net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92
  93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  94
  95 #define IPV6_MAX_STRLEN \
  96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  97
  98 static inline u32 cstamp_delta(unsigned long cstamp)
  99 {
 100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 101 }
 102
 103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 104 {
 105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 107         do_div(tmp, 1000000);
 108         return (s32)tmp;
 109 }
 110
 111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 112 {
 113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 119                 do_div(tmp, 1000000);
 120         }
 121         return (s32)tmp;
 122 }
 123
 124 #ifdef CONFIG_SYSCTL
 125 static int addrconf_sysctl_register(struct inet6_dev *idev);
 126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 127 #else
 128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 129 {
 130         return 0;
 131 }
 132
 133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 134 {
 135 }
 136 #endif
 137
 138 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 139
 140 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 141 static int ipv6_count_addresses(const struct inet6_dev *idev);
 142 static int ipv6_generate_stable_address(struct in6_addr *addr,
 143                                         u8 dad_count,
 144                                         const struct inet6_dev *idev);
 145
 146 #define IN6_ADDR_HSIZE_SHIFT    8
 147 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 148 /*
 149  *      Configured unicast address hash table
 150  */
 151 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 152 static DEFINE_SPINLOCK(addrconf_hash_lock);
 153
 154 static void addrconf_verify(void);
 155 static void addrconf_verify_rtnl(void);
 156 static void addrconf_verify_work(struct work_struct *);
 157
 158 static struct workqueue_struct *addrconf_wq;
 159 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 160
 161 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 162 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 163
 164 static void addrconf_type_change(struct net_device *dev,
 165                                  unsigned long event);
 166 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 167
 168 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 169                                                   int plen,
 170                                                   const struct net_device *dev,
 171                                                   u32 flags, u32 noflags,
 172                                                   bool no_gw);
 173
 174 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 175 static void addrconf_dad_work(struct work_struct *w);
 176 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 177                                    bool send_na);
 178 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 179 static void addrconf_rs_timer(struct timer_list *t);
 180 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 181 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182
 183 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 184                                 struct prefix_info *pinfo);
 185
 186 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 187         .forwarding             = 0,
 188         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 189         .mtu6                   = IPV6_MIN_MTU,
 190         .accept_ra              = 1,
 191         .accept_redirects       = 1,
 192         .autoconf               = 1,
 193         .force_mld_version      = 0,
 194         .mldv1_unsolicited_report_interval = 10 * HZ,
 195         .mldv2_unsolicited_report_interval = HZ,
 196         .dad_transmits          = 1,
 197         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 198         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 199         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 201         .use_tempaddr           = 0,
 202         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 203         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 204         .regen_max_retry        = REGEN_MAX_RETRY,
 205         .max_desync_factor      = MAX_DESYNC_FACTOR,
 206         .max_addresses          = IPV6_MAX_ADDRESSES,
 207         .accept_ra_defrtr       = 1,
 208         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 209         .accept_ra_from_local   = 0,
 210         .accept_ra_min_hop_limit= 1,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_min_plen = 0,
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 0,
 224         .suppress_frag_ndisc    = 1,
 225         .accept_ra_mtu          = 1,
 226         .stable_secret          = {
 227                 .initialized = false,
 228         },
 229         .use_oif_addrs_only     = 0,
 230         .ignore_routes_with_linkdown = 0,
 231         .keep_addr_on_down      = 0,
 232         .seg6_enabled           = 0,
 233 #ifdef CONFIG_IPV6_SEG6_HMAC
 234         .seg6_require_hmac      = 0,
 235 #endif
 236         .enhanced_dad           = 1,
 237         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 238         .disable_policy         = 0,
 239         .rpl_seg_enabled        = 0,
 240 };
 241
 242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 243         .forwarding             = 0,
 244         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 245         .mtu6                   = IPV6_MIN_MTU,
 246         .accept_ra              = 1,
 247         .accept_redirects       = 1,
 248         .autoconf               = 1,
 249         .force_mld_version      = 0,
 250         .mldv1_unsolicited_report_interval = 10 * HZ,
 251         .mldv2_unsolicited_report_interval = HZ,
 252         .dad_transmits          = 1,
 253         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 254         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 255         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 256         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 257         .use_tempaddr           = 0,
 258         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 259         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 260         .regen_max_retry        = REGEN_MAX_RETRY,
 261         .max_desync_factor      = MAX_DESYNC_FACTOR,
 262         .max_addresses          = IPV6_MAX_ADDRESSES,
 263         .accept_ra_defrtr       = 1,
 264         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 265         .accept_ra_from_local   = 0,
 266         .accept_ra_min_hop_limit= 1,
 267         .accept_ra_pinfo        = 1,
 268 #ifdef CONFIG_IPV6_ROUTER_PREF
 269         .accept_ra_rtr_pref     = 1,
 270         .rtr_probe_interval     = 60 * HZ,
 271 #ifdef CONFIG_IPV6_ROUTE_INFO
 272         .accept_ra_rt_info_min_plen = 0,
 273         .accept_ra_rt_info_max_plen = 0,
 274 #endif
 275 #endif
 276         .proxy_ndp              = 0,
 277         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 278         .disable_ipv6           = 0,
 279         .accept_dad             = 1,
 280         .suppress_frag_ndisc    = 1,
 281         .accept_ra_mtu          = 1,
 282         .stable_secret          = {
 283                 .initialized = false,
 284         },
 285         .use_oif_addrs_only     = 0,
 286         .ignore_routes_with_linkdown = 0,
 287         .keep_addr_on_down      = 0,
 288         .seg6_enabled           = 0,
 289 #ifdef CONFIG_IPV6_SEG6_HMAC
 290         .seg6_require_hmac      = 0,
 291 #endif
 292         .enhanced_dad           = 1,
 293         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 294         .disable_policy         = 0,
 295         .rpl_seg_enabled        = 0,
 296 };
 297
 298 /* Check if link is ready: is it up and is a valid qdisc available */
 299 static inline bool addrconf_link_ready(const struct net_device *dev)
 300 {
 301         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 302 }
 303
 304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 305 {
 306         if (del_timer(&idev->rs_timer))
 307                 __in6_dev_put(idev);
 308 }
 309
 310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 311 {
 312         if (cancel_delayed_work(&ifp->dad_work))
 313                 __in6_ifa_put(ifp);
 314 }
 315
 316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 317                                   unsigned long when)
 318 {
 319         if (!timer_pending(&idev->rs_timer))
 320                 in6_dev_hold(idev);
 321         mod_timer(&idev->rs_timer, jiffies + when);
 322 }
 323
 324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 325                                    unsigned long delay)
 326 {
 327         in6_ifa_hold(ifp);
 328         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 329                 in6_ifa_put(ifp);
 330 }
 331
 332 static int snmp6_alloc_dev(struct inet6_dev *idev)
 333 {
 334         int i;
 335
 336         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 337         if (!idev->stats.ipv6)
 338                 goto err_ip;
 339
 340         for_each_possible_cpu(i) {
 341                 struct ipstats_mib *addrconf_stats;
 342                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 343                 u64_stats_init(&addrconf_stats->syncp);
 344         }
 345
 346
 347         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 348                                         GFP_KERNEL);
 349         if (!idev->stats.icmpv6dev)
 350                 goto err_icmp;
 351         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 352                                            GFP_KERNEL);
 353         if (!idev->stats.icmpv6msgdev)
 354                 goto err_icmpmsg;
 355
 356         return 0;
 357
 358 err_icmpmsg:
 359         kfree(idev->stats.icmpv6dev);
 360 err_icmp:
 361         free_percpu(idev->stats.ipv6);
 362 err_ip:
 363         return -ENOMEM;
 364 }
 365
 366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 367 {
 368         struct inet6_dev *ndev;
 369         int err = -ENOMEM;
 370
 371         ASSERT_RTNL();
 372
 373         if (dev->mtu < IPV6_MIN_MTU)
 374                 return ERR_PTR(-EINVAL);
 375
 376         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 377         if (!ndev)
 378                 return ERR_PTR(err);
 379
 380         rwlock_init(&ndev->lock);
 381         ndev->dev = dev;
 382         INIT_LIST_HEAD(&ndev->addr_list);
 383         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 384         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 385
 386         if (ndev->cnf.stable_secret.initialized)
 387                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 388
 389         ndev->cnf.mtu6 = dev->mtu;
 390         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 391         if (!ndev->nd_parms) {
 392                 kfree(ndev);
 393                 return ERR_PTR(err);
 394         }
 395         if (ndev->cnf.forwarding)
 396                 dev_disable_lro(dev);
 397         /* We refer to the device */
 398         dev_hold(dev);
 399
 400         if (snmp6_alloc_dev(ndev) < 0) {
 401                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 402                            __func__);
 403                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 404                 dev_put(dev);
 405                 kfree(ndev);
 406                 return ERR_PTR(err);
 407         }
 408
 409         if (snmp6_register_dev(ndev) < 0) {
 410                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 411                            __func__, dev->name);
 412                 goto err_release;
 413         }
 414
 415         /* One reference from device. */
 416         refcount_set(&ndev->refcnt, 1);
 417
 418         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 419                 ndev->cnf.accept_dad = -1;
 420
 421 #if IS_ENABLED(CONFIG_IPV6_SIT)
 422         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 423                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 424                 ndev->cnf.rtr_solicits = 0;
 425         }
 426 #endif
 427
 428         INIT_LIST_HEAD(&ndev->tempaddr_list);
 429         ndev->desync_factor = U32_MAX;
 430         if ((dev->flags&IFF_LOOPBACK) ||
 431             dev->type == ARPHRD_TUNNEL ||
 432             dev->type == ARPHRD_TUNNEL6 ||
 433             dev->type == ARPHRD_SIT ||
 434             dev->type == ARPHRD_NONE) {
 435                 ndev->cnf.use_tempaddr = -1;
 436         }
 437
 438         ndev->token = in6addr_any;
 439
 440         if (netif_running(dev) && addrconf_link_ready(dev))
 441                 ndev->if_flags |= IF_READY;
 442
 443         ipv6_mc_init_dev(ndev);
 444         ndev->tstamp = jiffies;
 445         err = addrconf_sysctl_register(ndev);
 446         if (err) {
 447                 ipv6_mc_destroy_dev(ndev);
 448                 snmp6_unregister_dev(ndev);
 449                 goto err_release;
 450         }
 451         /* protected by rtnl_lock */
 452         rcu_assign_pointer(dev->ip6_ptr, ndev);
 453
 454         /* Join interface-local all-node multicast group */
 455         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 456
 457         /* Join all-node multicast group */
 458         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 459
 460         /* Join all-router multicast group if forwarding is set */
 461         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 462                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 463
 464         return ndev;
 465
 466 err_release:
 467         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 468         ndev->dead = 1;
 469         in6_dev_finish_destroy(ndev);
 470         return ERR_PTR(err);
 471 }
 472
 473 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 474 {
 475         struct inet6_dev *idev;
 476
 477         ASSERT_RTNL();
 478
 479         idev = __in6_dev_get(dev);
 480         if (!idev) {
 481                 idev = ipv6_add_dev(dev);
 482                 if (IS_ERR(idev))
 483                         return idev;
 484         }
 485
 486         if (dev->flags&IFF_UP)
 487                 ipv6_mc_up(idev);
 488         return idev;
 489 }
 490
 491 static int inet6_netconf_msgsize_devconf(int type)
 492 {
 493         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 494                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 495         bool all = false;
 496
 497         if (type == NETCONFA_ALL)
 498                 all = true;
 499
 500         if (all || type == NETCONFA_FORWARDING)
 501                 size += nla_total_size(4);
 502 #ifdef CONFIG_IPV6_MROUTE
 503         if (all || type == NETCONFA_MC_FORWARDING)
 504                 size += nla_total_size(4);
 505 #endif
 506         if (all || type == NETCONFA_PROXY_NEIGH)
 507                 size += nla_total_size(4);
 508
 509         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 510                 size += nla_total_size(4);
 511
 512         return size;
 513 }
 514
 515 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 516                                       struct ipv6_devconf *devconf, u32 portid,
 517                                       u32 seq, int event, unsigned int flags,
 518                                       int type)
 519 {
 520         struct nlmsghdr  *nlh;
 521         struct netconfmsg *ncm;
 522         bool all = false;
 523
 524         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 525                         flags);
 526         if (!nlh)
 527                 return -EMSGSIZE;
 528
 529         if (type == NETCONFA_ALL)
 530                 all = true;
 531
 532         ncm = nlmsg_data(nlh);
 533         ncm->ncm_family = AF_INET6;
 534
 535         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 536                 goto nla_put_failure;
 537
 538         if (!devconf)
 539                 goto out;
 540
 541         if ((all || type == NETCONFA_FORWARDING) &&
 542             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 543                 goto nla_put_failure;
 544 #ifdef CONFIG_IPV6_MROUTE
 545         if ((all || type == NETCONFA_MC_FORWARDING) &&
 546             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 547                         devconf->mc_forwarding) < 0)
 548                 goto nla_put_failure;
 549 #endif
 550         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 551             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 552                 goto nla_put_failure;
 553
 554         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 555             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 556                         devconf->ignore_routes_with_linkdown) < 0)
 557                 goto nla_put_failure;
 558
 559 out:
 560         nlmsg_end(skb, nlh);
 561         return 0;
 562
 563 nla_put_failure:
 564         nlmsg_cancel(skb, nlh);
 565         return -EMSGSIZE;
 566 }
 567
 568 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 569                                   int ifindex, struct ipv6_devconf *devconf)
 570 {
 571         struct sk_buff *skb;
 572         int err = -ENOBUFS;
 573
 574         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 575         if (!skb)
 576                 goto errout;
 577
 578         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 579                                          event, 0, type);
 580         if (err < 0) {
 581                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 582                 WARN_ON(err == -EMSGSIZE);
 583                 kfree_skb(skb);
 584                 goto errout;
 585         }
 586         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 587         return;
 588 errout:
 589         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 590 }
 591
 592 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 593         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 594         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 595         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 596         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 597 };
 598
 599 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 600                                        const struct nlmsghdr *nlh,
 601                                        struct nlattr **tb,
 602                                        struct netlink_ext_ack *extack)
 603 {
 604         int i, err;
 605
 606         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 607                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 608                 return -EINVAL;
 609         }
 610
 611         if (!netlink_strict_get_check(skb))
 612                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 613                                               tb, NETCONFA_MAX,
 614                                               devconf_ipv6_policy, extack);
 615
 616         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 617                                             tb, NETCONFA_MAX,
 618                                             devconf_ipv6_policy, extack);
 619         if (err)
 620                 return err;
 621
 622         for (i = 0; i <= NETCONFA_MAX; i++) {
 623                 if (!tb[i])
 624                         continue;
 625
 626                 switch (i) {
 627                 case NETCONFA_IFINDEX:
 628                         break;
 629                 default:
 630                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 631                         return -EINVAL;
 632                 }
 633         }
 634
 635         return 0;
 636 }
 637
 638 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 639                                      struct nlmsghdr *nlh,
 640                                      struct netlink_ext_ack *extack)
 641 {
 642         struct net *net = sock_net(in_skb->sk);
 643         struct nlattr *tb[NETCONFA_MAX+1];
 644         struct inet6_dev *in6_dev = NULL;
 645         struct net_device *dev = NULL;
 646         struct sk_buff *skb;
 647         struct ipv6_devconf *devconf;
 648         int ifindex;
 649         int err;
 650
 651         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 652         if (err < 0)
 653                 return err;
 654
 655         if (!tb[NETCONFA_IFINDEX])
 656                 return -EINVAL;
 657
 658         err = -EINVAL;
 659         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 660         switch (ifindex) {
 661         case NETCONFA_IFINDEX_ALL:
 662                 devconf = net->ipv6.devconf_all;
 663                 break;
 664         case NETCONFA_IFINDEX_DEFAULT:
 665                 devconf = net->ipv6.devconf_dflt;
 666                 break;
 667         default:
 668                 dev = dev_get_by_index(net, ifindex);
 669                 if (!dev)
 670                         return -EINVAL;
 671                 in6_dev = in6_dev_get(dev);
 672                 if (!in6_dev)
 673                         goto errout;
 674                 devconf = &in6_dev->cnf;
 675                 break;
 676         }
 677
 678         err = -ENOBUFS;
 679         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 680         if (!skb)
 681                 goto errout;
 682
 683         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 684                                          NETLINK_CB(in_skb).portid,
 685                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 686                                          NETCONFA_ALL);
 687         if (err < 0) {
 688                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 689                 WARN_ON(err == -EMSGSIZE);
 690                 kfree_skb(skb);
 691                 goto errout;
 692         }
 693         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 694 errout:
 695         if (in6_dev)
 696                 in6_dev_put(in6_dev);
 697         if (dev)
 698                 dev_put(dev);
 699         return err;
 700 }
 701
 702 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 703                                       struct netlink_callback *cb)
 704 {
 705         const struct nlmsghdr *nlh = cb->nlh;
 706         struct net *net = sock_net(skb->sk);
 707         int h, s_h;
 708         int idx, s_idx;
 709         struct net_device *dev;
 710         struct inet6_dev *idev;
 711         struct hlist_head *head;
 712
 713         if (cb->strict_check) {
 714                 struct netlink_ext_ack *extack = cb->extack;
 715                 struct netconfmsg *ncm;
 716
 717                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 718                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 719                         return -EINVAL;
 720                 }
 721
 722                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 723                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 724                         return -EINVAL;
 725                 }
 726         }
 727
 728         s_h = cb->args[0];
 729         s_idx = idx = cb->args[1];
 730
 731         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 732                 idx = 0;
 733                 head = &net->dev_index_head[h];
 734                 rcu_read_lock();
 735                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 736                           net->dev_base_seq;
 737                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 738                         if (idx < s_idx)
 739                                 goto cont;
 740                         idev = __in6_dev_get(dev);
 741                         if (!idev)
 742                                 goto cont;
 743
 744                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 745                                                        &idev->cnf,
 746                                                        NETLINK_CB(cb->skb).portid,
 747                                                        nlh->nlmsg_seq,
 748                                                        RTM_NEWNETCONF,
 749                                                        NLM_F_MULTI,
 750                                                        NETCONFA_ALL) < 0) {
 751                                 rcu_read_unlock();
 752                                 goto done;
 753                         }
 754                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 755 cont:
 756                         idx++;
 757                 }
 758                 rcu_read_unlock();
 759         }
 760         if (h == NETDEV_HASHENTRIES) {
 761                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 762                                                net->ipv6.devconf_all,
 763                                                NETLINK_CB(cb->skb).portid,
 764                                                nlh->nlmsg_seq,
 765                                                RTM_NEWNETCONF, NLM_F_MULTI,
 766                                                NETCONFA_ALL) < 0)
 767                         goto done;
 768                 else
 769                         h++;
 770         }
 771         if (h == NETDEV_HASHENTRIES + 1) {
 772                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 773                                                net->ipv6.devconf_dflt,
 774                                                NETLINK_CB(cb->skb).portid,
 775                                                nlh->nlmsg_seq,
 776                                                RTM_NEWNETCONF, NLM_F_MULTI,
 777                                                NETCONFA_ALL) < 0)
 778                         goto done;
 779                 else
 780                         h++;
 781         }
 782 done:
 783         cb->args[0] = h;
 784         cb->args[1] = idx;
 785
 786         return skb->len;
 787 }
 788
 789 #ifdef CONFIG_SYSCTL
 790 static void dev_forward_change(struct inet6_dev *idev)
 791 {
 792         struct net_device *dev;
 793         struct inet6_ifaddr *ifa;
 794
 795         if (!idev)
 796                 return;
 797         dev = idev->dev;
 798         if (idev->cnf.forwarding)
 799                 dev_disable_lro(dev);
 800         if (dev->flags & IFF_MULTICAST) {
 801                 if (idev->cnf.forwarding) {
 802                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 803                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 804                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 805                 } else {
 806                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 807                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 808                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 809                 }
 810         }
 811
 812         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 813                 if (ifa->flags&IFA_F_TENTATIVE)
 814                         continue;
 815                 if (idev->cnf.forwarding)
 816                         addrconf_join_anycast(ifa);
 817                 else
 818                         addrconf_leave_anycast(ifa);
 819         }
 820         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 821                                      NETCONFA_FORWARDING,
 822                                      dev->ifindex, &idev->cnf);
 823 }
 824
 825
 826 static void addrconf_forward_change(struct net *net, __s32 newf)
 827 {
 828         struct net_device *dev;
 829         struct inet6_dev *idev;
 830
 831         for_each_netdev(net, dev) {
 832                 idev = __in6_dev_get(dev);
 833                 if (idev) {
 834                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 835                         idev->cnf.forwarding = newf;
 836                         if (changed)
 837                                 dev_forward_change(idev);
 838                 }
 839         }
 840 }
 841
 842 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 843 {
 844         struct net *net;
 845         int old;
 846
 847         if (!rtnl_trylock())
 848                 return restart_syscall();
 849
 850         net = (struct net *)table->extra2;
 851         old = *p;
 852         *p = newf;
 853
 854         if (p == &net->ipv6.devconf_dflt->forwarding) {
 855                 if ((!newf) ^ (!old))
 856                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 857                                                      NETCONFA_FORWARDING,
 858                                                      NETCONFA_IFINDEX_DEFAULT,
 859                                                      net->ipv6.devconf_dflt);
 860                 rtnl_unlock();
 861                 return 0;
 862         }
 863
 864         if (p == &net->ipv6.devconf_all->forwarding) {
 865                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 866
 867                 net->ipv6.devconf_dflt->forwarding = newf;
 868                 if ((!newf) ^ (!old_dflt))
 869                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 870                                                      NETCONFA_FORWARDING,
 871                                                      NETCONFA_IFINDEX_DEFAULT,
 872                                                      net->ipv6.devconf_dflt);
 873
 874                 addrconf_forward_change(net, newf);
 875                 if ((!newf) ^ (!old))
 876                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 877                                                      NETCONFA_FORWARDING,
 878                                                      NETCONFA_IFINDEX_ALL,
 879                                                      net->ipv6.devconf_all);
 880         } else if ((!newf) ^ (!old))
 881                 dev_forward_change((struct inet6_dev *)table->extra1);
 882         rtnl_unlock();
 883
 884         if (newf)
 885                 rt6_purge_dflt_routers(net);
 886         return 1;
 887 }
 888
 889 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 890 {
 891         struct net_device *dev;
 892         struct inet6_dev *idev;
 893
 894         for_each_netdev(net, dev) {
 895                 idev = __in6_dev_get(dev);
 896                 if (idev) {
 897                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 898
 899                         idev->cnf.ignore_routes_with_linkdown = newf;
 900                         if (changed)
 901                                 inet6_netconf_notify_devconf(dev_net(dev),
 902                                                              RTM_NEWNETCONF,
 903                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 904                                                              dev->ifindex,
 905                                                              &idev->cnf);
 906                 }
 907         }
 908 }
 909
 910 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 911 {
 912         struct net *net;
 913         int old;
 914
 915         if (!rtnl_trylock())
 916                 return restart_syscall();
 917
 918         net = (struct net *)table->extra2;
 919         old = *p;
 920         *p = newf;
 921
 922         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 923                 if ((!newf) ^ (!old))
 924                         inet6_netconf_notify_devconf(net,
 925                                                      RTM_NEWNETCONF,
 926                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 927                                                      NETCONFA_IFINDEX_DEFAULT,
 928                                                      net->ipv6.devconf_dflt);
 929                 rtnl_unlock();
 930                 return 0;
 931         }
 932
 933         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 934                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 935                 addrconf_linkdown_change(net, newf);
 936                 if ((!newf) ^ (!old))
 937                         inet6_netconf_notify_devconf(net,
 938                                                      RTM_NEWNETCONF,
 939                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 940                                                      NETCONFA_IFINDEX_ALL,
 941                                                      net->ipv6.devconf_all);
 942         }
 943         rtnl_unlock();
 944
 945         return 1;
 946 }
 947
 948 #endif
 949
 950 /* Nobody refers to this ifaddr, destroy it */
 951 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 952 {
 953         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 954
 955 #ifdef NET_REFCNT_DEBUG
 956         pr_debug("%s\n", __func__);
 957 #endif
 958
 959         in6_dev_put(ifp->idev);
 960
 961         if (cancel_delayed_work(&ifp->dad_work))
 962                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 963                           ifp);
 964
 965         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 966                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 967                 return;
 968         }
 969
 970         kfree_rcu(ifp, rcu);
 971 }
 972
 973 static void
 974 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 975 {
 976         struct list_head *p;
 977         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 978
 979         /*
 980          * Each device address list is sorted in order of scope -
 981          * global before linklocal.
 982          */
 983         list_for_each(p, &idev->addr_list) {
 984                 struct inet6_ifaddr *ifa
 985                         = list_entry(p, struct inet6_ifaddr, if_list);
 986                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 987                         break;
 988         }
 989
 990         list_add_tail_rcu(&ifp->if_list, p);
 991 }
 992
 993 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 994 {
 995         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 996
 997         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 998 }
 999
1000 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1001                                struct net_device *dev, unsigned int hash)
1002 {
1003         struct inet6_ifaddr *ifp;
1004
1005         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1006                 if (!net_eq(dev_net(ifp->idev->dev), net))
1007                         continue;
1008                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1009                         if (!dev || ifp->idev->dev == dev)
1010                                 return true;
1011                 }
1012         }
1013         return false;
1014 }
1015
1016 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1017 {
1018         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1019         int err = 0;
1020
1021         spin_lock(&addrconf_hash_lock);
1022
1023         /* Ignore adding duplicate addresses on an interface */
1024         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1025                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1026                 err = -EEXIST;
1027         } else {
1028                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1029         }
1030
1031         spin_unlock(&addrconf_hash_lock);
1032
1033         return err;
1034 }
1035
1036 /* On success it returns ifp with increased reference count */
1037
1038 static struct inet6_ifaddr *
1039 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1040               bool can_block, struct netlink_ext_ack *extack)
1041 {
1042         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1043         int addr_type = ipv6_addr_type(cfg->pfx);
1044         struct net *net = dev_net(idev->dev);
1045         struct inet6_ifaddr *ifa = NULL;
1046         struct fib6_info *f6i = NULL;
1047         int err = 0;
1048
1049         if (addr_type == IPV6_ADDR_ANY ||
1050             (addr_type & IPV6_ADDR_MULTICAST &&
1051              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1052             (!(idev->dev->flags & IFF_LOOPBACK) &&
1053              !netif_is_l3_master(idev->dev) &&
1054              addr_type & IPV6_ADDR_LOOPBACK))
1055                 return ERR_PTR(-EADDRNOTAVAIL);
1056
1057         if (idev->dead) {
1058                 err = -ENODEV;                  /*XXX*/
1059                 goto out;
1060         }
1061
1062         if (idev->cnf.disable_ipv6) {
1063                 err = -EACCES;
1064                 goto out;
1065         }
1066
1067         /* validator notifier needs to be blocking;
1068          * do not call in atomic context
1069          */
1070         if (can_block) {
1071                 struct in6_validator_info i6vi = {
1072                         .i6vi_addr = *cfg->pfx,
1073                         .i6vi_dev = idev,
1074                         .extack = extack,
1075                 };
1076
1077                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1078                 err = notifier_to_errno(err);
1079                 if (err < 0)
1080                         goto out;
1081         }
1082
1083         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1084         if (!ifa) {
1085                 err = -ENOBUFS;
1086                 goto out;
1087         }
1088
1089         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1090         if (IS_ERR(f6i)) {
1091                 err = PTR_ERR(f6i);
1092                 f6i = NULL;
1093                 goto out;
1094         }
1095
1096         if (net->ipv6.devconf_all->disable_policy ||
1097             idev->cnf.disable_policy)
1098                 f6i->dst_nopolicy = true;
1099
1100         neigh_parms_data_state_setall(idev->nd_parms);
1101
1102         ifa->addr = *cfg->pfx;
1103         if (cfg->peer_pfx)
1104                 ifa->peer_addr = *cfg->peer_pfx;
1105
1106         spin_lock_init(&ifa->lock);
1107         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1108         INIT_HLIST_NODE(&ifa->addr_lst);
1109         ifa->scope = cfg->scope;
1110         ifa->prefix_len = cfg->plen;
1111         ifa->rt_priority = cfg->rt_priority;
1112         ifa->flags = cfg->ifa_flags;
1113         /* No need to add the TENTATIVE flag for addresses with NODAD */
1114         if (!(cfg->ifa_flags & IFA_F_NODAD))
1115                 ifa->flags |= IFA_F_TENTATIVE;
1116         ifa->valid_lft = cfg->valid_lft;
1117         ifa->prefered_lft = cfg->preferred_lft;
1118         ifa->cstamp = ifa->tstamp = jiffies;
1119         ifa->tokenized = false;
1120
1121         ifa->rt = f6i;
1122
1123         ifa->idev = idev;
1124         in6_dev_hold(idev);
1125
1126         /* For caller */
1127         refcount_set(&ifa->refcnt, 1);
1128
1129         rcu_read_lock_bh();
1130
1131         err = ipv6_add_addr_hash(idev->dev, ifa);
1132         if (err < 0) {
1133                 rcu_read_unlock_bh();
1134                 goto out;
1135         }
1136
1137         write_lock(&idev->lock);
1138
1139         /* Add to inet6_dev unicast addr list. */
1140         ipv6_link_dev_addr(idev, ifa);
1141
1142         if (ifa->flags&IFA_F_TEMPORARY) {
1143                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1144                 in6_ifa_hold(ifa);
1145         }
1146
1147         in6_ifa_hold(ifa);
1148         write_unlock(&idev->lock);
1149
1150         rcu_read_unlock_bh();
1151
1152         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1153 out:
1154         if (unlikely(err < 0)) {
1155                 fib6_info_release(f6i);
1156
1157                 if (ifa) {
1158                         if (ifa->idev)
1159                                 in6_dev_put(ifa->idev);
1160                         kfree(ifa);
1161                 }
1162                 ifa = ERR_PTR(err);
1163         }
1164
1165         return ifa;
1166 }
1167
1168 enum cleanup_prefix_rt_t {
1169         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1170         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1171         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1172 };
1173
1174 /*
1175  * Check, whether the prefix for ifp would still need a prefix route
1176  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1177  * constants.
1178  *
1179  * 1) we don't purge prefix if address was not permanent.
1180  *    prefix is managed by its own lifetime.
1181  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1182  * 3) if there are no addresses, delete prefix.
1183  * 4) if there are still other permanent address(es),
1184  *    corresponding prefix is still permanent.
1185  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1186  *    don't purge the prefix, assume user space is managing it.
1187  * 6) otherwise, update prefix lifetime to the
1188  *    longest valid lifetime among the corresponding
1189  *    addresses on the device.
1190  *    Note: subsequent RA will update lifetime.
1191  **/
1192 static enum cleanup_prefix_rt_t
1193 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1194 {
1195         struct inet6_ifaddr *ifa;
1196         struct inet6_dev *idev = ifp->idev;
1197         unsigned long lifetime;
1198         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1199
1200         *expires = jiffies;
1201
1202         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1203                 if (ifa == ifp)
1204                         continue;
1205                 if (ifa->prefix_len != ifp->prefix_len ||
1206                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1207                                        ifp->prefix_len))
1208                         continue;
1209                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1210                         return CLEANUP_PREFIX_RT_NOP;
1211
1212                 action = CLEANUP_PREFIX_RT_EXPIRE;
1213
1214                 spin_lock(&ifa->lock);
1215
1216                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1217                 /*
1218                  * Note: Because this address is
1219                  * not permanent, lifetime <
1220                  * LONG_MAX / HZ here.
1221                  */
1222                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1223                         *expires = ifa->tstamp + lifetime * HZ;
1224                 spin_unlock(&ifa->lock);
1225         }
1226
1227         return action;
1228 }
1229
1230 static void
1231 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1232                      bool del_rt, bool del_peer)
1233 {
1234         struct fib6_info *f6i;
1235
1236         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1237                                         ifp->prefix_len,
1238                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1239         if (f6i) {
1240                 if (del_rt)
1241                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1242                 else {
1243                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1244                                 fib6_set_expires(f6i, expires);
1245                         fib6_info_release(f6i);
1246                 }
1247         }
1248 }
1249
1250
1251 /* This function wants to get referenced ifp and releases it before return */
1252
1253 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1254 {
1255         int state;
1256         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1257         unsigned long expires;
1258
1259         ASSERT_RTNL();
1260
1261         spin_lock_bh(&ifp->lock);
1262         state = ifp->state;
1263         ifp->state = INET6_IFADDR_STATE_DEAD;
1264         spin_unlock_bh(&ifp->lock);
1265
1266         if (state == INET6_IFADDR_STATE_DEAD)
1267                 goto out;
1268
1269         spin_lock_bh(&addrconf_hash_lock);
1270         hlist_del_init_rcu(&ifp->addr_lst);
1271         spin_unlock_bh(&addrconf_hash_lock);
1272
1273         write_lock_bh(&ifp->idev->lock);
1274
1275         if (ifp->flags&IFA_F_TEMPORARY) {
1276                 list_del(&ifp->tmp_list);
1277                 if (ifp->ifpub) {
1278                         in6_ifa_put(ifp->ifpub);
1279                         ifp->ifpub = NULL;
1280                 }
1281                 __in6_ifa_put(ifp);
1282         }
1283
1284         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1285                 action = check_cleanup_prefix_route(ifp, &expires);
1286
1287         list_del_rcu(&ifp->if_list);
1288         __in6_ifa_put(ifp);
1289
1290         write_unlock_bh(&ifp->idev->lock);
1291
1292         addrconf_del_dad_work(ifp);
1293
1294         ipv6_ifa_notify(RTM_DELADDR, ifp);
1295
1296         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1297
1298         if (action != CLEANUP_PREFIX_RT_NOP) {
1299                 cleanup_prefix_route(ifp, expires,
1300                         action == CLEANUP_PREFIX_RT_DEL, false);
1301         }
1302
1303         /* clean up prefsrc entries */
1304         rt6_remove_prefsrc(ifp);
1305 out:
1306         in6_ifa_put(ifp);
1307 }
1308
1309 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1310 {
1311         struct inet6_dev *idev = ifp->idev;
1312         unsigned long tmp_tstamp, age;
1313         unsigned long regen_advance;
1314         unsigned long now = jiffies;
1315         s32 cnf_temp_preferred_lft;
1316         struct inet6_ifaddr *ift;
1317         struct ifa6_config cfg;
1318         long max_desync_factor;
1319         struct in6_addr addr;
1320         int ret = 0;
1321
1322         write_lock_bh(&idev->lock);
1323
1324 retry:
1325         in6_dev_hold(idev);
1326         if (idev->cnf.use_tempaddr <= 0) {
1327                 write_unlock_bh(&idev->lock);
1328                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1329                 in6_dev_put(idev);
1330                 ret = -1;
1331                 goto out;
1332         }
1333         spin_lock_bh(&ifp->lock);
1334         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1335                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1336                 spin_unlock_bh(&ifp->lock);
1337                 write_unlock_bh(&idev->lock);
1338                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1339                         __func__);
1340                 in6_dev_put(idev);
1341                 ret = -1;
1342                 goto out;
1343         }
1344         in6_ifa_hold(ifp);
1345         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1346         ipv6_gen_rnd_iid(&addr);
1347
1348         age = (now - ifp->tstamp) / HZ;
1349
1350         regen_advance = idev->cnf.regen_max_retry *
1351                         idev->cnf.dad_transmits *
1352                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1353
1354         /* recalculate max_desync_factor each time and update
1355          * idev->desync_factor if it's larger
1356          */
1357         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1358         max_desync_factor = min_t(__u32,
1359                                   idev->cnf.max_desync_factor,
1360                                   cnf_temp_preferred_lft - regen_advance);
1361
1362         if (unlikely(idev->desync_factor > max_desync_factor)) {
1363                 if (max_desync_factor > 0) {
1364                         get_random_bytes(&idev->desync_factor,
1365                                          sizeof(idev->desync_factor));
1366                         idev->desync_factor %= max_desync_factor;
1367                 } else {
1368                         idev->desync_factor = 0;
1369                 }
1370         }
1371
1372         memset(&cfg, 0, sizeof(cfg));
1373         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1374                               idev->cnf.temp_valid_lft + age);
1375         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1376         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1377
1378         cfg.plen = ifp->prefix_len;
1379         tmp_tstamp = ifp->tstamp;
1380         spin_unlock_bh(&ifp->lock);
1381
1382         write_unlock_bh(&idev->lock);
1383
1384         /* A temporary address is created only if this calculated Preferred
1385          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1386          * an implementation must not create a temporary address with a zero
1387          * Preferred Lifetime.
1388          * Use age calculation as in addrconf_verify to avoid unnecessary
1389          * temporary addresses being generated.
1390          */
1391         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1392         if (cfg.preferred_lft <= regen_advance + age) {
1393                 in6_ifa_put(ifp);
1394                 in6_dev_put(idev);
1395                 ret = -1;
1396                 goto out;
1397         }
1398
1399         cfg.ifa_flags = IFA_F_TEMPORARY;
1400         /* set in addrconf_prefix_rcv() */
1401         if (ifp->flags & IFA_F_OPTIMISTIC)
1402                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1403
1404         cfg.pfx = &addr;
1405         cfg.scope = ipv6_addr_scope(cfg.pfx);
1406
1407         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1408         if (IS_ERR(ift)) {
1409                 in6_ifa_put(ifp);
1410                 in6_dev_put(idev);
1411                 pr_info("%s: retry temporary address regeneration\n", __func__);
1412                 write_lock_bh(&idev->lock);
1413                 goto retry;
1414         }
1415
1416         spin_lock_bh(&ift->lock);
1417         ift->ifpub = ifp;
1418         ift->cstamp = now;
1419         ift->tstamp = tmp_tstamp;
1420         spin_unlock_bh(&ift->lock);
1421
1422         addrconf_dad_start(ift);
1423         in6_ifa_put(ift);
1424         in6_dev_put(idev);
1425 out:
1426         return ret;
1427 }
1428
1429 /*
1430  *      Choose an appropriate source address (RFC3484)
1431  */
1432 enum {
1433         IPV6_SADDR_RULE_INIT = 0,
1434         IPV6_SADDR_RULE_LOCAL,
1435         IPV6_SADDR_RULE_SCOPE,
1436         IPV6_SADDR_RULE_PREFERRED,
1437 #ifdef CONFIG_IPV6_MIP6
1438         IPV6_SADDR_RULE_HOA,
1439 #endif
1440         IPV6_SADDR_RULE_OIF,
1441         IPV6_SADDR_RULE_LABEL,
1442         IPV6_SADDR_RULE_PRIVACY,
1443         IPV6_SADDR_RULE_ORCHID,
1444         IPV6_SADDR_RULE_PREFIX,
1445 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1446         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1447 #endif
1448         IPV6_SADDR_RULE_MAX
1449 };
1450
1451 struct ipv6_saddr_score {
1452         int                     rule;
1453         int                     addr_type;
1454         struct inet6_ifaddr     *ifa;
1455         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1456         int                     scopedist;
1457         int                     matchlen;
1458 };
1459
1460 struct ipv6_saddr_dst {
1461         const struct in6_addr *addr;
1462         int ifindex;
1463         int scope;
1464         int label;
1465         unsigned int prefs;
1466 };
1467
1468 static inline int ipv6_saddr_preferred(int type)
1469 {
1470         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1471                 return 1;
1472         return 0;
1473 }
1474
1475 static bool ipv6_use_optimistic_addr(struct net *net,
1476                                      struct inet6_dev *idev)
1477 {
1478 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1479         if (!idev)
1480                 return false;
1481         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1482                 return false;
1483         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1484                 return false;
1485
1486         return true;
1487 #else
1488         return false;
1489 #endif
1490 }
1491
1492 static bool ipv6_allow_optimistic_dad(struct net *net,
1493                                       struct inet6_dev *idev)
1494 {
1495 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1496         if (!idev)
1497                 return false;
1498         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1499                 return false;
1500
1501         return true;
1502 #else
1503         return false;
1504 #endif
1505 }
1506
1507 static int ipv6_get_saddr_eval(struct net *net,
1508                                struct ipv6_saddr_score *score,
1509                                struct ipv6_saddr_dst *dst,
1510                                int i)
1511 {
1512         int ret;
1513
1514         if (i <= score->rule) {
1515                 switch (i) {
1516                 case IPV6_SADDR_RULE_SCOPE:
1517                         ret = score->scopedist;
1518                         break;
1519                 case IPV6_SADDR_RULE_PREFIX:
1520                         ret = score->matchlen;
1521                         break;
1522                 default:
1523                         ret = !!test_bit(i, score->scorebits);
1524                 }
1525                 goto out;
1526         }
1527
1528         switch (i) {
1529         case IPV6_SADDR_RULE_INIT:
1530                 /* Rule 0: remember if hiscore is not ready yet */
1531                 ret = !!score->ifa;
1532                 break;
1533         case IPV6_SADDR_RULE_LOCAL:
1534                 /* Rule 1: Prefer same address */
1535                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1536                 break;
1537         case IPV6_SADDR_RULE_SCOPE:
1538                 /* Rule 2: Prefer appropriate scope
1539                  *
1540                  *      ret
1541                  *       ^
1542                  *    -1 |  d 15
1543                  *    ---+--+-+---> scope
1544                  *       |
1545                  *       |             d is scope of the destination.
1546                  *  B-d  |  \
1547                  *       |   \      <- smaller scope is better if
1548                  *  B-15 |    \        if scope is enough for destination.
1549                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1550                  * d-C-1 | /
1551                  *       |/         <- greater is better
1552                  *   -C  /             if scope is not enough for destination.
1553                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1554                  *
1555                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1556                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1557                  * Assume B = 0 and we get C > 29.
1558                  */
1559                 ret = __ipv6_addr_src_scope(score->addr_type);
1560                 if (ret >= dst->scope)
1561                         ret = -ret;
1562                 else
1563                         ret -= 128;     /* 30 is enough */
1564                 score->scopedist = ret;
1565                 break;
1566         case IPV6_SADDR_RULE_PREFERRED:
1567             {
1568                 /* Rule 3: Avoid deprecated and optimistic addresses */
1569                 u8 avoid = IFA_F_DEPRECATED;
1570
1571                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1572                         avoid |= IFA_F_OPTIMISTIC;
1573                 ret = ipv6_saddr_preferred(score->addr_type) ||
1574                       !(score->ifa->flags & avoid);
1575                 break;
1576             }
1577 #ifdef CONFIG_IPV6_MIP6
1578         case IPV6_SADDR_RULE_HOA:
1579             {
1580                 /* Rule 4: Prefer home address */
1581                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1582                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1583                 break;
1584             }
1585 #endif
1586         case IPV6_SADDR_RULE_OIF:
1587                 /* Rule 5: Prefer outgoing interface */
1588                 ret = (!dst->ifindex ||
1589                        dst->ifindex == score->ifa->idev->dev->ifindex);
1590                 break;
1591         case IPV6_SADDR_RULE_LABEL:
1592                 /* Rule 6: Prefer matching label */
1593                 ret = ipv6_addr_label(net,
1594                                       &score->ifa->addr, score->addr_type,
1595                                       score->ifa->idev->dev->ifindex) == dst->label;
1596                 break;
1597         case IPV6_SADDR_RULE_PRIVACY:
1598             {
1599                 /* Rule 7: Prefer public address
1600                  * Note: prefer temporary address if use_tempaddr >= 2
1601                  */
1602                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1603                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1604                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1605                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1606                 break;
1607             }
1608         case IPV6_SADDR_RULE_ORCHID:
1609                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1610                  *          non-ORCHID vs non-ORCHID
1611                  */
1612                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1613                         ipv6_addr_orchid(dst->addr));
1614                 break;
1615         case IPV6_SADDR_RULE_PREFIX:
1616                 /* Rule 8: Use longest matching prefix */
1617                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1618                 if (ret > score->ifa->prefix_len)
1619                         ret = score->ifa->prefix_len;
1620                 score->matchlen = ret;
1621                 break;
1622 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1623         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1624                 /* Optimistic addresses still have lower precedence than other
1625                  * preferred addresses.
1626                  */
1627                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1628                 break;
1629 #endif
1630         default:
1631                 ret = 0;
1632         }
1633
1634         if (ret)
1635                 __set_bit(i, score->scorebits);
1636         score->rule = i;
1637 out:
1638         return ret;
1639 }
1640
1641 static int __ipv6_dev_get_saddr(struct net *net,
1642                                 struct ipv6_saddr_dst *dst,
1643                                 struct inet6_dev *idev,
1644                                 struct ipv6_saddr_score *scores,
1645                                 int hiscore_idx)
1646 {
1647         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1648
1649         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1650                 int i;
1651
1652                 /*
1653                  * - Tentative Address (RFC2462 section 5.4)
1654                  *  - A tentative address is not considered
1655                  *    "assigned to an interface" in the traditional
1656                  *    sense, unless it is also flagged as optimistic.
1657                  * - Candidate Source Address (section 4)
1658                  *  - In any case, anycast addresses, multicast
1659                  *    addresses, and the unspecified address MUST
1660                  *    NOT be included in a candidate set.
1661                  */
1662                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1663                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1664                         continue;
1665
1666                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1667
1668                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1669                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1670                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1671                                             idev->dev->name);
1672                         continue;
1673                 }
1674
1675                 score->rule = -1;
1676                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1677
1678                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1679                         int minihiscore, miniscore;
1680
1681                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1682                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1683
1684                         if (minihiscore > miniscore) {
1685                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1686                                     score->scopedist > 0) {
1687                                         /*
1688                                          * special case:
1689                                          * each remaining entry
1690                                          * has too small (not enough)
1691                                          * scope, because ifa entries
1692                                          * are sorted by their scope
1693                                          * values.
1694                                          */
1695                                         goto out;
1696                                 }
1697                                 break;
1698                         } else if (minihiscore < miniscore) {
1699                                 swap(hiscore, score);
1700                                 hiscore_idx = 1 - hiscore_idx;
1701
1702                                 /* restore our iterator */
1703                                 score->ifa = hiscore->ifa;
1704
1705                                 break;
1706                         }
1707                 }
1708         }
1709 out:
1710         return hiscore_idx;
1711 }
1712
1713 static int ipv6_get_saddr_master(struct net *net,
1714                                  const struct net_device *dst_dev,
1715                                  const struct net_device *master,
1716                                  struct ipv6_saddr_dst *dst,
1717                                  struct ipv6_saddr_score *scores,
1718                                  int hiscore_idx)
1719 {
1720         struct inet6_dev *idev;
1721
1722         idev = __in6_dev_get(dst_dev);
1723         if (idev)
1724                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1725                                                    scores, hiscore_idx);
1726
1727         idev = __in6_dev_get(master);
1728         if (idev)
1729                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1730                                                    scores, hiscore_idx);
1731
1732         return hiscore_idx;
1733 }
1734
1735 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1736                        const struct in6_addr *daddr, unsigned int prefs,
1737                        struct in6_addr *saddr)
1738 {
1739         struct ipv6_saddr_score scores[2], *hiscore;
1740         struct ipv6_saddr_dst dst;
1741         struct inet6_dev *idev;
1742         struct net_device *dev;
1743         int dst_type;
1744         bool use_oif_addr = false;
1745         int hiscore_idx = 0;
1746         int ret = 0;
1747
1748         dst_type = __ipv6_addr_type(daddr);
1749         dst.addr = daddr;
1750         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1751         dst.scope = __ipv6_addr_src_scope(dst_type);
1752         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1753         dst.prefs = prefs;
1754
1755         scores[hiscore_idx].rule = -1;
1756         scores[hiscore_idx].ifa = NULL;
1757
1758         rcu_read_lock();
1759
1760         /* Candidate Source Address (section 4)
1761          *  - multicast and link-local destination address,
1762          *    the set of candidate source address MUST only
1763          *    include addresses assigned to interfaces
1764          *    belonging to the same link as the outgoing
1765          *    interface.
1766          * (- For site-local destination addresses, the
1767          *    set of candidate source addresses MUST only
1768          *    include addresses assigned to interfaces
1769          *    belonging to the same site as the outgoing
1770          *    interface.)
1771          *  - "It is RECOMMENDED that the candidate source addresses
1772          *    be the set of unicast addresses assigned to the
1773          *    interface that will be used to send to the destination
1774          *    (the 'outgoing' interface)." (RFC 6724)
1775          */
1776         if (dst_dev) {
1777                 idev = __in6_dev_get(dst_dev);
1778                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1779                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1780                     (idev && idev->cnf.use_oif_addrs_only)) {
1781                         use_oif_addr = true;
1782                 }
1783         }
1784
1785         if (use_oif_addr) {
1786                 if (idev)
1787                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1788         } else {
1789                 const struct net_device *master;
1790                 int master_idx = 0;
1791
1792                 /* if dst_dev exists and is enslaved to an L3 device, then
1793                  * prefer addresses from dst_dev and then the master over
1794                  * any other enslaved devices in the L3 domain.
1795                  */
1796                 master = l3mdev_master_dev_rcu(dst_dev);
1797                 if (master) {
1798                         master_idx = master->ifindex;
1799
1800                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1801                                                             master, &dst,
1802                                                             scores, hiscore_idx);
1803
1804                         if (scores[hiscore_idx].ifa)
1805                                 goto out;
1806                 }
1807
1808                 for_each_netdev_rcu(net, dev) {
1809                         /* only consider addresses on devices in the
1810                          * same L3 domain
1811                          */
1812                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1813                                 continue;
1814                         idev = __in6_dev_get(dev);
1815                         if (!idev)
1816                                 continue;
1817                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1818                 }
1819         }
1820
1821 out:
1822         hiscore = &scores[hiscore_idx];
1823         if (!hiscore->ifa)
1824                 ret = -EADDRNOTAVAIL;
1825         else
1826                 *saddr = hiscore->ifa->addr;
1827
1828         rcu_read_unlock();
1829         return ret;
1830 }
1831 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1832
1833 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1834                       u32 banned_flags)
1835 {
1836         struct inet6_ifaddr *ifp;
1837         int err = -EADDRNOTAVAIL;
1838
1839         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1840                 if (ifp->scope > IFA_LINK)
1841                         break;
1842                 if (ifp->scope == IFA_LINK &&
1843                     !(ifp->flags & banned_flags)) {
1844                         *addr = ifp->addr;
1845                         err = 0;
1846                         break;
1847                 }
1848         }
1849         return err;
1850 }
1851
1852 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1853                     u32 banned_flags)
1854 {
1855         struct inet6_dev *idev;
1856         int err = -EADDRNOTAVAIL;
1857
1858         rcu_read_lock();
1859         idev = __in6_dev_get(dev);
1860         if (idev) {
1861                 read_lock_bh(&idev->lock);
1862                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1863                 read_unlock_bh(&idev->lock);
1864         }
1865         rcu_read_unlock();
1866         return err;
1867 }
1868
1869 static int ipv6_count_addresses(const struct inet6_dev *idev)
1870 {
1871         const struct inet6_ifaddr *ifp;
1872         int cnt = 0;
1873
1874         rcu_read_lock();
1875         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1876                 cnt++;
1877         rcu_read_unlock();
1878         return cnt;
1879 }
1880
1881 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1882                   const struct net_device *dev, int strict)
1883 {
1884         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1885                                        strict, IFA_F_TENTATIVE);
1886 }
1887 EXPORT_SYMBOL(ipv6_chk_addr);
1888
1889 /* device argument is used to find the L3 domain of interest. If
1890  * skip_dev_check is set, then the ifp device is not checked against
1891  * the passed in dev argument. So the 2 cases for addresses checks are:
1892  *   1. does the address exist in the L3 domain that dev is part of
1893  *      (skip_dev_check = true), or
1894  *
1895  *   2. does the address exist on the specific device
1896  *      (skip_dev_check = false)
1897  */
1898 static struct net_device *
1899 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1900                           const struct net_device *dev, bool skip_dev_check,
1901                           int strict, u32 banned_flags)
1902 {
1903         unsigned int hash = inet6_addr_hash(net, addr);
1904         struct net_device *l3mdev, *ndev;
1905         struct inet6_ifaddr *ifp;
1906         u32 ifp_flags;
1907
1908         rcu_read_lock();
1909
1910         l3mdev = l3mdev_master_dev_rcu(dev);
1911         if (skip_dev_check)
1912                 dev = NULL;
1913
1914         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1915                 ndev = ifp->idev->dev;
1916                 if (!net_eq(dev_net(ndev), net))
1917                         continue;
1918
1919                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1920                         continue;
1921
1922                 /* Decouple optimistic from tentative for evaluation here.
1923                  * Ban optimistic addresses explicitly, when required.
1924                  */
1925                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1926                             ? (ifp->flags&~IFA_F_TENTATIVE)
1927                             : ifp->flags;
1928                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1929                     !(ifp_flags&banned_flags) &&
1930                     (!dev || ndev == dev ||
1931                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1932                         rcu_read_unlock();
1933                         return ndev;
1934                 }
1935         }
1936
1937         rcu_read_unlock();
1938         return NULL;
1939 }
1940
1941 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1942                             const struct net_device *dev, bool skip_dev_check,
1943                             int strict, u32 banned_flags)
1944 {
1945         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1946                                          strict, banned_flags) ? 1 : 0;
1947 }
1948 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1949
1950
1951 /* Compares an address/prefix_len with addresses on device @dev.
1952  * If one is found it returns true.
1953  */
1954 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1955         const unsigned int prefix_len, struct net_device *dev)
1956 {
1957         const struct inet6_ifaddr *ifa;
1958         const struct inet6_dev *idev;
1959         bool ret = false;
1960
1961         rcu_read_lock();
1962         idev = __in6_dev_get(dev);
1963         if (idev) {
1964                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1965                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1966                         if (ret)
1967                                 break;
1968                 }
1969         }
1970         rcu_read_unlock();
1971
1972         return ret;
1973 }
1974 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1975
1976 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1977 {
1978         const struct inet6_ifaddr *ifa;
1979         const struct inet6_dev *idev;
1980         int     onlink;
1981
1982         onlink = 0;
1983         rcu_read_lock();
1984         idev = __in6_dev_get(dev);
1985         if (idev) {
1986                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1987                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1988                                                    ifa->prefix_len);
1989                         if (onlink)
1990                                 break;
1991                 }
1992         }
1993         rcu_read_unlock();
1994         return onlink;
1995 }
1996 EXPORT_SYMBOL(ipv6_chk_prefix);
1997
1998 /**
1999  * ipv6_dev_find - find the first device with a given source address.
2000  * @net: the net namespace
2001  * @addr: the source address
2002  * @dev: used to find the L3 domain of interest
2003  *
2004  * The caller should be protected by RCU, or RTNL.
2005  */
2006 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2007                                  struct net_device *dev)
2008 {
2009         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2010                                          IFA_F_TENTATIVE);
2011 }
2012 EXPORT_SYMBOL(ipv6_dev_find);
2013
2014 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2015                                      struct net_device *dev, int strict)
2016 {
2017         unsigned int hash = inet6_addr_hash(net, addr);
2018         struct inet6_ifaddr *ifp, *result = NULL;
2019
2020         rcu_read_lock();
2021         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2022                 if (!net_eq(dev_net(ifp->idev->dev), net))
2023                         continue;
2024                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2025                         if (!dev || ifp->idev->dev == dev ||
2026                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2027                                 result = ifp;
2028                                 in6_ifa_hold(ifp);
2029                                 break;
2030                         }
2031                 }
2032         }
2033         rcu_read_unlock();
2034
2035         return result;
2036 }
2037
2038 /* Gets referenced address, destroys ifaddr */
2039
2040 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2041 {
2042         if (dad_failed)
2043                 ifp->flags |= IFA_F_DADFAILED;
2044
2045         if (ifp->flags&IFA_F_TEMPORARY) {
2046                 struct inet6_ifaddr *ifpub;
2047                 spin_lock_bh(&ifp->lock);
2048                 ifpub = ifp->ifpub;
2049                 if (ifpub) {
2050                         in6_ifa_hold(ifpub);
2051                         spin_unlock_bh(&ifp->lock);
2052                         ipv6_create_tempaddr(ifpub, true);
2053                         in6_ifa_put(ifpub);
2054                 } else {
2055                         spin_unlock_bh(&ifp->lock);
2056                 }
2057                 ipv6_del_addr(ifp);
2058         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2059                 spin_lock_bh(&ifp->lock);
2060                 addrconf_del_dad_work(ifp);
2061                 ifp->flags |= IFA_F_TENTATIVE;
2062                 if (dad_failed)
2063                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2064                 spin_unlock_bh(&ifp->lock);
2065                 if (dad_failed)
2066                         ipv6_ifa_notify(0, ifp);
2067                 in6_ifa_put(ifp);
2068         } else {
2069                 ipv6_del_addr(ifp);
2070         }
2071 }
2072
2073 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2074 {
2075         int err = -ENOENT;
2076
2077         spin_lock_bh(&ifp->lock);
2078         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2079                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2080                 err = 0;
2081         }
2082         spin_unlock_bh(&ifp->lock);
2083
2084         return err;
2085 }
2086
2087 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2088 {
2089         struct inet6_dev *idev = ifp->idev;
2090         struct net *net = dev_net(ifp->idev->dev);
2091
2092         if (addrconf_dad_end(ifp)) {
2093                 in6_ifa_put(ifp);
2094                 return;
2095         }
2096
2097         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2098                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2099
2100         spin_lock_bh(&ifp->lock);
2101
2102         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2103                 struct in6_addr new_addr;
2104                 struct inet6_ifaddr *ifp2;
2105                 int retries = ifp->stable_privacy_retry + 1;
2106                 struct ifa6_config cfg = {
2107                         .pfx = &new_addr,
2108                         .plen = ifp->prefix_len,
2109                         .ifa_flags = ifp->flags,
2110                         .valid_lft = ifp->valid_lft,
2111                         .preferred_lft = ifp->prefered_lft,
2112                         .scope = ifp->scope,
2113                 };
2114
2115                 if (retries > net->ipv6.sysctl.idgen_retries) {
2116                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2117                                              ifp->idev->dev->name);
2118                         goto errdad;
2119                 }
2120
2121                 new_addr = ifp->addr;
2122                 if (ipv6_generate_stable_address(&new_addr, retries,
2123                                                  idev))
2124                         goto errdad;
2125
2126                 spin_unlock_bh(&ifp->lock);
2127
2128                 if (idev->cnf.max_addresses &&
2129                     ipv6_count_addresses(idev) >=
2130                     idev->cnf.max_addresses)
2131                         goto lock_errdad;
2132
2133                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2134                                      ifp->idev->dev->name);
2135
2136                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2137                 if (IS_ERR(ifp2))
2138                         goto lock_errdad;
2139
2140                 spin_lock_bh(&ifp2->lock);
2141                 ifp2->stable_privacy_retry = retries;
2142                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2143                 spin_unlock_bh(&ifp2->lock);
2144
2145                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2146                 in6_ifa_put(ifp2);
2147 lock_errdad:
2148                 spin_lock_bh(&ifp->lock);
2149         }
2150
2151 errdad:
2152         /* transition from _POSTDAD to _ERRDAD */
2153         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2154         spin_unlock_bh(&ifp->lock);
2155
2156         addrconf_mod_dad_work(ifp, 0);
2157         in6_ifa_put(ifp);
2158 }
2159
2160 /* Join to solicited addr multicast group.
2161  * caller must hold RTNL */
2162 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2163 {
2164         struct in6_addr maddr;
2165
2166         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2167                 return;
2168
2169         addrconf_addr_solict_mult(addr, &maddr);
2170         ipv6_dev_mc_inc(dev, &maddr);
2171 }
2172
2173 /* caller must hold RTNL */
2174 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2175 {
2176         struct in6_addr maddr;
2177
2178         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2179                 return;
2180
2181         addrconf_addr_solict_mult(addr, &maddr);
2182         __ipv6_dev_mc_dec(idev, &maddr);
2183 }
2184
2185 /* caller must hold RTNL */
2186 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2187 {
2188         struct in6_addr addr;
2189
2190         if (ifp->prefix_len >= 127) /* RFC 6164 */
2191                 return;
2192         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2193         if (ipv6_addr_any(&addr))
2194                 return;
2195         __ipv6_dev_ac_inc(ifp->idev, &addr);
2196 }
2197
2198 /* caller must hold RTNL */
2199 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2200 {
2201         struct in6_addr addr;
2202
2203         if (ifp->prefix_len >= 127) /* RFC 6164 */
2204                 return;
2205         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2206         if (ipv6_addr_any(&addr))
2207                 return;
2208         __ipv6_dev_ac_dec(ifp->idev, &addr);
2209 }
2210
2211 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2212 {
2213         switch (dev->addr_len) {
2214         case ETH_ALEN:
2215                 memcpy(eui, dev->dev_addr, 3);
2216                 eui[3] = 0xFF;
2217                 eui[4] = 0xFE;
2218                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2219                 break;
2220         case EUI64_ADDR_LEN:
2221                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2222                 eui[0] ^= 2;
2223                 break;
2224         default:
2225                 return -1;
2226         }
2227
2228         return 0;
2229 }
2230
2231 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2232 {
2233         union fwnet_hwaddr *ha;
2234
2235         if (dev->addr_len != FWNET_ALEN)
2236                 return -1;
2237
2238         ha = (union fwnet_hwaddr *)dev->dev_addr;
2239
2240         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2241         eui[0] ^= 2;
2242         return 0;
2243 }
2244
2245 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2246 {
2247         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2248         if (dev->addr_len != ARCNET_ALEN)
2249                 return -1;
2250         memset(eui, 0, 7);
2251         eui[7] = *(u8 *)dev->dev_addr;
2252         return 0;
2253 }
2254
2255 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2256 {
2257         if (dev->addr_len != INFINIBAND_ALEN)
2258                 return -1;
2259         memcpy(eui, dev->dev_addr + 12, 8);
2260         eui[0] |= 2;
2261         return 0;
2262 }
2263
2264 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2265 {
2266         if (addr == 0)
2267                 return -1;
2268         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2269                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2270                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2271                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2272                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2273                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2274         eui[1] = 0;
2275         eui[2] = 0x5E;
2276         eui[3] = 0xFE;
2277         memcpy(eui + 4, &addr, 4);
2278         return 0;
2279 }
2280
2281 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2282 {
2283         if (dev->priv_flags & IFF_ISATAP)
2284                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2285         return -1;
2286 }
2287
2288 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2289 {
2290         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2291 }
2292
2293 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2294 {
2295         memcpy(eui, dev->perm_addr, 3);
2296         memcpy(eui + 5, dev->perm_addr + 3, 3);
2297         eui[3] = 0xFF;
2298         eui[4] = 0xFE;
2299         eui[0] ^= 2;
2300         return 0;
2301 }
2302
2303 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2304 {
2305         switch (dev->type) {
2306         case ARPHRD_ETHER:
2307         case ARPHRD_FDDI:
2308                 return addrconf_ifid_eui48(eui, dev);
2309         case ARPHRD_ARCNET:
2310                 return addrconf_ifid_arcnet(eui, dev);
2311         case ARPHRD_INFINIBAND:
2312                 return addrconf_ifid_infiniband(eui, dev);
2313         case ARPHRD_SIT:
2314                 return addrconf_ifid_sit(eui, dev);
2315         case ARPHRD_IPGRE:
2316         case ARPHRD_TUNNEL:
2317                 return addrconf_ifid_gre(eui, dev);
2318         case ARPHRD_6LOWPAN:
2319                 return addrconf_ifid_6lowpan(eui, dev);
2320         case ARPHRD_IEEE1394:
2321                 return addrconf_ifid_ieee1394(eui, dev);
2322         case ARPHRD_TUNNEL6:
2323         case ARPHRD_IP6GRE:
2324         case ARPHRD_RAWIP:
2325                 return addrconf_ifid_ip6tnl(eui, dev);
2326         }
2327         return -1;
2328 }
2329
2330 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2331 {
2332         int err = -1;
2333         struct inet6_ifaddr *ifp;
2334
2335         read_lock_bh(&idev->lock);
2336         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2337                 if (ifp->scope > IFA_LINK)
2338                         break;
2339                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2340                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2341                         err = 0;
2342                         break;
2343                 }
2344         }
2345         read_unlock_bh(&idev->lock);
2346         return err;
2347 }
2348
2349 /* Generation of a randomized Interface Identifier
2350  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2351  */
2352
2353 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2354 {
2355 regen:
2356         get_random_bytes(&addr->s6_addr[8], 8);
2357
2358         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2359          * check if generated address is not inappropriate:
2360          *
2361          * - Reserved IPv6 Interface Identifiers
2362          * - XXX: already assigned to an address on the device
2363          */
2364
2365         /* Subnet-router anycast: 0000:0000:0000:0000 */
2366         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2367                 goto regen;
2368
2369         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2370          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2371          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2372          */
2373         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2374             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2375                 goto regen;
2376
2377         /* Reserved subnet anycast addresses */
2378         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2379             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2380                 goto regen;
2381 }
2382
2383 /*
2384  *      Add prefix route.
2385  */
2386
2387 static void
2388 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2389                       struct net_device *dev, unsigned long expires,
2390                       u32 flags, gfp_t gfp_flags)
2391 {
2392         struct fib6_config cfg = {
2393                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2394                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2395                 .fc_ifindex = dev->ifindex,
2396                 .fc_expires = expires,
2397                 .fc_dst_len = plen,
2398                 .fc_flags = RTF_UP | flags,
2399                 .fc_nlinfo.nl_net = dev_net(dev),
2400                 .fc_protocol = RTPROT_KERNEL,
2401                 .fc_type = RTN_UNICAST,
2402         };
2403
2404         cfg.fc_dst = *pfx;
2405
2406         /* Prevent useless cloning on PtP SIT.
2407            This thing is done here expecting that the whole
2408            class of non-broadcast devices need not cloning.
2409          */
2410 #if IS_ENABLED(CONFIG_IPV6_SIT)
2411         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2412                 cfg.fc_flags |= RTF_NONEXTHOP;
2413 #endif
2414
2415         ip6_route_add(&cfg, gfp_flags, NULL);
2416 }
2417
2418
2419 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2420                                                   int plen,
2421                                                   const struct net_device *dev,
2422                                                   u32 flags, u32 noflags,
2423                                                   bool no_gw)
2424 {
2425         struct fib6_node *fn;
2426         struct fib6_info *rt = NULL;
2427         struct fib6_table *table;
2428         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2429
2430         table = fib6_get_table(dev_net(dev), tb_id);
2431         if (!table)
2432                 return NULL;
2433
2434         rcu_read_lock();
2435         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2436         if (!fn)
2437                 goto out;
2438
2439         for_each_fib6_node_rt_rcu(fn) {
2440                 /* prefix routes only use builtin fib6_nh */
2441                 if (rt->nh)
2442                         continue;
2443
2444                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2445                         continue;
2446                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2447                         continue;
2448                 if ((rt->fib6_flags & flags) != flags)
2449                         continue;
2450                 if ((rt->fib6_flags & noflags) != 0)
2451                         continue;
2452                 if (!fib6_info_hold_safe(rt))
2453                         continue;
2454                 break;
2455         }
2456 out:
2457         rcu_read_unlock();
2458         return rt;
2459 }
2460
2461
2462 /* Create "default" multicast route to the interface */
2463
2464 static void addrconf_add_mroute(struct net_device *dev)
2465 {
2466         struct fib6_config cfg = {
2467                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2468                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2469                 .fc_ifindex = dev->ifindex,
2470                 .fc_dst_len = 8,
2471                 .fc_flags = RTF_UP,
2472                 .fc_type = RTN_MULTICAST,
2473                 .fc_nlinfo.nl_net = dev_net(dev),
2474                 .fc_protocol = RTPROT_KERNEL,
2475         };
2476
2477         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2478
2479         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2480 }
2481
2482 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2483 {
2484         struct inet6_dev *idev;
2485
2486         ASSERT_RTNL();
2487
2488         idev = ipv6_find_idev(dev);
2489         if (IS_ERR(idev))
2490                 return idev;
2491
2492         if (idev->cnf.disable_ipv6)
2493                 return ERR_PTR(-EACCES);
2494
2495         /* Add default multicast route */
2496         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2497                 addrconf_add_mroute(dev);
2498
2499         return idev;
2500 }
2501
2502 static void manage_tempaddrs(struct inet6_dev *idev,
2503                              struct inet6_ifaddr *ifp,
2504                              __u32 valid_lft, __u32 prefered_lft,
2505                              bool create, unsigned long now)
2506 {
2507         u32 flags;
2508         struct inet6_ifaddr *ift;
2509
2510         read_lock_bh(&idev->lock);
2511         /* update all temporary addresses in the list */
2512         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2513                 int age, max_valid, max_prefered;
2514
2515                 if (ifp != ift->ifpub)
2516                         continue;
2517
2518                 /* RFC 4941 section 3.3:
2519                  * If a received option will extend the lifetime of a public
2520                  * address, the lifetimes of temporary addresses should
2521                  * be extended, subject to the overall constraint that no
2522                  * temporary addresses should ever remain "valid" or "preferred"
2523                  * for a time longer than (TEMP_VALID_LIFETIME) or
2524                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2525                  */
2526                 age = (now - ift->cstamp) / HZ;
2527                 max_valid = idev->cnf.temp_valid_lft - age;
2528                 if (max_valid < 0)
2529                         max_valid = 0;
2530
2531                 max_prefered = idev->cnf.temp_prefered_lft -
2532                                idev->desync_factor - age;
2533                 if (max_prefered < 0)
2534                         max_prefered = 0;
2535
2536                 if (valid_lft > max_valid)
2537                         valid_lft = max_valid;
2538
2539                 if (prefered_lft > max_prefered)
2540                         prefered_lft = max_prefered;
2541
2542                 spin_lock(&ift->lock);
2543                 flags = ift->flags;
2544                 ift->valid_lft = valid_lft;
2545                 ift->prefered_lft = prefered_lft;
2546                 ift->tstamp = now;
2547                 if (prefered_lft > 0)
2548                         ift->flags &= ~IFA_F_DEPRECATED;
2549
2550                 spin_unlock(&ift->lock);
2551                 if (!(flags&IFA_F_TENTATIVE))
2552                         ipv6_ifa_notify(0, ift);
2553         }
2554
2555         if ((create || list_empty(&idev->tempaddr_list)) &&
2556             idev->cnf.use_tempaddr > 0) {
2557                 /* When a new public address is created as described
2558                  * in [ADDRCONF], also create a new temporary address.
2559                  * Also create a temporary address if it's enabled but
2560                  * no temporary address currently exists.
2561                  */
2562                 read_unlock_bh(&idev->lock);
2563                 ipv6_create_tempaddr(ifp, false);
2564         } else {
2565                 read_unlock_bh(&idev->lock);
2566         }
2567 }
2568
2569 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2570 {
2571         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2572                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2573 }
2574
2575 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2576                                  const struct prefix_info *pinfo,
2577                                  struct inet6_dev *in6_dev,
2578                                  const struct in6_addr *addr, int addr_type,
2579                                  u32 addr_flags, bool sllao, bool tokenized,
2580                                  __u32 valid_lft, u32 prefered_lft)
2581 {
2582         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2583         int create = 0;
2584
2585         if (!ifp && valid_lft) {
2586                 int max_addresses = in6_dev->cnf.max_addresses;
2587                 struct ifa6_config cfg = {
2588                         .pfx = addr,
2589                         .plen = pinfo->prefix_len,
2590                         .ifa_flags = addr_flags,
2591                         .valid_lft = valid_lft,
2592                         .preferred_lft = prefered_lft,
2593                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2594                 };
2595
2596 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2597                 if ((net->ipv6.devconf_all->optimistic_dad ||
2598                      in6_dev->cnf.optimistic_dad) &&
2599                     !net->ipv6.devconf_all->forwarding && sllao)
2600                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2601 #endif
2602
2603                 /* Do not allow to create too much of autoconfigured
2604                  * addresses; this would be too easy way to crash kernel.
2605                  */
2606                 if (!max_addresses ||
2607                     ipv6_count_addresses(in6_dev) < max_addresses)
2608                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2609
2610                 if (IS_ERR_OR_NULL(ifp))
2611                         return -1;
2612
2613                 create = 1;
2614                 spin_lock_bh(&ifp->lock);
2615                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2616                 ifp->cstamp = jiffies;
2617                 ifp->tokenized = tokenized;
2618                 spin_unlock_bh(&ifp->lock);
2619                 addrconf_dad_start(ifp);
2620         }
2621
2622         if (ifp) {
2623                 u32 flags;
2624                 unsigned long now;
2625                 u32 stored_lft;
2626
2627                 /* Update lifetime (RFC4862 5.5.3 e)
2628                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2629                  * improve the reaction of SLAAC to renumbering events
2630                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2631                  */
2632                 spin_lock_bh(&ifp->lock);
2633                 now = jiffies;
2634                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2635                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2636                 else
2637                         stored_lft = 0;
2638
2639                 if (!create && stored_lft) {
2640                         ifp->valid_lft = valid_lft;
2641                         ifp->prefered_lft = prefered_lft;
2642                         ifp->tstamp = now;
2643                         flags = ifp->flags;
2644                         ifp->flags &= ~IFA_F_DEPRECATED;
2645                         spin_unlock_bh(&ifp->lock);
2646
2647                         if (!(flags&IFA_F_TENTATIVE))
2648                                 ipv6_ifa_notify(0, ifp);
2649                 } else
2650                         spin_unlock_bh(&ifp->lock);
2651
2652                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2653                                  create, now);
2654
2655                 in6_ifa_put(ifp);
2656                 addrconf_verify();
2657         }
2658
2659         return 0;
2660 }
2661 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2662
2663 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2664 {
2665         struct prefix_info *pinfo;
2666         __u32 valid_lft;
2667         __u32 prefered_lft;
2668         int addr_type, err;
2669         u32 addr_flags = 0;
2670         struct inet6_dev *in6_dev;
2671         struct net *net = dev_net(dev);
2672
2673         pinfo = (struct prefix_info *) opt;
2674
2675         if (len < sizeof(struct prefix_info)) {
2676                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2677                 return;
2678         }
2679
2680         /*
2681          *      Validation checks ([ADDRCONF], page 19)
2682          */
2683
2684         addr_type = ipv6_addr_type(&pinfo->prefix);
2685
2686         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2687                 return;
2688
2689         valid_lft = ntohl(pinfo->valid);
2690         prefered_lft = ntohl(pinfo->prefered);
2691
2692         if (prefered_lft > valid_lft) {
2693                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2694                 return;
2695         }
2696
2697         in6_dev = in6_dev_get(dev);
2698
2699         if (!in6_dev) {
2700                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2701                                     dev->name);
2702                 return;
2703         }
2704
2705         /*
2706          *      Two things going on here:
2707          *      1) Add routes for on-link prefixes
2708          *      2) Configure prefixes with the auto flag set
2709          */
2710
2711         if (pinfo->onlink) {
2712                 struct fib6_info *rt;
2713                 unsigned long rt_expires;
2714
2715                 /* Avoid arithmetic overflow. Really, we could
2716                  * save rt_expires in seconds, likely valid_lft,
2717                  * but it would require division in fib gc, that it
2718                  * not good.
2719                  */
2720                 if (HZ > USER_HZ)
2721                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2722                 else
2723                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2724
2725                 if (addrconf_finite_timeout(rt_expires))
2726                         rt_expires *= HZ;
2727
2728                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2729                                                pinfo->prefix_len,
2730                                                dev,
2731                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2732                                                RTF_DEFAULT, true);
2733
2734                 if (rt) {
2735                         /* Autoconf prefix route */
2736                         if (valid_lft == 0) {
2737                                 ip6_del_rt(net, rt, false);
2738                                 rt = NULL;
2739                         } else if (addrconf_finite_timeout(rt_expires)) {
2740                                 /* not infinity */
2741                                 fib6_set_expires(rt, jiffies + rt_expires);
2742                         } else {
2743                                 fib6_clean_expires(rt);
2744                         }
2745                 } else if (valid_lft) {
2746                         clock_t expires = 0;
2747                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2748                         if (addrconf_finite_timeout(rt_expires)) {
2749                                 /* not infinity */
2750                                 flags |= RTF_EXPIRES;
2751                                 expires = jiffies_to_clock_t(rt_expires);
2752                         }
2753                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2754                                               0, dev, expires, flags,
2755                                               GFP_ATOMIC);
2756                 }
2757                 fib6_info_release(rt);
2758         }
2759
2760         /* Try to figure out our local address for this prefix */
2761
2762         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2763                 struct in6_addr addr;
2764                 bool tokenized = false, dev_addr_generated = false;
2765
2766                 if (pinfo->prefix_len == 64) {
2767                         memcpy(&addr, &pinfo->prefix, 8);
2768
2769                         if (!ipv6_addr_any(&in6_dev->token)) {
2770                                 read_lock_bh(&in6_dev->lock);
2771                                 memcpy(addr.s6_addr + 8,
2772                                        in6_dev->token.s6_addr + 8, 8);
2773                                 read_unlock_bh(&in6_dev->lock);
2774                                 tokenized = true;
2775                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2776                                    !ipv6_generate_stable_address(&addr, 0,
2777                                                                  in6_dev)) {
2778                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2779                                 goto ok;
2780                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2781                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2782                                 goto put;
2783                         } else {
2784                                 dev_addr_generated = true;
2785                         }
2786                         goto ok;
2787                 }
2788                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2789                                     pinfo->prefix_len);
2790                 goto put;
2791
2792 ok:
2793                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2794                                                    &addr, addr_type,
2795                                                    addr_flags, sllao,
2796                                                    tokenized, valid_lft,
2797                                                    prefered_lft);
2798                 if (err)
2799                         goto put;
2800
2801                 /* Ignore error case here because previous prefix add addr was
2802                  * successful which will be notified.
2803                  */
2804                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2805                                               addr_type, addr_flags, sllao,
2806                                               tokenized, valid_lft,
2807                                               prefered_lft,
2808                                               dev_addr_generated);
2809         }
2810         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2811 put:
2812         in6_dev_put(in6_dev);
2813 }
2814
2815 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2816                 struct in6_ifreq *ireq)
2817 {
2818         struct ip_tunnel_parm p = { };
2819         int err;
2820
2821         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2822                 return -EADDRNOTAVAIL;
2823
2824         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2825         p.iph.version = 4;
2826         p.iph.ihl = 5;
2827         p.iph.protocol = IPPROTO_IPV6;
2828         p.iph.ttl = 64;
2829
2830         if (!dev->netdev_ops->ndo_tunnel_ctl)
2831                 return -EOPNOTSUPP;
2832         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2833         if (err)
2834                 return err;
2835
2836         dev = __dev_get_by_name(net, p.name);
2837         if (!dev)
2838                 return -ENOBUFS;
2839         return dev_open(dev, NULL);
2840 }
2841
2842 /*
2843  *      Set destination address.
2844  *      Special case for SIT interfaces where we create a new "virtual"
2845  *      device.
2846  */
2847 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2848 {
2849         struct net_device *dev;
2850         struct in6_ifreq ireq;
2851         int err = -ENODEV;
2852
2853         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2854                 return -ENODEV;
2855         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2856                 return -EFAULT;
2857
2858         rtnl_lock();
2859         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2860         if (dev && dev->type == ARPHRD_SIT)
2861                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2862         rtnl_unlock();
2863         return err;
2864 }
2865
2866 static int ipv6_mc_config(struct sock *sk, bool join,
2867                           const struct in6_addr *addr, int ifindex)
2868 {
2869         int ret;
2870
2871         ASSERT_RTNL();
2872
2873         lock_sock(sk);
2874         if (join)
2875                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2876         else
2877                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2878         release_sock(sk);
2879
2880         return ret;
2881 }
2882
2883 /*
2884  *      Manual configuration of address on an interface
2885  */
2886 static int inet6_addr_add(struct net *net, int ifindex,
2887                           struct ifa6_config *cfg,
2888                           struct netlink_ext_ack *extack)
2889 {
2890         struct inet6_ifaddr *ifp;
2891         struct inet6_dev *idev;
2892         struct net_device *dev;
2893         unsigned long timeout;
2894         clock_t expires;
2895         u32 flags;
2896
2897         ASSERT_RTNL();
2898
2899         if (cfg->plen > 128)
2900                 return -EINVAL;
2901
2902         /* check the lifetime */
2903         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2904                 return -EINVAL;
2905
2906         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2907                 return -EINVAL;
2908
2909         dev = __dev_get_by_index(net, ifindex);
2910         if (!dev)
2911                 return -ENODEV;
2912
2913         idev = addrconf_add_dev(dev);
2914         if (IS_ERR(idev))
2915                 return PTR_ERR(idev);
2916
2917         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2918                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2919                                          true, cfg->pfx, ifindex);
2920
2921                 if (ret < 0)
2922                         return ret;
2923         }
2924
2925         cfg->scope = ipv6_addr_scope(cfg->pfx);
2926
2927         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2928         if (addrconf_finite_timeout(timeout)) {
2929                 expires = jiffies_to_clock_t(timeout * HZ);
2930                 cfg->valid_lft = timeout;
2931                 flags = RTF_EXPIRES;
2932         } else {
2933                 expires = 0;
2934                 flags = 0;
2935                 cfg->ifa_flags |= IFA_F_PERMANENT;
2936         }
2937
2938         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2939         if (addrconf_finite_timeout(timeout)) {
2940                 if (timeout == 0)
2941                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2942                 cfg->preferred_lft = timeout;
2943         }
2944
2945         ifp = ipv6_add_addr(idev, cfg, true, extack);
2946         if (!IS_ERR(ifp)) {
2947                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2948                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2949                                               ifp->rt_priority, dev, expires,
2950                                               flags, GFP_KERNEL);
2951                 }
2952
2953                 /* Send a netlink notification if DAD is enabled and
2954                  * optimistic flag is not set
2955                  */
2956                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2957                         ipv6_ifa_notify(0, ifp);
2958                 /*
2959                  * Note that section 3.1 of RFC 4429 indicates
2960                  * that the Optimistic flag should not be set for
2961                  * manually configured addresses
2962                  */
2963                 addrconf_dad_start(ifp);
2964                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2965                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2966                                          cfg->preferred_lft, true, jiffies);
2967                 in6_ifa_put(ifp);
2968                 addrconf_verify_rtnl();
2969                 return 0;
2970         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2971                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2972                                cfg->pfx, ifindex);
2973         }
2974
2975         return PTR_ERR(ifp);
2976 }
2977
2978 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2979                           const struct in6_addr *pfx, unsigned int plen)
2980 {
2981         struct inet6_ifaddr *ifp;
2982         struct inet6_dev *idev;
2983         struct net_device *dev;
2984
2985         if (plen > 128)
2986                 return -EINVAL;
2987
2988         dev = __dev_get_by_index(net, ifindex);
2989         if (!dev)
2990                 return -ENODEV;
2991
2992         idev = __in6_dev_get(dev);
2993         if (!idev)
2994                 return -ENXIO;
2995
2996         read_lock_bh(&idev->lock);
2997         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2998                 if (ifp->prefix_len == plen &&
2999                     ipv6_addr_equal(pfx, &ifp->addr)) {
3000                         in6_ifa_hold(ifp);
3001                         read_unlock_bh(&idev->lock);
3002
3003                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3004                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3005                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3006                                                  jiffies);
3007                         ipv6_del_addr(ifp);
3008                         addrconf_verify_rtnl();
3009                         if (ipv6_addr_is_multicast(pfx)) {
3010                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3011                                                false, pfx, dev->ifindex);
3012                         }
3013                         return 0;
3014                 }
3015         }
3016         read_unlock_bh(&idev->lock);
3017         return -EADDRNOTAVAIL;
3018 }
3019
3020
3021 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3022 {
3023         struct ifa6_config cfg = {
3024                 .ifa_flags = IFA_F_PERMANENT,
3025                 .preferred_lft = INFINITY_LIFE_TIME,
3026                 .valid_lft = INFINITY_LIFE_TIME,
3027         };
3028         struct in6_ifreq ireq;
3029         int err;
3030
3031         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3032                 return -EPERM;
3033
3034         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3035                 return -EFAULT;
3036
3037         cfg.pfx = &ireq.ifr6_addr;
3038         cfg.plen = ireq.ifr6_prefixlen;
3039
3040         rtnl_lock();
3041         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3042         rtnl_unlock();
3043         return err;
3044 }
3045
3046 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3047 {
3048         struct in6_ifreq ireq;
3049         int err;
3050
3051         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3052                 return -EPERM;
3053
3054         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3055                 return -EFAULT;
3056
3057         rtnl_lock();
3058         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3059                              ireq.ifr6_prefixlen);
3060         rtnl_unlock();
3061         return err;
3062 }
3063
3064 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3065                      int plen, int scope)
3066 {
3067         struct inet6_ifaddr *ifp;
3068         struct ifa6_config cfg = {
3069                 .pfx = addr,
3070                 .plen = plen,
3071                 .ifa_flags = IFA_F_PERMANENT,
3072                 .valid_lft = INFINITY_LIFE_TIME,
3073                 .preferred_lft = INFINITY_LIFE_TIME,
3074                 .scope = scope
3075         };
3076
3077         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3078         if (!IS_ERR(ifp)) {
3079                 spin_lock_bh(&ifp->lock);
3080                 ifp->flags &= ~IFA_F_TENTATIVE;
3081                 spin_unlock_bh(&ifp->lock);
3082                 rt_genid_bump_ipv6(dev_net(idev->dev));
3083                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3084                 in6_ifa_put(ifp);
3085         }
3086 }
3087
3088 #if IS_ENABLED(CONFIG_IPV6_SIT)
3089 static void sit_add_v4_addrs(struct inet6_dev *idev)
3090 {
3091         struct in6_addr addr;
3092         struct net_device *dev;
3093         struct net *net = dev_net(idev->dev);
3094         int scope, plen;
3095         u32 pflags = 0;
3096
3097         ASSERT_RTNL();
3098
3099         memset(&addr, 0, sizeof(struct in6_addr));
3100         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3101
3102         if (idev->dev->flags&IFF_POINTOPOINT) {
3103                 addr.s6_addr32[0] = htonl(0xfe800000);
3104                 scope = IFA_LINK;
3105                 plen = 64;
3106         } else {
3107                 scope = IPV6_ADDR_COMPATv4;
3108                 plen = 96;
3109                 pflags |= RTF_NONEXTHOP;
3110         }
3111
3112         if (addr.s6_addr32[3]) {
3113                 add_addr(idev, &addr, plen, scope);
3114                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3115                                       GFP_KERNEL);
3116                 return;
3117         }
3118
3119         for_each_netdev(net, dev) {
3120                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3121                 if (in_dev && (dev->flags & IFF_UP)) {
3122                         struct in_ifaddr *ifa;
3123                         int flag = scope;
3124
3125                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3126                                 addr.s6_addr32[3] = ifa->ifa_local;
3127
3128                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3129                                         continue;
3130                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3131                                         if (idev->dev->flags&IFF_POINTOPOINT)
3132                                                 continue;
3133                                         flag |= IFA_HOST;
3134                                 }
3135
3136                                 add_addr(idev, &addr, plen, flag);
3137                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3138                                                       0, pflags, GFP_KERNEL);
3139                         }
3140                 }
3141         }
3142 }
3143 #endif
3144
3145 static void init_loopback(struct net_device *dev)
3146 {
3147         struct inet6_dev  *idev;
3148
3149         /* ::1 */
3150
3151         ASSERT_RTNL();
3152
3153         idev = ipv6_find_idev(dev);
3154         if (IS_ERR(idev)) {
3155                 pr_debug("%s: add_dev failed\n", __func__);
3156                 return;
3157         }
3158
3159         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3160 }
3161
3162 void addrconf_add_linklocal(struct inet6_dev *idev,
3163                             const struct in6_addr *addr, u32 flags)
3164 {
3165         struct ifa6_config cfg = {
3166                 .pfx = addr,
3167                 .plen = 64,
3168                 .ifa_flags = flags | IFA_F_PERMANENT,
3169                 .valid_lft = INFINITY_LIFE_TIME,
3170                 .preferred_lft = INFINITY_LIFE_TIME,
3171                 .scope = IFA_LINK
3172         };
3173         struct inet6_ifaddr *ifp;
3174
3175 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3176         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3177              idev->cnf.optimistic_dad) &&
3178             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3179                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3180 #endif
3181
3182         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3183         if (!IS_ERR(ifp)) {
3184                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3185                                       0, 0, GFP_ATOMIC);
3186                 addrconf_dad_start(ifp);
3187                 in6_ifa_put(ifp);
3188         }
3189 }
3190 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3191
3192 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3193 {
3194         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3195                 return true;
3196
3197         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3198             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3199                 return true;
3200
3201         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3202             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3203                 return true;
3204
3205         return false;
3206 }
3207
3208 static int ipv6_generate_stable_address(struct in6_addr *address,
3209                                         u8 dad_count,
3210                                         const struct inet6_dev *idev)
3211 {
3212         static DEFINE_SPINLOCK(lock);
3213         static __u32 digest[SHA1_DIGEST_WORDS];
3214         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3215
3216         static union {
3217                 char __data[SHA1_BLOCK_SIZE];
3218                 struct {
3219                         struct in6_addr secret;
3220                         __be32 prefix[2];
3221                         unsigned char hwaddr[MAX_ADDR_LEN];
3222                         u8 dad_count;
3223                 } __packed;
3224         } data;
3225
3226         struct in6_addr secret;
3227         struct in6_addr temp;
3228         struct net *net = dev_net(idev->dev);
3229
3230         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3231
3232         if (idev->cnf.stable_secret.initialized)
3233                 secret = idev->cnf.stable_secret.secret;
3234         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3235                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3236         else
3237                 return -1;
3238
3239 retry:
3240         spin_lock_bh(&lock);
3241
3242         sha1_init(digest);
3243         memset(&data, 0, sizeof(data));
3244         memset(workspace, 0, sizeof(workspace));
3245         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3246         data.prefix[0] = address->s6_addr32[0];
3247         data.prefix[1] = address->s6_addr32[1];
3248         data.secret = secret;
3249         data.dad_count = dad_count;
3250
3251         sha1_transform(digest, data.__data, workspace);
3252
3253         temp = *address;
3254         temp.s6_addr32[2] = (__force __be32)digest[0];
3255         temp.s6_addr32[3] = (__force __be32)digest[1];
3256
3257         spin_unlock_bh(&lock);
3258
3259         if (ipv6_reserved_interfaceid(temp)) {
3260                 dad_count++;
3261                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3262                         return -1;
3263                 goto retry;
3264         }
3265
3266         *address = temp;
3267         return 0;
3268 }
3269
3270 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3271 {
3272         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3273
3274         if (s->initialized)
3275                 return;
3276         s = &idev->cnf.stable_secret;
3277         get_random_bytes(&s->secret, sizeof(s->secret));
3278         s->initialized = true;
3279 }
3280
3281 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3282 {
3283         struct in6_addr addr;
3284
3285         /* no link local addresses on L3 master devices */
3286         if (netif_is_l3_master(idev->dev))
3287                 return;
3288
3289         /* no link local addresses on devices flagged as slaves */
3290         if (idev->dev->flags & IFF_SLAVE)
3291                 return;
3292
3293         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3294
3295         switch (idev->cnf.addr_gen_mode) {
3296         case IN6_ADDR_GEN_MODE_RANDOM:
3297                 ipv6_gen_mode_random_init(idev);
3298                 fallthrough;
3299         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3300                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3301                         addrconf_add_linklocal(idev, &addr,
3302                                                IFA_F_STABLE_PRIVACY);
3303                 else if (prefix_route)
3304                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3305                                               0, 0, GFP_KERNEL);
3306                 break;
3307         case IN6_ADDR_GEN_MODE_EUI64:
3308                 /* addrconf_add_linklocal also adds a prefix_route and we
3309                  * only need to care about prefix routes if ipv6_generate_eui64
3310                  * couldn't generate one.
3311                  */
3312                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3313                         addrconf_add_linklocal(idev, &addr, 0);
3314                 else if (prefix_route)
3315                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3316                                               0, 0, GFP_KERNEL);
3317                 break;
3318         case IN6_ADDR_GEN_MODE_NONE:
3319         default:
3320                 /* will not add any link local address */
3321                 break;
3322         }
3323 }
3324
3325 static void addrconf_dev_config(struct net_device *dev)
3326 {
3327         struct inet6_dev *idev;
3328
3329         ASSERT_RTNL();
3330
3331         if ((dev->type != ARPHRD_ETHER) &&
3332             (dev->type != ARPHRD_FDDI) &&
3333             (dev->type != ARPHRD_ARCNET) &&
3334             (dev->type != ARPHRD_INFINIBAND) &&
3335             (dev->type != ARPHRD_IEEE1394) &&
3336             (dev->type != ARPHRD_TUNNEL6) &&
3337             (dev->type != ARPHRD_6LOWPAN) &&
3338             (dev->type != ARPHRD_IP6GRE) &&
3339             (dev->type != ARPHRD_IPGRE) &&
3340             (dev->type != ARPHRD_TUNNEL) &&
3341             (dev->type != ARPHRD_NONE) &&
3342             (dev->type != ARPHRD_RAWIP)) {
3343                 /* Alas, we support only Ethernet autoconfiguration. */
3344                 idev = __in6_dev_get(dev);
3345                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3346                     dev->flags & IFF_MULTICAST)
3347                         ipv6_mc_up(idev);
3348                 return;
3349         }
3350
3351         idev = addrconf_add_dev(dev);
3352         if (IS_ERR(idev))
3353                 return;
3354
3355         /* this device type has no EUI support */
3356         if (dev->type == ARPHRD_NONE &&
3357             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3358                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3359
3360         addrconf_addr_gen(idev, false);
3361 }
3362
3363 #if IS_ENABLED(CONFIG_IPV6_SIT)
3364 static void addrconf_sit_config(struct net_device *dev)
3365 {
3366         struct inet6_dev *idev;
3367
3368         ASSERT_RTNL();
3369
3370         /*
3371          * Configure the tunnel with one of our IPv4
3372          * addresses... we should configure all of
3373          * our v4 addrs in the tunnel
3374          */
3375
3376         idev = ipv6_find_idev(dev);
3377         if (IS_ERR(idev)) {
3378                 pr_debug("%s: add_dev failed\n", __func__);
3379                 return;
3380         }
3381
3382         if (dev->priv_flags & IFF_ISATAP) {
3383                 addrconf_addr_gen(idev, false);
3384                 return;
3385         }
3386
3387         sit_add_v4_addrs(idev);
3388
3389         if (dev->flags&IFF_POINTOPOINT)
3390                 addrconf_add_mroute(dev);
3391 }
3392 #endif
3393
3394 #if IS_ENABLED(CONFIG_NET_IPGRE)
3395 static void addrconf_gre_config(struct net_device *dev)
3396 {
3397         struct inet6_dev *idev;
3398
3399         ASSERT_RTNL();
3400
3401         idev = ipv6_find_idev(dev);
3402         if (IS_ERR(idev)) {
3403                 pr_debug("%s: add_dev failed\n", __func__);
3404                 return;
3405         }
3406
3407         addrconf_addr_gen(idev, true);
3408         if (dev->flags & IFF_POINTOPOINT)
3409                 addrconf_add_mroute(dev);
3410 }
3411 #endif
3412
3413 static int fixup_permanent_addr(struct net *net,
3414                                 struct inet6_dev *idev,
3415                                 struct inet6_ifaddr *ifp)
3416 {
3417         /* !fib6_node means the host route was removed from the
3418          * FIB, for example, if 'lo' device is taken down. In that
3419          * case regenerate the host route.
3420          */
3421         if (!ifp->rt || !ifp->rt->fib6_node) {
3422                 struct fib6_info *f6i, *prev;
3423
3424                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3425                                          GFP_ATOMIC);
3426                 if (IS_ERR(f6i))
3427                         return PTR_ERR(f6i);
3428
3429                 /* ifp->rt can be accessed outside of rtnl */
3430                 spin_lock(&ifp->lock);
3431                 prev = ifp->rt;
3432                 ifp->rt = f6i;
3433                 spin_unlock(&ifp->lock);
3434
3435                 fib6_info_release(prev);
3436         }
3437
3438         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3439                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3440                                       ifp->rt_priority, idev->dev, 0, 0,
3441                                       GFP_ATOMIC);
3442         }
3443
3444         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3445                 addrconf_dad_start(ifp);
3446
3447         return 0;
3448 }
3449
3450 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3451 {
3452         struct inet6_ifaddr *ifp, *tmp;
3453         struct inet6_dev *idev;
3454
3455         idev = __in6_dev_get(dev);
3456         if (!idev)
3457                 return;
3458
3459         write_lock_bh(&idev->lock);
3460
3461         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3462                 if ((ifp->flags & IFA_F_PERMANENT) &&
3463                     fixup_permanent_addr(net, idev, ifp) < 0) {
3464                         write_unlock_bh(&idev->lock);
3465                         in6_ifa_hold(ifp);
3466                         ipv6_del_addr(ifp);
3467                         write_lock_bh(&idev->lock);
3468
3469                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3470                                              idev->dev->name, &ifp->addr);
3471                 }
3472         }
3473
3474         write_unlock_bh(&idev->lock);
3475 }
3476
3477 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3478                            void *ptr)
3479 {
3480         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3481         struct netdev_notifier_change_info *change_info;
3482         struct netdev_notifier_changeupper_info *info;
3483         struct inet6_dev *idev = __in6_dev_get(dev);
3484         struct net *net = dev_net(dev);
3485         int run_pending = 0;
3486         int err;
3487
3488         switch (event) {
3489         case NETDEV_REGISTER:
3490                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3491                         idev = ipv6_add_dev(dev);
3492                         if (IS_ERR(idev))
3493                                 return notifier_from_errno(PTR_ERR(idev));
3494                 }
3495                 break;
3496
3497         case NETDEV_CHANGEMTU:
3498                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3499                 if (dev->mtu < IPV6_MIN_MTU) {
3500                         addrconf_ifdown(dev, dev != net->loopback_dev);
3501                         break;
3502                 }
3503
3504                 if (idev) {
3505                         rt6_mtu_change(dev, dev->mtu);
3506                         idev->cnf.mtu6 = dev->mtu;
3507                         break;
3508                 }
3509
3510                 /* allocate new idev */
3511                 idev = ipv6_add_dev(dev);
3512                 if (IS_ERR(idev))
3513                         break;
3514
3515                 /* device is still not ready */
3516                 if (!(idev->if_flags & IF_READY))
3517                         break;
3518
3519                 run_pending = 1;
3520                 fallthrough;
3521         case NETDEV_UP:
3522         case NETDEV_CHANGE:
3523                 if (dev->flags & IFF_SLAVE)
3524                         break;
3525
3526                 if (idev && idev->cnf.disable_ipv6)
3527                         break;
3528
3529                 if (event == NETDEV_UP) {
3530                         /* restore routes for permanent addresses */
3531                         addrconf_permanent_addr(net, dev);
3532
3533                         if (!addrconf_link_ready(dev)) {
3534                                 /* device is not ready yet. */
3535                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3536                                          dev->name);
3537                                 break;
3538                         }
3539
3540                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3541                                 idev = ipv6_add_dev(dev);
3542
3543                         if (!IS_ERR_OR_NULL(idev)) {
3544                                 idev->if_flags |= IF_READY;
3545                                 run_pending = 1;
3546                         }
3547                 } else if (event == NETDEV_CHANGE) {
3548                         if (!addrconf_link_ready(dev)) {
3549                                 /* device is still not ready. */
3550                                 rt6_sync_down_dev(dev, event);
3551                                 break;
3552                         }
3553
3554                         if (!IS_ERR_OR_NULL(idev)) {
3555                                 if (idev->if_flags & IF_READY) {
3556                                         /* device is already configured -
3557                                          * but resend MLD reports, we might
3558                                          * have roamed and need to update
3559                                          * multicast snooping switches
3560                                          */
3561                                         ipv6_mc_up(idev);
3562                                         change_info = ptr;
3563                                         if (change_info->flags_changed & IFF_NOARP)
3564                                                 addrconf_dad_run(idev, true);
3565                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3566                                         break;
3567                                 }
3568                                 idev->if_flags |= IF_READY;
3569                         }
3570
3571                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3572                                 dev->name);
3573
3574                         run_pending = 1;
3575                 }
3576
3577                 switch (dev->type) {
3578 #if IS_ENABLED(CONFIG_IPV6_SIT)
3579                 case ARPHRD_SIT:
3580                         addrconf_sit_config(dev);
3581                         break;
3582 #endif
3583 #if IS_ENABLED(CONFIG_NET_IPGRE)
3584                 case ARPHRD_IPGRE:
3585                         addrconf_gre_config(dev);
3586                         break;
3587 #endif
3588                 case ARPHRD_LOOPBACK:
3589                         init_loopback(dev);
3590                         break;
3591
3592                 default:
3593                         addrconf_dev_config(dev);
3594                         break;
3595                 }
3596
3597                 if (!IS_ERR_OR_NULL(idev)) {
3598                         if (run_pending)
3599                                 addrconf_dad_run(idev, false);
3600
3601                         /* Device has an address by now */
3602                         rt6_sync_up(dev, RTNH_F_DEAD);
3603
3604                         /*
3605                          * If the MTU changed during the interface down,
3606                          * when the interface up, the changed MTU must be
3607                          * reflected in the idev as well as routers.
3608                          */
3609                         if (idev->cnf.mtu6 != dev->mtu &&
3610                             dev->mtu >= IPV6_MIN_MTU) {
3611                                 rt6_mtu_change(dev, dev->mtu);
3612                                 idev->cnf.mtu6 = dev->mtu;
3613                         }
3614                         idev->tstamp = jiffies;
3615                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3616
3617                         /*
3618                          * If the changed mtu during down is lower than
3619                          * IPV6_MIN_MTU stop IPv6 on this interface.
3620                          */
3621                         if (dev->mtu < IPV6_MIN_MTU)
3622                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3623                 }
3624                 break;
3625
3626         case NETDEV_DOWN:
3627         case NETDEV_UNREGISTER:
3628                 /*
3629                  *      Remove all addresses from this interface.
3630                  */
3631                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3632                 break;
3633
3634         case NETDEV_CHANGENAME:
3635                 if (idev) {
3636                         snmp6_unregister_dev(idev);
3637                         addrconf_sysctl_unregister(idev);
3638                         err = addrconf_sysctl_register(idev);
3639                         if (err)
3640                                 return notifier_from_errno(err);
3641                         err = snmp6_register_dev(idev);
3642                         if (err) {
3643                                 addrconf_sysctl_unregister(idev);
3644                                 return notifier_from_errno(err);
3645                         }
3646                 }
3647                 break;
3648
3649         case NETDEV_PRE_TYPE_CHANGE:
3650         case NETDEV_POST_TYPE_CHANGE:
3651                 if (idev)
3652                         addrconf_type_change(dev, event);
3653                 break;
3654
3655         case NETDEV_CHANGEUPPER:
3656                 info = ptr;
3657
3658                 /* flush all routes if dev is linked to or unlinked from
3659                  * an L3 master device (e.g., VRF)
3660                  */
3661                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3662                         addrconf_ifdown(dev, false);
3663         }
3664
3665         return NOTIFY_OK;
3666 }
3667
3668 /*
3669  *      addrconf module should be notified of a device going up
3670  */
3671 static struct notifier_block ipv6_dev_notf = {
3672         .notifier_call = addrconf_notify,
3673         .priority = ADDRCONF_NOTIFY_PRIORITY,
3674 };
3675
3676 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3677 {
3678         struct inet6_dev *idev;
3679         ASSERT_RTNL();
3680
3681         idev = __in6_dev_get(dev);
3682
3683         if (event == NETDEV_POST_TYPE_CHANGE)
3684                 ipv6_mc_remap(idev);
3685         else if (event == NETDEV_PRE_TYPE_CHANGE)
3686                 ipv6_mc_unmap(idev);
3687 }
3688
3689 static bool addr_is_local(const struct in6_addr *addr)
3690 {
3691         return ipv6_addr_type(addr) &
3692                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3693 }
3694
3695 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3696 {
3697         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3698         struct net *net = dev_net(dev);
3699         struct inet6_dev *idev;
3700         struct inet6_ifaddr *ifa, *tmp;
3701         bool keep_addr = false;
3702         int state, i;
3703
3704         ASSERT_RTNL();
3705
3706         rt6_disable_ip(dev, event);
3707
3708         idev = __in6_dev_get(dev);
3709         if (!idev)
3710                 return -ENODEV;
3711
3712         /*
3713          * Step 1: remove reference to ipv6 device from parent device.
3714          *         Do not dev_put!
3715          */
3716         if (unregister) {
3717                 idev->dead = 1;
3718
3719                 /* protected by rtnl_lock */
3720                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3721
3722                 /* Step 1.5: remove snmp6 entry */
3723                 snmp6_unregister_dev(idev);
3724
3725         }
3726
3727         /* combine the user config with event to determine if permanent
3728          * addresses are to be removed from address hash table
3729          */
3730         if (!unregister && !idev->cnf.disable_ipv6) {
3731                 /* aggregate the system setting and interface setting */
3732                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3733
3734                 if (!_keep_addr)
3735                         _keep_addr = idev->cnf.keep_addr_on_down;
3736
3737                 keep_addr = (_keep_addr > 0);
3738         }
3739
3740         /* Step 2: clear hash table */
3741         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3742                 struct hlist_head *h = &inet6_addr_lst[i];
3743
3744                 spin_lock_bh(&addrconf_hash_lock);
3745 restart:
3746                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3747                         if (ifa->idev == idev) {
3748                                 addrconf_del_dad_work(ifa);
3749                                 /* combined flag + permanent flag decide if
3750                                  * address is retained on a down event
3751                                  */
3752                                 if (!keep_addr ||
3753                                     !(ifa->flags & IFA_F_PERMANENT) ||
3754                                     addr_is_local(&ifa->addr)) {
3755                                         hlist_del_init_rcu(&ifa->addr_lst);
3756                                         goto restart;
3757                                 }
3758                         }
3759                 }
3760                 spin_unlock_bh(&addrconf_hash_lock);
3761         }
3762
3763         write_lock_bh(&idev->lock);
3764
3765         addrconf_del_rs_timer(idev);
3766
3767         /* Step 2: clear flags for stateless addrconf */
3768         if (!unregister)
3769                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3770
3771         /* Step 3: clear tempaddr list */
3772         while (!list_empty(&idev->tempaddr_list)) {
3773                 ifa = list_first_entry(&idev->tempaddr_list,
3774                                        struct inet6_ifaddr, tmp_list);
3775                 list_del(&ifa->tmp_list);
3776                 write_unlock_bh(&idev->lock);
3777                 spin_lock_bh(&ifa->lock);
3778
3779                 if (ifa->ifpub) {
3780                         in6_ifa_put(ifa->ifpub);
3781                         ifa->ifpub = NULL;
3782                 }
3783                 spin_unlock_bh(&ifa->lock);
3784                 in6_ifa_put(ifa);
3785                 write_lock_bh(&idev->lock);
3786         }
3787
3788         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3789                 struct fib6_info *rt = NULL;
3790                 bool keep;
3791
3792                 addrconf_del_dad_work(ifa);
3793
3794                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3795                         !addr_is_local(&ifa->addr);
3796
3797                 write_unlock_bh(&idev->lock);
3798                 spin_lock_bh(&ifa->lock);
3799
3800                 if (keep) {
3801                         /* set state to skip the notifier below */
3802                         state = INET6_IFADDR_STATE_DEAD;
3803                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3804                         if (!(ifa->flags & IFA_F_NODAD))
3805                                 ifa->flags |= IFA_F_TENTATIVE;
3806
3807                         rt = ifa->rt;
3808                         ifa->rt = NULL;
3809                 } else {
3810                         state = ifa->state;
3811                         ifa->state = INET6_IFADDR_STATE_DEAD;
3812                 }
3813
3814                 spin_unlock_bh(&ifa->lock);
3815
3816                 if (rt)
3817                         ip6_del_rt(net, rt, false);
3818
3819                 if (state != INET6_IFADDR_STATE_DEAD) {
3820                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3821                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3822                 } else {
3823                         if (idev->cnf.forwarding)
3824                                 addrconf_leave_anycast(ifa);
3825                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3826                 }
3827
3828                 write_lock_bh(&idev->lock);
3829                 if (!keep) {
3830                         list_del_rcu(&ifa->if_list);
3831                         in6_ifa_put(ifa);
3832                 }
3833         }
3834
3835         write_unlock_bh(&idev->lock);
3836
3837         /* Step 5: Discard anycast and multicast list */
3838         if (unregister) {
3839                 ipv6_ac_destroy_dev(idev);
3840                 ipv6_mc_destroy_dev(idev);
3841         } else {
3842                 ipv6_mc_down(idev);
3843         }
3844
3845         idev->tstamp = jiffies;
3846
3847         /* Last: Shot the device (if unregistered) */
3848         if (unregister) {
3849                 addrconf_sysctl_unregister(idev);
3850                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3851                 neigh_ifdown(&nd_tbl, dev);
3852                 in6_dev_put(idev);
3853         }
3854         return 0;
3855 }
3856
3857 static void addrconf_rs_timer(struct timer_list *t)
3858 {
3859         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3860         struct net_device *dev = idev->dev;
3861         struct in6_addr lladdr;
3862
3863         write_lock(&idev->lock);
3864         if (idev->dead || !(idev->if_flags & IF_READY))
3865                 goto out;
3866
3867         if (!ipv6_accept_ra(idev))
3868                 goto out;
3869
3870         /* Announcement received after solicitation was sent */
3871         if (idev->if_flags & IF_RA_RCVD)
3872                 goto out;
3873
3874         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3875                 write_unlock(&idev->lock);
3876                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3877                         ndisc_send_rs(dev, &lladdr,
3878                                       &in6addr_linklocal_allrouters);
3879                 else
3880                         goto put;
3881
3882                 write_lock(&idev->lock);
3883                 idev->rs_interval = rfc3315_s14_backoff_update(
3884                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3885                 /* The wait after the last probe can be shorter */
3886                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3887                                              idev->cnf.rtr_solicits) ?
3888                                       idev->cnf.rtr_solicit_delay :
3889                                       idev->rs_interval);
3890         } else {
3891                 /*
3892                  * Note: we do not support deprecated "all on-link"
3893                  * assumption any longer.
3894                  */
3895                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3896         }
3897
3898 out:
3899         write_unlock(&idev->lock);
3900 put:
3901         in6_dev_put(idev);
3902 }
3903
3904 /*
3905  *      Duplicate Address Detection
3906  */
3907 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3908 {
3909         unsigned long rand_num;
3910         struct inet6_dev *idev = ifp->idev;
3911         u64 nonce;
3912
3913         if (ifp->flags & IFA_F_OPTIMISTIC)
3914                 rand_num = 0;
3915         else
3916                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3917
3918         nonce = 0;
3919         if (idev->cnf.enhanced_dad ||
3920             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3921                 do
3922                         get_random_bytes(&nonce, 6);
3923                 while (nonce == 0);
3924         }
3925         ifp->dad_nonce = nonce;
3926         ifp->dad_probes = idev->cnf.dad_transmits;
3927         addrconf_mod_dad_work(ifp, rand_num);
3928 }
3929
3930 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3931 {
3932         struct inet6_dev *idev = ifp->idev;
3933         struct net_device *dev = idev->dev;
3934         bool bump_id, notify = false;
3935         struct net *net;
3936
3937         addrconf_join_solict(dev, &ifp->addr);
3938
3939         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3940
3941         read_lock_bh(&idev->lock);
3942         spin_lock(&ifp->lock);
3943         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3944                 goto out;
3945
3946         net = dev_net(dev);
3947         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3948             (net->ipv6.devconf_all->accept_dad < 1 &&
3949              idev->cnf.accept_dad < 1) ||
3950             !(ifp->flags&IFA_F_TENTATIVE) ||
3951             ifp->flags & IFA_F_NODAD) {
3952                 bool send_na = false;
3953
3954                 if (ifp->flags & IFA_F_TENTATIVE &&
3955                     !(ifp->flags & IFA_F_OPTIMISTIC))
3956                         send_na = true;
3957                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3958                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3959                 spin_unlock(&ifp->lock);
3960                 read_unlock_bh(&idev->lock);
3961
3962                 addrconf_dad_completed(ifp, bump_id, send_na);
3963                 return;
3964         }
3965
3966         if (!(idev->if_flags & IF_READY)) {
3967                 spin_unlock(&ifp->lock);
3968                 read_unlock_bh(&idev->lock);
3969                 /*
3970                  * If the device is not ready:
3971                  * - keep it tentative if it is a permanent address.
3972                  * - otherwise, kill it.
3973                  */
3974                 in6_ifa_hold(ifp);
3975                 addrconf_dad_stop(ifp, 0);
3976                 return;
3977         }
3978
3979         /*
3980          * Optimistic nodes can start receiving
3981          * Frames right away
3982          */
3983         if (ifp->flags & IFA_F_OPTIMISTIC) {
3984                 ip6_ins_rt(net, ifp->rt);
3985                 if (ipv6_use_optimistic_addr(net, idev)) {
3986                         /* Because optimistic nodes can use this address,
3987                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3988                          */
3989                         notify = true;
3990                 }
3991         }
3992
3993         addrconf_dad_kick(ifp);
3994 out:
3995         spin_unlock(&ifp->lock);
3996         read_unlock_bh(&idev->lock);
3997         if (notify)
3998                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3999 }
4000
4001 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4002 {
4003         bool begin_dad = false;
4004
4005         spin_lock_bh(&ifp->lock);
4006         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4007                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4008                 begin_dad = true;
4009         }
4010         spin_unlock_bh(&ifp->lock);
4011
4012         if (begin_dad)
4013                 addrconf_mod_dad_work(ifp, 0);
4014 }
4015
4016 static void addrconf_dad_work(struct work_struct *w)
4017 {
4018         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4019                                                 struct inet6_ifaddr,
4020                                                 dad_work);
4021         struct inet6_dev *idev = ifp->idev;
4022         bool bump_id, disable_ipv6 = false;
4023         struct in6_addr mcaddr;
4024
4025         enum {
4026                 DAD_PROCESS,
4027                 DAD_BEGIN,
4028                 DAD_ABORT,
4029         } action = DAD_PROCESS;
4030
4031         rtnl_lock();
4032
4033         spin_lock_bh(&ifp->lock);
4034         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4035                 action = DAD_BEGIN;
4036                 ifp->state = INET6_IFADDR_STATE_DAD;
4037         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4038                 action = DAD_ABORT;
4039                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4040
4041                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4042                      idev->cnf.accept_dad > 1) &&
4043                     !idev->cnf.disable_ipv6 &&
4044                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4045                         struct in6_addr addr;
4046
4047                         addr.s6_addr32[0] = htonl(0xfe800000);
4048                         addr.s6_addr32[1] = 0;
4049
4050                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4051                             ipv6_addr_equal(&ifp->addr, &addr)) {
4052                                 /* DAD failed for link-local based on MAC */
4053                                 idev->cnf.disable_ipv6 = 1;
4054
4055                                 pr_info("%s: IPv6 being disabled!\n",
4056                                         ifp->idev->dev->name);
4057                                 disable_ipv6 = true;
4058                         }
4059                 }
4060         }
4061         spin_unlock_bh(&ifp->lock);
4062
4063         if (action == DAD_BEGIN) {
4064                 addrconf_dad_begin(ifp);
4065                 goto out;
4066         } else if (action == DAD_ABORT) {
4067                 in6_ifa_hold(ifp);
4068                 addrconf_dad_stop(ifp, 1);
4069                 if (disable_ipv6)
4070                         addrconf_ifdown(idev->dev, false);
4071                 goto out;
4072         }
4073
4074         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4075                 goto out;
4076
4077         write_lock_bh(&idev->lock);
4078         if (idev->dead || !(idev->if_flags & IF_READY)) {
4079                 write_unlock_bh(&idev->lock);
4080                 goto out;
4081         }
4082
4083         spin_lock(&ifp->lock);
4084         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4085                 spin_unlock(&ifp->lock);
4086                 write_unlock_bh(&idev->lock);
4087                 goto out;
4088         }
4089
4090         if (ifp->dad_probes == 0) {
4091                 bool send_na = false;
4092
4093                 /*
4094                  * DAD was successful
4095                  */
4096
4097                 if (ifp->flags & IFA_F_TENTATIVE &&
4098                     !(ifp->flags & IFA_F_OPTIMISTIC))
4099                         send_na = true;
4100                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4101                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4102                 spin_unlock(&ifp->lock);
4103                 write_unlock_bh(&idev->lock);
4104
4105                 addrconf_dad_completed(ifp, bump_id, send_na);
4106
4107                 goto out;
4108         }
4109
4110         ifp->dad_probes--;
4111         addrconf_mod_dad_work(ifp,
4112                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4113                                   HZ/100));
4114         spin_unlock(&ifp->lock);
4115         write_unlock_bh(&idev->lock);
4116
4117         /* send a neighbour solicitation for our addr */
4118         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4119         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4120                       ifp->dad_nonce);
4121 out:
4122         in6_ifa_put(ifp);
4123         rtnl_unlock();
4124 }
4125
4126 /* ifp->idev must be at least read locked */
4127 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4128 {
4129         struct inet6_ifaddr *ifpiter;
4130         struct inet6_dev *idev = ifp->idev;
4131
4132         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4133                 if (ifpiter->scope > IFA_LINK)
4134                         break;
4135                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4136                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4137                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4138                     IFA_F_PERMANENT)
4139                         return false;
4140         }
4141         return true;
4142 }
4143
4144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4145                                    bool send_na)
4146 {
4147         struct net_device *dev = ifp->idev->dev;
4148         struct in6_addr lladdr;
4149         bool send_rs, send_mld;
4150
4151         addrconf_del_dad_work(ifp);
4152
4153         /*
4154          *      Configure the address for reception. Now it is valid.
4155          */
4156
4157         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4158
4159         /* If added prefix is link local and we are prepared to process
4160            router advertisements, start sending router solicitations.
4161          */
4162
4163         read_lock_bh(&ifp->idev->lock);
4164         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4165         send_rs = send_mld &&
4166                   ipv6_accept_ra(ifp->idev) &&
4167                   ifp->idev->cnf.rtr_solicits != 0 &&
4168                   (dev->flags&IFF_LOOPBACK) == 0;
4169         read_unlock_bh(&ifp->idev->lock);
4170
4171         /* While dad is in progress mld report's source address is in6_addrany.
4172          * Resend with proper ll now.
4173          */
4174         if (send_mld)
4175                 ipv6_mc_dad_complete(ifp->idev);
4176
4177         /* send unsolicited NA if enabled */
4178         if (send_na &&
4179             (ifp->idev->cnf.ndisc_notify ||
4180              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4181                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4182                               /*router=*/ !!ifp->idev->cnf.forwarding,
4183                               /*solicited=*/ false, /*override=*/ true,
4184                               /*inc_opt=*/ true);
4185         }
4186
4187         if (send_rs) {
4188                 /*
4189                  *      If a host as already performed a random delay
4190                  *      [...] as part of DAD [...] there is no need
4191                  *      to delay again before sending the first RS
4192                  */
4193                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4194                         return;
4195                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4196
4197                 write_lock_bh(&ifp->idev->lock);
4198                 spin_lock(&ifp->lock);
4199                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4200                         ifp->idev->cnf.rtr_solicit_interval);
4201                 ifp->idev->rs_probes = 1;
4202                 ifp->idev->if_flags |= IF_RS_SENT;
4203                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4204                 spin_unlock(&ifp->lock);
4205                 write_unlock_bh(&ifp->idev->lock);
4206         }
4207
4208         if (bump_id)
4209                 rt_genid_bump_ipv6(dev_net(dev));
4210
4211         /* Make sure that a new temporary address will be created
4212          * before this temporary address becomes deprecated.
4213          */
4214         if (ifp->flags & IFA_F_TEMPORARY)
4215                 addrconf_verify_rtnl();
4216 }
4217
4218 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4219 {
4220         struct inet6_ifaddr *ifp;
4221
4222         read_lock_bh(&idev->lock);
4223         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4224                 spin_lock(&ifp->lock);
4225                 if ((ifp->flags & IFA_F_TENTATIVE &&
4226                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4227                         if (restart)
4228                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4229                         addrconf_dad_kick(ifp);
4230                 }
4231                 spin_unlock(&ifp->lock);
4232         }
4233         read_unlock_bh(&idev->lock);
4234 }
4235
4236 #ifdef CONFIG_PROC_FS
4237 struct if6_iter_state {
4238         struct seq_net_private p;
4239         int bucket;
4240         int offset;
4241 };
4242
4243 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4244 {
4245         struct if6_iter_state *state = seq->private;
4246         struct net *net = seq_file_net(seq);
4247         struct inet6_ifaddr *ifa = NULL;
4248         int p = 0;
4249
4250         /* initial bucket if pos is 0 */
4251         if (pos == 0) {
4252                 state->bucket = 0;
4253                 state->offset = 0;
4254         }
4255
4256         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4257                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4258                                          addr_lst) {
4259                         if (!net_eq(dev_net(ifa->idev->dev), net))
4260                                 continue;
4261                         /* sync with offset */
4262                         if (p < state->offset) {
4263                                 p++;
4264                                 continue;
4265                         }
4266                         return ifa;
4267                 }
4268
4269                 /* prepare for next bucket */
4270                 state->offset = 0;
4271                 p = 0;
4272         }
4273         return NULL;
4274 }
4275
4276 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4277                                          struct inet6_ifaddr *ifa)
4278 {
4279         struct if6_iter_state *state = seq->private;
4280         struct net *net = seq_file_net(seq);
4281
4282         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4283                 if (!net_eq(dev_net(ifa->idev->dev), net))
4284                         continue;
4285                 state->offset++;
4286                 return ifa;
4287         }
4288
4289         state->offset = 0;
4290         while (++state->bucket < IN6_ADDR_HSIZE) {
4291                 hlist_for_each_entry_rcu(ifa,
4292                                      &inet6_addr_lst[state->bucket], addr_lst) {
4293                         if (!net_eq(dev_net(ifa->idev->dev), net))
4294                                 continue;
4295                         return ifa;
4296                 }
4297         }
4298
4299         return NULL;
4300 }
4301
4302 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4303         __acquires(rcu)
4304 {
4305         rcu_read_lock();
4306         return if6_get_first(seq, *pos);
4307 }
4308
4309 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4310 {
4311         struct inet6_ifaddr *ifa;
4312
4313         ifa = if6_get_next(seq, v);
4314         ++*pos;
4315         return ifa;
4316 }
4317
4318 static void if6_seq_stop(struct seq_file *seq, void *v)
4319         __releases(rcu)
4320 {
4321         rcu_read_unlock();
4322 }
4323
4324 static int if6_seq_show(struct seq_file *seq, void *v)
4325 {
4326         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4327         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4328                    &ifp->addr,
4329                    ifp->idev->dev->ifindex,
4330                    ifp->prefix_len,
4331                    ifp->scope,
4332                    (u8) ifp->flags,
4333                    ifp->idev->dev->name);
4334         return 0;
4335 }
4336
4337 static const struct seq_operations if6_seq_ops = {
4338         .start  = if6_seq_start,
4339         .next   = if6_seq_next,
4340         .show   = if6_seq_show,
4341         .stop   = if6_seq_stop,
4342 };
4343
4344 static int __net_init if6_proc_net_init(struct net *net)
4345 {
4346         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4347                         sizeof(struct if6_iter_state)))
4348                 return -ENOMEM;
4349         return 0;
4350 }
4351
4352 static void __net_exit if6_proc_net_exit(struct net *net)
4353 {
4354         remove_proc_entry("if_inet6", net->proc_net);
4355 }
4356
4357 static struct pernet_operations if6_proc_net_ops = {
4358         .init = if6_proc_net_init,
4359         .exit = if6_proc_net_exit,
4360 };
4361
4362 int __init if6_proc_init(void)
4363 {
4364         return register_pernet_subsys(&if6_proc_net_ops);
4365 }
4366
4367 void if6_proc_exit(void)
4368 {
4369         unregister_pernet_subsys(&if6_proc_net_ops);
4370 }
4371 #endif  /* CONFIG_PROC_FS */
4372
4373 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4374 /* Check if address is a home address configured on any interface. */
4375 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4376 {
4377         unsigned int hash = inet6_addr_hash(net, addr);
4378         struct inet6_ifaddr *ifp = NULL;
4379         int ret = 0;
4380
4381         rcu_read_lock();
4382         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4383                 if (!net_eq(dev_net(ifp->idev->dev), net))
4384                         continue;
4385                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4386                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4387                         ret = 1;
4388                         break;
4389                 }
4390         }
4391         rcu_read_unlock();
4392         return ret;
4393 }
4394 #endif
4395
4396 /* RFC6554 has some algorithm to avoid loops in segment routing by
4397  * checking if the segments contains any of a local interface address.
4398  *
4399  * Quote:
4400  *
4401  * To detect loops in the SRH, a router MUST determine if the SRH
4402  * includes multiple addresses assigned to any interface on that router.
4403  * If such addresses appear more than once and are separated by at least
4404  * one address not assigned to that router.
4405  */
4406 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4407                           unsigned char nsegs)
4408 {
4409         const struct in6_addr *addr;
4410         int i, ret = 0, found = 0;
4411         struct inet6_ifaddr *ifp;
4412         bool separated = false;
4413         unsigned int hash;
4414         bool hash_found;
4415
4416         rcu_read_lock();
4417         for (i = 0; i < nsegs; i++) {
4418                 addr = &segs[i];
4419                 hash = inet6_addr_hash(net, addr);
4420
4421                 hash_found = false;
4422                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4423                         if (!net_eq(dev_net(ifp->idev->dev), net))
4424                                 continue;
4425
4426                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4427                                 hash_found = true;
4428                                 break;
4429                         }
4430                 }
4431
4432                 if (hash_found) {
4433                         if (found > 1 && separated) {
4434                                 ret = 1;
4435                                 break;
4436                         }
4437
4438                         separated = false;
4439                         found++;
4440                 } else {
4441                         separated = true;
4442                 }
4443         }
4444         rcu_read_unlock();
4445
4446         return ret;
4447 }
4448
4449 /*
4450  *      Periodic address status verification
4451  */
4452
4453 static void addrconf_verify_rtnl(void)
4454 {
4455         unsigned long now, next, next_sec, next_sched;
4456         struct inet6_ifaddr *ifp;
4457         int i;
4458
4459         ASSERT_RTNL();
4460
4461         rcu_read_lock_bh();
4462         now = jiffies;
4463         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4464
4465         cancel_delayed_work(&addr_chk_work);
4466
4467         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4468 restart:
4469                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4470                         unsigned long age;
4471
4472                         /* When setting preferred_lft to a value not zero or
4473                          * infinity, while valid_lft is infinity
4474                          * IFA_F_PERMANENT has a non-infinity life time.
4475                          */
4476                         if ((ifp->flags & IFA_F_PERMANENT) &&
4477                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4478                                 continue;
4479
4480                         spin_lock(&ifp->lock);
4481                         /* We try to batch several events at once. */
4482                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4483
4484                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4485                             age >= ifp->valid_lft) {
4486                                 spin_unlock(&ifp->lock);
4487                                 in6_ifa_hold(ifp);
4488                                 ipv6_del_addr(ifp);
4489                                 goto restart;
4490                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4491                                 spin_unlock(&ifp->lock);
4492                                 continue;
4493                         } else if (age >= ifp->prefered_lft) {
4494                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4495                                 int deprecate = 0;
4496
4497                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4498                                         deprecate = 1;
4499                                         ifp->flags |= IFA_F_DEPRECATED;
4500                                 }
4501
4502                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4503                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4504                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4505
4506                                 spin_unlock(&ifp->lock);
4507
4508                                 if (deprecate) {
4509                                         in6_ifa_hold(ifp);
4510
4511                                         ipv6_ifa_notify(0, ifp);
4512                                         in6_ifa_put(ifp);
4513                                         goto restart;
4514                                 }
4515                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4516                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4517                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4518                                         ifp->idev->cnf.dad_transmits *
4519                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4520
4521                                 if (age >= ifp->prefered_lft - regen_advance) {
4522                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4523                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4524                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4525                                         if (!ifp->regen_count && ifpub) {
4526                                                 ifp->regen_count++;
4527                                                 in6_ifa_hold(ifp);
4528                                                 in6_ifa_hold(ifpub);
4529                                                 spin_unlock(&ifp->lock);
4530
4531                                                 spin_lock(&ifpub->lock);
4532                                                 ifpub->regen_count = 0;
4533                                                 spin_unlock(&ifpub->lock);
4534                                                 rcu_read_unlock_bh();
4535                                                 ipv6_create_tempaddr(ifpub, true);
4536                                                 in6_ifa_put(ifpub);
4537                                                 in6_ifa_put(ifp);
4538                                                 rcu_read_lock_bh();
4539                                                 goto restart;
4540                                         }
4541                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4542                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4543                                 spin_unlock(&ifp->lock);
4544                         } else {
4545                                 /* ifp->prefered_lft <= ifp->valid_lft */
4546                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4547                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4548                                 spin_unlock(&ifp->lock);
4549                         }
4550                 }
4551         }
4552
4553         next_sec = round_jiffies_up(next);
4554         next_sched = next;
4555
4556         /* If rounded timeout is accurate enough, accept it. */
4557         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4558                 next_sched = next_sec;
4559
4560         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4561         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4562                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4563
4564         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4565                  now, next, next_sec, next_sched);
4566         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4567         rcu_read_unlock_bh();
4568 }
4569
4570 static void addrconf_verify_work(struct work_struct *w)
4571 {
4572         rtnl_lock();
4573         addrconf_verify_rtnl();
4574         rtnl_unlock();
4575 }
4576
4577 static void addrconf_verify(void)
4578 {
4579         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4580 }
4581
4582 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4583                                      struct in6_addr **peer_pfx)
4584 {
4585         struct in6_addr *pfx = NULL;
4586
4587         *peer_pfx = NULL;
4588
4589         if (addr)
4590                 pfx = nla_data(addr);
4591
4592         if (local) {
4593                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4594                         *peer_pfx = pfx;
4595                 pfx = nla_data(local);
4596         }
4597
4598         return pfx;
4599 }
4600
4601 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4602         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4603         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4604         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4605         [IFA_FLAGS]             = { .len = sizeof(u32) },
4606         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4607         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4608 };
4609
4610 static int
4611 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4612                   struct netlink_ext_ack *extack)
4613 {
4614         struct net *net = sock_net(skb->sk);
4615         struct ifaddrmsg *ifm;
4616         struct nlattr *tb[IFA_MAX+1];
4617         struct in6_addr *pfx, *peer_pfx;
4618         u32 ifa_flags;
4619         int err;
4620
4621         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4622                                      ifa_ipv6_policy, extack);
4623         if (err < 0)
4624                 return err;
4625
4626         ifm = nlmsg_data(nlh);
4627         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4628         if (!pfx)
4629                 return -EINVAL;
4630
4631         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4632
4633         /* We ignore other flags so far. */
4634         ifa_flags &= IFA_F_MANAGETEMPADDR;
4635
4636         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4637                               ifm->ifa_prefixlen);
4638 }
4639
4640 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4641                                unsigned long expires, u32 flags,
4642                                bool modify_peer)
4643 {
4644         struct fib6_info *f6i;
4645         u32 prio;
4646
4647         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4648                                         ifp->prefix_len,
4649                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4650         if (!f6i)
4651                 return -ENOENT;
4652
4653         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4654         if (f6i->fib6_metric != prio) {
4655                 /* delete old one */
4656                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4657
4658                 /* add new one */
4659                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4660                                       ifp->prefix_len,
4661                                       ifp->rt_priority, ifp->idev->dev,
4662                                       expires, flags, GFP_KERNEL);
4663         } else {
4664                 if (!expires)
4665                         fib6_clean_expires(f6i);
4666                 else
4667                         fib6_set_expires(f6i, expires);
4668
4669                 fib6_info_release(f6i);
4670         }
4671
4672         return 0;
4673 }
4674
4675 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4676 {
4677         u32 flags;
4678         clock_t expires;
4679         unsigned long timeout;
4680         bool was_managetempaddr;
4681         bool had_prefixroute;
4682         bool new_peer = false;
4683
4684         ASSERT_RTNL();
4685
4686         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4687                 return -EINVAL;
4688
4689         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4690             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4691                 return -EINVAL;
4692
4693         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4694                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4695
4696         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4697         if (addrconf_finite_timeout(timeout)) {
4698                 expires = jiffies_to_clock_t(timeout * HZ);
4699                 cfg->valid_lft = timeout;
4700                 flags = RTF_EXPIRES;
4701         } else {
4702                 expires = 0;
4703                 flags = 0;
4704                 cfg->ifa_flags |= IFA_F_PERMANENT;
4705         }
4706
4707         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4708         if (addrconf_finite_timeout(timeout)) {
4709                 if (timeout == 0)
4710                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4711                 cfg->preferred_lft = timeout;
4712         }
4713
4714         if (cfg->peer_pfx &&
4715             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4716                 if (!ipv6_addr_any(&ifp->peer_addr))
4717                         cleanup_prefix_route(ifp, expires, true, true);
4718                 new_peer = true;
4719         }
4720
4721         spin_lock_bh(&ifp->lock);
4722         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4723         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4724                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4725         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4726                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4727                         IFA_F_NOPREFIXROUTE);
4728         ifp->flags |= cfg->ifa_flags;
4729         ifp->tstamp = jiffies;
4730         ifp->valid_lft = cfg->valid_lft;
4731         ifp->prefered_lft = cfg->preferred_lft;
4732
4733         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4734                 ifp->rt_priority = cfg->rt_priority;
4735
4736         if (new_peer)
4737                 ifp->peer_addr = *cfg->peer_pfx;
4738
4739         spin_unlock_bh(&ifp->lock);
4740         if (!(ifp->flags&IFA_F_TENTATIVE))
4741                 ipv6_ifa_notify(0, ifp);
4742
4743         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4744                 int rc = -ENOENT;
4745
4746                 if (had_prefixroute)
4747                         rc = modify_prefix_route(ifp, expires, flags, false);
4748
4749                 /* prefix route could have been deleted; if so restore it */
4750                 if (rc == -ENOENT) {
4751                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4752                                               ifp->rt_priority, ifp->idev->dev,
4753                                               expires, flags, GFP_KERNEL);
4754                 }
4755
4756                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4757                         rc = modify_prefix_route(ifp, expires, flags, true);
4758
4759                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4760                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4761                                               ifp->rt_priority, ifp->idev->dev,
4762                                               expires, flags, GFP_KERNEL);
4763                 }
4764         } else if (had_prefixroute) {
4765                 enum cleanup_prefix_rt_t action;
4766                 unsigned long rt_expires;
4767
4768                 write_lock_bh(&ifp->idev->lock);
4769                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4770                 write_unlock_bh(&ifp->idev->lock);
4771
4772                 if (action != CLEANUP_PREFIX_RT_NOP) {
4773                         cleanup_prefix_route(ifp, rt_expires,
4774                                 action == CLEANUP_PREFIX_RT_DEL, false);
4775                 }
4776         }
4777
4778         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4779                 if (was_managetempaddr &&
4780                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4781                         cfg->valid_lft = 0;
4782                         cfg->preferred_lft = 0;
4783                 }
4784                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4785                                  cfg->preferred_lft, !was_managetempaddr,
4786                                  jiffies);
4787         }
4788
4789         addrconf_verify_rtnl();
4790
4791         return 0;
4792 }
4793
4794 static int
4795 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4796                   struct netlink_ext_ack *extack)
4797 {
4798         struct net *net = sock_net(skb->sk);
4799         struct ifaddrmsg *ifm;
4800         struct nlattr *tb[IFA_MAX+1];
4801         struct in6_addr *peer_pfx;
4802         struct inet6_ifaddr *ifa;
4803         struct net_device *dev;
4804         struct inet6_dev *idev;
4805         struct ifa6_config cfg;
4806         int err;
4807
4808         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4809                                      ifa_ipv6_policy, extack);
4810         if (err < 0)
4811                 return err;
4812
4813         memset(&cfg, 0, sizeof(cfg));
4814
4815         ifm = nlmsg_data(nlh);
4816         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4817         if (!cfg.pfx)
4818                 return -EINVAL;
4819
4820         cfg.peer_pfx = peer_pfx;
4821         cfg.plen = ifm->ifa_prefixlen;
4822         if (tb[IFA_RT_PRIORITY])
4823                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4824
4825         cfg.valid_lft = INFINITY_LIFE_TIME;
4826         cfg.preferred_lft = INFINITY_LIFE_TIME;
4827
4828         if (tb[IFA_CACHEINFO]) {
4829                 struct ifa_cacheinfo *ci;
4830
4831                 ci = nla_data(tb[IFA_CACHEINFO]);
4832                 cfg.valid_lft = ci->ifa_valid;
4833                 cfg.preferred_lft = ci->ifa_prefered;
4834         }
4835
4836         dev =  __dev_get_by_index(net, ifm->ifa_index);
4837         if (!dev)
4838                 return -ENODEV;
4839
4840         if (tb[IFA_FLAGS])
4841                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4842         else
4843                 cfg.ifa_flags = ifm->ifa_flags;
4844
4845         /* We ignore other flags so far. */
4846         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4847                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4848                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4849
4850         idev = ipv6_find_idev(dev);
4851         if (IS_ERR(idev))
4852                 return PTR_ERR(idev);
4853
4854         if (!ipv6_allow_optimistic_dad(net, idev))
4855                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4856
4857         if (cfg.ifa_flags & IFA_F_NODAD &&
4858             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4859                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4860                 return -EINVAL;
4861         }
4862
4863         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4864         if (!ifa) {
4865                 /*
4866                  * It would be best to check for !NLM_F_CREATE here but
4867                  * userspace already relies on not having to provide this.
4868                  */
4869                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4870         }
4871
4872         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4873             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4874                 err = -EEXIST;
4875         else
4876                 err = inet6_addr_modify(ifa, &cfg);
4877
4878         in6_ifa_put(ifa);
4879
4880         return err;
4881 }
4882
4883 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4884                           u8 scope, int ifindex)
4885 {
4886         struct ifaddrmsg *ifm;
4887
4888         ifm = nlmsg_data(nlh);
4889         ifm->ifa_family = AF_INET6;
4890         ifm->ifa_prefixlen = prefixlen;
4891         ifm->ifa_flags = flags;
4892         ifm->ifa_scope = scope;
4893         ifm->ifa_index = ifindex;
4894 }
4895
4896 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4897                          unsigned long tstamp, u32 preferred, u32 valid)
4898 {
4899         struct ifa_cacheinfo ci;
4900
4901         ci.cstamp = cstamp_delta(cstamp);
4902         ci.tstamp = cstamp_delta(tstamp);
4903         ci.ifa_prefered = preferred;
4904         ci.ifa_valid = valid;
4905
4906         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4907 }
4908
4909 static inline int rt_scope(int ifa_scope)
4910 {
4911         if (ifa_scope & IFA_HOST)
4912                 return RT_SCOPE_HOST;
4913         else if (ifa_scope & IFA_LINK)
4914                 return RT_SCOPE_LINK;
4915         else if (ifa_scope & IFA_SITE)
4916                 return RT_SCOPE_SITE;
4917         else
4918                 return RT_SCOPE_UNIVERSE;
4919 }
4920
4921 static inline int inet6_ifaddr_msgsize(void)
4922 {
4923         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4924                + nla_total_size(16) /* IFA_LOCAL */
4925                + nla_total_size(16) /* IFA_ADDRESS */
4926                + nla_total_size(sizeof(struct ifa_cacheinfo))
4927                + nla_total_size(4)  /* IFA_FLAGS */
4928                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4929 }
4930
4931 enum addr_type_t {
4932         UNICAST_ADDR,
4933         MULTICAST_ADDR,
4934         ANYCAST_ADDR,
4935 };
4936
4937 struct inet6_fill_args {
4938         u32 portid;
4939         u32 seq;
4940         int event;
4941         unsigned int flags;
4942         int netnsid;
4943         int ifindex;
4944         enum addr_type_t type;
4945 };
4946
4947 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4948                              struct inet6_fill_args *args)
4949 {
4950         struct nlmsghdr  *nlh;
4951         u32 preferred, valid;
4952
4953         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4954                         sizeof(struct ifaddrmsg), args->flags);
4955         if (!nlh)
4956                 return -EMSGSIZE;
4957
4958         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4959                       ifa->idev->dev->ifindex);
4960
4961         if (args->netnsid >= 0 &&
4962             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4963                 goto error;
4964
4965         if (!((ifa->flags&IFA_F_PERMANENT) &&
4966               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4967                 preferred = ifa->prefered_lft;
4968                 valid = ifa->valid_lft;
4969                 if (preferred != INFINITY_LIFE_TIME) {
4970                         long tval = (jiffies - ifa->tstamp)/HZ;
4971                         if (preferred > tval)
4972                                 preferred -= tval;
4973                         else
4974                                 preferred = 0;
4975                         if (valid != INFINITY_LIFE_TIME) {
4976                                 if (valid > tval)
4977                                         valid -= tval;
4978                                 else
4979                                         valid = 0;
4980                         }
4981                 }
4982         } else {
4983                 preferred = INFINITY_LIFE_TIME;
4984                 valid = INFINITY_LIFE_TIME;
4985         }
4986
4987         if (!ipv6_addr_any(&ifa->peer_addr)) {
4988                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4989                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4990                         goto error;
4991         } else
4992                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4993                         goto error;
4994
4995         if (ifa->rt_priority &&
4996             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4997                 goto error;
4998
4999         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5000                 goto error;
5001
5002         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5003                 goto error;
5004
5005         nlmsg_end(skb, nlh);
5006         return 0;
5007
5008 error:
5009         nlmsg_cancel(skb, nlh);
5010         return -EMSGSIZE;
5011 }
5012
5013 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5014                                struct inet6_fill_args *args)
5015 {
5016         struct nlmsghdr  *nlh;
5017         u8 scope = RT_SCOPE_UNIVERSE;
5018         int ifindex = ifmca->idev->dev->ifindex;
5019
5020         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5021                 scope = RT_SCOPE_SITE;
5022
5023         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5024                         sizeof(struct ifaddrmsg), args->flags);
5025         if (!nlh)
5026                 return -EMSGSIZE;
5027
5028         if (args->netnsid >= 0 &&
5029             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5030                 nlmsg_cancel(skb, nlh);
5031                 return -EMSGSIZE;
5032         }
5033
5034         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5035         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5036             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5037                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5038                 nlmsg_cancel(skb, nlh);
5039                 return -EMSGSIZE;
5040         }
5041
5042         nlmsg_end(skb, nlh);
5043         return 0;
5044 }
5045
5046 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5047                                struct inet6_fill_args *args)
5048 {
5049         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5050         int ifindex = dev ? dev->ifindex : 1;
5051         struct nlmsghdr  *nlh;
5052         u8 scope = RT_SCOPE_UNIVERSE;
5053
5054         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5055                 scope = RT_SCOPE_SITE;
5056
5057         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5058                         sizeof(struct ifaddrmsg), args->flags);
5059         if (!nlh)
5060                 return -EMSGSIZE;
5061
5062         if (args->netnsid >= 0 &&
5063             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5064                 nlmsg_cancel(skb, nlh);
5065                 return -EMSGSIZE;
5066         }
5067
5068         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5069         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5070             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5071                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5072                 nlmsg_cancel(skb, nlh);
5073                 return -EMSGSIZE;
5074         }
5075
5076         nlmsg_end(skb, nlh);
5077         return 0;
5078 }
5079
5080 /* called with rcu_read_lock() */
5081 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5082                           struct netlink_callback *cb, int s_ip_idx,
5083                           struct inet6_fill_args *fillargs)
5084 {
5085         struct ifmcaddr6 *ifmca;
5086         struct ifacaddr6 *ifaca;
5087         int ip_idx = 0;
5088         int err = 1;
5089
5090         read_lock_bh(&idev->lock);
5091         switch (fillargs->type) {
5092         case UNICAST_ADDR: {
5093                 struct inet6_ifaddr *ifa;
5094                 fillargs->event = RTM_NEWADDR;
5095
5096                 /* unicast address incl. temp addr */
5097                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5098                         if (ip_idx < s_ip_idx)
5099                                 goto next;
5100                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5101                         if (err < 0)
5102                                 break;
5103                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5104 next:
5105                         ip_idx++;
5106                 }
5107                 break;
5108         }
5109         case MULTICAST_ADDR:
5110                 read_unlock_bh(&idev->lock);
5111                 fillargs->event = RTM_GETMULTICAST;
5112
5113                 /* multicast address */
5114                 for (ifmca = rcu_dereference(idev->mc_list);
5115                      ifmca;
5116                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5117                         if (ip_idx < s_ip_idx)
5118                                 continue;
5119                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5120                         if (err < 0)
5121                                 break;
5122                 }
5123                 read_lock_bh(&idev->lock);
5124                 break;
5125         case ANYCAST_ADDR:
5126                 fillargs->event = RTM_GETANYCAST;
5127                 /* anycast address */
5128                 for (ifaca = idev->ac_list; ifaca;
5129                      ifaca = ifaca->aca_next, ip_idx++) {
5130                         if (ip_idx < s_ip_idx)
5131                                 continue;
5132                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5133                         if (err < 0)
5134                                 break;
5135                 }
5136                 break;
5137         default:
5138                 break;
5139         }
5140         read_unlock_bh(&idev->lock);
5141         cb->args[2] = ip_idx;
5142         return err;
5143 }
5144
5145 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5146                                        struct inet6_fill_args *fillargs,
5147                                        struct net **tgt_net, struct sock *sk,
5148                                        struct netlink_callback *cb)
5149 {
5150         struct netlink_ext_ack *extack = cb->extack;
5151         struct nlattr *tb[IFA_MAX+1];
5152         struct ifaddrmsg *ifm;
5153         int err, i;
5154
5155         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5156                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5157                 return -EINVAL;
5158         }
5159
5160         ifm = nlmsg_data(nlh);
5161         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5162                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5163                 return -EINVAL;
5164         }
5165
5166         fillargs->ifindex = ifm->ifa_index;
5167         if (fillargs->ifindex) {
5168                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5169                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5170         }
5171
5172         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5173                                             ifa_ipv6_policy, extack);
5174         if (err < 0)
5175                 return err;
5176
5177         for (i = 0; i <= IFA_MAX; ++i) {
5178                 if (!tb[i])
5179                         continue;
5180
5181                 if (i == IFA_TARGET_NETNSID) {
5182                         struct net *net;
5183
5184                         fillargs->netnsid = nla_get_s32(tb[i]);
5185                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5186                         if (IS_ERR(net)) {
5187                                 fillargs->netnsid = -1;
5188                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5189                                 return PTR_ERR(net);
5190                         }
5191                         *tgt_net = net;
5192                 } else {
5193                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5194                         return -EINVAL;
5195                 }
5196         }
5197
5198         return 0;
5199 }
5200
5201 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5202                            enum addr_type_t type)
5203 {
5204         const struct nlmsghdr *nlh = cb->nlh;
5205         struct inet6_fill_args fillargs = {
5206                 .portid = NETLINK_CB(cb->skb).portid,
5207                 .seq = cb->nlh->nlmsg_seq,
5208                 .flags = NLM_F_MULTI,
5209                 .netnsid = -1,
5210                 .type = type,
5211         };
5212         struct net *net = sock_net(skb->sk);
5213         struct net *tgt_net = net;
5214         int idx, s_idx, s_ip_idx;
5215         int h, s_h;
5216         struct net_device *dev;
5217         struct inet6_dev *idev;
5218         struct hlist_head *head;
5219         int err = 0;
5220
5221         s_h = cb->args[0];
5222         s_idx = idx = cb->args[1];
5223         s_ip_idx = cb->args[2];
5224
5225         if (cb->strict_check) {
5226                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5227                                                   skb->sk, cb);
5228                 if (err < 0)
5229                         goto put_tgt_net;
5230
5231                 err = 0;
5232                 if (fillargs.ifindex) {
5233                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5234                         if (!dev) {
5235                                 err = -ENODEV;
5236                                 goto put_tgt_net;
5237                         }
5238                         idev = __in6_dev_get(dev);
5239                         if (idev) {
5240                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5241                                                      &fillargs);
5242                                 if (err > 0)
5243                                         err = 0;
5244                         }
5245                         goto put_tgt_net;
5246                 }
5247         }
5248
5249         rcu_read_lock();
5250         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5251         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5252                 idx = 0;
5253                 head = &tgt_net->dev_index_head[h];
5254                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5255                         if (idx < s_idx)
5256                                 goto cont;
5257                         if (h > s_h || idx > s_idx)
5258                                 s_ip_idx = 0;
5259                         idev = __in6_dev_get(dev);
5260                         if (!idev)
5261                                 goto cont;
5262
5263                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5264                                            &fillargs) < 0)
5265                                 goto done;
5266 cont:
5267                         idx++;
5268                 }
5269         }
5270 done:
5271         rcu_read_unlock();
5272         cb->args[0] = h;
5273         cb->args[1] = idx;
5274 put_tgt_net:
5275         if (fillargs.netnsid >= 0)
5276                 put_net(tgt_net);
5277
5278         return skb->len ? : err;
5279 }
5280
5281 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5282 {
5283         enum addr_type_t type = UNICAST_ADDR;
5284
5285         return inet6_dump_addr(skb, cb, type);
5286 }
5287
5288 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5289 {
5290         enum addr_type_t type = MULTICAST_ADDR;
5291
5292         return inet6_dump_addr(skb, cb, type);
5293 }
5294
5295
5296 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5297 {
5298         enum addr_type_t type = ANYCAST_ADDR;
5299
5300         return inet6_dump_addr(skb, cb, type);
5301 }
5302
5303 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5304                                        const struct nlmsghdr *nlh,
5305                                        struct nlattr **tb,
5306                                        struct netlink_ext_ack *extack)
5307 {
5308         struct ifaddrmsg *ifm;
5309         int i, err;
5310
5311         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5312                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5313                 return -EINVAL;
5314         }
5315
5316         if (!netlink_strict_get_check(skb))
5317                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5318                                               ifa_ipv6_policy, extack);
5319
5320         ifm = nlmsg_data(nlh);
5321         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5322                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5323                 return -EINVAL;
5324         }
5325
5326         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5327                                             ifa_ipv6_policy, extack);
5328         if (err)
5329                 return err;
5330
5331         for (i = 0; i <= IFA_MAX; i++) {
5332                 if (!tb[i])
5333                         continue;
5334
5335                 switch (i) {
5336                 case IFA_TARGET_NETNSID:
5337                 case IFA_ADDRESS:
5338                 case IFA_LOCAL:
5339                         break;
5340                 default:
5341                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5342                         return -EINVAL;
5343                 }
5344         }
5345
5346         return 0;
5347 }
5348
5349 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5350                              struct netlink_ext_ack *extack)
5351 {
5352         struct net *net = sock_net(in_skb->sk);
5353         struct inet6_fill_args fillargs = {
5354                 .portid = NETLINK_CB(in_skb).portid,
5355                 .seq = nlh->nlmsg_seq,
5356                 .event = RTM_NEWADDR,
5357                 .flags = 0,
5358                 .netnsid = -1,
5359         };
5360         struct net *tgt_net = net;
5361         struct ifaddrmsg *ifm;
5362         struct nlattr *tb[IFA_MAX+1];
5363         struct in6_addr *addr = NULL, *peer;
5364         struct net_device *dev = NULL;
5365         struct inet6_ifaddr *ifa;
5366         struct sk_buff *skb;
5367         int err;
5368
5369         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5370         if (err < 0)
5371                 return err;
5372
5373         if (tb[IFA_TARGET_NETNSID]) {
5374                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5375
5376                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5377                                                   fillargs.netnsid);
5378                 if (IS_ERR(tgt_net))
5379                         return PTR_ERR(tgt_net);
5380         }
5381
5382         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5383         if (!addr)
5384                 return -EINVAL;
5385
5386         ifm = nlmsg_data(nlh);
5387         if (ifm->ifa_index)
5388                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5389
5390         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5391         if (!ifa) {
5392                 err = -EADDRNOTAVAIL;
5393                 goto errout;
5394         }
5395
5396         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5397         if (!skb) {
5398                 err = -ENOBUFS;
5399                 goto errout_ifa;
5400         }
5401
5402         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5403         if (err < 0) {
5404                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5405                 WARN_ON(err == -EMSGSIZE);
5406                 kfree_skb(skb);
5407                 goto errout_ifa;
5408         }
5409         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5410 errout_ifa:
5411         in6_ifa_put(ifa);
5412 errout:
5413         if (dev)
5414                 dev_put(dev);
5415         if (fillargs.netnsid >= 0)
5416                 put_net(tgt_net);
5417
5418         return err;
5419 }
5420
5421 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5422 {
5423         struct sk_buff *skb;
5424         struct net *net = dev_net(ifa->idev->dev);
5425         struct inet6_fill_args fillargs = {
5426                 .portid = 0,
5427                 .seq = 0,
5428                 .event = event,
5429                 .flags = 0,
5430                 .netnsid = -1,
5431         };
5432         int err = -ENOBUFS;
5433
5434         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5435         if (!skb)
5436                 goto errout;
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;
5444         }
5445         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5446         return;
5447 errout:
5448         if (err < 0)
5449                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5450 }
5451
5452 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5453                                 __s32 *array, int bytes)
5454 {
5455         BUG_ON(bytes < (DEVCONF_MAX * 4));
5456
5457         memset(array, 0, bytes);
5458         array[DEVCONF_FORWARDING] = cnf->forwarding;
5459         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5460         array[DEVCONF_MTU6] = cnf->mtu6;
5461         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5462         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5463         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5464         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5465         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5466         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5467                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5468         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5469                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5470         array[DEVCONF_RTR_SOLICIT_DELAY] =
5471                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5472         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5473         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5474                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5475         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5476                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5477         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5478         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5479         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5480         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5481         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5482         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5483         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5484         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5485         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5486         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5487 #ifdef CONFIG_IPV6_ROUTER_PREF
5488         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5489         array[DEVCONF_RTR_PROBE_INTERVAL] =
5490                 jiffies_to_msecs(cnf->rtr_probe_interval);
5491 #ifdef CONFIG_IPV6_ROUTE_INFO
5492         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5493         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5494 #endif
5495 #endif
5496         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5497         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5498 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5499         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5500         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5501 #endif
5502 #ifdef CONFIG_IPV6_MROUTE
5503         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5504 #endif
5505         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5506         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5507         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5508         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5509         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5510         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5511         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5512         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5513         /* we omit DEVCONF_STABLE_SECRET for now */
5514         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5515         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5516         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5517         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5518         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5519 #ifdef CONFIG_IPV6_SEG6_HMAC
5520         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5521 #endif
5522         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5523         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5524         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5525         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5526         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5527 }
5528
5529 static inline size_t inet6_ifla6_size(void)
5530 {
5531         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5532              + nla_total_size(sizeof(struct ifla_cacheinfo))
5533              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5534              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5535              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5536              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5537              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5538              + 0;
5539 }
5540
5541 static inline size_t inet6_if_nlmsg_size(void)
5542 {
5543         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5544                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5545                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5546                + nla_total_size(4) /* IFLA_MTU */
5547                + nla_total_size(4) /* IFLA_LINK */
5548                + nla_total_size(1) /* IFLA_OPERSTATE */
5549                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5550 }
5551
5552 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5553                                         int bytes)
5554 {
5555         int i;
5556         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5557         BUG_ON(pad < 0);
5558
5559         /* Use put_unaligned() because stats may not be aligned for u64. */
5560         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5561         for (i = 1; i < ICMP6_MIB_MAX; i++)
5562                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5563
5564         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5565 }
5566
5567 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5568                                         int bytes, size_t syncpoff)
5569 {
5570         int i, c;
5571         u64 buff[IPSTATS_MIB_MAX];
5572         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5573
5574         BUG_ON(pad < 0);
5575
5576         memset(buff, 0, sizeof(buff));
5577         buff[0] = IPSTATS_MIB_MAX;
5578
5579         for_each_possible_cpu(c) {
5580                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5581                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5582         }
5583
5584         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5585         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5586 }
5587
5588 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5589                              int bytes)
5590 {
5591         switch (attrtype) {
5592         case IFLA_INET6_STATS:
5593                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5594                                      offsetof(struct ipstats_mib, syncp));
5595                 break;
5596         case IFLA_INET6_ICMP6STATS:
5597                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5598                 break;
5599         }
5600 }
5601
5602 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5603                                   u32 ext_filter_mask)
5604 {
5605         struct nlattr *nla;
5606         struct ifla_cacheinfo ci;
5607
5608         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5609                 goto nla_put_failure;
5610         ci.max_reasm_len = IPV6_MAXPLEN;
5611         ci.tstamp = cstamp_delta(idev->tstamp);
5612         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5613         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5614         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5615                 goto nla_put_failure;
5616         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5617         if (!nla)
5618                 goto nla_put_failure;
5619         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5620
5621         /* XXX - MC not implemented */
5622
5623         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5624                 return 0;
5625
5626         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5627         if (!nla)
5628                 goto nla_put_failure;
5629         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5630
5631         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5632         if (!nla)
5633                 goto nla_put_failure;
5634         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5635
5636         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5637         if (!nla)
5638                 goto nla_put_failure;
5639         read_lock_bh(&idev->lock);
5640         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5641         read_unlock_bh(&idev->lock);
5642
5643         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5644                 goto nla_put_failure;
5645
5646         return 0;
5647
5648 nla_put_failure:
5649         return -EMSGSIZE;
5650 }
5651
5652 static size_t inet6_get_link_af_size(const struct net_device *dev,
5653                                      u32 ext_filter_mask)
5654 {
5655         if (!__in6_dev_get(dev))
5656                 return 0;
5657
5658         return inet6_ifla6_size();
5659 }
5660
5661 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5662                               u32 ext_filter_mask)
5663 {
5664         struct inet6_dev *idev = __in6_dev_get(dev);
5665
5666         if (!idev)
5667                 return -ENODATA;
5668
5669         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5670                 return -EMSGSIZE;
5671
5672         return 0;
5673 }
5674
5675 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5676 {
5677         struct inet6_ifaddr *ifp;
5678         struct net_device *dev = idev->dev;
5679         bool clear_token, update_rs = false;
5680         struct in6_addr ll_addr;
5681
5682         ASSERT_RTNL();
5683
5684         if (!token)
5685                 return -EINVAL;
5686         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5687                 return -EINVAL;
5688         if (!ipv6_accept_ra(idev))
5689                 return -EINVAL;
5690         if (idev->cnf.rtr_solicits == 0)
5691                 return -EINVAL;
5692
5693         write_lock_bh(&idev->lock);
5694
5695         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5696         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5697
5698         write_unlock_bh(&idev->lock);
5699
5700         clear_token = ipv6_addr_any(token);
5701         if (clear_token)
5702                 goto update_lft;
5703
5704         if (!idev->dead && (idev->if_flags & IF_READY) &&
5705             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5706                              IFA_F_OPTIMISTIC)) {
5707                 /* If we're not ready, then normal ifup will take care
5708                  * of this. Otherwise, we need to request our rs here.
5709                  */
5710                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5711                 update_rs = true;
5712         }
5713
5714 update_lft:
5715         write_lock_bh(&idev->lock);
5716
5717         if (update_rs) {
5718                 idev->if_flags |= IF_RS_SENT;
5719                 idev->rs_interval = rfc3315_s14_backoff_init(
5720                         idev->cnf.rtr_solicit_interval);
5721                 idev->rs_probes = 1;
5722                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5723         }
5724
5725         /* Well, that's kinda nasty ... */
5726         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5727                 spin_lock(&ifp->lock);
5728                 if (ifp->tokenized) {
5729                         ifp->valid_lft = 0;
5730                         ifp->prefered_lft = 0;
5731                 }
5732                 spin_unlock(&ifp->lock);
5733         }
5734
5735         write_unlock_bh(&idev->lock);
5736         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5737         addrconf_verify_rtnl();
5738         return 0;
5739 }
5740
5741 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5742         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5743         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5744 };
5745
5746 static int check_addr_gen_mode(int mode)
5747 {
5748         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5749             mode != IN6_ADDR_GEN_MODE_NONE &&
5750             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5751             mode != IN6_ADDR_GEN_MODE_RANDOM)
5752                 return -EINVAL;
5753         return 1;
5754 }
5755
5756 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5757                                 int mode)
5758 {
5759         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5760             !idev->cnf.stable_secret.initialized &&
5761             !net->ipv6.devconf_dflt->stable_secret.initialized)
5762                 return -EINVAL;
5763         return 1;
5764 }
5765
5766 static int inet6_validate_link_af(const struct net_device *dev,
5767                                   const struct nlattr *nla)
5768 {
5769         struct nlattr *tb[IFLA_INET6_MAX + 1];
5770         struct inet6_dev *idev = NULL;
5771         int err;
5772
5773         if (dev) {
5774                 idev = __in6_dev_get(dev);
5775                 if (!idev)
5776                         return -EAFNOSUPPORT;
5777         }
5778
5779         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5780                                           inet6_af_policy, NULL);
5781         if (err)
5782                 return err;
5783
5784         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5785                 return -EINVAL;
5786
5787         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5788                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5789
5790                 if (check_addr_gen_mode(mode) < 0)
5791                         return -EINVAL;
5792                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5793                         return -EINVAL;
5794         }
5795
5796         return 0;
5797 }
5798
5799 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5800 {
5801         struct inet6_dev *idev = __in6_dev_get(dev);
5802         struct nlattr *tb[IFLA_INET6_MAX + 1];
5803         int err;
5804
5805         if (!idev)
5806                 return -EAFNOSUPPORT;
5807
5808         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5809                 BUG();
5810
5811         if (tb[IFLA_INET6_TOKEN]) {
5812                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5813                 if (err)
5814                         return err;
5815         }
5816
5817         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5818                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5819
5820                 idev->cnf.addr_gen_mode = mode;
5821         }
5822
5823         return 0;
5824 }
5825
5826 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5827                              u32 portid, u32 seq, int event, unsigned int flags)
5828 {
5829         struct net_device *dev = idev->dev;
5830         struct ifinfomsg *hdr;
5831         struct nlmsghdr *nlh;
5832         void *protoinfo;
5833
5834         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5835         if (!nlh)
5836                 return -EMSGSIZE;
5837
5838         hdr = nlmsg_data(nlh);
5839         hdr->ifi_family = AF_INET6;
5840         hdr->__ifi_pad = 0;
5841         hdr->ifi_type = dev->type;
5842         hdr->ifi_index = dev->ifindex;
5843         hdr->ifi_flags = dev_get_flags(dev);
5844         hdr->ifi_change = 0;
5845
5846         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5847             (dev->addr_len &&
5848              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5849             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5850             (dev->ifindex != dev_get_iflink(dev) &&
5851              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5852             nla_put_u8(skb, IFLA_OPERSTATE,
5853                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5854                 goto nla_put_failure;
5855         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5856         if (!protoinfo)
5857                 goto nla_put_failure;
5858
5859         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5860                 goto nla_put_failure;
5861
5862         nla_nest_end(skb, protoinfo);
5863         nlmsg_end(skb, nlh);
5864         return 0;
5865
5866 nla_put_failure:
5867         nlmsg_cancel(skb, nlh);
5868         return -EMSGSIZE;
5869 }
5870
5871 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5872                                    struct netlink_ext_ack *extack)
5873 {
5874         struct ifinfomsg *ifm;
5875
5876         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5877                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5878                 return -EINVAL;
5879         }
5880
5881         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5882                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5883                 return -EINVAL;
5884         }
5885
5886         ifm = nlmsg_data(nlh);
5887         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5888             ifm->ifi_change || ifm->ifi_index) {
5889                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5890                 return -EINVAL;
5891         }
5892
5893         return 0;
5894 }
5895
5896 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5897 {
5898         struct net *net = sock_net(skb->sk);
5899         int h, s_h;
5900         int idx = 0, s_idx;
5901         struct net_device *dev;
5902         struct inet6_dev *idev;
5903         struct hlist_head *head;
5904
5905         /* only requests using strict checking can pass data to
5906          * influence the dump
5907          */
5908         if (cb->strict_check) {
5909                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5910
5911                 if (err < 0)
5912                         return err;
5913         }
5914
5915         s_h = cb->args[0];
5916         s_idx = cb->args[1];
5917
5918         rcu_read_lock();
5919         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5920                 idx = 0;
5921                 head = &net->dev_index_head[h];
5922                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5923                         if (idx < s_idx)
5924                                 goto cont;
5925                         idev = __in6_dev_get(dev);
5926                         if (!idev)
5927                                 goto cont;
5928                         if (inet6_fill_ifinfo(skb, idev,
5929                                               NETLINK_CB(cb->skb).portid,
5930                                               cb->nlh->nlmsg_seq,
5931                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5932                                 goto out;
5933 cont:
5934                         idx++;
5935                 }
5936         }
5937 out:
5938         rcu_read_unlock();
5939         cb->args[1] = idx;
5940         cb->args[0] = h;
5941
5942         return skb->len;
5943 }
5944
5945 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5946 {
5947         struct sk_buff *skb;
5948         struct net *net = dev_net(idev->dev);
5949         int err = -ENOBUFS;
5950
5951         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5952         if (!skb)
5953                 goto errout;
5954
5955         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5956         if (err < 0) {
5957                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5958                 WARN_ON(err == -EMSGSIZE);
5959                 kfree_skb(skb);
5960                 goto errout;
5961         }
5962         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5963         return;
5964 errout:
5965         if (err < 0)
5966                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5967 }
5968
5969 static inline size_t inet6_prefix_nlmsg_size(void)
5970 {
5971         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5972                + nla_total_size(sizeof(struct in6_addr))
5973                + nla_total_size(sizeof(struct prefix_cacheinfo));
5974 }
5975
5976 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5977                              struct prefix_info *pinfo, u32 portid, u32 seq,
5978                              int event, unsigned int flags)
5979 {
5980         struct prefixmsg *pmsg;
5981         struct nlmsghdr *nlh;
5982         struct prefix_cacheinfo ci;
5983
5984         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5985         if (!nlh)
5986                 return -EMSGSIZE;
5987
5988         pmsg = nlmsg_data(nlh);
5989         pmsg->prefix_family = AF_INET6;
5990         pmsg->prefix_pad1 = 0;
5991         pmsg->prefix_pad2 = 0;
5992         pmsg->prefix_ifindex = idev->dev->ifindex;
5993         pmsg->prefix_len = pinfo->prefix_len;
5994         pmsg->prefix_type = pinfo->type;
5995         pmsg->prefix_pad3 = 0;
5996         pmsg->prefix_flags = 0;
5997         if (pinfo->onlink)
5998                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5999         if (pinfo->autoconf)
6000                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6001
6002         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6003                 goto nla_put_failure;
6004         ci.preferred_time = ntohl(pinfo->prefered);
6005         ci.valid_time = ntohl(pinfo->valid);
6006         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6007                 goto nla_put_failure;
6008         nlmsg_end(skb, nlh);
6009         return 0;
6010
6011 nla_put_failure:
6012         nlmsg_cancel(skb, nlh);
6013         return -EMSGSIZE;
6014 }
6015
6016 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6017                          struct prefix_info *pinfo)
6018 {
6019         struct sk_buff *skb;
6020         struct net *net = dev_net(idev->dev);
6021         int err = -ENOBUFS;
6022
6023         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6024         if (!skb)
6025                 goto errout;
6026
6027         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6028         if (err < 0) {
6029                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6030                 WARN_ON(err == -EMSGSIZE);
6031                 kfree_skb(skb);
6032                 goto errout;
6033         }
6034         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6035         return;
6036 errout:
6037         if (err < 0)
6038                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6039 }
6040
6041 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6042 {
6043         struct net *net = dev_net(ifp->idev->dev);
6044
6045         if (event)
6046                 ASSERT_RTNL();
6047
6048         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6049
6050         switch (event) {
6051         case RTM_NEWADDR:
6052                 /*
6053                  * If the address was optimistic we inserted the route at the
6054                  * start of our DAD process, so we don't need to do it again.
6055                  * If the device was taken down in the middle of the DAD
6056                  * cycle there is a race where we could get here without a
6057                  * host route, so nothing to insert. That will be fixed when
6058                  * the device is brought up.
6059                  */
6060                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6061                         ip6_ins_rt(net, ifp->rt);
6062                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6063                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6064                                 &ifp->addr, ifp->idev->dev->name);
6065                 }
6066
6067                 if (ifp->idev->cnf.forwarding)
6068                         addrconf_join_anycast(ifp);
6069                 if (!ipv6_addr_any(&ifp->peer_addr))
6070                         addrconf_prefix_route(&ifp->peer_addr, 128,
6071                                               ifp->rt_priority, ifp->idev->dev,
6072                                               0, 0, GFP_ATOMIC);
6073                 break;
6074         case RTM_DELADDR:
6075                 if (ifp->idev->cnf.forwarding)
6076                         addrconf_leave_anycast(ifp);
6077                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6078                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6079                         struct fib6_info *rt;
6080
6081                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6082                                                        ifp->idev->dev, 0, 0,
6083                                                        false);
6084                         if (rt)
6085                                 ip6_del_rt(net, rt, false);
6086                 }
6087                 if (ifp->rt) {
6088                         ip6_del_rt(net, ifp->rt, false);
6089                         ifp->rt = NULL;
6090                 }
6091                 rt_genid_bump_ipv6(net);
6092                 break;
6093         }
6094         atomic_inc(&net->ipv6.dev_addr_genid);
6095 }
6096
6097 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6098 {
6099         if (likely(ifp->idev->dead == 0))
6100                 __ipv6_ifa_notify(event, ifp);
6101 }
6102
6103 #ifdef CONFIG_SYSCTL
6104
6105 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6106                 void *buffer, size_t *lenp, loff_t *ppos)
6107 {
6108         int *valp = ctl->data;
6109         int val = *valp;
6110         loff_t pos = *ppos;
6111         struct ctl_table lctl;
6112         int ret;
6113
6114         /*
6115          * ctl->data points to idev->cnf.forwarding, we should
6116          * not modify it until we get the rtnl lock.
6117          */
6118         lctl = *ctl;
6119         lctl.data = &val;
6120
6121         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6122
6123         if (write)
6124                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6125         if (ret)
6126                 *ppos = pos;
6127         return ret;
6128 }
6129
6130 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6131                 void *buffer, size_t *lenp, loff_t *ppos)
6132 {
6133         struct inet6_dev *idev = ctl->extra1;
6134         int min_mtu = IPV6_MIN_MTU;
6135         struct ctl_table lctl;
6136
6137         lctl = *ctl;
6138         lctl.extra1 = &min_mtu;
6139         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6140
6141         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6142 }
6143
6144 static void dev_disable_change(struct inet6_dev *idev)
6145 {
6146         struct netdev_notifier_info info;
6147
6148         if (!idev || !idev->dev)
6149                 return;
6150
6151         netdev_notifier_info_init(&info, idev->dev);
6152         if (idev->cnf.disable_ipv6)
6153                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6154         else
6155                 addrconf_notify(NULL, NETDEV_UP, &info);
6156 }
6157
6158 static void addrconf_disable_change(struct net *net, __s32 newf)
6159 {
6160         struct net_device *dev;
6161         struct inet6_dev *idev;
6162
6163         for_each_netdev(net, dev) {
6164                 idev = __in6_dev_get(dev);
6165                 if (idev) {
6166                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6167                         idev->cnf.disable_ipv6 = newf;
6168                         if (changed)
6169                                 dev_disable_change(idev);
6170                 }
6171         }
6172 }
6173
6174 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6175 {
6176         struct net *net;
6177         int old;
6178
6179         if (!rtnl_trylock())
6180                 return restart_syscall();
6181
6182         net = (struct net *)table->extra2;
6183         old = *p;
6184         *p = newf;
6185
6186         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6187                 rtnl_unlock();
6188                 return 0;
6189         }
6190
6191         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6192                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6193                 addrconf_disable_change(net, newf);
6194         } else if ((!newf) ^ (!old))
6195                 dev_disable_change((struct inet6_dev *)table->extra1);
6196
6197         rtnl_unlock();
6198         return 0;
6199 }
6200
6201 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6202                 void *buffer, size_t *lenp, loff_t *ppos)
6203 {
6204         int *valp = ctl->data;
6205         int val = *valp;
6206         loff_t pos = *ppos;
6207         struct ctl_table lctl;
6208         int ret;
6209
6210         /*
6211          * ctl->data points to idev->cnf.disable_ipv6, we should
6212          * not modify it until we get the rtnl lock.
6213          */
6214         lctl = *ctl;
6215         lctl.data = &val;
6216
6217         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6218
6219         if (write)
6220                 ret = addrconf_disable_ipv6(ctl, valp, val);
6221         if (ret)
6222                 *ppos = pos;
6223         return ret;
6224 }
6225
6226 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6227                 void *buffer, size_t *lenp, loff_t *ppos)
6228 {
6229         int *valp = ctl->data;
6230         int ret;
6231         int old, new;
6232
6233         old = *valp;
6234         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6235         new = *valp;
6236
6237         if (write && old != new) {
6238                 struct net *net = ctl->extra2;
6239
6240                 if (!rtnl_trylock())
6241                         return restart_syscall();
6242
6243                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6244                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6245                                                      NETCONFA_PROXY_NEIGH,
6246                                                      NETCONFA_IFINDEX_DEFAULT,
6247                                                      net->ipv6.devconf_dflt);
6248                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6249                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6250                                                      NETCONFA_PROXY_NEIGH,
6251                                                      NETCONFA_IFINDEX_ALL,
6252                                                      net->ipv6.devconf_all);
6253                 else {
6254                         struct inet6_dev *idev = ctl->extra1;
6255
6256                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6257                                                      NETCONFA_PROXY_NEIGH,
6258                                                      idev->dev->ifindex,
6259                                                      &idev->cnf);
6260                 }
6261                 rtnl_unlock();
6262         }
6263
6264         return ret;
6265 }
6266
6267 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6268                                          void *buffer, size_t *lenp,
6269                                          loff_t *ppos)
6270 {
6271         int ret = 0;
6272         u32 new_val;
6273         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6274         struct net *net = (struct net *)ctl->extra2;
6275         struct ctl_table tmp = {
6276                 .data = &new_val,
6277                 .maxlen = sizeof(new_val),
6278                 .mode = ctl->mode,
6279         };
6280
6281         if (!rtnl_trylock())
6282                 return restart_syscall();
6283
6284         new_val = *((u32 *)ctl->data);
6285
6286         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6287         if (ret != 0)
6288                 goto out;
6289
6290         if (write) {
6291                 if (check_addr_gen_mode(new_val) < 0) {
6292                         ret = -EINVAL;
6293                         goto out;
6294                 }
6295
6296                 if (idev) {
6297                         if (check_stable_privacy(idev, net, new_val) < 0) {
6298                                 ret = -EINVAL;
6299                                 goto out;
6300                         }
6301
6302                         if (idev->cnf.addr_gen_mode != new_val) {
6303                                 idev->cnf.addr_gen_mode = new_val;
6304                                 addrconf_dev_config(idev->dev);
6305                         }
6306                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6307                         struct net_device *dev;
6308
6309                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6310                         for_each_netdev(net, dev) {
6311                                 idev = __in6_dev_get(dev);
6312                                 if (idev &&
6313                                     idev->cnf.addr_gen_mode != new_val) {
6314                                         idev->cnf.addr_gen_mode = new_val;
6315                                         addrconf_dev_config(idev->dev);
6316                                 }
6317                         }
6318                 }
6319
6320                 *((u32 *)ctl->data) = new_val;
6321         }
6322
6323 out:
6324         rtnl_unlock();
6325
6326         return ret;
6327 }
6328
6329 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6330                                          void *buffer, size_t *lenp,
6331                                          loff_t *ppos)
6332 {
6333         int err;
6334         struct in6_addr addr;
6335         char str[IPV6_MAX_STRLEN];
6336         struct ctl_table lctl = *ctl;
6337         struct net *net = ctl->extra2;
6338         struct ipv6_stable_secret *secret = ctl->data;
6339
6340         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6341                 return -EIO;
6342
6343         lctl.maxlen = IPV6_MAX_STRLEN;
6344         lctl.data = str;
6345
6346         if (!rtnl_trylock())
6347                 return restart_syscall();
6348
6349         if (!write && !secret->initialized) {
6350                 err = -EIO;
6351                 goto out;
6352         }
6353
6354         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6355         if (err >= sizeof(str)) {
6356                 err = -EIO;
6357                 goto out;
6358         }
6359
6360         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6361         if (err || !write)
6362                 goto out;
6363
6364         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6365                 err = -EIO;
6366                 goto out;
6367         }
6368
6369         secret->initialized = true;
6370         secret->secret = addr;
6371
6372         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6373                 struct net_device *dev;
6374
6375                 for_each_netdev(net, dev) {
6376                         struct inet6_dev *idev = __in6_dev_get(dev);
6377
6378                         if (idev) {
6379                                 idev->cnf.addr_gen_mode =
6380                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6381                         }
6382                 }
6383         } else {
6384                 struct inet6_dev *idev = ctl->extra1;
6385
6386                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6387         }
6388
6389 out:
6390         rtnl_unlock();
6391
6392         return err;
6393 }
6394
6395 static
6396 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6397                                                 int write, void *buffer,
6398                                                 size_t *lenp,
6399                                                 loff_t *ppos)
6400 {
6401         int *valp = ctl->data;
6402         int val = *valp;
6403         loff_t pos = *ppos;
6404         struct ctl_table lctl;
6405         int ret;
6406
6407         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6408          * we should not modify it until we get the rtnl lock.
6409          */
6410         lctl = *ctl;
6411         lctl.data = &val;
6412
6413         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6414
6415         if (write)
6416                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6417         if (ret)
6418                 *ppos = pos;
6419         return ret;
6420 }
6421
6422 static
6423 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6424 {
6425         if (rt) {
6426                 if (action)
6427                         rt->dst.flags |= DST_NOPOLICY;
6428                 else
6429                         rt->dst.flags &= ~DST_NOPOLICY;
6430         }
6431 }
6432
6433 static
6434 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6435 {
6436         struct inet6_ifaddr *ifa;
6437
6438         read_lock_bh(&idev->lock);
6439         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6440                 spin_lock(&ifa->lock);
6441                 if (ifa->rt) {
6442                         /* host routes only use builtin fib6_nh */
6443                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6444                         int cpu;
6445
6446                         rcu_read_lock();
6447                         ifa->rt->dst_nopolicy = val ? true : false;
6448                         if (nh->rt6i_pcpu) {
6449                                 for_each_possible_cpu(cpu) {
6450                                         struct rt6_info **rtp;
6451
6452                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6453                                         addrconf_set_nopolicy(*rtp, val);
6454                                 }
6455                         }
6456                         rcu_read_unlock();
6457                 }
6458                 spin_unlock(&ifa->lock);
6459         }
6460         read_unlock_bh(&idev->lock);
6461 }
6462
6463 static
6464 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6465 {
6466         struct inet6_dev *idev;
6467         struct net *net;
6468
6469         if (!rtnl_trylock())
6470                 return restart_syscall();
6471
6472         *valp = val;
6473
6474         net = (struct net *)ctl->extra2;
6475         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6476                 rtnl_unlock();
6477                 return 0;
6478         }
6479
6480         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6481                 struct net_device *dev;
6482
6483                 for_each_netdev(net, dev) {
6484                         idev = __in6_dev_get(dev);
6485                         if (idev)
6486                                 addrconf_disable_policy_idev(idev, val);
6487                 }
6488         } else {
6489                 idev = (struct inet6_dev *)ctl->extra1;
6490                 addrconf_disable_policy_idev(idev, val);
6491         }
6492
6493         rtnl_unlock();
6494         return 0;
6495 }
6496
6497 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6498                                    void *buffer, size_t *lenp, loff_t *ppos)
6499 {
6500         int *valp = ctl->data;
6501         int val = *valp;
6502         loff_t pos = *ppos;
6503         struct ctl_table lctl;
6504         int ret;
6505
6506         lctl = *ctl;
6507         lctl.data = &val;
6508         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6509
6510         if (write && (*valp != val))
6511                 ret = addrconf_disable_policy(ctl, valp, val);
6512
6513         if (ret)
6514                 *ppos = pos;
6515
6516         return ret;
6517 }
6518
6519 static int minus_one = -1;
6520 static const int two_five_five = 255;
6521
6522 static const struct ctl_table addrconf_sysctl[] = {
6523         {
6524                 .procname       = "forwarding",
6525                 .data           = &ipv6_devconf.forwarding,
6526                 .maxlen         = sizeof(int),
6527                 .mode           = 0644,
6528                 .proc_handler   = addrconf_sysctl_forward,
6529         },
6530         {
6531                 .procname       = "hop_limit",
6532                 .data           = &ipv6_devconf.hop_limit,
6533                 .maxlen         = sizeof(int),
6534                 .mode           = 0644,
6535                 .proc_handler   = proc_dointvec_minmax,
6536                 .extra1         = (void *)SYSCTL_ONE,
6537                 .extra2         = (void *)&two_five_five,
6538         },
6539         {
6540                 .procname       = "mtu",
6541                 .data           = &ipv6_devconf.mtu6,
6542                 .maxlen         = sizeof(int),
6543                 .mode           = 0644,
6544                 .proc_handler   = addrconf_sysctl_mtu,
6545         },
6546         {
6547                 .procname       = "accept_ra",
6548                 .data           = &ipv6_devconf.accept_ra,
6549                 .maxlen         = sizeof(int),
6550                 .mode           = 0644,
6551                 .proc_handler   = proc_dointvec,
6552         },
6553         {
6554                 .procname       = "accept_redirects",
6555                 .data           = &ipv6_devconf.accept_redirects,
6556                 .maxlen         = sizeof(int),
6557                 .mode           = 0644,
6558                 .proc_handler   = proc_dointvec,
6559         },
6560         {
6561                 .procname       = "autoconf",
6562                 .data           = &ipv6_devconf.autoconf,
6563                 .maxlen         = sizeof(int),
6564                 .mode           = 0644,
6565                 .proc_handler   = proc_dointvec,
6566         },
6567         {
6568                 .procname       = "dad_transmits",
6569                 .data           = &ipv6_devconf.dad_transmits,
6570                 .maxlen         = sizeof(int),
6571                 .mode           = 0644,
6572                 .proc_handler   = proc_dointvec,
6573         },
6574         {
6575                 .procname       = "router_solicitations",
6576                 .data           = &ipv6_devconf.rtr_solicits,
6577                 .maxlen         = sizeof(int),
6578                 .mode           = 0644,
6579                 .proc_handler   = proc_dointvec_minmax,
6580                 .extra1         = &minus_one,
6581         },
6582         {
6583                 .procname       = "router_solicitation_interval",
6584                 .data           = &ipv6_devconf.rtr_solicit_interval,
6585                 .maxlen         = sizeof(int),
6586                 .mode           = 0644,
6587                 .proc_handler   = proc_dointvec_jiffies,
6588         },
6589         {
6590                 .procname       = "router_solicitation_max_interval",
6591                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6592                 .maxlen         = sizeof(int),
6593                 .mode           = 0644,
6594                 .proc_handler   = proc_dointvec_jiffies,
6595         },
6596         {
6597                 .procname       = "router_solicitation_delay",
6598                 .data           = &ipv6_devconf.rtr_solicit_delay,
6599                 .maxlen         = sizeof(int),
6600                 .mode           = 0644,
6601                 .proc_handler   = proc_dointvec_jiffies,
6602         },
6603         {
6604                 .procname       = "force_mld_version",
6605                 .data           = &ipv6_devconf.force_mld_version,
6606                 .maxlen         = sizeof(int),
6607                 .mode           = 0644,
6608                 .proc_handler   = proc_dointvec,
6609         },
6610         {
6611                 .procname       = "mldv1_unsolicited_report_interval",
6612                 .data           =
6613                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6614                 .maxlen         = sizeof(int),
6615                 .mode           = 0644,
6616                 .proc_handler   = proc_dointvec_ms_jiffies,
6617         },
6618         {
6619                 .procname       = "mldv2_unsolicited_report_interval",
6620                 .data           =
6621                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6622                 .maxlen         = sizeof(int),
6623                 .mode           = 0644,
6624                 .proc_handler   = proc_dointvec_ms_jiffies,
6625         },
6626         {
6627                 .procname       = "use_tempaddr",
6628                 .data           = &ipv6_devconf.use_tempaddr,
6629                 .maxlen         = sizeof(int),
6630                 .mode           = 0644,
6631                 .proc_handler   = proc_dointvec,
6632         },
6633         {
6634                 .procname       = "temp_valid_lft",
6635                 .data           = &ipv6_devconf.temp_valid_lft,
6636                 .maxlen         = sizeof(int),
6637                 .mode           = 0644,
6638                 .proc_handler   = proc_dointvec,
6639         },
6640         {
6641                 .procname       = "temp_prefered_lft",
6642                 .data           = &ipv6_devconf.temp_prefered_lft,
6643                 .maxlen         = sizeof(int),
6644                 .mode           = 0644,
6645                 .proc_handler   = proc_dointvec,
6646         },
6647         {
6648                 .procname       = "regen_max_retry",
6649                 .data           = &ipv6_devconf.regen_max_retry,
6650                 .maxlen         = sizeof(int),
6651                 .mode           = 0644,
6652                 .proc_handler   = proc_dointvec,
6653         },
6654         {
6655                 .procname       = "max_desync_factor",
6656                 .data           = &ipv6_devconf.max_desync_factor,
6657                 .maxlen         = sizeof(int),
6658                 .mode           = 0644,
6659                 .proc_handler   = proc_dointvec,
6660         },
6661         {
6662                 .procname       = "max_addresses",
6663                 .data           = &ipv6_devconf.max_addresses,
6664                 .maxlen         = sizeof(int),
6665                 .mode           = 0644,
6666                 .proc_handler   = proc_dointvec,
6667         },
6668         {
6669                 .procname       = "accept_ra_defrtr",
6670                 .data           = &ipv6_devconf.accept_ra_defrtr,
6671                 .maxlen         = sizeof(int),
6672                 .mode           = 0644,
6673                 .proc_handler   = proc_dointvec,
6674         },
6675         {
6676                 .procname       = "ra_defrtr_metric",
6677                 .data           = &ipv6_devconf.ra_defrtr_metric,
6678                 .maxlen         = sizeof(u32),
6679                 .mode           = 0644,
6680                 .proc_handler   = proc_douintvec_minmax,
6681                 .extra1         = (void *)SYSCTL_ONE,
6682         },
6683         {
6684                 .procname       = "accept_ra_min_hop_limit",
6685                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6686                 .maxlen         = sizeof(int),
6687                 .mode           = 0644,
6688                 .proc_handler   = proc_dointvec,
6689         },
6690         {
6691                 .procname       = "accept_ra_pinfo",
6692                 .data           = &ipv6_devconf.accept_ra_pinfo,
6693                 .maxlen         = sizeof(int),
6694                 .mode           = 0644,
6695                 .proc_handler   = proc_dointvec,
6696         },
6697 #ifdef CONFIG_IPV6_ROUTER_PREF
6698         {
6699                 .procname       = "accept_ra_rtr_pref",
6700                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6701                 .maxlen         = sizeof(int),
6702                 .mode           = 0644,
6703                 .proc_handler   = proc_dointvec,
6704         },
6705         {
6706                 .procname       = "router_probe_interval",
6707                 .data           = &ipv6_devconf.rtr_probe_interval,
6708                 .maxlen         = sizeof(int),
6709                 .mode           = 0644,
6710                 .proc_handler   = proc_dointvec_jiffies,
6711         },
6712 #ifdef CONFIG_IPV6_ROUTE_INFO
6713         {
6714                 .procname       = "accept_ra_rt_info_min_plen",
6715                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6716                 .maxlen         = sizeof(int),
6717                 .mode           = 0644,
6718                 .proc_handler   = proc_dointvec,
6719         },
6720         {
6721                 .procname       = "accept_ra_rt_info_max_plen",
6722                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6723                 .maxlen         = sizeof(int),
6724                 .mode           = 0644,
6725                 .proc_handler   = proc_dointvec,
6726         },
6727 #endif
6728 #endif
6729         {
6730                 .procname       = "proxy_ndp",
6731                 .data           = &ipv6_devconf.proxy_ndp,
6732                 .maxlen         = sizeof(int),
6733                 .mode           = 0644,
6734                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6735         },
6736         {
6737                 .procname       = "accept_source_route",
6738                 .data           = &ipv6_devconf.accept_source_route,
6739                 .maxlen         = sizeof(int),
6740                 .mode           = 0644,
6741                 .proc_handler   = proc_dointvec,
6742         },
6743 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6744         {
6745                 .procname       = "optimistic_dad",
6746                 .data           = &ipv6_devconf.optimistic_dad,
6747                 .maxlen         = sizeof(int),
6748                 .mode           = 0644,
6749                 .proc_handler   = proc_dointvec,
6750         },
6751         {
6752                 .procname       = "use_optimistic",
6753                 .data           = &ipv6_devconf.use_optimistic,
6754                 .maxlen         = sizeof(int),
6755                 .mode           = 0644,
6756                 .proc_handler   = proc_dointvec,
6757         },
6758 #endif
6759 #ifdef CONFIG_IPV6_MROUTE
6760         {
6761                 .procname       = "mc_forwarding",
6762                 .data           = &ipv6_devconf.mc_forwarding,
6763                 .maxlen         = sizeof(int),
6764                 .mode           = 0444,
6765                 .proc_handler   = proc_dointvec,
6766         },
6767 #endif
6768         {
6769                 .procname       = "disable_ipv6",
6770                 .data           = &ipv6_devconf.disable_ipv6,
6771                 .maxlen         = sizeof(int),
6772                 .mode           = 0644,
6773                 .proc_handler   = addrconf_sysctl_disable,
6774         },
6775         {
6776                 .procname       = "accept_dad",
6777                 .data           = &ipv6_devconf.accept_dad,
6778                 .maxlen         = sizeof(int),
6779                 .mode           = 0644,
6780                 .proc_handler   = proc_dointvec,
6781         },
6782         {
6783                 .procname       = "force_tllao",
6784                 .data           = &ipv6_devconf.force_tllao,
6785                 .maxlen         = sizeof(int),
6786                 .mode           = 0644,
6787                 .proc_handler   = proc_dointvec
6788         },
6789         {
6790                 .procname       = "ndisc_notify",
6791                 .data           = &ipv6_devconf.ndisc_notify,
6792                 .maxlen         = sizeof(int),
6793                 .mode           = 0644,
6794                 .proc_handler   = proc_dointvec
6795         },
6796         {
6797                 .procname       = "suppress_frag_ndisc",
6798                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6799                 .maxlen         = sizeof(int),
6800                 .mode           = 0644,
6801                 .proc_handler   = proc_dointvec
6802         },
6803         {
6804                 .procname       = "accept_ra_from_local",
6805                 .data           = &ipv6_devconf.accept_ra_from_local,
6806                 .maxlen         = sizeof(int),
6807                 .mode           = 0644,
6808                 .proc_handler   = proc_dointvec,
6809         },
6810         {
6811                 .procname       = "accept_ra_mtu",
6812                 .data           = &ipv6_devconf.accept_ra_mtu,
6813                 .maxlen         = sizeof(int),
6814                 .mode           = 0644,
6815                 .proc_handler   = proc_dointvec,
6816         },
6817         {
6818                 .procname       = "stable_secret",
6819                 .data           = &ipv6_devconf.stable_secret,
6820                 .maxlen         = IPV6_MAX_STRLEN,
6821                 .mode           = 0600,
6822                 .proc_handler   = addrconf_sysctl_stable_secret,
6823         },
6824         {
6825                 .procname       = "use_oif_addrs_only",
6826                 .data           = &ipv6_devconf.use_oif_addrs_only,
6827                 .maxlen         = sizeof(int),
6828                 .mode           = 0644,
6829                 .proc_handler   = proc_dointvec,
6830         },
6831         {
6832                 .procname       = "ignore_routes_with_linkdown",
6833                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6834                 .maxlen         = sizeof(int),
6835                 .mode           = 0644,
6836                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6837         },
6838         {
6839                 .procname       = "drop_unicast_in_l2_multicast",
6840                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6841                 .maxlen         = sizeof(int),
6842                 .mode           = 0644,
6843                 .proc_handler   = proc_dointvec,
6844         },
6845         {
6846                 .procname       = "drop_unsolicited_na",
6847                 .data           = &ipv6_devconf.drop_unsolicited_na,
6848                 .maxlen         = sizeof(int),
6849                 .mode           = 0644,
6850                 .proc_handler   = proc_dointvec,
6851         },
6852         {
6853                 .procname       = "keep_addr_on_down",
6854                 .data           = &ipv6_devconf.keep_addr_on_down,
6855                 .maxlen         = sizeof(int),
6856                 .mode           = 0644,
6857                 .proc_handler   = proc_dointvec,
6858
6859         },
6860         {
6861                 .procname       = "seg6_enabled",
6862                 .data           = &ipv6_devconf.seg6_enabled,
6863                 .maxlen         = sizeof(int),
6864                 .mode           = 0644,
6865                 .proc_handler   = proc_dointvec,
6866         },
6867 #ifdef CONFIG_IPV6_SEG6_HMAC
6868         {
6869                 .procname       = "seg6_require_hmac",
6870                 .data           = &ipv6_devconf.seg6_require_hmac,
6871                 .maxlen         = sizeof(int),
6872                 .mode           = 0644,
6873                 .proc_handler   = proc_dointvec,
6874         },
6875 #endif
6876         {
6877                 .procname       = "enhanced_dad",
6878                 .data           = &ipv6_devconf.enhanced_dad,
6879                 .maxlen         = sizeof(int),
6880                 .mode           = 0644,
6881                 .proc_handler   = proc_dointvec,
6882         },
6883         {
6884                 .procname               = "addr_gen_mode",
6885                 .data                   = &ipv6_devconf.addr_gen_mode,
6886                 .maxlen                 = sizeof(int),
6887                 .mode                   = 0644,
6888                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6889         },
6890         {
6891                 .procname       = "disable_policy",
6892                 .data           = &ipv6_devconf.disable_policy,
6893                 .maxlen         = sizeof(int),
6894                 .mode           = 0644,
6895                 .proc_handler   = addrconf_sysctl_disable_policy,
6896         },
6897         {
6898                 .procname       = "ndisc_tclass",
6899                 .data           = &ipv6_devconf.ndisc_tclass,
6900                 .maxlen         = sizeof(int),
6901                 .mode           = 0644,
6902                 .proc_handler   = proc_dointvec_minmax,
6903                 .extra1         = (void *)SYSCTL_ZERO,
6904                 .extra2         = (void *)&two_five_five,
6905         },
6906         {
6907                 .procname       = "rpl_seg_enabled",
6908                 .data           = &ipv6_devconf.rpl_seg_enabled,
6909                 .maxlen         = sizeof(int),
6910                 .mode           = 0644,
6911                 .proc_handler   = proc_dointvec,
6912         },
6913         {
6914                 /* sentinel */
6915         }
6916 };
6917
6918 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6919                 struct inet6_dev *idev, struct ipv6_devconf *p)
6920 {
6921         int i, ifindex;
6922         struct ctl_table *table;
6923         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6924
6925         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6926         if (!table)
6927                 goto out;
6928
6929         for (i = 0; table[i].data; i++) {
6930                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6931                 /* If one of these is already set, then it is not safe to
6932                  * overwrite either of them: this makes proc_dointvec_minmax
6933                  * usable.
6934                  */
6935                 if (!table[i].extra1 && !table[i].extra2) {
6936                         table[i].extra1 = idev; /* embedded; no ref */
6937                         table[i].extra2 = net;
6938                 }
6939         }
6940
6941         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6942
6943         p->sysctl_header = register_net_sysctl(net, path, table);
6944         if (!p->sysctl_header)
6945                 goto free;
6946
6947         if (!strcmp(dev_name, "all"))
6948                 ifindex = NETCONFA_IFINDEX_ALL;
6949         else if (!strcmp(dev_name, "default"))
6950                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6951         else
6952                 ifindex = idev->dev->ifindex;
6953         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6954                                      ifindex, p);
6955         return 0;
6956
6957 free:
6958         kfree(table);
6959 out:
6960         return -ENOBUFS;
6961 }
6962
6963 static void __addrconf_sysctl_unregister(struct net *net,
6964                                          struct ipv6_devconf *p, int ifindex)
6965 {
6966         struct ctl_table *table;
6967
6968         if (!p->sysctl_header)
6969                 return;
6970
6971         table = p->sysctl_header->ctl_table_arg;
6972         unregister_net_sysctl_table(p->sysctl_header);
6973         p->sysctl_header = NULL;
6974         kfree(table);
6975
6976         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6977 }
6978
6979 static int addrconf_sysctl_register(struct inet6_dev *idev)
6980 {
6981         int err;
6982
6983         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6984                 return -EINVAL;
6985
6986         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6987                                     &ndisc_ifinfo_sysctl_change);
6988         if (err)
6989                 return err;
6990         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6991                                          idev, &idev->cnf);
6992         if (err)
6993                 neigh_sysctl_unregister(idev->nd_parms);
6994
6995         return err;
6996 }
6997
6998 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6999 {
7000         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7001                                      idev->dev->ifindex);
7002         neigh_sysctl_unregister(idev->nd_parms);
7003 }
7004
7005
7006 #endif
7007
7008 static int __net_init addrconf_init_net(struct net *net)
7009 {
7010         int err = -ENOMEM;
7011         struct ipv6_devconf *all, *dflt;
7012
7013         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7014         if (!all)
7015                 goto err_alloc_all;
7016
7017         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7018         if (!dflt)
7019                 goto err_alloc_dflt;
7020
7021         if (IS_ENABLED(CONFIG_SYSCTL) &&
7022             !net_eq(net, &init_net)) {
7023                 switch (sysctl_devconf_inherit_init_net) {
7024                 case 1:  /* copy from init_net */
7025                         memcpy(all, init_net.ipv6.devconf_all,
7026                                sizeof(ipv6_devconf));
7027                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7028                                sizeof(ipv6_devconf_dflt));
7029                         break;
7030                 case 3: /* copy from the current netns */
7031                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7032                                sizeof(ipv6_devconf));
7033                         memcpy(dflt,
7034                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7035                                sizeof(ipv6_devconf_dflt));
7036                         break;
7037                 case 0:
7038                 case 2:
7039                         /* use compiled values */
7040                         break;
7041                 }
7042         }
7043
7044         /* these will be inherited by all namespaces */
7045         dflt->autoconf = ipv6_defaults.autoconf;
7046         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7047
7048         dflt->stable_secret.initialized = false;
7049         all->stable_secret.initialized = false;
7050
7051         net->ipv6.devconf_all = all;
7052         net->ipv6.devconf_dflt = dflt;
7053
7054 #ifdef CONFIG_SYSCTL
7055         err = __addrconf_sysctl_register(net, "all", NULL, all);
7056         if (err < 0)
7057                 goto err_reg_all;
7058
7059         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7060         if (err < 0)
7061                 goto err_reg_dflt;
7062 #endif
7063         return 0;
7064
7065 #ifdef CONFIG_SYSCTL
7066 err_reg_dflt:
7067         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7068 err_reg_all:
7069         kfree(dflt);
7070 #endif
7071 err_alloc_dflt:
7072         kfree(all);
7073 err_alloc_all:
7074         return err;
7075 }
7076
7077 static void __net_exit addrconf_exit_net(struct net *net)
7078 {
7079 #ifdef CONFIG_SYSCTL
7080         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7081                                      NETCONFA_IFINDEX_DEFAULT);
7082         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7083                                      NETCONFA_IFINDEX_ALL);
7084 #endif
7085         kfree(net->ipv6.devconf_dflt);
7086         kfree(net->ipv6.devconf_all);
7087 }
7088
7089 static struct pernet_operations addrconf_ops = {
7090         .init = addrconf_init_net,
7091         .exit = addrconf_exit_net,
7092 };
7093
7094 static struct rtnl_af_ops inet6_ops __read_mostly = {
7095         .family           = AF_INET6,
7096         .fill_link_af     = inet6_fill_link_af,
7097         .get_link_af_size = inet6_get_link_af_size,
7098         .validate_link_af = inet6_validate_link_af,
7099         .set_link_af      = inet6_set_link_af,
7100 };
7101
7102 /*
7103  *      Init / cleanup code
7104  */
7105
7106 int __init addrconf_init(void)
7107 {
7108         struct inet6_dev *idev;
7109         int i, err;
7110
7111         err = ipv6_addr_label_init();
7112         if (err < 0) {
7113                 pr_crit("%s: cannot initialize default policy table: %d\n",
7114                         __func__, err);
7115                 goto out;
7116         }
7117
7118         err = register_pernet_subsys(&addrconf_ops);
7119         if (err < 0)
7120                 goto out_addrlabel;
7121
7122         addrconf_wq = create_workqueue("ipv6_addrconf");
7123         if (!addrconf_wq) {
7124                 err = -ENOMEM;
7125                 goto out_nowq;
7126         }
7127
7128         /* The addrconf netdev notifier requires that loopback_dev
7129          * has it's ipv6 private information allocated and setup
7130          * before it can bring up and give link-local addresses
7131          * to other devices which are up.
7132          *
7133          * Unfortunately, loopback_dev is not necessarily the first
7134          * entry in the global dev_base list of net devices.  In fact,
7135          * it is likely to be the very last entry on that list.
7136          * So this causes the notifier registry below to try and
7137          * give link-local addresses to all devices besides loopback_dev
7138          * first, then loopback_dev, which cases all the non-loopback_dev
7139          * devices to fail to get a link-local address.
7140          *
7141          * So, as a temporary fix, allocate the ipv6 structure for
7142          * loopback_dev first by hand.
7143          * Longer term, all of the dependencies ipv6 has upon the loopback
7144          * device and it being up should be removed.
7145          */
7146         rtnl_lock();
7147         idev = ipv6_add_dev(init_net.loopback_dev);
7148         rtnl_unlock();
7149         if (IS_ERR(idev)) {
7150                 err = PTR_ERR(idev);
7151                 goto errlo;
7152         }
7153
7154         ip6_route_init_special_entries();
7155
7156         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7157                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7158
7159         register_netdevice_notifier(&ipv6_dev_notf);
7160
7161         addrconf_verify();
7162
7163         rtnl_af_register(&inet6_ops);
7164
7165         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7166                                    NULL, inet6_dump_ifinfo, 0);
7167         if (err < 0)
7168                 goto errout;
7169
7170         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7171                                    inet6_rtm_newaddr, NULL, 0);
7172         if (err < 0)
7173                 goto errout;
7174         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7175                                    inet6_rtm_deladdr, NULL, 0);
7176         if (err < 0)
7177                 goto errout;
7178         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7179                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7180                                    RTNL_FLAG_DOIT_UNLOCKED);
7181         if (err < 0)
7182                 goto errout;
7183         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7184                                    NULL, inet6_dump_ifmcaddr, 0);
7185         if (err < 0)
7186                 goto errout;
7187         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7188                                    NULL, inet6_dump_ifacaddr, 0);
7189         if (err < 0)
7190                 goto errout;
7191         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7192                                    inet6_netconf_get_devconf,
7193                                    inet6_netconf_dump_devconf,
7194                                    RTNL_FLAG_DOIT_UNLOCKED);
7195         if (err < 0)
7196                 goto errout;
7197         err = ipv6_addr_label_rtnl_register();
7198         if (err < 0)
7199                 goto errout;
7200
7201         return 0;
7202 errout:
7203         rtnl_unregister_all(PF_INET6);
7204         rtnl_af_unregister(&inet6_ops);
7205         unregister_netdevice_notifier(&ipv6_dev_notf);
7206 errlo:
7207         destroy_workqueue(addrconf_wq);
7208 out_nowq:
7209         unregister_pernet_subsys(&addrconf_ops);
7210 out_addrlabel:
7211         ipv6_addr_label_cleanup();
7212 out:
7213         return err;
7214 }
7215
7216 void addrconf_cleanup(void)
7217 {
7218         struct net_device *dev;
7219         int i;
7220
7221         unregister_netdevice_notifier(&ipv6_dev_notf);
7222         unregister_pernet_subsys(&addrconf_ops);
7223         ipv6_addr_label_cleanup();
7224
7225         rtnl_af_unregister(&inet6_ops);
7226
7227         rtnl_lock();
7228
7229         /* clean dev list */
7230         for_each_netdev(&init_net, dev) {
7231                 if (__in6_dev_get(dev) == NULL)
7232                         continue;
7233                 addrconf_ifdown(dev, true);
7234         }
7235         addrconf_ifdown(init_net.loopback_dev, true);
7236
7237         /*
7238          *      Check hash table.
7239          */
7240         spin_lock_bh(&addrconf_hash_lock);
7241         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7242                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7243         spin_unlock_bh(&addrconf_hash_lock);
7244         cancel_delayed_work(&addr_chk_work);
7245         rtnl_unlock();
7246
7247         destroy_workqueue(addrconf_wq);
7248 }