, regardless of
2659                          *  whether the valid lifetime is also reset or
2660                          *  ignored."
2661                          *
2662                          * So we should always update prefered_lft here.
2663                          */
2664                         update_lft = 1;
2665                 }
2666
2667                 if (update_lft) {
2668                         ifp->valid_lft = valid_lft;
2669                         ifp->prefered_lft = prefered_lft;
2670                         ifp->tstamp = now;
2671                         flags = ifp->flags;
2672                         ifp->flags &= ~IFA_F_DEPRECATED;
2673                         spin_unlock_bh(&ifp->lock);
2674
2675                         if (!(flags&IFA_F_TENTATIVE))
2676                                 ipv6_ifa_notify(0, ifp);
2677                 } else
2678                         spin_unlock_bh(&ifp->lock);
2679
2680                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2681                                  create, now);
2682
2683                 in6_ifa_put(ifp);
2684                 addrconf_verify(net);
2685         }
2686
2687         return 0;
2688 }
2689 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2690
2691 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2692 {
2693         struct prefix_info *pinfo;
2694         __u32 valid_lft;
2695         __u32 prefered_lft;
2696         int addr_type, err;
2697         u32 addr_flags = 0;
2698         struct inet6_dev *in6_dev;
2699         struct net *net = dev_net(dev);
2700
2701         pinfo = (struct prefix_info *) opt;
2702
2703         if (len < sizeof(struct prefix_info)) {
2704                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2705                 return;
2706         }
2707
2708         /*
2709          *      Validation checks ([ADDRCONF], page 19)
2710          */
2711
2712         addr_type = ipv6_addr_type(&pinfo->prefix);
2713
2714         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2715                 return;
2716
2717         valid_lft = ntohl(pinfo->valid);
2718         prefered_lft = ntohl(pinfo->prefered);
2719
2720         if (prefered_lft > valid_lft) {
2721                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2722                 return;
2723         }
2724
2725         in6_dev = in6_dev_get(dev);
2726
2727         if (!in6_dev) {
2728                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2729                                     dev->name);
2730                 return;
2731         }
2732
2733         /*
2734          *      Two things going on here:
2735          *      1) Add routes for on-link prefixes
2736          *      2) Configure prefixes with the auto flag set
2737          */
2738
2739         if (pinfo->onlink) {
2740                 struct fib6_info *rt;
2741                 unsigned long rt_expires;
2742
2743                 /* Avoid arithmetic overflow. Really, we could
2744                  * save rt_expires in seconds, likely valid_lft,
2745                  * but it would require division in fib gc, that it
2746                  * not good.
2747                  */
2748                 if (HZ > USER_HZ)
2749                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2750                 else
2751                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2752
2753                 if (addrconf_finite_timeout(rt_expires))
2754                         rt_expires *= HZ;
2755
2756                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2757                                                pinfo->prefix_len,
2758                                                dev,
2759                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2760                                                RTF_DEFAULT, true);
2761
2762                 if (rt) {
2763                         /* Autoconf prefix route */
2764                         if (valid_lft == 0) {
2765                                 ip6_del_rt(net, rt, false);
2766                                 rt = NULL;
2767                         } else if (addrconf_finite_timeout(rt_expires)) {
2768                                 /* not infinity */
2769                                 fib6_set_expires(rt, jiffies + rt_expires);
2770                         } else {
2771                                 fib6_clean_expires(rt);
2772                         }
2773                 } else if (valid_lft) {
2774                         clock_t expires = 0;
2775                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2776                         if (addrconf_finite_timeout(rt_expires)) {
2777                                 /* not infinity */
2778                                 flags |= RTF_EXPIRES;
2779                                 expires = jiffies_to_clock_t(rt_expires);
2780                         }
2781                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2782                                               0, dev, expires, flags,
2783                                               GFP_ATOMIC);
2784                 }
2785                 fib6_info_release(rt);
2786         }
2787
2788         /* Try to figure out our local address for this prefix */
2789
2790         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2791                 struct in6_addr addr;
2792                 bool tokenized = false, dev_addr_generated = false;
2793
2794                 if (pinfo->prefix_len == 64) {
2795                         memcpy(&addr, &pinfo->prefix, 8);
2796
2797                         if (!ipv6_addr_any(&in6_dev->token)) {
2798                                 read_lock_bh(&in6_dev->lock);
2799                                 memcpy(addr.s6_addr + 8,
2800                                        in6_dev->token.s6_addr + 8, 8);
2801                                 read_unlock_bh(&in6_dev->lock);
2802                                 tokenized = true;
2803                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2804                                    !ipv6_generate_stable_address(&addr, 0,
2805                                                                  in6_dev)) {
2806                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2807                                 goto ok;
2808                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2809                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2810                                 goto put;
2811                         } else {
2812                                 dev_addr_generated = true;
2813                         }
2814                         goto ok;
2815                 }
2816                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2817                                     pinfo->prefix_len);
2818                 goto put;
2819
2820 ok:
2821                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2822                                                    &addr, addr_type,
2823                                                    addr_flags, sllao,
2824                                                    tokenized, valid_lft,
2825                                                    prefered_lft);
2826                 if (err)
2827                         goto put;
2828
2829                 /* Ignore error case here because previous prefix add addr was
2830                  * successful which will be notified.
2831                  */
2832                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2833                                               addr_type, addr_flags, sllao,
2834                                               tokenized, valid_lft,
2835                                               prefered_lft,
2836                                               dev_addr_generated);
2837         }
2838         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2839 put:
2840         in6_dev_put(in6_dev);
2841 }
2842
2843 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2844                 struct in6_ifreq *ireq)
2845 {
2846         struct ip_tunnel_parm p = { };
2847         int err;
2848
2849         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2850                 return -EADDRNOTAVAIL;
2851
2852         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2853         p.iph.version = 4;
2854         p.iph.ihl = 5;
2855         p.iph.protocol = IPPROTO_IPV6;
2856         p.iph.ttl = 64;
2857
2858         if (!dev->netdev_ops->ndo_tunnel_ctl)
2859                 return -EOPNOTSUPP;
2860         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2861         if (err)
2862                 return err;
2863
2864         dev = __dev_get_by_name(net, p.name);
2865         if (!dev)
2866                 return -ENOBUFS;
2867         return dev_open(dev, NULL);
2868 }
2869
2870 /*
2871  *      Set destination address.
2872  *      Special case for SIT interfaces where we create a new "virtual"
2873  *      device.
2874  */
2875 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2876 {
2877         struct net_device *dev;
2878         struct in6_ifreq ireq;
2879         int err = -ENODEV;
2880
2881         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2882                 return -ENODEV;
2883         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2884                 return -EFAULT;
2885
2886         rtnl_lock();
2887         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2888         if (dev && dev->type == ARPHRD_SIT)
2889                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2890         rtnl_unlock();
2891         return err;
2892 }
2893
2894 static int ipv6_mc_config(struct sock *sk, bool join,
2895                           const struct in6_addr *addr, int ifindex)
2896 {
2897         int ret;
2898
2899         ASSERT_RTNL();
2900
2901         lock_sock(sk);
2902         if (join)
2903                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2904         else
2905                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2906         release_sock(sk);
2907
2908         return ret;
2909 }
2910
2911 /*
2912  *      Manual configuration of address on an interface
2913  */
2914 static int inet6_addr_add(struct net *net, int ifindex,
2915                           struct ifa6_config *cfg,
2916                           struct netlink_ext_ack *extack)
2917 {
2918         struct inet6_ifaddr *ifp;
2919         struct inet6_dev *idev;
2920         struct net_device *dev;
2921         unsigned long timeout;
2922         clock_t expires;
2923         u32 flags;
2924
2925         ASSERT_RTNL();
2926
2927         if (cfg->plen > 128)
2928                 return -EINVAL;
2929
2930         /* check the lifetime */
2931         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2932                 return -EINVAL;
2933
2934         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2935                 return -EINVAL;
2936
2937         dev = __dev_get_by_index(net, ifindex);
2938         if (!dev)
2939                 return -ENODEV;
2940
2941         idev = addrconf_add_dev(dev);
2942         if (IS_ERR(idev))
2943                 return PTR_ERR(idev);
2944
2945         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2946                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2947                                          true, cfg->pfx, ifindex);
2948
2949                 if (ret < 0)
2950                         return ret;
2951         }
2952
2953         cfg->scope = ipv6_addr_scope(cfg->pfx);
2954
2955         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2956         if (addrconf_finite_timeout(timeout)) {
2957                 expires = jiffies_to_clock_t(timeout * HZ);
2958                 cfg->valid_lft = timeout;
2959                 flags = RTF_EXPIRES;
2960         } else {
2961                 expires = 0;
2962                 flags = 0;
2963                 cfg->ifa_flags |= IFA_F_PERMANENT;
2964         }
2965
2966         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2967         if (addrconf_finite_timeout(timeout)) {
2968                 if (timeout == 0)
2969                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2970                 cfg->preferred_lft = timeout;
2971         }
2972
2973         ifp = ipv6_add_addr(idev, cfg, true, extack);
2974         if (!IS_ERR(ifp)) {
2975                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2976                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2977                                               ifp->rt_priority, dev, expires,
2978                                               flags, GFP_KERNEL);
2979                 }
2980
2981                 /* Send a netlink notification if DAD is enabled and
2982                  * optimistic flag is not set
2983                  */
2984                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2985                         ipv6_ifa_notify(0, ifp);
2986                 /*
2987                  * Note that section 3.1 of RFC 4429 indicates
2988                  * that the Optimistic flag should not be set for
2989                  * manually configured addresses
2990                  */
2991                 addrconf_dad_start(ifp);
2992                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2993                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2994                                          cfg->preferred_lft, true, jiffies);
2995                 in6_ifa_put(ifp);
2996                 addrconf_verify_rtnl(net);
2997                 return 0;
2998         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2999                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3000                                cfg->pfx, ifindex);
3001         }
3002
3003         return PTR_ERR(ifp);
3004 }
3005
3006 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3007                           const struct in6_addr *pfx, unsigned int plen)
3008 {
3009         struct inet6_ifaddr *ifp;
3010         struct inet6_dev *idev;
3011         struct net_device *dev;
3012
3013         if (plen > 128)
3014                 return -EINVAL;
3015
3016         dev = __dev_get_by_index(net, ifindex);
3017         if (!dev)
3018                 return -ENODEV;
3019
3020         idev = __in6_dev_get(dev);
3021         if (!idev)
3022                 return -ENXIO;
3023
3024         read_lock_bh(&idev->lock);
3025         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3026                 if (ifp->prefix_len == plen &&
3027                     ipv6_addr_equal(pfx, &ifp->addr)) {
3028                         in6_ifa_hold(ifp);
3029                         read_unlock_bh(&idev->lock);
3030
3031                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3032                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3033                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3034                                                  jiffies);
3035                         ipv6_del_addr(ifp);
3036                         addrconf_verify_rtnl(net);
3037                         if (ipv6_addr_is_multicast(pfx)) {
3038                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3039                                                false, pfx, dev->ifindex);
3040                         }
3041                         return 0;
3042                 }
3043         }
3044         read_unlock_bh(&idev->lock);
3045         return -EADDRNOTAVAIL;
3046 }
3047
3048
3049 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3050 {
3051         struct ifa6_config cfg = {
3052                 .ifa_flags = IFA_F_PERMANENT,
3053                 .preferred_lft = INFINITY_LIFE_TIME,
3054                 .valid_lft = INFINITY_LIFE_TIME,
3055         };
3056         struct in6_ifreq ireq;
3057         int err;
3058
3059         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3060                 return -EPERM;
3061
3062         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3063                 return -EFAULT;
3064
3065         cfg.pfx = &ireq.ifr6_addr;
3066         cfg.plen = ireq.ifr6_prefixlen;
3067
3068         rtnl_lock();
3069         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3070         rtnl_unlock();
3071         return err;
3072 }
3073
3074 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3075 {
3076         struct in6_ifreq ireq;
3077         int err;
3078
3079         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3080                 return -EPERM;
3081
3082         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3083                 return -EFAULT;
3084
3085         rtnl_lock();
3086         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3087                              ireq.ifr6_prefixlen);
3088         rtnl_unlock();
3089         return err;
3090 }
3091
3092 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3093                      int plen, int scope, u8 proto)
3094 {
3095         struct inet6_ifaddr *ifp;
3096         struct ifa6_config cfg = {
3097                 .pfx = addr,
3098                 .plen = plen,
3099                 .ifa_flags = IFA_F_PERMANENT,
3100                 .valid_lft = INFINITY_LIFE_TIME,
3101                 .preferred_lft = INFINITY_LIFE_TIME,
3102                 .scope = scope,
3103                 .ifa_proto = proto
3104         };
3105
3106         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3107         if (!IS_ERR(ifp)) {
3108                 spin_lock_bh(&ifp->lock);
3109                 ifp->flags &= ~IFA_F_TENTATIVE;
3110                 spin_unlock_bh(&ifp->lock);
3111                 rt_genid_bump_ipv6(dev_net(idev->dev));
3112                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3113                 in6_ifa_put(ifp);
3114         }
3115 }
3116
3117 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3118 static void add_v4_addrs(struct inet6_dev *idev)
3119 {
3120         struct in6_addr addr;
3121         struct net_device *dev;
3122         struct net *net = dev_net(idev->dev);
3123         int scope, plen, offset = 0;
3124         u32 pflags = 0;
3125
3126         ASSERT_RTNL();
3127
3128         memset(&addr, 0, sizeof(struct in6_addr));
3129         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3130         if (idev->dev->addr_len == sizeof(struct in6_addr))
3131                 offset = sizeof(struct in6_addr) - 4;
3132         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3133
3134         if (idev->dev->flags&IFF_POINTOPOINT) {
3135                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3136                         return;
3137
3138                 addr.s6_addr32[0] = htonl(0xfe800000);
3139                 scope = IFA_LINK;
3140                 plen = 64;
3141         } else {
3142                 scope = IPV6_ADDR_COMPATv4;
3143                 plen = 96;
3144                 pflags |= RTF_NONEXTHOP;
3145         }
3146
3147         if (addr.s6_addr32[3]) {
3148                 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3149                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3150                                       GFP_KERNEL);
3151                 return;
3152         }
3153
3154         for_each_netdev(net, dev) {
3155                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3156                 if (in_dev && (dev->flags & IFF_UP)) {
3157                         struct in_ifaddr *ifa;
3158                         int flag = scope;
3159
3160                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3161                                 addr.s6_addr32[3] = ifa->ifa_local;
3162
3163                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3164                                         continue;
3165                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3166                                         if (idev->dev->flags&IFF_POINTOPOINT)
3167                                                 continue;
3168                                         flag |= IFA_HOST;
3169                                 }
3170
3171                                 add_addr(idev, &addr, plen, flag,
3172                                          IFAPROT_UNSPEC);
3173                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3174                                                       0, pflags, GFP_KERNEL);
3175                         }
3176                 }
3177         }
3178 }
3179 #endif
3180
3181 static void init_loopback(struct net_device *dev)
3182 {
3183         struct inet6_dev  *idev;
3184
3185         /* ::1 */
3186
3187         ASSERT_RTNL();
3188
3189         idev = ipv6_find_idev(dev);
3190         if (IS_ERR(idev)) {
3191                 pr_debug("%s: add_dev failed\n", __func__);
3192                 return;
3193         }
3194
3195         add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3196 }
3197
3198 void addrconf_add_linklocal(struct inet6_dev *idev,
3199                             const struct in6_addr *addr, u32 flags)
3200 {
3201         struct ifa6_config cfg = {
3202                 .pfx = addr,
3203                 .plen = 64,
3204                 .ifa_flags = flags | IFA_F_PERMANENT,
3205                 .valid_lft = INFINITY_LIFE_TIME,
3206                 .preferred_lft = INFINITY_LIFE_TIME,
3207                 .scope = IFA_LINK,
3208                 .ifa_proto = IFAPROT_KERNEL_LL
3209         };
3210         struct inet6_ifaddr *ifp;
3211
3212 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3213         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3214              idev->cnf.optimistic_dad) &&
3215             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3216                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3217 #endif
3218
3219         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3220         if (!IS_ERR(ifp)) {
3221                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3222                                       0, 0, GFP_ATOMIC);
3223                 addrconf_dad_start(ifp);
3224                 in6_ifa_put(ifp);
3225         }
3226 }
3227 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3228
3229 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3230 {
3231         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3232                 return true;
3233
3234         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3235             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3236                 return true;
3237
3238         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3239             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3240                 return true;
3241
3242         return false;
3243 }
3244
3245 static int ipv6_generate_stable_address(struct in6_addr *address,
3246                                         u8 dad_count,
3247                                         const struct inet6_dev *idev)
3248 {
3249         static DEFINE_SPINLOCK(lock);
3250         static __u32 digest[SHA1_DIGEST_WORDS];
3251         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3252
3253         static union {
3254                 char __data[SHA1_BLOCK_SIZE];
3255                 struct {
3256                         struct in6_addr secret;
3257                         __be32 prefix[2];
3258                         unsigned char hwaddr[MAX_ADDR_LEN];
3259                         u8 dad_count;
3260                 } __packed;
3261         } data;
3262
3263         struct in6_addr secret;
3264         struct in6_addr temp;
3265         struct net *net = dev_net(idev->dev);
3266
3267         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3268
3269         if (idev->cnf.stable_secret.initialized)
3270                 secret = idev->cnf.stable_secret.secret;
3271         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3272                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3273         else
3274                 return -1;
3275
3276 retry:
3277         spin_lock_bh(&lock);
3278
3279         sha1_init(digest);
3280         memset(&data, 0, sizeof(data));
3281         memset(workspace, 0, sizeof(workspace));
3282         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3283         data.prefix[0] = address->s6_addr32[0];
3284         data.prefix[1] = address->s6_addr32[1];
3285         data.secret = secret;
3286         data.dad_count = dad_count;
3287
3288         sha1_transform(digest, data.__data, workspace);
3289
3290         temp = *address;
3291         temp.s6_addr32[2] = (__force __be32)digest[0];
3292         temp.s6_addr32[3] = (__force __be32)digest[1];
3293
3294         spin_unlock_bh(&lock);
3295
3296         if (ipv6_reserved_interfaceid(temp)) {
3297                 dad_count++;
3298                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3299                         return -1;
3300                 goto retry;
3301         }
3302
3303         *address = temp;
3304         return 0;
3305 }
3306
3307 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3308 {
3309         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3310
3311         if (s->initialized)
3312                 return;
3313         s = &idev->cnf.stable_secret;
3314         get_random_bytes(&s->secret, sizeof(s->secret));
3315         s->initialized = true;
3316 }
3317
3318 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3319 {
3320         struct in6_addr addr;
3321
3322         /* no link local addresses on L3 master devices */
3323         if (netif_is_l3_master(idev->dev))
3324                 return;
3325
3326         /* no link local addresses on devices flagged as slaves */
3327         if (idev->dev->flags & IFF_SLAVE)
3328                 return;
3329
3330         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3331
3332         switch (idev->cnf.addr_gen_mode) {
3333         case IN6_ADDR_GEN_MODE_RANDOM:
3334                 ipv6_gen_mode_random_init(idev);
3335                 fallthrough;
3336         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3337                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3338                         addrconf_add_linklocal(idev, &addr,
3339                                                IFA_F_STABLE_PRIVACY);
3340                 else if (prefix_route)
3341                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3342                                               0, 0, GFP_KERNEL);
3343                 break;
3344         case IN6_ADDR_GEN_MODE_EUI64:
3345                 /* addrconf_add_linklocal also adds a prefix_route and we
3346                  * only need to care about prefix routes if ipv6_generate_eui64
3347                  * couldn't generate one.
3348                  */
3349                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3350                         addrconf_add_linklocal(idev, &addr, 0);
3351                 else if (prefix_route)
3352                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3353                                               0, 0, GFP_KERNEL);
3354                 break;
3355         case IN6_ADDR_GEN_MODE_NONE:
3356         default:
3357                 /* will not add any link local address */
3358                 break;
3359         }
3360 }
3361
3362 static void addrconf_dev_config(struct net_device *dev)
3363 {
3364         struct inet6_dev *idev;
3365
3366         ASSERT_RTNL();
3367
3368         if ((dev->type != ARPHRD_ETHER) &&
3369             (dev->type != ARPHRD_FDDI) &&
3370             (dev->type != ARPHRD_ARCNET) &&
3371             (dev->type != ARPHRD_INFINIBAND) &&
3372             (dev->type != ARPHRD_IEEE1394) &&
3373             (dev->type != ARPHRD_TUNNEL6) &&
3374             (dev->type != ARPHRD_6LOWPAN) &&
3375             (dev->type != ARPHRD_TUNNEL) &&
3376             (dev->type != ARPHRD_NONE) &&
3377             (dev->type != ARPHRD_RAWIP)) {
3378                 /* Alas, we support only Ethernet autoconfiguration. */
3379                 idev = __in6_dev_get(dev);
3380                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3381                     dev->flags & IFF_MULTICAST)
3382                         ipv6_mc_up(idev);
3383                 return;
3384         }
3385
3386         idev = addrconf_add_dev(dev);
3387         if (IS_ERR(idev))
3388                 return;
3389
3390         /* this device type has no EUI support */
3391         if (dev->type == ARPHRD_NONE &&
3392             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3393                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3394
3395         addrconf_addr_gen(idev, false);
3396 }
3397
3398 #if IS_ENABLED(CONFIG_IPV6_SIT)
3399 static void addrconf_sit_config(struct net_device *dev)
3400 {
3401         struct inet6_dev *idev;
3402
3403         ASSERT_RTNL();
3404
3405         /*
3406          * Configure the tunnel with one of our IPv4
3407          * addresses... we should configure all of
3408          * our v4 addrs in the tunnel
3409          */
3410
3411         idev = ipv6_find_idev(dev);
3412         if (IS_ERR(idev)) {
3413                 pr_debug("%s: add_dev failed\n", __func__);
3414                 return;
3415         }
3416
3417         if (dev->priv_flags & IFF_ISATAP) {
3418                 addrconf_addr_gen(idev, false);
3419                 return;
3420         }
3421
3422         add_v4_addrs(idev);
3423
3424         if (dev->flags&IFF_POINTOPOINT)
3425                 addrconf_add_mroute(dev);
3426 }
3427 #endif
3428
3429 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3430 static void addrconf_gre_config(struct net_device *dev)
3431 {
3432         struct inet6_dev *idev;
3433
3434         ASSERT_RTNL();
3435
3436         idev = ipv6_find_idev(dev);
3437         if (IS_ERR(idev)) {
3438                 pr_debug("%s: add_dev failed\n", __func__);
3439                 return;
3440         }
3441
3442         if (dev->type == ARPHRD_ETHER) {
3443                 addrconf_addr_gen(idev, true);
3444                 return;
3445         }
3446
3447         add_v4_addrs(idev);
3448
3449         if (dev->flags & IFF_POINTOPOINT)
3450                 addrconf_add_mroute(dev);
3451 }
3452 #endif
3453
3454 static int fixup_permanent_addr(struct net *net,
3455                                 struct inet6_dev *idev,
3456                                 struct inet6_ifaddr *ifp)
3457 {
3458         /* !fib6_node means the host route was removed from the
3459          * FIB, for example, if 'lo' device is taken down. In that
3460          * case regenerate the host route.
3461          */
3462         if (!ifp->rt || !ifp->rt->fib6_node) {
3463                 struct fib6_info *f6i, *prev;
3464
3465                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3466                                          GFP_ATOMIC);
3467                 if (IS_ERR(f6i))
3468                         return PTR_ERR(f6i);
3469
3470                 /* ifp->rt can be accessed outside of rtnl */
3471                 spin_lock(&ifp->lock);
3472                 prev = ifp->rt;
3473                 ifp->rt = f6i;
3474                 spin_unlock(&ifp->lock);
3475
3476                 fib6_info_release(prev);
3477         }
3478
3479         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3480                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3481                                       ifp->rt_priority, idev->dev, 0, 0,
3482                                       GFP_ATOMIC);
3483         }
3484
3485         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3486                 addrconf_dad_start(ifp);
3487
3488         return 0;
3489 }
3490
3491 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3492 {
3493         struct inet6_ifaddr *ifp, *tmp;
3494         struct inet6_dev *idev;
3495
3496         idev = __in6_dev_get(dev);
3497         if (!idev)
3498                 return;
3499
3500         write_lock_bh(&idev->lock);
3501
3502         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3503                 if ((ifp->flags & IFA_F_PERMANENT) &&
3504                     fixup_permanent_addr(net, idev, ifp) < 0) {
3505                         write_unlock_bh(&idev->lock);
3506                         in6_ifa_hold(ifp);
3507                         ipv6_del_addr(ifp);
3508                         write_lock_bh(&idev->lock);
3509
3510                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3511                                              idev->dev->name, &ifp->addr);
3512                 }
3513         }
3514
3515         write_unlock_bh(&idev->lock);
3516 }
3517
3518 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3519                            void *ptr)
3520 {
3521         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3522         struct netdev_notifier_change_info *change_info;
3523         struct netdev_notifier_changeupper_info *info;
3524         struct inet6_dev *idev = __in6_dev_get(dev);
3525         struct net *net = dev_net(dev);
3526         int run_pending = 0;
3527         int err;
3528
3529         switch (event) {
3530         case NETDEV_REGISTER:
3531                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3532                         idev = ipv6_add_dev(dev);
3533                         if (IS_ERR(idev))
3534                                 return notifier_from_errno(PTR_ERR(idev));
3535                 }
3536                 break;
3537
3538         case NETDEV_CHANGEMTU:
3539                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3540                 if (dev->mtu < IPV6_MIN_MTU) {
3541                         addrconf_ifdown(dev, dev != net->loopback_dev);
3542                         break;
3543                 }
3544
3545                 if (idev) {
3546                         rt6_mtu_change(dev, dev->mtu);
3547                         idev->cnf.mtu6 = dev->mtu;
3548                         break;
3549                 }
3550
3551                 /* allocate new idev */
3552                 idev = ipv6_add_dev(dev);
3553                 if (IS_ERR(idev))
3554                         break;
3555
3556                 /* device is still not ready */
3557                 if (!(idev->if_flags & IF_READY))
3558                         break;
3559
3560                 run_pending = 1;
3561                 fallthrough;
3562         case NETDEV_UP:
3563         case NETDEV_CHANGE:
3564                 if (dev->flags & IFF_SLAVE)
3565                         break;
3566
3567                 if (idev && idev->cnf.disable_ipv6)
3568                         break;
3569
3570                 if (event == NETDEV_UP) {
3571                         /* restore routes for permanent addresses */
3572                         addrconf_permanent_addr(net, dev);
3573
3574                         if (!addrconf_link_ready(dev)) {
3575                                 /* device is not ready yet. */
3576                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3577                                          dev->name);
3578                                 break;
3579                         }
3580
3581                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3582                                 idev = ipv6_add_dev(dev);
3583
3584                         if (!IS_ERR_OR_NULL(idev)) {
3585                                 idev->if_flags |= IF_READY;
3586                                 run_pending = 1;
3587                         }
3588                 } else if (event == NETDEV_CHANGE) {
3589                         if (!addrconf_link_ready(dev)) {
3590                                 /* device is still not ready. */
3591                                 rt6_sync_down_dev(dev, event);
3592                                 break;
3593                         }
3594
3595                         if (!IS_ERR_OR_NULL(idev)) {
3596                                 if (idev->if_flags & IF_READY) {
3597                                         /* device is already configured -
3598                                          * but resend MLD reports, we might
3599                                          * have roamed and need to update
3600                                          * multicast snooping switches
3601                                          */
3602                                         ipv6_mc_up(idev);
3603                                         change_info = ptr;
3604                                         if (change_info->flags_changed & IFF_NOARP)
3605                                                 addrconf_dad_run(idev, true);
3606                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3607                                         break;
3608                                 }
3609                                 idev->if_flags |= IF_READY;
3610                         }
3611
3612                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3613                                 dev->name);
3614
3615                         run_pending = 1;
3616                 }
3617
3618                 switch (dev->type) {
3619 #if IS_ENABLED(CONFIG_IPV6_SIT)
3620                 case ARPHRD_SIT:
3621                         addrconf_sit_config(dev);
3622                         break;
3623 #endif
3624 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3625                 case ARPHRD_IP6GRE:
3626                 case ARPHRD_IPGRE:
3627                         addrconf_gre_config(dev);
3628                         break;
3629 #endif
3630                 case ARPHRD_LOOPBACK:
3631                         init_loopback(dev);
3632                         break;
3633
3634                 default:
3635                         addrconf_dev_config(dev);
3636                         break;
3637                 }
3638
3639                 if (!IS_ERR_OR_NULL(idev)) {
3640                         if (run_pending)
3641                                 addrconf_dad_run(idev, false);
3642
3643                         /* Device has an address by now */
3644                         rt6_sync_up(dev, RTNH_F_DEAD);
3645
3646                         /*
3647                          * If the MTU changed during the interface down,
3648                          * when the interface up, the changed MTU must be
3649                          * reflected in the idev as well as routers.
3650                          */
3651                         if (idev->cnf.mtu6 != dev->mtu &&
3652                             dev->mtu >= IPV6_MIN_MTU) {
3653                                 rt6_mtu_change(dev, dev->mtu);
3654                                 idev->cnf.mtu6 = dev->mtu;
3655                         }
3656                         idev->tstamp = jiffies;
3657                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3658
3659                         /*
3660                          * If the changed mtu during down is lower than
3661                          * IPV6_MIN_MTU stop IPv6 on this interface.
3662                          */
3663                         if (dev->mtu < IPV6_MIN_MTU)
3664                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3665                 }
3666                 break;
3667
3668         case NETDEV_DOWN:
3669         case NETDEV_UNREGISTER:
3670                 /*
3671                  *      Remove all addresses from this interface.
3672                  */
3673                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3674                 break;
3675
3676         case NETDEV_CHANGENAME:
3677                 if (idev) {
3678                         snmp6_unregister_dev(idev);
3679                         addrconf_sysctl_unregister(idev);
3680                         err = addrconf_sysctl_register(idev);
3681                         if (err)
3682                                 return notifier_from_errno(err);
3683                         err = snmp6_register_dev(idev);
3684                         if (err) {
3685                                 addrconf_sysctl_unregister(idev);
3686                                 return notifier_from_errno(err);
3687                         }
3688                 }
3689                 break;
3690
3691         case NETDEV_PRE_TYPE_CHANGE:
3692         case NETDEV_POST_TYPE_CHANGE:
3693                 if (idev)
3694                         addrconf_type_change(dev, event);
3695                 break;
3696
3697         case NETDEV_CHANGEUPPER:
3698                 info = ptr;
3699
3700                 /* flush all routes if dev is linked to or unlinked from
3701                  * an L3 master device (e.g., VRF)
3702                  */
3703                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3704                         addrconf_ifdown(dev, false);
3705         }
3706
3707         return NOTIFY_OK;
3708 }
3709
3710 /*
3711  *      addrconf module should be notified of a device going up
3712  */
3713 static struct notifier_block ipv6_dev_notf = {
3714         .notifier_call = addrconf_notify,
3715         .priority = ADDRCONF_NOTIFY_PRIORITY,
3716 };
3717
3718 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3719 {
3720         struct inet6_dev *idev;
3721         ASSERT_RTNL();
3722
3723         idev = __in6_dev_get(dev);
3724
3725         if (event == NETDEV_POST_TYPE_CHANGE)
3726                 ipv6_mc_remap(idev);
3727         else if (event == NETDEV_PRE_TYPE_CHANGE)
3728                 ipv6_mc_unmap(idev);
3729 }
3730
3731 static bool addr_is_local(const struct in6_addr *addr)
3732 {
3733         return ipv6_addr_type(addr) &
3734                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3735 }
3736
3737 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3738 {
3739         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3740         struct net *net = dev_net(dev);
3741         struct inet6_dev *idev;
3742         struct inet6_ifaddr *ifa;
3743         LIST_HEAD(tmp_addr_list);
3744         bool keep_addr = false;
3745         bool was_ready;
3746         int state, i;
3747
3748         ASSERT_RTNL();
3749
3750         rt6_disable_ip(dev, event);
3751
3752         idev = __in6_dev_get(dev);
3753         if (!idev)
3754                 return -ENODEV;
3755
3756         /*
3757          * Step 1: remove reference to ipv6 device from parent device.
3758          *         Do not dev_put!
3759          */
3760         if (unregister) {
3761                 idev->dead = 1;
3762
3763                 /* protected by rtnl_lock */
3764                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3765
3766                 /* Step 1.5: remove snmp6 entry */
3767                 snmp6_unregister_dev(idev);
3768
3769         }
3770
3771         /* combine the user config with event to determine if permanent
3772          * addresses are to be removed from address hash table
3773          */
3774         if (!unregister && !idev->cnf.disable_ipv6) {
3775                 /* aggregate the system setting and interface setting */
3776                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3777
3778                 if (!_keep_addr)
3779                         _keep_addr = idev->cnf.keep_addr_on_down;
3780
3781                 keep_addr = (_keep_addr > 0);
3782         }
3783
3784         /* Step 2: clear hash table */
3785         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3786                 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3787
3788                 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3789 restart:
3790                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3791                         if (ifa->idev == idev) {
3792                                 addrconf_del_dad_work(ifa);
3793                                 /* combined flag + permanent flag decide if
3794                                  * address is retained on a down event
3795                                  */
3796                                 if (!keep_addr ||
3797                                     !(ifa->flags & IFA_F_PERMANENT) ||
3798                                     addr_is_local(&ifa->addr)) {
3799                                         hlist_del_init_rcu(&ifa->addr_lst);
3800                                         goto restart;
3801                                 }
3802                         }
3803                 }
3804                 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3805         }
3806
3807         write_lock_bh(&idev->lock);
3808
3809         addrconf_del_rs_timer(idev);
3810
3811         /* Step 2: clear flags for stateless addrconf, repeated down
3812          *         detection
3813          */
3814         was_ready = idev->if_flags & IF_READY;
3815         if (!unregister)
3816                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3817
3818         /* Step 3: clear tempaddr list */
3819         while (!list_empty(&idev->tempaddr_list)) {
3820                 ifa = list_first_entry(&idev->tempaddr_list,
3821                                        struct inet6_ifaddr, tmp_list);
3822                 list_del(&ifa->tmp_list);
3823                 write_unlock_bh(&idev->lock);
3824                 spin_lock_bh(&ifa->lock);
3825
3826                 if (ifa->ifpub) {
3827                         in6_ifa_put(ifa->ifpub);
3828                         ifa->ifpub = NULL;
3829                 }
3830                 spin_unlock_bh(&ifa->lock);
3831                 in6_ifa_put(ifa);
3832                 write_lock_bh(&idev->lock);
3833         }
3834
3835         list_for_each_entry(ifa, &idev->addr_list, if_list)
3836                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3837         write_unlock_bh(&idev->lock);
3838
3839         while (!list_empty(&tmp_addr_list)) {
3840                 struct fib6_info *rt = NULL;
3841                 bool keep;
3842
3843                 ifa = list_first_entry(&tmp_addr_list,
3844                                        struct inet6_ifaddr, if_list_aux);
3845                 list_del(&ifa->if_list_aux);
3846
3847                 addrconf_del_dad_work(ifa);
3848
3849                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3850                         !addr_is_local(&ifa->addr);
3851
3852                 spin_lock_bh(&ifa->lock);
3853
3854                 if (keep) {
3855                         /* set state to skip the notifier below */
3856                         state = INET6_IFADDR_STATE_DEAD;
3857                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3858                         if (!(ifa->flags & IFA_F_NODAD))
3859                                 ifa->flags |= IFA_F_TENTATIVE;
3860
3861                         rt = ifa->rt;
3862                         ifa->rt = NULL;
3863                 } else {
3864                         state = ifa->state;
3865                         ifa->state = INET6_IFADDR_STATE_DEAD;
3866                 }
3867
3868                 spin_unlock_bh(&ifa->lock);
3869
3870                 if (rt)
3871                         ip6_del_rt(net, rt, false);
3872
3873                 if (state != INET6_IFADDR_STATE_DEAD) {
3874                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3875                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3876                 } else {
3877                         if (idev->cnf.forwarding)
3878                                 addrconf_leave_anycast(ifa);
3879                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3880                 }
3881
3882                 if (!keep) {
3883                         write_lock_bh(&idev->lock);
3884                         list_del_rcu(&ifa->if_list);
3885                         write_unlock_bh(&idev->lock);
3886                         in6_ifa_put(ifa);
3887                 }
3888         }
3889
3890         /* Step 5: Discard anycast and multicast list */
3891         if (unregister) {
3892                 ipv6_ac_destroy_dev(idev);
3893                 ipv6_mc_destroy_dev(idev);
3894         } else if (was_ready) {
3895                 ipv6_mc_down(idev);
3896         }
3897
3898         idev->tstamp = jiffies;
3899         idev->ra_mtu = 0;
3900
3901         /* Last: Shot the device (if unregistered) */
3902         if (unregister) {
3903                 addrconf_sysctl_unregister(idev);
3904                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3905                 neigh_ifdown(&nd_tbl, dev);
3906                 in6_dev_put(idev);
3907         }
3908         return 0;
3909 }
3910
3911 static void addrconf_rs_timer(struct timer_list *t)
3912 {
3913         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3914         struct net_device *dev = idev->dev;
3915         struct in6_addr lladdr;
3916
3917         write_lock(&idev->lock);
3918         if (idev->dead || !(idev->if_flags & IF_READY))
3919                 goto out;
3920
3921         if (!ipv6_accept_ra(idev))
3922                 goto out;
3923
3924         /* Announcement received after solicitation was sent */
3925         if (idev->if_flags & IF_RA_RCVD)
3926                 goto out;
3927
3928         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3929                 write_unlock(&idev->lock);
3930                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3931                         ndisc_send_rs(dev, &lladdr,
3932                                       &in6addr_linklocal_allrouters);
3933                 else
3934                         goto put;
3935
3936                 write_lock(&idev->lock);
3937                 idev->rs_interval = rfc3315_s14_backoff_update(
3938                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3939                 /* The wait after the last probe can be shorter */
3940                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3941                                              idev->cnf.rtr_solicits) ?
3942                                       idev->cnf.rtr_solicit_delay :
3943                                       idev->rs_interval);
3944         } else {
3945                 /*
3946                  * Note: we do not support deprecated "all on-link"
3947                  * assumption any longer.
3948                  */
3949                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3950         }
3951
3952 out:
3953         write_unlock(&idev->lock);
3954 put:
3955         in6_dev_put(idev);
3956 }
3957
3958 /*
3959  *      Duplicate Address Detection
3960  */
3961 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3962 {
3963         unsigned long rand_num;
3964         struct inet6_dev *idev = ifp->idev;
3965         u64 nonce;
3966
3967         if (ifp->flags & IFA_F_OPTIMISTIC)
3968                 rand_num = 0;
3969         else
3970                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3971
3972         nonce = 0;
3973         if (idev->cnf.enhanced_dad ||
3974             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3975                 do
3976                         get_random_bytes(&nonce, 6);
3977                 while (nonce == 0);
3978         }
3979         ifp->dad_nonce = nonce;
3980         ifp->dad_probes = idev->cnf.dad_transmits;
3981         addrconf_mod_dad_work(ifp, rand_num);
3982 }
3983
3984 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3985 {
3986         struct inet6_dev *idev = ifp->idev;
3987         struct net_device *dev = idev->dev;
3988         bool bump_id, notify = false;
3989         struct net *net;
3990
3991         addrconf_join_solict(dev, &ifp->addr);
3992
3993         read_lock_bh(&idev->lock);
3994         spin_lock(&ifp->lock);
3995         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3996                 goto out;
3997
3998         net = dev_net(dev);
3999         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4000             (net->ipv6.devconf_all->accept_dad < 1 &&
4001              idev->cnf.accept_dad < 1) ||
4002             !(ifp->flags&IFA_F_TENTATIVE) ||
4003             ifp->flags & IFA_F_NODAD) {
4004                 bool send_na = false;
4005
4006                 if (ifp->flags & IFA_F_TENTATIVE &&
4007                     !(ifp->flags & IFA_F_OPTIMISTIC))
4008                         send_na = true;
4009                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4010                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4011                 spin_unlock(&ifp->lock);
4012                 read_unlock_bh(&idev->lock);
4013
4014                 addrconf_dad_completed(ifp, bump_id, send_na);
4015                 return;
4016         }
4017
4018         if (!(idev->if_flags & IF_READY)) {
4019                 spin_unlock(&ifp->lock);
4020                 read_unlock_bh(&idev->lock);
4021                 /*
4022                  * If the device is not ready:
4023                  * - keep it tentative if it is a permanent address.
4024                  * - otherwise, kill it.
4025                  */
4026                 in6_ifa_hold(ifp);
4027                 addrconf_dad_stop(ifp, 0);
4028                 return;
4029         }
4030
4031         /*
4032          * Optimistic nodes can start receiving
4033          * Frames right away
4034          */
4035         if (ifp->flags & IFA_F_OPTIMISTIC) {
4036                 ip6_ins_rt(net, ifp->rt);
4037                 if (ipv6_use_optimistic_addr(net, idev)) {
4038                         /* Because optimistic nodes can use this address,
4039                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4040                          */
4041                         notify = true;
4042                 }
4043         }
4044
4045         addrconf_dad_kick(ifp);
4046 out:
4047         spin_unlock(&ifp->lock);
4048         read_unlock_bh(&idev->lock);
4049         if (notify)
4050                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4051 }
4052
4053 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4054 {
4055         bool begin_dad = false;
4056
4057         spin_lock_bh(&ifp->lock);
4058         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4059                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4060                 begin_dad = true;
4061         }
4062         spin_unlock_bh(&ifp->lock);
4063
4064         if (begin_dad)
4065                 addrconf_mod_dad_work(ifp, 0);
4066 }
4067
4068 static void addrconf_dad_work(struct work_struct *w)
4069 {
4070         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4071                                                 struct inet6_ifaddr,
4072                                                 dad_work);
4073         struct inet6_dev *idev = ifp->idev;
4074         bool bump_id, disable_ipv6 = false;
4075         struct in6_addr mcaddr;
4076
4077         enum {
4078                 DAD_PROCESS,
4079                 DAD_BEGIN,
4080                 DAD_ABORT,
4081         } action = DAD_PROCESS;
4082
4083         rtnl_lock();
4084
4085         spin_lock_bh(&ifp->lock);
4086         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4087                 action = DAD_BEGIN;
4088                 ifp->state = INET6_IFADDR_STATE_DAD;
4089         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4090                 action = DAD_ABORT;
4091                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4092
4093                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4094                      idev->cnf.accept_dad > 1) &&
4095                     !idev->cnf.disable_ipv6 &&
4096                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4097                         struct in6_addr addr;
4098
4099                         addr.s6_addr32[0] = htonl(0xfe800000);
4100                         addr.s6_addr32[1] = 0;
4101
4102                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4103                             ipv6_addr_equal(&ifp->addr, &addr)) {
4104                                 /* DAD failed for link-local based on MAC */
4105                                 idev->cnf.disable_ipv6 = 1;
4106
4107                                 pr_info("%s: IPv6 being disabled!\n",
4108                                         ifp->idev->dev->name);
4109                                 disable_ipv6 = true;
4110                         }
4111                 }
4112         }
4113         spin_unlock_bh(&ifp->lock);
4114
4115         if (action == DAD_BEGIN) {
4116                 addrconf_dad_begin(ifp);
4117                 goto out;
4118         } else if (action == DAD_ABORT) {
4119                 in6_ifa_hold(ifp);
4120                 addrconf_dad_stop(ifp, 1);
4121                 if (disable_ipv6)
4122                         addrconf_ifdown(idev->dev, false);
4123                 goto out;
4124         }
4125
4126         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4127                 goto out;
4128
4129         write_lock_bh(&idev->lock);
4130         if (idev->dead || !(idev->if_flags & IF_READY)) {
4131                 write_unlock_bh(&idev->lock);
4132                 goto out;
4133         }
4134
4135         spin_lock(&ifp->lock);
4136         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4137                 spin_unlock(&ifp->lock);
4138                 write_unlock_bh(&idev->lock);
4139                 goto out;
4140         }
4141
4142         if (ifp->dad_probes == 0) {
4143                 bool send_na = false;
4144
4145                 /*
4146                  * DAD was successful
4147                  */
4148
4149                 if (ifp->flags & IFA_F_TENTATIVE &&
4150                     !(ifp->flags & IFA_F_OPTIMISTIC))
4151                         send_na = true;
4152                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4153                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4154                 spin_unlock(&ifp->lock);
4155                 write_unlock_bh(&idev->lock);
4156
4157                 addrconf_dad_completed(ifp, bump_id, send_na);
4158
4159                 goto out;
4160         }
4161
4162         ifp->dad_probes--;
4163         addrconf_mod_dad_work(ifp,
4164                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4165                                   HZ/100));
4166         spin_unlock(&ifp->lock);
4167         write_unlock_bh(&idev->lock);
4168
4169         /* send a neighbour solicitation for our addr */
4170         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4171         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4172                       ifp->dad_nonce);
4173 out:
4174         in6_ifa_put(ifp);
4175         rtnl_unlock();
4176 }
4177
4178 /* ifp->idev must be at least read locked */
4179 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4180 {
4181         struct inet6_ifaddr *ifpiter;
4182         struct inet6_dev *idev = ifp->idev;
4183
4184         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4185                 if (ifpiter->scope > IFA_LINK)
4186                         break;
4187                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4188                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4189                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4190                     IFA_F_PERMANENT)
4191                         return false;
4192         }
4193         return true;
4194 }
4195
4196 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4197                                    bool send_na)
4198 {
4199         struct net_device *dev = ifp->idev->dev;
4200         struct in6_addr lladdr;
4201         bool send_rs, send_mld;
4202
4203         addrconf_del_dad_work(ifp);
4204
4205         /*
4206          *      Configure the address for reception. Now it is valid.
4207          */
4208
4209         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4210
4211         /* If added prefix is link local and we are prepared to process
4212            router advertisements, start sending router solicitations.
4213          */
4214
4215         read_lock_bh(&ifp->idev->lock);
4216         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4217         send_rs = send_mld &&
4218                   ipv6_accept_ra(ifp->idev) &&
4219                   ifp->idev->cnf.rtr_solicits != 0 &&
4220                   (dev->flags & IFF_LOOPBACK) == 0 &&
4221                   (dev->type != ARPHRD_TUNNEL);
4222         read_unlock_bh(&ifp->idev->lock);
4223
4224         /* While dad is in progress mld report's source address is in6_addrany.
4225          * Resend with proper ll now.
4226          */
4227         if (send_mld)
4228                 ipv6_mc_dad_complete(ifp->idev);
4229
4230         /* send unsolicited NA if enabled */
4231         if (send_na &&
4232             (ifp->idev->cnf.ndisc_notify ||
4233              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4234                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4235                               /*router=*/ !!ifp->idev->cnf.forwarding,
4236                               /*solicited=*/ false, /*override=*/ true,
4237                               /*inc_opt=*/ true);
4238         }
4239
4240         if (send_rs) {
4241                 /*
4242                  *      If a host as already performed a random delay
4243                  *      [...] as part of DAD [...] there is no need
4244                  *      to delay again before sending the first RS
4245                  */
4246                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4247                         return;
4248                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4249
4250                 write_lock_bh(&ifp->idev->lock);
4251                 spin_lock(&ifp->lock);
4252                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4253                         ifp->idev->cnf.rtr_solicit_interval);
4254                 ifp->idev->rs_probes = 1;
4255                 ifp->idev->if_flags |= IF_RS_SENT;
4256                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4257                 spin_unlock(&ifp->lock);
4258                 write_unlock_bh(&ifp->idev->lock);
4259         }
4260
4261         if (bump_id)
4262                 rt_genid_bump_ipv6(dev_net(dev));
4263
4264         /* Make sure that a new temporary address will be created
4265          * before this temporary address becomes deprecated.
4266          */
4267         if (ifp->flags & IFA_F_TEMPORARY)
4268                 addrconf_verify_rtnl(dev_net(dev));
4269 }
4270
4271 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4272 {
4273         struct inet6_ifaddr *ifp;
4274
4275         read_lock_bh(&idev->lock);
4276         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4277                 spin_lock(&ifp->lock);
4278                 if ((ifp->flags & IFA_F_TENTATIVE &&
4279                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4280                         if (restart)
4281                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4282                         addrconf_dad_kick(ifp);
4283                 }
4284                 spin_unlock(&ifp->lock);
4285         }
4286         read_unlock_bh(&idev->lock);
4287 }
4288
4289 #ifdef CONFIG_PROC_FS
4290 struct if6_iter_state {
4291         struct seq_net_private p;
4292         int bucket;
4293         int offset;
4294 };
4295
4296 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4297 {
4298         struct if6_iter_state *state = seq->private;
4299         struct net *net = seq_file_net(seq);
4300         struct inet6_ifaddr *ifa = NULL;
4301         int p = 0;
4302
4303         /* initial bucket if pos is 0 */
4304         if (pos == 0) {
4305                 state->bucket = 0;
4306                 state->offset = 0;
4307         }
4308
4309         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4310                 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4311                                          addr_lst) {
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                 state->offset++;
4335                 return ifa;
4336         }
4337
4338         state->offset = 0;
4339         while (++state->bucket < IN6_ADDR_HSIZE) {
4340                 hlist_for_each_entry_rcu(ifa,
4341                                      &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4342                         return ifa;
4343                 }
4344         }
4345
4346         return NULL;
4347 }
4348
4349 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4350         __acquires(rcu)
4351 {
4352         rcu_read_lock();
4353         return if6_get_first(seq, *pos);
4354 }
4355
4356 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4357 {
4358         struct inet6_ifaddr *ifa;
4359
4360         ifa = if6_get_next(seq, v);
4361         ++*pos;
4362         return ifa;
4363 }
4364
4365 static void if6_seq_stop(struct seq_file *seq, void *v)
4366         __releases(rcu)
4367 {
4368         rcu_read_unlock();
4369 }
4370
4371 static int if6_seq_show(struct seq_file *seq, void *v)
4372 {
4373         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4374         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4375                    &ifp->addr,
4376                    ifp->idev->dev->ifindex,
4377                    ifp->prefix_len,
4378                    ifp->scope,
4379                    (u8) ifp->flags,
4380                    ifp->idev->dev->name);
4381         return 0;
4382 }
4383
4384 static const struct seq_operations if6_seq_ops = {
4385         .start  = if6_seq_start,
4386         .next   = if6_seq_next,
4387         .show   = if6_seq_show,
4388         .stop   = if6_seq_stop,
4389 };
4390
4391 static int __net_init if6_proc_net_init(struct net *net)
4392 {
4393         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4394                         sizeof(struct if6_iter_state)))
4395                 return -ENOMEM;
4396         return 0;
4397 }
4398
4399 static void __net_exit if6_proc_net_exit(struct net *net)
4400 {
4401         remove_proc_entry("if_inet6", net->proc_net);
4402 }
4403
4404 static struct pernet_operations if6_proc_net_ops = {
4405         .init = if6_proc_net_init,
4406         .exit = if6_proc_net_exit,
4407 };
4408
4409 int __init if6_proc_init(void)
4410 {
4411         return register_pernet_subsys(&if6_proc_net_ops);
4412 }
4413
4414 void if6_proc_exit(void)
4415 {
4416         unregister_pernet_subsys(&if6_proc_net_ops);
4417 }
4418 #endif  /* CONFIG_PROC_FS */
4419
4420 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4421 /* Check if address is a home address configured on any interface. */
4422 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4423 {
4424         unsigned int hash = inet6_addr_hash(net, addr);
4425         struct inet6_ifaddr *ifp = NULL;
4426         int ret = 0;
4427
4428         rcu_read_lock();
4429         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4430                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4431                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4432                         ret = 1;
4433                         break;
4434                 }
4435         }
4436         rcu_read_unlock();
4437         return ret;
4438 }
4439 #endif
4440
4441 /* RFC6554 has some algorithm to avoid loops in segment routing by
4442  * checking if the segments contains any of a local interface address.
4443  *
4444  * Quote:
4445  *
4446  * To detect loops in the SRH, a router MUST determine if the SRH
4447  * includes multiple addresses assigned to any interface on that router.
4448  * If such addresses appear more than once and are separated by at least
4449  * one address not assigned to that router.
4450  */
4451 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4452                           unsigned char nsegs)
4453 {
4454         const struct in6_addr *addr;
4455         int i, ret = 0, found = 0;
4456         struct inet6_ifaddr *ifp;
4457         bool separated = false;
4458         unsigned int hash;
4459         bool hash_found;
4460
4461         rcu_read_lock();
4462         for (i = 0; i < nsegs; i++) {
4463                 addr = &segs[i];
4464                 hash = inet6_addr_hash(net, addr);
4465
4466                 hash_found = false;
4467                 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4468
4469                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4470                                 hash_found = true;
4471                                 break;
4472                         }
4473                 }
4474
4475                 if (hash_found) {
4476                         if (found > 1 && separated) {
4477                                 ret = 1;
4478                                 break;
4479                         }
4480
4481                         separated = false;
4482                         found++;
4483                 } else {
4484                         separated = true;
4485                 }
4486         }
4487         rcu_read_unlock();
4488
4489         return ret;
4490 }
4491
4492 /*
4493  *      Periodic address status verification
4494  */
4495
4496 static void addrconf_verify_rtnl(struct net *net)
4497 {
4498         unsigned long now, next, next_sec, next_sched;
4499         struct inet6_ifaddr *ifp;
4500         int i;
4501
4502         ASSERT_RTNL();
4503
4504         rcu_read_lock_bh();
4505         now = jiffies;
4506         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4507
4508         cancel_delayed_work(&net->ipv6.addr_chk_work);
4509
4510         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4511 restart:
4512                 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4513                         unsigned long age;
4514
4515                         /* When setting preferred_lft to a value not zero or
4516                          * infinity, while valid_lft is infinity
4517                          * IFA_F_PERMANENT has a non-infinity life time.
4518                          */
4519                         if ((ifp->flags & IFA_F_PERMANENT) &&
4520                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4521                                 continue;
4522
4523                         spin_lock(&ifp->lock);
4524                         /* We try to batch several events at once. */
4525                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4526
4527                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4528                             age >= ifp->valid_lft) {
4529                                 spin_unlock(&ifp->lock);
4530                                 in6_ifa_hold(ifp);
4531                                 rcu_read_unlock_bh();
4532                                 ipv6_del_addr(ifp);
4533                                 rcu_read_lock_bh();
4534                                 goto restart;
4535                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4536                                 spin_unlock(&ifp->lock);
4537                                 continue;
4538                         } else if (age >= ifp->prefered_lft) {
4539                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4540                                 int deprecate = 0;
4541
4542                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4543                                         deprecate = 1;
4544                                         ifp->flags |= IFA_F_DEPRECATED;
4545                                 }
4546
4547                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4548                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4549                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4550
4551                                 spin_unlock(&ifp->lock);
4552
4553                                 if (deprecate) {
4554                                         in6_ifa_hold(ifp);
4555
4556                                         ipv6_ifa_notify(0, ifp);
4557                                         in6_ifa_put(ifp);
4558                                         goto restart;
4559                                 }
4560                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4561                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4562                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4563                                         ifp->idev->cnf.dad_transmits *
4564                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4565
4566                                 if (age >= ifp->prefered_lft - regen_advance) {
4567                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4568                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4569                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4570                                         if (!ifp->regen_count && ifpub) {
4571                                                 ifp->regen_count++;
4572                                                 in6_ifa_hold(ifp);
4573                                                 in6_ifa_hold(ifpub);
4574                                                 spin_unlock(&ifp->lock);
4575
4576                                                 spin_lock(&ifpub->lock);
4577                                                 ifpub->regen_count = 0;
4578                                                 spin_unlock(&ifpub->lock);
4579                                                 rcu_read_unlock_bh();
4580                                                 ipv6_create_tempaddr(ifpub, true);
4581                                                 in6_ifa_put(ifpub);
4582                                                 in6_ifa_put(ifp);
4583                                                 rcu_read_lock_bh();
4584                                                 goto restart;
4585                                         }
4586                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4587                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4588                                 spin_unlock(&ifp->lock);
4589                         } else {
4590                                 /* ifp->prefered_lft <= ifp->valid_lft */
4591                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4592                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4593                                 spin_unlock(&ifp->lock);
4594                         }
4595                 }
4596         }
4597
4598         next_sec = round_jiffies_up(next);
4599         next_sched = next;
4600
4601         /* If rounded timeout is accurate enough, accept it. */
4602         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4603                 next_sched = next_sec;
4604
4605         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4606         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4607                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4608
4609         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4610                  now, next, next_sec, next_sched);
4611         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4612         rcu_read_unlock_bh();
4613 }
4614
4615 static void addrconf_verify_work(struct work_struct *w)
4616 {
4617         struct net *net = container_of(to_delayed_work(w), struct net,
4618                                        ipv6.addr_chk_work);
4619
4620         rtnl_lock();
4621         addrconf_verify_rtnl(net);
4622         rtnl_unlock();
4623 }
4624
4625 static void addrconf_verify(struct net *net)
4626 {
4627         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4628 }
4629
4630 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4631                                      struct in6_addr **peer_pfx)
4632 {
4633         struct in6_addr *pfx = NULL;
4634
4635         *peer_pfx = NULL;
4636
4637         if (addr)
4638                 pfx = nla_data(addr);
4639
4640         if (local) {
4641                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4642                         *peer_pfx = pfx;
4643                 pfx = nla_data(local);
4644         }
4645
4646         return pfx;
4647 }
4648
4649 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4650         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4651         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4652         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4653         [IFA_FLAGS]             = { .len = sizeof(u32) },
4654         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4655         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4656         [IFA_PROTO]             = { .type = NLA_U8 },
4657 };
4658
4659 static int
4660 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4661                   struct netlink_ext_ack *extack)
4662 {
4663         struct net *net = sock_net(skb->sk);
4664         struct ifaddrmsg *ifm;
4665         struct nlattr *tb[IFA_MAX+1];
4666         struct in6_addr *pfx, *peer_pfx;
4667         u32 ifa_flags;
4668         int err;
4669
4670         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4671                                      ifa_ipv6_policy, extack);
4672         if (err < 0)
4673                 return err;
4674
4675         ifm = nlmsg_data(nlh);
4676         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4677         if (!pfx)
4678                 return -EINVAL;
4679
4680         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4681
4682         /* We ignore other flags so far. */
4683         ifa_flags &= IFA_F_MANAGETEMPADDR;
4684
4685         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4686                               ifm->ifa_prefixlen);
4687 }
4688
4689 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4690                                unsigned long expires, u32 flags,
4691                                bool modify_peer)
4692 {
4693         struct fib6_info *f6i;
4694         u32 prio;
4695
4696         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4697                                         ifp->prefix_len,
4698                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4699         if (!f6i)
4700                 return -ENOENT;
4701
4702         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4703         if (f6i->fib6_metric != prio) {
4704                 /* delete old one */
4705                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4706
4707                 /* add new one */
4708                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4709                                       ifp->prefix_len,
4710                                       ifp->rt_priority, ifp->idev->dev,
4711                                       expires, flags, GFP_KERNEL);
4712         } else {
4713                 if (!expires)
4714                         fib6_clean_expires(f6i);
4715                 else
4716                         fib6_set_expires(f6i, expires);
4717
4718                 fib6_info_release(f6i);
4719         }
4720
4721         return 0;
4722 }
4723
4724 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4725                              struct ifa6_config *cfg)
4726 {
4727         u32 flags;
4728         clock_t expires;
4729         unsigned long timeout;
4730         bool was_managetempaddr;
4731         bool had_prefixroute;
4732         bool new_peer = false;
4733
4734         ASSERT_RTNL();
4735
4736         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4737                 return -EINVAL;
4738
4739         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4740             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4741                 return -EINVAL;
4742
4743         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4744                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4745
4746         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4747         if (addrconf_finite_timeout(timeout)) {
4748                 expires = jiffies_to_clock_t(timeout * HZ);
4749                 cfg->valid_lft = timeout;
4750                 flags = RTF_EXPIRES;
4751         } else {
4752                 expires = 0;
4753                 flags = 0;
4754                 cfg->ifa_flags |= IFA_F_PERMANENT;
4755         }
4756
4757         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4758         if (addrconf_finite_timeout(timeout)) {
4759                 if (timeout == 0)
4760                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4761                 cfg->preferred_lft = timeout;
4762         }
4763
4764         if (cfg->peer_pfx &&
4765             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4766                 if (!ipv6_addr_any(&ifp->peer_addr))
4767                         cleanup_prefix_route(ifp, expires, true, true);
4768                 new_peer = true;
4769         }
4770
4771         spin_lock_bh(&ifp->lock);
4772         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4773         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4774                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4775         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4776                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4777                         IFA_F_NOPREFIXROUTE);
4778         ifp->flags |= cfg->ifa_flags;
4779         ifp->tstamp = jiffies;
4780         ifp->valid_lft = cfg->valid_lft;
4781         ifp->prefered_lft = cfg->preferred_lft;
4782         ifp->ifa_proto = cfg->ifa_proto;
4783
4784         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4785                 ifp->rt_priority = cfg->rt_priority;
4786
4787         if (new_peer)
4788                 ifp->peer_addr = *cfg->peer_pfx;
4789
4790         spin_unlock_bh(&ifp->lock);
4791         if (!(ifp->flags&IFA_F_TENTATIVE))
4792                 ipv6_ifa_notify(0, ifp);
4793
4794         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4795                 int rc = -ENOENT;
4796
4797                 if (had_prefixroute)
4798                         rc = modify_prefix_route(ifp, expires, flags, false);
4799
4800                 /* prefix route could have been deleted; if so restore it */
4801                 if (rc == -ENOENT) {
4802                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4803                                               ifp->rt_priority, ifp->idev->dev,
4804                                               expires, flags, GFP_KERNEL);
4805                 }
4806
4807                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4808                         rc = modify_prefix_route(ifp, expires, flags, true);
4809
4810                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4811                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4812                                               ifp->rt_priority, ifp->idev->dev,
4813                                               expires, flags, GFP_KERNEL);
4814                 }
4815         } else if (had_prefixroute) {
4816                 enum cleanup_prefix_rt_t action;
4817                 unsigned long rt_expires;
4818
4819                 write_lock_bh(&ifp->idev->lock);
4820                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4821                 write_unlock_bh(&ifp->idev->lock);
4822
4823                 if (action != CLEANUP_PREFIX_RT_NOP) {
4824                         cleanup_prefix_route(ifp, rt_expires,
4825                                 action == CLEANUP_PREFIX_RT_DEL, false);
4826                 }
4827         }
4828
4829         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4830                 if (was_managetempaddr &&
4831                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4832                         cfg->valid_lft = 0;
4833                         cfg->preferred_lft = 0;
4834                 }
4835                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4836                                  cfg->preferred_lft, !was_managetempaddr,
4837                                  jiffies);
4838         }
4839
4840         addrconf_verify_rtnl(net);
4841
4842         return 0;
4843 }
4844
4845 static int
4846 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4847                   struct netlink_ext_ack *extack)
4848 {
4849         struct net *net = sock_net(skb->sk);
4850         struct ifaddrmsg *ifm;
4851         struct nlattr *tb[IFA_MAX+1];
4852         struct in6_addr *peer_pfx;
4853         struct inet6_ifaddr *ifa;
4854         struct net_device *dev;
4855         struct inet6_dev *idev;
4856         struct ifa6_config cfg;
4857         int err;
4858
4859         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4860                                      ifa_ipv6_policy, extack);
4861         if (err < 0)
4862                 return err;
4863
4864         memset(&cfg, 0, sizeof(cfg));
4865
4866         ifm = nlmsg_data(nlh);
4867         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4868         if (!cfg.pfx)
4869                 return -EINVAL;
4870
4871         cfg.peer_pfx = peer_pfx;
4872         cfg.plen = ifm->ifa_prefixlen;
4873         if (tb[IFA_RT_PRIORITY])
4874                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4875
4876         if (tb[IFA_PROTO])
4877                 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4878
4879         cfg.valid_lft = INFINITY_LIFE_TIME;
4880         cfg.preferred_lft = INFINITY_LIFE_TIME;
4881
4882         if (tb[IFA_CACHEINFO]) {
4883                 struct ifa_cacheinfo *ci;
4884
4885                 ci = nla_data(tb[IFA_CACHEINFO]);
4886                 cfg.valid_lft = ci->ifa_valid;
4887                 cfg.preferred_lft = ci->ifa_prefered;
4888         }
4889
4890         dev =  __dev_get_by_index(net, ifm->ifa_index);
4891         if (!dev)
4892                 return -ENODEV;
4893
4894         if (tb[IFA_FLAGS])
4895                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4896         else
4897                 cfg.ifa_flags = ifm->ifa_flags;
4898
4899         /* We ignore other flags so far. */
4900         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4901                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4902                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4903
4904         idev = ipv6_find_idev(dev);
4905         if (IS_ERR(idev))
4906                 return PTR_ERR(idev);
4907
4908         if (!ipv6_allow_optimistic_dad(net, idev))
4909                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4910
4911         if (cfg.ifa_flags & IFA_F_NODAD &&
4912             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4913                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4914                 return -EINVAL;
4915         }
4916
4917         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4918         if (!ifa) {
4919                 /*
4920                  * It would be best to check for !NLM_F_CREATE here but
4921                  * userspace already relies on not having to provide this.
4922                  */
4923                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4924         }
4925
4926         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4927             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4928                 err = -EEXIST;
4929         else
4930                 err = inet6_addr_modify(net, ifa, &cfg);
4931
4932         in6_ifa_put(ifa);
4933
4934         return err;
4935 }
4936
4937 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4938                           u8 scope, int ifindex)
4939 {
4940         struct ifaddrmsg *ifm;
4941
4942         ifm = nlmsg_data(nlh);
4943         ifm->ifa_family = AF_INET6;
4944         ifm->ifa_prefixlen = prefixlen;
4945         ifm->ifa_flags = flags;
4946         ifm->ifa_scope = scope;
4947         ifm->ifa_index = ifindex;
4948 }
4949
4950 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4951                          unsigned long tstamp, u32 preferred, u32 valid)
4952 {
4953         struct ifa_cacheinfo ci;
4954
4955         ci.cstamp = cstamp_delta(cstamp);
4956         ci.tstamp = cstamp_delta(tstamp);
4957         ci.ifa_prefered = preferred;
4958         ci.ifa_valid = valid;
4959
4960         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4961 }
4962
4963 static inline int rt_scope(int ifa_scope)
4964 {
4965         if (ifa_scope & IFA_HOST)
4966                 return RT_SCOPE_HOST;
4967         else if (ifa_scope & IFA_LINK)
4968                 return RT_SCOPE_LINK;
4969         else if (ifa_scope & IFA_SITE)
4970                 return RT_SCOPE_SITE;
4971         else
4972                 return RT_SCOPE_UNIVERSE;
4973 }
4974
4975 static inline int inet6_ifaddr_msgsize(void)
4976 {
4977         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4978                + nla_total_size(16) /* IFA_LOCAL */
4979                + nla_total_size(16) /* IFA_ADDRESS */
4980                + nla_total_size(sizeof(struct ifa_cacheinfo))
4981                + nla_total_size(4)  /* IFA_FLAGS */
4982                + nla_total_size(1)  /* IFA_PROTO */
4983                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4984 }
4985
4986 enum addr_type_t {
4987         UNICAST_ADDR,
4988         MULTICAST_ADDR,
4989         ANYCAST_ADDR,
4990 };
4991
4992 struct inet6_fill_args {
4993         u32 portid;
4994         u32 seq;
4995         int event;
4996         unsigned int flags;
4997         int netnsid;
4998         int ifindex;
4999         enum addr_type_t type;
5000 };
5001
5002 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5003                              struct inet6_fill_args *args)
5004 {
5005         struct nlmsghdr  *nlh;
5006         u32 preferred, valid;
5007
5008         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5009                         sizeof(struct ifaddrmsg), args->flags);
5010         if (!nlh)
5011                 return -EMSGSIZE;
5012
5013         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5014                       ifa->idev->dev->ifindex);
5015
5016         if (args->netnsid >= 0 &&
5017             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5018                 goto error;
5019
5020         spin_lock_bh(&ifa->lock);
5021         if (!((ifa->flags&IFA_F_PERMANENT) &&
5022               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5023                 preferred = ifa->prefered_lft;
5024                 valid = ifa->valid_lft;
5025                 if (preferred != INFINITY_LIFE_TIME) {
5026                         long tval = (jiffies - ifa->tstamp)/HZ;
5027                         if (preferred > tval)
5028                                 preferred -= tval;
5029                         else
5030                                 preferred = 0;
5031                         if (valid != INFINITY_LIFE_TIME) {
5032                                 if (valid > tval)
5033                                         valid -= tval;
5034                                 else
5035                                         valid = 0;
5036                         }
5037                 }
5038         } else {
5039                 preferred = INFINITY_LIFE_TIME;
5040                 valid = INFINITY_LIFE_TIME;
5041         }
5042         spin_unlock_bh(&ifa->lock);
5043
5044         if (!ipv6_addr_any(&ifa->peer_addr)) {
5045                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5046                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5047                         goto error;
5048         } else
5049                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5050                         goto error;
5051
5052         if (ifa->rt_priority &&
5053             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5054                 goto error;
5055
5056         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5057                 goto error;
5058
5059         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5060                 goto error;
5061
5062         if (ifa->ifa_proto &&
5063             nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5064                 goto error;
5065
5066         nlmsg_end(skb, nlh);
5067         return 0;
5068
5069 error:
5070         nlmsg_cancel(skb, nlh);
5071         return -EMSGSIZE;
5072 }
5073
5074 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5075                                struct inet6_fill_args *args)
5076 {
5077         struct nlmsghdr  *nlh;
5078         u8 scope = RT_SCOPE_UNIVERSE;
5079         int ifindex = ifmca->idev->dev->ifindex;
5080
5081         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5082                 scope = RT_SCOPE_SITE;
5083
5084         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5085                         sizeof(struct ifaddrmsg), args->flags);
5086         if (!nlh)
5087                 return -EMSGSIZE;
5088
5089         if (args->netnsid >= 0 &&
5090             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5091                 nlmsg_cancel(skb, nlh);
5092                 return -EMSGSIZE;
5093         }
5094
5095         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5096         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5097             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5098                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5099                 nlmsg_cancel(skb, nlh);
5100                 return -EMSGSIZE;
5101         }
5102
5103         nlmsg_end(skb, nlh);
5104         return 0;
5105 }
5106
5107 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5108                                struct inet6_fill_args *args)
5109 {
5110         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5111         int ifindex = dev ? dev->ifindex : 1;
5112         struct nlmsghdr  *nlh;
5113         u8 scope = RT_SCOPE_UNIVERSE;
5114
5115         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5116                 scope = RT_SCOPE_SITE;
5117
5118         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5119                         sizeof(struct ifaddrmsg), args->flags);
5120         if (!nlh)
5121                 return -EMSGSIZE;
5122
5123         if (args->netnsid >= 0 &&
5124             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5125                 nlmsg_cancel(skb, nlh);
5126                 return -EMSGSIZE;
5127         }
5128
5129         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5130         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5131             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5132                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5133                 nlmsg_cancel(skb, nlh);
5134                 return -EMSGSIZE;
5135         }
5136
5137         nlmsg_end(skb, nlh);
5138         return 0;
5139 }
5140
5141 /* called with rcu_read_lock() */
5142 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5143                           struct netlink_callback *cb, int s_ip_idx,
5144                           struct inet6_fill_args *fillargs)
5145 {
5146         struct ifmcaddr6 *ifmca;
5147         struct ifacaddr6 *ifaca;
5148         int ip_idx = 0;
5149         int err = 1;
5150
5151         read_lock_bh(&idev->lock);
5152         switch (fillargs->type) {
5153         case UNICAST_ADDR: {
5154                 struct inet6_ifaddr *ifa;
5155                 fillargs->event = RTM_NEWADDR;
5156
5157                 /* unicast address incl. temp addr */
5158                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5159                         if (ip_idx < s_ip_idx)
5160                                 goto next;
5161                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5162                         if (err < 0)
5163                                 break;
5164                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5165 next:
5166                         ip_idx++;
5167                 }
5168                 break;
5169         }
5170         case MULTICAST_ADDR:
5171                 read_unlock_bh(&idev->lock);
5172                 fillargs->event = RTM_GETMULTICAST;
5173
5174                 /* multicast address */
5175                 for (ifmca = rcu_dereference(idev->mc_list);
5176                      ifmca;
5177                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5178                         if (ip_idx < s_ip_idx)
5179                                 continue;
5180                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5181                         if (err < 0)
5182                                 break;
5183                 }
5184                 read_lock_bh(&idev->lock);
5185                 break;
5186         case ANYCAST_ADDR:
5187                 fillargs->event = RTM_GETANYCAST;
5188                 /* anycast address */
5189                 for (ifaca = idev->ac_list; ifaca;
5190                      ifaca = ifaca->aca_next, ip_idx++) {
5191                         if (ip_idx < s_ip_idx)
5192                                 continue;
5193                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5194                         if (err < 0)
5195                                 break;
5196                 }
5197                 break;
5198         default:
5199                 break;
5200         }
5201         read_unlock_bh(&idev->lock);
5202         cb->args[2] = ip_idx;
5203         return err;
5204 }
5205
5206 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5207                                        struct inet6_fill_args *fillargs,
5208                                        struct net **tgt_net, struct sock *sk,
5209                                        struct netlink_callback *cb)
5210 {
5211         struct netlink_ext_ack *extack = cb->extack;
5212         struct nlattr *tb[IFA_MAX+1];
5213         struct ifaddrmsg *ifm;
5214         int err, i;
5215
5216         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5217                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5218                 return -EINVAL;
5219         }
5220
5221         ifm = nlmsg_data(nlh);
5222         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5223                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5224                 return -EINVAL;
5225         }
5226
5227         fillargs->ifindex = ifm->ifa_index;
5228         if (fillargs->ifindex) {
5229                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5230                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5231         }
5232
5233         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5234                                             ifa_ipv6_policy, extack);
5235         if (err < 0)
5236                 return err;
5237
5238         for (i = 0; i <= IFA_MAX; ++i) {
5239                 if (!tb[i])
5240                         continue;
5241
5242                 if (i == IFA_TARGET_NETNSID) {
5243                         struct net *net;
5244
5245                         fillargs->netnsid = nla_get_s32(tb[i]);
5246                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5247                         if (IS_ERR(net)) {
5248                                 fillargs->netnsid = -1;
5249                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5250                                 return PTR_ERR(net);
5251                         }
5252                         *tgt_net = net;
5253                 } else {
5254                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5255                         return -EINVAL;
5256                 }
5257         }
5258
5259         return 0;
5260 }
5261
5262 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5263                            enum addr_type_t type)
5264 {
5265         const struct nlmsghdr *nlh = cb->nlh;
5266         struct inet6_fill_args fillargs = {
5267                 .portid = NETLINK_CB(cb->skb).portid,
5268                 .seq = cb->nlh->nlmsg_seq,
5269                 .flags = NLM_F_MULTI,
5270                 .netnsid = -1,
5271                 .type = type,
5272         };
5273         struct net *tgt_net = sock_net(skb->sk);
5274         int idx, s_idx, s_ip_idx;
5275         int h, s_h;
5276         struct net_device *dev;
5277         struct inet6_dev *idev;
5278         struct hlist_head *head;
5279         int err = 0;
5280
5281         s_h = cb->args[0];
5282         s_idx = idx = cb->args[1];
5283         s_ip_idx = cb->args[2];
5284
5285         if (cb->strict_check) {
5286                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5287                                                   skb->sk, cb);
5288                 if (err < 0)
5289                         goto put_tgt_net;
5290
5291                 err = 0;
5292                 if (fillargs.ifindex) {
5293                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5294                         if (!dev) {
5295                                 err = -ENODEV;
5296                                 goto put_tgt_net;
5297                         }
5298                         idev = __in6_dev_get(dev);
5299                         if (idev) {
5300                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5301                                                      &fillargs);
5302                                 if (err > 0)
5303                                         err = 0;
5304                         }
5305                         goto put_tgt_net;
5306                 }
5307         }
5308
5309         rcu_read_lock();
5310         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5311         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5312                 idx = 0;
5313                 head = &tgt_net->dev_index_head[h];
5314                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5315                         if (idx < s_idx)
5316                                 goto cont;
5317                         if (h > s_h || idx > s_idx)
5318                                 s_ip_idx = 0;
5319                         idev = __in6_dev_get(dev);
5320                         if (!idev)
5321                                 goto cont;
5322
5323                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5324                                            &fillargs) < 0)
5325                                 goto done;
5326 cont:
5327                         idx++;
5328                 }
5329         }
5330 done:
5331         rcu_read_unlock();
5332         cb->args[0] = h;
5333         cb->args[1] = idx;
5334 put_tgt_net:
5335         if (fillargs.netnsid >= 0)
5336                 put_net(tgt_net);
5337
5338         return skb->len ? : err;
5339 }
5340
5341 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5342 {
5343         enum addr_type_t type = UNICAST_ADDR;
5344
5345         return inet6_dump_addr(skb, cb, type);
5346 }
5347
5348 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5349 {
5350         enum addr_type_t type = MULTICAST_ADDR;
5351
5352         return inet6_dump_addr(skb, cb, type);
5353 }
5354
5355
5356 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5357 {
5358         enum addr_type_t type = ANYCAST_ADDR;
5359
5360         return inet6_dump_addr(skb, cb, type);
5361 }
5362
5363 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5364                                        const struct nlmsghdr *nlh,
5365                                        struct nlattr **tb,
5366                                        struct netlink_ext_ack *extack)
5367 {
5368         struct ifaddrmsg *ifm;
5369         int i, err;
5370
5371         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5372                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5373                 return -EINVAL;
5374         }
5375
5376         if (!netlink_strict_get_check(skb))
5377                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5378                                               ifa_ipv6_policy, extack);
5379
5380         ifm = nlmsg_data(nlh);
5381         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5382                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5383                 return -EINVAL;
5384         }
5385
5386         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5387                                             ifa_ipv6_policy, extack);
5388         if (err)
5389                 return err;
5390
5391         for (i = 0; i <= IFA_MAX; i++) {
5392                 if (!tb[i])
5393                         continue;
5394
5395                 switch (i) {
5396                 case IFA_TARGET_NETNSID:
5397                 case IFA_ADDRESS:
5398                 case IFA_LOCAL:
5399                         break;
5400                 default:
5401                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5402                         return -EINVAL;
5403                 }
5404         }
5405
5406         return 0;
5407 }
5408
5409 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5410                              struct netlink_ext_ack *extack)
5411 {
5412         struct net *tgt_net = sock_net(in_skb->sk);
5413         struct inet6_fill_args fillargs = {
5414                 .portid = NETLINK_CB(in_skb).portid,
5415                 .seq = nlh->nlmsg_seq,
5416                 .event = RTM_NEWADDR,
5417                 .flags = 0,
5418                 .netnsid = -1,
5419         };
5420         struct ifaddrmsg *ifm;
5421         struct nlattr *tb[IFA_MAX+1];
5422         struct in6_addr *addr = NULL, *peer;
5423         struct net_device *dev = NULL;
5424         struct inet6_ifaddr *ifa;
5425         struct sk_buff *skb;
5426         int err;
5427
5428         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5429         if (err < 0)
5430                 return err;
5431
5432         if (tb[IFA_TARGET_NETNSID]) {
5433                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5434
5435                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5436                                                   fillargs.netnsid);
5437                 if (IS_ERR(tgt_net))
5438                         return PTR_ERR(tgt_net);
5439         }
5440
5441         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5442         if (!addr)
5443                 return -EINVAL;
5444
5445         ifm = nlmsg_data(nlh);
5446         if (ifm->ifa_index)
5447                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5448
5449         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5450         if (!ifa) {
5451                 err = -EADDRNOTAVAIL;
5452                 goto errout;
5453         }
5454
5455         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5456         if (!skb) {
5457                 err = -ENOBUFS;
5458                 goto errout_ifa;
5459         }
5460
5461         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5462         if (err < 0) {
5463                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5464                 WARN_ON(err == -EMSGSIZE);
5465                 kfree_skb(skb);
5466                 goto errout_ifa;
5467         }
5468         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5469 errout_ifa:
5470         in6_ifa_put(ifa);
5471 errout:
5472         dev_put(dev);
5473         if (fillargs.netnsid >= 0)
5474                 put_net(tgt_net);
5475
5476         return err;
5477 }
5478
5479 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5480 {
5481         struct sk_buff *skb;
5482         struct net *net = dev_net(ifa->idev->dev);
5483         struct inet6_fill_args fillargs = {
5484                 .portid = 0,
5485                 .seq = 0,
5486                 .event = event,
5487                 .flags = 0,
5488                 .netnsid = -1,
5489         };
5490         int err = -ENOBUFS;
5491
5492         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5493         if (!skb)
5494                 goto errout;
5495
5496         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5497         if (err < 0) {
5498                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5499                 WARN_ON(err == -EMSGSIZE);
5500                 kfree_skb(skb);
5501                 goto errout;
5502         }
5503         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5504         return;
5505 errout:
5506         if (err < 0)
5507                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5508 }
5509
5510 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5511                                 __s32 *array, int bytes)
5512 {
5513         BUG_ON(bytes < (DEVCONF_MAX * 4));
5514
5515         memset(array, 0, bytes);
5516         array[DEVCONF_FORWARDING] = cnf->forwarding;
5517         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5518         array[DEVCONF_MTU6] = cnf->mtu6;
5519         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5520         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5521         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5522         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5523         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5524         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5525                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5526         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5527                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5528         array[DEVCONF_RTR_SOLICIT_DELAY] =
5529                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5530         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5531         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5532                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5533         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5534                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5535         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5536         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5537         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5538         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5539         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5540         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5541         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5542         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5543         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5544         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5545 #ifdef CONFIG_IPV6_ROUTER_PREF
5546         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5547         array[DEVCONF_RTR_PROBE_INTERVAL] =
5548                 jiffies_to_msecs(cnf->rtr_probe_interval);
5549 #ifdef CONFIG_IPV6_ROUTE_INFO
5550         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5551         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5552 #endif
5553 #endif
5554         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5555         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5556 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5557         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5558         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5559 #endif
5560 #ifdef CONFIG_IPV6_MROUTE
5561         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5562 #endif
5563         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5564         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5565         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5566         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5567         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5568         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5569         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5570         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5571         /* we omit DEVCONF_STABLE_SECRET for now */
5572         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5573         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5574         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5575         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5576         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5577 #ifdef CONFIG_IPV6_SEG6_HMAC
5578         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5579 #endif
5580         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5581         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5582         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5583         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5584         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5585         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5586         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5587         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5588         array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5589         array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5590 }
5591
5592 static inline size_t inet6_ifla6_size(void)
5593 {
5594         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5595              + nla_total_size(sizeof(struct ifla_cacheinfo))
5596              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5597              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5598              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5599              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5600              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5601              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5602              + 0;
5603 }
5604
5605 static inline size_t inet6_if_nlmsg_size(void)
5606 {
5607         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5608                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5609                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5610                + nla_total_size(4) /* IFLA_MTU */
5611                + nla_total_size(4) /* IFLA_LINK */
5612                + nla_total_size(1) /* IFLA_OPERSTATE */
5613                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5614 }
5615
5616 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5617                                         int bytes)
5618 {
5619         int i;
5620         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5621         BUG_ON(pad < 0);
5622
5623         /* Use put_unaligned() because stats may not be aligned for u64. */
5624         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5625         for (i = 1; i < ICMP6_MIB_MAX; i++)
5626                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5627
5628         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5629 }
5630
5631 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5632                                         int bytes, size_t syncpoff)
5633 {
5634         int i, c;
5635         u64 buff[IPSTATS_MIB_MAX];
5636         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5637
5638         BUG_ON(pad < 0);
5639
5640         memset(buff, 0, sizeof(buff));
5641         buff[0] = IPSTATS_MIB_MAX;
5642
5643         for_each_possible_cpu(c) {
5644                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5645                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5646         }
5647
5648         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5649         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5650 }
5651
5652 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5653                              int bytes)
5654 {
5655         switch (attrtype) {
5656         case IFLA_INET6_STATS:
5657                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5658                                      offsetof(struct ipstats_mib, syncp));
5659                 break;
5660         case IFLA_INET6_ICMP6STATS:
5661                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5662                 break;
5663         }
5664 }
5665
5666 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5667                                   u32 ext_filter_mask)
5668 {
5669         struct nlattr *nla;
5670         struct ifla_cacheinfo ci;
5671
5672         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5673                 goto nla_put_failure;
5674         ci.max_reasm_len = IPV6_MAXPLEN;
5675         ci.tstamp = cstamp_delta(idev->tstamp);
5676         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5677         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5678         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5679                 goto nla_put_failure;
5680         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5681         if (!nla)
5682                 goto nla_put_failure;
5683         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5684
5685         /* XXX - MC not implemented */
5686
5687         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5688                 return 0;
5689
5690         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5691         if (!nla)
5692                 goto nla_put_failure;
5693         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5694
5695         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5696         if (!nla)
5697                 goto nla_put_failure;
5698         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5699
5700         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5701         if (!nla)
5702                 goto nla_put_failure;
5703         read_lock_bh(&idev->lock);
5704         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5705         read_unlock_bh(&idev->lock);
5706
5707         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5708                 goto nla_put_failure;
5709
5710         if (idev->ra_mtu &&
5711             nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5712                 goto nla_put_failure;
5713
5714         return 0;
5715
5716 nla_put_failure:
5717         return -EMSGSIZE;
5718 }
5719
5720 static size_t inet6_get_link_af_size(const struct net_device *dev,
5721                                      u32 ext_filter_mask)
5722 {
5723         if (!__in6_dev_get(dev))
5724                 return 0;
5725
5726         return inet6_ifla6_size();
5727 }
5728
5729 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5730                               u32 ext_filter_mask)
5731 {
5732         struct inet6_dev *idev = __in6_dev_get(dev);
5733
5734         if (!idev)
5735                 return -ENODATA;
5736
5737         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5738                 return -EMSGSIZE;
5739
5740         return 0;
5741 }
5742
5743 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5744                              struct netlink_ext_ack *extack)
5745 {
5746         struct inet6_ifaddr *ifp;
5747         struct net_device *dev = idev->dev;
5748         bool clear_token, update_rs = false;
5749         struct in6_addr ll_addr;
5750
5751         ASSERT_RTNL();
5752
5753         if (!token)
5754                 return -EINVAL;
5755
5756         if (dev->flags & IFF_LOOPBACK) {
5757                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5758                 return -EINVAL;
5759         }
5760
5761         if (dev->flags & IFF_NOARP) {
5762                 NL_SET_ERR_MSG_MOD(extack,
5763                                    "Device does not do neighbour discovery");
5764                 return -EINVAL;
5765         }
5766
5767         if (!ipv6_accept_ra(idev)) {
5768                 NL_SET_ERR_MSG_MOD(extack,
5769                                    "Router advertisement is disabled on device");
5770                 return -EINVAL;
5771         }
5772
5773         if (idev->cnf.rtr_solicits == 0) {
5774                 NL_SET_ERR_MSG(extack,
5775                                "Router solicitation is disabled on device");
5776                 return -EINVAL;
5777         }
5778
5779         write_lock_bh(&idev->lock);
5780
5781         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5782         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5783
5784         write_unlock_bh(&idev->lock);
5785
5786         clear_token = ipv6_addr_any(token);
5787         if (clear_token)
5788                 goto update_lft;
5789
5790         if (!idev->dead && (idev->if_flags & IF_READY) &&
5791             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5792                              IFA_F_OPTIMISTIC)) {
5793                 /* If we're not ready, then normal ifup will take care
5794                  * of this. Otherwise, we need to request our rs here.
5795                  */
5796                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5797                 update_rs = true;
5798         }
5799
5800 update_lft:
5801         write_lock_bh(&idev->lock);
5802
5803         if (update_rs) {
5804                 idev->if_flags |= IF_RS_SENT;
5805                 idev->rs_interval = rfc3315_s14_backoff_init(
5806                         idev->cnf.rtr_solicit_interval);
5807                 idev->rs_probes = 1;
5808                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5809         }
5810
5811         /* Well, that's kinda nasty ... */
5812         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5813                 spin_lock(&ifp->lock);
5814                 if (ifp->tokenized) {
5815                         ifp->valid_lft = 0;
5816                         ifp->prefered_lft = 0;
5817                 }
5818                 spin_unlock(&ifp->lock);
5819         }
5820
5821         write_unlock_bh(&idev->lock);
5822         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5823         addrconf_verify_rtnl(dev_net(dev));
5824         return 0;
5825 }
5826
5827 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5828         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5829         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5830         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5831                                             .reject_message =
5832                                                 "IFLA_INET6_RA_MTU can not be set" },
5833 };
5834
5835 static int check_addr_gen_mode(int mode)
5836 {
5837         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5838             mode != IN6_ADDR_GEN_MODE_NONE &&
5839             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5840             mode != IN6_ADDR_GEN_MODE_RANDOM)
5841                 return -EINVAL;
5842         return 1;
5843 }
5844
5845 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5846                                 int mode)
5847 {
5848         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5849             !idev->cnf.stable_secret.initialized &&
5850             !net->ipv6.devconf_dflt->stable_secret.initialized)
5851                 return -EINVAL;
5852         return 1;
5853 }
5854
5855 static int inet6_validate_link_af(const struct net_device *dev,
5856                                   const struct nlattr *nla,
5857                                   struct netlink_ext_ack *extack)
5858 {
5859         struct nlattr *tb[IFLA_INET6_MAX + 1];
5860         struct inet6_dev *idev = NULL;
5861         int err;
5862
5863         if (dev) {
5864                 idev = __in6_dev_get(dev);
5865                 if (!idev)
5866                         return -EAFNOSUPPORT;
5867         }
5868
5869         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5870                                           inet6_af_policy, extack);
5871         if (err)
5872                 return err;
5873
5874         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5875                 return -EINVAL;
5876
5877         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5878                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5879
5880                 if (check_addr_gen_mode(mode) < 0)
5881                         return -EINVAL;
5882                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5883                         return -EINVAL;
5884         }
5885
5886         return 0;
5887 }
5888
5889 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5890                              struct netlink_ext_ack *extack)
5891 {
5892         struct inet6_dev *idev = __in6_dev_get(dev);
5893         struct nlattr *tb[IFLA_INET6_MAX + 1];
5894         int err;
5895
5896         if (!idev)
5897                 return -EAFNOSUPPORT;
5898
5899         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5900                 return -EINVAL;
5901
5902         if (tb[IFLA_INET6_TOKEN]) {
5903                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5904                                         extack);
5905                 if (err)
5906                         return err;
5907         }
5908
5909         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5910                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5911
5912                 idev->cnf.addr_gen_mode = mode;
5913         }
5914
5915         return 0;
5916 }
5917
5918 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5919                              u32 portid, u32 seq, int event, unsigned int flags)
5920 {
5921         struct net_device *dev = idev->dev;
5922         struct ifinfomsg *hdr;
5923         struct nlmsghdr *nlh;
5924         void *protoinfo;
5925
5926         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5927         if (!nlh)
5928                 return -EMSGSIZE;
5929
5930         hdr = nlmsg_data(nlh);
5931         hdr->ifi_family = AF_INET6;
5932         hdr->__ifi_pad = 0;
5933         hdr->ifi_type = dev->type;
5934         hdr->ifi_index = dev->ifindex;
5935         hdr->ifi_flags = dev_get_flags(dev);
5936         hdr->ifi_change = 0;
5937
5938         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5939             (dev->addr_len &&
5940              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5941             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5942             (dev->ifindex != dev_get_iflink(dev) &&
5943              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5944             nla_put_u8(skb, IFLA_OPERSTATE,
5945                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5946                 goto nla_put_failure;
5947         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5948         if (!protoinfo)
5949                 goto nla_put_failure;
5950
5951         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5952                 goto nla_put_failure;
5953
5954         nla_nest_end(skb, protoinfo);
5955         nlmsg_end(skb, nlh);
5956         return 0;
5957
5958 nla_put_failure:
5959         nlmsg_cancel(skb, nlh);
5960         return -EMSGSIZE;
5961 }
5962
5963 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5964                                    struct netlink_ext_ack *extack)
5965 {
5966         struct ifinfomsg *ifm;
5967
5968         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5969                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5970                 return -EINVAL;
5971         }
5972
5973         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5974                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5975                 return -EINVAL;
5976         }
5977
5978         ifm = nlmsg_data(nlh);
5979         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5980             ifm->ifi_change || ifm->ifi_index) {
5981                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5982                 return -EINVAL;
5983         }
5984
5985         return 0;
5986 }
5987
5988 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5989 {
5990         struct net *net = sock_net(skb->sk);
5991         int h, s_h;
5992         int idx = 0, s_idx;
5993         struct net_device *dev;
5994         struct inet6_dev *idev;
5995         struct hlist_head *head;
5996
5997         /* only requests using strict checking can pass data to
5998          * influence the dump
5999          */
6000         if (cb->strict_check) {
6001                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6002
6003                 if (err < 0)
6004                         return err;
6005         }
6006
6007         s_h = cb->args[0];
6008         s_idx = cb->args[1];
6009
6010         rcu_read_lock();
6011         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6012                 idx = 0;
6013                 head = &net->dev_index_head[h];
6014                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6015                         if (idx < s_idx)
6016                                 goto cont;
6017                         idev = __in6_dev_get(dev);
6018                         if (!idev)
6019                                 goto cont;
6020                         if (inet6_fill_ifinfo(skb, idev,
6021                                               NETLINK_CB(cb->skb).portid,
6022                                               cb->nlh->nlmsg_seq,
6023                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
6024                                 goto out;
6025 cont:
6026                         idx++;
6027                 }
6028         }
6029 out:
6030         rcu_read_unlock();
6031         cb->args[1] = idx;
6032         cb->args[0] = h;
6033
6034         return skb->len;
6035 }
6036
6037 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6038 {
6039         struct sk_buff *skb;
6040         struct net *net = dev_net(idev->dev);
6041         int err = -ENOBUFS;
6042
6043         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6044         if (!skb)
6045                 goto errout;
6046
6047         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6048         if (err < 0) {
6049                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6050                 WARN_ON(err == -EMSGSIZE);
6051                 kfree_skb(skb);
6052                 goto errout;
6053         }
6054         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6055         return;
6056 errout:
6057         if (err < 0)
6058                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6059 }
6060
6061 static inline size_t inet6_prefix_nlmsg_size(void)
6062 {
6063         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6064                + nla_total_size(sizeof(struct in6_addr))
6065                + nla_total_size(sizeof(struct prefix_cacheinfo));
6066 }
6067
6068 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6069                              struct prefix_info *pinfo, u32 portid, u32 seq,
6070                              int event, unsigned int flags)
6071 {
6072         struct prefixmsg *pmsg;
6073         struct nlmsghdr *nlh;
6074         struct prefix_cacheinfo ci;
6075
6076         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6077         if (!nlh)
6078                 return -EMSGSIZE;
6079
6080         pmsg = nlmsg_data(nlh);
6081         pmsg->prefix_family = AF_INET6;
6082         pmsg->prefix_pad1 = 0;
6083         pmsg->prefix_pad2 = 0;
6084         pmsg->prefix_ifindex = idev->dev->ifindex;
6085         pmsg->prefix_len = pinfo->prefix_len;
6086         pmsg->prefix_type = pinfo->type;
6087         pmsg->prefix_pad3 = 0;
6088         pmsg->prefix_flags = 0;
6089         if (pinfo->onlink)
6090                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6091         if (pinfo->autoconf)
6092                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6093
6094         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6095                 goto nla_put_failure;
6096         ci.preferred_time = ntohl(pinfo->prefered);
6097         ci.valid_time = ntohl(pinfo->valid);
6098         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6099                 goto nla_put_failure;
6100         nlmsg_end(skb, nlh);
6101         return 0;
6102
6103 nla_put_failure:
6104         nlmsg_cancel(skb, nlh);
6105         return -EMSGSIZE;
6106 }
6107
6108 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6109                          struct prefix_info *pinfo)
6110 {
6111         struct sk_buff *skb;
6112         struct net *net = dev_net(idev->dev);
6113         int err = -ENOBUFS;
6114
6115         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6116         if (!skb)
6117                 goto errout;
6118
6119         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6120         if (err < 0) {
6121                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6122                 WARN_ON(err == -EMSGSIZE);
6123                 kfree_skb(skb);
6124                 goto errout;
6125         }
6126         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6127         return;
6128 errout:
6129         if (err < 0)
6130                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6131 }
6132
6133 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6134 {
6135         struct net *net = dev_net(ifp->idev->dev);
6136
6137         if (event)
6138                 ASSERT_RTNL();
6139
6140         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6141
6142         switch (event) {
6143         case RTM_NEWADDR:
6144                 /*
6145                  * If the address was optimistic we inserted the route at the
6146                  * start of our DAD process, so we don't need to do it again.
6147                  * If the device was taken down in the middle of the DAD
6148                  * cycle there is a race where we could get here without a
6149                  * host route, so nothing to insert. That will be fixed when
6150                  * the device is brought up.
6151                  */
6152                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6153                         ip6_ins_rt(net, ifp->rt);
6154                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6155                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6156                                 &ifp->addr, ifp->idev->dev->name);
6157                 }
6158
6159                 if (ifp->idev->cnf.forwarding)
6160                         addrconf_join_anycast(ifp);
6161                 if (!ipv6_addr_any(&ifp->peer_addr))
6162                         addrconf_prefix_route(&ifp->peer_addr, 128,
6163                                               ifp->rt_priority, ifp->idev->dev,
6164                                               0, 0, GFP_ATOMIC);
6165                 break;
6166         case RTM_DELADDR:
6167                 if (ifp->idev->cnf.forwarding)
6168                         addrconf_leave_anycast(ifp);
6169                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6170                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6171                         struct fib6_info *rt;
6172
6173                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6174                                                        ifp->idev->dev, 0, 0,
6175                                                        false);
6176                         if (rt)
6177                                 ip6_del_rt(net, rt, false);
6178                 }
6179                 if (ifp->rt) {
6180                         ip6_del_rt(net, ifp->rt, false);
6181                         ifp->rt = NULL;
6182                 }
6183                 rt_genid_bump_ipv6(net);
6184                 break;
6185         }
6186         atomic_inc(&net->ipv6.dev_addr_genid);
6187 }
6188
6189 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6190 {
6191         if (likely(ifp->idev->dead == 0))
6192                 __ipv6_ifa_notify(event, ifp);
6193 }
6194
6195 #ifdef CONFIG_SYSCTL
6196
6197 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6198                 void *buffer, size_t *lenp, loff_t *ppos)
6199 {
6200         int *valp = ctl->data;
6201         int val = *valp;
6202         loff_t pos = *ppos;
6203         struct ctl_table lctl;
6204         int ret;
6205
6206         /*
6207          * ctl->data points to idev->cnf.forwarding, we should
6208          * not modify it until we get the rtnl lock.
6209          */
6210         lctl = *ctl;
6211         lctl.data = &val;
6212
6213         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6214
6215         if (write)
6216                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6217         if (ret)
6218                 *ppos = pos;
6219         return ret;
6220 }
6221
6222 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6223                 void *buffer, size_t *lenp, loff_t *ppos)
6224 {
6225         struct inet6_dev *idev = ctl->extra1;
6226         int min_mtu = IPV6_MIN_MTU;
6227         struct ctl_table lctl;
6228
6229         lctl = *ctl;
6230         lctl.extra1 = &min_mtu;
6231         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6232
6233         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6234 }
6235
6236 static void dev_disable_change(struct inet6_dev *idev)
6237 {
6238         struct netdev_notifier_info info;
6239
6240         if (!idev || !idev->dev)
6241                 return;
6242
6243         netdev_notifier_info_init(&info, idev->dev);
6244         if (idev->cnf.disable_ipv6)
6245                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6246         else
6247                 addrconf_notify(NULL, NETDEV_UP, &info);
6248 }
6249
6250 static void addrconf_disable_change(struct net *net, __s32 newf)
6251 {
6252         struct net_device *dev;
6253         struct inet6_dev *idev;
6254
6255         for_each_netdev(net, dev) {
6256                 idev = __in6_dev_get(dev);
6257                 if (idev) {
6258                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6259                         idev->cnf.disable_ipv6 = newf;
6260                         if (changed)
6261                                 dev_disable_change(idev);
6262                 }
6263         }
6264 }
6265
6266 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6267 {
6268         struct net *net;
6269         int old;
6270
6271         if (!rtnl_trylock())
6272                 return restart_syscall();
6273
6274         net = (struct net *)table->extra2;
6275         old = *p;
6276         *p = newf;
6277
6278         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6279                 rtnl_unlock();
6280                 return 0;
6281         }
6282
6283         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6284                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6285                 addrconf_disable_change(net, newf);
6286         } else if ((!newf) ^ (!old))
6287                 dev_disable_change((struct inet6_dev *)table->extra1);
6288
6289         rtnl_unlock();
6290         return 0;
6291 }
6292
6293 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6294                 void *buffer, size_t *lenp, loff_t *ppos)
6295 {
6296         int *valp = ctl->data;
6297         int val = *valp;
6298         loff_t pos = *ppos;
6299         struct ctl_table lctl;
6300         int ret;
6301
6302         /*
6303          * ctl->data points to idev->cnf.disable_ipv6, we should
6304          * not modify it until we get the rtnl lock.
6305          */
6306         lctl = *ctl;
6307         lctl.data = &val;
6308
6309         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6310
6311         if (write)
6312                 ret = addrconf_disable_ipv6(ctl, valp, val);
6313         if (ret)
6314                 *ppos = pos;
6315         return ret;
6316 }
6317
6318 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6319                 void *buffer, size_t *lenp, loff_t *ppos)
6320 {
6321         int *valp = ctl->data;
6322         int ret;
6323         int old, new;
6324
6325         old = *valp;
6326         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6327         new = *valp;
6328
6329         if (write && old != new) {
6330                 struct net *net = ctl->extra2;
6331
6332                 if (!rtnl_trylock())
6333                         return restart_syscall();
6334
6335                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6336                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6337                                                      NETCONFA_PROXY_NEIGH,
6338                                                      NETCONFA_IFINDEX_DEFAULT,
6339                                                      net->ipv6.devconf_dflt);
6340                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6341                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6342                                                      NETCONFA_PROXY_NEIGH,
6343                                                      NETCONFA_IFINDEX_ALL,
6344                                                      net->ipv6.devconf_all);
6345                 else {
6346                         struct inet6_dev *idev = ctl->extra1;
6347
6348                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6349                                                      NETCONFA_PROXY_NEIGH,
6350                                                      idev->dev->ifindex,
6351                                                      &idev->cnf);
6352                 }
6353                 rtnl_unlock();
6354         }
6355
6356         return ret;
6357 }
6358
6359 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6360                                          void *buffer, size_t *lenp,
6361                                          loff_t *ppos)
6362 {
6363         int ret = 0;
6364         u32 new_val;
6365         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6366         struct net *net = (struct net *)ctl->extra2;
6367         struct ctl_table tmp = {
6368                 .data = &new_val,
6369                 .maxlen = sizeof(new_val),
6370                 .mode = ctl->mode,
6371         };
6372
6373         if (!rtnl_trylock())
6374                 return restart_syscall();
6375
6376         new_val = *((u32 *)ctl->data);
6377
6378         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6379         if (ret != 0)
6380                 goto out;
6381
6382         if (write) {
6383                 if (check_addr_gen_mode(new_val) < 0) {
6384                         ret = -EINVAL;
6385                         goto out;
6386                 }
6387
6388                 if (idev) {
6389                         if (check_stable_privacy(idev, net, new_val) < 0) {
6390                                 ret = -EINVAL;
6391                                 goto out;
6392                         }
6393
6394                         if (idev->cnf.addr_gen_mode != new_val) {
6395                                 idev->cnf.addr_gen_mode = new_val;
6396                                 addrconf_dev_config(idev->dev);
6397                         }
6398                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6399                         struct net_device *dev;
6400
6401                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6402                         for_each_netdev(net, dev) {
6403                                 idev = __in6_dev_get(dev);
6404                                 if (idev &&
6405                                     idev->cnf.addr_gen_mode != new_val) {
6406                                         idev->cnf.addr_gen_mode = new_val;
6407                                         addrconf_dev_config(idev->dev);
6408                                 }
6409                         }
6410                 }
6411
6412                 *((u32 *)ctl->data) = new_val;
6413         }
6414
6415 out:
6416         rtnl_unlock();
6417
6418         return ret;
6419 }
6420
6421 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6422                                          void *buffer, size_t *lenp,
6423                                          loff_t *ppos)
6424 {
6425         int err;
6426         struct in6_addr addr;
6427         char str[IPV6_MAX_STRLEN];
6428         struct ctl_table lctl = *ctl;
6429         struct net *net = ctl->extra2;
6430         struct ipv6_stable_secret *secret = ctl->data;
6431
6432         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6433                 return -EIO;
6434
6435         lctl.maxlen = IPV6_MAX_STRLEN;
6436         lctl.data = str;
6437
6438         if (!rtnl_trylock())
6439                 return restart_syscall();
6440
6441         if (!write && !secret->initialized) {
6442                 err = -EIO;
6443                 goto out;
6444         }
6445
6446         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6447         if (err >= sizeof(str)) {
6448                 err = -EIO;
6449                 goto out;
6450         }
6451
6452         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6453         if (err || !write)
6454                 goto out;
6455
6456         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6457                 err = -EIO;
6458                 goto out;
6459         }
6460
6461         secret->initialized = true;
6462         secret->secret = addr;
6463
6464         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6465                 struct net_device *dev;
6466
6467                 for_each_netdev(net, dev) {
6468                         struct inet6_dev *idev = __in6_dev_get(dev);
6469
6470                         if (idev) {
6471                                 idev->cnf.addr_gen_mode =
6472                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6473                         }
6474                 }
6475         } else {
6476                 struct inet6_dev *idev = ctl->extra1;
6477
6478                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6479         }
6480
6481 out:
6482         rtnl_unlock();
6483
6484         return err;
6485 }
6486
6487 static
6488 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6489                                                 int write, void *buffer,
6490                                                 size_t *lenp,
6491                                                 loff_t *ppos)
6492 {
6493         int *valp = ctl->data;
6494         int val = *valp;
6495         loff_t pos = *ppos;
6496         struct ctl_table lctl;
6497         int ret;
6498
6499         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6500          * we should not modify it until we get the rtnl lock.
6501          */
6502         lctl = *ctl;
6503         lctl.data = &val;
6504
6505         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6506
6507         if (write)
6508                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6509         if (ret)
6510                 *ppos = pos;
6511         return ret;
6512 }
6513
6514 static
6515 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6516 {
6517         if (rt) {
6518                 if (action)
6519                         rt->dst.flags |= DST_NOPOLICY;
6520                 else
6521                         rt->dst.flags &= ~DST_NOPOLICY;
6522         }
6523 }
6524
6525 static
6526 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6527 {
6528         struct inet6_ifaddr *ifa;
6529
6530         read_lock_bh(&idev->lock);
6531         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6532                 spin_lock(&ifa->lock);
6533                 if (ifa->rt) {
6534                         /* host routes only use builtin fib6_nh */
6535                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6536                         int cpu;
6537
6538                         rcu_read_lock();
6539                         ifa->rt->dst_nopolicy = val ? true : false;
6540                         if (nh->rt6i_pcpu) {
6541                                 for_each_possible_cpu(cpu) {
6542                                         struct rt6_info **rtp;
6543
6544                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6545                                         addrconf_set_nopolicy(*rtp, val);
6546                                 }
6547                         }
6548                         rcu_read_unlock();
6549                 }
6550                 spin_unlock(&ifa->lock);
6551         }
6552         read_unlock_bh(&idev->lock);
6553 }
6554
6555 static
6556 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6557 {
6558         struct inet6_dev *idev;
6559         struct net *net;
6560
6561         if (!rtnl_trylock())
6562                 return restart_syscall();
6563
6564         *valp = val;
6565
6566         net = (struct net *)ctl->extra2;
6567         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6568                 rtnl_unlock();
6569                 return 0;
6570         }
6571
6572         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6573                 struct net_device *dev;
6574
6575                 for_each_netdev(net, dev) {
6576                         idev = __in6_dev_get(dev);
6577                         if (idev)
6578                                 addrconf_disable_policy_idev(idev, val);
6579                 }
6580         } else {
6581                 idev = (struct inet6_dev *)ctl->extra1;
6582                 addrconf_disable_policy_idev(idev, val);
6583         }
6584
6585         rtnl_unlock();
6586         return 0;
6587 }
6588
6589 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6590                                    void *buffer, size_t *lenp, loff_t *ppos)
6591 {
6592         int *valp = ctl->data;
6593         int val = *valp;
6594         loff_t pos = *ppos;
6595         struct ctl_table lctl;
6596         int ret;
6597
6598         lctl = *ctl;
6599         lctl.data = &val;
6600         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6601
6602         if (write && (*valp != val))
6603                 ret = addrconf_disable_policy(ctl, valp, val);
6604
6605         if (ret)
6606                 *ppos = pos;
6607
6608         return ret;
6609 }
6610
6611 static int minus_one = -1;
6612 static const int two_five_five = 255;
6613 static u32 ioam6_if_id_max = U16_MAX;
6614
6615 static const struct ctl_table addrconf_sysctl[] = {
6616         {
6617                 .procname       = "forwarding",
6618                 .data           = &ipv6_devconf.forwarding,
6619                 .maxlen         = sizeof(int),
6620                 .mode           = 0644,
6621                 .proc_handler   = addrconf_sysctl_forward,
6622         },
6623         {
6624                 .procname       = "hop_limit",
6625                 .data           = &ipv6_devconf.hop_limit,
6626                 .maxlen         = sizeof(int),
6627                 .mode           = 0644,
6628                 .proc_handler   = proc_dointvec_minmax,
6629                 .extra1         = (void *)SYSCTL_ONE,
6630                 .extra2         = (void *)&two_five_five,
6631         },
6632         {
6633                 .procname       = "mtu",
6634                 .data           = &ipv6_devconf.mtu6,
6635                 .maxlen         = sizeof(int),
6636                 .mode           = 0644,
6637                 .proc_handler   = addrconf_sysctl_mtu,
6638         },
6639         {
6640                 .procname       = "accept_ra",
6641                 .data           = &ipv6_devconf.accept_ra,
6642                 .maxlen         = sizeof(int),
6643                 .mode           = 0644,
6644                 .proc_handler   = proc_dointvec,
6645         },
6646         {
6647                 .procname       = "accept_redirects",
6648                 .data           = &ipv6_devconf.accept_redirects,
6649                 .maxlen         = sizeof(int),
6650                 .mode           = 0644,
6651                 .proc_handler   = proc_dointvec,
6652         },
6653         {
6654                 .procname       = "autoconf",
6655                 .data           = &ipv6_devconf.autoconf,
6656                 .maxlen         = sizeof(int),
6657                 .mode           = 0644,
6658                 .proc_handler   = proc_dointvec,
6659         },
6660         {
6661                 .procname       = "dad_transmits",
6662                 .data           = &ipv6_devconf.dad_transmits,
6663                 .maxlen         = sizeof(int),
6664                 .mode           = 0644,
6665                 .proc_handler   = proc_dointvec,
6666         },
6667         {
6668                 .procname       = "router_solicitations",
6669                 .data           = &ipv6_devconf.rtr_solicits,
6670                 .maxlen         = sizeof(int),
6671                 .mode           = 0644,
6672                 .proc_handler   = proc_dointvec_minmax,
6673                 .extra1         = &minus_one,
6674         },
6675         {
6676                 .procname       = "router_solicitation_interval",
6677                 .data           = &ipv6_devconf.rtr_solicit_interval,
6678                 .maxlen         = sizeof(int),
6679                 .mode           = 0644,
6680                 .proc_handler   = proc_dointvec_jiffies,
6681         },
6682         {
6683                 .procname       = "router_solicitation_max_interval",
6684                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6685                 .maxlen         = sizeof(int),
6686                 .mode           = 0644,
6687                 .proc_handler   = proc_dointvec_jiffies,
6688         },
6689         {
6690                 .procname       = "router_solicitation_delay",
6691                 .data           = &ipv6_devconf.rtr_solicit_delay,
6692                 .maxlen         = sizeof(int),
6693                 .mode           = 0644,
6694                 .proc_handler   = proc_dointvec_jiffies,
6695         },
6696         {
6697                 .procname       = "force_mld_version",
6698                 .data           = &ipv6_devconf.force_mld_version,
6699                 .maxlen         = sizeof(int),
6700                 .mode           = 0644,
6701                 .proc_handler   = proc_dointvec,
6702         },
6703         {
6704                 .procname       = "mldv1_unsolicited_report_interval",
6705                 .data           =
6706                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6707                 .maxlen         = sizeof(int),
6708                 .mode           = 0644,
6709                 .proc_handler   = proc_dointvec_ms_jiffies,
6710         },
6711         {
6712                 .procname       = "mldv2_unsolicited_report_interval",
6713                 .data           =
6714                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6715                 .maxlen         = sizeof(int),
6716                 .mode           = 0644,
6717                 .proc_handler   = proc_dointvec_ms_jiffies,
6718         },
6719         {
6720                 .procname       = "use_tempaddr",
6721                 .data           = &ipv6_devconf.use_tempaddr,
6722                 .maxlen         = sizeof(int),
6723                 .mode           = 0644,
6724                 .proc_handler   = proc_dointvec,
6725         },
6726         {
6727                 .procname       = "temp_valid_lft",
6728                 .data           = &ipv6_devconf.temp_valid_lft,
6729                 .maxlen         = sizeof(int),
6730                 .mode           = 0644,
6731                 .proc_handler   = proc_dointvec,
6732         },
6733         {
6734                 .procname       = "temp_prefered_lft",
6735                 .data           = &ipv6_devconf.temp_prefered_lft,
6736                 .maxlen         = sizeof(int),
6737                 .mode           = 0644,
6738                 .proc_handler   = proc_dointvec,
6739         },
6740         {
6741                 .procname       = "regen_max_retry",
6742                 .data           = &ipv6_devconf.regen_max_retry,
6743                 .maxlen         = sizeof(int),
6744                 .mode           = 0644,
6745                 .proc_handler   = proc_dointvec,
6746         },
6747         {
6748                 .procname       = "max_desync_factor",
6749                 .data           = &ipv6_devconf.max_desync_factor,
6750                 .maxlen         = sizeof(int),
6751                 .mode           = 0644,
6752                 .proc_handler   = proc_dointvec,
6753         },
6754         {
6755                 .procname       = "max_addresses",
6756                 .data           = &ipv6_devconf.max_addresses,
6757                 .maxlen         = sizeof(int),
6758                 .mode           = 0644,
6759                 .proc_handler   = proc_dointvec,
6760         },
6761         {
6762                 .procname       = "accept_ra_defrtr",
6763                 .data           = &ipv6_devconf.accept_ra_defrtr,
6764                 .maxlen         = sizeof(int),
6765                 .mode           = 0644,
6766                 .proc_handler   = proc_dointvec,
6767         },
6768         {
6769                 .procname       = "ra_defrtr_metric",
6770                 .data           = &ipv6_devconf.ra_defrtr_metric,
6771                 .maxlen         = sizeof(u32),
6772                 .mode           = 0644,
6773                 .proc_handler   = proc_douintvec_minmax,
6774                 .extra1         = (void *)SYSCTL_ONE,
6775         },
6776         {
6777                 .procname       = "accept_ra_min_hop_limit",
6778                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6779                 .maxlen         = sizeof(int),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_dointvec,
6782         },
6783         {
6784                 .procname       = "accept_ra_pinfo",
6785                 .data           = &ipv6_devconf.accept_ra_pinfo,
6786                 .maxlen         = sizeof(int),
6787                 .mode           = 0644,
6788                 .proc_handler   = proc_dointvec,
6789         },
6790 #ifdef CONFIG_IPV6_ROUTER_PREF
6791         {
6792                 .procname       = "accept_ra_rtr_pref",
6793                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6794                 .maxlen         = sizeof(int),
6795                 .mode           = 0644,
6796                 .proc_handler   = proc_dointvec,
6797         },
6798         {
6799                 .procname       = "router_probe_interval",
6800                 .data           = &ipv6_devconf.rtr_probe_interval,
6801                 .maxlen         = sizeof(int),
6802                 .mode           = 0644,
6803                 .proc_handler   = proc_dointvec_jiffies,
6804         },
6805 #ifdef CONFIG_IPV6_ROUTE_INFO
6806         {
6807                 .procname       = "accept_ra_rt_info_min_plen",
6808                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6809                 .maxlen         = sizeof(int),
6810                 .mode           = 0644,
6811                 .proc_handler   = proc_dointvec,
6812         },
6813         {
6814                 .procname       = "accept_ra_rt_info_max_plen",
6815                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6816                 .maxlen         = sizeof(int),
6817                 .mode           = 0644,
6818                 .proc_handler   = proc_dointvec,
6819         },
6820 #endif
6821 #endif
6822         {
6823                 .procname       = "proxy_ndp",
6824                 .data           = &ipv6_devconf.proxy_ndp,
6825                 .maxlen         = sizeof(int),
6826                 .mode           = 0644,
6827                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6828         },
6829         {
6830                 .procname       = "accept_source_route",
6831                 .data           = &ipv6_devconf.accept_source_route,
6832                 .maxlen         = sizeof(int),
6833                 .mode           = 0644,
6834                 .proc_handler   = proc_dointvec,
6835         },
6836 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6837         {
6838                 .procname       = "optimistic_dad",
6839                 .data           = &ipv6_devconf.optimistic_dad,
6840                 .maxlen         = sizeof(int),
6841                 .mode           = 0644,
6842                 .proc_handler   = proc_dointvec,
6843         },
6844         {
6845                 .procname       = "use_optimistic",
6846                 .data           = &ipv6_devconf.use_optimistic,
6847                 .maxlen         = sizeof(int),
6848                 .mode           = 0644,
6849                 .proc_handler   = proc_dointvec,
6850         },
6851 #endif
6852 #ifdef CONFIG_IPV6_MROUTE
6853         {
6854                 .procname       = "mc_forwarding",
6855                 .data           = &ipv6_devconf.mc_forwarding,
6856                 .maxlen         = sizeof(int),
6857                 .mode           = 0444,
6858                 .proc_handler   = proc_dointvec,
6859         },
6860 #endif
6861         {
6862                 .procname       = "disable_ipv6",
6863                 .data           = &ipv6_devconf.disable_ipv6,
6864                 .maxlen         = sizeof(int),
6865                 .mode           = 0644,
6866                 .proc_handler   = addrconf_sysctl_disable,
6867         },
6868         {
6869                 .procname       = "accept_dad",
6870                 .data           = &ipv6_devconf.accept_dad,
6871                 .maxlen         = sizeof(int),
6872                 .mode           = 0644,
6873                 .proc_handler   = proc_dointvec,
6874         },
6875         {
6876                 .procname       = "force_tllao",
6877                 .data           = &ipv6_devconf.force_tllao,
6878                 .maxlen         = sizeof(int),
6879                 .mode           = 0644,
6880                 .proc_handler   = proc_dointvec
6881         },
6882         {
6883                 .procname       = "ndisc_notify",
6884                 .data           = &ipv6_devconf.ndisc_notify,
6885                 .maxlen         = sizeof(int),
6886                 .mode           = 0644,
6887                 .proc_handler   = proc_dointvec
6888         },
6889         {
6890                 .procname       = "suppress_frag_ndisc",
6891                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6892                 .maxlen         = sizeof(int),
6893                 .mode           = 0644,
6894                 .proc_handler   = proc_dointvec
6895         },
6896         {
6897                 .procname       = "accept_ra_from_local",
6898                 .data           = &ipv6_devconf.accept_ra_from_local,
6899                 .maxlen         = sizeof(int),
6900                 .mode           = 0644,
6901                 .proc_handler   = proc_dointvec,
6902         },
6903         {
6904                 .procname       = "accept_ra_mtu",
6905                 .data           = &ipv6_devconf.accept_ra_mtu,
6906                 .maxlen         = sizeof(int),
6907                 .mode           = 0644,
6908                 .proc_handler   = proc_dointvec,
6909         },
6910         {
6911                 .procname       = "stable_secret",
6912                 .data           = &ipv6_devconf.stable_secret,
6913                 .maxlen         = IPV6_MAX_STRLEN,
6914                 .mode           = 0600,
6915                 .proc_handler   = addrconf_sysctl_stable_secret,
6916         },
6917         {
6918                 .procname       = "use_oif_addrs_only",
6919                 .data           = &ipv6_devconf.use_oif_addrs_only,
6920                 .maxlen         = sizeof(int),
6921                 .mode           = 0644,
6922                 .proc_handler   = proc_dointvec,
6923         },
6924         {
6925                 .procname       = "ignore_routes_with_linkdown",
6926                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6927                 .maxlen         = sizeof(int),
6928                 .mode           = 0644,
6929                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6930         },
6931         {
6932                 .procname       = "drop_unicast_in_l2_multicast",
6933                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6934                 .maxlen         = sizeof(int),
6935                 .mode           = 0644,
6936                 .proc_handler   = proc_dointvec,
6937         },
6938         {
6939                 .procname       = "drop_unsolicited_na",
6940                 .data           = &ipv6_devconf.drop_unsolicited_na,
6941                 .maxlen         = sizeof(int),
6942                 .mode           = 0644,
6943                 .proc_handler   = proc_dointvec,
6944         },
6945         {
6946                 .procname       = "keep_addr_on_down",
6947                 .data           = &ipv6_devconf.keep_addr_on_down,
6948                 .maxlen         = sizeof(int),
6949                 .mode           = 0644,
6950                 .proc_handler   = proc_dointvec,
6951
6952         },
6953         {
6954                 .procname       = "seg6_enabled",
6955                 .data           = &ipv6_devconf.seg6_enabled,
6956                 .maxlen         = sizeof(int),
6957                 .mode           = 0644,
6958                 .proc_handler   = proc_dointvec,
6959         },
6960 #ifdef CONFIG_IPV6_SEG6_HMAC
6961         {
6962                 .procname       = "seg6_require_hmac",
6963                 .data           = &ipv6_devconf.seg6_require_hmac,
6964                 .maxlen         = sizeof(int),
6965                 .mode           = 0644,
6966                 .proc_handler   = proc_dointvec,
6967         },
6968 #endif
6969         {
6970                 .procname       = "enhanced_dad",
6971                 .data           = &ipv6_devconf.enhanced_dad,
6972                 .maxlen         = sizeof(int),
6973                 .mode           = 0644,
6974                 .proc_handler   = proc_dointvec,
6975         },
6976         {
6977                 .procname       = "addr_gen_mode",
6978                 .data           = &ipv6_devconf.addr_gen_mode,
6979                 .maxlen         = sizeof(int),
6980                 .mode           = 0644,
6981                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6982         },
6983         {
6984                 .procname       = "disable_policy",
6985                 .data           = &ipv6_devconf.disable_policy,
6986                 .maxlen         = sizeof(int),
6987                 .mode           = 0644,
6988                 .proc_handler   = addrconf_sysctl_disable_policy,
6989         },
6990         {
6991                 .procname       = "ndisc_tclass",
6992                 .data           = &ipv6_devconf.ndisc_tclass,
6993                 .maxlen         = sizeof(int),
6994                 .mode           = 0644,
6995                 .proc_handler   = proc_dointvec_minmax,
6996                 .extra1         = (void *)SYSCTL_ZERO,
6997                 .extra2         = (void *)&two_five_five,
6998         },
6999         {
7000                 .procname       = "rpl_seg_enabled",
7001                 .data           = &ipv6_devconf.rpl_seg_enabled,
7002                 .maxlen         = sizeof(int),
7003                 .mode           = 0644,
7004                 .proc_handler   = proc_dointvec,
7005         },
7006         {
7007                 .procname       = "ioam6_enabled",
7008                 .data           = &ipv6_devconf.ioam6_enabled,
7009                 .maxlen         = sizeof(u8),
7010                 .mode           = 0644,
7011                 .proc_handler   = proc_dou8vec_minmax,
7012                 .extra1         = (void *)SYSCTL_ZERO,
7013                 .extra2         = (void *)SYSCTL_ONE,
7014         },
7015         {
7016                 .procname       = "ioam6_id",
7017                 .data           = &ipv6_devconf.ioam6_id,
7018                 .maxlen         = sizeof(u32),
7019                 .mode           = 0644,
7020                 .proc_handler   = proc_douintvec_minmax,
7021                 .extra1         = (void *)SYSCTL_ZERO,
7022                 .extra2         = (void *)&ioam6_if_id_max,
7023         },
7024         {
7025                 .procname       = "ioam6_id_wide",
7026                 .data           = &ipv6_devconf.ioam6_id_wide,
7027                 .maxlen         = sizeof(u32),
7028                 .mode           = 0644,
7029                 .proc_handler   = proc_douintvec,
7030         },
7031         {
7032                 .procname       = "ndisc_evict_nocarrier",
7033                 .data           = &ipv6_devconf.ndisc_evict_nocarrier,
7034                 .maxlen         = sizeof(u8),
7035                 .mode           = 0644,
7036                 .proc_handler   = proc_dou8vec_minmax,
7037                 .extra1         = (void *)SYSCTL_ZERO,
7038                 .extra2         = (void *)SYSCTL_ONE,
7039         },
7040         {
7041                 .procname       = "accept_untracked_na",
7042                 .data           = &ipv6_devconf.accept_untracked_na,
7043                 .maxlen         = sizeof(int),
7044                 .mode           = 0644,
7045                 .proc_handler   = proc_dointvec_minmax,
7046                 .extra1         = (void *)SYSCTL_ZERO,
7047                 .extra2         = (void *)SYSCTL_ONE,
7048         },
7049         {
7050                 /* sentinel */
7051         }
7052 };
7053
7054 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7055                 struct inet6_dev *idev, struct ipv6_devconf *p)
7056 {
7057         int i, ifindex;
7058         struct ctl_table *table;
7059         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7060
7061         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7062         if (!table)
7063                 goto out;
7064
7065         for (i = 0; table[i].data; i++) {
7066                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7067                 /* If one of these is already set, then it is not safe to
7068                  * overwrite either of them: this makes proc_dointvec_minmax
7069                  * usable.
7070                  */
7071                 if (!table[i].extra1 && !table[i].extra2) {
7072                         table[i].extra1 = idev; /* embedded; no ref */
7073                         table[i].extra2 = net;
7074                 }
7075         }
7076
7077         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7078
7079         p->sysctl_header = register_net_sysctl(net, path, table);
7080         if (!p->sysctl_header)
7081                 goto free;
7082
7083         if (!strcmp(dev_name, "all"))
7084                 ifindex = NETCONFA_IFINDEX_ALL;
7085         else if (!strcmp(dev_name, "default"))
7086                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7087         else
7088                 ifindex = idev->dev->ifindex;
7089         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7090                                      ifindex, p);
7091         return 0;
7092
7093 free:
7094         kfree(table);
7095 out:
7096         return -ENOBUFS;
7097 }
7098
7099 static void __addrconf_sysctl_unregister(struct net *net,
7100                                          struct ipv6_devconf *p, int ifindex)
7101 {
7102         struct ctl_table *table;
7103
7104         if (!p->sysctl_header)
7105                 return;
7106
7107         table = p->sysctl_header->ctl_table_arg;
7108         unregister_net_sysctl_table(p->sysctl_header);
7109         p->sysctl_header = NULL;
7110         kfree(table);
7111
7112         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7113 }
7114
7115 static int addrconf_sysctl_register(struct inet6_dev *idev)
7116 {
7117         int err;
7118
7119         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7120                 return -EINVAL;
7121
7122         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7123                                     &ndisc_ifinfo_sysctl_change);
7124         if (err)
7125                 return err;
7126         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7127                                          idev, &idev->cnf);
7128         if (err)
7129                 neigh_sysctl_unregister(idev->nd_parms);
7130
7131         return err;
7132 }
7133
7134 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7135 {
7136         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7137                                      idev->dev->ifindex);
7138         neigh_sysctl_unregister(idev->nd_parms);
7139 }
7140
7141
7142 #endif
7143
7144 static int __net_init addrconf_init_net(struct net *net)
7145 {
7146         int err = -ENOMEM;
7147         struct ipv6_devconf *all, *dflt;
7148
7149         spin_lock_init(&net->ipv6.addrconf_hash_lock);
7150         INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7151         net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7152                                            sizeof(struct hlist_head),
7153                                            GFP_KERNEL);
7154         if (!net->ipv6.inet6_addr_lst)
7155                 goto err_alloc_addr;
7156
7157         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7158         if (!all)
7159                 goto err_alloc_all;
7160
7161         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7162         if (!dflt)
7163                 goto err_alloc_dflt;
7164
7165         if (IS_ENABLED(CONFIG_SYSCTL) &&
7166             !net_eq(net, &init_net)) {
7167                 switch (sysctl_devconf_inherit_init_net) {
7168                 case 1:  /* copy from init_net */
7169                         memcpy(all, init_net.ipv6.devconf_all,
7170                                sizeof(ipv6_devconf));
7171                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7172                                sizeof(ipv6_devconf_dflt));
7173                         break;
7174                 case 3: /* copy from the current netns */
7175                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7176                                sizeof(ipv6_devconf));
7177                         memcpy(dflt,
7178                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7179                                sizeof(ipv6_devconf_dflt));
7180                         break;
7181                 case 0:
7182                 case 2:
7183                         /* use compiled values */
7184                         break;
7185                 }
7186         }
7187
7188         /* these will be inherited by all namespaces */
7189         dflt->autoconf = ipv6_defaults.autoconf;
7190         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7191
7192         dflt->stable_secret.initialized = false;
7193         all->stable_secret.initialized = false;
7194
7195         net->ipv6.devconf_all = all;
7196         net->ipv6.devconf_dflt = dflt;
7197
7198 #ifdef CONFIG_SYSCTL
7199         err = __addrconf_sysctl_register(net, "all", NULL, all);
7200         if (err < 0)
7201                 goto err_reg_all;
7202
7203         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7204         if (err < 0)
7205                 goto err_reg_dflt;
7206 #endif
7207         return 0;
7208
7209 #ifdef CONFIG_SYSCTL
7210 err_reg_dflt:
7211         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7212 err_reg_all:
7213         kfree(dflt);
7214 #endif
7215 err_alloc_dflt:
7216         kfree(all);
7217 err_alloc_all:
7218         kfree(net->ipv6.inet6_addr_lst);
7219 err_alloc_addr:
7220         return err;
7221 }
7222
7223 static void __net_exit addrconf_exit_net(struct net *net)
7224 {
7225         int i;
7226
7227 #ifdef CONFIG_SYSCTL
7228         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7229                                      NETCONFA_IFINDEX_DEFAULT);
7230         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7231                                      NETCONFA_IFINDEX_ALL);
7232 #endif
7233         kfree(net->ipv6.devconf_dflt);
7234         net->ipv6.devconf_dflt = NULL;
7235         kfree(net->ipv6.devconf_all);
7236         net->ipv6.devconf_all = NULL;
7237
7238         cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7239         /*
7240          *      Check hash table, then free it.
7241          */
7242         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7243                 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7244
7245         kfree(net->ipv6.inet6_addr_lst);
7246         net->ipv6.inet6_addr_lst = NULL;
7247 }
7248
7249 static struct pernet_operations addrconf_ops = {
7250         .init = addrconf_init_net,
7251         .exit = addrconf_exit_net,
7252 };
7253
7254 static struct rtnl_af_ops inet6_ops __read_mostly = {
7255         .family           = AF_INET6,
7256         .fill_link_af     = inet6_fill_link_af,
7257         .get_link_af_size = inet6_get_link_af_size,
7258         .validate_link_af = inet6_validate_link_af,
7259         .set_link_af      = inet6_set_link_af,
7260 };
7261
7262 /*
7263  *      Init / cleanup code
7264  */
7265
7266 int __init addrconf_init(void)
7267 {
7268         struct inet6_dev *idev;
7269         int err;
7270
7271         err = ipv6_addr_label_init();
7272         if (err < 0) {
7273                 pr_crit("%s: cannot initialize default policy table: %d\n",
7274                         __func__, err);
7275                 goto out;
7276         }
7277
7278         err = register_pernet_subsys(&addrconf_ops);
7279         if (err < 0)
7280                 goto out_addrlabel;
7281
7282         addrconf_wq = create_workqueue("ipv6_addrconf");
7283         if (!addrconf_wq) {
7284                 err = -ENOMEM;
7285                 goto out_nowq;
7286         }
7287
7288         rtnl_lock();
7289         idev = ipv6_add_dev(blackhole_netdev);
7290         rtnl_unlock();
7291         if (IS_ERR(idev)) {
7292                 err = PTR_ERR(idev);
7293                 goto errlo;
7294         }
7295
7296         ip6_route_init_special_entries();
7297
7298         register_netdevice_notifier(&ipv6_dev_notf);
7299
7300         addrconf_verify(&init_net);
7301
7302         rtnl_af_register(&inet6_ops);
7303
7304         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7305                                    NULL, inet6_dump_ifinfo, 0);
7306         if (err < 0)
7307                 goto errout;
7308
7309         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7310                                    inet6_rtm_newaddr, NULL, 0);
7311         if (err < 0)
7312                 goto errout;
7313         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7314                                    inet6_rtm_deladdr, NULL, 0);
7315         if (err < 0)
7316                 goto errout;
7317         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7318                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7319                                    RTNL_FLAG_DOIT_UNLOCKED);
7320         if (err < 0)
7321                 goto errout;
7322         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7323                                    NULL, inet6_dump_ifmcaddr, 0);
7324         if (err < 0)
7325                 goto errout;
7326         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7327                                    NULL, inet6_dump_ifacaddr, 0);
7328         if (err < 0)
7329                 goto errout;
7330         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7331                                    inet6_netconf_get_devconf,
7332                                    inet6_netconf_dump_devconf,
7333                                    RTNL_FLAG_DOIT_UNLOCKED);
7334         if (err < 0)
7335                 goto errout;
7336         err = ipv6_addr_label_rtnl_register();
7337         if (err < 0)
7338                 goto errout;
7339
7340         return 0;
7341 errout:
7342         rtnl_unregister_all(PF_INET6);
7343         rtnl_af_unregister(&inet6_ops);
7344         unregister_netdevice_notifier(&ipv6_dev_notf);
7345 errlo:
7346         destroy_workqueue(addrconf_wq);
7347 out_nowq:
7348         unregister_pernet_subsys(&addrconf_ops);
7349 out_addrlabel:
7350         ipv6_addr_label_cleanup();
7351 out:
7352         return err;
7353 }
7354
7355 void addrconf_cleanup(void)
7356 {
7357         struct net_device *dev;
7358
7359         unregister_netdevice_notifier(&ipv6_dev_notf);
7360         unregister_pernet_subsys(&addrconf_ops);
7361         ipv6_addr_label_cleanup();
7362
7363         rtnl_af_unregister(&inet6_ops);
7364
7365         rtnl_lock();
7366
7367         /* clean dev list */
7368         for_each_netdev(&init_net, dev) {
7369                 if (__in6_dev_get(dev) == NULL)
7370                         continue;
7371                 addrconf_ifdown(dev, true);
7372         }
7373         addrconf_ifdown(init_net.loopback_dev, true);
7374
7375         rtnl_unlock();
7376
7377         destroy_workqueue(addrconf_wq);
7378 }