net/ipv6/addrconf.c

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