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 rt6_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         else
 389                 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
 390
 391         ndev->cnf.mtu6 = dev->mtu;
 392         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 393         if (!ndev->nd_parms) {
 394                 kfree(ndev);
 395                 return ERR_PTR(err);
 396         }
 397         if (ndev->cnf.forwarding)
 398                 dev_disable_lro(dev);
 399         /* We refer to the device */
 400         dev_hold(dev);
 401
 402         if (snmp6_alloc_dev(ndev) < 0) {
 403                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 404                            __func__);
 405                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 406                 dev_put(dev);
 407                 kfree(ndev);
 408                 return ERR_PTR(err);
 409         }
 410
 411         if (snmp6_register_dev(ndev) < 0) {
 412                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 413                            __func__, dev->name);
 414                 goto err_release;
 415         }
 416
 417         /* One reference from device. */
 418         refcount_set(&ndev->refcnt, 1);
 419
 420         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 421                 ndev->cnf.accept_dad = -1;
 422
 423 #if IS_ENABLED(CONFIG_IPV6_SIT)
 424         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 425                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 426                 ndev->cnf.rtr_solicits = 0;
 427         }
 428 #endif
 429
 430         INIT_LIST_HEAD(&ndev->tempaddr_list);
 431         ndev->desync_factor = U32_MAX;
 432         if ((dev->flags&IFF_LOOPBACK) ||
 433             dev->type == ARPHRD_TUNNEL ||
 434             dev->type == ARPHRD_TUNNEL6 ||
 435             dev->type == ARPHRD_SIT ||
 436             dev->type == ARPHRD_NONE) {
 437                 ndev->cnf.use_tempaddr = -1;
 438         } else
 439                 ipv6_regen_rndid(ndev);
 440
 441         ndev->token = in6addr_any;
 442
 443         if (netif_running(dev) && addrconf_link_ready(dev))
 444                 ndev->if_flags |= IF_READY;
 445
 446         ipv6_mc_init_dev(ndev);
 447         ndev->tstamp = jiffies;
 448         err = addrconf_sysctl_register(ndev);
 449         if (err) {
 450                 ipv6_mc_destroy_dev(ndev);
 451                 snmp6_unregister_dev(ndev);
 452                 goto err_release;
 453         }
 454         /* protected by rtnl_lock */
 455         rcu_assign_pointer(dev->ip6_ptr, ndev);
 456
 457         /* Join interface-local all-node multicast group */
 458         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 459
 460         /* Join all-node multicast group */
 461         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 462
 463         /* Join all-router multicast group if forwarding is set */
 464         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 465                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 466
 467         return ndev;
 468
 469 err_release:
 470         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 471         ndev->dead = 1;
 472         in6_dev_finish_destroy(ndev);
 473         return ERR_PTR(err);
 474 }
 475
 476 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 477 {
 478         struct inet6_dev *idev;
 479
 480         ASSERT_RTNL();
 481
 482         idev = __in6_dev_get(dev);
 483         if (!idev) {
 484                 idev = ipv6_add_dev(dev);
 485                 if (IS_ERR(idev))
 486                         return NULL;
 487         }
 488
 489         if (dev->flags&IFF_UP)
 490                 ipv6_mc_up(idev);
 491         return idev;
 492 }
 493
 494 static int inet6_netconf_msgsize_devconf(int type)
 495 {
 496         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 497                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 498         bool all = false;
 499
 500         if (type == NETCONFA_ALL)
 501                 all = true;
 502
 503         if (all || type == NETCONFA_FORWARDING)
 504                 size += nla_total_size(4);
 505 #ifdef CONFIG_IPV6_MROUTE
 506         if (all || type == NETCONFA_MC_FORWARDING)
 507                 size += nla_total_size(4);
 508 #endif
 509         if (all || type == NETCONFA_PROXY_NEIGH)
 510                 size += nla_total_size(4);
 511
 512         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 513                 size += nla_total_size(4);
 514
 515         return size;
 516 }
 517
 518 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 519                                       struct ipv6_devconf *devconf, u32 portid,
 520                                       u32 seq, int event, unsigned int flags,
 521                                       int type)
 522 {
 523         struct nlmsghdr  *nlh;
 524         struct netconfmsg *ncm;
 525         bool all = false;
 526
 527         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 528                         flags);
 529         if (!nlh)
 530                 return -EMSGSIZE;
 531
 532         if (type == NETCONFA_ALL)
 533                 all = true;
 534
 535         ncm = nlmsg_data(nlh);
 536         ncm->ncm_family = AF_INET6;
 537
 538         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 539                 goto nla_put_failure;
 540
 541         if (!devconf)
 542                 goto out;
 543
 544         if ((all || type == NETCONFA_FORWARDING) &&
 545             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 546                 goto nla_put_failure;
 547 #ifdef CONFIG_IPV6_MROUTE
 548         if ((all || type == NETCONFA_MC_FORWARDING) &&
 549             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 550                         devconf->mc_forwarding) < 0)
 551                 goto nla_put_failure;
 552 #endif
 553         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 554             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 555                 goto nla_put_failure;
 556
 557         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 558             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 559                         devconf->ignore_routes_with_linkdown) < 0)
 560                 goto nla_put_failure;
 561
 562 out:
 563         nlmsg_end(skb, nlh);
 564         return 0;
 565
 566 nla_put_failure:
 567         nlmsg_cancel(skb, nlh);
 568         return -EMSGSIZE;
 569 }
 570
 571 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 572                                   int ifindex, struct ipv6_devconf *devconf)
 573 {
 574         struct sk_buff *skb;
 575         int err = -ENOBUFS;
 576
 577         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 578         if (!skb)
 579                 goto errout;
 580
 581         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 582                                          event, 0, type);
 583         if (err < 0) {
 584                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 585                 WARN_ON(err == -EMSGSIZE);
 586                 kfree_skb(skb);
 587                 goto errout;
 588         }
 589         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 590         return;
 591 errout:
 592         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 593 }
 594
 595 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 596         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 597         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 598         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 599         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 600 };
 601
 602 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 603                                      struct nlmsghdr *nlh,
 604                                      struct netlink_ext_ack *extack)
 605 {
 606         struct net *net = sock_net(in_skb->sk);
 607         struct nlattr *tb[NETCONFA_MAX+1];
 608         struct inet6_dev *in6_dev = NULL;
 609         struct net_device *dev = NULL;
 610         struct netconfmsg *ncm;
 611         struct sk_buff *skb;
 612         struct ipv6_devconf *devconf;
 613         int ifindex;
 614         int err;
 615
 616         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 617                           devconf_ipv6_policy, extack);
 618         if (err < 0)
 619                 return err;
 620
 621         if (!tb[NETCONFA_IFINDEX])
 622                 return -EINVAL;
 623
 624         err = -EINVAL;
 625         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 626         switch (ifindex) {
 627         case NETCONFA_IFINDEX_ALL:
 628                 devconf = net->ipv6.devconf_all;
 629                 break;
 630         case NETCONFA_IFINDEX_DEFAULT:
 631                 devconf = net->ipv6.devconf_dflt;
 632                 break;
 633         default:
 634                 dev = dev_get_by_index(net, ifindex);
 635                 if (!dev)
 636                         return -EINVAL;
 637                 in6_dev = in6_dev_get(dev);
 638                 if (!in6_dev)
 639                         goto errout;
 640                 devconf = &in6_dev->cnf;
 641                 break;
 642         }
 643
 644         err = -ENOBUFS;
 645         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 646         if (!skb)
 647                 goto errout;
 648
 649         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 650                                          NETLINK_CB(in_skb).portid,
 651                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 652                                          NETCONFA_ALL);
 653         if (err < 0) {
 654                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 655                 WARN_ON(err == -EMSGSIZE);
 656                 kfree_skb(skb);
 657                 goto errout;
 658         }
 659         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 660 errout:
 661         if (in6_dev)
 662                 in6_dev_put(in6_dev);
 663         if (dev)
 664                 dev_put(dev);
 665         return err;
 666 }
 667
 668 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 669                                       struct netlink_callback *cb)
 670 {
 671         struct net *net = sock_net(skb->sk);
 672         int h, s_h;
 673         int idx, s_idx;
 674         struct net_device *dev;
 675         struct inet6_dev *idev;
 676         struct hlist_head *head;
 677
 678         s_h = cb->args[0];
 679         s_idx = idx = cb->args[1];
 680
 681         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 682                 idx = 0;
 683                 head = &net->dev_index_head[h];
 684                 rcu_read_lock();
 685                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 686                           net->dev_base_seq;
 687                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 688                         if (idx < s_idx)
 689                                 goto cont;
 690                         idev = __in6_dev_get(dev);
 691                         if (!idev)
 692                                 goto cont;
 693
 694                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 695                                                        &idev->cnf,
 696                                                        NETLINK_CB(cb->skb).portid,
 697                                                        cb->nlh->nlmsg_seq,
 698                                                        RTM_NEWNETCONF,
 699                                                        NLM_F_MULTI,
 700                                                        NETCONFA_ALL) < 0) {
 701                                 rcu_read_unlock();
 702                                 goto done;
 703                         }
 704                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 705 cont:
 706                         idx++;
 707                 }
 708                 rcu_read_unlock();
 709         }
 710         if (h == NETDEV_HASHENTRIES) {
 711                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 712                                                net->ipv6.devconf_all,
 713                                                NETLINK_CB(cb->skb).portid,
 714                                                cb->nlh->nlmsg_seq,
 715                                                RTM_NEWNETCONF, NLM_F_MULTI,
 716                                                NETCONFA_ALL) < 0)
 717                         goto done;
 718                 else
 719                         h++;
 720         }
 721         if (h == NETDEV_HASHENTRIES + 1) {
 722                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 723                                                net->ipv6.devconf_dflt,
 724                                                NETLINK_CB(cb->skb).portid,
 725                                                cb->nlh->nlmsg_seq,
 726                                                RTM_NEWNETCONF, NLM_F_MULTI,
 727                                                NETCONFA_ALL) < 0)
 728                         goto done;
 729                 else
 730                         h++;
 731         }
 732 done:
 733         cb->args[0] = h;
 734         cb->args[1] = idx;
 735
 736         return skb->len;
 737 }
 738
 739 #ifdef CONFIG_SYSCTL
 740 static void dev_forward_change(struct inet6_dev *idev)
 741 {
 742         struct net_device *dev;
 743         struct inet6_ifaddr *ifa;
 744
 745         if (!idev)
 746                 return;
 747         dev = idev->dev;
 748         if (idev->cnf.forwarding)
 749                 dev_disable_lro(dev);
 750         if (dev->flags & IFF_MULTICAST) {
 751                 if (idev->cnf.forwarding) {
 752                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 753                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 754                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 755                 } else {
 756                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 757                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 758                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 759                 }
 760         }
 761
 762         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 763                 if (ifa->flags&IFA_F_TENTATIVE)
 764                         continue;
 765                 if (idev->cnf.forwarding)
 766                         addrconf_join_anycast(ifa);
 767                 else
 768                         addrconf_leave_anycast(ifa);
 769         }
 770         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 771                                      NETCONFA_FORWARDING,
 772                                      dev->ifindex, &idev->cnf);
 773 }
 774
 775
 776 static void addrconf_forward_change(struct net *net, __s32 newf)
 777 {
 778         struct net_device *dev;
 779         struct inet6_dev *idev;
 780
 781         for_each_netdev(net, dev) {
 782                 idev = __in6_dev_get(dev);
 783                 if (idev) {
 784                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 785                         idev->cnf.forwarding = newf;
 786                         if (changed)
 787                                 dev_forward_change(idev);
 788                 }
 789         }
 790 }
 791
 792 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 793 {
 794         struct net *net;
 795         int old;
 796
 797         if (!rtnl_trylock())
 798                 return restart_syscall();
 799
 800         net = (struct net *)table->extra2;
 801         old = *p;
 802         *p = newf;
 803
 804         if (p == &net->ipv6.devconf_dflt->forwarding) {
 805                 if ((!newf) ^ (!old))
 806                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 807                                                      NETCONFA_FORWARDING,
 808                                                      NETCONFA_IFINDEX_DEFAULT,
 809                                                      net->ipv6.devconf_dflt);
 810                 rtnl_unlock();
 811                 return 0;
 812         }
 813
 814         if (p == &net->ipv6.devconf_all->forwarding) {
 815                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 816
 817                 net->ipv6.devconf_dflt->forwarding = newf;
 818                 if ((!newf) ^ (!old_dflt))
 819                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 820                                                      NETCONFA_FORWARDING,
 821                                                      NETCONFA_IFINDEX_DEFAULT,
 822                                                      net->ipv6.devconf_dflt);
 823
 824                 addrconf_forward_change(net, newf);
 825                 if ((!newf) ^ (!old))
 826                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 827                                                      NETCONFA_FORWARDING,
 828                                                      NETCONFA_IFINDEX_ALL,
 829                                                      net->ipv6.devconf_all);
 830         } else if ((!newf) ^ (!old))
 831                 dev_forward_change((struct inet6_dev *)table->extra1);
 832         rtnl_unlock();
 833
 834         if (newf)
 835                 rt6_purge_dflt_routers(net);
 836         return 1;
 837 }
 838
 839 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 840 {
 841         struct net_device *dev;
 842         struct inet6_dev *idev;
 843
 844         for_each_netdev(net, dev) {
 845                 idev = __in6_dev_get(dev);
 846                 if (idev) {
 847                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 848
 849                         idev->cnf.ignore_routes_with_linkdown = newf;
 850                         if (changed)
 851                                 inet6_netconf_notify_devconf(dev_net(dev),
 852                                                              RTM_NEWNETCONF,
 853                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 854                                                              dev->ifindex,
 855                                                              &idev->cnf);
 856                 }
 857         }
 858 }
 859
 860 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 861 {
 862         struct net *net;
 863         int old;
 864
 865         if (!rtnl_trylock())
 866                 return restart_syscall();
 867
 868         net = (struct net *)table->extra2;
 869         old = *p;
 870         *p = newf;
 871
 872         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 873                 if ((!newf) ^ (!old))
 874                         inet6_netconf_notify_devconf(net,
 875                                                      RTM_NEWNETCONF,
 876                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 877                                                      NETCONFA_IFINDEX_DEFAULT,
 878                                                      net->ipv6.devconf_dflt);
 879                 rtnl_unlock();
 880                 return 0;
 881         }
 882
 883         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 884                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 885                 addrconf_linkdown_change(net, newf);
 886                 if ((!newf) ^ (!old))
 887                         inet6_netconf_notify_devconf(net,
 888                                                      RTM_NEWNETCONF,
 889                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 890                                                      NETCONFA_IFINDEX_ALL,
 891                                                      net->ipv6.devconf_all);
 892         }
 893         rtnl_unlock();
 894
 895         return 1;
 896 }
 897
 898 #endif
 899
 900 /* Nobody refers to this ifaddr, destroy it */
 901 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 902 {
 903         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 904
 905 #ifdef NET_REFCNT_DEBUG
 906         pr_debug("%s\n", __func__);
 907 #endif
 908
 909         in6_dev_put(ifp->idev);
 910
 911         if (cancel_delayed_work(&ifp->dad_work))
 912                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 913                           ifp);
 914
 915         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 916                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 917                 return;
 918         }
 919         ip6_rt_put(ifp->rt);
 920
 921         kfree_rcu(ifp, rcu);
 922 }
 923
 924 static void
 925 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 926 {
 927         struct list_head *p;
 928         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 929
 930         /*
 931          * Each device address list is sorted in order of scope -
 932          * global before linklocal.
 933          */
 934         list_for_each(p, &idev->addr_list) {
 935                 struct inet6_ifaddr *ifa
 936                         = list_entry(p, struct inet6_ifaddr, if_list);
 937                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 938                         break;
 939         }
 940
 941         list_add_tail_rcu(&ifp->if_list, p);
 942 }
 943
 944 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 945 {
 946         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 947
 948         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 949 }
 950
 951 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 952                                struct net_device *dev, unsigned int hash)
 953 {
 954         struct inet6_ifaddr *ifp;
 955
 956         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
 957                 if (!net_eq(dev_net(ifp->idev->dev), net))
 958                         continue;
 959                 if (ipv6_addr_equal(&ifp->addr, addr)) {
 960                         if (!dev || ifp->idev->dev == dev)
 961                                 return true;
 962                 }
 963         }
 964         return false;
 965 }
 966
 967 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
 968 {
 969         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
 970         int err = 0;
 971
 972         spin_lock(&addrconf_hash_lock);
 973
 974         /* Ignore adding duplicate addresses on an interface */
 975         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
 976                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
 977                 err = -EEXIST;
 978         } else {
 979                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 980         }
 981
 982         spin_unlock(&addrconf_hash_lock);
 983
 984         return err;
 985 }
 986
 987 /* On success it returns ifp with increased reference count */
 988
 989 static struct inet6_ifaddr *
 990 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 991               const struct in6_addr *peer_addr, int pfxlen,
 992               int scope, u32 flags, u32 valid_lft, u32 prefered_lft,
 993               bool can_block, struct netlink_ext_ack *extack)
 994 {
 995         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
 996         struct net *net = dev_net(idev->dev);
 997         struct inet6_ifaddr *ifa = NULL;
 998         struct rt6_info *rt = NULL;
 999         int err = 0;
1000         int addr_type = ipv6_addr_type(addr);
1001
1002         if (addr_type == IPV6_ADDR_ANY ||
1003             addr_type & IPV6_ADDR_MULTICAST ||
1004             (!(idev->dev->flags & IFF_LOOPBACK) &&
1005              addr_type & IPV6_ADDR_LOOPBACK))
1006                 return ERR_PTR(-EADDRNOTAVAIL);
1007
1008         if (idev->dead) {
1009                 err = -ENODEV;                  /*XXX*/
1010                 goto out;
1011         }
1012
1013         if (idev->cnf.disable_ipv6) {
1014                 err = -EACCES;
1015                 goto out;
1016         }
1017
1018         /* validator notifier needs to be blocking;
1019          * do not call in atomic context
1020          */
1021         if (can_block) {
1022                 struct in6_validator_info i6vi = {
1023                         .i6vi_addr = *addr,
1024                         .i6vi_dev = idev,
1025                         .extack = extack,
1026                 };
1027
1028                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1029                 err = notifier_to_errno(err);
1030                 if (err < 0)
1031                         goto out;
1032         }
1033
1034         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1035         if (!ifa) {
1036                 err = -ENOBUFS;
1037                 goto out;
1038         }
1039
1040         rt = addrconf_dst_alloc(idev, addr, false);
1041         if (IS_ERR(rt)) {
1042                 err = PTR_ERR(rt);
1043                 rt = NULL;
1044                 goto out;
1045         }
1046
1047         if (net->ipv6.devconf_all->disable_policy ||
1048             idev->cnf.disable_policy)
1049                 rt->dst.flags |= DST_NOPOLICY;
1050
1051         neigh_parms_data_state_setall(idev->nd_parms);
1052
1053         ifa->addr = *addr;
1054         if (peer_addr)
1055                 ifa->peer_addr = *peer_addr;
1056
1057         spin_lock_init(&ifa->lock);
1058         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1059         INIT_HLIST_NODE(&ifa->addr_lst);
1060         ifa->scope = scope;
1061         ifa->prefix_len = pfxlen;
1062         ifa->flags = flags;
1063         /* No need to add the TENTATIVE flag for addresses with NODAD */
1064         if (!(flags & IFA_F_NODAD))
1065                 ifa->flags |= IFA_F_TENTATIVE;
1066         ifa->valid_lft = valid_lft;
1067         ifa->prefered_lft = prefered_lft;
1068         ifa->cstamp = ifa->tstamp = jiffies;
1069         ifa->tokenized = false;
1070
1071         ifa->rt = rt;
1072
1073         ifa->idev = idev;
1074         in6_dev_hold(idev);
1075
1076         /* For caller */
1077         refcount_set(&ifa->refcnt, 1);
1078
1079         rcu_read_lock_bh();
1080
1081         err = ipv6_add_addr_hash(idev->dev, ifa);
1082         if (err < 0) {
1083                 rcu_read_unlock_bh();
1084                 goto out;
1085         }
1086
1087         write_lock(&idev->lock);
1088
1089         /* Add to inet6_dev unicast addr list. */
1090         ipv6_link_dev_addr(idev, ifa);
1091
1092         if (ifa->flags&IFA_F_TEMPORARY) {
1093                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1094                 in6_ifa_hold(ifa);
1095         }
1096
1097         in6_ifa_hold(ifa);
1098         write_unlock(&idev->lock);
1099
1100         rcu_read_unlock_bh();
1101
1102         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1103 out:
1104         if (unlikely(err < 0)) {
1105                 if (rt)
1106                         ip6_rt_put(rt);
1107                 if (ifa) {
1108                         if (ifa->idev)
1109                                 in6_dev_put(ifa->idev);
1110                         kfree(ifa);
1111                 }
1112                 ifa = ERR_PTR(err);
1113         }
1114
1115         return ifa;
1116 }
1117
1118 enum cleanup_prefix_rt_t {
1119         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1120         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1121         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1122 };
1123
1124 /*
1125  * Check, whether the prefix for ifp would still need a prefix route
1126  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1127  * constants.
1128  *
1129  * 1) we don't purge prefix if address was not permanent.
1130  *    prefix is managed by its own lifetime.
1131  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1132  * 3) if there are no addresses, delete prefix.
1133  * 4) if there are still other permanent address(es),
1134  *    corresponding prefix is still permanent.
1135  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1136  *    don't purge the prefix, assume user space is managing it.
1137  * 6) otherwise, update prefix lifetime to the
1138  *    longest valid lifetime among the corresponding
1139  *    addresses on the device.
1140  *    Note: subsequent RA will update lifetime.
1141  **/
1142 static enum cleanup_prefix_rt_t
1143 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1144 {
1145         struct inet6_ifaddr *ifa;
1146         struct inet6_dev *idev = ifp->idev;
1147         unsigned long lifetime;
1148         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1149
1150         *expires = jiffies;
1151
1152         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1153                 if (ifa == ifp)
1154                         continue;
1155                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1156                                        ifp->prefix_len))
1157                         continue;
1158                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1159                         return CLEANUP_PREFIX_RT_NOP;
1160
1161                 action = CLEANUP_PREFIX_RT_EXPIRE;
1162
1163                 spin_lock(&ifa->lock);
1164
1165                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1166                 /*
1167                  * Note: Because this address is
1168                  * not permanent, lifetime <
1169                  * LONG_MAX / HZ here.
1170                  */
1171                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1172                         *expires = ifa->tstamp + lifetime * HZ;
1173                 spin_unlock(&ifa->lock);
1174         }
1175
1176         return action;
1177 }
1178
1179 static void
1180 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1181 {
1182         struct rt6_info *rt;
1183
1184         rt = addrconf_get_prefix_route(&ifp->addr,
1185                                        ifp->prefix_len,
1186                                        ifp->idev->dev,
1187                                        0, RTF_GATEWAY | RTF_DEFAULT);
1188         if (rt) {
1189                 if (del_rt)
1190                         ip6_del_rt(rt);
1191                 else {
1192                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1193                                 rt6_set_expires(rt, expires);
1194                         ip6_rt_put(rt);
1195                 }
1196         }
1197 }
1198
1199
1200 /* This function wants to get referenced ifp and releases it before return */
1201
1202 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1203 {
1204         int state;
1205         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1206         unsigned long expires;
1207
1208         ASSERT_RTNL();
1209
1210         spin_lock_bh(&ifp->lock);
1211         state = ifp->state;
1212         ifp->state = INET6_IFADDR_STATE_DEAD;
1213         spin_unlock_bh(&ifp->lock);
1214
1215         if (state == INET6_IFADDR_STATE_DEAD)
1216                 goto out;
1217
1218         spin_lock_bh(&addrconf_hash_lock);
1219         hlist_del_init_rcu(&ifp->addr_lst);
1220         spin_unlock_bh(&addrconf_hash_lock);
1221
1222         write_lock_bh(&ifp->idev->lock);
1223
1224         if (ifp->flags&IFA_F_TEMPORARY) {
1225                 list_del(&ifp->tmp_list);
1226                 if (ifp->ifpub) {
1227                         in6_ifa_put(ifp->ifpub);
1228                         ifp->ifpub = NULL;
1229                 }
1230                 __in6_ifa_put(ifp);
1231         }
1232
1233         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1234                 action = check_cleanup_prefix_route(ifp, &expires);
1235
1236         list_del_rcu(&ifp->if_list);
1237         __in6_ifa_put(ifp);
1238
1239         write_unlock_bh(&ifp->idev->lock);
1240
1241         addrconf_del_dad_work(ifp);
1242
1243         ipv6_ifa_notify(RTM_DELADDR, ifp);
1244
1245         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1246
1247         if (action != CLEANUP_PREFIX_RT_NOP) {
1248                 cleanup_prefix_route(ifp, expires,
1249                         action == CLEANUP_PREFIX_RT_DEL);
1250         }
1251
1252         /* clean up prefsrc entries */
1253         rt6_remove_prefsrc(ifp);
1254 out:
1255         in6_ifa_put(ifp);
1256 }
1257
1258 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1259                                 struct inet6_ifaddr *ift,
1260                                 bool block)
1261 {
1262         struct inet6_dev *idev = ifp->idev;
1263         struct in6_addr addr, *tmpaddr;
1264         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1265         unsigned long regen_advance;
1266         int tmp_plen;
1267         int ret = 0;
1268         u32 addr_flags;
1269         unsigned long now = jiffies;
1270         long max_desync_factor;
1271         s32 cnf_temp_preferred_lft;
1272
1273         write_lock_bh(&idev->lock);
1274         if (ift) {
1275                 spin_lock_bh(&ift->lock);
1276                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1277                 spin_unlock_bh(&ift->lock);
1278                 tmpaddr = &addr;
1279         } else {
1280                 tmpaddr = NULL;
1281         }
1282 retry:
1283         in6_dev_hold(idev);
1284         if (idev->cnf.use_tempaddr <= 0) {
1285                 write_unlock_bh(&idev->lock);
1286                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1287                 in6_dev_put(idev);
1288                 ret = -1;
1289                 goto out;
1290         }
1291         spin_lock_bh(&ifp->lock);
1292         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1293                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1294                 spin_unlock_bh(&ifp->lock);
1295                 write_unlock_bh(&idev->lock);
1296                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1297                         __func__);
1298                 in6_dev_put(idev);
1299                 ret = -1;
1300                 goto out;
1301         }
1302         in6_ifa_hold(ifp);
1303         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1304         ipv6_try_regen_rndid(idev, tmpaddr);
1305         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1306         age = (now - ifp->tstamp) / HZ;
1307
1308         regen_advance = idev->cnf.regen_max_retry *
1309                         idev->cnf.dad_transmits *
1310                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1311
1312         /* recalculate max_desync_factor each time and update
1313          * idev->desync_factor if it's larger
1314          */
1315         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1316         max_desync_factor = min_t(__u32,
1317                                   idev->cnf.max_desync_factor,
1318                                   cnf_temp_preferred_lft - regen_advance);
1319
1320         if (unlikely(idev->desync_factor > max_desync_factor)) {
1321                 if (max_desync_factor > 0) {
1322                         get_random_bytes(&idev->desync_factor,
1323                                          sizeof(idev->desync_factor));
1324                         idev->desync_factor %= max_desync_factor;
1325                 } else {
1326                         idev->desync_factor = 0;
1327                 }
1328         }
1329
1330         tmp_valid_lft = min_t(__u32,
1331                               ifp->valid_lft,
1332                               idev->cnf.temp_valid_lft + age);
1333         tmp_prefered_lft = cnf_temp_preferred_lft + age -
1334                             idev->desync_factor;
1335         tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1336         tmp_plen = ifp->prefix_len;
1337         tmp_tstamp = ifp->tstamp;
1338         spin_unlock_bh(&ifp->lock);
1339
1340         write_unlock_bh(&idev->lock);
1341
1342         /* A temporary address is created only if this calculated Preferred
1343          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1344          * an implementation must not create a temporary address with a zero
1345          * Preferred Lifetime.
1346          * Use age calculation as in addrconf_verify to avoid unnecessary
1347          * temporary addresses being generated.
1348          */
1349         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1350         if (tmp_prefered_lft <= regen_advance + age) {
1351                 in6_ifa_put(ifp);
1352                 in6_dev_put(idev);
1353                 ret = -1;
1354                 goto out;
1355         }
1356
1357         addr_flags = IFA_F_TEMPORARY;
1358         /* set in addrconf_prefix_rcv() */
1359         if (ifp->flags & IFA_F_OPTIMISTIC)
1360                 addr_flags |= IFA_F_OPTIMISTIC;
1361
1362         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1363                             ipv6_addr_scope(&addr), addr_flags,
1364                             tmp_valid_lft, tmp_prefered_lft, block, NULL);
1365         if (IS_ERR(ift)) {
1366                 in6_ifa_put(ifp);
1367                 in6_dev_put(idev);
1368                 pr_info("%s: retry temporary address regeneration\n", __func__);
1369                 tmpaddr = &addr;
1370                 write_lock_bh(&idev->lock);
1371                 goto retry;
1372         }
1373
1374         spin_lock_bh(&ift->lock);
1375         ift->ifpub = ifp;
1376         ift->cstamp = now;
1377         ift->tstamp = tmp_tstamp;
1378         spin_unlock_bh(&ift->lock);
1379
1380         addrconf_dad_start(ift);
1381         in6_ifa_put(ift);
1382         in6_dev_put(idev);
1383 out:
1384         return ret;
1385 }
1386
1387 /*
1388  *      Choose an appropriate source address (RFC3484)
1389  */
1390 enum {
1391         IPV6_SADDR_RULE_INIT = 0,
1392         IPV6_SADDR_RULE_LOCAL,
1393         IPV6_SADDR_RULE_SCOPE,
1394         IPV6_SADDR_RULE_PREFERRED,
1395 #ifdef CONFIG_IPV6_MIP6
1396         IPV6_SADDR_RULE_HOA,
1397 #endif
1398         IPV6_SADDR_RULE_OIF,
1399         IPV6_SADDR_RULE_LABEL,
1400         IPV6_SADDR_RULE_PRIVACY,
1401         IPV6_SADDR_RULE_ORCHID,
1402         IPV6_SADDR_RULE_PREFIX,
1403 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1404         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1405 #endif
1406         IPV6_SADDR_RULE_MAX
1407 };
1408
1409 struct ipv6_saddr_score {
1410         int                     rule;
1411         int                     addr_type;
1412         struct inet6_ifaddr     *ifa;
1413         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1414         int                     scopedist;
1415         int                     matchlen;
1416 };
1417
1418 struct ipv6_saddr_dst {
1419         const struct in6_addr *addr;
1420         int ifindex;
1421         int scope;
1422         int label;
1423         unsigned int prefs;
1424 };
1425
1426 static inline int ipv6_saddr_preferred(int type)
1427 {
1428         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1429                 return 1;
1430         return 0;
1431 }
1432
1433 static bool ipv6_use_optimistic_addr(struct net *net,
1434                                      struct inet6_dev *idev)
1435 {
1436 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1437         if (!idev)
1438                 return false;
1439         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1440                 return false;
1441         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1442                 return false;
1443
1444         return true;
1445 #else
1446         return false;
1447 #endif
1448 }
1449
1450 static bool ipv6_allow_optimistic_dad(struct net *net,
1451                                       struct inet6_dev *idev)
1452 {
1453 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1454         if (!idev)
1455                 return false;
1456         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1457                 return false;
1458
1459         return true;
1460 #else
1461         return false;
1462 #endif
1463 }
1464
1465 static int ipv6_get_saddr_eval(struct net *net,
1466                                struct ipv6_saddr_score *score,
1467                                struct ipv6_saddr_dst *dst,
1468                                int i)
1469 {
1470         int ret;
1471
1472         if (i <= score->rule) {
1473                 switch (i) {
1474                 case IPV6_SADDR_RULE_SCOPE:
1475                         ret = score->scopedist;
1476                         break;
1477                 case IPV6_SADDR_RULE_PREFIX:
1478                         ret = score->matchlen;
1479                         break;
1480                 default:
1481                         ret = !!test_bit(i, score->scorebits);
1482                 }
1483                 goto out;
1484         }
1485
1486         switch (i) {
1487         case IPV6_SADDR_RULE_INIT:
1488                 /* Rule 0: remember if hiscore is not ready yet */
1489                 ret = !!score->ifa;
1490                 break;
1491         case IPV6_SADDR_RULE_LOCAL:
1492                 /* Rule 1: Prefer same address */
1493                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1494                 break;
1495         case IPV6_SADDR_RULE_SCOPE:
1496                 /* Rule 2: Prefer appropriate scope
1497                  *
1498                  *      ret
1499                  *       ^
1500                  *    -1 |  d 15
1501                  *    ---+--+-+---> scope
1502                  *       |
1503                  *       |             d is scope of the destination.
1504                  *  B-d  |  \
1505                  *       |   \      <- smaller scope is better if
1506                  *  B-15 |    \        if scope is enough for destination.
1507                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1508                  * d-C-1 | /
1509                  *       |/         <- greater is better
1510                  *   -C  /             if scope is not enough for destination.
1511                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1512                  *
1513                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1514                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1515                  * Assume B = 0 and we get C > 29.
1516                  */
1517                 ret = __ipv6_addr_src_scope(score->addr_type);
1518                 if (ret >= dst->scope)
1519                         ret = -ret;
1520                 else
1521                         ret -= 128;     /* 30 is enough */
1522                 score->scopedist = ret;
1523                 break;
1524         case IPV6_SADDR_RULE_PREFERRED:
1525             {
1526                 /* Rule 3: Avoid deprecated and optimistic addresses */
1527                 u8 avoid = IFA_F_DEPRECATED;
1528
1529                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1530                         avoid |= IFA_F_OPTIMISTIC;
1531                 ret = ipv6_saddr_preferred(score->addr_type) ||
1532                       !(score->ifa->flags & avoid);
1533                 break;
1534             }
1535 #ifdef CONFIG_IPV6_MIP6
1536         case IPV6_SADDR_RULE_HOA:
1537             {
1538                 /* Rule 4: Prefer home address */
1539                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1540                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1541                 break;
1542             }
1543 #endif
1544         case IPV6_SADDR_RULE_OIF:
1545                 /* Rule 5: Prefer outgoing interface */
1546                 ret = (!dst->ifindex ||
1547                        dst->ifindex == score->ifa->idev->dev->ifindex);
1548                 break;
1549         case IPV6_SADDR_RULE_LABEL:
1550                 /* Rule 6: Prefer matching label */
1551                 ret = ipv6_addr_label(net,
1552                                       &score->ifa->addr, score->addr_type,
1553                                       score->ifa->idev->dev->ifindex) == dst->label;
1554                 break;
1555         case IPV6_SADDR_RULE_PRIVACY:
1556             {
1557                 /* Rule 7: Prefer public address
1558                  * Note: prefer temporary address if use_tempaddr >= 2
1559                  */
1560                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1561                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1562                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1563                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1564                 break;
1565             }
1566         case IPV6_SADDR_RULE_ORCHID:
1567                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1568                  *          non-ORCHID vs non-ORCHID
1569                  */
1570                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1571                         ipv6_addr_orchid(dst->addr));
1572                 break;
1573         case IPV6_SADDR_RULE_PREFIX:
1574                 /* Rule 8: Use longest matching prefix */
1575                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1576                 if (ret > score->ifa->prefix_len)
1577                         ret = score->ifa->prefix_len;
1578                 score->matchlen = ret;
1579                 break;
1580 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1581         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1582                 /* Optimistic addresses still have lower precedence than other
1583                  * preferred addresses.
1584                  */
1585                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1586                 break;
1587 #endif
1588         default:
1589                 ret = 0;
1590         }
1591
1592         if (ret)
1593                 __set_bit(i, score->scorebits);
1594         score->rule = i;
1595 out:
1596         return ret;
1597 }
1598
1599 static int __ipv6_dev_get_saddr(struct net *net,
1600                                 struct ipv6_saddr_dst *dst,
1601                                 struct inet6_dev *idev,
1602                                 struct ipv6_saddr_score *scores,
1603                                 int hiscore_idx)
1604 {
1605         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1606
1607         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1608                 int i;
1609
1610                 /*
1611                  * - Tentative Address (RFC2462 section 5.4)
1612                  *  - A tentative address is not considered
1613                  *    "assigned to an interface" in the traditional
1614                  *    sense, unless it is also flagged as optimistic.
1615                  * - Candidate Source Address (section 4)
1616                  *  - In any case, anycast addresses, multicast
1617                  *    addresses, and the unspecified address MUST
1618                  *    NOT be included in a candidate set.
1619                  */
1620                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1621                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1622                         continue;
1623
1624                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1625
1626                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1627                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1628                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1629                                             idev->dev->name);
1630                         continue;
1631                 }
1632
1633                 score->rule = -1;
1634                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1635
1636                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1637                         int minihiscore, miniscore;
1638
1639                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1640                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1641
1642                         if (minihiscore > miniscore) {
1643                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1644                                     score->scopedist > 0) {
1645                                         /*
1646                                          * special case:
1647                                          * each remaining entry
1648                                          * has too small (not enough)
1649                                          * scope, because ifa entries
1650                                          * are sorted by their scope
1651                                          * values.
1652                                          */
1653                                         goto out;
1654                                 }
1655                                 break;
1656                         } else if (minihiscore < miniscore) {
1657                                 swap(hiscore, score);
1658                                 hiscore_idx = 1 - hiscore_idx;
1659
1660                                 /* restore our iterator */
1661                                 score->ifa = hiscore->ifa;
1662
1663                                 break;
1664                         }
1665                 }
1666         }
1667 out:
1668         return hiscore_idx;
1669 }
1670
1671 static int ipv6_get_saddr_master(struct net *net,
1672                                  const struct net_device *dst_dev,
1673                                  const struct net_device *master,
1674                                  struct ipv6_saddr_dst *dst,
1675                                  struct ipv6_saddr_score *scores,
1676                                  int hiscore_idx)
1677 {
1678         struct inet6_dev *idev;
1679
1680         idev = __in6_dev_get(dst_dev);
1681         if (idev)
1682                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1683                                                    scores, hiscore_idx);
1684
1685         idev = __in6_dev_get(master);
1686         if (idev)
1687                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1688                                                    scores, hiscore_idx);
1689
1690         return hiscore_idx;
1691 }
1692
1693 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1694                        const struct in6_addr *daddr, unsigned int prefs,
1695                        struct in6_addr *saddr)
1696 {
1697         struct ipv6_saddr_score scores[2], *hiscore;
1698         struct ipv6_saddr_dst dst;
1699         struct inet6_dev *idev;
1700         struct net_device *dev;
1701         int dst_type;
1702         bool use_oif_addr = false;
1703         int hiscore_idx = 0;
1704         int ret = 0;
1705
1706         dst_type = __ipv6_addr_type(daddr);
1707         dst.addr = daddr;
1708         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1709         dst.scope = __ipv6_addr_src_scope(dst_type);
1710         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1711         dst.prefs = prefs;
1712
1713         scores[hiscore_idx].rule = -1;
1714         scores[hiscore_idx].ifa = NULL;
1715
1716         rcu_read_lock();
1717
1718         /* Candidate Source Address (section 4)
1719          *  - multicast and link-local destination address,
1720          *    the set of candidate source address MUST only
1721          *    include addresses assigned to interfaces
1722          *    belonging to the same link as the outgoing
1723          *    interface.
1724          * (- For site-local destination addresses, the
1725          *    set of candidate source addresses MUST only
1726          *    include addresses assigned to interfaces
1727          *    belonging to the same site as the outgoing
1728          *    interface.)
1729          *  - "It is RECOMMENDED that the candidate source addresses
1730          *    be the set of unicast addresses assigned to the
1731          *    interface that will be used to send to the destination
1732          *    (the 'outgoing' interface)." (RFC 6724)
1733          */
1734         if (dst_dev) {
1735                 idev = __in6_dev_get(dst_dev);
1736                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1737                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1738                     (idev && idev->cnf.use_oif_addrs_only)) {
1739                         use_oif_addr = true;
1740                 }
1741         }
1742
1743         if (use_oif_addr) {
1744                 if (idev)
1745                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1746         } else {
1747                 const struct net_device *master;
1748                 int master_idx = 0;
1749
1750                 /* if dst_dev exists and is enslaved to an L3 device, then
1751                  * prefer addresses from dst_dev and then the master over
1752                  * any other enslaved devices in the L3 domain.
1753                  */
1754                 master = l3mdev_master_dev_rcu(dst_dev);
1755                 if (master) {
1756                         master_idx = master->ifindex;
1757
1758                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1759                                                             master, &dst,
1760                                                             scores, hiscore_idx);
1761
1762                         if (scores[hiscore_idx].ifa)
1763                                 goto out;
1764                 }
1765
1766                 for_each_netdev_rcu(net, dev) {
1767                         /* only consider addresses on devices in the
1768                          * same L3 domain
1769                          */
1770                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1771                                 continue;
1772                         idev = __in6_dev_get(dev);
1773                         if (!idev)
1774                                 continue;
1775                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1776                 }
1777         }
1778
1779 out:
1780         hiscore = &scores[hiscore_idx];
1781         if (!hiscore->ifa)
1782                 ret = -EADDRNOTAVAIL;
1783         else
1784                 *saddr = hiscore->ifa->addr;
1785
1786         rcu_read_unlock();
1787         return ret;
1788 }
1789 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1790
1791 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1792                       u32 banned_flags)
1793 {
1794         struct inet6_ifaddr *ifp;
1795         int err = -EADDRNOTAVAIL;
1796
1797         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1798                 if (ifp->scope > IFA_LINK)
1799                         break;
1800                 if (ifp->scope == IFA_LINK &&
1801                     !(ifp->flags & banned_flags)) {
1802                         *addr = ifp->addr;
1803                         err = 0;
1804                         break;
1805                 }
1806         }
1807         return err;
1808 }
1809
1810 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1811                     u32 banned_flags)
1812 {
1813         struct inet6_dev *idev;
1814         int err = -EADDRNOTAVAIL;
1815
1816         rcu_read_lock();
1817         idev = __in6_dev_get(dev);
1818         if (idev) {
1819                 read_lock_bh(&idev->lock);
1820                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1821                 read_unlock_bh(&idev->lock);
1822         }
1823         rcu_read_unlock();
1824         return err;
1825 }
1826
1827 static int ipv6_count_addresses(const struct inet6_dev *idev)
1828 {
1829         const struct inet6_ifaddr *ifp;
1830         int cnt = 0;
1831
1832         rcu_read_lock();
1833         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1834                 cnt++;
1835         rcu_read_unlock();
1836         return cnt;
1837 }
1838
1839 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1840                   const struct net_device *dev, int strict)
1841 {
1842         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1843                                        strict, IFA_F_TENTATIVE);
1844 }
1845 EXPORT_SYMBOL(ipv6_chk_addr);
1846
1847 /* device argument is used to find the L3 domain of interest. If
1848  * skip_dev_check is set, then the ifp device is not checked against
1849  * the passed in dev argument. So the 2 cases for addresses checks are:
1850  *   1. does the address exist in the L3 domain that dev is part of
1851  *      (skip_dev_check = true), or
1852  *
1853  *   2. does the address exist on the specific device
1854  *      (skip_dev_check = false)
1855  */
1856 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1857                             const struct net_device *dev, bool skip_dev_check,
1858                             int strict, u32 banned_flags)
1859 {
1860         unsigned int hash = inet6_addr_hash(net, addr);
1861         const struct net_device *l3mdev;
1862         struct inet6_ifaddr *ifp;
1863         u32 ifp_flags;
1864
1865         rcu_read_lock();
1866
1867         l3mdev = l3mdev_master_dev_rcu(dev);
1868         if (skip_dev_check)
1869                 dev = NULL;
1870
1871         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1872                 if (!net_eq(dev_net(ifp->idev->dev), net))
1873                         continue;
1874
1875                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1876                         continue;
1877
1878                 /* Decouple optimistic from tentative for evaluation here.
1879                  * Ban optimistic addresses explicitly, when required.
1880                  */
1881                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1882                             ? (ifp->flags&~IFA_F_TENTATIVE)
1883                             : ifp->flags;
1884                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1885                     !(ifp_flags&banned_flags) &&
1886                     (!dev || ifp->idev->dev == dev ||
1887                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1888                         rcu_read_unlock();
1889                         return 1;
1890                 }
1891         }
1892
1893         rcu_read_unlock();
1894         return 0;
1895 }
1896 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1897
1898
1899 /* Compares an address/prefix_len with addresses on device @dev.
1900  * If one is found it returns true.
1901  */
1902 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1903         const unsigned int prefix_len, struct net_device *dev)
1904 {
1905         const struct inet6_ifaddr *ifa;
1906         const struct inet6_dev *idev;
1907         bool ret = false;
1908
1909         rcu_read_lock();
1910         idev = __in6_dev_get(dev);
1911         if (idev) {
1912                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1913                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1914                         if (ret)
1915                                 break;
1916                 }
1917         }
1918         rcu_read_unlock();
1919
1920         return ret;
1921 }
1922 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1923
1924 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1925 {
1926         const struct inet6_ifaddr *ifa;
1927         const struct inet6_dev *idev;
1928         int     onlink;
1929
1930         onlink = 0;
1931         rcu_read_lock();
1932         idev = __in6_dev_get(dev);
1933         if (idev) {
1934                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1935                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1936                                                    ifa->prefix_len);
1937                         if (onlink)
1938                                 break;
1939                 }
1940         }
1941         rcu_read_unlock();
1942         return onlink;
1943 }
1944 EXPORT_SYMBOL(ipv6_chk_prefix);
1945
1946 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1947                                      struct net_device *dev, int strict)
1948 {
1949         unsigned int hash = inet6_addr_hash(net, addr);
1950         struct inet6_ifaddr *ifp, *result = NULL;
1951
1952         rcu_read_lock();
1953         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1954                 if (!net_eq(dev_net(ifp->idev->dev), net))
1955                         continue;
1956                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1957                         if (!dev || ifp->idev->dev == dev ||
1958                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1959                                 result = ifp;
1960                                 in6_ifa_hold(ifp);
1961                                 break;
1962                         }
1963                 }
1964         }
1965         rcu_read_unlock();
1966
1967         return result;
1968 }
1969
1970 /* Gets referenced address, destroys ifaddr */
1971
1972 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1973 {
1974         if (dad_failed)
1975                 ifp->flags |= IFA_F_DADFAILED;
1976
1977         if (ifp->flags&IFA_F_TEMPORARY) {
1978                 struct inet6_ifaddr *ifpub;
1979                 spin_lock_bh(&ifp->lock);
1980                 ifpub = ifp->ifpub;
1981                 if (ifpub) {
1982                         in6_ifa_hold(ifpub);
1983                         spin_unlock_bh(&ifp->lock);
1984                         ipv6_create_tempaddr(ifpub, ifp, true);
1985                         in6_ifa_put(ifpub);
1986                 } else {
1987                         spin_unlock_bh(&ifp->lock);
1988                 }
1989                 ipv6_del_addr(ifp);
1990         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1991                 spin_lock_bh(&ifp->lock);
1992                 addrconf_del_dad_work(ifp);
1993                 ifp->flags |= IFA_F_TENTATIVE;
1994                 if (dad_failed)
1995                         ifp->flags &= ~IFA_F_OPTIMISTIC;
1996                 spin_unlock_bh(&ifp->lock);
1997                 if (dad_failed)
1998                         ipv6_ifa_notify(0, ifp);
1999                 in6_ifa_put(ifp);
2000         } else {
2001                 ipv6_del_addr(ifp);
2002         }
2003 }
2004
2005 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2006 {
2007         int err = -ENOENT;
2008
2009         spin_lock_bh(&ifp->lock);
2010         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2011                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2012                 err = 0;
2013         }
2014         spin_unlock_bh(&ifp->lock);
2015
2016         return err;
2017 }
2018
2019 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2020 {
2021         struct inet6_dev *idev = ifp->idev;
2022         struct net *net = dev_net(ifp->idev->dev);
2023
2024         if (addrconf_dad_end(ifp)) {
2025                 in6_ifa_put(ifp);
2026                 return;
2027         }
2028
2029         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2030                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2031
2032         spin_lock_bh(&ifp->lock);
2033
2034         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2035                 int scope = ifp->scope;
2036                 u32 flags = ifp->flags;
2037                 struct in6_addr new_addr;
2038                 struct inet6_ifaddr *ifp2;
2039                 u32 valid_lft, preferred_lft;
2040                 int pfxlen = ifp->prefix_len;
2041                 int retries = ifp->stable_privacy_retry + 1;
2042
2043                 if (retries > net->ipv6.sysctl.idgen_retries) {
2044                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2045                                              ifp->idev->dev->name);
2046                         goto errdad;
2047                 }
2048
2049                 new_addr = ifp->addr;
2050                 if (ipv6_generate_stable_address(&new_addr, retries,
2051                                                  idev))
2052                         goto errdad;
2053
2054                 valid_lft = ifp->valid_lft;
2055                 preferred_lft = ifp->prefered_lft;
2056
2057                 spin_unlock_bh(&ifp->lock);
2058
2059                 if (idev->cnf.max_addresses &&
2060                     ipv6_count_addresses(idev) >=
2061                     idev->cnf.max_addresses)
2062                         goto lock_errdad;
2063
2064                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2065                                      ifp->idev->dev->name);
2066
2067                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2068                                      scope, flags, valid_lft,
2069                                      preferred_lft, false, NULL);
2070                 if (IS_ERR(ifp2))
2071                         goto lock_errdad;
2072
2073                 spin_lock_bh(&ifp2->lock);
2074                 ifp2->stable_privacy_retry = retries;
2075                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2076                 spin_unlock_bh(&ifp2->lock);
2077
2078                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2079                 in6_ifa_put(ifp2);
2080 lock_errdad:
2081                 spin_lock_bh(&ifp->lock);
2082         }
2083
2084 errdad:
2085         /* transition from _POSTDAD to _ERRDAD */
2086         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2087         spin_unlock_bh(&ifp->lock);
2088
2089         addrconf_mod_dad_work(ifp, 0);
2090         in6_ifa_put(ifp);
2091 }
2092
2093 /* Join to solicited addr multicast group.
2094  * caller must hold RTNL */
2095 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2096 {
2097         struct in6_addr maddr;
2098
2099         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2100                 return;
2101
2102         addrconf_addr_solict_mult(addr, &maddr);
2103         ipv6_dev_mc_inc(dev, &maddr);
2104 }
2105
2106 /* caller must hold RTNL */
2107 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2108 {
2109         struct in6_addr maddr;
2110
2111         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2112                 return;
2113
2114         addrconf_addr_solict_mult(addr, &maddr);
2115         __ipv6_dev_mc_dec(idev, &maddr);
2116 }
2117
2118 /* caller must hold RTNL */
2119 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2120 {
2121         struct in6_addr addr;
2122
2123         if (ifp->prefix_len >= 127) /* RFC 6164 */
2124                 return;
2125         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2126         if (ipv6_addr_any(&addr))
2127                 return;
2128         __ipv6_dev_ac_inc(ifp->idev, &addr);
2129 }
2130
2131 /* caller must hold RTNL */
2132 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2133 {
2134         struct in6_addr addr;
2135
2136         if (ifp->prefix_len >= 127) /* RFC 6164 */
2137                 return;
2138         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2139         if (ipv6_addr_any(&addr))
2140                 return;
2141         __ipv6_dev_ac_dec(ifp->idev, &addr);
2142 }
2143
2144 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2145 {
2146         switch (dev->addr_len) {
2147         case ETH_ALEN:
2148                 memcpy(eui, dev->dev_addr, 3);
2149                 eui[3] = 0xFF;
2150                 eui[4] = 0xFE;
2151                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2152                 break;
2153         case EUI64_ADDR_LEN:
2154                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2155                 eui[0] ^= 2;
2156                 break;
2157         default:
2158                 return -1;
2159         }
2160
2161         return 0;
2162 }
2163
2164 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2165 {
2166         union fwnet_hwaddr *ha;
2167
2168         if (dev->addr_len != FWNET_ALEN)
2169                 return -1;
2170
2171         ha = (union fwnet_hwaddr *)dev->dev_addr;
2172
2173         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2174         eui[0] ^= 2;
2175         return 0;
2176 }
2177
2178 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2179 {
2180         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2181         if (dev->addr_len != ARCNET_ALEN)
2182                 return -1;
2183         memset(eui, 0, 7);
2184         eui[7] = *(u8 *)dev->dev_addr;
2185         return 0;
2186 }
2187
2188 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2189 {
2190         if (dev->addr_len != INFINIBAND_ALEN)
2191                 return -1;
2192         memcpy(eui, dev->dev_addr + 12, 8);
2193         eui[0] |= 2;
2194         return 0;
2195 }
2196
2197 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2198 {
2199         if (addr == 0)
2200                 return -1;
2201         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2202                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2203                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2204                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2205                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2206                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2207         eui[1] = 0;
2208         eui[2] = 0x5E;
2209         eui[3] = 0xFE;
2210         memcpy(eui + 4, &addr, 4);
2211         return 0;
2212 }
2213
2214 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2215 {
2216         if (dev->priv_flags & IFF_ISATAP)
2217                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2218         return -1;
2219 }
2220
2221 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2222 {
2223         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2224 }
2225
2226 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2227 {
2228         memcpy(eui, dev->perm_addr, 3);
2229         memcpy(eui + 5, dev->perm_addr + 3, 3);
2230         eui[3] = 0xFF;
2231         eui[4] = 0xFE;
2232         eui[0] ^= 2;
2233         return 0;
2234 }
2235
2236 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2237 {
2238         switch (dev->type) {
2239         case ARPHRD_ETHER:
2240         case ARPHRD_FDDI:
2241                 return addrconf_ifid_eui48(eui, dev);
2242         case ARPHRD_ARCNET:
2243                 return addrconf_ifid_arcnet(eui, dev);
2244         case ARPHRD_INFINIBAND:
2245                 return addrconf_ifid_infiniband(eui, dev);
2246         case ARPHRD_SIT:
2247                 return addrconf_ifid_sit(eui, dev);
2248         case ARPHRD_IPGRE:
2249         case ARPHRD_TUNNEL:
2250                 return addrconf_ifid_gre(eui, dev);
2251         case ARPHRD_6LOWPAN:
2252                 return addrconf_ifid_6lowpan(eui, dev);
2253         case ARPHRD_IEEE1394:
2254                 return addrconf_ifid_ieee1394(eui, dev);
2255         case ARPHRD_TUNNEL6:
2256         case ARPHRD_IP6GRE:
2257                 return addrconf_ifid_ip6tnl(eui, dev);
2258         }
2259         return -1;
2260 }
2261
2262 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2263 {
2264         int err = -1;
2265         struct inet6_ifaddr *ifp;
2266
2267         read_lock_bh(&idev->lock);
2268         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2269                 if (ifp->scope > IFA_LINK)
2270                         break;
2271                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2272                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2273                         err = 0;
2274                         break;
2275                 }
2276         }
2277         read_unlock_bh(&idev->lock);
2278         return err;
2279 }
2280
2281 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2282 static void ipv6_regen_rndid(struct inet6_dev *idev)
2283 {
2284 regen:
2285         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2286         idev->rndid[0] &= ~0x02;
2287
2288         /*
2289          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2290          * check if generated address is not inappropriate
2291          *
2292          *  - Reserved subnet anycast (RFC 2526)
2293          *      11111101 11....11 1xxxxxxx
2294          *  - ISATAP (RFC4214) 6.1
2295          *      00-00-5E-FE-xx-xx-xx-xx
2296          *  - value 0
2297          *  - XXX: already assigned to an address on the device
2298          */
2299         if (idev->rndid[0] == 0xfd &&
2300             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2301             (idev->rndid[7]&0x80))
2302                 goto regen;
2303         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2304                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2305                         goto regen;
2306                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2307                         goto regen;
2308         }
2309 }
2310
2311 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2312 {
2313         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2314                 ipv6_regen_rndid(idev);
2315 }
2316
2317 /*
2318  *      Add prefix route.
2319  */
2320
2321 static void
2322 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2323                       unsigned long expires, u32 flags)
2324 {
2325         struct fib6_config cfg = {
2326                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2327                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2328                 .fc_ifindex = dev->ifindex,
2329                 .fc_expires = expires,
2330                 .fc_dst_len = plen,
2331                 .fc_flags = RTF_UP | flags,
2332                 .fc_nlinfo.nl_net = dev_net(dev),
2333                 .fc_protocol = RTPROT_KERNEL,
2334         };
2335
2336         cfg.fc_dst = *pfx;
2337
2338         /* Prevent useless cloning on PtP SIT.
2339            This thing is done here expecting that the whole
2340            class of non-broadcast devices need not cloning.
2341          */
2342 #if IS_ENABLED(CONFIG_IPV6_SIT)
2343         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2344                 cfg.fc_flags |= RTF_NONEXTHOP;
2345 #endif
2346
2347         ip6_route_add(&cfg, NULL);
2348 }
2349
2350
2351 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2352                                                   int plen,
2353                                                   const struct net_device *dev,
2354                                                   u32 flags, u32 noflags)
2355 {
2356         struct fib6_node *fn;
2357         struct rt6_info *rt = NULL;
2358         struct fib6_table *table;
2359         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2360
2361         table = fib6_get_table(dev_net(dev), tb_id);
2362         if (!table)
2363                 return NULL;
2364
2365         rcu_read_lock();
2366         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2367         if (!fn)
2368                 goto out;
2369
2370         for_each_fib6_node_rt_rcu(fn) {
2371                 if (rt->dst.dev->ifindex != dev->ifindex)
2372                         continue;
2373                 if ((rt->rt6i_flags & flags) != flags)
2374                         continue;
2375                 if ((rt->rt6i_flags & noflags) != 0)
2376                         continue;
2377                 if (!dst_hold_safe(&rt->dst))
2378                         rt = NULL;
2379                 break;
2380         }
2381 out:
2382         rcu_read_unlock();
2383         return rt;
2384 }
2385
2386
2387 /* Create "default" multicast route to the interface */
2388
2389 static void addrconf_add_mroute(struct net_device *dev)
2390 {
2391         struct fib6_config cfg = {
2392                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2393                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2394                 .fc_ifindex = dev->ifindex,
2395                 .fc_dst_len = 8,
2396                 .fc_flags = RTF_UP,
2397                 .fc_nlinfo.nl_net = dev_net(dev),
2398         };
2399
2400         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2401
2402         ip6_route_add(&cfg, NULL);
2403 }
2404
2405 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2406 {
2407         struct inet6_dev *idev;
2408
2409         ASSERT_RTNL();
2410
2411         idev = ipv6_find_idev(dev);
2412         if (!idev)
2413                 return ERR_PTR(-ENOBUFS);
2414
2415         if (idev->cnf.disable_ipv6)
2416                 return ERR_PTR(-EACCES);
2417
2418         /* Add default multicast route */
2419         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2420                 addrconf_add_mroute(dev);
2421
2422         return idev;
2423 }
2424
2425 static void manage_tempaddrs(struct inet6_dev *idev,
2426                              struct inet6_ifaddr *ifp,
2427                              __u32 valid_lft, __u32 prefered_lft,
2428                              bool create, unsigned long now)
2429 {
2430         u32 flags;
2431         struct inet6_ifaddr *ift;
2432
2433         read_lock_bh(&idev->lock);
2434         /* update all temporary addresses in the list */
2435         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2436                 int age, max_valid, max_prefered;
2437
2438                 if (ifp != ift->ifpub)
2439                         continue;
2440
2441                 /* RFC 4941 section 3.3:
2442                  * If a received option will extend the lifetime of a public
2443                  * address, the lifetimes of temporary addresses should
2444                  * be extended, subject to the overall constraint that no
2445                  * temporary addresses should ever remain "valid" or "preferred"
2446                  * for a time longer than (TEMP_VALID_LIFETIME) or
2447                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2448                  */
2449                 age = (now - ift->cstamp) / HZ;
2450                 max_valid = idev->cnf.temp_valid_lft - age;
2451                 if (max_valid < 0)
2452                         max_valid = 0;
2453
2454                 max_prefered = idev->cnf.temp_prefered_lft -
2455                                idev->desync_factor - age;
2456                 if (max_prefered < 0)
2457                         max_prefered = 0;
2458
2459                 if (valid_lft > max_valid)
2460                         valid_lft = max_valid;
2461
2462                 if (prefered_lft > max_prefered)
2463                         prefered_lft = max_prefered;
2464
2465                 spin_lock(&ift->lock);
2466                 flags = ift->flags;
2467                 ift->valid_lft = valid_lft;
2468                 ift->prefered_lft = prefered_lft;
2469                 ift->tstamp = now;
2470                 if (prefered_lft > 0)
2471                         ift->flags &= ~IFA_F_DEPRECATED;
2472
2473                 spin_unlock(&ift->lock);
2474                 if (!(flags&IFA_F_TENTATIVE))
2475                         ipv6_ifa_notify(0, ift);
2476         }
2477
2478         if ((create || list_empty(&idev->tempaddr_list)) &&
2479             idev->cnf.use_tempaddr > 0) {
2480                 /* When a new public address is created as described
2481                  * in [ADDRCONF], also create a new temporary address.
2482                  * Also create a temporary address if it's enabled but
2483                  * no temporary address currently exists.
2484                  */
2485                 read_unlock_bh(&idev->lock);
2486                 ipv6_create_tempaddr(ifp, NULL, false);
2487         } else {
2488                 read_unlock_bh(&idev->lock);
2489         }
2490 }
2491
2492 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2493 {
2494         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2495                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2496 }
2497
2498 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2499                                  const struct prefix_info *pinfo,
2500                                  struct inet6_dev *in6_dev,
2501                                  const struct in6_addr *addr, int addr_type,
2502                                  u32 addr_flags, bool sllao, bool tokenized,
2503                                  __u32 valid_lft, u32 prefered_lft)
2504 {
2505         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2506         int create = 0, update_lft = 0;
2507
2508         if (!ifp && valid_lft) {
2509                 int max_addresses = in6_dev->cnf.max_addresses;
2510
2511 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2512                 if ((net->ipv6.devconf_all->optimistic_dad ||
2513                      in6_dev->cnf.optimistic_dad) &&
2514                     !net->ipv6.devconf_all->forwarding && sllao)
2515                         addr_flags |= IFA_F_OPTIMISTIC;
2516 #endif
2517
2518                 /* Do not allow to create too much of autoconfigured
2519                  * addresses; this would be too easy way to crash kernel.
2520                  */
2521                 if (!max_addresses ||
2522                     ipv6_count_addresses(in6_dev) < max_addresses)
2523                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2524                                             pinfo->prefix_len,
2525                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2526                                             addr_flags, valid_lft,
2527                                             prefered_lft, false, NULL);
2528
2529                 if (IS_ERR_OR_NULL(ifp))
2530                         return -1;
2531
2532                 create = 1;
2533                 spin_lock_bh(&ifp->lock);
2534                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2535                 ifp->cstamp = jiffies;
2536                 ifp->tokenized = tokenized;
2537                 spin_unlock_bh(&ifp->lock);
2538                 addrconf_dad_start(ifp);
2539         }
2540
2541         if (ifp) {
2542                 u32 flags;
2543                 unsigned long now;
2544                 u32 stored_lft;
2545
2546                 /* update lifetime (RFC2462 5.5.3 e) */
2547                 spin_lock_bh(&ifp->lock);
2548                 now = jiffies;
2549                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2550                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2551                 else
2552                         stored_lft = 0;
2553                 if (!create && stored_lft) {
2554                         const u32 minimum_lft = min_t(u32,
2555                                 stored_lft, MIN_VALID_LIFETIME);
2556                         valid_lft = max(valid_lft, minimum_lft);
2557
2558                         /* RFC4862 Section 5.5.3e:
2559                          * "Note that the preferred lifetime of the
2560                          *  corresponding address is always reset to
2561                          *  the Preferred Lifetime in the received
2562                          *  Prefix Information option, regardless of
2563                          *  whether the valid lifetime is also reset or
2564                          *  ignored."
2565                          *
2566                          * So we should always update prefered_lft here.
2567                          */
2568                         update_lft = 1;
2569                 }
2570
2571                 if (update_lft) {
2572                         ifp->valid_lft = valid_lft;
2573                         ifp->prefered_lft = prefered_lft;
2574                         ifp->tstamp = now;
2575                         flags = ifp->flags;
2576                         ifp->flags &= ~IFA_F_DEPRECATED;
2577                         spin_unlock_bh(&ifp->lock);
2578
2579                         if (!(flags&IFA_F_TENTATIVE))
2580                                 ipv6_ifa_notify(0, ifp);
2581                 } else
2582                         spin_unlock_bh(&ifp->lock);
2583
2584                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2585                                  create, now);
2586
2587                 in6_ifa_put(ifp);
2588                 addrconf_verify();
2589         }
2590
2591         return 0;
2592 }
2593 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2594
2595 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2596 {
2597         struct prefix_info *pinfo;
2598         __u32 valid_lft;
2599         __u32 prefered_lft;
2600         int addr_type, err;
2601         u32 addr_flags = 0;
2602         struct inet6_dev *in6_dev;
2603         struct net *net = dev_net(dev);
2604
2605         pinfo = (struct prefix_info *) opt;
2606
2607         if (len < sizeof(struct prefix_info)) {
2608                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2609                 return;
2610         }
2611
2612         /*
2613          *      Validation checks ([ADDRCONF], page 19)
2614          */
2615
2616         addr_type = ipv6_addr_type(&pinfo->prefix);
2617
2618         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2619                 return;
2620
2621         valid_lft = ntohl(pinfo->valid);
2622         prefered_lft = ntohl(pinfo->prefered);
2623
2624         if (prefered_lft > valid_lft) {
2625                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2626                 return;
2627         }
2628
2629         in6_dev = in6_dev_get(dev);
2630
2631         if (!in6_dev) {
2632                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2633                                     dev->name);
2634                 return;
2635         }
2636
2637         /*
2638          *      Two things going on here:
2639          *      1) Add routes for on-link prefixes
2640          *      2) Configure prefixes with the auto flag set
2641          */
2642
2643         if (pinfo->onlink) {
2644                 struct rt6_info *rt;
2645                 unsigned long rt_expires;
2646
2647                 /* Avoid arithmetic overflow. Really, we could
2648                  * save rt_expires in seconds, likely valid_lft,
2649                  * but it would require division in fib gc, that it
2650                  * not good.
2651                  */
2652                 if (HZ > USER_HZ)
2653                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2654                 else
2655                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2656
2657                 if (addrconf_finite_timeout(rt_expires))
2658                         rt_expires *= HZ;
2659
2660                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2661                                                pinfo->prefix_len,
2662                                                dev,
2663                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2664                                                RTF_GATEWAY | RTF_DEFAULT);
2665
2666                 if (rt) {
2667                         /* Autoconf prefix route */
2668                         if (valid_lft == 0) {
2669                                 ip6_del_rt(rt);
2670                                 rt = NULL;
2671                         } else if (addrconf_finite_timeout(rt_expires)) {
2672                                 /* not infinity */
2673                                 rt6_set_expires(rt, jiffies + rt_expires);
2674                         } else {
2675                                 rt6_clean_expires(rt);
2676                         }
2677                 } else if (valid_lft) {
2678                         clock_t expires = 0;
2679                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2680                         if (addrconf_finite_timeout(rt_expires)) {
2681                                 /* not infinity */
2682                                 flags |= RTF_EXPIRES;
2683                                 expires = jiffies_to_clock_t(rt_expires);
2684                         }
2685                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2686                                               dev, expires, flags);
2687                 }
2688                 ip6_rt_put(rt);
2689         }
2690
2691         /* Try to figure out our local address for this prefix */
2692
2693         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2694                 struct in6_addr addr;
2695                 bool tokenized = false, dev_addr_generated = false;
2696
2697                 if (pinfo->prefix_len == 64) {
2698                         memcpy(&addr, &pinfo->prefix, 8);
2699
2700                         if (!ipv6_addr_any(&in6_dev->token)) {
2701                                 read_lock_bh(&in6_dev->lock);
2702                                 memcpy(addr.s6_addr + 8,
2703                                        in6_dev->token.s6_addr + 8, 8);
2704                                 read_unlock_bh(&in6_dev->lock);
2705                                 tokenized = true;
2706                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2707                                    !ipv6_generate_stable_address(&addr, 0,
2708                                                                  in6_dev)) {
2709                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2710                                 goto ok;
2711                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2712                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2713                                 goto put;
2714                         } else {
2715                                 dev_addr_generated = true;
2716                         }
2717                         goto ok;
2718                 }
2719                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2720                                     pinfo->prefix_len);
2721                 goto put;
2722
2723 ok:
2724                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2725                                                    &addr, addr_type,
2726                                                    addr_flags, sllao,
2727                                                    tokenized, valid_lft,
2728                                                    prefered_lft);
2729                 if (err)
2730                         goto put;
2731
2732                 /* Ignore error case here because previous prefix add addr was
2733                  * successful which will be notified.
2734                  */
2735                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2736                                               addr_type, addr_flags, sllao,
2737                                               tokenized, valid_lft,
2738                                               prefered_lft,
2739                                               dev_addr_generated);
2740         }
2741         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2742 put:
2743         in6_dev_put(in6_dev);
2744 }
2745
2746 /*
2747  *      Set destination address.
2748  *      Special case for SIT interfaces where we create a new "virtual"
2749  *      device.
2750  */
2751 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2752 {
2753         struct in6_ifreq ireq;
2754         struct net_device *dev;
2755         int err = -EINVAL;
2756
2757         rtnl_lock();
2758
2759         err = -EFAULT;
2760         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2761                 goto err_exit;
2762
2763         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2764
2765         err = -ENODEV;
2766         if (!dev)
2767                 goto err_exit;
2768
2769 #if IS_ENABLED(CONFIG_IPV6_SIT)
2770         if (dev->type == ARPHRD_SIT) {
2771                 const struct net_device_ops *ops = dev->netdev_ops;
2772                 struct ifreq ifr;
2773                 struct ip_tunnel_parm p;
2774
2775                 err = -EADDRNOTAVAIL;
2776                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2777                         goto err_exit;
2778
2779                 memset(&p, 0, sizeof(p));
2780                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2781                 p.iph.saddr = 0;
2782                 p.iph.version = 4;
2783                 p.iph.ihl = 5;
2784                 p.iph.protocol = IPPROTO_IPV6;
2785                 p.iph.ttl = 64;
2786                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2787
2788                 if (ops->ndo_do_ioctl) {
2789                         mm_segment_t oldfs = get_fs();
2790
2791                         set_fs(KERNEL_DS);
2792                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2793                         set_fs(oldfs);
2794                 } else
2795                         err = -EOPNOTSUPP;
2796
2797                 if (err == 0) {
2798                         err = -ENOBUFS;
2799                         dev = __dev_get_by_name(net, p.name);
2800                         if (!dev)
2801                                 goto err_exit;
2802                         err = dev_open(dev);
2803                 }
2804         }
2805 #endif
2806
2807 err_exit:
2808         rtnl_unlock();
2809         return err;
2810 }
2811
2812 static int ipv6_mc_config(struct sock *sk, bool join,
2813                           const struct in6_addr *addr, int ifindex)
2814 {
2815         int ret;
2816
2817         ASSERT_RTNL();
2818
2819         lock_sock(sk);
2820         if (join)
2821                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2822         else
2823                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2824         release_sock(sk);
2825
2826         return ret;
2827 }
2828
2829 /*
2830  *      Manual configuration of address on an interface
2831  */
2832 static int inet6_addr_add(struct net *net, int ifindex,
2833                           const struct in6_addr *pfx,
2834                           const struct in6_addr *peer_pfx,
2835                           unsigned int plen, __u32 ifa_flags,
2836                           __u32 prefered_lft, __u32 valid_lft,
2837                           struct netlink_ext_ack *extack)
2838 {
2839         struct inet6_ifaddr *ifp;
2840         struct inet6_dev *idev;
2841         struct net_device *dev;
2842         unsigned long timeout;
2843         clock_t expires;
2844         int scope;
2845         u32 flags;
2846
2847         ASSERT_RTNL();
2848
2849         if (plen > 128)
2850                 return -EINVAL;
2851
2852         /* check the lifetime */
2853         if (!valid_lft || prefered_lft > valid_lft)
2854                 return -EINVAL;
2855
2856         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2857                 return -EINVAL;
2858
2859         dev = __dev_get_by_index(net, ifindex);
2860         if (!dev)
2861                 return -ENODEV;
2862
2863         idev = addrconf_add_dev(dev);
2864         if (IS_ERR(idev))
2865                 return PTR_ERR(idev);
2866
2867         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2868                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2869                                          true, pfx, ifindex);
2870
2871                 if (ret < 0)
2872                         return ret;
2873         }
2874
2875         scope = ipv6_addr_scope(pfx);
2876
2877         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2878         if (addrconf_finite_timeout(timeout)) {
2879                 expires = jiffies_to_clock_t(timeout * HZ);
2880                 valid_lft = timeout;
2881                 flags = RTF_EXPIRES;
2882         } else {
2883                 expires = 0;
2884                 flags = 0;
2885                 ifa_flags |= IFA_F_PERMANENT;
2886         }
2887
2888         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2889         if (addrconf_finite_timeout(timeout)) {
2890                 if (timeout == 0)
2891                         ifa_flags |= IFA_F_DEPRECATED;
2892                 prefered_lft = timeout;
2893         }
2894
2895         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2896                             valid_lft, prefered_lft, true, extack);
2897
2898         if (!IS_ERR(ifp)) {
2899                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2900                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2901                                               expires, flags);
2902                 }
2903
2904                 /* Send a netlink notification if DAD is enabled and
2905                  * optimistic flag is not set
2906                  */
2907                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2908                         ipv6_ifa_notify(0, ifp);
2909                 /*
2910                  * Note that section 3.1 of RFC 4429 indicates
2911                  * that the Optimistic flag should not be set for
2912                  * manually configured addresses
2913                  */
2914                 addrconf_dad_start(ifp);
2915                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2916                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2917                                          true, jiffies);
2918                 in6_ifa_put(ifp);
2919                 addrconf_verify_rtnl();
2920                 return 0;
2921         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2922                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2923                                false, pfx, ifindex);
2924         }
2925
2926         return PTR_ERR(ifp);
2927 }
2928
2929 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2930                           const struct in6_addr *pfx, unsigned int plen)
2931 {
2932         struct inet6_ifaddr *ifp;
2933         struct inet6_dev *idev;
2934         struct net_device *dev;
2935
2936         if (plen > 128)
2937                 return -EINVAL;
2938
2939         dev = __dev_get_by_index(net, ifindex);
2940         if (!dev)
2941                 return -ENODEV;
2942
2943         idev = __in6_dev_get(dev);
2944         if (!idev)
2945                 return -ENXIO;
2946
2947         read_lock_bh(&idev->lock);
2948         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2949                 if (ifp->prefix_len == plen &&
2950                     ipv6_addr_equal(pfx, &ifp->addr)) {
2951                         in6_ifa_hold(ifp);
2952                         read_unlock_bh(&idev->lock);
2953
2954                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2955                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2956                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2957                                                  jiffies);
2958                         ipv6_del_addr(ifp);
2959                         addrconf_verify_rtnl();
2960                         if (ipv6_addr_is_multicast(pfx)) {
2961                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2962                                                false, pfx, dev->ifindex);
2963                         }
2964                         return 0;
2965                 }
2966         }
2967         read_unlock_bh(&idev->lock);
2968         return -EADDRNOTAVAIL;
2969 }
2970
2971
2972 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2973 {
2974         struct in6_ifreq ireq;
2975         int err;
2976
2977         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2978                 return -EPERM;
2979
2980         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2981                 return -EFAULT;
2982
2983         rtnl_lock();
2984         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2985                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2986                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2987         rtnl_unlock();
2988         return err;
2989 }
2990
2991 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2992 {
2993         struct in6_ifreq ireq;
2994         int err;
2995
2996         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2997                 return -EPERM;
2998
2999         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3000                 return -EFAULT;
3001
3002         rtnl_lock();
3003         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3004                              ireq.ifr6_prefixlen);
3005         rtnl_unlock();
3006         return err;
3007 }
3008
3009 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3010                      int plen, int scope)
3011 {
3012         struct inet6_ifaddr *ifp;
3013
3014         ifp = ipv6_add_addr(idev, addr, NULL, plen,
3015                             scope, IFA_F_PERMANENT,
3016                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
3017                             true, NULL);
3018         if (!IS_ERR(ifp)) {
3019                 spin_lock_bh(&ifp->lock);
3020                 ifp->flags &= ~IFA_F_TENTATIVE;
3021                 spin_unlock_bh(&ifp->lock);
3022                 rt_genid_bump_ipv6(dev_net(idev->dev));
3023                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3024                 in6_ifa_put(ifp);
3025         }
3026 }
3027
3028 #if IS_ENABLED(CONFIG_IPV6_SIT)
3029 static void sit_add_v4_addrs(struct inet6_dev *idev)
3030 {
3031         struct in6_addr addr;
3032         struct net_device *dev;
3033         struct net *net = dev_net(idev->dev);
3034         int scope, plen;
3035         u32 pflags = 0;
3036
3037         ASSERT_RTNL();
3038
3039         memset(&addr, 0, sizeof(struct in6_addr));
3040         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3041
3042         if (idev->dev->flags&IFF_POINTOPOINT) {
3043                 addr.s6_addr32[0] = htonl(0xfe800000);
3044                 scope = IFA_LINK;
3045                 plen = 64;
3046         } else {
3047                 scope = IPV6_ADDR_COMPATv4;
3048                 plen = 96;
3049                 pflags |= RTF_NONEXTHOP;
3050         }
3051
3052         if (addr.s6_addr32[3]) {
3053                 add_addr(idev, &addr, plen, scope);
3054                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3055                 return;
3056         }
3057
3058         for_each_netdev(net, dev) {
3059                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3060                 if (in_dev && (dev->flags & IFF_UP)) {
3061                         struct in_ifaddr *ifa;
3062
3063                         int flag = scope;
3064
3065                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3066
3067                                 addr.s6_addr32[3] = ifa->ifa_local;
3068
3069                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3070                                         continue;
3071                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3072                                         if (idev->dev->flags&IFF_POINTOPOINT)
3073                                                 continue;
3074                                         flag |= IFA_HOST;
3075                                 }
3076
3077                                 add_addr(idev, &addr, plen, flag);
3078                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3079                                                       pflags);
3080                         }
3081                 }
3082         }
3083 }
3084 #endif
3085
3086 static void init_loopback(struct net_device *dev)
3087 {
3088         struct inet6_dev  *idev;
3089
3090         /* ::1 */
3091
3092         ASSERT_RTNL();
3093
3094         idev = ipv6_find_idev(dev);
3095         if (!idev) {
3096                 pr_debug("%s: add_dev failed\n", __func__);
3097                 return;
3098         }
3099
3100         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3101 }
3102
3103 void addrconf_add_linklocal(struct inet6_dev *idev,
3104                             const struct in6_addr *addr, u32 flags)
3105 {
3106         struct inet6_ifaddr *ifp;
3107         u32 addr_flags = flags | IFA_F_PERMANENT;
3108
3109 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3110         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3111              idev->cnf.optimistic_dad) &&
3112             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3113                 addr_flags |= IFA_F_OPTIMISTIC;
3114 #endif
3115
3116         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3117                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3118         if (!IS_ERR(ifp)) {
3119                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3120                 addrconf_dad_start(ifp);
3121                 in6_ifa_put(ifp);
3122         }
3123 }
3124 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3125
3126 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3127 {
3128         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3129                 return true;
3130
3131         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3132             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3133                 return true;
3134
3135         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3136             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3137                 return true;
3138
3139         return false;
3140 }
3141
3142 static int ipv6_generate_stable_address(struct in6_addr *address,
3143                                         u8 dad_count,
3144                                         const struct inet6_dev *idev)
3145 {
3146         static DEFINE_SPINLOCK(lock);
3147         static __u32 digest[SHA_DIGEST_WORDS];
3148         static __u32 workspace[SHA_WORKSPACE_WORDS];
3149
3150         static union {
3151                 char __data[SHA_MESSAGE_BYTES];
3152                 struct {
3153                         struct in6_addr secret;
3154                         __be32 prefix[2];
3155                         unsigned char hwaddr[MAX_ADDR_LEN];
3156                         u8 dad_count;
3157                 } __packed;
3158         } data;
3159
3160         struct in6_addr secret;
3161         struct in6_addr temp;
3162         struct net *net = dev_net(idev->dev);
3163
3164         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3165
3166         if (idev->cnf.stable_secret.initialized)
3167                 secret = idev->cnf.stable_secret.secret;
3168         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3169                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3170         else
3171                 return -1;
3172
3173 retry:
3174         spin_lock_bh(&lock);
3175
3176         sha_init(digest);
3177         memset(&data, 0, sizeof(data));
3178         memset(workspace, 0, sizeof(workspace));
3179         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3180         data.prefix[0] = address->s6_addr32[0];
3181         data.prefix[1] = address->s6_addr32[1];
3182         data.secret = secret;
3183         data.dad_count = dad_count;
3184
3185         sha_transform(digest, data.__data, workspace);
3186
3187         temp = *address;
3188         temp.s6_addr32[2] = (__force __be32)digest[0];
3189         temp.s6_addr32[3] = (__force __be32)digest[1];
3190
3191         spin_unlock_bh(&lock);
3192
3193         if (ipv6_reserved_interfaceid(temp)) {
3194                 dad_count++;
3195                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3196                         return -1;
3197                 goto retry;
3198         }
3199
3200         *address = temp;
3201         return 0;
3202 }
3203
3204 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3205 {
3206         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3207
3208         if (s->initialized)
3209                 return;
3210         s = &idev->cnf.stable_secret;
3211         get_random_bytes(&s->secret, sizeof(s->secret));
3212         s->initialized = true;
3213 }
3214
3215 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3216 {
3217         struct in6_addr addr;
3218
3219         /* no link local addresses on L3 master devices */
3220         if (netif_is_l3_master(idev->dev))
3221                 return;
3222
3223         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3224
3225         switch (idev->cnf.addr_gen_mode) {
3226         case IN6_ADDR_GEN_MODE_RANDOM:
3227                 ipv6_gen_mode_random_init(idev);
3228                 /* fallthrough */
3229         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3230                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3231                         addrconf_add_linklocal(idev, &addr,
3232                                                IFA_F_STABLE_PRIVACY);
3233                 else if (prefix_route)
3234                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3235                 break;
3236         case IN6_ADDR_GEN_MODE_EUI64:
3237                 /* addrconf_add_linklocal also adds a prefix_route and we
3238                  * only need to care about prefix routes if ipv6_generate_eui64
3239                  * couldn't generate one.
3240                  */
3241                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3242                         addrconf_add_linklocal(idev, &addr, 0);
3243                 else if (prefix_route)
3244                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3245                 break;
3246         case IN6_ADDR_GEN_MODE_NONE:
3247         default:
3248                 /* will not add any link local address */
3249                 break;
3250         }
3251 }
3252
3253 static void addrconf_dev_config(struct net_device *dev)
3254 {
3255         struct inet6_dev *idev;
3256
3257         ASSERT_RTNL();
3258
3259         if ((dev->type != ARPHRD_ETHER) &&
3260             (dev->type != ARPHRD_FDDI) &&
3261             (dev->type != ARPHRD_ARCNET) &&
3262             (dev->type != ARPHRD_INFINIBAND) &&
3263             (dev->type != ARPHRD_IEEE1394) &&
3264             (dev->type != ARPHRD_TUNNEL6) &&
3265             (dev->type != ARPHRD_6LOWPAN) &&
3266             (dev->type != ARPHRD_IP6GRE) &&
3267             (dev->type != ARPHRD_IPGRE) &&
3268             (dev->type != ARPHRD_TUNNEL) &&
3269             (dev->type != ARPHRD_NONE)) {
3270                 /* Alas, we support only Ethernet autoconfiguration. */
3271                 return;
3272         }
3273
3274         idev = addrconf_add_dev(dev);
3275         if (IS_ERR(idev))
3276                 return;
3277
3278         /* this device type has no EUI support */
3279         if (dev->type == ARPHRD_NONE &&
3280             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3281                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3282
3283         addrconf_addr_gen(idev, false);
3284 }
3285
3286 #if IS_ENABLED(CONFIG_IPV6_SIT)
3287 static void addrconf_sit_config(struct net_device *dev)
3288 {
3289         struct inet6_dev *idev;
3290
3291         ASSERT_RTNL();
3292
3293         /*
3294          * Configure the tunnel with one of our IPv4
3295          * addresses... we should configure all of
3296          * our v4 addrs in the tunnel
3297          */
3298
3299         idev = ipv6_find_idev(dev);
3300         if (!idev) {
3301                 pr_debug("%s: add_dev failed\n", __func__);
3302                 return;
3303         }
3304
3305         if (dev->priv_flags & IFF_ISATAP) {
3306                 addrconf_addr_gen(idev, false);
3307                 return;
3308         }
3309
3310         sit_add_v4_addrs(idev);
3311
3312         if (dev->flags&IFF_POINTOPOINT)
3313                 addrconf_add_mroute(dev);
3314 }
3315 #endif
3316
3317 #if IS_ENABLED(CONFIG_NET_IPGRE)
3318 static void addrconf_gre_config(struct net_device *dev)
3319 {
3320         struct inet6_dev *idev;
3321
3322         ASSERT_RTNL();
3323
3324         idev = ipv6_find_idev(dev);
3325         if (!idev) {
3326                 pr_debug("%s: add_dev failed\n", __func__);
3327                 return;
3328         }
3329
3330         addrconf_addr_gen(idev, true);
3331         if (dev->flags & IFF_POINTOPOINT)
3332                 addrconf_add_mroute(dev);
3333 }
3334 #endif
3335
3336 static int fixup_permanent_addr(struct inet6_dev *idev,
3337                                 struct inet6_ifaddr *ifp)
3338 {
3339         /* !rt6i_node means the host route was removed from the
3340          * FIB, for example, if 'lo' device is taken down. In that
3341          * case regenerate the host route.
3342          */
3343         if (!ifp->rt || !ifp->rt->rt6i_node) {
3344                 struct rt6_info *rt, *prev;
3345
3346                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3347                 if (IS_ERR(rt))
3348                         return PTR_ERR(rt);
3349
3350                 /* ifp->rt can be accessed outside of rtnl */
3351                 spin_lock(&ifp->lock);
3352                 prev = ifp->rt;
3353                 ifp->rt = rt;
3354                 spin_unlock(&ifp->lock);
3355
3356                 ip6_rt_put(prev);
3357         }
3358
3359         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3360                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3361                                       idev->dev, 0, 0);
3362         }
3363
3364         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3365                 addrconf_dad_start(ifp);
3366
3367         return 0;
3368 }
3369
3370 static void addrconf_permanent_addr(struct net_device *dev)
3371 {
3372         struct inet6_ifaddr *ifp, *tmp;
3373         struct inet6_dev *idev;
3374
3375         idev = __in6_dev_get(dev);
3376         if (!idev)
3377                 return;
3378
3379         write_lock_bh(&idev->lock);