net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92 #include <linux/ioam6.h>
  93
  94 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  95
  96 #define IPV6_MAX_STRLEN \
  97         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  98
  99 static inline u32 cstamp_delta(unsigned long cstamp)
 100 {
 101         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 102 }
 103
 104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 105 {
 106         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 107         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 108         do_div(tmp, 1000000);
 109         return (s32)tmp;
 110 }
 111
 112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 113 {
 114         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 115         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 116         do_div(tmp, 1000000);
 117         if ((s32)tmp > mrt) {
 118                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 119                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 120                 do_div(tmp, 1000000);
 121         }
 122         return (s32)tmp;
 123 }
 124
 125 #ifdef CONFIG_SYSCTL
 126 static int addrconf_sysctl_register(struct inet6_dev *idev);
 127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 128 #else
 129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 130 {
 131         return 0;
 132 }
 133
 134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 135 {
 136 }
 137 #endif
 138
 139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 140
 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 142 static int ipv6_count_addresses(const struct inet6_dev *idev);
 143 static int ipv6_generate_stable_address(struct in6_addr *addr,
 144                                         u8 dad_count,
 145                                         const struct inet6_dev *idev);
 146
 147 #define IN6_ADDR_HSIZE_SHIFT    8
 148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 149 /*
 150  *      Configured unicast address hash table
 151  */
 152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 153 static DEFINE_SPINLOCK(addrconf_hash_lock);
 154
 155 static void addrconf_verify(void);
 156 static void addrconf_verify_rtnl(void);
 157 static void addrconf_verify_work(struct work_struct *);
 158
 159 static struct workqueue_struct *addrconf_wq;
 160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 161
 162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 164
 165 static void addrconf_type_change(struct net_device *dev,
 166                                  unsigned long event);
 167 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 168
 169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 170                                                   int plen,
 171                                                   const struct net_device *dev,
 172                                                   u32 flags, u32 noflags,
 173                                                   bool no_gw);
 174
 175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 176 static void addrconf_dad_work(struct work_struct *w);
 177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 178                                    bool send_na);
 179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 180 static void addrconf_rs_timer(struct timer_list *t);
 181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 183
 184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 185                                 struct prefix_info *pinfo);
 186
 187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 188         .forwarding             = 0,
 189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 190         .mtu6                   = IPV6_MIN_MTU,
 191         .accept_ra              = 1,
 192         .accept_redirects       = 1,
 193         .autoconf               = 1,
 194         .force_mld_version      = 0,
 195         .mldv1_unsolicited_report_interval = 10 * HZ,
 196         .mldv2_unsolicited_report_interval = HZ,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 210         .accept_ra_from_local   = 0,
 211         .accept_ra_min_hop_limit= 1,
 212         .accept_ra_min_lft      = 0,
 213         .accept_ra_pinfo        = 1,
 214 #ifdef CONFIG_IPV6_ROUTER_PREF
 215         .accept_ra_rtr_pref     = 1,
 216         .rtr_probe_interval     = 60 * HZ,
 217 #ifdef CONFIG_IPV6_ROUTE_INFO
 218         .accept_ra_rt_info_min_plen = 0,
 219         .accept_ra_rt_info_max_plen = 0,
 220 #endif
 221 #endif
 222         .proxy_ndp              = 0,
 223         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 224         .disable_ipv6           = 0,
 225         .accept_dad             = 0,
 226         .suppress_frag_ndisc    = 1,
 227         .accept_ra_mtu          = 1,
 228         .stable_secret          = {
 229                 .initialized = false,
 230         },
 231         .use_oif_addrs_only     = 0,
 232         .ignore_routes_with_linkdown = 0,
 233         .keep_addr_on_down      = 0,
 234         .seg6_enabled           = 0,
 235 #ifdef CONFIG_IPV6_SEG6_HMAC
 236         .seg6_require_hmac      = 0,
 237 #endif
 238         .enhanced_dad           = 1,
 239         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 240         .disable_policy         = 0,
 241         .rpl_seg_enabled        = 0,
 242         .ioam6_enabled          = 0,
 243         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 244         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 245 };
 246
 247 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 248         .forwarding             = 0,
 249         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 250         .mtu6                   = IPV6_MIN_MTU,
 251         .accept_ra              = 1,
 252         .accept_redirects       = 1,
 253         .autoconf               = 1,
 254         .force_mld_version      = 0,
 255         .mldv1_unsolicited_report_interval = 10 * HZ,
 256         .mldv2_unsolicited_report_interval = HZ,
 257         .dad_transmits          = 1,
 258         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 259         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 260         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 261         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 262         .use_tempaddr           = 0,
 263         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 264         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 265         .regen_max_retry        = REGEN_MAX_RETRY,
 266         .max_desync_factor      = MAX_DESYNC_FACTOR,
 267         .max_addresses          = IPV6_MAX_ADDRESSES,
 268         .accept_ra_defrtr       = 1,
 269         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 270         .accept_ra_from_local   = 0,
 271         .accept_ra_min_hop_limit= 1,
 272         .accept_ra_min_lft      = 0,
 273         .accept_ra_pinfo        = 1,
 274 #ifdef CONFIG_IPV6_ROUTER_PREF
 275         .accept_ra_rtr_pref     = 1,
 276         .rtr_probe_interval     = 60 * HZ,
 277 #ifdef CONFIG_IPV6_ROUTE_INFO
 278         .accept_ra_rt_info_min_plen = 0,
 279         .accept_ra_rt_info_max_plen = 0,
 280 #endif
 281 #endif
 282         .proxy_ndp              = 0,
 283         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 284         .disable_ipv6           = 0,
 285         .accept_dad             = 1,
 286         .suppress_frag_ndisc    = 1,
 287         .accept_ra_mtu          = 1,
 288         .stable_secret          = {
 289                 .initialized = false,
 290         },
 291         .use_oif_addrs_only     = 0,
 292         .ignore_routes_with_linkdown = 0,
 293         .keep_addr_on_down      = 0,
 294         .seg6_enabled           = 0,
 295 #ifdef CONFIG_IPV6_SEG6_HMAC
 296         .seg6_require_hmac      = 0,
 297 #endif
 298         .enhanced_dad           = 1,
 299         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 300         .disable_policy         = 0,
 301         .rpl_seg_enabled        = 0,
 302         .ioam6_enabled          = 0,
 303         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 304         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 305 };
 306
 307 /* Check if link is ready: is it up and is a valid qdisc available */
 308 static inline bool addrconf_link_ready(const struct net_device *dev)
 309 {
 310         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 311 }
 312
 313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 314 {
 315         if (del_timer(&idev->rs_timer))
 316                 __in6_dev_put(idev);
 317 }
 318
 319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 320 {
 321         if (cancel_delayed_work(&ifp->dad_work))
 322                 __in6_ifa_put(ifp);
 323 }
 324
 325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 326                                   unsigned long when)
 327 {
 328         if (!mod_timer(&idev->rs_timer, jiffies + when))
 329                 in6_dev_hold(idev);
 330 }
 331
 332 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 333                                    unsigned long delay)
 334 {
 335         in6_ifa_hold(ifp);
 336         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 337                 in6_ifa_put(ifp);
 338 }
 339
 340 static int snmp6_alloc_dev(struct inet6_dev *idev)
 341 {
 342         int i;
 343
 344         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 345         if (!idev->stats.ipv6)
 346                 goto err_ip;
 347
 348         for_each_possible_cpu(i) {
 349                 struct ipstats_mib *addrconf_stats;
 350                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 351                 u64_stats_init(&addrconf_stats->syncp);
 352         }
 353
 354
 355         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 356                                         GFP_KERNEL);
 357         if (!idev->stats.icmpv6dev)
 358                 goto err_icmp;
 359         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 360                                            GFP_KERNEL);
 361         if (!idev->stats.icmpv6msgdev)
 362                 goto err_icmpmsg;
 363
 364         return 0;
 365
 366 err_icmpmsg:
 367         kfree(idev->stats.icmpv6dev);
 368 err_icmp:
 369         free_percpu(idev->stats.ipv6);
 370 err_ip:
 371         return -ENOMEM;
 372 }
 373
 374 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 375 {
 376         struct inet6_dev *ndev;
 377         int err = -ENOMEM;
 378
 379         ASSERT_RTNL();
 380
 381         if (dev->mtu < IPV6_MIN_MTU)
 382                 return ERR_PTR(-EINVAL);
 383
 384         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 385         if (!ndev)
 386                 return ERR_PTR(err);
 387
 388         rwlock_init(&ndev->lock);
 389         ndev->dev = dev;
 390         INIT_LIST_HEAD(&ndev->addr_list);
 391         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 392         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 393
 394         if (ndev->cnf.stable_secret.initialized)
 395                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 396
 397         ndev->cnf.mtu6 = dev->mtu;
 398         ndev->ra_mtu = 0;
 399         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 400         if (!ndev->nd_parms) {
 401                 kfree(ndev);
 402                 return ERR_PTR(err);
 403         }
 404         if (ndev->cnf.forwarding)
 405                 dev_disable_lro(dev);
 406         /* We refer to the device */
 407         dev_hold(dev);
 408
 409         if (snmp6_alloc_dev(ndev) < 0) {
 410                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 411                            __func__);
 412                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 413                 dev_put(dev);
 414                 kfree(ndev);
 415                 return ERR_PTR(err);
 416         }
 417
 418         if (snmp6_register_dev(ndev) < 0) {
 419                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 420                            __func__, dev->name);
 421                 goto err_release;
 422         }
 423
 424         /* One reference from device. */
 425         refcount_set(&ndev->refcnt, 1);
 426
 427         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 428                 ndev->cnf.accept_dad = -1;
 429
 430 #if IS_ENABLED(CONFIG_IPV6_SIT)
 431         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 432                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 433                 ndev->cnf.rtr_solicits = 0;
 434         }
 435 #endif
 436
 437         INIT_LIST_HEAD(&ndev->tempaddr_list);
 438         ndev->desync_factor = U32_MAX;
 439         if ((dev->flags&IFF_LOOPBACK) ||
 440             dev->type == ARPHRD_TUNNEL ||
 441             dev->type == ARPHRD_TUNNEL6 ||
 442             dev->type == ARPHRD_SIT ||
 443             dev->type == ARPHRD_NONE) {
 444                 ndev->cnf.use_tempaddr = -1;
 445         }
 446
 447         ndev->token = in6addr_any;
 448
 449         if (netif_running(dev) && addrconf_link_ready(dev))
 450                 ndev->if_flags |= IF_READY;
 451
 452         ipv6_mc_init_dev(ndev);
 453         ndev->tstamp = jiffies;
 454         err = addrconf_sysctl_register(ndev);
 455         if (err) {
 456                 ipv6_mc_destroy_dev(ndev);
 457                 snmp6_unregister_dev(ndev);
 458                 goto err_release;
 459         }
 460         /* protected by rtnl_lock */
 461         rcu_assign_pointer(dev->ip6_ptr, ndev);
 462
 463         /* Join interface-local all-node multicast group */
 464         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 465
 466         /* Join all-node multicast group */
 467         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 468
 469         /* Join all-router multicast group if forwarding is set */
 470         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 471                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 472
 473         return ndev;
 474
 475 err_release:
 476         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 477         ndev->dead = 1;
 478         in6_dev_finish_destroy(ndev);
 479         return ERR_PTR(err);
 480 }
 481
 482 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 483 {
 484         struct inet6_dev *idev;
 485
 486         ASSERT_RTNL();
 487
 488         idev = __in6_dev_get(dev);
 489         if (!idev) {
 490                 idev = ipv6_add_dev(dev);
 491                 if (IS_ERR(idev))
 492                         return idev;
 493         }
 494
 495         if (dev->flags&IFF_UP)
 496                 ipv6_mc_up(idev);
 497         return idev;
 498 }
 499
 500 static int inet6_netconf_msgsize_devconf(int type)
 501 {
 502         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 503                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 504         bool all = false;
 505
 506         if (type == NETCONFA_ALL)
 507                 all = true;
 508
 509         if (all || type == NETCONFA_FORWARDING)
 510                 size += nla_total_size(4);
 511 #ifdef CONFIG_IPV6_MROUTE
 512         if (all || type == NETCONFA_MC_FORWARDING)
 513                 size += nla_total_size(4);
 514 #endif
 515         if (all || type == NETCONFA_PROXY_NEIGH)
 516                 size += nla_total_size(4);
 517
 518         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 519                 size += nla_total_size(4);
 520
 521         return size;
 522 }
 523
 524 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 525                                       struct ipv6_devconf *devconf, u32 portid,
 526                                       u32 seq, int event, unsigned int flags,
 527                                       int type)
 528 {
 529         struct nlmsghdr  *nlh;
 530         struct netconfmsg *ncm;
 531         bool all = false;
 532
 533         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 534                         flags);
 535         if (!nlh)
 536                 return -EMSGSIZE;
 537
 538         if (type == NETCONFA_ALL)
 539                 all = true;
 540
 541         ncm = nlmsg_data(nlh);
 542         ncm->ncm_family = AF_INET6;
 543
 544         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 545                 goto nla_put_failure;
 546
 547         if (!devconf)
 548                 goto out;
 549
 550         if ((all || type == NETCONFA_FORWARDING) &&
 551             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 552                 goto nla_put_failure;
 553 #ifdef CONFIG_IPV6_MROUTE
 554         if ((all || type == NETCONFA_MC_FORWARDING) &&
 555             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 556                         atomic_read(&devconf->mc_forwarding)) < 0)
 557                 goto nla_put_failure;
 558 #endif
 559         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 560             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 561                 goto nla_put_failure;
 562
 563         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 564             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 565                         devconf->ignore_routes_with_linkdown) < 0)
 566                 goto nla_put_failure;
 567
 568 out:
 569         nlmsg_end(skb, nlh);
 570         return 0;
 571
 572 nla_put_failure:
 573         nlmsg_cancel(skb, nlh);
 574         return -EMSGSIZE;
 575 }
 576
 577 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 578                                   int ifindex, struct ipv6_devconf *devconf)
 579 {
 580         struct sk_buff *skb;
 581         int err = -ENOBUFS;
 582
 583         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 584         if (!skb)
 585                 goto errout;
 586
 587         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 588                                          event, 0, type);
 589         if (err < 0) {
 590                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 591                 WARN_ON(err == -EMSGSIZE);
 592                 kfree_skb(skb);
 593                 goto errout;
 594         }
 595         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 596         return;
 597 errout:
 598         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 599 }
 600
 601 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 602         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 603         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 604         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 605         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 606 };
 607
 608 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 609                                        const struct nlmsghdr *nlh,
 610                                        struct nlattr **tb,
 611                                        struct netlink_ext_ack *extack)
 612 {
 613         int i, err;
 614
 615         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 616                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 617                 return -EINVAL;
 618         }
 619
 620         if (!netlink_strict_get_check(skb))
 621                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 622                                               tb, NETCONFA_MAX,
 623                                               devconf_ipv6_policy, extack);
 624
 625         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 626                                             tb, NETCONFA_MAX,
 627                                             devconf_ipv6_policy, extack);
 628         if (err)
 629                 return err;
 630
 631         for (i = 0; i <= NETCONFA_MAX; i++) {
 632                 if (!tb[i])
 633                         continue;
 634
 635                 switch (i) {
 636                 case NETCONFA_IFINDEX:
 637                         break;
 638                 default:
 639                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 640                         return -EINVAL;
 641                 }
 642         }
 643
 644         return 0;
 645 }
 646
 647 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 648                                      struct nlmsghdr *nlh,
 649                                      struct netlink_ext_ack *extack)
 650 {
 651         struct net *net = sock_net(in_skb->sk);
 652         struct nlattr *tb[NETCONFA_MAX+1];
 653         struct inet6_dev *in6_dev = NULL;
 654         struct net_device *dev = NULL;
 655         struct sk_buff *skb;
 656         struct ipv6_devconf *devconf;
 657         int ifindex;
 658         int err;
 659
 660         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 661         if (err < 0)
 662                 return err;
 663
 664         if (!tb[NETCONFA_IFINDEX])
 665                 return -EINVAL;
 666
 667         err = -EINVAL;
 668         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 669         switch (ifindex) {
 670         case NETCONFA_IFINDEX_ALL:
 671                 devconf = net->ipv6.devconf_all;
 672                 break;
 673         case NETCONFA_IFINDEX_DEFAULT:
 674                 devconf = net->ipv6.devconf_dflt;
 675                 break;
 676         default:
 677                 dev = dev_get_by_index(net, ifindex);
 678                 if (!dev)
 679                         return -EINVAL;
 680                 in6_dev = in6_dev_get(dev);
 681                 if (!in6_dev)
 682                         goto errout;
 683                 devconf = &in6_dev->cnf;
 684                 break;
 685         }
 686
 687         err = -ENOBUFS;
 688         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 689         if (!skb)
 690                 goto errout;
 691
 692         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 693                                          NETLINK_CB(in_skb).portid,
 694                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 695                                          NETCONFA_ALL);
 696         if (err < 0) {
 697                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 698                 WARN_ON(err == -EMSGSIZE);
 699                 kfree_skb(skb);
 700                 goto errout;
 701         }
 702         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 703 errout:
 704         if (in6_dev)
 705                 in6_dev_put(in6_dev);
 706         dev_put(dev);
 707         return err;
 708 }
 709
 710 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 711                                       struct netlink_callback *cb)
 712 {
 713         const struct nlmsghdr *nlh = cb->nlh;
 714         struct net *net = sock_net(skb->sk);
 715         int h, s_h;
 716         int idx, s_idx;
 717         struct net_device *dev;
 718         struct inet6_dev *idev;
 719         struct hlist_head *head;
 720
 721         if (cb->strict_check) {
 722                 struct netlink_ext_ack *extack = cb->extack;
 723                 struct netconfmsg *ncm;
 724
 725                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 726                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 727                         return -EINVAL;
 728                 }
 729
 730                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 731                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 732                         return -EINVAL;
 733                 }
 734         }
 735
 736         s_h = cb->args[0];
 737         s_idx = idx = cb->args[1];
 738
 739         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 740                 idx = 0;
 741                 head = &net->dev_index_head[h];
 742                 rcu_read_lock();
 743                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 744                           net->dev_base_seq;
 745                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 746                         if (idx < s_idx)
 747                                 goto cont;
 748                         idev = __in6_dev_get(dev);
 749                         if (!idev)
 750                                 goto cont;
 751
 752                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 753                                                        &idev->cnf,
 754                                                        NETLINK_CB(cb->skb).portid,
 755                                                        nlh->nlmsg_seq,
 756                                                        RTM_NEWNETCONF,
 757                                                        NLM_F_MULTI,
 758                                                        NETCONFA_ALL) < 0) {
 759                                 rcu_read_unlock();
 760                                 goto done;
 761                         }
 762                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 763 cont:
 764                         idx++;
 765                 }
 766                 rcu_read_unlock();
 767         }
 768         if (h == NETDEV_HASHENTRIES) {
 769                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 770                                                net->ipv6.devconf_all,
 771                                                NETLINK_CB(cb->skb).portid,
 772                                                nlh->nlmsg_seq,
 773                                                RTM_NEWNETCONF, NLM_F_MULTI,
 774                                                NETCONFA_ALL) < 0)
 775                         goto done;
 776                 else
 777                         h++;
 778         }
 779         if (h == NETDEV_HASHENTRIES + 1) {
 780                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 781                                                net->ipv6.devconf_dflt,
 782                                                NETLINK_CB(cb->skb).portid,
 783                                                nlh->nlmsg_seq,
 784                                                RTM_NEWNETCONF, NLM_F_MULTI,
 785                                                NETCONFA_ALL) < 0)
 786                         goto done;
 787                 else
 788                         h++;
 789         }
 790 done:
 791         cb->args[0] = h;
 792         cb->args[1] = idx;
 793
 794         return skb->len;
 795 }
 796
 797 #ifdef CONFIG_SYSCTL
 798 static void dev_forward_change(struct inet6_dev *idev)
 799 {
 800         struct net_device *dev;
 801         struct inet6_ifaddr *ifa;
 802         LIST_HEAD(tmp_addr_list);
 803
 804         if (!idev)
 805                 return;
 806         dev = idev->dev;
 807         if (idev->cnf.forwarding)
 808                 dev_disable_lro(dev);
 809         if (dev->flags & IFF_MULTICAST) {
 810                 if (idev->cnf.forwarding) {
 811                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 812                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 813                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 814                 } else {
 815                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 816                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 817                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 818                 }
 819         }
 820
 821         read_lock_bh(&idev->lock);
 822         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 823                 if (ifa->flags&IFA_F_TENTATIVE)
 824                         continue;
 825                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
 826         }
 827         read_unlock_bh(&idev->lock);
 828
 829         while (!list_empty(&tmp_addr_list)) {
 830                 ifa = list_first_entry(&tmp_addr_list,
 831                                        struct inet6_ifaddr, if_list_aux);
 832                 list_del(&ifa->if_list_aux);
 833                 if (idev->cnf.forwarding)
 834                         addrconf_join_anycast(ifa);
 835                 else
 836                         addrconf_leave_anycast(ifa);
 837         }
 838
 839         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 840                                      NETCONFA_FORWARDING,
 841                                      dev->ifindex, &idev->cnf);
 842 }
 843
 844
 845 static void addrconf_forward_change(struct net *net, __s32 newf)
 846 {
 847         struct net_device *dev;
 848         struct inet6_dev *idev;
 849
 850         for_each_netdev(net, dev) {
 851                 idev = __in6_dev_get(dev);
 852                 if (idev) {
 853                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 854                         idev->cnf.forwarding = newf;
 855                         if (changed)
 856                                 dev_forward_change(idev);
 857                 }
 858         }
 859 }
 860
 861 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 862 {
 863         struct net *net;
 864         int old;
 865
 866         if (!rtnl_trylock())
 867                 return restart_syscall();
 868
 869         net = (struct net *)table->extra2;
 870         old = *p;
 871         *p = newf;
 872
 873         if (p == &net->ipv6.devconf_dflt->forwarding) {
 874                 if ((!newf) ^ (!old))
 875                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 876                                                      NETCONFA_FORWARDING,
 877                                                      NETCONFA_IFINDEX_DEFAULT,
 878                                                      net->ipv6.devconf_dflt);
 879                 rtnl_unlock();
 880                 return 0;
 881         }
 882
 883         if (p == &net->ipv6.devconf_all->forwarding) {
 884                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 885
 886                 net->ipv6.devconf_dflt->forwarding = newf;
 887                 if ((!newf) ^ (!old_dflt))
 888                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 889                                                      NETCONFA_FORWARDING,
 890                                                      NETCONFA_IFINDEX_DEFAULT,
 891                                                      net->ipv6.devconf_dflt);
 892
 893                 addrconf_forward_change(net, newf);
 894                 if ((!newf) ^ (!old))
 895                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 896                                                      NETCONFA_FORWARDING,
 897                                                      NETCONFA_IFINDEX_ALL,
 898                                                      net->ipv6.devconf_all);
 899         } else if ((!newf) ^ (!old))
 900                 dev_forward_change((struct inet6_dev *)table->extra1);
 901         rtnl_unlock();
 902
 903         if (newf)
 904                 rt6_purge_dflt_routers(net);
 905         return 1;
 906 }
 907
 908 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 909 {
 910         struct net_device *dev;
 911         struct inet6_dev *idev;
 912
 913         for_each_netdev(net, dev) {
 914                 idev = __in6_dev_get(dev);
 915                 if (idev) {
 916                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 917
 918                         idev->cnf.ignore_routes_with_linkdown = newf;
 919                         if (changed)
 920                                 inet6_netconf_notify_devconf(dev_net(dev),
 921                                                              RTM_NEWNETCONF,
 922                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 923                                                              dev->ifindex,
 924                                                              &idev->cnf);
 925                 }
 926         }
 927 }
 928
 929 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 930 {
 931         struct net *net;
 932         int old;
 933
 934         if (!rtnl_trylock())
 935                 return restart_syscall();
 936
 937         net = (struct net *)table->extra2;
 938         old = *p;
 939         *p = newf;
 940
 941         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 942                 if ((!newf) ^ (!old))
 943                         inet6_netconf_notify_devconf(net,
 944                                                      RTM_NEWNETCONF,
 945                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 946                                                      NETCONFA_IFINDEX_DEFAULT,
 947                                                      net->ipv6.devconf_dflt);
 948                 rtnl_unlock();
 949                 return 0;
 950         }
 951
 952         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 953                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 954                 addrconf_linkdown_change(net, newf);
 955                 if ((!newf) ^ (!old))
 956                         inet6_netconf_notify_devconf(net,
 957                                                      RTM_NEWNETCONF,
 958                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 959                                                      NETCONFA_IFINDEX_ALL,
 960                                                      net->ipv6.devconf_all);
 961         }
 962         rtnl_unlock();
 963
 964         return 1;
 965 }
 966
 967 #endif
 968
 969 /* Nobody refers to this ifaddr, destroy it */
 970 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 971 {
 972         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 973
 974 #ifdef NET_REFCNT_DEBUG
 975         pr_debug("%s\n", __func__);
 976 #endif
 977
 978         in6_dev_put(ifp->idev);
 979
 980         if (cancel_delayed_work(&ifp->dad_work))
 981                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 982                           ifp);
 983
 984         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 985                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 986                 return;
 987         }
 988
 989         kfree_rcu(ifp, rcu);
 990 }
 991
 992 static void
 993 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 994 {
 995         struct list_head *p;
 996         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 997
 998         /*
 999          * Each device address list is sorted in order of scope -
1000          * global before linklocal.
1001          */
1002         list_for_each(p, &idev->addr_list) {
1003                 struct inet6_ifaddr *ifa
1004                         = list_entry(p, struct inet6_ifaddr, if_list);
1005                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1006                         break;
1007         }
1008
1009         list_add_tail_rcu(&ifp->if_list, p);
1010 }
1011
1012 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1013 {
1014         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1015
1016         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1017 }
1018
1019 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1020                                struct net_device *dev, unsigned int hash)
1021 {
1022         struct inet6_ifaddr *ifp;
1023
1024         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1025                 if (!net_eq(dev_net(ifp->idev->dev), net))
1026                         continue;
1027                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1028                         if (!dev || ifp->idev->dev == dev)
1029                                 return true;
1030                 }
1031         }
1032         return false;
1033 }
1034
1035 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1036 {
1037         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1038         int err = 0;
1039
1040         spin_lock(&addrconf_hash_lock);
1041
1042         /* Ignore adding duplicate addresses on an interface */
1043         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1044                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1045                 err = -EEXIST;
1046         } else {
1047                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1048         }
1049
1050         spin_unlock(&addrconf_hash_lock);
1051
1052         return err;
1053 }
1054
1055 /* On success it returns ifp with increased reference count */
1056
1057 static struct inet6_ifaddr *
1058 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1059               bool can_block, struct netlink_ext_ack *extack)
1060 {
1061         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1062         int addr_type = ipv6_addr_type(cfg->pfx);
1063         struct net *net = dev_net(idev->dev);
1064         struct inet6_ifaddr *ifa = NULL;
1065         struct fib6_info *f6i = NULL;
1066         int err = 0;
1067
1068         if (addr_type == IPV6_ADDR_ANY ||
1069             (addr_type & IPV6_ADDR_MULTICAST &&
1070              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1071             (!(idev->dev->flags & IFF_LOOPBACK) &&
1072              !netif_is_l3_master(idev->dev) &&
1073              addr_type & IPV6_ADDR_LOOPBACK))
1074                 return ERR_PTR(-EADDRNOTAVAIL);
1075
1076         if (idev->dead) {
1077                 err = -ENODEV;                  /*XXX*/
1078                 goto out;
1079         }
1080
1081         if (idev->cnf.disable_ipv6) {
1082                 err = -EACCES;
1083                 goto out;
1084         }
1085
1086         /* validator notifier needs to be blocking;
1087          * do not call in atomic context
1088          */
1089         if (can_block) {
1090                 struct in6_validator_info i6vi = {
1091                         .i6vi_addr = *cfg->pfx,
1092                         .i6vi_dev = idev,
1093                         .extack = extack,
1094                 };
1095
1096                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1097                 err = notifier_to_errno(err);
1098                 if (err < 0)
1099                         goto out;
1100         }
1101
1102         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1103         if (!ifa) {
1104                 err = -ENOBUFS;
1105                 goto out;
1106         }
1107
1108         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1109         if (IS_ERR(f6i)) {
1110                 err = PTR_ERR(f6i);
1111                 f6i = NULL;
1112                 goto out;
1113         }
1114
1115         neigh_parms_data_state_setall(idev->nd_parms);
1116
1117         ifa->addr = *cfg->pfx;
1118         if (cfg->peer_pfx)
1119                 ifa->peer_addr = *cfg->peer_pfx;
1120
1121         spin_lock_init(&ifa->lock);
1122         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1123         INIT_HLIST_NODE(&ifa->addr_lst);
1124         ifa->scope = cfg->scope;
1125         ifa->prefix_len = cfg->plen;
1126         ifa->rt_priority = cfg->rt_priority;
1127         ifa->flags = cfg->ifa_flags;
1128         /* No need to add the TENTATIVE flag for addresses with NODAD */
1129         if (!(cfg->ifa_flags & IFA_F_NODAD))
1130                 ifa->flags |= IFA_F_TENTATIVE;
1131         ifa->valid_lft = cfg->valid_lft;
1132         ifa->prefered_lft = cfg->preferred_lft;
1133         ifa->cstamp = ifa->tstamp = jiffies;
1134         ifa->tokenized = false;
1135
1136         ifa->rt = f6i;
1137
1138         ifa->idev = idev;
1139         in6_dev_hold(idev);
1140
1141         /* For caller */
1142         refcount_set(&ifa->refcnt, 1);
1143
1144         rcu_read_lock_bh();
1145
1146         err = ipv6_add_addr_hash(idev->dev, ifa);
1147         if (err < 0) {
1148                 rcu_read_unlock_bh();
1149                 goto out;
1150         }
1151
1152         write_lock(&idev->lock);
1153
1154         /* Add to inet6_dev unicast addr list. */
1155         ipv6_link_dev_addr(idev, ifa);
1156
1157         if (ifa->flags&IFA_F_TEMPORARY) {
1158                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1159                 in6_ifa_hold(ifa);
1160         }
1161
1162         in6_ifa_hold(ifa);
1163         write_unlock(&idev->lock);
1164
1165         rcu_read_unlock_bh();
1166
1167         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1168 out:
1169         if (unlikely(err < 0)) {
1170                 fib6_info_release(f6i);
1171
1172                 if (ifa) {
1173                         if (ifa->idev)
1174                                 in6_dev_put(ifa->idev);
1175                         kfree(ifa);
1176                 }
1177                 ifa = ERR_PTR(err);
1178         }
1179
1180         return ifa;
1181 }
1182
1183 enum cleanup_prefix_rt_t {
1184         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1185         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1186         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1187 };
1188
1189 /*
1190  * Check, whether the prefix for ifp would still need a prefix route
1191  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1192  * constants.
1193  *
1194  * 1) we don't purge prefix if address was not permanent.
1195  *    prefix is managed by its own lifetime.
1196  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1197  * 3) if there are no addresses, delete prefix.
1198  * 4) if there are still other permanent address(es),
1199  *    corresponding prefix is still permanent.
1200  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1201  *    don't purge the prefix, assume user space is managing it.
1202  * 6) otherwise, update prefix lifetime to the
1203  *    longest valid lifetime among the corresponding
1204  *    addresses on the device.
1205  *    Note: subsequent RA will update lifetime.
1206  **/
1207 static enum cleanup_prefix_rt_t
1208 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1209 {
1210         struct inet6_ifaddr *ifa;
1211         struct inet6_dev *idev = ifp->idev;
1212         unsigned long lifetime;
1213         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1214
1215         *expires = jiffies;
1216
1217         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1218                 if (ifa == ifp)
1219                         continue;
1220                 if (ifa->prefix_len != ifp->prefix_len ||
1221                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1222                                        ifp->prefix_len))
1223                         continue;
1224                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1225                         return CLEANUP_PREFIX_RT_NOP;
1226
1227                 action = CLEANUP_PREFIX_RT_EXPIRE;
1228
1229                 spin_lock(&ifa->lock);
1230
1231                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1232                 /*
1233                  * Note: Because this address is
1234                  * not permanent, lifetime <
1235                  * LONG_MAX / HZ here.
1236                  */
1237                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1238                         *expires = ifa->tstamp + lifetime * HZ;
1239                 spin_unlock(&ifa->lock);
1240         }
1241
1242         return action;
1243 }
1244
1245 static void
1246 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1247                      bool del_rt, bool del_peer)
1248 {
1249         struct fib6_info *f6i;
1250
1251         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1252                                         ifp->prefix_len,
1253                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1254         if (f6i) {
1255                 if (del_rt)
1256                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1257                 else {
1258                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1259                                 fib6_set_expires(f6i, expires);
1260                         fib6_info_release(f6i);
1261                 }
1262         }
1263 }
1264
1265
1266 /* This function wants to get referenced ifp and releases it before return */
1267
1268 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1269 {
1270         int state;
1271         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1272         unsigned long expires;
1273
1274         ASSERT_RTNL();
1275
1276         spin_lock_bh(&ifp->lock);
1277         state = ifp->state;
1278         ifp->state = INET6_IFADDR_STATE_DEAD;
1279         spin_unlock_bh(&ifp->lock);
1280
1281         if (state == INET6_IFADDR_STATE_DEAD)
1282                 goto out;
1283
1284         spin_lock_bh(&addrconf_hash_lock);
1285         hlist_del_init_rcu(&ifp->addr_lst);
1286         spin_unlock_bh(&addrconf_hash_lock);
1287
1288         write_lock_bh(&ifp->idev->lock);
1289
1290         if (ifp->flags&IFA_F_TEMPORARY) {
1291                 list_del(&ifp->tmp_list);
1292                 if (ifp->ifpub) {
1293                         in6_ifa_put(ifp->ifpub);
1294                         ifp->ifpub = NULL;
1295                 }
1296                 __in6_ifa_put(ifp);
1297         }
1298
1299         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1300                 action = check_cleanup_prefix_route(ifp, &expires);
1301
1302         list_del_rcu(&ifp->if_list);
1303         __in6_ifa_put(ifp);
1304
1305         write_unlock_bh(&ifp->idev->lock);
1306
1307         addrconf_del_dad_work(ifp);
1308
1309         ipv6_ifa_notify(RTM_DELADDR, ifp);
1310
1311         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1312
1313         if (action != CLEANUP_PREFIX_RT_NOP) {
1314                 cleanup_prefix_route(ifp, expires,
1315                         action == CLEANUP_PREFIX_RT_DEL, false);
1316         }
1317
1318         /* clean up prefsrc entries */
1319         rt6_remove_prefsrc(ifp);
1320 out:
1321         in6_ifa_put(ifp);
1322 }
1323
1324 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1325 {
1326         struct inet6_dev *idev = ifp->idev;
1327         unsigned long tmp_tstamp, age;
1328         unsigned long regen_advance;
1329         unsigned long now = jiffies;
1330         s32 cnf_temp_preferred_lft;
1331         struct inet6_ifaddr *ift;
1332         struct ifa6_config cfg;
1333         long max_desync_factor;
1334         struct in6_addr addr;
1335         int ret = 0;
1336
1337         write_lock_bh(&idev->lock);
1338
1339 retry:
1340         in6_dev_hold(idev);
1341         if (idev->cnf.use_tempaddr <= 0) {
1342                 write_unlock_bh(&idev->lock);
1343                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1344                 in6_dev_put(idev);
1345                 ret = -1;
1346                 goto out;
1347         }
1348         spin_lock_bh(&ifp->lock);
1349         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1350                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1351                 spin_unlock_bh(&ifp->lock);
1352                 write_unlock_bh(&idev->lock);
1353                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1354                         __func__);
1355                 in6_dev_put(idev);
1356                 ret = -1;
1357                 goto out;
1358         }
1359         in6_ifa_hold(ifp);
1360         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1361         ipv6_gen_rnd_iid(&addr);
1362
1363         age = (now - ifp->tstamp) / HZ;
1364
1365         regen_advance = idev->cnf.regen_max_retry *
1366                         idev->cnf.dad_transmits *
1367                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1368
1369         /* recalculate max_desync_factor each time and update
1370          * idev->desync_factor if it's larger
1371          */
1372         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1373         max_desync_factor = min_t(long,
1374                                   idev->cnf.max_desync_factor,
1375                                   cnf_temp_preferred_lft - regen_advance);
1376
1377         if (unlikely(idev->desync_factor > max_desync_factor)) {
1378                 if (max_desync_factor > 0) {
1379                         get_random_bytes(&idev->desync_factor,
1380                                          sizeof(idev->desync_factor));
1381                         idev->desync_factor %= max_desync_factor;
1382                 } else {
1383                         idev->desync_factor = 0;
1384                 }
1385         }
1386
1387         memset(&cfg, 0, sizeof(cfg));
1388         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1389                               idev->cnf.temp_valid_lft + age);
1390         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1391         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1392
1393         cfg.plen = ifp->prefix_len;
1394         tmp_tstamp = ifp->tstamp;
1395         spin_unlock_bh(&ifp->lock);
1396
1397         write_unlock_bh(&idev->lock);
1398
1399         /* A temporary address is created only if this calculated Preferred
1400          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1401          * an implementation must not create a temporary address with a zero
1402          * Preferred Lifetime.
1403          * Use age calculation as in addrconf_verify to avoid unnecessary
1404          * temporary addresses being generated.
1405          */
1406         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1407         if (cfg.preferred_lft <= regen_advance + age) {
1408                 in6_ifa_put(ifp);
1409                 in6_dev_put(idev);
1410                 ret = -1;
1411                 goto out;
1412         }
1413
1414         cfg.ifa_flags = IFA_F_TEMPORARY;
1415         /* set in addrconf_prefix_rcv() */
1416         if (ifp->flags & IFA_F_OPTIMISTIC)
1417                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1418
1419         cfg.pfx = &addr;
1420         cfg.scope = ipv6_addr_scope(cfg.pfx);
1421
1422         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1423         if (IS_ERR(ift)) {
1424                 in6_ifa_put(ifp);
1425                 in6_dev_put(idev);
1426                 pr_info("%s: retry temporary address regeneration\n", __func__);
1427                 write_lock_bh(&idev->lock);
1428                 goto retry;
1429         }
1430
1431         spin_lock_bh(&ift->lock);
1432         ift->ifpub = ifp;
1433         ift->cstamp = now;
1434         ift->tstamp = tmp_tstamp;
1435         spin_unlock_bh(&ift->lock);
1436
1437         addrconf_dad_start(ift);
1438         in6_ifa_put(ift);
1439         in6_dev_put(idev);
1440 out:
1441         return ret;
1442 }
1443
1444 /*
1445  *      Choose an appropriate source address (RFC3484)
1446  */
1447 enum {
1448         IPV6_SADDR_RULE_INIT = 0,
1449         IPV6_SADDR_RULE_LOCAL,
1450         IPV6_SADDR_RULE_SCOPE,
1451         IPV6_SADDR_RULE_PREFERRED,
1452 #ifdef CONFIG_IPV6_MIP6
1453         IPV6_SADDR_RULE_HOA,
1454 #endif
1455         IPV6_SADDR_RULE_OIF,
1456         IPV6_SADDR_RULE_LABEL,
1457         IPV6_SADDR_RULE_PRIVACY,
1458         IPV6_SADDR_RULE_ORCHID,
1459         IPV6_SADDR_RULE_PREFIX,
1460 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1461         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1462 #endif
1463         IPV6_SADDR_RULE_MAX
1464 };
1465
1466 struct ipv6_saddr_score {
1467         int                     rule;
1468         int                     addr_type;
1469         struct inet6_ifaddr     *ifa;
1470         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1471         int                     scopedist;
1472         int                     matchlen;
1473 };
1474
1475 struct ipv6_saddr_dst {
1476         const struct in6_addr *addr;
1477         int ifindex;
1478         int scope;
1479         int label;
1480         unsigned int prefs;
1481 };
1482
1483 static inline int ipv6_saddr_preferred(int type)
1484 {
1485         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1486                 return 1;
1487         return 0;
1488 }
1489
1490 static bool ipv6_use_optimistic_addr(struct net *net,
1491                                      struct inet6_dev *idev)
1492 {
1493 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1494         if (!idev)
1495                 return false;
1496         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1497                 return false;
1498         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1499                 return false;
1500
1501         return true;
1502 #else
1503         return false;
1504 #endif
1505 }
1506
1507 static bool ipv6_allow_optimistic_dad(struct net *net,
1508                                       struct inet6_dev *idev)
1509 {
1510 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1511         if (!idev)
1512                 return false;
1513         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1514                 return false;
1515
1516         return true;
1517 #else
1518         return false;
1519 #endif
1520 }
1521
1522 static int ipv6_get_saddr_eval(struct net *net,
1523                                struct ipv6_saddr_score *score,
1524                                struct ipv6_saddr_dst *dst,
1525                                int i)
1526 {
1527         int ret;
1528
1529         if (i <= score->rule) {
1530                 switch (i) {
1531                 case IPV6_SADDR_RULE_SCOPE:
1532                         ret = score->scopedist;
1533                         break;
1534                 case IPV6_SADDR_RULE_PREFIX:
1535                         ret = score->matchlen;
1536                         break;
1537                 default:
1538                         ret = !!test_bit(i, score->scorebits);
1539                 }
1540                 goto out;
1541         }
1542
1543         switch (i) {
1544         case IPV6_SADDR_RULE_INIT:
1545                 /* Rule 0: remember if hiscore is not ready yet */
1546                 ret = !!score->ifa;
1547                 break;
1548         case IPV6_SADDR_RULE_LOCAL:
1549                 /* Rule 1: Prefer same address */
1550                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1551                 break;
1552         case IPV6_SADDR_RULE_SCOPE:
1553                 /* Rule 2: Prefer appropriate scope
1554                  *
1555                  *      ret
1556                  *       ^
1557                  *    -1 |  d 15
1558                  *    ---+--+-+---> scope
1559                  *       |
1560                  *       |             d is scope of the destination.
1561                  *  B-d  |  \
1562                  *       |   \      <- smaller scope is better if
1563                  *  B-15 |    \        if scope is enough for destination.
1564                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1565                  * d-C-1 | /
1566                  *       |/         <- greater is better
1567                  *   -C  /             if scope is not enough for destination.
1568                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1569                  *
1570                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1571                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1572                  * Assume B = 0 and we get C > 29.
1573                  */
1574                 ret = __ipv6_addr_src_scope(score->addr_type);
1575                 if (ret >= dst->scope)
1576                         ret = -ret;
1577                 else
1578                         ret -= 128;     /* 30 is enough */
1579                 score->scopedist = ret;
1580                 break;
1581         case IPV6_SADDR_RULE_PREFERRED:
1582             {
1583                 /* Rule 3: Avoid deprecated and optimistic addresses */
1584                 u8 avoid = IFA_F_DEPRECATED;
1585
1586                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1587                         avoid |= IFA_F_OPTIMISTIC;
1588                 ret = ipv6_saddr_preferred(score->addr_type) ||
1589                       !(score->ifa->flags & avoid);
1590                 break;
1591             }
1592 #ifdef CONFIG_IPV6_MIP6
1593         case IPV6_SADDR_RULE_HOA:
1594             {
1595                 /* Rule 4: Prefer home address */
1596                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1597                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1598                 break;
1599             }
1600 #endif
1601         case IPV6_SADDR_RULE_OIF:
1602                 /* Rule 5: Prefer outgoing interface */
1603                 ret = (!dst->ifindex ||
1604                        dst->ifindex == score->ifa->idev->dev->ifindex);
1605                 break;
1606         case IPV6_SADDR_RULE_LABEL:
1607                 /* Rule 6: Prefer matching label */
1608                 ret = ipv6_addr_label(net,
1609                                       &score->ifa->addr, score->addr_type,
1610                                       score->ifa->idev->dev->ifindex) == dst->label;
1611                 break;
1612         case IPV6_SADDR_RULE_PRIVACY:
1613             {
1614                 /* Rule 7: Prefer public address
1615                  * Note: prefer temporary address if use_tempaddr >= 2
1616                  */
1617                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1618                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1619                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1620                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1621                 break;
1622             }
1623         case IPV6_SADDR_RULE_ORCHID:
1624                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1625                  *          non-ORCHID vs non-ORCHID
1626                  */
1627                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1628                         ipv6_addr_orchid(dst->addr));
1629                 break;
1630         case IPV6_SADDR_RULE_PREFIX:
1631                 /* Rule 8: Use longest matching prefix */
1632                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1633                 if (ret > score->ifa->prefix_len)
1634                         ret = score->ifa->prefix_len;
1635                 score->matchlen = ret;
1636                 break;
1637 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1638         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1639                 /* Optimistic addresses still have lower precedence than other
1640                  * preferred addresses.
1641                  */
1642                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1643                 break;
1644 #endif
1645         default:
1646                 ret = 0;
1647         }
1648
1649         if (ret)
1650                 __set_bit(i, score->scorebits);
1651         score->rule = i;
1652 out:
1653         return ret;
1654 }
1655
1656 static int __ipv6_dev_get_saddr(struct net *net,
1657                                 struct ipv6_saddr_dst *dst,
1658                                 struct inet6_dev *idev,
1659                                 struct ipv6_saddr_score *scores,
1660                                 int hiscore_idx)
1661 {
1662         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1663
1664         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1665                 int i;
1666
1667                 /*
1668                  * - Tentative Address (RFC2462 section 5.4)
1669                  *  - A tentative address is not considered
1670                  *    "assigned to an interface" in the traditional
1671                  *    sense, unless it is also flagged as optimistic.
1672                  * - Candidate Source Address (section 4)
1673                  *  - In any case, anycast addresses, multicast
1674                  *    addresses, and the unspecified address MUST
1675                  *    NOT be included in a candidate set.
1676                  */
1677                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1678                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1679                         continue;
1680
1681                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1682
1683                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1684                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1685                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1686                                             idev->dev->name);
1687                         continue;
1688                 }
1689
1690                 score->rule = -1;
1691                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1692
1693                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1694                         int minihiscore, miniscore;
1695
1696                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1697                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1698
1699                         if (minihiscore > miniscore) {
1700                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1701                                     score->scopedist > 0) {
1702                                         /*
1703                                          * special case:
1704                                          * each remaining entry
1705                                          * has too small (not enough)
1706                                          * scope, because ifa entries
1707                                          * are sorted by their scope
1708                                          * values.
1709                                          */
1710                                         goto out;
1711                                 }
1712                                 break;
1713                         } else if (minihiscore < miniscore) {
1714                                 swap(hiscore, score);
1715                                 hiscore_idx = 1 - hiscore_idx;
1716
1717                                 /* restore our iterator */
1718                                 score->ifa = hiscore->ifa;
1719
1720                                 break;
1721                         }
1722                 }
1723         }
1724 out:
1725         return hiscore_idx;
1726 }
1727
1728 static int ipv6_get_saddr_master(struct net *net,
1729                                  const struct net_device *dst_dev,
1730                                  const struct net_device *master,
1731                                  struct ipv6_saddr_dst *dst,
1732                                  struct ipv6_saddr_score *scores,
1733                                  int hiscore_idx)
1734 {
1735         struct inet6_dev *idev;
1736
1737         idev = __in6_dev_get(dst_dev);
1738         if (idev)
1739                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1740                                                    scores, hiscore_idx);
1741
1742         idev = __in6_dev_get(master);
1743         if (idev)
1744                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1745                                                    scores, hiscore_idx);
1746
1747         return hiscore_idx;
1748 }
1749
1750 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1751                        const struct in6_addr *daddr, unsigned int prefs,
1752                        struct in6_addr *saddr)
1753 {
1754         struct ipv6_saddr_score scores[2], *hiscore;
1755         struct ipv6_saddr_dst dst;
1756         struct inet6_dev *idev;
1757         struct net_device *dev;
1758         int dst_type;
1759         bool use_oif_addr = false;
1760         int hiscore_idx = 0;
1761         int ret = 0;
1762
1763         dst_type = __ipv6_addr_type(daddr);
1764         dst.addr = daddr;
1765         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1766         dst.scope = __ipv6_addr_src_scope(dst_type);
1767         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1768         dst.prefs = prefs;
1769
1770         scores[hiscore_idx].rule = -1;
1771         scores[hiscore_idx].ifa = NULL;
1772
1773         rcu_read_lock();
1774
1775         /* Candidate Source Address (section 4)
1776          *  - multicast and link-local destination address,
1777          *    the set of candidate source address MUST only
1778          *    include addresses assigned to interfaces
1779          *    belonging to the same link as the outgoing
1780          *    interface.
1781          * (- For site-local destination addresses, the
1782          *    set of candidate source addresses MUST only
1783          *    include addresses assigned to interfaces
1784          *    belonging to the same site as the outgoing
1785          *    interface.)
1786          *  - "It is RECOMMENDED that the candidate source addresses
1787          *    be the set of unicast addresses assigned to the
1788          *    interface that will be used to send to the destination
1789          *    (the 'outgoing' interface)." (RFC 6724)
1790          */
1791         if (dst_dev) {
1792                 idev = __in6_dev_get(dst_dev);
1793                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1794                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1795                     (idev && idev->cnf.use_oif_addrs_only)) {
1796                         use_oif_addr = true;
1797                 }
1798         }
1799
1800         if (use_oif_addr) {
1801                 if (idev)
1802                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1803         } else {
1804                 const struct net_device *master;
1805                 int master_idx = 0;
1806
1807                 /* if dst_dev exists and is enslaved to an L3 device, then
1808                  * prefer addresses from dst_dev and then the master over
1809                  * any other enslaved devices in the L3 domain.
1810                  */
1811                 master = l3mdev_master_dev_rcu(dst_dev);
1812                 if (master) {
1813                         master_idx = master->ifindex;
1814
1815                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1816                                                             master, &dst,
1817                                                             scores, hiscore_idx);
1818
1819                         if (scores[hiscore_idx].ifa)
1820                                 goto out;
1821                 }
1822
1823                 for_each_netdev_rcu(net, dev) {
1824                         /* only consider addresses on devices in the
1825                          * same L3 domain
1826                          */
1827                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1828                                 continue;
1829                         idev = __in6_dev_get(dev);
1830                         if (!idev)
1831                                 continue;
1832                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1833                 }
1834         }
1835
1836 out:
1837         hiscore = &scores[hiscore_idx];
1838         if (!hiscore->ifa)
1839                 ret = -EADDRNOTAVAIL;
1840         else
1841                 *saddr = hiscore->ifa->addr;
1842
1843         rcu_read_unlock();
1844         return ret;
1845 }
1846 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1847
1848 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1849                               u32 banned_flags)
1850 {
1851         struct inet6_ifaddr *ifp;
1852         int err = -EADDRNOTAVAIL;
1853
1854         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1855                 if (ifp->scope > IFA_LINK)
1856                         break;
1857                 if (ifp->scope == IFA_LINK &&
1858                     !(ifp->flags & banned_flags)) {
1859                         *addr = ifp->addr;
1860                         err = 0;
1861                         break;
1862                 }
1863         }
1864         return err;
1865 }
1866
1867 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1868                     u32 banned_flags)
1869 {
1870         struct inet6_dev *idev;
1871         int err = -EADDRNOTAVAIL;
1872
1873         rcu_read_lock();
1874         idev = __in6_dev_get(dev);
1875         if (idev) {
1876                 read_lock_bh(&idev->lock);
1877                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1878                 read_unlock_bh(&idev->lock);
1879         }
1880         rcu_read_unlock();
1881         return err;
1882 }
1883
1884 static int ipv6_count_addresses(const struct inet6_dev *idev)
1885 {
1886         const struct inet6_ifaddr *ifp;
1887         int cnt = 0;
1888
1889         rcu_read_lock();
1890         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1891                 cnt++;
1892         rcu_read_unlock();
1893         return cnt;
1894 }
1895
1896 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1897                   const struct net_device *dev, int strict)
1898 {
1899         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1900                                        strict, IFA_F_TENTATIVE);
1901 }
1902 EXPORT_SYMBOL(ipv6_chk_addr);
1903
1904 /* device argument is used to find the L3 domain of interest. If
1905  * skip_dev_check is set, then the ifp device is not checked against
1906  * the passed in dev argument. So the 2 cases for addresses checks are:
1907  *   1. does the address exist in the L3 domain that dev is part of
1908  *      (skip_dev_check = true), or
1909  *
1910  *   2. does the address exist on the specific device
1911  *      (skip_dev_check = false)
1912  */
1913 static struct net_device *
1914 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1915                           const struct net_device *dev, bool skip_dev_check,
1916                           int strict, u32 banned_flags)
1917 {
1918         unsigned int hash = inet6_addr_hash(net, addr);
1919         struct net_device *l3mdev, *ndev;
1920         struct inet6_ifaddr *ifp;
1921         u32 ifp_flags;
1922
1923         rcu_read_lock();
1924
1925         l3mdev = l3mdev_master_dev_rcu(dev);
1926         if (skip_dev_check)
1927                 dev = NULL;
1928
1929         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1930                 ndev = ifp->idev->dev;
1931                 if (!net_eq(dev_net(ndev), net))
1932                         continue;
1933
1934                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1935                         continue;
1936
1937                 /* Decouple optimistic from tentative for evaluation here.
1938                  * Ban optimistic addresses explicitly, when required.
1939                  */
1940                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1941                             ? (ifp->flags&~IFA_F_TENTATIVE)
1942                             : ifp->flags;
1943                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1944                     !(ifp_flags&banned_flags) &&
1945                     (!dev || ndev == dev ||
1946                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1947                         rcu_read_unlock();
1948                         return ndev;
1949                 }
1950         }
1951
1952         rcu_read_unlock();
1953         return NULL;
1954 }
1955
1956 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1957                             const struct net_device *dev, bool skip_dev_check,
1958                             int strict, u32 banned_flags)
1959 {
1960         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1961                                          strict, banned_flags) ? 1 : 0;
1962 }
1963 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1964
1965
1966 /* Compares an address/prefix_len with addresses on device @dev.
1967  * If one is found it returns true.
1968  */
1969 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1970         const unsigned int prefix_len, struct net_device *dev)
1971 {
1972         const struct inet6_ifaddr *ifa;
1973         const struct inet6_dev *idev;
1974         bool ret = false;
1975
1976         rcu_read_lock();
1977         idev = __in6_dev_get(dev);
1978         if (idev) {
1979                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1980                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1981                         if (ret)
1982                                 break;
1983                 }
1984         }
1985         rcu_read_unlock();
1986
1987         return ret;
1988 }
1989 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1990
1991 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1992 {
1993         const struct inet6_ifaddr *ifa;
1994         const struct inet6_dev *idev;
1995         int     onlink;
1996
1997         onlink = 0;
1998         rcu_read_lock();
1999         idev = __in6_dev_get(dev);
2000         if (idev) {
2001                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2002                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
2003                                                    ifa->prefix_len);
2004                         if (onlink)
2005                                 break;
2006                 }
2007         }
2008         rcu_read_unlock();
2009         return onlink;
2010 }
2011 EXPORT_SYMBOL(ipv6_chk_prefix);
2012
2013 /**
2014  * ipv6_dev_find - find the first device with a given source address.
2015  * @net: the net namespace
2016  * @addr: the source address
2017  * @dev: used to find the L3 domain of interest
2018  *
2019  * The caller should be protected by RCU, or RTNL.
2020  */
2021 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2022                                  struct net_device *dev)
2023 {
2024         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2025                                          IFA_F_TENTATIVE);
2026 }
2027 EXPORT_SYMBOL(ipv6_dev_find);
2028
2029 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2030                                      struct net_device *dev, int strict)
2031 {
2032         unsigned int hash = inet6_addr_hash(net, addr);
2033         struct inet6_ifaddr *ifp, *result = NULL;
2034
2035         rcu_read_lock();
2036         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2037                 if (!net_eq(dev_net(ifp->idev->dev), net))
2038                         continue;
2039                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2040                         if (!dev || ifp->idev->dev == dev ||
2041                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2042                                 result = ifp;
2043                                 in6_ifa_hold(ifp);
2044                                 break;
2045                         }
2046                 }
2047         }
2048         rcu_read_unlock();
2049
2050         return result;
2051 }
2052
2053 /* Gets referenced address, destroys ifaddr */
2054
2055 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2056 {
2057         if (dad_failed)
2058                 ifp->flags |= IFA_F_DADFAILED;
2059
2060         if (ifp->flags&IFA_F_TEMPORARY) {
2061                 struct inet6_ifaddr *ifpub;
2062                 spin_lock_bh(&ifp->lock);
2063                 ifpub = ifp->ifpub;
2064                 if (ifpub) {
2065                         in6_ifa_hold(ifpub);
2066                         spin_unlock_bh(&ifp->lock);
2067                         ipv6_create_tempaddr(ifpub, true);
2068                         in6_ifa_put(ifpub);
2069                 } else {
2070                         spin_unlock_bh(&ifp->lock);
2071                 }
2072                 ipv6_del_addr(ifp);
2073         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2074                 spin_lock_bh(&ifp->lock);
2075                 addrconf_del_dad_work(ifp);
2076                 ifp->flags |= IFA_F_TENTATIVE;
2077                 if (dad_failed)
2078                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2079                 spin_unlock_bh(&ifp->lock);
2080                 if (dad_failed)
2081                         ipv6_ifa_notify(0, ifp);
2082                 in6_ifa_put(ifp);
2083         } else {
2084                 ipv6_del_addr(ifp);
2085         }
2086 }
2087
2088 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2089 {
2090         int err = -ENOENT;
2091
2092         spin_lock_bh(&ifp->lock);
2093         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2094                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2095                 err = 0;
2096         }
2097         spin_unlock_bh(&ifp->lock);
2098
2099         return err;
2100 }
2101
2102 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2103 {
2104         struct inet6_dev *idev = ifp->idev;
2105         struct net *net = dev_net(ifp->idev->dev);
2106
2107         if (addrconf_dad_end(ifp)) {
2108                 in6_ifa_put(ifp);
2109                 return;
2110         }
2111
2112         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2113                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2114
2115         spin_lock_bh(&ifp->lock);
2116
2117         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2118                 struct in6_addr new_addr;
2119                 struct inet6_ifaddr *ifp2;
2120                 int retries = ifp->stable_privacy_retry + 1;
2121                 struct ifa6_config cfg = {
2122                         .pfx = &new_addr,
2123                         .plen = ifp->prefix_len,
2124                         .ifa_flags = ifp->flags,
2125                         .valid_lft = ifp->valid_lft,
2126                         .preferred_lft = ifp->prefered_lft,
2127                         .scope = ifp->scope,
2128                 };
2129
2130                 if (retries > net->ipv6.sysctl.idgen_retries) {
2131                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2132                                              ifp->idev->dev->name);
2133                         goto errdad;
2134                 }
2135
2136                 new_addr = ifp->addr;
2137                 if (ipv6_generate_stable_address(&new_addr, retries,
2138                                                  idev))
2139                         goto errdad;
2140
2141                 spin_unlock_bh(&ifp->lock);
2142
2143                 if (idev->cnf.max_addresses &&
2144                     ipv6_count_addresses(idev) >=
2145                     idev->cnf.max_addresses)
2146                         goto lock_errdad;
2147
2148                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2149                                      ifp->idev->dev->name);
2150
2151                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2152                 if (IS_ERR(ifp2))
2153                         goto lock_errdad;
2154
2155                 spin_lock_bh(&ifp2->lock);
2156                 ifp2->stable_privacy_retry = retries;
2157                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2158                 spin_unlock_bh(&ifp2->lock);
2159
2160                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2161                 in6_ifa_put(ifp2);
2162 lock_errdad:
2163                 spin_lock_bh(&ifp->lock);
2164         }
2165
2166 errdad:
2167         /* transition from _POSTDAD to _ERRDAD */
2168         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2169         spin_unlock_bh(&ifp->lock);
2170
2171         addrconf_mod_dad_work(ifp, 0);
2172         in6_ifa_put(ifp);
2173 }
2174
2175 /* Join to solicited addr multicast group.
2176  * caller must hold RTNL */
2177 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2178 {
2179         struct in6_addr maddr;
2180
2181         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2182                 return;
2183
2184         addrconf_addr_solict_mult(addr, &maddr);
2185         ipv6_dev_mc_inc(dev, &maddr);
2186 }
2187
2188 /* caller must hold RTNL */
2189 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2190 {
2191         struct in6_addr maddr;
2192
2193         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2194                 return;
2195
2196         addrconf_addr_solict_mult(addr, &maddr);
2197         __ipv6_dev_mc_dec(idev, &maddr);
2198 }
2199
2200 /* caller must hold RTNL */
2201 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2202 {
2203         struct in6_addr addr;
2204
2205         if (ifp->prefix_len >= 127) /* RFC 6164 */
2206                 return;
2207         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2208         if (ipv6_addr_any(&addr))
2209                 return;
2210         __ipv6_dev_ac_inc(ifp->idev, &addr);
2211 }
2212
2213 /* caller must hold RTNL */
2214 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2215 {
2216         struct in6_addr addr;
2217
2218         if (ifp->prefix_len >= 127) /* RFC 6164 */
2219                 return;
2220         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2221         if (ipv6_addr_any(&addr))
2222                 return;
2223         __ipv6_dev_ac_dec(ifp->idev, &addr);
2224 }
2225
2226 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2227 {
2228         switch (dev->addr_len) {
2229         case ETH_ALEN:
2230                 memcpy(eui, dev->dev_addr, 3);
2231                 eui[3] = 0xFF;
2232                 eui[4] = 0xFE;
2233                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2234                 break;
2235         case EUI64_ADDR_LEN:
2236                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2237                 eui[0] ^= 2;
2238                 break;
2239         default:
2240                 return -1;
2241         }
2242
2243         return 0;
2244 }
2245
2246 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2247 {
2248         union fwnet_hwaddr *ha;
2249
2250         if (dev->addr_len != FWNET_ALEN)
2251                 return -1;
2252
2253         ha = (union fwnet_hwaddr *)dev->dev_addr;
2254
2255         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2256         eui[0] ^= 2;
2257         return 0;
2258 }
2259
2260 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2261 {
2262         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2263         if (dev->addr_len != ARCNET_ALEN)
2264                 return -1;
2265         memset(eui, 0, 7);
2266         eui[7] = *(u8 *)dev->dev_addr;
2267         return 0;
2268 }
2269
2270 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2271 {
2272         if (dev->addr_len != INFINIBAND_ALEN)
2273                 return -1;
2274         memcpy(eui, dev->dev_addr + 12, 8);
2275         eui[0] |= 2;
2276         return 0;
2277 }
2278
2279 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2280 {
2281         if (addr == 0)
2282                 return -1;
2283         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2284                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2285                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2286                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2287                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2288                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2289         eui[1] = 0;
2290         eui[2] = 0x5E;
2291         eui[3] = 0xFE;
2292         memcpy(eui + 4, &addr, 4);
2293         return 0;
2294 }
2295
2296 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2297 {
2298         if (dev->priv_flags & IFF_ISATAP)
2299                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2300         return -1;
2301 }
2302
2303 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2304 {
2305         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2306 }
2307
2308 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2309 {
2310         memcpy(eui, dev->perm_addr, 3);
2311         memcpy(eui + 5, dev->perm_addr + 3, 3);
2312         eui[3] = 0xFF;
2313         eui[4] = 0xFE;
2314         eui[0] ^= 2;
2315         return 0;
2316 }
2317
2318 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2319 {
2320         switch (dev->type) {
2321         case ARPHRD_ETHER:
2322         case ARPHRD_FDDI:
2323                 return addrconf_ifid_eui48(eui, dev);
2324         case ARPHRD_ARCNET:
2325                 return addrconf_ifid_arcnet(eui, dev);
2326         case ARPHRD_INFINIBAND:
2327                 return addrconf_ifid_infiniband(eui, dev);
2328         case ARPHRD_SIT:
2329                 return addrconf_ifid_sit(eui, dev);
2330         case ARPHRD_IPGRE:
2331         case ARPHRD_TUNNEL:
2332                 return addrconf_ifid_gre(eui, dev);
2333         case ARPHRD_6LOWPAN:
2334                 return addrconf_ifid_6lowpan(eui, dev);
2335         case ARPHRD_IEEE1394:
2336                 return addrconf_ifid_ieee1394(eui, dev);
2337         case ARPHRD_TUNNEL6:
2338         case ARPHRD_IP6GRE:
2339         case ARPHRD_RAWIP:
2340                 return addrconf_ifid_ip6tnl(eui, dev);
2341         }
2342         return -1;
2343 }
2344
2345 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2346 {
2347         int err = -1;
2348         struct inet6_ifaddr *ifp;
2349
2350         read_lock_bh(&idev->lock);
2351         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2352                 if (ifp->scope > IFA_LINK)
2353                         break;
2354                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2355                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2356                         err = 0;
2357                         break;
2358                 }
2359         }
2360         read_unlock_bh(&idev->lock);
2361         return err;
2362 }
2363
2364 /* Generation of a randomized Interface Identifier
2365  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2366  */
2367
2368 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2369 {
2370 regen:
2371         get_random_bytes(&addr->s6_addr[8], 8);
2372
2373         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2374          * check if generated address is not inappropriate:
2375          *
2376          * - Reserved IPv6 Interface Identifiers
2377          * - XXX: already assigned to an address on the device
2378          */
2379
2380         /* Subnet-router anycast: 0000:0000:0000:0000 */
2381         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2382                 goto regen;
2383
2384         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2385          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2386          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2387          */
2388         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2389             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2390                 goto regen;
2391
2392         /* Reserved subnet anycast addresses */
2393         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2394             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2395                 goto regen;
2396 }
2397
2398 /*
2399  *      Add prefix route.
2400  */
2401
2402 static void
2403 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2404                       struct net_device *dev, unsigned long expires,
2405                       u32 flags, gfp_t gfp_flags)
2406 {
2407         struct fib6_config cfg = {
2408                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2409                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2410                 .fc_ifindex = dev->ifindex,
2411                 .fc_expires = expires,
2412                 .fc_dst_len = plen,
2413                 .fc_flags = RTF_UP | flags,
2414                 .fc_nlinfo.nl_net = dev_net(dev),
2415                 .fc_protocol = RTPROT_KERNEL,
2416                 .fc_type = RTN_UNICAST,
2417         };
2418
2419         cfg.fc_dst = *pfx;
2420
2421         /* Prevent useless cloning on PtP SIT.
2422            This thing is done here expecting that the whole
2423            class of non-broadcast devices need not cloning.
2424          */
2425 #if IS_ENABLED(CONFIG_IPV6_SIT)
2426         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2427                 cfg.fc_flags |= RTF_NONEXTHOP;
2428 #endif
2429
2430         ip6_route_add(&cfg, gfp_flags, NULL);
2431 }
2432
2433
2434 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2435                                                   int plen,
2436                                                   const struct net_device *dev,
2437                                                   u32 flags, u32 noflags,
2438                                                   bool no_gw)
2439 {
2440         struct fib6_node *fn;
2441         struct fib6_info *rt = NULL;
2442         struct fib6_table *table;
2443         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2444
2445         table = fib6_get_table(dev_net(dev), tb_id);
2446         if (!table)
2447                 return NULL;
2448
2449         rcu_read_lock();
2450         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2451         if (!fn)
2452                 goto out;
2453
2454         for_each_fib6_node_rt_rcu(fn) {
2455                 /* prefix routes only use builtin fib6_nh */
2456                 if (rt->nh)
2457                         continue;
2458
2459                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2460                         continue;
2461                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2462                         continue;
2463                 if ((rt->fib6_flags & flags) != flags)
2464                         continue;
2465                 if ((rt->fib6_flags & noflags) != 0)
2466                         continue;
2467                 if (!fib6_info_hold_safe(rt))
2468                         continue;
2469                 break;
2470         }
2471 out:
2472         rcu_read_unlock();
2473         return rt;
2474 }
2475
2476
2477 /* Create "default" multicast route to the interface */
2478
2479 static void addrconf_add_mroute(struct net_device *dev)
2480 {
2481         struct fib6_config cfg = {
2482                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2483                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2484                 .fc_ifindex = dev->ifindex,
2485                 .fc_dst_len = 8,
2486                 .fc_flags = RTF_UP,
2487                 .fc_type = RTN_MULTICAST,
2488                 .fc_nlinfo.nl_net = dev_net(dev),
2489                 .fc_protocol = RTPROT_KERNEL,
2490         };
2491
2492         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2493
2494         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2495 }
2496
2497 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2498 {
2499         struct inet6_dev *idev;
2500
2501         ASSERT_RTNL();
2502
2503         idev = ipv6_find_idev(dev);
2504         if (IS_ERR(idev))
2505                 return idev;
2506
2507         if (idev->cnf.disable_ipv6)
2508                 return ERR_PTR(-EACCES);
2509
2510         /* Add default multicast route */
2511         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2512                 addrconf_add_mroute(dev);
2513
2514         return idev;
2515 }
2516
2517 static void manage_tempaddrs(struct inet6_dev *idev,
2518                              struct inet6_ifaddr *ifp,
2519                              __u32 valid_lft, __u32 prefered_lft,
2520                              bool create, unsigned long now)
2521 {
2522         u32 flags;
2523         struct inet6_ifaddr *ift;
2524
2525         read_lock_bh(&idev->lock);
2526         /* update all temporary addresses in the list */
2527         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2528                 int age, max_valid, max_prefered;
2529
2530                 if (ifp != ift->ifpub)
2531                         continue;
2532
2533                 /* RFC 4941 section 3.3:
2534                  * If a received option will extend the lifetime of a public
2535                  * address, the lifetimes of temporary addresses should
2536                  * be extended, subject to the overall constraint that no
2537                  * temporary addresses should ever remain "valid" or "preferred"
2538                  * for a time longer than (TEMP_VALID_LIFETIME) or
2539                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2540                  */
2541                 age = (now - ift->cstamp) / HZ;
2542                 max_valid = idev->cnf.temp_valid_lft - age;
2543                 if (max_valid < 0)
2544                         max_valid = 0;
2545
2546                 max_prefered = idev->cnf.temp_prefered_lft -
2547                                idev->desync_factor - age;
2548                 if (max_prefered < 0)
2549                         max_prefered = 0;
2550
2551                 if (valid_lft > max_valid)
2552                         valid_lft = max_valid;
2553
2554                 if (prefered_lft > max_prefered)
2555                         prefered_lft = max_prefered;
2556
2557                 spin_lock(&ift->lock);
2558                 flags = ift->flags;
2559                 ift->valid_lft = valid_lft;
2560                 ift->prefered_lft = prefered_lft;
2561                 ift->tstamp = now;
2562                 if (prefered_lft > 0)
2563                         ift->flags &= ~IFA_F_DEPRECATED;
2564
2565                 spin_unlock(&ift->lock);
2566                 if (!(flags&IFA_F_TENTATIVE))
2567                         ipv6_ifa_notify(0, ift);
2568         }
2569
2570         /* Also create a temporary address if it's enabled but no temporary
2571          * address currently exists.
2572          * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2573          * as part of cleanup (ie. deleting the mngtmpaddr).
2574          * We don't want that to result in creating a new temporary ip address.
2575          */
2576         if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2577                 create = true;
2578
2579         if (create && idev->cnf.use_tempaddr > 0) {
2580                 /* When a new public address is created as described
2581                  * in [ADDRCONF], also create a new temporary address.
2582                  */
2583                 read_unlock_bh(&idev->lock);
2584                 ipv6_create_tempaddr(ifp, false);
2585         } else {
2586                 read_unlock_bh(&idev->lock);
2587         }
2588 }
2589
2590 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2591 {
2592         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2593                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2594 }
2595
2596 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2597                                  const struct prefix_info *pinfo,
2598                                  struct inet6_dev *in6_dev,
2599                                  const struct in6_addr *addr, int addr_type,
2600                                  u32 addr_flags, bool sllao, bool tokenized,
2601                                  __u32 valid_lft, u32 prefered_lft)
2602 {
2603         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2604         int create = 0, update_lft = 0;
2605
2606         if (!ifp && valid_lft) {
2607                 int max_addresses = in6_dev->cnf.max_addresses;
2608                 struct ifa6_config cfg = {
2609                         .pfx = addr,
2610                         .plen = pinfo->prefix_len,
2611                         .ifa_flags = addr_flags,
2612                         .valid_lft = valid_lft,
2613                         .preferred_lft = prefered_lft,
2614                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2615                 };
2616
2617 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2618                 if ((net->ipv6.devconf_all->optimistic_dad ||
2619                      in6_dev->cnf.optimistic_dad) &&
2620                     !net->ipv6.devconf_all->forwarding && sllao)
2621                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2622 #endif
2623
2624                 /* Do not allow to create too much of autoconfigured
2625                  * addresses; this would be too easy way to crash kernel.
2626                  */
2627                 if (!max_addresses ||
2628                     ipv6_count_addresses(in6_dev) < max_addresses)
2629                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2630
2631                 if (IS_ERR_OR_NULL(ifp))
2632                         return -1;
2633
2634                 create = 1;
2635                 spin_lock_bh(&ifp->lock);
2636                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2637                 ifp->cstamp = jiffies;
2638                 ifp->tokenized = tokenized;
2639                 spin_unlock_bh(&ifp->lock);
2640                 addrconf_dad_start(ifp);
2641         }
2642
2643         if (ifp) {
2644                 u32 flags;
2645                 unsigned long now;
2646                 u32 stored_lft;
2647
2648                 /* update lifetime (RFC2462 5.5.3 e) */
2649                 spin_lock_bh(&ifp->lock);
2650                 now = jiffies;
2651                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2652                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2653                 else
2654                         stored_lft = 0;
2655                 if (!create && stored_lft) {
2656                         const u32 minimum_lft = min_t(u32,
2657                                 stored_lft, MIN_VALID_LIFETIME);
2658                         valid_lft = max(valid_lft, minimum_lft);
2659
2660                         /* RFC4862 Section 5.5.3e:
2661                          * "Note that the preferred lifetime of the
2662                          *  corresponding address is always reset to
2663                          *  the Preferred Lifetime in the received
2664                          *  Prefix Information option, regardless of
2665                          *  whether the valid lifetime is also reset or
2666                          *  ignored."
2667                          *
2668                          * So we should always update prefered_lft here.
2669                          */
2670                         update_lft = 1;
2671                 }
2672
2673                 if (update_lft) {
2674                         ifp->valid_lft = valid_lft;
2675                         ifp->prefered_lft = prefered_lft;
2676                         ifp->tstamp = now;
2677                         flags = ifp->flags;
2678                         ifp->flags &= ~IFA_F_DEPRECATED;
2679                         spin_unlock_bh(&ifp->lock);
2680
2681                         if (!(flags&IFA_F_TENTATIVE))
2682                                 ipv6_ifa_notify(0, ifp);
2683                 } else
2684                         spin_unlock_bh(&ifp->lock);
2685
2686                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2687                                  create, now);
2688
2689                 in6_ifa_put(ifp);
2690                 addrconf_verify();
2691         }
2692
2693         return 0;
2694 }
2695 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2696
2697 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2698 {
2699         struct prefix_info *pinfo;
2700         __u32 valid_lft;
2701         __u32 prefered_lft;
2702         int addr_type, err;
2703         u32 addr_flags = 0;
2704         struct inet6_dev *in6_dev;
2705         struct net *net = dev_net(dev);
2706
2707         pinfo = (struct prefix_info *) opt;
2708
2709         if (len < sizeof(struct prefix_info)) {
2710                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2711                 return;
2712         }
2713
2714         /*
2715          *      Validation checks ([ADDRCONF], page 19)
2716          */
2717
2718         addr_type = ipv6_addr_type(&pinfo->prefix);
2719
2720         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2721                 return;
2722
2723         valid_lft = ntohl(pinfo->valid);
2724         prefered_lft = ntohl(pinfo->prefered);
2725
2726         if (prefered_lft > valid_lft) {
2727                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2728                 return;
2729         }
2730
2731         in6_dev = in6_dev_get(dev);
2732
2733         if (!in6_dev) {
2734                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2735                                     dev->name);
2736                 return;
2737         }
2738
2739         if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2740                 goto put;
2741
2742         /*
2743          *      Two things going on here:
2744          *      1) Add routes for on-link prefixes
2745          *      2) Configure prefixes with the auto flag set
2746          */
2747
2748         if (pinfo->onlink) {
2749                 struct fib6_info *rt;
2750                 unsigned long rt_expires;
2751
2752                 /* Avoid arithmetic overflow. Really, we could
2753                  * save rt_expires in seconds, likely valid_lft,
2754                  * but it would require division in fib gc, that it
2755                  * not good.
2756                  */
2757                 if (HZ > USER_HZ)
2758                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2759                 else
2760                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2761
2762                 if (addrconf_finite_timeout(rt_expires))
2763                         rt_expires *= HZ;
2764
2765                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2766                                                pinfo->prefix_len,
2767                                                dev,
2768                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2769                                                RTF_DEFAULT, true);
2770
2771                 if (rt) {
2772                         /* Autoconf prefix route */
2773                         if (valid_lft == 0) {
2774                                 ip6_del_rt(net, rt, false);
2775                                 rt = NULL;
2776                         } else if (addrconf_finite_timeout(rt_expires)) {
2777                                 /* not infinity */
2778                                 fib6_set_expires(rt, jiffies + rt_expires);
2779                         } else {
2780                                 fib6_clean_expires(rt);
2781                         }
2782                 } else if (valid_lft) {
2783                         clock_t expires = 0;
2784                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2785                         if (addrconf_finite_timeout(rt_expires)) {
2786                                 /* not infinity */
2787                                 flags |= RTF_EXPIRES;
2788                                 expires = jiffies_to_clock_t(rt_expires);
2789                         }
2790                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2791                                               0, dev, expires, flags,
2792                                               GFP_ATOMIC);
2793                 }
2794                 fib6_info_release(rt);
2795         }
2796
2797         /* Try to figure out our local address for this prefix */
2798
2799         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2800                 struct in6_addr addr;
2801                 bool tokenized = false, dev_addr_generated = false;
2802
2803                 if (pinfo->prefix_len == 64) {
2804                         memcpy(&addr, &pinfo->prefix, 8);
2805
2806                         if (!ipv6_addr_any(&in6_dev->token)) {
2807                                 read_lock_bh(&in6_dev->lock);
2808                                 memcpy(addr.s6_addr + 8,
2809                                        in6_dev->token.s6_addr + 8, 8);
2810                                 read_unlock_bh(&in6_dev->lock);
2811                                 tokenized = true;
2812                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2813                                    !ipv6_generate_stable_address(&addr, 0,
2814                                                                  in6_dev)) {
2815                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2816                                 goto ok;
2817                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2818                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2819                                 goto put;
2820                         } else {
2821                                 dev_addr_generated = true;
2822                         }
2823                         goto ok;
2824                 }
2825                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2826                                     pinfo->prefix_len);
2827                 goto put;
2828
2829 ok:
2830                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2831                                                    &addr, addr_type,
2832                                                    addr_flags, sllao,
2833                                                    tokenized, valid_lft,
2834                                                    prefered_lft);
2835                 if (err)
2836                         goto put;
2837
2838                 /* Ignore error case here because previous prefix add addr was
2839                  * successful which will be notified.
2840                  */
2841                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2842                                               addr_type, addr_flags, sllao,
2843                                               tokenized, valid_lft,
2844                                               prefered_lft,
2845                                               dev_addr_generated);
2846         }
2847         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2848 put:
2849         in6_dev_put(in6_dev);
2850 }
2851
2852 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2853                 struct in6_ifreq *ireq)
2854 {
2855         struct ip_tunnel_parm p = { };
2856         int err;
2857
2858         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2859                 return -EADDRNOTAVAIL;
2860
2861         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2862         p.iph.version = 4;
2863         p.iph.ihl = 5;
2864         p.iph.protocol = IPPROTO_IPV6;
2865         p.iph.ttl = 64;
2866
2867         if (!dev->netdev_ops->ndo_tunnel_ctl)
2868                 return -EOPNOTSUPP;
2869         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2870         if (err)
2871                 return err;
2872
2873         dev = __dev_get_by_name(net, p.name);
2874         if (!dev)
2875                 return -ENOBUFS;
2876         return dev_open(dev, NULL);
2877 }
2878
2879 /*
2880  *      Set destination address.
2881  *      Special case for SIT interfaces where we create a new "virtual"
2882  *      device.
2883  */
2884 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2885 {
2886         struct net_device *dev;
2887         struct in6_ifreq ireq;
2888         int err = -ENODEV;
2889
2890         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2891                 return -ENODEV;
2892         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2893                 return -EFAULT;
2894
2895         rtnl_lock();
2896         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2897         if (dev && dev->type == ARPHRD_SIT)
2898                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2899         rtnl_unlock();
2900         return err;
2901 }
2902
2903 static int ipv6_mc_config(struct sock *sk, bool join,
2904                           const struct in6_addr *addr, int ifindex)
2905 {
2906         int ret;
2907
2908         ASSERT_RTNL();
2909
2910         lock_sock(sk);
2911         if (join)
2912                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2913         else
2914                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2915         release_sock(sk);
2916
2917         return ret;
2918 }
2919
2920 /*
2921  *      Manual configuration of address on an interface
2922  */
2923 static int inet6_addr_add(struct net *net, int ifindex,
2924                           struct ifa6_config *cfg,
2925                           struct netlink_ext_ack *extack)
2926 {
2927         struct inet6_ifaddr *ifp;
2928         struct inet6_dev *idev;
2929         struct net_device *dev;
2930         unsigned long timeout;
2931         clock_t expires;
2932         u32 flags;
2933
2934         ASSERT_RTNL();
2935
2936         if (cfg->plen > 128)
2937                 return -EINVAL;
2938
2939         /* check the lifetime */
2940         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2941                 return -EINVAL;
2942
2943         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2944                 return -EINVAL;
2945
2946         dev = __dev_get_by_index(net, ifindex);
2947         if (!dev)
2948                 return -ENODEV;
2949
2950         idev = addrconf_add_dev(dev);
2951         if (IS_ERR(idev))
2952                 return PTR_ERR(idev);
2953
2954         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2955                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2956                                          true, cfg->pfx, ifindex);
2957
2958                 if (ret < 0)
2959                         return ret;
2960         }
2961
2962         cfg->scope = ipv6_addr_scope(cfg->pfx);
2963
2964         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2965         if (addrconf_finite_timeout(timeout)) {
2966                 expires = jiffies_to_clock_t(timeout * HZ);
2967                 cfg->valid_lft = timeout;
2968                 flags = RTF_EXPIRES;
2969         } else {
2970                 expires = 0;
2971                 flags = 0;
2972                 cfg->ifa_flags |= IFA_F_PERMANENT;
2973         }
2974
2975         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2976         if (addrconf_finite_timeout(timeout)) {
2977                 if (timeout == 0)
2978                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2979                 cfg->preferred_lft = timeout;
2980         }
2981
2982         ifp = ipv6_add_addr(idev, cfg, true, extack);
2983         if (!IS_ERR(ifp)) {
2984                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2985                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2986                                               ifp->rt_priority, dev, expires,
2987                                               flags, GFP_KERNEL);
2988                 }
2989
2990                 /* Send a netlink notification if DAD is enabled and
2991                  * optimistic flag is not set
2992                  */
2993                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2994                         ipv6_ifa_notify(0, ifp);
2995                 /*
2996                  * Note that section 3.1 of RFC 4429 indicates
2997                  * that the Optimistic flag should not be set for
2998                  * manually configured addresses
2999                  */
3000                 addrconf_dad_start(ifp);
3001                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3002                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
3003                                          cfg->preferred_lft, true, jiffies);
3004                 in6_ifa_put(ifp);
3005                 addrconf_verify_rtnl();
3006                 return 0;
3007         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3008                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3009                                cfg->pfx, ifindex);
3010         }
3011
3012         return PTR_ERR(ifp);
3013 }
3014
3015 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3016                           const struct in6_addr *pfx, unsigned int plen)
3017 {
3018         struct inet6_ifaddr *ifp;
3019         struct inet6_dev *idev;
3020         struct net_device *dev;
3021
3022         if (plen > 128)
3023                 return -EINVAL;
3024
3025         dev = __dev_get_by_index(net, ifindex);
3026         if (!dev)
3027                 return -ENODEV;
3028
3029         idev = __in6_dev_get(dev);
3030         if (!idev)
3031                 return -ENXIO;
3032
3033         read_lock_bh(&idev->lock);
3034         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3035                 if (ifp->prefix_len == plen &&
3036                     ipv6_addr_equal(pfx, &ifp->addr)) {
3037                         in6_ifa_hold(ifp);
3038                         read_unlock_bh(&idev->lock);
3039
3040                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3041                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3042                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3043                                                  jiffies);
3044                         ipv6_del_addr(ifp);
3045                         addrconf_verify_rtnl();
3046                         if (ipv6_addr_is_multicast(pfx)) {
3047                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3048                                                false, pfx, dev->ifindex);
3049                         }
3050                         return 0;
3051                 }
3052         }
3053         read_unlock_bh(&idev->lock);
3054         return -EADDRNOTAVAIL;
3055 }
3056
3057
3058 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3059 {
3060         struct ifa6_config cfg = {
3061                 .ifa_flags = IFA_F_PERMANENT,
3062                 .preferred_lft = INFINITY_LIFE_TIME,
3063                 .valid_lft = INFINITY_LIFE_TIME,
3064         };
3065         struct in6_ifreq ireq;
3066         int err;
3067
3068         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3069                 return -EPERM;
3070
3071         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3072                 return -EFAULT;
3073
3074         cfg.pfx = &ireq.ifr6_addr;
3075         cfg.plen = ireq.ifr6_prefixlen;
3076
3077         rtnl_lock();
3078         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3079         rtnl_unlock();
3080         return err;
3081 }
3082
3083 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3084 {
3085         struct in6_ifreq ireq;
3086         int err;
3087
3088         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3089                 return -EPERM;
3090
3091         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3092                 return -EFAULT;
3093
3094         rtnl_lock();
3095         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3096                              ireq.ifr6_prefixlen);
3097         rtnl_unlock();
3098         return err;
3099 }
3100
3101 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3102                      int plen, int scope)
3103 {
3104         struct inet6_ifaddr *ifp;
3105         struct ifa6_config cfg = {
3106                 .pfx = addr,
3107                 .plen = plen,
3108                 .ifa_flags = IFA_F_PERMANENT,
3109                 .valid_lft = INFINITY_LIFE_TIME,
3110                 .preferred_lft = INFINITY_LIFE_TIME,
3111                 .scope = scope
3112         };
3113
3114         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3115         if (!IS_ERR(ifp)) {
3116                 spin_lock_bh(&ifp->lock);
3117                 ifp->flags &= ~IFA_F_TENTATIVE;
3118                 spin_unlock_bh(&ifp->lock);
3119                 rt_genid_bump_ipv6(dev_net(idev->dev));
3120                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3121                 in6_ifa_put(ifp);
3122         }
3123 }
3124
3125 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3126 static void add_v4_addrs(struct inet6_dev *idev)
3127 {
3128         struct in6_addr addr;
3129         struct net_device *dev;
3130         struct net *net = dev_net(idev->dev);
3131         int scope, plen, offset = 0;
3132         u32 pflags = 0;
3133
3134         ASSERT_RTNL();
3135
3136         memset(&addr, 0, sizeof(struct in6_addr));
3137         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3138         if (idev->dev->addr_len == sizeof(struct in6_addr))
3139                 offset = sizeof(struct in6_addr) - 4;
3140         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3141
3142         if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3143                 scope = IPV6_ADDR_COMPATv4;
3144                 plen = 96;
3145                 pflags |= RTF_NONEXTHOP;
3146         } else {
3147                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3148                         return;
3149
3150                 addr.s6_addr32[0] = htonl(0xfe800000);
3151                 scope = IFA_LINK;
3152                 plen = 64;
3153         }
3154
3155         if (addr.s6_addr32[3]) {
3156                 add_addr(idev, &addr, plen, scope);
3157                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3158                                       GFP_KERNEL);
3159                 return;
3160         }
3161
3162         for_each_netdev(net, dev) {
3163                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3164                 if (in_dev && (dev->flags & IFF_UP)) {
3165                         struct in_ifaddr *ifa;
3166                         int flag = scope;
3167
3168                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3169                                 addr.s6_addr32[3] = ifa->ifa_local;
3170
3171                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3172                                         continue;
3173                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3174                                         if (idev->dev->flags&IFF_POINTOPOINT)
3175                                                 continue;
3176                                         flag |= IFA_HOST;
3177                                 }
3178
3179                                 add_addr(idev, &addr, plen, flag);
3180                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3181                                                       0, pflags, GFP_KERNEL);
3182                         }
3183                 }
3184         }
3185 }
3186 #endif
3187
3188 static void init_loopback(struct net_device *dev)
3189 {
3190         struct inet6_dev  *idev;
3191
3192         /* ::1 */
3193
3194         ASSERT_RTNL();
3195
3196         idev = ipv6_find_idev(dev);
3197         if (IS_ERR(idev)) {
3198                 pr_debug("%s: add_dev failed\n", __func__);
3199                 return;
3200         }
3201
3202         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3203 }
3204
3205 void addrconf_add_linklocal(struct inet6_dev *idev,
3206                             const struct in6_addr *addr, u32 flags)
3207 {
3208         struct ifa6_config cfg = {
3209                 .pfx = addr,
3210                 .plen = 64,
3211                 .ifa_flags = flags | IFA_F_PERMANENT,
3212                 .valid_lft = INFINITY_LIFE_TIME,
3213                 .preferred_lft = INFINITY_LIFE_TIME,
3214                 .scope = IFA_LINK
3215         };
3216         struct inet6_ifaddr *ifp;
3217
3218 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3219         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3220              idev->cnf.optimistic_dad) &&
3221             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3222                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3223 #endif
3224
3225         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3226         if (!IS_ERR(ifp)) {
3227                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3228                                       0, 0, GFP_ATOMIC);
3229                 addrconf_dad_start(ifp);
3230                 in6_ifa_put(ifp);
3231         }
3232 }
3233 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3234
3235 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3236 {
3237         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3238                 return true;
3239
3240         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3241             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3242                 return true;
3243
3244         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3245             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3246                 return true;
3247
3248         return false;
3249 }
3250
3251 static int ipv6_generate_stable_address(struct in6_addr *address,
3252                                         u8 dad_count,
3253                                         const struct inet6_dev *idev)
3254 {
3255         static DEFINE_SPINLOCK(lock);
3256         static __u32 digest[SHA1_DIGEST_WORDS];
3257         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3258
3259         static union {
3260                 char __data[SHA1_BLOCK_SIZE];
3261                 struct {
3262                         struct in6_addr secret;
3263                         __be32 prefix[2];
3264                         unsigned char hwaddr[MAX_ADDR_LEN];
3265                         u8 dad_count;
3266                 } __packed;
3267         } data;
3268
3269         struct in6_addr secret;
3270         struct in6_addr temp;
3271         struct net *net = dev_net(idev->dev);
3272
3273         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3274
3275         if (idev->cnf.stable_secret.initialized)
3276                 secret = idev->cnf.stable_secret.secret;
3277         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3278                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3279         else
3280                 return -1;
3281
3282 retry:
3283         spin_lock_bh(&lock);
3284
3285         sha1_init(digest);
3286         memset(&data, 0, sizeof(data));
3287         memset(workspace, 0, sizeof(workspace));
3288         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3289         data.prefix[0] = address->s6_addr32[0];
3290         data.prefix[1] = address->s6_addr32[1];
3291         data.secret = secret;
3292         data.dad_count = dad_count;
3293
3294         sha1_transform(digest, data.__data, workspace);
3295
3296         temp = *address;
3297         temp.s6_addr32[2] = (__force __be32)digest[0];
3298         temp.s6_addr32[3] = (__force __be32)digest[1];
3299
3300         spin_unlock_bh(&lock);
3301
3302         if (ipv6_reserved_interfaceid(temp)) {
3303                 dad_count++;
3304                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3305                         return -1;
3306                 goto retry;
3307         }
3308
3309         *address = temp;
3310         return 0;
3311 }
3312
3313 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3314 {
3315         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3316
3317         if (s->initialized)
3318                 return;
3319         s = &idev->cnf.stable_secret;
3320         get_random_bytes(&s->secret, sizeof(s->secret));
3321         s->initialized = true;
3322 }
3323
3324 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3325 {
3326         struct in6_addr addr;
3327
3328         /* no link local addresses on L3 master devices */
3329         if (netif_is_l3_master(idev->dev))
3330                 return;
3331
3332         /* no link local addresses on devices flagged as slaves */
3333         if (idev->dev->flags & IFF_SLAVE)
3334                 return;
3335
3336         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3337
3338         switch (idev->cnf.addr_gen_mode) {
3339         case IN6_ADDR_GEN_MODE_RANDOM:
3340                 ipv6_gen_mode_random_init(idev);
3341                 fallthrough;
3342         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3343                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3344                         addrconf_add_linklocal(idev, &addr,
3345                                                IFA_F_STABLE_PRIVACY);
3346                 else if (prefix_route)
3347                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3348                                               0, 0, GFP_KERNEL);
3349                 break;
3350         case IN6_ADDR_GEN_MODE_EUI64:
3351                 /* addrconf_add_linklocal also adds a prefix_route and we
3352                  * only need to care about prefix routes if ipv6_generate_eui64
3353                  * couldn't generate one.
3354                  */
3355                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3356                         addrconf_add_linklocal(idev, &addr, 0);
3357                 else if (prefix_route)
3358                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3359                                               0, 0, GFP_KERNEL);
3360                 break;
3361         case IN6_ADDR_GEN_MODE_NONE:
3362         default:
3363                 /* will not add any link local address */
3364                 break;
3365         }
3366 }
3367
3368 static void addrconf_dev_config(struct net_device *dev)
3369 {
3370         struct inet6_dev *idev;
3371
3372         ASSERT_RTNL();
3373
3374         if ((dev->type != ARPHRD_ETHER) &&
3375             (dev->type != ARPHRD_FDDI) &&
3376             (dev->type != ARPHRD_ARCNET) &&
3377             (dev->type != ARPHRD_INFINIBAND) &&
3378             (dev->type != ARPHRD_IEEE1394) &&
3379             (dev->type != ARPHRD_TUNNEL6) &&
3380             (dev->type != ARPHRD_6LOWPAN) &&
3381             (dev->type != ARPHRD_TUNNEL) &&
3382             (dev->type != ARPHRD_NONE) &&
3383             (dev->type != ARPHRD_RAWIP)) {
3384                 /* Alas, we support only Ethernet autoconfiguration. */
3385                 idev = __in6_dev_get(dev);
3386                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3387                     dev->flags & IFF_MULTICAST)
3388                         ipv6_mc_up(idev);
3389                 return;
3390         }
3391
3392         idev = addrconf_add_dev(dev);
3393         if (IS_ERR(idev))
3394                 return;
3395
3396         /* this device type has no EUI support */
3397         if (dev->type == ARPHRD_NONE &&
3398             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3399                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3400
3401         addrconf_addr_gen(idev, false);
3402 }
3403
3404 #if IS_ENABLED(CONFIG_IPV6_SIT)
3405 static void addrconf_sit_config(struct net_device *dev)
3406 {
3407         struct inet6_dev *idev;
3408
3409         ASSERT_RTNL();
3410
3411         /*
3412          * Configure the tunnel with one of our IPv4
3413          * addresses... we should configure all of
3414          * our v4 addrs in the tunnel
3415          */
3416
3417         idev = ipv6_find_idev(dev);
3418         if (IS_ERR(idev)) {
3419                 pr_debug("%s: add_dev failed\n", __func__);
3420                 return;
3421         }
3422
3423         if (dev->priv_flags & IFF_ISATAP) {
3424                 addrconf_addr_gen(idev, false);
3425                 return;
3426         }
3427
3428         add_v4_addrs(idev);
3429
3430         if (dev->flags&IFF_POINTOPOINT)
3431                 addrconf_add_mroute(dev);
3432 }
3433 #endif
3434
3435 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3436 static void addrconf_gre_config(struct net_device *dev)
3437 {
3438         struct inet6_dev *idev;
3439
3440         ASSERT_RTNL();
3441
3442         idev = ipv6_find_idev(dev);
3443         if (IS_ERR(idev)) {
3444                 pr_debug("%s: add_dev failed\n", __func__);
3445                 return;
3446         }
3447
3448         if (dev->type == ARPHRD_ETHER) {
3449                 addrconf_addr_gen(idev, true);
3450                 return;
3451         }
3452
3453         add_v4_addrs(idev);
3454
3455         if (dev->flags & IFF_POINTOPOINT)
3456                 addrconf_add_mroute(dev);
3457 }
3458 #endif
3459
3460 static void addrconf_init_auto_addrs(struct net_device *dev)
3461 {
3462         switch (dev->type) {
3463 #if IS_ENABLED(CONFIG_IPV6_SIT)
3464         case ARPHRD_SIT:
3465                 addrconf_sit_config(dev);
3466                 break;
3467 #endif
3468 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3469         case ARPHRD_IP6GRE:
3470         case ARPHRD_IPGRE:
3471                 addrconf_gre_config(dev);
3472                 break;
3473 #endif
3474         case ARPHRD_LOOPBACK:
3475                 init_loopback(dev);
3476                 break;
3477
3478         default:
3479                 addrconf_dev_config(dev);
3480                 break;
3481         }
3482 }
3483
3484 static int fixup_permanent_addr(struct net *net,
3485                                 struct inet6_dev *idev,
3486                                 struct inet6_ifaddr *ifp)
3487 {
3488         /* !fib6_node means the host route was removed from the
3489          * FIB, for example, if 'lo' device is taken down. In that
3490          * case regenerate the host route.
3491          */
3492         if (!ifp->rt || !ifp->rt->fib6_node) {
3493                 struct fib6_info *f6i, *prev;
3494
3495                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3496                                          GFP_ATOMIC);
3497                 if (IS_ERR(f6i))
3498                         return PTR_ERR(f6i);
3499
3500                 /* ifp->rt can be accessed outside of rtnl */
3501                 spin_lock(&ifp->lock);
3502                 prev = ifp->rt;
3503                 ifp->rt = f6i;
3504                 spin_unlock(&ifp->lock);
3505
3506                 fib6_info_release(prev);
3507         }
3508
3509         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3510                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3511                                       ifp->rt_priority, idev->dev, 0, 0,
3512                                       GFP_ATOMIC);
3513         }
3514
3515         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3516                 addrconf_dad_start(ifp);
3517
3518         return 0;
3519 }
3520
3521 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3522 {
3523         struct inet6_ifaddr *ifp, *tmp;
3524         struct inet6_dev *idev;
3525
3526         idev = __in6_dev_get(dev);
3527         if (!idev)
3528                 return;
3529
3530         write_lock_bh(&idev->lock);
3531
3532         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3533                 if ((ifp->flags & IFA_F_PERMANENT) &&
3534                     fixup_permanent_addr(net, idev, ifp) < 0) {
3535                         write_unlock_bh(&idev->lock);
3536                         in6_ifa_hold(ifp);
3537                         ipv6_del_addr(ifp);
3538                         write_lock_bh(&idev->lock);
3539
3540                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3541                                              idev->dev->name, &ifp->addr);
3542                 }
3543         }
3544
3545         write_unlock_bh(&idev->lock);
3546 }
3547
3548 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3549                            void *ptr)
3550 {
3551         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3552         struct netdev_notifier_change_info *change_info;
3553         struct netdev_notifier_changeupper_info *info;
3554         struct inet6_dev *idev = __in6_dev_get(dev);
3555         struct net *net = dev_net(dev);
3556         int run_pending = 0;
3557         int err;
3558
3559         switch (event) {
3560         case NETDEV_REGISTER:
3561                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3562                         idev = ipv6_add_dev(dev);
3563                         if (IS_ERR(idev))
3564                                 return notifier_from_errno(PTR_ERR(idev));
3565                 }
3566                 break;
3567
3568         case NETDEV_CHANGEMTU:
3569                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3570                 if (dev->mtu < IPV6_MIN_MTU) {
3571                         addrconf_ifdown(dev, dev != net->loopback_dev);
3572                         break;
3573                 }
3574
3575                 if (idev) {
3576                         rt6_mtu_change(dev, dev->mtu);
3577                         idev->cnf.mtu6 = dev->mtu;
3578                         break;
3579                 }
3580
3581                 /* allocate new idev */
3582                 idev = ipv6_add_dev(dev);
3583                 if (IS_ERR(idev))
3584                         break;
3585
3586                 /* device is still not ready */
3587                 if (!(idev->if_flags & IF_READY))
3588                         break;
3589
3590                 run_pending = 1;
3591                 fallthrough;
3592         case NETDEV_UP:
3593         case NETDEV_CHANGE:
3594                 if (dev->flags & IFF_SLAVE)
3595                         break;
3596
3597                 if (idev && idev->cnf.disable_ipv6)
3598                         break;
3599
3600                 if (event == NETDEV_UP) {
3601                         /* restore routes for permanent addresses */
3602                         addrconf_permanent_addr(net, dev);
3603
3604                         if (!addrconf_link_ready(dev)) {
3605                                 /* device is not ready yet. */
3606                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3607                                          dev->name);
3608                                 break;
3609                         }
3610
3611                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3612                                 idev = ipv6_add_dev(dev);
3613
3614                         if (!IS_ERR_OR_NULL(idev)) {
3615                                 idev->if_flags |= IF_READY;
3616                                 run_pending = 1;
3617                         }
3618                 } else if (event == NETDEV_CHANGE) {
3619                         if (!addrconf_link_ready(dev)) {
3620                                 /* device is still not ready. */
3621                                 rt6_sync_down_dev(dev, event);
3622                                 break;
3623                         }
3624
3625                         if (!IS_ERR_OR_NULL(idev)) {
3626                                 if (idev->if_flags & IF_READY) {
3627                                         /* device is already configured -
3628                                          * but resend MLD reports, we might
3629                                          * have roamed and need to update
3630                                          * multicast snooping switches
3631                                          */
3632                                         ipv6_mc_up(idev);
3633                                         change_info = ptr;
3634                                         if (change_info->flags_changed & IFF_NOARP)
3635                                                 addrconf_dad_run(idev, true);
3636                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3637                                         break;
3638                                 }
3639                                 idev->if_flags |= IF_READY;
3640                         }
3641
3642                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3643                                 dev->name);
3644
3645                         run_pending = 1;
3646                 }
3647
3648                 addrconf_init_auto_addrs(dev);
3649
3650                 if (!IS_ERR_OR_NULL(idev)) {
3651                         if (run_pending)
3652                                 addrconf_dad_run(idev, false);
3653
3654                         /* Device has an address by now */
3655                         rt6_sync_up(dev, RTNH_F_DEAD);
3656
3657                         /*
3658                          * If the MTU changed during the interface down,
3659                          * when the interface up, the changed MTU must be
3660                          * reflected in the idev as well as routers.
3661                          */
3662                         if (idev->cnf.mtu6 != dev->mtu &&
3663                             dev->mtu >= IPV6_MIN_MTU) {
3664                                 rt6_mtu_change(dev, dev->mtu);
3665                                 idev->cnf.mtu6 = dev->mtu;
3666                         }
3667                         idev->tstamp = jiffies;
3668                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3669
3670                         /*
3671                          * If the changed mtu during down is lower than
3672                          * IPV6_MIN_MTU stop IPv6 on this interface.
3673                          */
3674                         if (dev->mtu < IPV6_MIN_MTU)
3675                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3676                 }
3677                 break;
3678
3679         case NETDEV_DOWN:
3680         case NETDEV_UNREGISTER:
3681                 /*
3682                  *      Remove all addresses from this interface.
3683                  */
3684                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3685                 break;
3686
3687         case NETDEV_CHANGENAME:
3688                 if (idev) {
3689                         snmp6_unregister_dev(idev);
3690                         addrconf_sysctl_unregister(idev);
3691                         err = addrconf_sysctl_register(idev);
3692                         if (err)
3693                                 return notifier_from_errno(err);
3694                         err = snmp6_register_dev(idev);
3695                         if (err) {
3696                                 addrconf_sysctl_unregister(idev);
3697                                 return notifier_from_errno(err);
3698                         }
3699                 }
3700                 break;
3701
3702         case NETDEV_PRE_TYPE_CHANGE:
3703         case NETDEV_POST_TYPE_CHANGE:
3704                 if (idev)
3705                         addrconf_type_change(dev, event);
3706                 break;
3707
3708         case NETDEV_CHANGEUPPER:
3709                 info = ptr;
3710
3711                 /* flush all routes if dev is linked to or unlinked from
3712                  * an L3 master device (e.g., VRF)
3713                  */
3714                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3715                         addrconf_ifdown(dev, false);
3716         }
3717
3718         return NOTIFY_OK;
3719 }
3720
3721 /*
3722  *      addrconf module should be notified of a device going up
3723  */
3724 static struct notifier_block ipv6_dev_notf = {
3725         .notifier_call = addrconf_notify,
3726         .priority = ADDRCONF_NOTIFY_PRIORITY,
3727 };
3728
3729 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3730 {
3731         struct inet6_dev *idev;
3732         ASSERT_RTNL();
3733
3734         idev = __in6_dev_get(dev);
3735
3736         if (event == NETDEV_POST_TYPE_CHANGE)
3737                 ipv6_mc_remap(idev);
3738         else if (event == NETDEV_PRE_TYPE_CHANGE)
3739                 ipv6_mc_unmap(idev);
3740 }
3741
3742 static bool addr_is_local(const struct in6_addr *addr)
3743 {
3744         return ipv6_addr_type(addr) &
3745                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3746 }
3747
3748 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3749 {
3750         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3751         struct net *net = dev_net(dev);
3752         struct inet6_dev *idev;
3753         struct inet6_ifaddr *ifa;
3754         LIST_HEAD(tmp_addr_list);
3755         bool keep_addr = false;
3756         bool was_ready;
3757         int state, i;
3758
3759         ASSERT_RTNL();
3760
3761         rt6_disable_ip(dev, event);
3762
3763         idev = __in6_dev_get(dev);
3764         if (!idev)
3765                 return -ENODEV;
3766
3767         /*
3768          * Step 1: remove reference to ipv6 device from parent device.
3769          *         Do not dev_put!
3770          */
3771         if (unregister) {
3772                 idev->dead = 1;
3773
3774                 /* protected by rtnl_lock */
3775                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3776
3777                 /* Step 1.5: remove snmp6 entry */
3778                 snmp6_unregister_dev(idev);
3779
3780         }
3781
3782         /* combine the user config with event to determine if permanent
3783          * addresses are to be removed from address hash table
3784          */
3785         if (!unregister && !idev->cnf.disable_ipv6) {
3786                 /* aggregate the system setting and interface setting */
3787                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3788
3789                 if (!_keep_addr)
3790                         _keep_addr = idev->cnf.keep_addr_on_down;
3791
3792                 keep_addr = (_keep_addr > 0);
3793         }
3794
3795         /* Step 2: clear hash table */
3796         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3797                 struct hlist_head *h = &inet6_addr_lst[i];
3798
3799                 spin_lock_bh(&addrconf_hash_lock);
3800 restart:
3801                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3802                         if (ifa->idev == idev) {
3803                                 addrconf_del_dad_work(ifa);
3804                                 /* combined flag + permanent flag decide if
3805                                  * address is retained on a down event
3806                                  */
3807                                 if (!keep_addr ||
3808                                     !(ifa->flags & IFA_F_PERMANENT) ||
3809                                     addr_is_local(&ifa->addr)) {
3810                                         hlist_del_init_rcu(&ifa->addr_lst);
3811                                         goto restart;
3812                                 }
3813                         }
3814                 }
3815                 spin_unlock_bh(&addrconf_hash_lock);
3816         }
3817
3818         write_lock_bh(&idev->lock);
3819
3820         addrconf_del_rs_timer(idev);
3821
3822         /* Step 2: clear flags for stateless addrconf, repeated down
3823          *         detection
3824          */
3825         was_ready = idev->if_flags & IF_READY;
3826         if (!unregister)
3827                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3828
3829         /* Step 3: clear tempaddr list */
3830         while (!list_empty(&idev->tempaddr_list)) {
3831                 ifa = list_first_entry(&idev->tempaddr_list,
3832                                        struct inet6_ifaddr, tmp_list);
3833                 list_del(&ifa->tmp_list);
3834                 write_unlock_bh(&idev->lock);
3835                 spin_lock_bh(&ifa->lock);
3836
3837                 if (ifa->ifpub) {
3838                         in6_ifa_put(ifa->ifpub);
3839                         ifa->ifpub = NULL;
3840                 }
3841                 spin_unlock_bh(&ifa->lock);
3842                 in6_ifa_put(ifa);
3843                 write_lock_bh(&idev->lock);
3844         }
3845
3846         list_for_each_entry(ifa, &idev->addr_list, if_list)
3847                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3848         write_unlock_bh(&idev->lock);
3849
3850         while (!list_empty(&tmp_addr_list)) {
3851                 struct fib6_info *rt = NULL;
3852                 bool keep;
3853
3854                 ifa = list_first_entry(&tmp_addr_list,
3855                                        struct inet6_ifaddr, if_list_aux);
3856                 list_del(&ifa->if_list_aux);
3857
3858                 addrconf_del_dad_work(ifa);
3859
3860                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3861                         !addr_is_local(&ifa->addr);
3862
3863                 spin_lock_bh(&ifa->lock);
3864
3865                 if (keep) {
3866                         /* set state to skip the notifier below */
3867                         state = INET6_IFADDR_STATE_DEAD;
3868                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3869                         if (!(ifa->flags & IFA_F_NODAD))
3870                                 ifa->flags |= IFA_F_TENTATIVE;
3871
3872                         rt = ifa->rt;
3873                         ifa->rt = NULL;
3874                 } else {
3875                         state = ifa->state;
3876                         ifa->state = INET6_IFADDR_STATE_DEAD;
3877                 }
3878
3879                 spin_unlock_bh(&ifa->lock);
3880
3881                 if (rt)
3882                         ip6_del_rt(net, rt, false);
3883
3884                 if (state != INET6_IFADDR_STATE_DEAD) {
3885                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3886                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3887                 } else {
3888                         if (idev->cnf.forwarding)
3889                                 addrconf_leave_anycast(ifa);
3890                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3891                 }
3892
3893                 if (!keep) {
3894                         write_lock_bh(&idev->lock);
3895                         list_del_rcu(&ifa->if_list);
3896                         write_unlock_bh(&idev->lock);
3897                         in6_ifa_put(ifa);
3898                 }
3899         }
3900
3901         /* Step 5: Discard anycast and multicast list */
3902         if (unregister) {
3903                 ipv6_ac_destroy_dev(idev);
3904                 ipv6_mc_destroy_dev(idev);
3905         } else if (was_ready) {
3906                 ipv6_mc_down(idev);
3907         }
3908
3909         idev->tstamp = jiffies;
3910         idev->ra_mtu = 0;
3911
3912         /* Last: Shot the device (if unregistered) */
3913         if (unregister) {
3914                 addrconf_sysctl_unregister(idev);
3915                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3916                 neigh_ifdown(&nd_tbl, dev);
3917                 in6_dev_put(idev);
3918         }
3919         return 0;
3920 }
3921
3922 static void addrconf_rs_timer(struct timer_list *t)
3923 {
3924         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3925         struct net_device *dev = idev->dev;
3926         struct in6_addr lladdr;
3927
3928         write_lock(&idev->lock);
3929         if (idev->dead || !(idev->if_flags & IF_READY))
3930                 goto out;
3931
3932         if (!ipv6_accept_ra(idev))
3933                 goto out;
3934
3935         /* Announcement received after solicitation was sent */
3936         if (idev->if_flags & IF_RA_RCVD)
3937                 goto out;
3938
3939         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3940                 write_unlock(&idev->lock);
3941                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3942                         ndisc_send_rs(dev, &lladdr,
3943                                       &in6addr_linklocal_allrouters);
3944                 else
3945                         goto put;
3946
3947                 write_lock(&idev->lock);
3948                 idev->rs_interval = rfc3315_s14_backoff_update(
3949                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3950                 /* The wait after the last probe can be shorter */
3951                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3952                                              idev->cnf.rtr_solicits) ?
3953                                       idev->cnf.rtr_solicit_delay :
3954                                       idev->rs_interval);
3955         } else {
3956                 /*
3957                  * Note: we do not support deprecated "all on-link"
3958                  * assumption any longer.
3959                  */
3960                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3961         }
3962
3963 out:
3964         write_unlock(&idev->lock);
3965 put:
3966         in6_dev_put(idev);
3967 }
3968
3969 /*
3970  *      Duplicate Address Detection
3971  */
3972 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3973 {
3974         unsigned long rand_num;
3975         struct inet6_dev *idev = ifp->idev;
3976         u64 nonce;
3977
3978         if (ifp->flags & IFA_F_OPTIMISTIC)
3979                 rand_num = 0;
3980         else
3981                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3982
3983         nonce = 0;
3984         if (idev->cnf.enhanced_dad ||
3985             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3986                 do
3987                         get_random_bytes(&nonce, 6);
3988                 while (nonce == 0);
3989         }
3990         ifp->dad_nonce = nonce;
3991         ifp->dad_probes = idev->cnf.dad_transmits;
3992         addrconf_mod_dad_work(ifp, rand_num);
3993 }
3994
3995 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3996 {
3997         struct inet6_dev *idev = ifp->idev;
3998         struct net_device *dev = idev->dev;
3999         bool bump_id, notify = false;
4000         struct net *net;
4001
4002         addrconf_join_solict(dev, &ifp->addr);
4003
4004         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
4005
4006         read_lock_bh(&idev->lock);
4007         spin_lock(&ifp->lock);
4008         if (ifp->state == INET6_IFADDR_STATE_DEAD)
4009                 goto out;
4010
4011         net = dev_net(dev);
4012         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4013             (net->ipv6.devconf_all->accept_dad < 1 &&
4014              idev->cnf.accept_dad < 1) ||
4015             !(ifp->flags&IFA_F_TENTATIVE) ||
4016             ifp->flags & IFA_F_NODAD) {
4017                 bool send_na = false;
4018
4019                 if (ifp->flags & IFA_F_TENTATIVE &&
4020                     !(ifp->flags & IFA_F_OPTIMISTIC))
4021                         send_na = true;
4022                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4023                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4024                 spin_unlock(&ifp->lock);
4025                 read_unlock_bh(&idev->lock);
4026
4027                 addrconf_dad_completed(ifp, bump_id, send_na);
4028                 return;
4029         }
4030
4031         if (!(idev->if_flags & IF_READY)) {
4032                 spin_unlock(&ifp->lock);
4033                 read_unlock_bh(&idev->lock);
4034                 /*
4035                  * If the device is not ready:
4036                  * - keep it tentative if it is a permanent address.
4037                  * - otherwise, kill it.
4038                  */
4039                 in6_ifa_hold(ifp);
4040                 addrconf_dad_stop(ifp, 0);
4041                 return;
4042         }
4043
4044         /*
4045          * Optimistic nodes can start receiving
4046          * Frames right away
4047          */
4048         if (ifp->flags & IFA_F_OPTIMISTIC) {
4049                 ip6_ins_rt(net, ifp->rt);
4050                 if (ipv6_use_optimistic_addr(net, idev)) {
4051                         /* Because optimistic nodes can use this address,
4052                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4053                          */
4054                         notify = true;
4055                 }
4056         }
4057
4058         addrconf_dad_kick(ifp);
4059 out:
4060         spin_unlock(&ifp->lock);
4061         read_unlock_bh(&idev->lock);
4062         if (notify)
4063                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4064 }
4065
4066 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4067 {
4068         bool begin_dad = false;
4069
4070         spin_lock_bh(&ifp->lock);
4071         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4072                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4073                 begin_dad = true;
4074         }
4075         spin_unlock_bh(&ifp->lock);
4076
4077         if (begin_dad)
4078                 addrconf_mod_dad_work(ifp, 0);
4079 }
4080
4081 static void addrconf_dad_work(struct work_struct *w)
4082 {
4083         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4084                                                 struct inet6_ifaddr,
4085                                                 dad_work);
4086         struct inet6_dev *idev = ifp->idev;
4087         bool bump_id, disable_ipv6 = false;
4088         struct in6_addr mcaddr;
4089
4090         enum {
4091                 DAD_PROCESS,
4092                 DAD_BEGIN,
4093                 DAD_ABORT,
4094         } action = DAD_PROCESS;
4095
4096         rtnl_lock();
4097
4098         spin_lock_bh(&ifp->lock);
4099         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4100                 action = DAD_BEGIN;
4101                 ifp->state = INET6_IFADDR_STATE_DAD;
4102         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4103                 action = DAD_ABORT;
4104                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4105
4106                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4107                      idev->cnf.accept_dad > 1) &&
4108                     !idev->cnf.disable_ipv6 &&
4109                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4110                         struct in6_addr addr;
4111
4112                         addr.s6_addr32[0] = htonl(0xfe800000);
4113                         addr.s6_addr32[1] = 0;
4114
4115                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4116                             ipv6_addr_equal(&ifp->addr, &addr)) {
4117                                 /* DAD failed for link-local based on MAC */
4118                                 idev->cnf.disable_ipv6 = 1;
4119
4120                                 pr_info("%s: IPv6 being disabled!\n",
4121                                         ifp->idev->dev->name);
4122                                 disable_ipv6 = true;
4123                         }
4124                 }
4125         }
4126         spin_unlock_bh(&ifp->lock);
4127
4128         if (action == DAD_BEGIN) {
4129                 addrconf_dad_begin(ifp);
4130                 goto out;
4131         } else if (action == DAD_ABORT) {
4132                 in6_ifa_hold(ifp);
4133                 addrconf_dad_stop(ifp, 1);
4134                 if (disable_ipv6)
4135                         addrconf_ifdown(idev->dev, false);
4136                 goto out;
4137         }
4138
4139         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4140                 goto out;
4141
4142         write_lock_bh(&idev->lock);
4143         if (idev->dead || !(idev->if_flags & IF_READY)) {
4144                 write_unlock_bh(&idev->lock);
4145                 goto out;
4146         }
4147
4148         spin_lock(&ifp->lock);
4149         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4150                 spin_unlock(&ifp->lock);
4151                 write_unlock_bh(&idev->lock);
4152                 goto out;
4153         }
4154
4155         if (ifp->dad_probes == 0) {
4156                 bool send_na = false;
4157
4158                 /*
4159                  * DAD was successful
4160                  */
4161
4162                 if (ifp->flags & IFA_F_TENTATIVE &&
4163                     !(ifp->flags & IFA_F_OPTIMISTIC))
4164                         send_na = true;
4165                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4166                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4167                 spin_unlock(&ifp->lock);
4168                 write_unlock_bh(&idev->lock);
4169
4170                 addrconf_dad_completed(ifp, bump_id, send_na);
4171
4172                 goto out;
4173         }
4174
4175         ifp->dad_probes--;
4176         addrconf_mod_dad_work(ifp,
4177                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4178                                   HZ/100));
4179         spin_unlock(&ifp->lock);
4180         write_unlock_bh(&idev->lock);
4181
4182         /* send a neighbour solicitation for our addr */
4183         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4184         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4185                       ifp->dad_nonce);
4186 out:
4187         in6_ifa_put(ifp);
4188         rtnl_unlock();
4189 }
4190
4191 /* ifp->idev must be at least read locked */
4192 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4193 {
4194         struct inet6_ifaddr *ifpiter;
4195         struct inet6_dev *idev = ifp->idev;
4196
4197         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4198                 if (ifpiter->scope > IFA_LINK)
4199                         break;
4200                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4201                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4202                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4203                     IFA_F_PERMANENT)
4204                         return false;
4205         }
4206         return true;
4207 }
4208
4209 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4210                                    bool send_na)
4211 {
4212         struct net_device *dev = ifp->idev->dev;
4213         struct in6_addr lladdr;
4214         bool send_rs, send_mld;
4215
4216         addrconf_del_dad_work(ifp);
4217
4218         /*
4219          *      Configure the address for reception. Now it is valid.
4220          */
4221
4222         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4223
4224         /* If added prefix is link local and we are prepared to process
4225            router advertisements, start sending router solicitations.
4226          */
4227
4228         read_lock_bh(&ifp->idev->lock);
4229         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4230         send_rs = send_mld &&
4231                   ipv6_accept_ra(ifp->idev) &&
4232                   ifp->idev->cnf.rtr_solicits != 0 &&
4233                   (dev->flags & IFF_LOOPBACK) == 0 &&
4234                   (dev->type != ARPHRD_TUNNEL);
4235         read_unlock_bh(&ifp->idev->lock);
4236
4237         /* While dad is in progress mld report's source address is in6_addrany.
4238          * Resend with proper ll now.
4239          */
4240         if (send_mld)
4241                 ipv6_mc_dad_complete(ifp->idev);
4242
4243         /* send unsolicited NA if enabled */
4244         if (send_na &&
4245             (ifp->idev->cnf.ndisc_notify ||
4246              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4247                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4248                               /*router=*/ !!ifp->idev->cnf.forwarding,
4249                               /*solicited=*/ false, /*override=*/ true,
4250                               /*inc_opt=*/ true);
4251         }
4252
4253         if (send_rs) {
4254                 /*
4255                  *      If a host as already performed a random delay
4256                  *      [...] as part of DAD [...] there is no need
4257                  *      to delay again before sending the first RS
4258                  */
4259                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4260                         return;
4261                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4262
4263                 write_lock_bh(&ifp->idev->lock);
4264                 spin_lock(&ifp->lock);
4265                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4266                         ifp->idev->cnf.rtr_solicit_interval);
4267                 ifp->idev->rs_probes = 1;
4268                 ifp->idev->if_flags |= IF_RS_SENT;
4269                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4270                 spin_unlock(&ifp->lock);
4271                 write_unlock_bh(&ifp->idev->lock);
4272         }
4273
4274         if (bump_id)
4275                 rt_genid_bump_ipv6(dev_net(dev));
4276
4277         /* Make sure that a new temporary address will be created
4278          * before this temporary address becomes deprecated.
4279          */
4280         if (ifp->flags & IFA_F_TEMPORARY)
4281                 addrconf_verify_rtnl();
4282 }
4283
4284 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4285 {
4286         struct inet6_ifaddr *ifp;
4287
4288         read_lock_bh(&idev->lock);
4289         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4290                 spin_lock(&ifp->lock);
4291                 if ((ifp->flags & IFA_F_TENTATIVE &&
4292                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4293                         if (restart)
4294                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4295                         addrconf_dad_kick(ifp);
4296                 }
4297                 spin_unlock(&ifp->lock);
4298         }
4299         read_unlock_bh(&idev->lock);
4300 }
4301
4302 #ifdef CONFIG_PROC_FS
4303 struct if6_iter_state {
4304         struct seq_net_private p;
4305         int bucket;
4306         int offset;
4307 };
4308
4309 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4310 {
4311         struct if6_iter_state *state = seq->private;
4312         struct net *net = seq_file_net(seq);
4313         struct inet6_ifaddr *ifa = NULL;
4314         int p = 0;
4315
4316         /* initial bucket if pos is 0 */
4317         if (pos == 0) {
4318                 state->bucket = 0;
4319                 state->offset = 0;
4320         }
4321
4322         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4323                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4324                                          addr_lst) {
4325                         if (!net_eq(dev_net(ifa->idev->dev), net))
4326                                 continue;
4327                         /* sync with offset */
4328                         if (p < state->offset) {
4329                                 p++;
4330                                 continue;
4331                         }
4332                         return ifa;
4333                 }
4334
4335                 /* prepare for next bucket */
4336                 state->offset = 0;
4337                 p = 0;
4338         }
4339         return NULL;
4340 }
4341
4342 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4343                                          struct inet6_ifaddr *ifa)
4344 {
4345         struct if6_iter_state *state = seq->private;
4346         struct net *net = seq_file_net(seq);
4347
4348         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4349                 if (!net_eq(dev_net(ifa->idev->dev), net))
4350                         continue;
4351                 state->offset++;
4352                 return ifa;
4353         }
4354
4355         state->offset = 0;
4356         while (++state->bucket < IN6_ADDR_HSIZE) {
4357                 hlist_for_each_entry_rcu(ifa,
4358                                      &inet6_addr_lst[state->bucket], addr_lst) {
4359                         if (!net_eq(dev_net(ifa->idev->dev), net))
4360                                 continue;
4361                         return ifa;
4362                 }
4363         }
4364
4365         return NULL;
4366 }
4367
4368 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4369         __acquires(rcu)
4370 {
4371         rcu_read_lock();
4372         return if6_get_first(seq, *pos);
4373 }
4374
4375 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4376 {
4377         struct inet6_ifaddr *ifa;
4378
4379         ifa = if6_get_next(seq, v);
4380         ++*pos;
4381         return ifa;
4382 }
4383
4384 static void if6_seq_stop(struct seq_file *seq, void *v)
4385         __releases(rcu)
4386 {
4387         rcu_read_unlock();
4388 }
4389
4390 static int if6_seq_show(struct seq_file *seq, void *v)
4391 {
4392         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4393         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4394                    &ifp->addr,
4395                    ifp->idev->dev->ifindex,
4396                    ifp->prefix_len,
4397                    ifp->scope,
4398                    (u8) ifp->flags,
4399                    ifp->idev->dev->name);
4400         return 0;
4401 }
4402
4403 static const struct seq_operations if6_seq_ops = {
4404         .start  = if6_seq_start,
4405         .next   = if6_seq_next,
4406         .show   = if6_seq_show,
4407         .stop   = if6_seq_stop,
4408 };
4409
4410 static int __net_init if6_proc_net_init(struct net *net)
4411 {
4412         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4413                         sizeof(struct if6_iter_state)))
4414                 return -ENOMEM;
4415         return 0;
4416 }
4417
4418 static void __net_exit if6_proc_net_exit(struct net *net)
4419 {
4420         remove_proc_entry("if_inet6", net->proc_net);
4421 }
4422
4423 static struct pernet_operations if6_proc_net_ops = {
4424         .init = if6_proc_net_init,
4425         .exit = if6_proc_net_exit,
4426 };
4427
4428 int __init if6_proc_init(void)
4429 {
4430         return register_pernet_subsys(&if6_proc_net_ops);
4431 }
4432
4433 void if6_proc_exit(void)
4434 {
4435         unregister_pernet_subsys(&if6_proc_net_ops);
4436 }
4437 #endif  /* CONFIG_PROC_FS */
4438
4439 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4440 /* Check if address is a home address configured on any interface. */
4441 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4442 {
4443         unsigned int hash = inet6_addr_hash(net, addr);
4444         struct inet6_ifaddr *ifp = NULL;
4445         int ret = 0;
4446
4447         rcu_read_lock();
4448         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4449                 if (!net_eq(dev_net(ifp->idev->dev), net))
4450                         continue;
4451                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4452                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4453                         ret = 1;
4454                         break;
4455                 }
4456         }
4457         rcu_read_unlock();
4458         return ret;
4459 }
4460 #endif
4461
4462 /* RFC6554 has some algorithm to avoid loops in segment routing by
4463  * checking if the segments contains any of a local interface address.
4464  *
4465  * Quote:
4466  *
4467  * To detect loops in the SRH, a router MUST determine if the SRH
4468  * includes multiple addresses assigned to any interface on that router.
4469  * If such addresses appear more than once and are separated by at least
4470  * one address not assigned to that router.
4471  */
4472 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4473                           unsigned char nsegs)
4474 {
4475         const struct in6_addr *addr;
4476         int i, ret = 0, found = 0;
4477         struct inet6_ifaddr *ifp;
4478         bool separated = false;
4479         unsigned int hash;
4480         bool hash_found;
4481
4482         rcu_read_lock();
4483         for (i = 0; i < nsegs; i++) {
4484                 addr = &segs[i];
4485                 hash = inet6_addr_hash(net, addr);
4486
4487                 hash_found = false;
4488                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4489                         if (!net_eq(dev_net(ifp->idev->dev), net))
4490                                 continue;
4491
4492                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4493                                 hash_found = true;
4494                                 break;
4495                         }
4496                 }
4497
4498                 if (hash_found) {
4499                         if (found > 1 && separated) {
4500                                 ret = 1;
4501                                 break;
4502                         }
4503
4504                         separated = false;
4505                         found++;
4506                 } else {
4507                         separated = true;
4508                 }
4509         }
4510         rcu_read_unlock();
4511
4512         return ret;
4513 }
4514
4515 /*
4516  *      Periodic address status verification
4517  */
4518
4519 static void addrconf_verify_rtnl(void)
4520 {
4521         unsigned long now, next, next_sec, next_sched;
4522         struct inet6_ifaddr *ifp;
4523         int i;
4524
4525         ASSERT_RTNL();
4526
4527         rcu_read_lock_bh();
4528         now = jiffies;
4529         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4530
4531         cancel_delayed_work(&addr_chk_work);
4532
4533         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4534 restart:
4535                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4536                         unsigned long age;
4537
4538                         /* When setting preferred_lft to a value not zero or
4539                          * infinity, while valid_lft is infinity
4540                          * IFA_F_PERMANENT has a non-infinity life time.
4541                          */
4542                         if ((ifp->flags & IFA_F_PERMANENT) &&
4543                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4544                                 continue;
4545
4546                         spin_lock(&ifp->lock);
4547                         /* We try to batch several events at once. */
4548                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4549
4550                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4551                             age >= ifp->valid_lft) {
4552                                 spin_unlock(&ifp->lock);
4553                                 in6_ifa_hold(ifp);
4554                                 rcu_read_unlock_bh();
4555                                 ipv6_del_addr(ifp);
4556                                 rcu_read_lock_bh();
4557                                 goto restart;
4558                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4559                                 spin_unlock(&ifp->lock);
4560                                 continue;
4561                         } else if (age >= ifp->prefered_lft) {
4562                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4563                                 int deprecate = 0;
4564
4565                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4566                                         deprecate = 1;
4567                                         ifp->flags |= IFA_F_DEPRECATED;
4568                                 }
4569
4570                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4571                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4572                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4573
4574                                 spin_unlock(&ifp->lock);
4575
4576                                 if (deprecate) {
4577                                         in6_ifa_hold(ifp);
4578
4579                                         ipv6_ifa_notify(0, ifp);
4580                                         in6_ifa_put(ifp);
4581                                         goto restart;
4582                                 }
4583                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4584                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4585                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4586                                         ifp->idev->cnf.dad_transmits *
4587                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4588
4589                                 if (age >= ifp->prefered_lft - regen_advance) {
4590                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4591                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4592                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4593                                         if (!ifp->regen_count && ifpub) {
4594                                                 ifp->regen_count++;
4595                                                 in6_ifa_hold(ifp);
4596                                                 in6_ifa_hold(ifpub);
4597                                                 spin_unlock(&ifp->lock);
4598
4599                                                 spin_lock(&ifpub->lock);
4600                                                 ifpub->regen_count = 0;
4601                                                 spin_unlock(&ifpub->lock);
4602                                                 rcu_read_unlock_bh();
4603                                                 ipv6_create_tempaddr(ifpub, true);
4604                                                 in6_ifa_put(ifpub);
4605                                                 in6_ifa_put(ifp);
4606                                                 rcu_read_lock_bh();
4607                                                 goto restart;
4608                                         }
4609                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4610                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4611                                 spin_unlock(&ifp->lock);
4612                         } else {
4613                                 /* ifp->prefered_lft <= ifp->valid_lft */
4614                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4615                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4616                                 spin_unlock(&ifp->lock);
4617                         }
4618                 }
4619         }
4620
4621         next_sec = round_jiffies_up(next);
4622         next_sched = next;
4623
4624         /* If rounded timeout is accurate enough, accept it. */
4625         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4626                 next_sched = next_sec;
4627
4628         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4629         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4630                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4631
4632         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4633                  now, next, next_sec, next_sched);
4634         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4635         rcu_read_unlock_bh();
4636 }
4637
4638 static void addrconf_verify_work(struct work_struct *w)
4639 {
4640         rtnl_lock();
4641         addrconf_verify_rtnl();
4642         rtnl_unlock();
4643 }
4644
4645 static void addrconf_verify(void)
4646 {
4647         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4648 }
4649
4650 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4651                                      struct in6_addr **peer_pfx)
4652 {
4653         struct in6_addr *pfx = NULL;
4654
4655         *peer_pfx = NULL;
4656
4657         if (addr)
4658                 pfx = nla_data(addr);
4659
4660         if (local) {
4661                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4662                         *peer_pfx = pfx;
4663                 pfx = nla_data(local);
4664         }
4665
4666         return pfx;
4667 }
4668
4669 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4670         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4671         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4672         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4673         [IFA_FLAGS]             = { .len = sizeof(u32) },
4674         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4675         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4676 };
4677
4678 static int
4679 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4680                   struct netlink_ext_ack *extack)
4681 {
4682         struct net *net = sock_net(skb->sk);
4683         struct ifaddrmsg *ifm;
4684         struct nlattr *tb[IFA_MAX+1];
4685         struct in6_addr *pfx, *peer_pfx;
4686         u32 ifa_flags;
4687         int err;
4688
4689         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4690                                      ifa_ipv6_policy, extack);
4691         if (err < 0)
4692                 return err;
4693
4694         ifm = nlmsg_data(nlh);
4695         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4696         if (!pfx)
4697                 return -EINVAL;
4698
4699         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4700
4701         /* We ignore other flags so far. */
4702         ifa_flags &= IFA_F_MANAGETEMPADDR;
4703
4704         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4705                               ifm->ifa_prefixlen);
4706 }
4707
4708 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4709                                unsigned long expires, u32 flags,
4710                                bool modify_peer)
4711 {
4712         struct fib6_info *f6i;
4713         u32 prio;
4714
4715         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4716                                         ifp->prefix_len,
4717                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4718         if (!f6i)
4719                 return -ENOENT;
4720
4721         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4722         if (f6i->fib6_metric != prio) {
4723                 /* delete old one */
4724                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4725
4726                 /* add new one */
4727                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4728                                       ifp->prefix_len,
4729                                       ifp->rt_priority, ifp->idev->dev,
4730                                       expires, flags, GFP_KERNEL);
4731         } else {
4732                 if (!expires)
4733                         fib6_clean_expires(f6i);
4734                 else
4735                         fib6_set_expires(f6i, expires);
4736
4737                 fib6_info_release(f6i);
4738         }
4739
4740         return 0;
4741 }
4742
4743 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4744 {
4745         u32 flags;
4746         clock_t expires;
4747         unsigned long timeout;
4748         bool was_managetempaddr;
4749         bool had_prefixroute;
4750         bool new_peer = false;
4751
4752         ASSERT_RTNL();
4753
4754         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4755                 return -EINVAL;
4756
4757         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4758             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4759                 return -EINVAL;
4760
4761         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4762                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4763
4764         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4765         if (addrconf_finite_timeout(timeout)) {
4766                 expires = jiffies_to_clock_t(timeout * HZ);
4767                 cfg->valid_lft = timeout;
4768                 flags = RTF_EXPIRES;
4769         } else {
4770                 expires = 0;
4771                 flags = 0;
4772                 cfg->ifa_flags |= IFA_F_PERMANENT;
4773         }
4774
4775         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4776         if (addrconf_finite_timeout(timeout)) {
4777                 if (timeout == 0)
4778                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4779                 cfg->preferred_lft = timeout;
4780         }
4781
4782         if (cfg->peer_pfx &&
4783             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4784                 if (!ipv6_addr_any(&ifp->peer_addr))
4785                         cleanup_prefix_route(ifp, expires, true, true);
4786                 new_peer = true;
4787         }
4788
4789         spin_lock_bh(&ifp->lock);
4790         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4791         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4792                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4793         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4794                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4795                         IFA_F_NOPREFIXROUTE);
4796         ifp->flags |= cfg->ifa_flags;
4797         ifp->tstamp = jiffies;
4798         ifp->valid_lft = cfg->valid_lft;
4799         ifp->prefered_lft = cfg->preferred_lft;
4800
4801         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4802                 ifp->rt_priority = cfg->rt_priority;
4803
4804         if (new_peer)
4805                 ifp->peer_addr = *cfg->peer_pfx;
4806
4807         spin_unlock_bh(&ifp->lock);
4808         if (!(ifp->flags&IFA_F_TENTATIVE))
4809                 ipv6_ifa_notify(0, ifp);
4810
4811         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4812                 int rc = -ENOENT;
4813
4814                 if (had_prefixroute)
4815                         rc = modify_prefix_route(ifp, expires, flags, false);
4816
4817                 /* prefix route could have been deleted; if so restore it */
4818                 if (rc == -ENOENT) {
4819                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4820                                               ifp->rt_priority, ifp->idev->dev,
4821                                               expires, flags, GFP_KERNEL);
4822                 }
4823
4824                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4825                         rc = modify_prefix_route(ifp, expires, flags, true);
4826
4827                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4828                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4829                                               ifp->rt_priority, ifp->idev->dev,
4830                                               expires, flags, GFP_KERNEL);
4831                 }
4832         } else if (had_prefixroute) {
4833                 enum cleanup_prefix_rt_t action;
4834                 unsigned long rt_expires;
4835
4836                 write_lock_bh(&ifp->idev->lock);
4837                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4838                 write_unlock_bh(&ifp->idev->lock);
4839
4840                 if (action != CLEANUP_PREFIX_RT_NOP) {
4841                         cleanup_prefix_route(ifp, rt_expires,
4842                                 action == CLEANUP_PREFIX_RT_DEL, false);
4843                 }
4844         }
4845
4846         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4847                 if (was_managetempaddr &&
4848                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4849                         cfg->valid_lft = 0;
4850                         cfg->preferred_lft = 0;
4851                 }
4852                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4853                                  cfg->preferred_lft, !was_managetempaddr,
4854                                  jiffies);
4855         }
4856
4857         addrconf_verify_rtnl();
4858
4859         return 0;
4860 }
4861
4862 static int
4863 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4864                   struct netlink_ext_ack *extack)
4865 {
4866         struct net *net = sock_net(skb->sk);
4867         struct ifaddrmsg *ifm;
4868         struct nlattr *tb[IFA_MAX+1];
4869         struct in6_addr *peer_pfx;
4870         struct inet6_ifaddr *ifa;
4871         struct net_device *dev;
4872         struct inet6_dev *idev;
4873         struct ifa6_config cfg;
4874         int err;
4875
4876         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4877                                      ifa_ipv6_policy, extack);
4878         if (err < 0)
4879                 return err;
4880
4881         memset(&cfg, 0, sizeof(cfg));
4882
4883         ifm = nlmsg_data(nlh);
4884         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4885         if (!cfg.pfx)
4886                 return -EINVAL;
4887
4888         cfg.peer_pfx = peer_pfx;
4889         cfg.plen = ifm->ifa_prefixlen;
4890         if (tb[IFA_RT_PRIORITY])
4891                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4892
4893         cfg.valid_lft = INFINITY_LIFE_TIME;
4894         cfg.preferred_lft = INFINITY_LIFE_TIME;
4895
4896         if (tb[IFA_CACHEINFO]) {
4897                 struct ifa_cacheinfo *ci;
4898
4899                 ci = nla_data(tb[IFA_CACHEINFO]);
4900                 cfg.valid_lft = ci->ifa_valid;
4901                 cfg.preferred_lft = ci->ifa_prefered;
4902         }
4903
4904         dev =  __dev_get_by_index(net, ifm->ifa_index);
4905         if (!dev)
4906                 return -ENODEV;
4907
4908         if (tb[IFA_FLAGS])
4909                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4910         else
4911                 cfg.ifa_flags = ifm->ifa_flags;
4912
4913         /* We ignore other flags so far. */
4914         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4915                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4916                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4917
4918         idev = ipv6_find_idev(dev);
4919         if (IS_ERR(idev))
4920                 return PTR_ERR(idev);
4921
4922         if (!ipv6_allow_optimistic_dad(net, idev))
4923                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4924
4925         if (cfg.ifa_flags & IFA_F_NODAD &&
4926             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4927                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4928                 return -EINVAL;
4929         }
4930
4931         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4932         if (!ifa) {
4933                 /*
4934                  * It would be best to check for !NLM_F_CREATE here but
4935                  * userspace already relies on not having to provide this.
4936                  */
4937                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4938         }
4939
4940         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4941             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4942                 err = -EEXIST;
4943         else
4944                 err = inet6_addr_modify(ifa, &cfg);
4945
4946         in6_ifa_put(ifa);
4947
4948         return err;
4949 }
4950
4951 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4952                           u8 scope, int ifindex)
4953 {
4954         struct ifaddrmsg *ifm;
4955
4956         ifm = nlmsg_data(nlh);
4957         ifm->ifa_family = AF_INET6;
4958         ifm->ifa_prefixlen = prefixlen;
4959         ifm->ifa_flags = flags;
4960         ifm->ifa_scope = scope;
4961         ifm->ifa_index = ifindex;
4962 }
4963
4964 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4965                          unsigned long tstamp, u32 preferred, u32 valid)
4966 {
4967         struct ifa_cacheinfo ci;
4968
4969         ci.cstamp = cstamp_delta(cstamp);
4970         ci.tstamp = cstamp_delta(tstamp);
4971         ci.ifa_prefered = preferred;
4972         ci.ifa_valid = valid;
4973
4974         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4975 }
4976
4977 static inline int rt_scope(int ifa_scope)
4978 {
4979         if (ifa_scope & IFA_HOST)
4980                 return RT_SCOPE_HOST;
4981         else if (ifa_scope & IFA_LINK)
4982                 return RT_SCOPE_LINK;
4983         else if (ifa_scope & IFA_SITE)
4984                 return RT_SCOPE_SITE;
4985         else
4986                 return RT_SCOPE_UNIVERSE;
4987 }
4988
4989 static inline int inet6_ifaddr_msgsize(void)
4990 {
4991         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4992                + nla_total_size(16) /* IFA_LOCAL */
4993                + nla_total_size(16) /* IFA_ADDRESS */
4994                + nla_total_size(sizeof(struct ifa_cacheinfo))
4995                + nla_total_size(4)  /* IFA_FLAGS */
4996                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4997 }
4998
4999 enum addr_type_t {
5000         UNICAST_ADDR,
5001         MULTICAST_ADDR,
5002         ANYCAST_ADDR,
5003 };
5004
5005 struct inet6_fill_args {
5006         u32 portid;
5007         u32 seq;
5008         int event;
5009         unsigned int flags;
5010         int netnsid;
5011         int ifindex;
5012         enum addr_type_t type;
5013 };
5014
5015 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5016                              struct inet6_fill_args *args)
5017 {
5018         struct nlmsghdr  *nlh;
5019         u32 preferred, valid;
5020
5021         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5022                         sizeof(struct ifaddrmsg), args->flags);
5023         if (!nlh)
5024                 return -EMSGSIZE;
5025
5026         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5027                       ifa->idev->dev->ifindex);
5028
5029         if (args->netnsid >= 0 &&
5030             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5031                 goto error;
5032
5033         spin_lock_bh(&ifa->lock);
5034         if (!((ifa->flags&IFA_F_PERMANENT) &&
5035               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5036                 preferred = ifa->prefered_lft;
5037                 valid = ifa->valid_lft;
5038                 if (preferred != INFINITY_LIFE_TIME) {
5039                         long tval = (jiffies - ifa->tstamp)/HZ;
5040                         if (preferred > tval)
5041                                 preferred -= tval;
5042                         else
5043                                 preferred = 0;
5044                         if (valid != INFINITY_LIFE_TIME) {
5045                                 if (valid > tval)
5046                                         valid -= tval;
5047                                 else
5048                                         valid = 0;
5049                         }
5050                 }
5051         } else {
5052                 preferred = INFINITY_LIFE_TIME;
5053                 valid = INFINITY_LIFE_TIME;
5054         }
5055         spin_unlock_bh(&ifa->lock);
5056
5057         if (!ipv6_addr_any(&ifa->peer_addr)) {
5058                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5059                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5060                         goto error;
5061         } else
5062                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5063                         goto error;
5064
5065         if (ifa->rt_priority &&
5066             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5067                 goto error;
5068
5069         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5070                 goto error;
5071
5072         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5073                 goto error;
5074
5075         nlmsg_end(skb, nlh);
5076         return 0;
5077
5078 error:
5079         nlmsg_cancel(skb, nlh);
5080         return -EMSGSIZE;
5081 }
5082
5083 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5084                                struct inet6_fill_args *args)
5085 {
5086         struct nlmsghdr  *nlh;
5087         u8 scope = RT_SCOPE_UNIVERSE;
5088         int ifindex = ifmca->idev->dev->ifindex;
5089
5090         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5091                 scope = RT_SCOPE_SITE;
5092
5093         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5094                         sizeof(struct ifaddrmsg), args->flags);
5095         if (!nlh)
5096                 return -EMSGSIZE;
5097
5098         if (args->netnsid >= 0 &&
5099             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5100                 nlmsg_cancel(skb, nlh);
5101                 return -EMSGSIZE;
5102         }
5103
5104         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5105         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5106             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5107                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5108                 nlmsg_cancel(skb, nlh);
5109                 return -EMSGSIZE;
5110         }
5111
5112         nlmsg_end(skb, nlh);
5113         return 0;
5114 }
5115
5116 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5117                                struct inet6_fill_args *args)
5118 {
5119         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5120         int ifindex = dev ? dev->ifindex : 1;
5121         struct nlmsghdr  *nlh;
5122         u8 scope = RT_SCOPE_UNIVERSE;
5123
5124         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5125                 scope = RT_SCOPE_SITE;
5126
5127         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5128                         sizeof(struct ifaddrmsg), args->flags);
5129         if (!nlh)
5130                 return -EMSGSIZE;
5131
5132         if (args->netnsid >= 0 &&
5133             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5134                 nlmsg_cancel(skb, nlh);
5135                 return -EMSGSIZE;
5136         }
5137
5138         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5139         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5140             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5141                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5142                 nlmsg_cancel(skb, nlh);
5143                 return -EMSGSIZE;
5144         }
5145
5146         nlmsg_end(skb, nlh);
5147         return 0;
5148 }
5149
5150 /* called with rcu_read_lock() */
5151 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5152                           struct netlink_callback *cb, int s_ip_idx,
5153                           struct inet6_fill_args *fillargs)
5154 {
5155         struct ifmcaddr6 *ifmca;
5156         struct ifacaddr6 *ifaca;
5157         int ip_idx = 0;
5158         int err = 1;
5159
5160         read_lock_bh(&idev->lock);
5161         switch (fillargs->type) {
5162         case UNICAST_ADDR: {
5163                 struct inet6_ifaddr *ifa;
5164                 fillargs->event = RTM_NEWADDR;
5165
5166                 /* unicast address incl. temp addr */
5167                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5168                         if (ip_idx < s_ip_idx)
5169                                 goto next;
5170                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5171                         if (err < 0)
5172                                 break;
5173                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5174 next:
5175                         ip_idx++;
5176                 }
5177                 break;
5178         }
5179         case MULTICAST_ADDR:
5180                 read_unlock_bh(&idev->lock);
5181                 fillargs->event = RTM_GETMULTICAST;
5182
5183                 /* multicast address */
5184                 for (ifmca = rtnl_dereference(idev->mc_list);
5185                      ifmca;
5186                      ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5187                         if (ip_idx < s_ip_idx)
5188                                 continue;
5189                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5190                         if (err < 0)
5191                                 break;
5192                 }
5193                 read_lock_bh(&idev->lock);
5194                 break;
5195         case ANYCAST_ADDR:
5196                 fillargs->event = RTM_GETANYCAST;
5197                 /* anycast address */
5198                 for (ifaca = idev->ac_list; ifaca;
5199                      ifaca = ifaca->aca_next, ip_idx++) {
5200                         if (ip_idx < s_ip_idx)
5201                                 continue;
5202                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5203                         if (err < 0)
5204                                 break;
5205                 }
5206                 break;
5207         default:
5208                 break;
5209         }
5210         read_unlock_bh(&idev->lock);
5211         cb->args[2] = ip_idx;
5212         return err;
5213 }
5214
5215 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5216                                        struct inet6_fill_args *fillargs,
5217                                        struct net **tgt_net, struct sock *sk,
5218                                        struct netlink_callback *cb)
5219 {
5220         struct netlink_ext_ack *extack = cb->extack;
5221         struct nlattr *tb[IFA_MAX+1];
5222         struct ifaddrmsg *ifm;
5223         int err, i;
5224
5225         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5226                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5227                 return -EINVAL;
5228         }
5229
5230         ifm = nlmsg_data(nlh);
5231         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5232                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5233                 return -EINVAL;
5234         }
5235
5236         fillargs->ifindex = ifm->ifa_index;
5237         if (fillargs->ifindex) {
5238                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5239                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5240         }
5241
5242         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5243                                             ifa_ipv6_policy, extack);
5244         if (err < 0)
5245                 return err;
5246
5247         for (i = 0; i <= IFA_MAX; ++i) {
5248                 if (!tb[i])
5249                         continue;
5250
5251                 if (i == IFA_TARGET_NETNSID) {
5252                         struct net *net;
5253
5254                         fillargs->netnsid = nla_get_s32(tb[i]);
5255                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5256                         if (IS_ERR(net)) {
5257                                 fillargs->netnsid = -1;
5258                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5259                                 return PTR_ERR(net);
5260                         }
5261                         *tgt_net = net;
5262                 } else {
5263                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5264                         return -EINVAL;
5265                 }
5266         }
5267
5268         return 0;
5269 }
5270
5271 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5272                            enum addr_type_t type)
5273 {
5274         const struct nlmsghdr *nlh = cb->nlh;
5275         struct inet6_fill_args fillargs = {
5276                 .portid = NETLINK_CB(cb->skb).portid,
5277                 .seq = cb->nlh->nlmsg_seq,
5278                 .flags = NLM_F_MULTI,
5279                 .netnsid = -1,
5280                 .type = type,
5281         };
5282         struct net *tgt_net = sock_net(skb->sk);
5283         int idx, s_idx, s_ip_idx;
5284         int h, s_h;
5285         struct net_device *dev;
5286         struct inet6_dev *idev;
5287         struct hlist_head *head;
5288         int err = 0;
5289
5290         s_h = cb->args[0];
5291         s_idx = idx = cb->args[1];
5292         s_ip_idx = cb->args[2];
5293
5294         if (cb->strict_check) {
5295                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5296                                                   skb->sk, cb);
5297                 if (err < 0)
5298                         goto put_tgt_net;
5299
5300                 err = 0;
5301                 if (fillargs.ifindex) {
5302                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5303                         if (!dev) {
5304                                 err = -ENODEV;
5305                                 goto put_tgt_net;
5306                         }
5307                         idev = __in6_dev_get(dev);
5308                         if (idev) {
5309                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5310                                                      &fillargs);
5311                                 if (err > 0)
5312                                         err = 0;
5313                         }
5314                         goto put_tgt_net;
5315                 }
5316         }
5317
5318         rcu_read_lock();
5319         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5320         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5321                 idx = 0;
5322                 head = &tgt_net->dev_index_head[h];
5323                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5324                         if (idx < s_idx)
5325                                 goto cont;
5326                         if (h > s_h || idx > s_idx)
5327                                 s_ip_idx = 0;
5328                         idev = __in6_dev_get(dev);
5329                         if (!idev)
5330                                 goto cont;
5331
5332                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5333                                            &fillargs) < 0)
5334                                 goto done;
5335 cont:
5336                         idx++;
5337                 }
5338         }
5339 done:
5340         rcu_read_unlock();
5341         cb->args[0] = h;
5342         cb->args[1] = idx;
5343 put_tgt_net:
5344         if (fillargs.netnsid >= 0)
5345                 put_net(tgt_net);
5346
5347         return skb->len ? : err;
5348 }
5349
5350 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5351 {
5352         enum addr_type_t type = UNICAST_ADDR;
5353
5354         return inet6_dump_addr(skb, cb, type);
5355 }
5356
5357 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5358 {
5359         enum addr_type_t type = MULTICAST_ADDR;
5360
5361         return inet6_dump_addr(skb, cb, type);
5362 }
5363
5364
5365 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5366 {
5367         enum addr_type_t type = ANYCAST_ADDR;
5368
5369         return inet6_dump_addr(skb, cb, type);
5370 }
5371
5372 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5373                                        const struct nlmsghdr *nlh,
5374                                        struct nlattr **tb,
5375                                        struct netlink_ext_ack *extack)
5376 {
5377         struct ifaddrmsg *ifm;
5378         int i, err;
5379
5380         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5381                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5382                 return -EINVAL;
5383         }
5384
5385         if (!netlink_strict_get_check(skb))
5386                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5387                                               ifa_ipv6_policy, extack);
5388
5389         ifm = nlmsg_data(nlh);
5390         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5391                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5392                 return -EINVAL;
5393         }
5394
5395         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5396                                             ifa_ipv6_policy, extack);
5397         if (err)
5398                 return err;
5399
5400         for (i = 0; i <= IFA_MAX; i++) {
5401                 if (!tb[i])
5402                         continue;
5403
5404                 switch (i) {
5405                 case IFA_TARGET_NETNSID:
5406                 case IFA_ADDRESS:
5407                 case IFA_LOCAL:
5408                         break;
5409                 default:
5410                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5411                         return -EINVAL;
5412                 }
5413         }
5414
5415         return 0;
5416 }
5417
5418 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5419                              struct netlink_ext_ack *extack)
5420 {
5421         struct net *tgt_net = sock_net(in_skb->sk);
5422         struct inet6_fill_args fillargs = {
5423                 .portid = NETLINK_CB(in_skb).portid,
5424                 .seq = nlh->nlmsg_seq,
5425                 .event = RTM_NEWADDR,
5426                 .flags = 0,
5427                 .netnsid = -1,
5428         };
5429         struct ifaddrmsg *ifm;
5430         struct nlattr *tb[IFA_MAX+1];
5431         struct in6_addr *addr = NULL, *peer;
5432         struct net_device *dev = NULL;
5433         struct inet6_ifaddr *ifa;
5434         struct sk_buff *skb;
5435         int err;
5436
5437         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5438         if (err < 0)
5439                 return err;
5440
5441         if (tb[IFA_TARGET_NETNSID]) {
5442                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5443
5444                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5445                                                   fillargs.netnsid);
5446                 if (IS_ERR(tgt_net))
5447                         return PTR_ERR(tgt_net);
5448         }
5449
5450         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5451         if (!addr)
5452                 return -EINVAL;
5453
5454         ifm = nlmsg_data(nlh);
5455         if (ifm->ifa_index)
5456                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5457
5458         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5459         if (!ifa) {
5460                 err = -EADDRNOTAVAIL;
5461                 goto errout;
5462         }
5463
5464         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5465         if (!skb) {
5466                 err = -ENOBUFS;
5467                 goto errout_ifa;
5468         }
5469
5470         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5471         if (err < 0) {
5472                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5473                 WARN_ON(err == -EMSGSIZE);
5474                 kfree_skb(skb);
5475                 goto errout_ifa;
5476         }
5477         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5478 errout_ifa:
5479         in6_ifa_put(ifa);
5480 errout:
5481         dev_put(dev);
5482         if (fillargs.netnsid >= 0)
5483                 put_net(tgt_net);
5484
5485         return err;
5486 }
5487
5488 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5489 {
5490         struct sk_buff *skb;
5491         struct net *net = dev_net(ifa->idev->dev);
5492         struct inet6_fill_args fillargs = {
5493                 .portid = 0,
5494                 .seq = 0,
5495                 .event = event,
5496                 .flags = 0,
5497                 .netnsid = -1,
5498         };
5499         int err = -ENOBUFS;
5500
5501         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5502         if (!skb)
5503                 goto errout;
5504
5505         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5506         if (err < 0) {
5507                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5508                 WARN_ON(err == -EMSGSIZE);
5509                 kfree_skb(skb);
5510                 goto errout;
5511         }
5512         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5513         return;
5514 errout:
5515         if (err < 0)
5516                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5517 }
5518
5519 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5520                                 __s32 *array, int bytes)
5521 {
5522         BUG_ON(bytes < (DEVCONF_MAX * 4));
5523
5524         memset(array, 0, bytes);
5525         array[DEVCONF_FORWARDING] = cnf->forwarding;
5526         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5527         array[DEVCONF_MTU6] = cnf->mtu6;
5528         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5529         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5530         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5531         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5532         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5533         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5534                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5535         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5536                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5537         array[DEVCONF_RTR_SOLICIT_DELAY] =
5538                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5539         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5540         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5541                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5542         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5543                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5544         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5545         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5546         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5547         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5548         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5549         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5550         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5551         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5552         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5553         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5554 #ifdef CONFIG_IPV6_ROUTER_PREF
5555         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5556         array[DEVCONF_RTR_PROBE_INTERVAL] =
5557                 jiffies_to_msecs(cnf->rtr_probe_interval);
5558 #ifdef CONFIG_IPV6_ROUTE_INFO
5559         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5560         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5561 #endif
5562 #endif
5563         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5564         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5565 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5566         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5567         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5568 #endif
5569 #ifdef CONFIG_IPV6_MROUTE
5570         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5571 #endif
5572         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5573         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5574         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5575         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5576         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5577         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5578         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5579         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5580         /* we omit DEVCONF_STABLE_SECRET for now */
5581         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5582         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5583         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5584         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5585         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5586 #ifdef CONFIG_IPV6_SEG6_HMAC
5587         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5588 #endif
5589         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5590         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5591         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5592         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5593         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5594         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5595         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5596         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5597         array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft;
5598 }
5599
5600 static inline size_t inet6_ifla6_size(void)
5601 {
5602         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5603              + nla_total_size(sizeof(struct ifla_cacheinfo))
5604              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5605              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5606              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5607              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5608              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5609              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5610              + 0;
5611 }
5612
5613 static inline size_t inet6_if_nlmsg_size(void)
5614 {
5615         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5616                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5617                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5618                + nla_total_size(4) /* IFLA_MTU */
5619                + nla_total_size(4) /* IFLA_LINK */
5620                + nla_total_size(1) /* IFLA_OPERSTATE */
5621                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5622 }
5623
5624 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5625                                         int bytes)
5626 {
5627         int i;
5628         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5629         BUG_ON(pad < 0);
5630
5631         /* Use put_unaligned() because stats may not be aligned for u64. */
5632         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5633         for (i = 1; i < ICMP6_MIB_MAX; i++)
5634                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5635
5636         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5637 }
5638
5639 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5640                                         int bytes, size_t syncpoff)
5641 {
5642         int i, c;
5643         u64 buff[IPSTATS_MIB_MAX];
5644         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5645
5646         BUG_ON(pad < 0);
5647
5648         memset(buff, 0, sizeof(buff));
5649         buff[0] = IPSTATS_MIB_MAX;
5650
5651         for_each_possible_cpu(c) {
5652                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5653                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5654         }
5655
5656         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5657         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5658 }
5659
5660 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5661                              int bytes)
5662 {
5663         switch (attrtype) {
5664         case IFLA_INET6_STATS:
5665                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5666                                      offsetof(struct ipstats_mib, syncp));
5667                 break;
5668         case IFLA_INET6_ICMP6STATS:
5669                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5670                 break;
5671         }
5672 }
5673
5674 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5675                                   u32 ext_filter_mask)
5676 {
5677         struct nlattr *nla;
5678         struct ifla_cacheinfo ci;
5679
5680         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5681                 goto nla_put_failure;
5682         ci.max_reasm_len = IPV6_MAXPLEN;
5683         ci.tstamp = cstamp_delta(idev->tstamp);
5684         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5685         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5686         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5687                 goto nla_put_failure;
5688         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5689         if (!nla)
5690                 goto nla_put_failure;
5691         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5692
5693         /* XXX - MC not implemented */
5694
5695         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5696                 return 0;
5697
5698         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5699         if (!nla)
5700                 goto nla_put_failure;
5701         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5702
5703         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5704         if (!nla)
5705                 goto nla_put_failure;
5706         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5707
5708         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5709         if (!nla)
5710                 goto nla_put_failure;
5711         read_lock_bh(&idev->lock);
5712         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5713         read_unlock_bh(&idev->lock);
5714
5715         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5716                 goto nla_put_failure;
5717
5718         if (idev->ra_mtu &&
5719             nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5720                 goto nla_put_failure;
5721
5722         return 0;
5723
5724 nla_put_failure:
5725         return -EMSGSIZE;
5726 }
5727
5728 static size_t inet6_get_link_af_size(const struct net_device *dev,
5729                                      u32 ext_filter_mask)
5730 {
5731         if (!__in6_dev_get(dev))
5732                 return 0;
5733
5734         return inet6_ifla6_size();
5735 }
5736
5737 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5738                               u32 ext_filter_mask)
5739 {
5740         struct inet6_dev *idev = __in6_dev_get(dev);
5741
5742         if (!idev)
5743                 return -ENODATA;
5744
5745         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5746                 return -EMSGSIZE;
5747
5748         return 0;
5749 }
5750
5751 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5752                              struct netlink_ext_ack *extack)
5753 {
5754         struct inet6_ifaddr *ifp;
5755         struct net_device *dev = idev->dev;
5756         bool clear_token, update_rs = false;
5757         struct in6_addr ll_addr;
5758
5759         ASSERT_RTNL();
5760
5761         if (!token)
5762                 return -EINVAL;
5763
5764         if (dev->flags & IFF_LOOPBACK) {
5765                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5766                 return -EINVAL;
5767         }
5768
5769         if (dev->flags & IFF_NOARP) {
5770                 NL_SET_ERR_MSG_MOD(extack,
5771                                    "Device does not do neighbour discovery");
5772                 return -EINVAL;
5773         }
5774
5775         if (!ipv6_accept_ra(idev)) {
5776                 NL_SET_ERR_MSG_MOD(extack,
5777                                    "Router advertisement is disabled on device");
5778                 return -EINVAL;
5779         }
5780
5781         if (idev->cnf.rtr_solicits == 0) {
5782                 NL_SET_ERR_MSG(extack,
5783                                "Router solicitation is disabled on device");
5784                 return -EINVAL;
5785         }
5786
5787         write_lock_bh(&idev->lock);
5788
5789         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5790         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5791
5792         write_unlock_bh(&idev->lock);
5793
5794         clear_token = ipv6_addr_any(token);
5795         if (clear_token)
5796                 goto update_lft;
5797
5798         if (!idev->dead && (idev->if_flags & IF_READY) &&
5799             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5800                              IFA_F_OPTIMISTIC)) {
5801                 /* If we're not ready, then normal ifup will take care
5802                  * of this. Otherwise, we need to request our rs here.
5803                  */
5804                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5805                 update_rs = true;
5806         }
5807
5808 update_lft:
5809         write_lock_bh(&idev->lock);
5810
5811         if (update_rs) {
5812                 idev->if_flags |= IF_RS_SENT;
5813                 idev->rs_interval = rfc3315_s14_backoff_init(
5814                         idev->cnf.rtr_solicit_interval);
5815                 idev->rs_probes = 1;
5816                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5817         }
5818
5819         /* Well, that's kinda nasty ... */
5820         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5821                 spin_lock(&ifp->lock);
5822                 if (ifp->tokenized) {
5823                         ifp->valid_lft = 0;
5824                         ifp->prefered_lft = 0;
5825                 }
5826                 spin_unlock(&ifp->lock);
5827         }
5828
5829         write_unlock_bh(&idev->lock);
5830         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5831         addrconf_verify_rtnl();
5832         return 0;
5833 }
5834
5835 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5836         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5837         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5838         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5839                                             .reject_message =
5840                                                 "IFLA_INET6_RA_MTU can not be set" },
5841 };
5842
5843 static int check_addr_gen_mode(int mode)
5844 {
5845         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5846             mode != IN6_ADDR_GEN_MODE_NONE &&
5847             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5848             mode != IN6_ADDR_GEN_MODE_RANDOM)
5849                 return -EINVAL;
5850         return 1;
5851 }
5852
5853 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5854                                 int mode)
5855 {
5856         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5857             !idev->cnf.stable_secret.initialized &&
5858             !net->ipv6.devconf_dflt->stable_secret.initialized)
5859                 return -EINVAL;
5860         return 1;
5861 }
5862
5863 static int inet6_validate_link_af(const struct net_device *dev,
5864                                   const struct nlattr *nla,
5865                                   struct netlink_ext_ack *extack)
5866 {
5867         struct nlattr *tb[IFLA_INET6_MAX + 1];
5868         struct inet6_dev *idev = NULL;
5869         int err;
5870
5871         if (dev) {
5872                 idev = __in6_dev_get(dev);
5873                 if (!idev)
5874                         return -EAFNOSUPPORT;
5875         }
5876
5877         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5878                                           inet6_af_policy, extack);
5879         if (err)
5880                 return err;
5881
5882         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5883                 return -EINVAL;
5884
5885         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5886                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5887
5888                 if (check_addr_gen_mode(mode) < 0)
5889                         return -EINVAL;
5890                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5891                         return -EINVAL;
5892         }
5893
5894         return 0;
5895 }
5896
5897 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5898                              struct netlink_ext_ack *extack)
5899 {
5900         struct inet6_dev *idev = __in6_dev_get(dev);
5901         struct nlattr *tb[IFLA_INET6_MAX + 1];
5902         int err;
5903
5904         if (!idev)
5905                 return -EAFNOSUPPORT;
5906
5907         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5908                 return -EINVAL;
5909
5910         if (tb[IFLA_INET6_TOKEN]) {
5911                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5912                                         extack);
5913                 if (err)
5914                         return err;
5915         }
5916
5917         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5918                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5919
5920                 idev->cnf.addr_gen_mode = mode;
5921         }
5922
5923         return 0;
5924 }
5925
5926 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5927                              u32 portid, u32 seq, int event, unsigned int flags)
5928 {
5929         struct net_device *dev = idev->dev;
5930         struct ifinfomsg *hdr;
5931         struct nlmsghdr *nlh;
5932         void *protoinfo;
5933
5934         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5935         if (!nlh)
5936                 return -EMSGSIZE;
5937
5938         hdr = nlmsg_data(nlh);
5939         hdr->ifi_family = AF_INET6;
5940         hdr->__ifi_pad = 0;
5941         hdr->ifi_type = dev->type;
5942         hdr->ifi_index = dev->ifindex;
5943         hdr->ifi_flags = dev_get_flags(dev);
5944         hdr->ifi_change = 0;
5945
5946         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5947             (dev->addr_len &&
5948              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5949             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5950             (dev->ifindex != dev_get_iflink(dev) &&
5951              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5952             nla_put_u8(skb, IFLA_OPERSTATE,
5953                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5954                 goto nla_put_failure;
5955         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5956         if (!protoinfo)
5957                 goto nla_put_failure;
5958
5959         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5960                 goto nla_put_failure;
5961
5962         nla_nest_end(skb, protoinfo);
5963         nlmsg_end(skb, nlh);
5964         return 0;
5965
5966 nla_put_failure:
5967         nlmsg_cancel(skb, nlh);
5968         return -EMSGSIZE;
5969 }
5970
5971 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5972                                    struct netlink_ext_ack *extack)
5973 {
5974         struct ifinfomsg *ifm;
5975
5976         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5977                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5978                 return -EINVAL;
5979         }
5980
5981         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5982                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5983                 return -EINVAL;
5984         }
5985
5986         ifm = nlmsg_data(nlh);
5987         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5988             ifm->ifi_change || ifm->ifi_index) {
5989                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5990                 return -EINVAL;
5991         }
5992
5993         return 0;
5994 }
5995
5996 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5997 {
5998         struct net *net = sock_net(skb->sk);
5999         int h, s_h;
6000         int idx = 0, s_idx;
6001         struct net_device *dev;
6002         struct inet6_dev *idev;
6003         struct hlist_head *head;
6004
6005         /* only requests using strict checking can pass data to
6006          * influence the dump
6007          */
6008         if (cb->strict_check) {
6009                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6010
6011                 if (err < 0)
6012                         return err;
6013         }
6014
6015         s_h = cb->args[0];
6016         s_idx = cb->args[1];
6017
6018         rcu_read_lock();
6019         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6020                 idx = 0;
6021                 head = &net->dev_index_head[h];
6022                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
6023                         if (idx < s_idx)
6024                                 goto cont;
6025                         idev = __in6_dev_get(dev);
6026                         if (!idev)
6027                                 goto cont;
6028                         if (inet6_fill_ifinfo(skb, idev,
6029                                               NETLINK_CB(cb->skb).portid,
6030                                               cb->nlh->nlmsg_seq,
6031                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
6032                                 goto out;
6033 cont:
6034                         idx++;
6035                 }
6036         }
6037 out:
6038         rcu_read_unlock();
6039         cb->args[1] = idx;
6040         cb->args[0] = h;
6041
6042         return skb->len;
6043 }
6044
6045 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6046 {
6047         struct sk_buff *skb;
6048         struct net *net = dev_net(idev->dev);
6049         int err = -ENOBUFS;
6050
6051         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6052         if (!skb)
6053                 goto errout;
6054
6055         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6056         if (err < 0) {
6057                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6058                 WARN_ON(err == -EMSGSIZE);
6059                 kfree_skb(skb);
6060                 goto errout;
6061         }
6062         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6063         return;
6064 errout:
6065         if (err < 0)
6066                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6067 }
6068
6069 static inline size_t inet6_prefix_nlmsg_size(void)
6070 {
6071         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6072                + nla_total_size(sizeof(struct in6_addr))
6073                + nla_total_size(sizeof(struct prefix_cacheinfo));
6074 }
6075
6076 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6077                              struct prefix_info *pinfo, u32 portid, u32 seq,
6078                              int event, unsigned int flags)
6079 {
6080         struct prefixmsg *pmsg;
6081         struct nlmsghdr *nlh;
6082         struct prefix_cacheinfo ci;
6083
6084         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6085         if (!nlh)
6086                 return -EMSGSIZE;
6087
6088         pmsg = nlmsg_data(nlh);
6089         pmsg->prefix_family = AF_INET6;
6090         pmsg->prefix_pad1 = 0;
6091         pmsg->prefix_pad2 = 0;
6092         pmsg->prefix_ifindex = idev->dev->ifindex;
6093         pmsg->prefix_len = pinfo->prefix_len;
6094         pmsg->prefix_type = pinfo->type;
6095         pmsg->prefix_pad3 = 0;
6096         pmsg->prefix_flags = 0;
6097         if (pinfo->onlink)
6098                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6099         if (pinfo->autoconf)
6100                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6101
6102         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6103                 goto nla_put_failure;
6104         ci.preferred_time = ntohl(pinfo->prefered);
6105         ci.valid_time = ntohl(pinfo->valid);
6106         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6107                 goto nla_put_failure;
6108         nlmsg_end(skb, nlh);
6109         return 0;
6110
6111 nla_put_failure:
6112         nlmsg_cancel(skb, nlh);
6113         return -EMSGSIZE;
6114 }
6115
6116 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6117                          struct prefix_info *pinfo)
6118 {
6119         struct sk_buff *skb;
6120         struct net *net = dev_net(idev->dev);
6121         int err = -ENOBUFS;
6122
6123         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6124         if (!skb)
6125                 goto errout;
6126
6127         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6128         if (err < 0) {
6129                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6130                 WARN_ON(err == -EMSGSIZE);
6131                 kfree_skb(skb);
6132                 goto errout;
6133         }
6134         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6135         return;
6136 errout:
6137         if (err < 0)
6138                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6139 }
6140
6141 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6142 {
6143         struct net *net = dev_net(ifp->idev->dev);
6144
6145         if (event)
6146                 ASSERT_RTNL();
6147
6148         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6149
6150         switch (event) {
6151         case RTM_NEWADDR:
6152                 /*
6153                  * If the address was optimistic we inserted the route at the
6154                  * start of our DAD process, so we don't need to do it again.
6155                  * If the device was taken down in the middle of the DAD
6156                  * cycle there is a race where we could get here without a
6157                  * host route, so nothing to insert. That will be fixed when
6158                  * the device is brought up.
6159                  */
6160                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6161                         ip6_ins_rt(net, ifp->rt);
6162                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6163                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6164                                 &ifp->addr, ifp->idev->dev->name);
6165                 }
6166
6167                 if (ifp->idev->cnf.forwarding)
6168                         addrconf_join_anycast(ifp);
6169                 if (!ipv6_addr_any(&ifp->peer_addr))
6170                         addrconf_prefix_route(&ifp->peer_addr, 128,
6171                                               ifp->rt_priority, ifp->idev->dev,
6172                                               0, 0, GFP_ATOMIC);
6173                 break;
6174         case RTM_DELADDR:
6175                 if (ifp->idev->cnf.forwarding)
6176                         addrconf_leave_anycast(ifp);
6177                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6178                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6179                         struct fib6_info *rt;
6180
6181                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6182                                                        ifp->idev->dev, 0, 0,
6183                                                        false);
6184                         if (rt)
6185                                 ip6_del_rt(net, rt, false);
6186                 }
6187                 if (ifp->rt) {
6188                         ip6_del_rt(net, ifp->rt, false);
6189                         ifp->rt = NULL;
6190                 }
6191                 rt_genid_bump_ipv6(net);
6192                 break;
6193         }
6194         atomic_inc(&net->ipv6.dev_addr_genid);
6195 }
6196
6197 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6198 {
6199         if (likely(ifp->idev->dead == 0))
6200                 __ipv6_ifa_notify(event, ifp);
6201 }
6202
6203 #ifdef CONFIG_SYSCTL
6204
6205 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6206                 void *buffer, size_t *lenp, loff_t *ppos)
6207 {
6208         int *valp = ctl->data;
6209         int val = *valp;
6210         loff_t pos = *ppos;
6211         struct ctl_table lctl;
6212         int ret;
6213
6214         /*
6215          * ctl->data points to idev->cnf.forwarding, we should
6216          * not modify it until we get the rtnl lock.
6217          */
6218         lctl = *ctl;
6219         lctl.data = &val;
6220
6221         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6222
6223         if (write)
6224                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6225         if (ret)
6226                 *ppos = pos;
6227         return ret;
6228 }
6229
6230 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6231                 void *buffer, size_t *lenp, loff_t *ppos)
6232 {
6233         struct inet6_dev *idev = ctl->extra1;
6234         int min_mtu = IPV6_MIN_MTU;
6235         struct ctl_table lctl;
6236
6237         lctl = *ctl;
6238         lctl.extra1 = &min_mtu;
6239         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6240
6241         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6242 }
6243
6244 static void dev_disable_change(struct inet6_dev *idev)
6245 {
6246         struct netdev_notifier_info info;
6247
6248         if (!idev || !idev->dev)
6249                 return;
6250
6251         netdev_notifier_info_init(&info, idev->dev);
6252         if (idev->cnf.disable_ipv6)
6253                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6254         else
6255                 addrconf_notify(NULL, NETDEV_UP, &info);
6256 }
6257
6258 static void addrconf_disable_change(struct net *net, __s32 newf)
6259 {
6260         struct net_device *dev;
6261         struct inet6_dev *idev;
6262
6263         for_each_netdev(net, dev) {
6264                 idev = __in6_dev_get(dev);
6265                 if (idev) {
6266                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6267                         idev->cnf.disable_ipv6 = newf;
6268                         if (changed)
6269                                 dev_disable_change(idev);
6270                 }
6271         }
6272 }
6273
6274 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6275 {
6276         struct net *net;
6277         int old;
6278
6279         if (!rtnl_trylock())
6280                 return restart_syscall();
6281
6282         net = (struct net *)table->extra2;
6283         old = *p;
6284         *p = newf;
6285
6286         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6287                 rtnl_unlock();
6288                 return 0;
6289         }
6290
6291         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6292                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6293                 addrconf_disable_change(net, newf);
6294         } else if ((!newf) ^ (!old))
6295                 dev_disable_change((struct inet6_dev *)table->extra1);
6296
6297         rtnl_unlock();
6298         return 0;
6299 }
6300
6301 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6302                 void *buffer, size_t *lenp, loff_t *ppos)
6303 {
6304         int *valp = ctl->data;
6305         int val = *valp;
6306         loff_t pos = *ppos;
6307         struct ctl_table lctl;
6308         int ret;
6309
6310         /*
6311          * ctl->data points to idev->cnf.disable_ipv6, we should
6312          * not modify it until we get the rtnl lock.
6313          */
6314         lctl = *ctl;
6315         lctl.data = &val;
6316
6317         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6318
6319         if (write)
6320                 ret = addrconf_disable_ipv6(ctl, valp, val);
6321         if (ret)
6322                 *ppos = pos;
6323         return ret;
6324 }
6325
6326 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6327                 void *buffer, size_t *lenp, loff_t *ppos)
6328 {
6329         int *valp = ctl->data;
6330         int ret;
6331         int old, new;
6332
6333         old = *valp;
6334         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6335         new = *valp;
6336
6337         if (write && old != new) {
6338                 struct net *net = ctl->extra2;
6339
6340                 if (!rtnl_trylock())
6341                         return restart_syscall();
6342
6343                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6344                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6345                                                      NETCONFA_PROXY_NEIGH,
6346                                                      NETCONFA_IFINDEX_DEFAULT,
6347                                                      net->ipv6.devconf_dflt);
6348                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6349                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6350                                                      NETCONFA_PROXY_NEIGH,
6351                                                      NETCONFA_IFINDEX_ALL,
6352                                                      net->ipv6.devconf_all);
6353                 else {
6354                         struct inet6_dev *idev = ctl->extra1;
6355
6356                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6357                                                      NETCONFA_PROXY_NEIGH,
6358                                                      idev->dev->ifindex,
6359                                                      &idev->cnf);
6360                 }
6361                 rtnl_unlock();
6362         }
6363
6364         return ret;
6365 }
6366
6367 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6368                                          void *buffer, size_t *lenp,
6369                                          loff_t *ppos)
6370 {
6371         int ret = 0;
6372         u32 new_val;
6373         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6374         struct net *net = (struct net *)ctl->extra2;
6375         struct ctl_table tmp = {
6376                 .data = &new_val,
6377                 .maxlen = sizeof(new_val),
6378                 .mode = ctl->mode,
6379         };
6380
6381         if (!rtnl_trylock())
6382                 return restart_syscall();
6383
6384         new_val = *((u32 *)ctl->data);
6385
6386         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6387         if (ret != 0)
6388                 goto out;
6389
6390         if (write) {
6391                 if (check_addr_gen_mode(new_val) < 0) {
6392                         ret = -EINVAL;
6393                         goto out;
6394                 }
6395
6396                 if (idev) {
6397                         if (check_stable_privacy(idev, net, new_val) < 0) {
6398                                 ret = -EINVAL;
6399                                 goto out;
6400                         }
6401
6402                         if (idev->cnf.addr_gen_mode != new_val) {
6403                                 idev->cnf.addr_gen_mode = new_val;
6404                                 addrconf_init_auto_addrs(idev->dev);
6405                         }
6406                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6407                         struct net_device *dev;
6408
6409                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6410                         for_each_netdev(net, dev) {
6411                                 idev = __in6_dev_get(dev);
6412                                 if (idev &&
6413                                     idev->cnf.addr_gen_mode != new_val) {
6414                                         idev->cnf.addr_gen_mode = new_val;
6415                                         addrconf_init_auto_addrs(idev->dev);
6416                                 }
6417                         }
6418                 }
6419
6420                 *((u32 *)ctl->data) = new_val;
6421         }
6422
6423 out:
6424         rtnl_unlock();
6425
6426         return ret;
6427 }
6428
6429 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6430                                          void *buffer, size_t *lenp,
6431                                          loff_t *ppos)
6432 {
6433         int err;
6434         struct in6_addr addr;
6435         char str[IPV6_MAX_STRLEN];
6436         struct ctl_table lctl = *ctl;
6437         struct net *net = ctl->extra2;
6438         struct ipv6_stable_secret *secret = ctl->data;
6439
6440         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6441                 return -EIO;
6442
6443         lctl.maxlen = IPV6_MAX_STRLEN;
6444         lctl.data = str;
6445
6446         if (!rtnl_trylock())
6447                 return restart_syscall();
6448
6449         if (!write && !secret->initialized) {
6450                 err = -EIO;
6451                 goto out;
6452         }
6453
6454         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6455         if (err >= sizeof(str)) {
6456                 err = -EIO;
6457                 goto out;
6458         }
6459
6460         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6461         if (err || !write)
6462                 goto out;
6463
6464         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6465                 err = -EIO;
6466                 goto out;
6467         }
6468
6469         secret->initialized = true;
6470         secret->secret = addr;
6471
6472         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6473                 struct net_device *dev;
6474
6475                 for_each_netdev(net, dev) {
6476                         struct inet6_dev *idev = __in6_dev_get(dev);
6477
6478                         if (idev) {
6479                                 idev->cnf.addr_gen_mode =
6480                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6481                         }
6482                 }
6483         } else {
6484                 struct inet6_dev *idev = ctl->extra1;
6485
6486                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6487         }
6488
6489 out:
6490         rtnl_unlock();
6491
6492         return err;
6493 }
6494
6495 static
6496 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6497                                                 int write, void *buffer,
6498                                                 size_t *lenp,
6499                                                 loff_t *ppos)
6500 {
6501         int *valp = ctl->data;
6502         int val = *valp;
6503         loff_t pos = *ppos;
6504         struct ctl_table lctl;
6505         int ret;
6506
6507         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6508          * we should not modify it until we get the rtnl lock.
6509          */
6510         lctl = *ctl;
6511         lctl.data = &val;
6512
6513         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6514
6515         if (write)
6516                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6517         if (ret)
6518                 *ppos = pos;
6519         return ret;
6520 }
6521
6522 static
6523 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6524 {
6525         if (rt) {
6526                 if (action)
6527                         rt->dst.flags |= DST_NOPOLICY;
6528                 else
6529                         rt->dst.flags &= ~DST_NOPOLICY;
6530         }
6531 }
6532
6533 static
6534 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6535 {
6536         struct inet6_ifaddr *ifa;
6537
6538         read_lock_bh(&idev->lock);
6539         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6540                 spin_lock(&ifa->lock);
6541                 if (ifa->rt) {
6542                         /* host routes only use builtin fib6_nh */
6543                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6544                         int cpu;
6545
6546                         rcu_read_lock();
6547                         ifa->rt->dst_nopolicy = val ? true : false;
6548                         if (nh->rt6i_pcpu) {
6549                                 for_each_possible_cpu(cpu) {
6550                                         struct rt6_info **rtp;
6551
6552                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6553                                         addrconf_set_nopolicy(*rtp, val);
6554                                 }
6555                         }
6556                         rcu_read_unlock();
6557                 }
6558                 spin_unlock(&ifa->lock);
6559         }
6560         read_unlock_bh(&idev->lock);
6561 }
6562
6563 static
6564 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6565 {
6566         struct inet6_dev *idev;
6567         struct net *net;
6568
6569         if (!rtnl_trylock())
6570                 return restart_syscall();
6571
6572         *valp = val;
6573
6574         net = (struct net *)ctl->extra2;
6575         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6576                 rtnl_unlock();
6577                 return 0;
6578         }
6579
6580         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6581                 struct net_device *dev;
6582
6583                 for_each_netdev(net, dev) {
6584                         idev = __in6_dev_get(dev);
6585                         if (idev)
6586                                 addrconf_disable_policy_idev(idev, val);
6587                 }
6588         } else {
6589                 idev = (struct inet6_dev *)ctl->extra1;
6590                 addrconf_disable_policy_idev(idev, val);
6591         }
6592
6593         rtnl_unlock();
6594         return 0;
6595 }
6596
6597 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6598                                    void *buffer, size_t *lenp, loff_t *ppos)
6599 {
6600         int *valp = ctl->data;
6601         int val = *valp;
6602         loff_t pos = *ppos;
6603         struct ctl_table lctl;
6604         int ret;
6605
6606         lctl = *ctl;
6607         lctl.data = &val;
6608         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6609
6610         if (write && (*valp != val))
6611                 ret = addrconf_disable_policy(ctl, valp, val);
6612
6613         if (ret)
6614                 *ppos = pos;
6615
6616         return ret;
6617 }
6618
6619 static int minus_one = -1;
6620 static const int two_five_five = 255;
6621 static u32 ioam6_if_id_max = U16_MAX;
6622
6623 static const struct ctl_table addrconf_sysctl[] = {
6624         {
6625                 .procname       = "forwarding",
6626                 .data           = &ipv6_devconf.forwarding,
6627                 .maxlen         = sizeof(int),
6628                 .mode           = 0644,
6629                 .proc_handler   = addrconf_sysctl_forward,
6630         },
6631         {
6632                 .procname       = "hop_limit",
6633                 .data           = &ipv6_devconf.hop_limit,
6634                 .maxlen         = sizeof(int),
6635                 .mode           = 0644,
6636                 .proc_handler   = proc_dointvec_minmax,
6637                 .extra1         = (void *)SYSCTL_ONE,
6638                 .extra2         = (void *)&two_five_five,
6639         },
6640         {
6641                 .procname       = "mtu",
6642                 .data           = &ipv6_devconf.mtu6,
6643                 .maxlen         = sizeof(int),
6644                 .mode           = 0644,
6645                 .proc_handler   = addrconf_sysctl_mtu,
6646         },
6647         {
6648                 .procname       = "accept_ra",
6649                 .data           = &ipv6_devconf.accept_ra,
6650                 .maxlen         = sizeof(int),
6651                 .mode           = 0644,
6652                 .proc_handler   = proc_dointvec,
6653         },
6654         {
6655                 .procname       = "accept_redirects",
6656                 .data           = &ipv6_devconf.accept_redirects,
6657                 .maxlen         = sizeof(int),
6658                 .mode           = 0644,
6659                 .proc_handler   = proc_dointvec,
6660         },
6661         {
6662                 .procname       = "autoconf",
6663                 .data           = &ipv6_devconf.autoconf,
6664                 .maxlen         = sizeof(int),
6665                 .mode           = 0644,
6666                 .proc_handler   = proc_dointvec,
6667         },
6668         {
6669                 .procname       = "dad_transmits",
6670                 .data           = &ipv6_devconf.dad_transmits,
6671                 .maxlen         = sizeof(int),
6672                 .mode           = 0644,
6673                 .proc_handler   = proc_dointvec,
6674         },
6675         {
6676                 .procname       = "router_solicitations",
6677                 .data           = &ipv6_devconf.rtr_solicits,
6678                 .maxlen         = sizeof(int),
6679                 .mode           = 0644,
6680                 .proc_handler   = proc_dointvec_minmax,
6681                 .extra1         = &minus_one,
6682         },
6683         {
6684                 .procname       = "router_solicitation_interval",
6685                 .data           = &ipv6_devconf.rtr_solicit_interval,
6686                 .maxlen         = sizeof(int),
6687                 .mode           = 0644,
6688                 .proc_handler   = proc_dointvec_jiffies,
6689         },
6690         {
6691                 .procname       = "router_solicitation_max_interval",
6692                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6693                 .maxlen         = sizeof(int),
6694                 .mode           = 0644,
6695                 .proc_handler   = proc_dointvec_jiffies,
6696         },
6697         {
6698                 .procname       = "router_solicitation_delay",
6699                 .data           = &ipv6_devconf.rtr_solicit_delay,
6700                 .maxlen         = sizeof(int),
6701                 .mode           = 0644,
6702                 .proc_handler   = proc_dointvec_jiffies,
6703         },
6704         {
6705                 .procname       = "force_mld_version",
6706                 .data           = &ipv6_devconf.force_mld_version,
6707                 .maxlen         = sizeof(int),
6708                 .mode           = 0644,
6709                 .proc_handler   = proc_dointvec,
6710         },
6711         {
6712                 .procname       = "mldv1_unsolicited_report_interval",
6713                 .data           =
6714                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6715                 .maxlen         = sizeof(int),
6716                 .mode           = 0644,
6717                 .proc_handler   = proc_dointvec_ms_jiffies,
6718         },
6719         {
6720                 .procname       = "mldv2_unsolicited_report_interval",
6721                 .data           =
6722                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6723                 .maxlen         = sizeof(int),
6724                 .mode           = 0644,
6725                 .proc_handler   = proc_dointvec_ms_jiffies,
6726         },
6727         {
6728                 .procname       = "use_tempaddr",
6729                 .data           = &ipv6_devconf.use_tempaddr,
6730                 .maxlen         = sizeof(int),
6731                 .mode           = 0644,
6732                 .proc_handler   = proc_dointvec,
6733         },
6734         {
6735                 .procname       = "temp_valid_lft",
6736                 .data           = &ipv6_devconf.temp_valid_lft,
6737                 .maxlen         = sizeof(int),
6738                 .mode           = 0644,
6739                 .proc_handler   = proc_dointvec,
6740         },
6741         {
6742                 .procname       = "temp_prefered_lft",
6743                 .data           = &ipv6_devconf.temp_prefered_lft,
6744                 .maxlen         = sizeof(int),
6745                 .mode           = 0644,
6746                 .proc_handler   = proc_dointvec,
6747         },
6748         {
6749                 .procname       = "regen_max_retry",
6750                 .data           = &ipv6_devconf.regen_max_retry,
6751                 .maxlen         = sizeof(int),
6752                 .mode           = 0644,
6753                 .proc_handler   = proc_dointvec,
6754         },
6755         {
6756                 .procname       = "max_desync_factor",
6757                 .data           = &ipv6_devconf.max_desync_factor,
6758                 .maxlen         = sizeof(int),
6759                 .mode           = 0644,
6760                 .proc_handler   = proc_dointvec,
6761         },
6762         {
6763                 .procname       = "max_addresses",
6764                 .data           = &ipv6_devconf.max_addresses,
6765                 .maxlen         = sizeof(int),
6766                 .mode           = 0644,
6767                 .proc_handler   = proc_dointvec,
6768         },
6769         {
6770                 .procname       = "accept_ra_defrtr",
6771                 .data           = &ipv6_devconf.accept_ra_defrtr,
6772                 .maxlen         = sizeof(int),
6773                 .mode           = 0644,
6774                 .proc_handler   = proc_dointvec,
6775         },
6776         {
6777                 .procname       = "ra_defrtr_metric",
6778                 .data           = &ipv6_devconf.ra_defrtr_metric,
6779                 .maxlen         = sizeof(u32),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_douintvec_minmax,
6782                 .extra1         = (void *)SYSCTL_ONE,
6783         },
6784         {
6785                 .procname       = "accept_ra_min_hop_limit",
6786                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6787                 .maxlen         = sizeof(int),
6788                 .mode           = 0644,
6789                 .proc_handler   = proc_dointvec,
6790         },
6791         {
6792                 .procname       = "accept_ra_min_lft",
6793                 .data           = &ipv6_devconf.accept_ra_min_lft,
6794                 .maxlen         = sizeof(int),
6795                 .mode           = 0644,
6796                 .proc_handler   = proc_dointvec,
6797         },
6798         {
6799                 .procname       = "accept_ra_pinfo",
6800                 .data           = &ipv6_devconf.accept_ra_pinfo,
6801                 .maxlen         = sizeof(int),
6802                 .mode           = 0644,
6803                 .proc_handler   = proc_dointvec,
6804         },
6805 #ifdef CONFIG_IPV6_ROUTER_PREF
6806         {
6807                 .procname       = "accept_ra_rtr_pref",
6808                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6809                 .maxlen         = sizeof(int),
6810                 .mode           = 0644,
6811                 .proc_handler   = proc_dointvec,
6812         },
6813         {
6814                 .procname       = "router_probe_interval",
6815                 .data           = &ipv6_devconf.rtr_probe_interval,
6816                 .maxlen         = sizeof(int),
6817                 .mode           = 0644,
6818                 .proc_handler   = proc_dointvec_jiffies,
6819         },
6820 #ifdef CONFIG_IPV6_ROUTE_INFO
6821         {
6822                 .procname       = "accept_ra_rt_info_min_plen",
6823                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6824                 .maxlen         = sizeof(int),
6825                 .mode           = 0644,
6826                 .proc_handler   = proc_dointvec,
6827         },
6828         {
6829                 .procname       = "accept_ra_rt_info_max_plen",
6830                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6831                 .maxlen         = sizeof(int),
6832                 .mode           = 0644,
6833                 .proc_handler   = proc_dointvec,
6834         },
6835 #endif
6836 #endif
6837         {
6838                 .procname       = "proxy_ndp",
6839                 .data           = &ipv6_devconf.proxy_ndp,
6840                 .maxlen         = sizeof(int),
6841                 .mode           = 0644,
6842                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6843         },
6844         {
6845                 .procname       = "accept_source_route",
6846                 .data           = &ipv6_devconf.accept_source_route,
6847                 .maxlen         = sizeof(int),
6848                 .mode           = 0644,
6849                 .proc_handler   = proc_dointvec,
6850         },
6851 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6852         {
6853                 .procname       = "optimistic_dad",
6854                 .data           = &ipv6_devconf.optimistic_dad,
6855                 .maxlen         = sizeof(int),
6856                 .mode           = 0644,
6857                 .proc_handler   = proc_dointvec,
6858         },
6859         {
6860                 .procname       = "use_optimistic",
6861                 .data           = &ipv6_devconf.use_optimistic,
6862                 .maxlen         = sizeof(int),
6863                 .mode           = 0644,
6864                 .proc_handler   = proc_dointvec,
6865         },
6866 #endif
6867 #ifdef CONFIG_IPV6_MROUTE
6868         {
6869                 .procname       = "mc_forwarding",
6870                 .data           = &ipv6_devconf.mc_forwarding,
6871                 .maxlen         = sizeof(int),
6872                 .mode           = 0444,
6873                 .proc_handler   = proc_dointvec,
6874         },
6875 #endif
6876         {
6877                 .procname       = "disable_ipv6",
6878                 .data           = &ipv6_devconf.disable_ipv6,
6879                 .maxlen         = sizeof(int),
6880                 .mode           = 0644,
6881                 .proc_handler   = addrconf_sysctl_disable,
6882         },
6883         {
6884                 .procname       = "accept_dad",
6885                 .data           = &ipv6_devconf.accept_dad,
6886                 .maxlen         = sizeof(int),
6887                 .mode           = 0644,
6888                 .proc_handler   = proc_dointvec,
6889         },
6890         {
6891                 .procname       = "force_tllao",
6892                 .data           = &ipv6_devconf.force_tllao,
6893                 .maxlen         = sizeof(int),
6894                 .mode           = 0644,
6895                 .proc_handler   = proc_dointvec
6896         },
6897         {
6898                 .procname       = "ndisc_notify",
6899                 .data           = &ipv6_devconf.ndisc_notify,
6900                 .maxlen         = sizeof(int),
6901                 .mode           = 0644,
6902                 .proc_handler   = proc_dointvec
6903         },
6904         {
6905                 .procname       = "suppress_frag_ndisc",
6906                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6907                 .maxlen         = sizeof(int),
6908                 .mode           = 0644,
6909                 .proc_handler   = proc_dointvec
6910         },
6911         {
6912                 .procname       = "accept_ra_from_local",
6913                 .data           = &ipv6_devconf.accept_ra_from_local,
6914                 .maxlen         = sizeof(int),
6915                 .mode           = 0644,
6916                 .proc_handler   = proc_dointvec,
6917         },
6918         {
6919                 .procname       = "accept_ra_mtu",
6920                 .data           = &ipv6_devconf.accept_ra_mtu,
6921                 .maxlen         = sizeof(int),
6922                 .mode           = 0644,
6923                 .proc_handler   = proc_dointvec,
6924         },
6925         {
6926                 .procname       = "stable_secret",
6927                 .data           = &ipv6_devconf.stable_secret,
6928                 .maxlen         = IPV6_MAX_STRLEN,
6929                 .mode           = 0600,
6930                 .proc_handler   = addrconf_sysctl_stable_secret,
6931         },
6932         {
6933                 .procname       = "use_oif_addrs_only",
6934                 .data           = &ipv6_devconf.use_oif_addrs_only,
6935                 .maxlen         = sizeof(int),
6936                 .mode           = 0644,
6937                 .proc_handler   = proc_dointvec,
6938         },
6939         {
6940                 .procname       = "ignore_routes_with_linkdown",
6941                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6942                 .maxlen         = sizeof(int),
6943                 .mode           = 0644,
6944                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6945         },
6946         {
6947                 .procname       = "drop_unicast_in_l2_multicast",
6948                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6949                 .maxlen         = sizeof(int),
6950                 .mode           = 0644,
6951                 .proc_handler   = proc_dointvec,
6952         },
6953         {
6954                 .procname       = "drop_unsolicited_na",
6955                 .data           = &ipv6_devconf.drop_unsolicited_na,
6956                 .maxlen         = sizeof(int),
6957                 .mode           = 0644,
6958                 .proc_handler   = proc_dointvec,
6959         },
6960         {
6961                 .procname       = "keep_addr_on_down",
6962                 .data           = &ipv6_devconf.keep_addr_on_down,
6963                 .maxlen         = sizeof(int),
6964                 .mode           = 0644,
6965                 .proc_handler   = proc_dointvec,
6966
6967         },
6968         {
6969                 .procname       = "seg6_enabled",
6970                 .data           = &ipv6_devconf.seg6_enabled,
6971                 .maxlen         = sizeof(int),
6972                 .mode           = 0644,
6973                 .proc_handler   = proc_dointvec,
6974         },
6975 #ifdef CONFIG_IPV6_SEG6_HMAC
6976         {
6977                 .procname       = "seg6_require_hmac",
6978                 .data           = &ipv6_devconf.seg6_require_hmac,
6979                 .maxlen         = sizeof(int),
6980                 .mode           = 0644,
6981                 .proc_handler   = proc_dointvec,
6982         },
6983 #endif
6984         {
6985                 .procname       = "enhanced_dad",
6986                 .data           = &ipv6_devconf.enhanced_dad,
6987                 .maxlen         = sizeof(int),
6988                 .mode           = 0644,
6989                 .proc_handler   = proc_dointvec,
6990         },
6991         {
6992                 .procname       = "addr_gen_mode",
6993                 .data           = &ipv6_devconf.addr_gen_mode,
6994                 .maxlen         = sizeof(int),
6995                 .mode           = 0644,
6996                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6997         },
6998         {
6999                 .procname       = "disable_policy",
7000                 .data           = &ipv6_devconf.disable_policy,
7001                 .maxlen         = sizeof(int),
7002                 .mode           = 0644,
7003                 .proc_handler   = addrconf_sysctl_disable_policy,
7004         },
7005         {
7006                 .procname       = "ndisc_tclass",
7007                 .data           = &ipv6_devconf.ndisc_tclass,
7008                 .maxlen         = sizeof(int),
7009                 .mode           = 0644,
7010                 .proc_handler   = proc_dointvec_minmax,
7011                 .extra1         = (void *)SYSCTL_ZERO,
7012                 .extra2         = (void *)&two_five_five,
7013         },
7014         {
7015                 .procname       = "rpl_seg_enabled",
7016                 .data           = &ipv6_devconf.rpl_seg_enabled,
7017                 .maxlen         = sizeof(int),
7018                 .mode           = 0644,
7019                 .proc_handler   = proc_dointvec,
7020         },
7021         {
7022                 .procname       = "ioam6_enabled",
7023                 .data           = &ipv6_devconf.ioam6_enabled,
7024                 .maxlen         = sizeof(u8),
7025                 .mode           = 0644,
7026                 .proc_handler   = proc_dou8vec_minmax,
7027                 .extra1         = (void *)SYSCTL_ZERO,
7028                 .extra2         = (void *)SYSCTL_ONE,
7029         },
7030         {
7031                 .procname       = "ioam6_id",
7032                 .data           = &ipv6_devconf.ioam6_id,
7033                 .maxlen         = sizeof(u32),
7034                 .mode           = 0644,
7035                 .proc_handler   = proc_douintvec_minmax,
7036                 .extra1         = (void *)SYSCTL_ZERO,
7037                 .extra2         = (void *)&ioam6_if_id_max,
7038         },
7039         {
7040                 .procname       = "ioam6_id_wide",
7041                 .data           = &ipv6_devconf.ioam6_id_wide,
7042                 .maxlen         = sizeof(u32),
7043                 .mode           = 0644,
7044                 .proc_handler   = proc_douintvec,
7045         },
7046         {
7047                 /* sentinel */
7048         }
7049 };
7050
7051 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7052                 struct inet6_dev *idev, struct ipv6_devconf *p)
7053 {
7054         int i, ifindex;
7055         struct ctl_table *table;
7056         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7057
7058         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
7059         if (!table)
7060                 goto out;
7061
7062         for (i = 0; table[i].data; i++) {
7063                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7064                 /* If one of these is already set, then it is not safe to
7065                  * overwrite either of them: this makes proc_dointvec_minmax
7066                  * usable.
7067                  */
7068                 if (!table[i].extra1 && !table[i].extra2) {
7069                         table[i].extra1 = idev; /* embedded; no ref */
7070                         table[i].extra2 = net;
7071                 }
7072         }
7073
7074         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7075
7076         p->sysctl_header = register_net_sysctl(net, path, table);
7077         if (!p->sysctl_header)
7078                 goto free;
7079
7080         if (!strcmp(dev_name, "all"))
7081                 ifindex = NETCONFA_IFINDEX_ALL;
7082         else if (!strcmp(dev_name, "default"))
7083                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7084         else
7085                 ifindex = idev->dev->ifindex;
7086         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7087                                      ifindex, p);
7088         return 0;
7089
7090 free:
7091         kfree(table);
7092 out:
7093         return -ENOBUFS;
7094 }
7095
7096 static void __addrconf_sysctl_unregister(struct net *net,
7097                                          struct ipv6_devconf *p, int ifindex)
7098 {
7099         struct ctl_table *table;
7100
7101         if (!p->sysctl_header)
7102                 return;
7103
7104         table = p->sysctl_header->ctl_table_arg;
7105         unregister_net_sysctl_table(p->sysctl_header);
7106         p->sysctl_header = NULL;
7107         kfree(table);
7108
7109         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7110 }
7111
7112 static int addrconf_sysctl_register(struct inet6_dev *idev)
7113 {
7114         int err;
7115
7116         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7117                 return -EINVAL;
7118
7119         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7120                                     &ndisc_ifinfo_sysctl_change);
7121         if (err)
7122                 return err;
7123         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7124                                          idev, &idev->cnf);
7125         if (err)
7126                 neigh_sysctl_unregister(idev->nd_parms);
7127
7128         return err;
7129 }
7130
7131 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7132 {
7133         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7134                                      idev->dev->ifindex);
7135         neigh_sysctl_unregister(idev->nd_parms);
7136 }
7137
7138
7139 #endif
7140
7141 static int __net_init addrconf_init_net(struct net *net)
7142 {
7143         int err = -ENOMEM;
7144         struct ipv6_devconf *all, *dflt;
7145
7146         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7147         if (!all)
7148                 goto err_alloc_all;
7149
7150         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7151         if (!dflt)
7152                 goto err_alloc_dflt;
7153
7154         if (!net_eq(net, &init_net)) {
7155                 switch (net_inherit_devconf()) {
7156                 case 1:  /* copy from init_net */
7157                         memcpy(all, init_net.ipv6.devconf_all,
7158                                sizeof(ipv6_devconf));
7159                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7160                                sizeof(ipv6_devconf_dflt));
7161                         break;
7162                 case 3: /* copy from the current netns */
7163                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7164                                sizeof(ipv6_devconf));
7165                         memcpy(dflt,
7166                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7167                                sizeof(ipv6_devconf_dflt));
7168                         break;
7169                 case 0:
7170                 case 2:
7171                         /* use compiled values */
7172                         break;
7173                 }
7174         }
7175
7176         /* these will be inherited by all namespaces */
7177         dflt->autoconf = ipv6_defaults.autoconf;
7178         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7179
7180         dflt->stable_secret.initialized = false;
7181         all->stable_secret.initialized = false;
7182
7183         net->ipv6.devconf_all = all;
7184         net->ipv6.devconf_dflt = dflt;
7185
7186 #ifdef CONFIG_SYSCTL
7187         err = __addrconf_sysctl_register(net, "all", NULL, all);
7188         if (err < 0)
7189                 goto err_reg_all;
7190
7191         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7192         if (err < 0)
7193                 goto err_reg_dflt;
7194 #endif
7195         return 0;
7196
7197 #ifdef CONFIG_SYSCTL
7198 err_reg_dflt:
7199         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7200 err_reg_all:
7201         kfree(dflt);
7202 #endif
7203 err_alloc_dflt:
7204         kfree(all);
7205 err_alloc_all:
7206         return err;
7207 }
7208
7209 static void __net_exit addrconf_exit_net(struct net *net)
7210 {
7211 #ifdef CONFIG_SYSCTL
7212         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7213                                      NETCONFA_IFINDEX_DEFAULT);
7214         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7215                                      NETCONFA_IFINDEX_ALL);
7216 #endif
7217         kfree(net->ipv6.devconf_dflt);
7218         kfree(net->ipv6.devconf_all);
7219 }
7220
7221 static struct pernet_operations addrconf_ops = {
7222         .init = addrconf_init_net,
7223         .exit = addrconf_exit_net,
7224 };
7225
7226 static struct rtnl_af_ops inet6_ops __read_mostly = {
7227         .family           = AF_INET6,
7228         .fill_link_af     = inet6_fill_link_af,
7229         .get_link_af_size = inet6_get_link_af_size,
7230         .validate_link_af = inet6_validate_link_af,
7231         .set_link_af      = inet6_set_link_af,
7232 };
7233
7234 /*
7235  *      Init / cleanup code
7236  */
7237
7238 int __init addrconf_init(void)
7239 {
7240         struct inet6_dev *idev;
7241         int i, err;
7242
7243         err = ipv6_addr_label_init();
7244         if (err < 0) {
7245                 pr_crit("%s: cannot initialize default policy table: %d\n",
7246                         __func__, err);
7247                 goto out;
7248         }
7249
7250         err = register_pernet_subsys(&addrconf_ops);
7251         if (err < 0)
7252                 goto out_addrlabel;
7253
7254         addrconf_wq = create_workqueue("ipv6_addrconf");
7255         if (!addrconf_wq) {
7256                 err = -ENOMEM;
7257                 goto out_nowq;
7258         }
7259
7260         /* The addrconf netdev notifier requires that loopback_dev
7261          * has it's ipv6 private information allocated and setup
7262          * before it can bring up and give link-local addresses
7263          * to other devices which are up.
7264          *
7265          * Unfortunately, loopback_dev is not necessarily the first
7266          * entry in the global dev_base list of net devices.  In fact,
7267          * it is likely to be the very last entry on that list.
7268          * So this causes the notifier registry below to try and
7269          * give link-local addresses to all devices besides loopback_dev
7270          * first, then loopback_dev, which cases all the non-loopback_dev
7271          * devices to fail to get a link-local address.
7272          *
7273          * So, as a temporary fix, allocate the ipv6 structure for
7274          * loopback_dev first by hand.
7275          * Longer term, all of the dependencies ipv6 has upon the loopback
7276          * device and it being up should be removed.
7277          */
7278         rtnl_lock();
7279         idev = ipv6_add_dev(init_net.loopback_dev);
7280         rtnl_unlock();
7281         if (IS_ERR(idev)) {
7282                 err = PTR_ERR(idev);
7283                 goto errlo;
7284         }
7285
7286         ip6_route_init_special_entries();
7287
7288         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7289                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7290
7291         register_netdevice_notifier(&ipv6_dev_notf);
7292
7293         addrconf_verify();
7294
7295         rtnl_af_register(&inet6_ops);
7296
7297         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7298                                    NULL, inet6_dump_ifinfo, 0);
7299         if (err < 0)
7300                 goto errout;
7301
7302         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7303                                    inet6_rtm_newaddr, NULL, 0);
7304         if (err < 0)
7305                 goto errout;
7306         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7307                                    inet6_rtm_deladdr, NULL, 0);
7308         if (err < 0)
7309                 goto errout;
7310         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7311                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7312                                    RTNL_FLAG_DOIT_UNLOCKED);
7313         if (err < 0)
7314                 goto errout;
7315         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7316                                    NULL, inet6_dump_ifmcaddr, 0);
7317         if (err < 0)
7318                 goto errout;
7319         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7320                                    NULL, inet6_dump_ifacaddr, 0);
7321         if (err < 0)
7322                 goto errout;
7323         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7324                                    inet6_netconf_get_devconf,
7325                                    inet6_netconf_dump_devconf,
7326                                    RTNL_FLAG_DOIT_UNLOCKED);
7327         if (err < 0)
7328                 goto errout;
7329         err = ipv6_addr_label_rtnl_register();
7330         if (err < 0)
7331                 goto errout;
7332
7333         return 0;
7334 errout:
7335         rtnl_unregister_all(PF_INET6);
7336         rtnl_af_unregister(&inet6_ops);
7337         unregister_netdevice_notifier(&ipv6_dev_notf);
7338 errlo:
7339         destroy_workqueue(addrconf_wq);
7340 out_nowq:
7341         unregister_pernet_subsys(&addrconf_ops);
7342 out_addrlabel:
7343         ipv6_addr_label_cleanup();
7344 out:
7345         return err;
7346 }
7347
7348 void addrconf_cleanup(void)
7349 {
7350         struct net_device *dev;
7351         int i;
7352
7353         unregister_netdevice_notifier(&ipv6_dev_notf);
7354         unregister_pernet_subsys(&addrconf_ops);
7355         ipv6_addr_label_cleanup();
7356
7357         rtnl_af_unregister(&inet6_ops);
7358
7359         rtnl_lock();
7360
7361         /* clean dev list */
7362         for_each_netdev(&init_net, dev) {
7363                 if (__in6_dev_get(dev) == NULL)
7364                         continue;
7365                 addrconf_ifdown(dev, true);
7366         }
7367         addrconf_ifdown(init_net.loopback_dev, true);
7368
7369         /*
7370          *      Check hash table.
7371          */
7372         spin_lock_bh(&addrconf_hash_lock);
7373         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7374                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7375         spin_unlock_bh(&addrconf_hash_lock);
7376         cancel_delayed_work(&addr_chk_work);
7377         rtnl_unlock();
7378
7379         destroy_workqueue(addrconf_wq);
7380 }