net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/sched/signal.h>
  47 #include <linux/socket.h>
  48 #include <linux/sockios.h>
  49 #include <linux/net.h>
  50 #include <linux/inet.h>
  51 #include <linux/in6.h>
  52 #include <linux/netdevice.h>
  53 #include <linux/if_addr.h>
  54 #include <linux/if_arp.h>
  55 #include <linux/if_arcnet.h>
  56 #include <linux/if_infiniband.h>
  57 #include <linux/route.h>
  58 #include <linux/inetdevice.h>
  59 #include <linux/init.h>
  60 #include <linux/slab.h>
  61 #ifdef CONFIG_SYSCTL
  62 #include <linux/sysctl.h>
  63 #endif
  64 #include <linux/capability.h>
  65 #include <linux/delay.h>
  66 #include <linux/notifier.h>
  67 #include <linux/string.h>
  68 #include <linux/hash.h>
  69
  70 #include <net/net_namespace.h>
  71 #include <net/sock.h>
  72 #include <net/snmp.h>
  73
  74 #include <net/6lowpan.h>
  75 #include <net/firewire.h>
  76 #include <net/ipv6.h>
  77 #include <net/protocol.h>
  78 #include <net/ndisc.h>
  79 #include <net/ip6_route.h>
  80 #include <net/addrconf.h>
  81 #include <net/tcp.h>
  82 #include <net/ip.h>
  83 #include <net/netlink.h>
  84 #include <net/pkt_sched.h>
  85 #include <net/l3mdev.h>
  86 #include <linux/if_tunnel.h>
  87 #include <linux/rtnetlink.h>
  88 #include <linux/netconf.h>
  89 #include <linux/random.h>
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  98
  99 #define IPV6_MAX_STRLEN \
 100         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 108 {
 109         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 110         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 111         do_div(tmp, 1000000);
 112         return (s32)tmp;
 113 }
 114
 115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 116 {
 117         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 118         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 119         do_div(tmp, 1000000);
 120         if ((s32)tmp > mrt) {
 121                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 122                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 123                 do_div(tmp, 1000000);
 124         }
 125         return (s32)tmp;
 126 }
 127
 128 #ifdef CONFIG_SYSCTL
 129 static int addrconf_sysctl_register(struct inet6_dev *idev);
 130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 131 #else
 132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 133 {
 134         return 0;
 135 }
 136
 137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 138 {
 139 }
 140 #endif
 141
 142 static void ipv6_regen_rndid(struct inet6_dev *idev);
 143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 144
 145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 146 static int ipv6_count_addresses(const struct inet6_dev *idev);
 147 static int ipv6_generate_stable_address(struct in6_addr *addr,
 148                                         u8 dad_count,
 149                                         const struct inet6_dev *idev);
 150
 151 #define IN6_ADDR_HSIZE_SHIFT    8
 152 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 153 /*
 154  *      Configured unicast address hash table
 155  */
 156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 157 static DEFINE_SPINLOCK(addrconf_hash_lock);
 158
 159 static void addrconf_verify(void);
 160 static void addrconf_verify_rtnl(void);
 161 static void addrconf_verify_work(struct work_struct *);
 162
 163 static struct workqueue_struct *addrconf_wq;
 164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 165
 166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 168
 169 static void addrconf_type_change(struct net_device *dev,
 170                                  unsigned long event);
 171 static int addrconf_ifdown(struct net_device *dev, int how);
 172
 173 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 174                                                   int plen,
 175                                                   const struct net_device *dev,
 176                                                   u32 flags, u32 noflags);
 177
 178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 179 static void addrconf_dad_work(struct work_struct *w);
 180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 181                                    bool send_na);
 182 static void addrconf_dad_run(struct inet6_dev *idev);
 183 static void addrconf_rs_timer(struct timer_list *t);
 184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 186
 187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 188                                 struct prefix_info *pinfo);
 189
 190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 191         .forwarding             = 0,
 192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 193         .mtu6                   = IPV6_MIN_MTU,
 194         .accept_ra              = 1,
 195         .accept_redirects       = 1,
 196         .autoconf               = 1,
 197         .force_mld_version      = 0,
 198         .mldv1_unsolicited_report_interval = 10 * HZ,
 199         .mldv2_unsolicited_report_interval = HZ,
 200         .dad_transmits          = 1,
 201         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 202         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 203         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205         .use_tempaddr           = 0,
 206         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 207         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 208         .regen_max_retry        = REGEN_MAX_RETRY,
 209         .max_desync_factor      = MAX_DESYNC_FACTOR,
 210         .max_addresses          = IPV6_MAX_ADDRESSES,
 211         .accept_ra_defrtr       = 1,
 212         .accept_ra_from_local   = 0,
 213         .accept_ra_min_hop_limit= 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_min_plen = 0,
 220         .accept_ra_rt_info_max_plen = 0,
 221 #endif
 222 #endif
 223         .proxy_ndp              = 0,
 224         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 225         .disable_ipv6           = 0,
 226         .accept_dad             = 0,
 227         .suppress_frag_ndisc    = 1,
 228         .accept_ra_mtu          = 1,
 229         .stable_secret          = {
 230                 .initialized = false,
 231         },
 232         .use_oif_addrs_only     = 0,
 233         .ignore_routes_with_linkdown = 0,
 234         .keep_addr_on_down      = 0,
 235         .seg6_enabled           = 0,
 236 #ifdef CONFIG_IPV6_SEG6_HMAC
 237         .seg6_require_hmac      = 0,
 238 #endif
 239         .enhanced_dad           = 1,
 240         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 241         .disable_policy         = 0,
 242 };
 243
 244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 245         .forwarding             = 0,
 246         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 247         .mtu6                   = IPV6_MIN_MTU,
 248         .accept_ra              = 1,
 249         .accept_redirects       = 1,
 250         .autoconf               = 1,
 251         .force_mld_version      = 0,
 252         .mldv1_unsolicited_report_interval = 10 * HZ,
 253         .mldv2_unsolicited_report_interval = HZ,
 254         .dad_transmits          = 1,
 255         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 256         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 257         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 258         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 259         .use_tempaddr           = 0,
 260         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 261         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 262         .regen_max_retry        = REGEN_MAX_RETRY,
 263         .max_desync_factor      = MAX_DESYNC_FACTOR,
 264         .max_addresses          = IPV6_MAX_ADDRESSES,
 265         .accept_ra_defrtr       = 1,
 266         .accept_ra_from_local   = 0,
 267         .accept_ra_min_hop_limit= 1,
 268         .accept_ra_pinfo        = 1,
 269 #ifdef CONFIG_IPV6_ROUTER_PREF
 270         .accept_ra_rtr_pref     = 1,
 271         .rtr_probe_interval     = 60 * HZ,
 272 #ifdef CONFIG_IPV6_ROUTE_INFO
 273         .accept_ra_rt_info_min_plen = 0,
 274         .accept_ra_rt_info_max_plen = 0,
 275 #endif
 276 #endif
 277         .proxy_ndp              = 0,
 278         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 279         .disable_ipv6           = 0,
 280         .accept_dad             = 1,
 281         .suppress_frag_ndisc    = 1,
 282         .accept_ra_mtu          = 1,
 283         .stable_secret          = {
 284                 .initialized = false,
 285         },
 286         .use_oif_addrs_only     = 0,
 287         .ignore_routes_with_linkdown = 0,
 288         .keep_addr_on_down      = 0,
 289         .seg6_enabled           = 0,
 290 #ifdef CONFIG_IPV6_SEG6_HMAC
 291         .seg6_require_hmac      = 0,
 292 #endif
 293         .enhanced_dad           = 1,
 294         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 295         .disable_policy         = 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         } else
 437                 ipv6_regen_rndid(ndev);
 438
 439         ndev->token = in6addr_any;
 440
 441         if (netif_running(dev) && addrconf_link_ready(dev))
 442                 ndev->if_flags |= IF_READY;
 443
 444         ipv6_mc_init_dev(ndev);
 445         ndev->tstamp = jiffies;
 446         err = addrconf_sysctl_register(ndev);
 447         if (err) {
 448                 ipv6_mc_destroy_dev(ndev);
 449                 snmp6_unregister_dev(ndev);
 450                 goto err_release;
 451         }
 452         /* protected by rtnl_lock */
 453         rcu_assign_pointer(dev->ip6_ptr, ndev);
 454
 455         /* Join interface-local all-node multicast group */
 456         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 457
 458         /* Join all-node multicast group */
 459         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 460
 461         /* Join all-router multicast group if forwarding is set */
 462         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 463                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 464
 465         return ndev;
 466
 467 err_release:
 468         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 469         ndev->dead = 1;
 470         in6_dev_finish_destroy(ndev);
 471         return ERR_PTR(err);
 472 }
 473
 474 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 475 {
 476         struct inet6_dev *idev;
 477
 478         ASSERT_RTNL();
 479
 480         idev = __in6_dev_get(dev);
 481         if (!idev) {
 482                 idev = ipv6_add_dev(dev);
 483                 if (IS_ERR(idev))
 484                         return NULL;
 485         }
 486
 487         if (dev->flags&IFF_UP)
 488                 ipv6_mc_up(idev);
 489         return idev;
 490 }
 491
 492 static int inet6_netconf_msgsize_devconf(int type)
 493 {
 494         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 495                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 496         bool all = false;
 497
 498         if (type == NETCONFA_ALL)
 499                 all = true;
 500
 501         if (all || type == NETCONFA_FORWARDING)
 502                 size += nla_total_size(4);
 503 #ifdef CONFIG_IPV6_MROUTE
 504         if (all || type == NETCONFA_MC_FORWARDING)
 505                 size += nla_total_size(4);
 506 #endif
 507         if (all || type == NETCONFA_PROXY_NEIGH)
 508                 size += nla_total_size(4);
 509
 510         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 511                 size += nla_total_size(4);
 512
 513         return size;
 514 }
 515
 516 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 517                                       struct ipv6_devconf *devconf, u32 portid,
 518                                       u32 seq, int event, unsigned int flags,
 519                                       int type)
 520 {
 521         struct nlmsghdr  *nlh;
 522         struct netconfmsg *ncm;
 523         bool all = false;
 524
 525         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 526                         flags);
 527         if (!nlh)
 528                 return -EMSGSIZE;
 529
 530         if (type == NETCONFA_ALL)
 531                 all = true;
 532
 533         ncm = nlmsg_data(nlh);
 534         ncm->ncm_family = AF_INET6;
 535
 536         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 537                 goto nla_put_failure;
 538
 539         if (!devconf)
 540                 goto out;
 541
 542         if ((all || type == NETCONFA_FORWARDING) &&
 543             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 544                 goto nla_put_failure;
 545 #ifdef CONFIG_IPV6_MROUTE
 546         if ((all || type == NETCONFA_MC_FORWARDING) &&
 547             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 548                         devconf->mc_forwarding) < 0)
 549                 goto nla_put_failure;
 550 #endif
 551         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 552             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 553                 goto nla_put_failure;
 554
 555         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 556             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 557                         devconf->ignore_routes_with_linkdown) < 0)
 558                 goto nla_put_failure;
 559
 560 out:
 561         nlmsg_end(skb, nlh);
 562         return 0;
 563
 564 nla_put_failure:
 565         nlmsg_cancel(skb, nlh);
 566         return -EMSGSIZE;
 567 }
 568
 569 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 570                                   int ifindex, struct ipv6_devconf *devconf)
 571 {
 572         struct sk_buff *skb;
 573         int err = -ENOBUFS;
 574
 575         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 576         if (!skb)
 577                 goto errout;
 578
 579         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 580                                          event, 0, type);
 581         if (err < 0) {
 582                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 583                 WARN_ON(err == -EMSGSIZE);
 584                 kfree_skb(skb);
 585                 goto errout;
 586         }
 587         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 588         return;
 589 errout:
 590         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 591 }
 592
 593 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 594         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 595         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 596         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 597         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 598 };
 599
 600 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 601                                      struct nlmsghdr *nlh,
 602                                      struct netlink_ext_ack *extack)
 603 {
 604         struct net *net = sock_net(in_skb->sk);
 605         struct nlattr *tb[NETCONFA_MAX+1];
 606         struct inet6_dev *in6_dev = NULL;
 607         struct net_device *dev = NULL;
 608         struct netconfmsg *ncm;
 609         struct sk_buff *skb;
 610         struct ipv6_devconf *devconf;
 611         int ifindex;
 612         int err;
 613
 614         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 615                           devconf_ipv6_policy, extack);
 616         if (err < 0)
 617                 return err;
 618
 619         if (!tb[NETCONFA_IFINDEX])
 620                 return -EINVAL;
 621
 622         err = -EINVAL;
 623         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 624         switch (ifindex) {
 625         case NETCONFA_IFINDEX_ALL:
 626                 devconf = net->ipv6.devconf_all;
 627                 break;
 628         case NETCONFA_IFINDEX_DEFAULT:
 629                 devconf = net->ipv6.devconf_dflt;
 630                 break;
 631         default:
 632                 dev = dev_get_by_index(net, ifindex);
 633                 if (!dev)
 634                         return -EINVAL;
 635                 in6_dev = in6_dev_get(dev);
 636                 if (!in6_dev)
 637                         goto errout;
 638                 devconf = &in6_dev->cnf;
 639                 break;
 640         }
 641
 642         err = -ENOBUFS;
 643         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 644         if (!skb)
 645                 goto errout;
 646
 647         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 648                                          NETLINK_CB(in_skb).portid,
 649                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 650                                          NETCONFA_ALL);
 651         if (err < 0) {
 652                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 653                 WARN_ON(err == -EMSGSIZE);
 654                 kfree_skb(skb);
 655                 goto errout;
 656         }
 657         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 658 errout:
 659         if (in6_dev)
 660                 in6_dev_put(in6_dev);
 661         if (dev)
 662                 dev_put(dev);
 663         return err;
 664 }
 665
 666 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 667                                       struct netlink_callback *cb)
 668 {
 669         const struct nlmsghdr *nlh = cb->nlh;
 670         struct net *net = sock_net(skb->sk);
 671         int h, s_h;
 672         int idx, s_idx;
 673         struct net_device *dev;
 674         struct inet6_dev *idev;
 675         struct hlist_head *head;
 676
 677         if (cb->strict_check) {
 678                 struct netlink_ext_ack *extack = cb->extack;
 679                 struct netconfmsg *ncm;
 680
 681                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 682                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 683                         return -EINVAL;
 684                 }
 685
 686                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 687                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 688                         return -EINVAL;
 689                 }
 690         }
 691
 692         s_h = cb->args[0];
 693         s_idx = idx = cb->args[1];
 694
 695         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 696                 idx = 0;
 697                 head = &net->dev_index_head[h];
 698                 rcu_read_lock();
 699                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 700                           net->dev_base_seq;
 701                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 702                         if (idx < s_idx)
 703                                 goto cont;
 704                         idev = __in6_dev_get(dev);
 705                         if (!idev)
 706                                 goto cont;
 707
 708                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 709                                                        &idev->cnf,
 710                                                        NETLINK_CB(cb->skb).portid,
 711                                                        nlh->nlmsg_seq,
 712                                                        RTM_NEWNETCONF,
 713                                                        NLM_F_MULTI,
 714                                                        NETCONFA_ALL) < 0) {
 715                                 rcu_read_unlock();
 716                                 goto done;
 717                         }
 718                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 719 cont:
 720                         idx++;
 721                 }
 722                 rcu_read_unlock();
 723         }
 724         if (h == NETDEV_HASHENTRIES) {
 725                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 726                                                net->ipv6.devconf_all,
 727                                                NETLINK_CB(cb->skb).portid,
 728                                                nlh->nlmsg_seq,
 729                                                RTM_NEWNETCONF, NLM_F_MULTI,
 730                                                NETCONFA_ALL) < 0)
 731                         goto done;
 732                 else
 733                         h++;
 734         }
 735         if (h == NETDEV_HASHENTRIES + 1) {
 736                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 737                                                net->ipv6.devconf_dflt,
 738                                                NETLINK_CB(cb->skb).portid,
 739                                                nlh->nlmsg_seq,
 740                                                RTM_NEWNETCONF, NLM_F_MULTI,
 741                                                NETCONFA_ALL) < 0)
 742                         goto done;
 743                 else
 744                         h++;
 745         }
 746 done:
 747         cb->args[0] = h;
 748         cb->args[1] = idx;
 749
 750         return skb->len;
 751 }
 752
 753 #ifdef CONFIG_SYSCTL
 754 static void dev_forward_change(struct inet6_dev *idev)
 755 {
 756         struct net_device *dev;
 757         struct inet6_ifaddr *ifa;
 758
 759         if (!idev)
 760                 return;
 761         dev = idev->dev;
 762         if (idev->cnf.forwarding)
 763                 dev_disable_lro(dev);
 764         if (dev->flags & IFF_MULTICAST) {
 765                 if (idev->cnf.forwarding) {
 766                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 767                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 768                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 769                 } else {
 770                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 771                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 772                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 773                 }
 774         }
 775
 776         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 777                 if (ifa->flags&IFA_F_TENTATIVE)
 778                         continue;
 779                 if (idev->cnf.forwarding)
 780                         addrconf_join_anycast(ifa);
 781                 else
 782                         addrconf_leave_anycast(ifa);
 783         }
 784         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 785                                      NETCONFA_FORWARDING,
 786                                      dev->ifindex, &idev->cnf);
 787 }
 788
 789
 790 static void addrconf_forward_change(struct net *net, __s32 newf)
 791 {
 792         struct net_device *dev;
 793         struct inet6_dev *idev;
 794
 795         for_each_netdev(net, dev) {
 796                 idev = __in6_dev_get(dev);
 797                 if (idev) {
 798                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 799                         idev->cnf.forwarding = newf;
 800                         if (changed)
 801                                 dev_forward_change(idev);
 802                 }
 803         }
 804 }
 805
 806 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 807 {
 808         struct net *net;
 809         int old;
 810
 811         if (!rtnl_trylock())
 812                 return restart_syscall();
 813
 814         net = (struct net *)table->extra2;
 815         old = *p;
 816         *p = newf;
 817
 818         if (p == &net->ipv6.devconf_dflt->forwarding) {
 819                 if ((!newf) ^ (!old))
 820                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 821                                                      NETCONFA_FORWARDING,
 822                                                      NETCONFA_IFINDEX_DEFAULT,
 823                                                      net->ipv6.devconf_dflt);
 824                 rtnl_unlock();
 825                 return 0;
 826         }
 827
 828         if (p == &net->ipv6.devconf_all->forwarding) {
 829                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 830
 831                 net->ipv6.devconf_dflt->forwarding = newf;
 832                 if ((!newf) ^ (!old_dflt))
 833                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 834                                                      NETCONFA_FORWARDING,
 835                                                      NETCONFA_IFINDEX_DEFAULT,
 836                                                      net->ipv6.devconf_dflt);
 837
 838                 addrconf_forward_change(net, newf);
 839                 if ((!newf) ^ (!old))
 840                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 841                                                      NETCONFA_FORWARDING,
 842                                                      NETCONFA_IFINDEX_ALL,
 843                                                      net->ipv6.devconf_all);
 844         } else if ((!newf) ^ (!old))
 845                 dev_forward_change((struct inet6_dev *)table->extra1);
 846         rtnl_unlock();
 847
 848         if (newf)
 849                 rt6_purge_dflt_routers(net);
 850         return 1;
 851 }
 852
 853 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 854 {
 855         struct net_device *dev;
 856         struct inet6_dev *idev;
 857
 858         for_each_netdev(net, dev) {
 859                 idev = __in6_dev_get(dev);
 860                 if (idev) {
 861                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 862
 863                         idev->cnf.ignore_routes_with_linkdown = newf;
 864                         if (changed)
 865                                 inet6_netconf_notify_devconf(dev_net(dev),
 866                                                              RTM_NEWNETCONF,
 867                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 868                                                              dev->ifindex,
 869                                                              &idev->cnf);
 870                 }
 871         }
 872 }
 873
 874 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 875 {
 876         struct net *net;
 877         int old;
 878
 879         if (!rtnl_trylock())
 880                 return restart_syscall();
 881
 882         net = (struct net *)table->extra2;
 883         old = *p;
 884         *p = newf;
 885
 886         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 887                 if ((!newf) ^ (!old))
 888                         inet6_netconf_notify_devconf(net,
 889                                                      RTM_NEWNETCONF,
 890                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 891                                                      NETCONFA_IFINDEX_DEFAULT,
 892                                                      net->ipv6.devconf_dflt);
 893                 rtnl_unlock();
 894                 return 0;
 895         }
 896
 897         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 898                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 899                 addrconf_linkdown_change(net, newf);
 900                 if ((!newf) ^ (!old))
 901                         inet6_netconf_notify_devconf(net,
 902                                                      RTM_NEWNETCONF,
 903                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 904                                                      NETCONFA_IFINDEX_ALL,
 905                                                      net->ipv6.devconf_all);
 906         }
 907         rtnl_unlock();
 908
 909         return 1;
 910 }
 911
 912 #endif
 913
 914 /* Nobody refers to this ifaddr, destroy it */
 915 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 916 {
 917         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 918
 919 #ifdef NET_REFCNT_DEBUG
 920         pr_debug("%s\n", __func__);
 921 #endif
 922
 923         in6_dev_put(ifp->idev);
 924
 925         if (cancel_delayed_work(&ifp->dad_work))
 926                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 927                           ifp);
 928
 929         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 930                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 931                 return;
 932         }
 933
 934         kfree_rcu(ifp, rcu);
 935 }
 936
 937 static void
 938 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 939 {
 940         struct list_head *p;
 941         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 942
 943         /*
 944          * Each device address list is sorted in order of scope -
 945          * global before linklocal.
 946          */
 947         list_for_each(p, &idev->addr_list) {
 948                 struct inet6_ifaddr *ifa
 949                         = list_entry(p, struct inet6_ifaddr, if_list);
 950                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 951                         break;
 952         }
 953
 954         list_add_tail_rcu(&ifp->if_list, p);
 955 }
 956
 957 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 958 {
 959         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 960
 961         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 962 }
 963
 964 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 965                                struct net_device *dev, unsigned int hash)
 966 {
 967         struct inet6_ifaddr *ifp;
 968
 969         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
 970                 if (!net_eq(dev_net(ifp->idev->dev), net))
 971                         continue;
 972                 if (ipv6_addr_equal(&ifp->addr, addr)) {
 973                         if (!dev || ifp->idev->dev == dev)
 974                                 return true;
 975                 }
 976         }
 977         return false;
 978 }
 979
 980 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
 981 {
 982         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
 983         int err = 0;
 984
 985         spin_lock(&addrconf_hash_lock);
 986
 987         /* Ignore adding duplicate addresses on an interface */
 988         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
 989                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
 990                 err = -EEXIST;
 991         } else {
 992                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 993         }
 994
 995         spin_unlock(&addrconf_hash_lock);
 996
 997         return err;
 998 }
 999
1000 /* On success it returns ifp with increased reference count */
1001
1002 static struct inet6_ifaddr *
1003 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1004               bool can_block, struct netlink_ext_ack *extack)
1005 {
1006         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1007         int addr_type = ipv6_addr_type(cfg->pfx);
1008         struct net *net = dev_net(idev->dev);
1009         struct inet6_ifaddr *ifa = NULL;
1010         struct fib6_info *f6i = NULL;
1011         int err = 0;
1012
1013         if (addr_type == IPV6_ADDR_ANY ||
1014             addr_type & IPV6_ADDR_MULTICAST ||
1015             (!(idev->dev->flags & IFF_LOOPBACK) &&
1016              !netif_is_l3_master(idev->dev) &&
1017              addr_type & IPV6_ADDR_LOOPBACK))
1018                 return ERR_PTR(-EADDRNOTAVAIL);
1019
1020         if (idev->dead) {
1021                 err = -ENODEV;                  /*XXX*/
1022                 goto out;
1023         }
1024
1025         if (idev->cnf.disable_ipv6) {
1026                 err = -EACCES;
1027                 goto out;
1028         }
1029
1030         /* validator notifier needs to be blocking;
1031          * do not call in atomic context
1032          */
1033         if (can_block) {
1034                 struct in6_validator_info i6vi = {
1035                         .i6vi_addr = *cfg->pfx,
1036                         .i6vi_dev = idev,
1037                         .extack = extack,
1038                 };
1039
1040                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1041                 err = notifier_to_errno(err);
1042                 if (err < 0)
1043                         goto out;
1044         }
1045
1046         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1047         if (!ifa) {
1048                 err = -ENOBUFS;
1049                 goto out;
1050         }
1051
1052         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1053         if (IS_ERR(f6i)) {
1054                 err = PTR_ERR(f6i);
1055                 f6i = NULL;
1056                 goto out;
1057         }
1058
1059         if (net->ipv6.devconf_all->disable_policy ||
1060             idev->cnf.disable_policy)
1061                 f6i->dst_nopolicy = true;
1062
1063         neigh_parms_data_state_setall(idev->nd_parms);
1064
1065         ifa->addr = *cfg->pfx;
1066         if (cfg->peer_pfx)
1067                 ifa->peer_addr = *cfg->peer_pfx;
1068
1069         spin_lock_init(&ifa->lock);
1070         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1071         INIT_HLIST_NODE(&ifa->addr_lst);
1072         ifa->scope = cfg->scope;
1073         ifa->prefix_len = cfg->plen;
1074         ifa->rt_priority = cfg->rt_priority;
1075         ifa->flags = cfg->ifa_flags;
1076         /* No need to add the TENTATIVE flag for addresses with NODAD */
1077         if (!(cfg->ifa_flags & IFA_F_NODAD))
1078                 ifa->flags |= IFA_F_TENTATIVE;
1079         ifa->valid_lft = cfg->valid_lft;
1080         ifa->prefered_lft = cfg->preferred_lft;
1081         ifa->cstamp = ifa->tstamp = jiffies;
1082         ifa->tokenized = false;
1083
1084         ifa->rt = f6i;
1085
1086         ifa->idev = idev;
1087         in6_dev_hold(idev);
1088
1089         /* For caller */
1090         refcount_set(&ifa->refcnt, 1);
1091
1092         rcu_read_lock_bh();
1093
1094         err = ipv6_add_addr_hash(idev->dev, ifa);
1095         if (err < 0) {
1096                 rcu_read_unlock_bh();
1097                 goto out;
1098         }
1099
1100         write_lock(&idev->lock);
1101
1102         /* Add to inet6_dev unicast addr list. */
1103         ipv6_link_dev_addr(idev, ifa);
1104
1105         if (ifa->flags&IFA_F_TEMPORARY) {
1106                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1107                 in6_ifa_hold(ifa);
1108         }
1109
1110         in6_ifa_hold(ifa);
1111         write_unlock(&idev->lock);
1112
1113         rcu_read_unlock_bh();
1114
1115         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1116 out:
1117         if (unlikely(err < 0)) {
1118                 fib6_info_release(f6i);
1119
1120                 if (ifa) {
1121                         if (ifa->idev)
1122                                 in6_dev_put(ifa->idev);
1123                         kfree(ifa);
1124                 }
1125                 ifa = ERR_PTR(err);
1126         }
1127
1128         return ifa;
1129 }
1130
1131 enum cleanup_prefix_rt_t {
1132         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1133         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1134         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1135 };
1136
1137 /*
1138  * Check, whether the prefix for ifp would still need a prefix route
1139  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1140  * constants.
1141  *
1142  * 1) we don't purge prefix if address was not permanent.
1143  *    prefix is managed by its own lifetime.
1144  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1145  * 3) if there are no addresses, delete prefix.
1146  * 4) if there are still other permanent address(es),
1147  *    corresponding prefix is still permanent.
1148  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1149  *    don't purge the prefix, assume user space is managing it.
1150  * 6) otherwise, update prefix lifetime to the
1151  *    longest valid lifetime among the corresponding
1152  *    addresses on the device.
1153  *    Note: subsequent RA will update lifetime.
1154  **/
1155 static enum cleanup_prefix_rt_t
1156 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1157 {
1158         struct inet6_ifaddr *ifa;
1159         struct inet6_dev *idev = ifp->idev;
1160         unsigned long lifetime;
1161         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1162
1163         *expires = jiffies;
1164
1165         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1166                 if (ifa == ifp)
1167                         continue;
1168                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1169                                        ifp->prefix_len))
1170                         continue;
1171                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1172                         return CLEANUP_PREFIX_RT_NOP;
1173
1174                 action = CLEANUP_PREFIX_RT_EXPIRE;
1175
1176                 spin_lock(&ifa->lock);
1177
1178                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1179                 /*
1180                  * Note: Because this address is
1181                  * not permanent, lifetime <
1182                  * LONG_MAX / HZ here.
1183                  */
1184                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1185                         *expires = ifa->tstamp + lifetime * HZ;
1186                 spin_unlock(&ifa->lock);
1187         }
1188
1189         return action;
1190 }
1191
1192 static void
1193 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1194 {
1195         struct fib6_info *f6i;
1196
1197         f6i = addrconf_get_prefix_route(&ifp->addr,
1198                                        ifp->prefix_len,
1199                                        ifp->idev->dev,
1200                                        0, RTF_GATEWAY | RTF_DEFAULT);
1201         if (f6i) {
1202                 if (del_rt)
1203                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1204                 else {
1205                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1206                                 fib6_set_expires(f6i, expires);
1207                         fib6_info_release(f6i);
1208                 }
1209         }
1210 }
1211
1212
1213 /* This function wants to get referenced ifp and releases it before return */
1214
1215 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1216 {
1217         int state;
1218         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1219         unsigned long expires;
1220
1221         ASSERT_RTNL();
1222
1223         spin_lock_bh(&ifp->lock);
1224         state = ifp->state;
1225         ifp->state = INET6_IFADDR_STATE_DEAD;
1226         spin_unlock_bh(&ifp->lock);
1227
1228         if (state == INET6_IFADDR_STATE_DEAD)
1229                 goto out;
1230
1231         spin_lock_bh(&addrconf_hash_lock);
1232         hlist_del_init_rcu(&ifp->addr_lst);
1233         spin_unlock_bh(&addrconf_hash_lock);
1234
1235         write_lock_bh(&ifp->idev->lock);
1236
1237         if (ifp->flags&IFA_F_TEMPORARY) {
1238                 list_del(&ifp->tmp_list);
1239                 if (ifp->ifpub) {
1240                         in6_ifa_put(ifp->ifpub);
1241                         ifp->ifpub = NULL;
1242                 }
1243                 __in6_ifa_put(ifp);
1244         }
1245
1246         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1247                 action = check_cleanup_prefix_route(ifp, &expires);
1248
1249         list_del_rcu(&ifp->if_list);
1250         __in6_ifa_put(ifp);
1251
1252         write_unlock_bh(&ifp->idev->lock);
1253
1254         addrconf_del_dad_work(ifp);
1255
1256         ipv6_ifa_notify(RTM_DELADDR, ifp);
1257
1258         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1259
1260         if (action != CLEANUP_PREFIX_RT_NOP) {
1261                 cleanup_prefix_route(ifp, expires,
1262                         action == CLEANUP_PREFIX_RT_DEL);
1263         }
1264
1265         /* clean up prefsrc entries */
1266         rt6_remove_prefsrc(ifp);
1267 out:
1268         in6_ifa_put(ifp);
1269 }
1270
1271 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1272                                 struct inet6_ifaddr *ift,
1273                                 bool block)
1274 {
1275         struct inet6_dev *idev = ifp->idev;
1276         struct in6_addr addr, *tmpaddr;
1277         unsigned long tmp_tstamp, age;
1278         unsigned long regen_advance;
1279         struct ifa6_config cfg;
1280         int ret = 0;
1281         unsigned long now = jiffies;
1282         long max_desync_factor;
1283         s32 cnf_temp_preferred_lft;
1284
1285         write_lock_bh(&idev->lock);
1286         if (ift) {
1287                 spin_lock_bh(&ift->lock);
1288                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1289                 spin_unlock_bh(&ift->lock);
1290                 tmpaddr = &addr;
1291         } else {
1292                 tmpaddr = NULL;
1293         }
1294 retry:
1295         in6_dev_hold(idev);
1296         if (idev->cnf.use_tempaddr <= 0) {
1297                 write_unlock_bh(&idev->lock);
1298                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1299                 in6_dev_put(idev);
1300                 ret = -1;
1301                 goto out;
1302         }
1303         spin_lock_bh(&ifp->lock);
1304         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1305                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1306                 spin_unlock_bh(&ifp->lock);
1307                 write_unlock_bh(&idev->lock);
1308                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1309                         __func__);
1310                 in6_dev_put(idev);
1311                 ret = -1;
1312                 goto out;
1313         }
1314         in6_ifa_hold(ifp);
1315         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1316         ipv6_try_regen_rndid(idev, tmpaddr);
1317         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1318         age = (now - ifp->tstamp) / HZ;
1319
1320         regen_advance = idev->cnf.regen_max_retry *
1321                         idev->cnf.dad_transmits *
1322                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1323
1324         /* recalculate max_desync_factor each time and update
1325          * idev->desync_factor if it's larger
1326          */
1327         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1328         max_desync_factor = min_t(__u32,
1329                                   idev->cnf.max_desync_factor,
1330                                   cnf_temp_preferred_lft - regen_advance);
1331
1332         if (unlikely(idev->desync_factor > max_desync_factor)) {
1333                 if (max_desync_factor > 0) {
1334                         get_random_bytes(&idev->desync_factor,
1335                                          sizeof(idev->desync_factor));
1336                         idev->desync_factor %= max_desync_factor;
1337                 } else {
1338                         idev->desync_factor = 0;
1339                 }
1340         }
1341
1342         memset(&cfg, 0, sizeof(cfg));
1343         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1344                               idev->cnf.temp_valid_lft + age);
1345         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1346         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1347
1348         cfg.plen = ifp->prefix_len;
1349         tmp_tstamp = ifp->tstamp;
1350         spin_unlock_bh(&ifp->lock);
1351
1352         write_unlock_bh(&idev->lock);
1353
1354         /* A temporary address is created only if this calculated Preferred
1355          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1356          * an implementation must not create a temporary address with a zero
1357          * Preferred Lifetime.
1358          * Use age calculation as in addrconf_verify to avoid unnecessary
1359          * temporary addresses being generated.
1360          */
1361         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1362         if (cfg.preferred_lft <= regen_advance + age) {
1363                 in6_ifa_put(ifp);
1364                 in6_dev_put(idev);
1365                 ret = -1;
1366                 goto out;
1367         }
1368
1369         cfg.ifa_flags = IFA_F_TEMPORARY;
1370         /* set in addrconf_prefix_rcv() */
1371         if (ifp->flags & IFA_F_OPTIMISTIC)
1372                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1373
1374         cfg.pfx = &addr;
1375         cfg.scope = ipv6_addr_scope(cfg.pfx);
1376
1377         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1378         if (IS_ERR(ift)) {
1379                 in6_ifa_put(ifp);
1380                 in6_dev_put(idev);
1381                 pr_info("%s: retry temporary address regeneration\n", __func__);
1382                 tmpaddr = &addr;
1383                 write_lock_bh(&idev->lock);
1384                 goto retry;
1385         }
1386
1387         spin_lock_bh(&ift->lock);
1388         ift->ifpub = ifp;
1389         ift->cstamp = now;
1390         ift->tstamp = tmp_tstamp;
1391         spin_unlock_bh(&ift->lock);
1392
1393         addrconf_dad_start(ift);
1394         in6_ifa_put(ift);
1395         in6_dev_put(idev);
1396 out:
1397         return ret;
1398 }
1399
1400 /*
1401  *      Choose an appropriate source address (RFC3484)
1402  */
1403 enum {
1404         IPV6_SADDR_RULE_INIT = 0,
1405         IPV6_SADDR_RULE_LOCAL,
1406         IPV6_SADDR_RULE_SCOPE,
1407         IPV6_SADDR_RULE_PREFERRED,
1408 #ifdef CONFIG_IPV6_MIP6
1409         IPV6_SADDR_RULE_HOA,
1410 #endif
1411         IPV6_SADDR_RULE_OIF,
1412         IPV6_SADDR_RULE_LABEL,
1413         IPV6_SADDR_RULE_PRIVACY,
1414         IPV6_SADDR_RULE_ORCHID,
1415         IPV6_SADDR_RULE_PREFIX,
1416 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1417         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1418 #endif
1419         IPV6_SADDR_RULE_MAX
1420 };
1421
1422 struct ipv6_saddr_score {
1423         int                     rule;
1424         int                     addr_type;
1425         struct inet6_ifaddr     *ifa;
1426         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1427         int                     scopedist;
1428         int                     matchlen;
1429 };
1430
1431 struct ipv6_saddr_dst {
1432         const struct in6_addr *addr;
1433         int ifindex;
1434         int scope;
1435         int label;
1436         unsigned int prefs;
1437 };
1438
1439 static inline int ipv6_saddr_preferred(int type)
1440 {
1441         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1442                 return 1;
1443         return 0;
1444 }
1445
1446 static bool ipv6_use_optimistic_addr(struct net *net,
1447                                      struct inet6_dev *idev)
1448 {
1449 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1450         if (!idev)
1451                 return false;
1452         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1453                 return false;
1454         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1455                 return false;
1456
1457         return true;
1458 #else
1459         return false;
1460 #endif
1461 }
1462
1463 static bool ipv6_allow_optimistic_dad(struct net *net,
1464                                       struct inet6_dev *idev)
1465 {
1466 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1467         if (!idev)
1468                 return false;
1469         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1470                 return false;
1471
1472         return true;
1473 #else
1474         return false;
1475 #endif
1476 }
1477
1478 static int ipv6_get_saddr_eval(struct net *net,
1479                                struct ipv6_saddr_score *score,
1480                                struct ipv6_saddr_dst *dst,
1481                                int i)
1482 {
1483         int ret;
1484
1485         if (i <= score->rule) {
1486                 switch (i) {
1487                 case IPV6_SADDR_RULE_SCOPE:
1488                         ret = score->scopedist;
1489                         break;
1490                 case IPV6_SADDR_RULE_PREFIX:
1491                         ret = score->matchlen;
1492                         break;
1493                 default:
1494                         ret = !!test_bit(i, score->scorebits);
1495                 }
1496                 goto out;
1497         }
1498
1499         switch (i) {
1500         case IPV6_SADDR_RULE_INIT:
1501                 /* Rule 0: remember if hiscore is not ready yet */
1502                 ret = !!score->ifa;
1503                 break;
1504         case IPV6_SADDR_RULE_LOCAL:
1505                 /* Rule 1: Prefer same address */
1506                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1507                 break;
1508         case IPV6_SADDR_RULE_SCOPE:
1509                 /* Rule 2: Prefer appropriate scope
1510                  *
1511                  *      ret
1512                  *       ^
1513                  *    -1 |  d 15
1514                  *    ---+--+-+---> scope
1515                  *       |
1516                  *       |             d is scope of the destination.
1517                  *  B-d  |  \
1518                  *       |   \      <- smaller scope is better if
1519                  *  B-15 |    \        if scope is enough for destination.
1520                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1521                  * d-C-1 | /
1522                  *       |/         <- greater is better
1523                  *   -C  /             if scope is not enough for destination.
1524                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1525                  *
1526                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1527                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1528                  * Assume B = 0 and we get C > 29.
1529                  */
1530                 ret = __ipv6_addr_src_scope(score->addr_type);
1531                 if (ret >= dst->scope)
1532                         ret = -ret;
1533                 else
1534                         ret -= 128;     /* 30 is enough */
1535                 score->scopedist = ret;
1536                 break;
1537         case IPV6_SADDR_RULE_PREFERRED:
1538             {
1539                 /* Rule 3: Avoid deprecated and optimistic addresses */
1540                 u8 avoid = IFA_F_DEPRECATED;
1541
1542                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1543                         avoid |= IFA_F_OPTIMISTIC;
1544                 ret = ipv6_saddr_preferred(score->addr_type) ||
1545                       !(score->ifa->flags & avoid);
1546                 break;
1547             }
1548 #ifdef CONFIG_IPV6_MIP6
1549         case IPV6_SADDR_RULE_HOA:
1550             {
1551                 /* Rule 4: Prefer home address */
1552                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1553                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1554                 break;
1555             }
1556 #endif
1557         case IPV6_SADDR_RULE_OIF:
1558                 /* Rule 5: Prefer outgoing interface */
1559                 ret = (!dst->ifindex ||
1560                        dst->ifindex == score->ifa->idev->dev->ifindex);
1561                 break;
1562         case IPV6_SADDR_RULE_LABEL:
1563                 /* Rule 6: Prefer matching label */
1564                 ret = ipv6_addr_label(net,
1565                                       &score->ifa->addr, score->addr_type,
1566                                       score->ifa->idev->dev->ifindex) == dst->label;
1567                 break;
1568         case IPV6_SADDR_RULE_PRIVACY:
1569             {
1570                 /* Rule 7: Prefer public address
1571                  * Note: prefer temporary address if use_tempaddr >= 2
1572                  */
1573                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1574                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1575                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1576                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1577                 break;
1578             }
1579         case IPV6_SADDR_RULE_ORCHID:
1580                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1581                  *          non-ORCHID vs non-ORCHID
1582                  */
1583                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1584                         ipv6_addr_orchid(dst->addr));
1585                 break;
1586         case IPV6_SADDR_RULE_PREFIX:
1587                 /* Rule 8: Use longest matching prefix */
1588                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1589                 if (ret > score->ifa->prefix_len)
1590                         ret = score->ifa->prefix_len;
1591                 score->matchlen = ret;
1592                 break;
1593 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1594         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1595                 /* Optimistic addresses still have lower precedence than other
1596                  * preferred addresses.
1597                  */
1598                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1599                 break;
1600 #endif
1601         default:
1602                 ret = 0;
1603         }
1604
1605         if (ret)
1606                 __set_bit(i, score->scorebits);
1607         score->rule = i;
1608 out:
1609         return ret;
1610 }
1611
1612 static int __ipv6_dev_get_saddr(struct net *net,
1613                                 struct ipv6_saddr_dst *dst,
1614                                 struct inet6_dev *idev,
1615                                 struct ipv6_saddr_score *scores,
1616                                 int hiscore_idx)
1617 {
1618         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1619
1620         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1621                 int i;
1622
1623                 /*
1624                  * - Tentative Address (RFC2462 section 5.4)
1625                  *  - A tentative address is not considered
1626                  *    "assigned to an interface" in the traditional
1627                  *    sense, unless it is also flagged as optimistic.
1628                  * - Candidate Source Address (section 4)
1629                  *  - In any case, anycast addresses, multicast
1630                  *    addresses, and the unspecified address MUST
1631                  *    NOT be included in a candidate set.
1632                  */
1633                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1634                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1635                         continue;
1636
1637                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1638
1639                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1640                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1641                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1642                                             idev->dev->name);
1643                         continue;
1644                 }
1645
1646                 score->rule = -1;
1647                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1648
1649                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1650                         int minihiscore, miniscore;
1651
1652                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1653                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1654
1655                         if (minihiscore > miniscore) {
1656                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1657                                     score->scopedist > 0) {
1658                                         /*
1659                                          * special case:
1660                                          * each remaining entry
1661                                          * has too small (not enough)
1662                                          * scope, because ifa entries
1663                                          * are sorted by their scope
1664                                          * values.
1665                                          */
1666                                         goto out;
1667                                 }
1668                                 break;
1669                         } else if (minihiscore < miniscore) {
1670                                 swap(hiscore, score);
1671                                 hiscore_idx = 1 - hiscore_idx;
1672
1673                                 /* restore our iterator */
1674                                 score->ifa = hiscore->ifa;
1675
1676                                 break;
1677                         }
1678                 }
1679         }
1680 out:
1681         return hiscore_idx;
1682 }
1683
1684 static int ipv6_get_saddr_master(struct net *net,
1685                                  const struct net_device *dst_dev,
1686                                  const struct net_device *master,
1687                                  struct ipv6_saddr_dst *dst,
1688                                  struct ipv6_saddr_score *scores,
1689                                  int hiscore_idx)
1690 {
1691         struct inet6_dev *idev;
1692
1693         idev = __in6_dev_get(dst_dev);
1694         if (idev)
1695                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1696                                                    scores, hiscore_idx);
1697
1698         idev = __in6_dev_get(master);
1699         if (idev)
1700                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1701                                                    scores, hiscore_idx);
1702
1703         return hiscore_idx;
1704 }
1705
1706 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1707                        const struct in6_addr *daddr, unsigned int prefs,
1708                        struct in6_addr *saddr)
1709 {
1710         struct ipv6_saddr_score scores[2], *hiscore;
1711         struct ipv6_saddr_dst dst;
1712         struct inet6_dev *idev;
1713         struct net_device *dev;
1714         int dst_type;
1715         bool use_oif_addr = false;
1716         int hiscore_idx = 0;
1717         int ret = 0;
1718
1719         dst_type = __ipv6_addr_type(daddr);
1720         dst.addr = daddr;
1721         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1722         dst.scope = __ipv6_addr_src_scope(dst_type);
1723         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1724         dst.prefs = prefs;
1725
1726         scores[hiscore_idx].rule = -1;
1727         scores[hiscore_idx].ifa = NULL;
1728
1729         rcu_read_lock();
1730
1731         /* Candidate Source Address (section 4)
1732          *  - multicast and link-local destination address,
1733          *    the set of candidate source address MUST only
1734          *    include addresses assigned to interfaces
1735          *    belonging to the same link as the outgoing
1736          *    interface.
1737          * (- For site-local destination addresses, the
1738          *    set of candidate source addresses MUST only
1739          *    include addresses assigned to interfaces
1740          *    belonging to the same site as the outgoing
1741          *    interface.)
1742          *  - "It is RECOMMENDED that the candidate source addresses
1743          *    be the set of unicast addresses assigned to the
1744          *    interface that will be used to send to the destination
1745          *    (the 'outgoing' interface)." (RFC 6724)
1746          */
1747         if (dst_dev) {
1748                 idev = __in6_dev_get(dst_dev);
1749                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1750                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1751                     (idev && idev->cnf.use_oif_addrs_only)) {
1752                         use_oif_addr = true;
1753                 }
1754         }
1755
1756         if (use_oif_addr) {
1757                 if (idev)
1758                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1759         } else {
1760                 const struct net_device *master;
1761                 int master_idx = 0;
1762
1763                 /* if dst_dev exists and is enslaved to an L3 device, then
1764                  * prefer addresses from dst_dev and then the master over
1765                  * any other enslaved devices in the L3 domain.
1766                  */
1767                 master = l3mdev_master_dev_rcu(dst_dev);
1768                 if (master) {
1769                         master_idx = master->ifindex;
1770
1771                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1772                                                             master, &dst,
1773                                                             scores, hiscore_idx);
1774
1775                         if (scores[hiscore_idx].ifa)
1776                                 goto out;
1777                 }
1778
1779                 for_each_netdev_rcu(net, dev) {
1780                         /* only consider addresses on devices in the
1781                          * same L3 domain
1782                          */
1783                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1784                                 continue;
1785                         idev = __in6_dev_get(dev);
1786                         if (!idev)
1787                                 continue;
1788                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1789                 }
1790         }
1791
1792 out:
1793         hiscore = &scores[hiscore_idx];
1794         if (!hiscore->ifa)
1795                 ret = -EADDRNOTAVAIL;
1796         else
1797                 *saddr = hiscore->ifa->addr;
1798
1799         rcu_read_unlock();
1800         return ret;
1801 }
1802 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1803
1804 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1805                       u32 banned_flags)
1806 {
1807         struct inet6_ifaddr *ifp;
1808         int err = -EADDRNOTAVAIL;
1809
1810         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1811                 if (ifp->scope > IFA_LINK)
1812                         break;
1813                 if (ifp->scope == IFA_LINK &&
1814                     !(ifp->flags & banned_flags)) {
1815                         *addr = ifp->addr;
1816                         err = 0;
1817                         break;
1818                 }
1819         }
1820         return err;
1821 }
1822
1823 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1824                     u32 banned_flags)
1825 {
1826         struct inet6_dev *idev;
1827         int err = -EADDRNOTAVAIL;
1828
1829         rcu_read_lock();
1830         idev = __in6_dev_get(dev);
1831         if (idev) {
1832                 read_lock_bh(&idev->lock);
1833                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1834                 read_unlock_bh(&idev->lock);
1835         }
1836         rcu_read_unlock();
1837         return err;
1838 }
1839
1840 static int ipv6_count_addresses(const struct inet6_dev *idev)
1841 {
1842         const struct inet6_ifaddr *ifp;
1843         int cnt = 0;
1844
1845         rcu_read_lock();
1846         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1847                 cnt++;
1848         rcu_read_unlock();
1849         return cnt;
1850 }
1851
1852 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1853                   const struct net_device *dev, int strict)
1854 {
1855         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1856                                        strict, IFA_F_TENTATIVE);
1857 }
1858 EXPORT_SYMBOL(ipv6_chk_addr);
1859
1860 /* device argument is used to find the L3 domain of interest. If
1861  * skip_dev_check is set, then the ifp device is not checked against
1862  * the passed in dev argument. So the 2 cases for addresses checks are:
1863  *   1. does the address exist in the L3 domain that dev is part of
1864  *      (skip_dev_check = true), or
1865  *
1866  *   2. does the address exist on the specific device
1867  *      (skip_dev_check = false)
1868  */
1869 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1870                             const struct net_device *dev, bool skip_dev_check,
1871                             int strict, u32 banned_flags)
1872 {
1873         unsigned int hash = inet6_addr_hash(net, addr);
1874         const struct net_device *l3mdev;
1875         struct inet6_ifaddr *ifp;
1876         u32 ifp_flags;
1877
1878         rcu_read_lock();
1879
1880         l3mdev = l3mdev_master_dev_rcu(dev);
1881         if (skip_dev_check)
1882                 dev = NULL;
1883
1884         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1885                 if (!net_eq(dev_net(ifp->idev->dev), net))
1886                         continue;
1887
1888                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1889                         continue;
1890
1891                 /* Decouple optimistic from tentative for evaluation here.
1892                  * Ban optimistic addresses explicitly, when required.
1893                  */
1894                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1895                             ? (ifp->flags&~IFA_F_TENTATIVE)
1896                             : ifp->flags;
1897                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1898                     !(ifp_flags&banned_flags) &&
1899                     (!dev || ifp->idev->dev == dev ||
1900                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1901                         rcu_read_unlock();
1902                         return 1;
1903                 }
1904         }
1905
1906         rcu_read_unlock();
1907         return 0;
1908 }
1909 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1910
1911
1912 /* Compares an address/prefix_len with addresses on device @dev.
1913  * If one is found it returns true.
1914  */
1915 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1916         const unsigned int prefix_len, struct net_device *dev)
1917 {
1918         const struct inet6_ifaddr *ifa;
1919         const struct inet6_dev *idev;
1920         bool ret = false;
1921
1922         rcu_read_lock();
1923         idev = __in6_dev_get(dev);
1924         if (idev) {
1925                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1926                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1927                         if (ret)
1928                                 break;
1929                 }
1930         }
1931         rcu_read_unlock();
1932
1933         return ret;
1934 }
1935 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1936
1937 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1938 {
1939         const struct inet6_ifaddr *ifa;
1940         const struct inet6_dev *idev;
1941         int     onlink;
1942
1943         onlink = 0;
1944         rcu_read_lock();
1945         idev = __in6_dev_get(dev);
1946         if (idev) {
1947                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1948                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1949                                                    ifa->prefix_len);
1950                         if (onlink)
1951                                 break;
1952                 }
1953         }
1954         rcu_read_unlock();
1955         return onlink;
1956 }
1957 EXPORT_SYMBOL(ipv6_chk_prefix);
1958
1959 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1960                                      struct net_device *dev, int strict)
1961 {
1962         unsigned int hash = inet6_addr_hash(net, addr);
1963         struct inet6_ifaddr *ifp, *result = NULL;
1964
1965         rcu_read_lock();
1966         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1967                 if (!net_eq(dev_net(ifp->idev->dev), net))
1968                         continue;
1969                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1970                         if (!dev || ifp->idev->dev == dev ||
1971                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1972                                 result = ifp;
1973                                 in6_ifa_hold(ifp);
1974                                 break;
1975                         }
1976                 }
1977         }
1978         rcu_read_unlock();
1979
1980         return result;
1981 }
1982
1983 /* Gets referenced address, destroys ifaddr */
1984
1985 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1986 {
1987         if (dad_failed)
1988                 ifp->flags |= IFA_F_DADFAILED;
1989
1990         if (ifp->flags&IFA_F_TEMPORARY) {
1991                 struct inet6_ifaddr *ifpub;
1992                 spin_lock_bh(&ifp->lock);
1993                 ifpub = ifp->ifpub;
1994                 if (ifpub) {
1995                         in6_ifa_hold(ifpub);
1996                         spin_unlock_bh(&ifp->lock);
1997                         ipv6_create_tempaddr(ifpub, ifp, true);
1998                         in6_ifa_put(ifpub);
1999                 } else {
2000                         spin_unlock_bh(&ifp->lock);
2001                 }
2002                 ipv6_del_addr(ifp);
2003         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2004                 spin_lock_bh(&ifp->lock);
2005                 addrconf_del_dad_work(ifp);
2006                 ifp->flags |= IFA_F_TENTATIVE;
2007                 if (dad_failed)
2008                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2009                 spin_unlock_bh(&ifp->lock);
2010                 if (dad_failed)
2011                         ipv6_ifa_notify(0, ifp);
2012                 in6_ifa_put(ifp);
2013         } else {
2014                 ipv6_del_addr(ifp);
2015         }
2016 }
2017
2018 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2019 {
2020         int err = -ENOENT;
2021
2022         spin_lock_bh(&ifp->lock);
2023         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2024                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2025                 err = 0;
2026         }
2027         spin_unlock_bh(&ifp->lock);
2028
2029         return err;
2030 }
2031
2032 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2033 {
2034         struct inet6_dev *idev = ifp->idev;
2035         struct net *net = dev_net(ifp->idev->dev);
2036
2037         if (addrconf_dad_end(ifp)) {
2038                 in6_ifa_put(ifp);
2039                 return;
2040         }
2041
2042         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2043                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2044
2045         spin_lock_bh(&ifp->lock);
2046
2047         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2048                 struct in6_addr new_addr;
2049                 struct inet6_ifaddr *ifp2;
2050                 int retries = ifp->stable_privacy_retry + 1;
2051                 struct ifa6_config cfg = {
2052                         .pfx = &new_addr,
2053                         .plen = ifp->prefix_len,
2054                         .ifa_flags = ifp->flags,
2055                         .valid_lft = ifp->valid_lft,
2056                         .preferred_lft = ifp->prefered_lft,
2057                         .scope = ifp->scope,
2058                 };
2059
2060                 if (retries > net->ipv6.sysctl.idgen_retries) {
2061                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2062                                              ifp->idev->dev->name);
2063                         goto errdad;
2064                 }
2065
2066                 new_addr = ifp->addr;
2067                 if (ipv6_generate_stable_address(&new_addr, retries,
2068                                                  idev))
2069                         goto errdad;
2070
2071                 spin_unlock_bh(&ifp->lock);
2072
2073                 if (idev->cnf.max_addresses &&
2074                     ipv6_count_addresses(idev) >=
2075                     idev->cnf.max_addresses)
2076                         goto lock_errdad;
2077
2078                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2079                                      ifp->idev->dev->name);
2080
2081                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2082                 if (IS_ERR(ifp2))
2083                         goto lock_errdad;
2084
2085                 spin_lock_bh(&ifp2->lock);
2086                 ifp2->stable_privacy_retry = retries;
2087                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2088                 spin_unlock_bh(&ifp2->lock);
2089
2090                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2091                 in6_ifa_put(ifp2);
2092 lock_errdad:
2093                 spin_lock_bh(&ifp->lock);
2094         }
2095
2096 errdad:
2097         /* transition from _POSTDAD to _ERRDAD */
2098         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2099         spin_unlock_bh(&ifp->lock);
2100
2101         addrconf_mod_dad_work(ifp, 0);
2102         in6_ifa_put(ifp);
2103 }
2104
2105 /* Join to solicited addr multicast group.
2106  * caller must hold RTNL */
2107 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2108 {
2109         struct in6_addr maddr;
2110
2111         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2112                 return;
2113
2114         addrconf_addr_solict_mult(addr, &maddr);
2115         ipv6_dev_mc_inc(dev, &maddr);
2116 }
2117
2118 /* caller must hold RTNL */
2119 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2120 {
2121         struct in6_addr maddr;
2122
2123         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2124                 return;
2125
2126         addrconf_addr_solict_mult(addr, &maddr);
2127         __ipv6_dev_mc_dec(idev, &maddr);
2128 }
2129
2130 /* caller must hold RTNL */
2131 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2132 {
2133         struct in6_addr addr;
2134
2135         if (ifp->prefix_len >= 127) /* RFC 6164 */
2136                 return;
2137         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2138         if (ipv6_addr_any(&addr))
2139                 return;
2140         __ipv6_dev_ac_inc(ifp->idev, &addr);
2141 }
2142
2143 /* caller must hold RTNL */
2144 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2145 {
2146         struct in6_addr addr;
2147
2148         if (ifp->prefix_len >= 127) /* RFC 6164 */
2149                 return;
2150         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2151         if (ipv6_addr_any(&addr))
2152                 return;
2153         __ipv6_dev_ac_dec(ifp->idev, &addr);
2154 }
2155
2156 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2157 {
2158         switch (dev->addr_len) {
2159         case ETH_ALEN:
2160                 memcpy(eui, dev->dev_addr, 3);
2161                 eui[3] = 0xFF;
2162                 eui[4] = 0xFE;
2163                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2164                 break;
2165         case EUI64_ADDR_LEN:
2166                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2167                 eui[0] ^= 2;
2168                 break;
2169         default:
2170                 return -1;
2171         }
2172
2173         return 0;
2174 }
2175
2176 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2177 {
2178         union fwnet_hwaddr *ha;
2179
2180         if (dev->addr_len != FWNET_ALEN)
2181                 return -1;
2182
2183         ha = (union fwnet_hwaddr *)dev->dev_addr;
2184
2185         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2186         eui[0] ^= 2;
2187         return 0;
2188 }
2189
2190 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2191 {
2192         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2193         if (dev->addr_len != ARCNET_ALEN)
2194                 return -1;
2195         memset(eui, 0, 7);
2196         eui[7] = *(u8 *)dev->dev_addr;
2197         return 0;
2198 }
2199
2200 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2201 {
2202         if (dev->addr_len != INFINIBAND_ALEN)
2203                 return -1;
2204         memcpy(eui, dev->dev_addr + 12, 8);
2205         eui[0] |= 2;
2206         return 0;
2207 }
2208
2209 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2210 {
2211         if (addr == 0)
2212                 return -1;
2213         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2214                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2215                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2216                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2217                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2218                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2219         eui[1] = 0;
2220         eui[2] = 0x5E;
2221         eui[3] = 0xFE;
2222         memcpy(eui + 4, &addr, 4);
2223         return 0;
2224 }
2225
2226 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2227 {
2228         if (dev->priv_flags & IFF_ISATAP)
2229                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2230         return -1;
2231 }
2232
2233 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2234 {
2235         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2236 }
2237
2238 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2239 {
2240         memcpy(eui, dev->perm_addr, 3);
2241         memcpy(eui + 5, dev->perm_addr + 3, 3);
2242         eui[3] = 0xFF;
2243         eui[4] = 0xFE;
2244         eui[0] ^= 2;
2245         return 0;
2246 }
2247
2248 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2249 {
2250         switch (dev->type) {
2251         case ARPHRD_ETHER:
2252         case ARPHRD_FDDI:
2253                 return addrconf_ifid_eui48(eui, dev);
2254         case ARPHRD_ARCNET:
2255                 return addrconf_ifid_arcnet(eui, dev);
2256         case ARPHRD_INFINIBAND:
2257                 return addrconf_ifid_infiniband(eui, dev);
2258         case ARPHRD_SIT:
2259                 return addrconf_ifid_sit(eui, dev);
2260         case ARPHRD_IPGRE:
2261         case ARPHRD_TUNNEL:
2262                 return addrconf_ifid_gre(eui, dev);
2263         case ARPHRD_6LOWPAN:
2264                 return addrconf_ifid_6lowpan(eui, dev);
2265         case ARPHRD_IEEE1394:
2266                 return addrconf_ifid_ieee1394(eui, dev);
2267         case ARPHRD_TUNNEL6:
2268         case ARPHRD_IP6GRE:
2269         case ARPHRD_RAWIP:
2270                 return addrconf_ifid_ip6tnl(eui, dev);
2271         }
2272         return -1;
2273 }
2274
2275 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2276 {
2277         int err = -1;
2278         struct inet6_ifaddr *ifp;
2279
2280         read_lock_bh(&idev->lock);
2281         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2282                 if (ifp->scope > IFA_LINK)
2283                         break;
2284                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2285                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2286                         err = 0;
2287                         break;
2288                 }
2289         }
2290         read_unlock_bh(&idev->lock);
2291         return err;
2292 }
2293
2294 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2295 static void ipv6_regen_rndid(struct inet6_dev *idev)
2296 {
2297 regen:
2298         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2299         idev->rndid[0] &= ~0x02;
2300
2301         /*
2302          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2303          * check if generated address is not inappropriate
2304          *
2305          *  - Reserved subnet anycast (RFC 2526)
2306          *      11111101 11....11 1xxxxxxx
2307          *  - ISATAP (RFC4214) 6.1
2308          *      00-00-5E-FE-xx-xx-xx-xx
2309          *  - value 0
2310          *  - XXX: already assigned to an address on the device
2311          */
2312         if (idev->rndid[0] == 0xfd &&
2313             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2314             (idev->rndid[7]&0x80))
2315                 goto regen;
2316         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2317                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2318                         goto regen;
2319                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2320                         goto regen;
2321         }
2322 }
2323
2324 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2325 {
2326         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2327                 ipv6_regen_rndid(idev);
2328 }
2329
2330 /*
2331  *      Add prefix route.
2332  */
2333
2334 static void
2335 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2336                       struct net_device *dev, unsigned long expires,
2337                       u32 flags, gfp_t gfp_flags)
2338 {
2339         struct fib6_config cfg = {
2340                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2341                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2342                 .fc_ifindex = dev->ifindex,
2343                 .fc_expires = expires,
2344                 .fc_dst_len = plen,
2345                 .fc_flags = RTF_UP | flags,
2346                 .fc_nlinfo.nl_net = dev_net(dev),
2347                 .fc_protocol = RTPROT_KERNEL,
2348                 .fc_type = RTN_UNICAST,
2349         };
2350
2351         cfg.fc_dst = *pfx;
2352
2353         /* Prevent useless cloning on PtP SIT.
2354            This thing is done here expecting that the whole
2355            class of non-broadcast devices need not cloning.
2356          */
2357 #if IS_ENABLED(CONFIG_IPV6_SIT)
2358         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2359                 cfg.fc_flags |= RTF_NONEXTHOP;
2360 #endif
2361
2362         ip6_route_add(&cfg, gfp_flags, NULL);
2363 }
2364
2365
2366 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2367                                                   int plen,
2368                                                   const struct net_device *dev,
2369                                                   u32 flags, u32 noflags)
2370 {
2371         struct fib6_node *fn;
2372         struct fib6_info *rt = NULL;
2373         struct fib6_table *table;
2374         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2375
2376         table = fib6_get_table(dev_net(dev), tb_id);
2377         if (!table)
2378                 return NULL;
2379
2380         rcu_read_lock();
2381         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2382         if (!fn)
2383                 goto out;
2384
2385         for_each_fib6_node_rt_rcu(fn) {
2386                 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2387                         continue;
2388                 if ((rt->fib6_flags & flags) != flags)
2389                         continue;
2390                 if ((rt->fib6_flags & noflags) != 0)
2391                         continue;
2392                 if (!fib6_info_hold_safe(rt))
2393                         continue;
2394                 break;
2395         }
2396 out:
2397         rcu_read_unlock();
2398         return rt;
2399 }
2400
2401
2402 /* Create "default" multicast route to the interface */
2403
2404 static void addrconf_add_mroute(struct net_device *dev)
2405 {
2406         struct fib6_config cfg = {
2407                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2408                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2409                 .fc_ifindex = dev->ifindex,
2410                 .fc_dst_len = 8,
2411                 .fc_flags = RTF_UP,
2412                 .fc_type = RTN_UNICAST,
2413                 .fc_nlinfo.nl_net = dev_net(dev),
2414         };
2415
2416         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2417
2418         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2419 }
2420
2421 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2422 {
2423         struct inet6_dev *idev;
2424
2425         ASSERT_RTNL();
2426
2427         idev = ipv6_find_idev(dev);
2428         if (!idev)
2429                 return ERR_PTR(-ENOBUFS);
2430
2431         if (idev->cnf.disable_ipv6)
2432                 return ERR_PTR(-EACCES);
2433
2434         /* Add default multicast route */
2435         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2436                 addrconf_add_mroute(dev);
2437
2438         return idev;
2439 }
2440
2441 static void manage_tempaddrs(struct inet6_dev *idev,
2442                              struct inet6_ifaddr *ifp,
2443                              __u32 valid_lft, __u32 prefered_lft,
2444                              bool create, unsigned long now)
2445 {
2446         u32 flags;
2447         struct inet6_ifaddr *ift;
2448
2449         read_lock_bh(&idev->lock);
2450         /* update all temporary addresses in the list */
2451         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2452                 int age, max_valid, max_prefered;
2453
2454                 if (ifp != ift->ifpub)
2455                         continue;
2456
2457                 /* RFC 4941 section 3.3:
2458                  * If a received option will extend the lifetime of a public
2459                  * address, the lifetimes of temporary addresses should
2460                  * be extended, subject to the overall constraint that no
2461                  * temporary addresses should ever remain "valid" or "preferred"
2462                  * for a time longer than (TEMP_VALID_LIFETIME) or
2463                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2464                  */
2465                 age = (now - ift->cstamp) / HZ;
2466                 max_valid = idev->cnf.temp_valid_lft - age;
2467                 if (max_valid < 0)
2468                         max_valid = 0;
2469
2470                 max_prefered = idev->cnf.temp_prefered_lft -
2471                                idev->desync_factor - age;
2472                 if (max_prefered < 0)
2473                         max_prefered = 0;
2474
2475                 if (valid_lft > max_valid)
2476                         valid_lft = max_valid;
2477
2478                 if (prefered_lft > max_prefered)
2479                         prefered_lft = max_prefered;
2480
2481                 spin_lock(&ift->lock);
2482                 flags = ift->flags;
2483                 ift->valid_lft = valid_lft;
2484                 ift->prefered_lft = prefered_lft;
2485                 ift->tstamp = now;
2486                 if (prefered_lft > 0)
2487                         ift->flags &= ~IFA_F_DEPRECATED;
2488
2489                 spin_unlock(&ift->lock);
2490                 if (!(flags&IFA_F_TENTATIVE))
2491                         ipv6_ifa_notify(0, ift);
2492         }
2493
2494         if ((create || list_empty(&idev->tempaddr_list)) &&
2495             idev->cnf.use_tempaddr > 0) {
2496                 /* When a new public address is created as described
2497                  * in [ADDRCONF], also create a new temporary address.
2498                  * Also create a temporary address if it's enabled but
2499                  * no temporary address currently exists.
2500                  */
2501                 read_unlock_bh(&idev->lock);
2502                 ipv6_create_tempaddr(ifp, NULL, false);
2503         } else {
2504                 read_unlock_bh(&idev->lock);
2505         }
2506 }
2507
2508 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2509 {
2510         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2511                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2512 }
2513
2514 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2515                                  const struct prefix_info *pinfo,
2516                                  struct inet6_dev *in6_dev,
2517                                  const struct in6_addr *addr, int addr_type,
2518                                  u32 addr_flags, bool sllao, bool tokenized,
2519                                  __u32 valid_lft, u32 prefered_lft)
2520 {
2521         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2522         int create = 0, update_lft = 0;
2523
2524         if (!ifp && valid_lft) {
2525                 int max_addresses = in6_dev->cnf.max_addresses;
2526                 struct ifa6_config cfg = {
2527                         .pfx = addr,
2528                         .plen = pinfo->prefix_len,
2529                         .ifa_flags = addr_flags,
2530                         .valid_lft = valid_lft,
2531                         .preferred_lft = prefered_lft,
2532                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2533                 };
2534
2535 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2536                 if ((net->ipv6.devconf_all->optimistic_dad ||
2537                      in6_dev->cnf.optimistic_dad) &&
2538                     !net->ipv6.devconf_all->forwarding && sllao)
2539                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2540 #endif
2541
2542                 /* Do not allow to create too much of autoconfigured
2543                  * addresses; this would be too easy way to crash kernel.
2544                  */
2545                 if (!max_addresses ||
2546                     ipv6_count_addresses(in6_dev) < max_addresses)
2547                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2548
2549                 if (IS_ERR_OR_NULL(ifp))
2550                         return -1;
2551
2552                 create = 1;
2553                 spin_lock_bh(&ifp->lock);
2554                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2555                 ifp->cstamp = jiffies;
2556                 ifp->tokenized = tokenized;
2557                 spin_unlock_bh(&ifp->lock);
2558                 addrconf_dad_start(ifp);
2559         }
2560
2561         if (ifp) {
2562                 u32 flags;
2563                 unsigned long now;
2564                 u32 stored_lft;
2565
2566                 /* update lifetime (RFC2462 5.5.3 e) */
2567                 spin_lock_bh(&ifp->lock);
2568                 now = jiffies;
2569                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2570                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2571                 else
2572                         stored_lft = 0;
2573                 if (!create && stored_lft) {
2574                         const u32 minimum_lft = min_t(u32,
2575                                 stored_lft, MIN_VALID_LIFETIME);
2576                         valid_lft = max(valid_lft, minimum_lft);
2577
2578                         /* RFC4862 Section 5.5.3e:
2579                          * "Note that the preferred lifetime of the
2580                          *  corresponding address is always reset to
2581                          *  the Preferred Lifetime in the received
2582                          *  Prefix Information option, regardless of
2583                          *  whether the valid lifetime is also reset or
2584                          *  ignored."
2585                          *
2586                          * So we should always update prefered_lft here.
2587                          */
2588                         update_lft = 1;
2589                 }
2590
2591                 if (update_lft) {
2592                         ifp->valid_lft = valid_lft;
2593                         ifp->prefered_lft = prefered_lft;
2594                         ifp->tstamp = now;
2595                         flags = ifp->flags;
2596                         ifp->flags &= ~IFA_F_DEPRECATED;
2597                         spin_unlock_bh(&ifp->lock);
2598
2599                         if (!(flags&IFA_F_TENTATIVE))
2600                                 ipv6_ifa_notify(0, ifp);
2601                 } else
2602                         spin_unlock_bh(&ifp->lock);
2603
2604                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2605                                  create, now);
2606
2607                 in6_ifa_put(ifp);
2608                 addrconf_verify();
2609         }
2610
2611         return 0;
2612 }
2613 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2614
2615 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2616 {
2617         struct prefix_info *pinfo;
2618         __u32 valid_lft;
2619         __u32 prefered_lft;
2620         int addr_type, err;
2621         u32 addr_flags = 0;
2622         struct inet6_dev *in6_dev;
2623         struct net *net = dev_net(dev);
2624
2625         pinfo = (struct prefix_info *) opt;
2626
2627         if (len < sizeof(struct prefix_info)) {
2628                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2629                 return;
2630         }
2631
2632         /*
2633          *      Validation checks ([ADDRCONF], page 19)
2634          */
2635
2636         addr_type = ipv6_addr_type(&pinfo->prefix);
2637
2638         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2639                 return;
2640
2641         valid_lft = ntohl(pinfo->valid);
2642         prefered_lft = ntohl(pinfo->prefered);
2643
2644         if (prefered_lft > valid_lft) {
2645                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2646                 return;
2647         }
2648
2649         in6_dev = in6_dev_get(dev);
2650
2651         if (!in6_dev) {
2652                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2653                                     dev->name);
2654                 return;
2655         }
2656
2657         /*
2658          *      Two things going on here:
2659          *      1) Add routes for on-link prefixes
2660          *      2) Configure prefixes with the auto flag set
2661          */
2662
2663         if (pinfo->onlink) {
2664                 struct fib6_info *rt;
2665                 unsigned long rt_expires;
2666
2667                 /* Avoid arithmetic overflow. Really, we could
2668                  * save rt_expires in seconds, likely valid_lft,
2669                  * but it would require division in fib gc, that it
2670                  * not good.
2671                  */
2672                 if (HZ > USER_HZ)
2673                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2674                 else
2675                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2676
2677                 if (addrconf_finite_timeout(rt_expires))
2678                         rt_expires *= HZ;
2679
2680                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2681                                                pinfo->prefix_len,
2682                                                dev,
2683                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2684                                                RTF_GATEWAY | RTF_DEFAULT);
2685
2686                 if (rt) {
2687                         /* Autoconf prefix route */
2688                         if (valid_lft == 0) {
2689                                 ip6_del_rt(net, rt);
2690                                 rt = NULL;
2691                         } else if (addrconf_finite_timeout(rt_expires)) {
2692                                 /* not infinity */
2693                                 fib6_set_expires(rt, jiffies + rt_expires);
2694                         } else {
2695                                 fib6_clean_expires(rt);
2696                         }
2697                 } else if (valid_lft) {
2698                         clock_t expires = 0;
2699                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2700                         if (addrconf_finite_timeout(rt_expires)) {
2701                                 /* not infinity */
2702                                 flags |= RTF_EXPIRES;
2703                                 expires = jiffies_to_clock_t(rt_expires);
2704                         }
2705                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2706                                               0, dev, expires, flags,
2707                                               GFP_ATOMIC);
2708                 }
2709                 fib6_info_release(rt);
2710         }
2711
2712         /* Try to figure out our local address for this prefix */
2713
2714         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2715                 struct in6_addr addr;
2716                 bool tokenized = false, dev_addr_generated = false;
2717
2718                 if (pinfo->prefix_len == 64) {
2719                         memcpy(&addr, &pinfo->prefix, 8);
2720
2721                         if (!ipv6_addr_any(&in6_dev->token)) {
2722                                 read_lock_bh(&in6_dev->lock);
2723                                 memcpy(addr.s6_addr + 8,
2724                                        in6_dev->token.s6_addr + 8, 8);
2725                                 read_unlock_bh(&in6_dev->lock);
2726                                 tokenized = true;
2727                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2728                                    !ipv6_generate_stable_address(&addr, 0,
2729                                                                  in6_dev)) {
2730                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2731                                 goto ok;
2732                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2733                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2734                                 goto put;
2735                         } else {
2736                                 dev_addr_generated = true;
2737                         }
2738                         goto ok;
2739                 }
2740                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2741                                     pinfo->prefix_len);
2742                 goto put;
2743
2744 ok:
2745                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2746                                                    &addr, addr_type,
2747                                                    addr_flags, sllao,
2748                                                    tokenized, valid_lft,
2749                                                    prefered_lft);
2750                 if (err)
2751                         goto put;
2752
2753                 /* Ignore error case here because previous prefix add addr was
2754                  * successful which will be notified.
2755                  */
2756                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2757                                               addr_type, addr_flags, sllao,
2758                                               tokenized, valid_lft,
2759                                               prefered_lft,
2760                                               dev_addr_generated);
2761         }
2762         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2763 put:
2764         in6_dev_put(in6_dev);
2765 }
2766
2767 /*
2768  *      Set destination address.
2769  *      Special case for SIT interfaces where we create a new "virtual"
2770  *      device.
2771  */
2772 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2773 {
2774         struct in6_ifreq ireq;
2775         struct net_device *dev;
2776         int err = -EINVAL;
2777
2778         rtnl_lock();
2779
2780         err = -EFAULT;
2781         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2782                 goto err_exit;
2783
2784         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2785
2786         err = -ENODEV;
2787         if (!dev)
2788                 goto err_exit;
2789
2790 #if IS_ENABLED(CONFIG_IPV6_SIT)
2791         if (dev->type == ARPHRD_SIT) {
2792                 const struct net_device_ops *ops = dev->netdev_ops;
2793                 struct ifreq ifr;
2794                 struct ip_tunnel_parm p;
2795
2796                 err = -EADDRNOTAVAIL;
2797                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2798                         goto err_exit;
2799
2800                 memset(&p, 0, sizeof(p));
2801                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2802                 p.iph.saddr = 0;
2803                 p.iph.version = 4;
2804                 p.iph.ihl = 5;
2805                 p.iph.protocol = IPPROTO_IPV6;
2806                 p.iph.ttl = 64;
2807                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2808
2809                 if (ops->ndo_do_ioctl) {
2810                         mm_segment_t oldfs = get_fs();
2811
2812                         set_fs(KERNEL_DS);
2813                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2814                         set_fs(oldfs);
2815                 } else
2816                         err = -EOPNOTSUPP;
2817
2818                 if (err == 0) {
2819                         err = -ENOBUFS;
2820                         dev = __dev_get_by_name(net, p.name);
2821                         if (!dev)
2822                                 goto err_exit;
2823                         err = dev_open(dev);
2824                 }
2825         }
2826 #endif
2827
2828 err_exit:
2829         rtnl_unlock();
2830         return err;
2831 }
2832
2833 static int ipv6_mc_config(struct sock *sk, bool join,
2834                           const struct in6_addr *addr, int ifindex)
2835 {
2836         int ret;
2837
2838         ASSERT_RTNL();
2839
2840         lock_sock(sk);
2841         if (join)
2842                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2843         else
2844                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2845         release_sock(sk);
2846
2847         return ret;
2848 }
2849
2850 /*
2851  *      Manual configuration of address on an interface
2852  */
2853 static int inet6_addr_add(struct net *net, int ifindex,
2854                           struct ifa6_config *cfg,
2855                           struct netlink_ext_ack *extack)
2856 {
2857         struct inet6_ifaddr *ifp;
2858         struct inet6_dev *idev;
2859         struct net_device *dev;
2860         unsigned long timeout;
2861         clock_t expires;
2862         u32 flags;
2863
2864         ASSERT_RTNL();
2865
2866         if (cfg->plen > 128)
2867                 return -EINVAL;
2868
2869         /* check the lifetime */
2870         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2871                 return -EINVAL;
2872
2873         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2874                 return -EINVAL;
2875
2876         dev = __dev_get_by_index(net, ifindex);
2877         if (!dev)
2878                 return -ENODEV;
2879
2880         idev = addrconf_add_dev(dev);
2881         if (IS_ERR(idev))
2882                 return PTR_ERR(idev);
2883
2884         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2885                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2886                                          true, cfg->pfx, ifindex);
2887
2888                 if (ret < 0)
2889                         return ret;
2890         }
2891
2892         cfg->scope = ipv6_addr_scope(cfg->pfx);
2893
2894         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2895         if (addrconf_finite_timeout(timeout)) {
2896                 expires = jiffies_to_clock_t(timeout * HZ);
2897                 cfg->valid_lft = timeout;
2898                 flags = RTF_EXPIRES;
2899         } else {
2900                 expires = 0;
2901                 flags = 0;
2902                 cfg->ifa_flags |= IFA_F_PERMANENT;
2903         }
2904
2905         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2906         if (addrconf_finite_timeout(timeout)) {
2907                 if (timeout == 0)
2908                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2909                 cfg->preferred_lft = timeout;
2910         }
2911
2912         ifp = ipv6_add_addr(idev, cfg, true, extack);
2913         if (!IS_ERR(ifp)) {
2914                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2915                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2916                                               ifp->rt_priority, dev, expires,
2917                                               flags, GFP_KERNEL);
2918                 }
2919
2920                 /* Send a netlink notification if DAD is enabled and
2921                  * optimistic flag is not set
2922                  */
2923                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2924                         ipv6_ifa_notify(0, ifp);
2925                 /*
2926                  * Note that section 3.1 of RFC 4429 indicates
2927                  * that the Optimistic flag should not be set for
2928                  * manually configured addresses
2929                  */
2930                 addrconf_dad_start(ifp);
2931                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2932                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2933                                          cfg->preferred_lft, true, jiffies);
2934                 in6_ifa_put(ifp);
2935                 addrconf_verify_rtnl();
2936                 return 0;
2937         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2938                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2939                                cfg->pfx, ifindex);
2940         }
2941
2942         return PTR_ERR(ifp);
2943 }
2944
2945 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2946                           const struct in6_addr *pfx, unsigned int plen)
2947 {
2948         struct inet6_ifaddr *ifp;
2949         struct inet6_dev *idev;
2950         struct net_device *dev;
2951
2952         if (plen > 128)
2953                 return -EINVAL;
2954
2955         dev = __dev_get_by_index(net, ifindex);
2956         if (!dev)
2957                 return -ENODEV;
2958
2959         idev = __in6_dev_get(dev);
2960         if (!idev)
2961                 return -ENXIO;
2962
2963         read_lock_bh(&idev->lock);
2964         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2965                 if (ifp->prefix_len == plen &&
2966                     ipv6_addr_equal(pfx, &ifp->addr)) {
2967                         in6_ifa_hold(ifp);
2968                         read_unlock_bh(&idev->lock);
2969
2970                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2971                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2972                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2973                                                  jiffies);
2974                         ipv6_del_addr(ifp);
2975                         addrconf_verify_rtnl();
2976                         if (ipv6_addr_is_multicast(pfx)) {
2977                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2978                                                false, pfx, dev->ifindex);
2979                         }
2980                         return 0;
2981                 }
2982         }
2983         read_unlock_bh(&idev->lock);
2984         return -EADDRNOTAVAIL;
2985 }
2986
2987
2988 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2989 {
2990         struct ifa6_config cfg = {
2991                 .ifa_flags = IFA_F_PERMANENT,
2992                 .preferred_lft = INFINITY_LIFE_TIME,
2993                 .valid_lft = INFINITY_LIFE_TIME,
2994         };
2995         struct in6_ifreq ireq;
2996         int err;
2997
2998         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2999                 return -EPERM;
3000
3001         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3002                 return -EFAULT;
3003
3004         cfg.pfx = &ireq.ifr6_addr;
3005         cfg.plen = ireq.ifr6_prefixlen;
3006
3007         rtnl_lock();
3008         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3009         rtnl_unlock();
3010         return err;
3011 }
3012
3013 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3014 {
3015         struct in6_ifreq ireq;
3016         int err;
3017
3018         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3019                 return -EPERM;
3020
3021         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3022                 return -EFAULT;
3023
3024         rtnl_lock();
3025         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3026                              ireq.ifr6_prefixlen);
3027         rtnl_unlock();
3028         return err;
3029 }
3030
3031 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3032                      int plen, int scope)
3033 {
3034         struct inet6_ifaddr *ifp;
3035         struct ifa6_config cfg = {
3036                 .pfx = addr,
3037                 .plen = plen,
3038                 .ifa_flags = IFA_F_PERMANENT,
3039                 .valid_lft = INFINITY_LIFE_TIME,
3040                 .preferred_lft = INFINITY_LIFE_TIME,
3041                 .scope = scope
3042         };
3043
3044         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3045         if (!IS_ERR(ifp)) {
3046                 spin_lock_bh(&ifp->lock);
3047                 ifp->flags &= ~IFA_F_TENTATIVE;
3048                 spin_unlock_bh(&ifp->lock);
3049                 rt_genid_bump_ipv6(dev_net(idev->dev));
3050                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3051                 in6_ifa_put(ifp);
3052         }
3053 }
3054
3055 #if IS_ENABLED(CONFIG_IPV6_SIT)
3056 static void sit_add_v4_addrs(struct inet6_dev *idev)
3057 {
3058         struct in6_addr addr;
3059         struct net_device *dev;
3060         struct net *net = dev_net(idev->dev);
3061         int scope, plen;
3062         u32 pflags = 0;
3063
3064         ASSERT_RTNL();
3065
3066         memset(&addr, 0, sizeof(struct in6_addr));
3067         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3068
3069         if (idev->dev->flags&IFF_POINTOPOINT) {
3070                 addr.s6_addr32[0] = htonl(0xfe800000);
3071                 scope = IFA_LINK;
3072                 plen = 64;
3073         } else {
3074                 scope = IPV6_ADDR_COMPATv4;
3075                 plen = 96;
3076                 pflags |= RTF_NONEXTHOP;
3077         }
3078
3079         if (addr.s6_addr32[3]) {
3080                 add_addr(idev, &addr, plen, scope);
3081                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3082                                       GFP_KERNEL);
3083                 return;
3084         }
3085
3086         for_each_netdev(net, dev) {
3087                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3088                 if (in_dev && (dev->flags & IFF_UP)) {
3089                         struct in_ifaddr *ifa;
3090
3091                         int flag = scope;
3092
3093                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3094
3095                                 addr.s6_addr32[3] = ifa->ifa_local;
3096
3097                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3098                                         continue;
3099                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3100                                         if (idev->dev->flags&IFF_POINTOPOINT)
3101                                                 continue;
3102                                         flag |= IFA_HOST;
3103                                 }
3104
3105                                 add_addr(idev, &addr, plen, flag);
3106                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3107                                                       0, pflags, GFP_KERNEL);
3108                         }
3109                 }
3110         }
3111 }
3112 #endif
3113
3114 static void init_loopback(struct net_device *dev)
3115 {
3116         struct inet6_dev  *idev;
3117
3118         /* ::1 */
3119
3120         ASSERT_RTNL();
3121
3122         idev = ipv6_find_idev(dev);
3123         if (!idev) {
3124                 pr_debug("%s: add_dev failed\n", __func__);
3125                 return;
3126         }
3127
3128         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3129 }
3130
3131 void addrconf_add_linklocal(struct inet6_dev *idev,
3132                             const struct in6_addr *addr, u32 flags)
3133 {
3134         struct ifa6_config cfg = {
3135                 .pfx = addr,
3136                 .plen = 64,
3137                 .ifa_flags = flags | IFA_F_PERMANENT,
3138                 .valid_lft = INFINITY_LIFE_TIME,
3139                 .preferred_lft = INFINITY_LIFE_TIME,
3140                 .scope = IFA_LINK
3141         };
3142         struct inet6_ifaddr *ifp;
3143
3144 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3145         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3146              idev->cnf.optimistic_dad) &&
3147             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3148                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3149 #endif
3150
3151         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3152         if (!IS_ERR(ifp)) {
3153                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3154                                       0, 0, GFP_ATOMIC);
3155                 addrconf_dad_start(ifp);
3156                 in6_ifa_put(ifp);
3157         }
3158 }
3159 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3160
3161 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3162 {
3163         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3164                 return true;
3165
3166         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3167             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3168                 return true;
3169
3170         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3171             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3172                 return true;
3173
3174         return false;
3175 }
3176
3177 static int ipv6_generate_stable_address(struct in6_addr *address,
3178                                         u8 dad_count,
3179                                         const struct inet6_dev *idev)
3180 {
3181         static DEFINE_SPINLOCK(lock);
3182         static __u32 digest[SHA_DIGEST_WORDS];
3183         static __u32 workspace[SHA_WORKSPACE_WORDS];
3184
3185         static union {
3186                 char __data[SHA_MESSAGE_BYTES];
3187                 struct {
3188                         struct in6_addr secret;
3189                         __be32 prefix[2];
3190                         unsigned char hwaddr[MAX_ADDR_LEN];
3191                         u8 dad_count;
3192                 } __packed;
3193         } data;
3194
3195         struct in6_addr secret;
3196         struct in6_addr temp;
3197         struct net *net = dev_net(idev->dev);
3198
3199         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3200
3201         if (idev->cnf.stable_secret.initialized)
3202                 secret = idev->cnf.stable_secret.secret;
3203         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3204                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3205         else
3206                 return -1;
3207
3208 retry:
3209         spin_lock_bh(&lock);
3210
3211         sha_init(digest);
3212         memset(&data, 0, sizeof(data));
3213         memset(workspace, 0, sizeof(workspace));
3214         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3215         data.prefix[0] = address->s6_addr32[0];
3216         data.prefix[1] = address->s6_addr32[1];
3217         data.secret = secret;
3218         data.dad_count = dad_count;
3219
3220         sha_transform(digest, data.__data, workspace);
3221
3222         temp = *address;
3223         temp.s6_addr32[2] = (__force __be32)digest[0];
3224         temp.s6_addr32[3] = (__force __be32)digest[1];
3225
3226         spin_unlock_bh(&lock);
3227
3228         if (ipv6_reserved_interfaceid(temp)) {
3229                 dad_count++;
3230                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3231                         return -1;
3232                 goto retry;
3233         }
3234
3235         *address = temp;
3236         return 0;
3237 }
3238
3239 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3240 {
3241         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3242
3243         if (s->initialized)
3244                 return;
3245         s = &idev->cnf.stable_secret;
3246         get_random_bytes(&s->secret, sizeof(s->secret));
3247         s->initialized = true;
3248 }
3249
3250 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3251 {
3252         struct in6_addr addr;
3253
3254         /* no link local addresses on L3 master devices */
3255         if (netif_is_l3_master(idev->dev))
3256                 return;
3257
3258         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3259
3260         switch (idev->cnf.addr_gen_mode) {
3261         case IN6_ADDR_GEN_MODE_RANDOM:
3262                 ipv6_gen_mode_random_init(idev);
3263                 /* fallthrough */
3264         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3265                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3266                         addrconf_add_linklocal(idev, &addr,
3267                                                IFA_F_STABLE_PRIVACY);
3268                 else if (prefix_route)
3269                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3270                                               0, 0, GFP_KERNEL);
3271                 break;
3272         case IN6_ADDR_GEN_MODE_EUI64:
3273                 /* addrconf_add_linklocal also adds a prefix_route and we
3274                  * only need to care about prefix routes if ipv6_generate_eui64
3275                  * couldn't generate one.
3276                  */
3277                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3278                         addrconf_add_linklocal(idev, &addr, 0);
3279                 else if (prefix_route)
3280                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3281                                               0, 0, GFP_KERNEL);
3282                 break;
3283         case IN6_ADDR_GEN_MODE_NONE:
3284         default:
3285                 /* will not add any link local address */
3286                 break;
3287         }
3288 }
3289
3290 static void addrconf_dev_config(struct net_device *dev)
3291 {
3292         struct inet6_dev *idev;
3293
3294         ASSERT_RTNL();
3295
3296         if ((dev->type != ARPHRD_ETHER) &&
3297             (dev->type != ARPHRD_FDDI) &&
3298             (dev->type != ARPHRD_ARCNET) &&
3299             (dev->type != ARPHRD_INFINIBAND) &&
3300             (dev->type != ARPHRD_IEEE1394) &&
3301             (dev->type != ARPHRD_TUNNEL6) &&
3302             (dev->type != ARPHRD_6LOWPAN) &&
3303             (dev->type != ARPHRD_IP6GRE) &&
3304             (dev->type != ARPHRD_IPGRE) &&
3305             (dev->type != ARPHRD_TUNNEL) &&
3306             (dev->type != ARPHRD_NONE) &&
3307             (dev->type != ARPHRD_RAWIP)) {
3308                 /* Alas, we support only Ethernet autoconfiguration. */
3309                 return;
3310         }
3311
3312         idev = addrconf_add_dev(dev);
3313         if (IS_ERR(idev))
3314                 return;
3315
3316         /* this device type has no EUI support */
3317         if (dev->type == ARPHRD_NONE &&
3318             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3319                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3320
3321         addrconf_addr_gen(idev, false);
3322 }
3323
3324 #if IS_ENABLED(CONFIG_IPV6_SIT)
3325 static void addrconf_sit_config(struct net_device *dev)
3326 {
3327         struct inet6_dev *idev;
3328
3329         ASSERT_RTNL();
3330
3331         /*
3332          * Configure the tunnel with one of our IPv4
3333          * addresses... we should configure all of
3334          * our v4 addrs in the tunnel
3335          */
3336
3337         idev = ipv6_find_idev(dev);
3338         if (!idev) {
3339                 pr_debug("%s: add_dev failed\n", __func__);
3340                 return;
3341         }
3342
3343         if (dev->priv_flags & IFF_ISATAP) {
3344                 addrconf_addr_gen(idev, false);
3345                 return;
3346         }
3347
3348         sit_add_v4_addrs(idev);
3349
3350         if (dev->flags&IFF_POINTOPOINT)
3351                 addrconf_add_mroute(dev);
3352 }
3353 #endif
3354
3355 #if IS_ENABLED(CONFIG_NET_IPGRE)
3356 static void addrconf_gre_config(struct net_device *dev)
3357 {
3358         struct inet6_dev *idev;
3359
3360         ASSERT_RTNL();
3361
3362         idev = ipv6_find_idev(dev);
3363         if (!idev) {
3364                 pr_debug("%s: add_dev failed\n", __func__);
3365                 return;
3366         }
3367
3368         addrconf_addr_gen(idev, true);
3369         if (dev->flags & IFF_POINTOPOINT)
3370                 addrconf_add_mroute(dev);
3371 }
3372 #endif
3373
3374 static int fixup_permanent_addr(struct net *net,
3375                                 struct inet6_dev *idev,
3376                                 struct&nbs