net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92
  93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  94
  95 #define IPV6_MAX_STRLEN \
  96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  97
  98 static inline u32 cstamp_delta(unsigned long cstamp)
  99 {
 100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 101 }
 102
 103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 104 {
 105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 107         do_div(tmp, 1000000);
 108         return (s32)tmp;
 109 }
 110
 111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 112 {
 113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 119                 do_div(tmp, 1000000);
 120         }
 121         return (s32)tmp;
 122 }
 123
 124 #ifdef CONFIG_SYSCTL
 125 static int addrconf_sysctl_register(struct inet6_dev *idev);
 126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 127 #else
 128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 129 {
 130         return 0;
 131 }
 132
 133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 134 {
 135 }
 136 #endif
 137
 138 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 139
 140 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 141 static int ipv6_count_addresses(const struct inet6_dev *idev);
 142 static int ipv6_generate_stable_address(struct in6_addr *addr,
 143                                         u8 dad_count,
 144                                         const struct inet6_dev *idev);
 145
 146 #define IN6_ADDR_HSIZE_SHIFT    8
 147 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 148 /*
 149  *      Configured unicast address hash table
 150  */
 151 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 152 static DEFINE_SPINLOCK(addrconf_hash_lock);
 153
 154 static void addrconf_verify(void);
 155 static void addrconf_verify_rtnl(void);
 156 static void addrconf_verify_work(struct work_struct *);
 157
 158 static struct workqueue_struct *addrconf_wq;
 159 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 160
 161 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 162 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 163
 164 static void addrconf_type_change(struct net_device *dev,
 165                                  unsigned long event);
 166 static int addrconf_ifdown(struct net_device *dev, int how);
 167
 168 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 169                                                   int plen,
 170                                                   const struct net_device *dev,
 171                                                   u32 flags, u32 noflags,
 172                                                   bool no_gw);
 173
 174 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 175 static void addrconf_dad_work(struct work_struct *w);
 176 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 177                                    bool send_na);
 178 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 179 static void addrconf_rs_timer(struct timer_list *t);
 180 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 181 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182
 183 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 184                                 struct prefix_info *pinfo);
 185
 186 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 187         .forwarding             = 0,
 188         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 189         .mtu6                   = IPV6_MIN_MTU,
 190         .accept_ra              = 1,
 191         .accept_redirects       = 1,
 192         .autoconf               = 1,
 193         .force_mld_version      = 0,
 194         .mldv1_unsolicited_report_interval = 10 * HZ,
 195         .mldv2_unsolicited_report_interval = HZ,
 196         .dad_transmits          = 1,
 197         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 198         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 199         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 201         .use_tempaddr           = 0,
 202         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 203         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 204         .regen_max_retry        = REGEN_MAX_RETRY,
 205         .max_desync_factor      = MAX_DESYNC_FACTOR,
 206         .max_addresses          = IPV6_MAX_ADDRESSES,
 207         .accept_ra_defrtr       = 1,
 208         .accept_ra_from_local   = 0,
 209         .accept_ra_min_hop_limit= 1,
 210         .accept_ra_pinfo        = 1,
 211 #ifdef CONFIG_IPV6_ROUTER_PREF
 212         .accept_ra_rtr_pref     = 1,
 213         .rtr_probe_interval     = 60 * HZ,
 214 #ifdef CONFIG_IPV6_ROUTE_INFO
 215         .accept_ra_rt_info_min_plen = 0,
 216         .accept_ra_rt_info_max_plen = 0,
 217 #endif
 218 #endif
 219         .proxy_ndp              = 0,
 220         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 221         .disable_ipv6           = 0,
 222         .accept_dad             = 0,
 223         .suppress_frag_ndisc    = 1,
 224         .accept_ra_mtu          = 1,
 225         .stable_secret          = {
 226                 .initialized = false,
 227         },
 228         .use_oif_addrs_only     = 0,
 229         .ignore_routes_with_linkdown = 0,
 230         .keep_addr_on_down      = 0,
 231         .seg6_enabled           = 0,
 232 #ifdef CONFIG_IPV6_SEG6_HMAC
 233         .seg6_require_hmac      = 0,
 234 #endif
 235         .enhanced_dad           = 1,
 236         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 237         .disable_policy         = 0,
 238         .rpl_seg_enabled        = 0,
 239 };
 240
 241 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 242         .forwarding             = 0,
 243         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 244         .mtu6                   = IPV6_MIN_MTU,
 245         .accept_ra              = 1,
 246         .accept_redirects       = 1,
 247         .autoconf               = 1,
 248         .force_mld_version      = 0,
 249         .mldv1_unsolicited_report_interval = 10 * HZ,
 250         .mldv2_unsolicited_report_interval = HZ,
 251         .dad_transmits          = 1,
 252         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 253         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 254         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 255         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 256         .use_tempaddr           = 0,
 257         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 258         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 259         .regen_max_retry        = REGEN_MAX_RETRY,
 260         .max_desync_factor      = MAX_DESYNC_FACTOR,
 261         .max_addresses          = IPV6_MAX_ADDRESSES,
 262         .accept_ra_defrtr       = 1,
 263         .accept_ra_from_local   = 0,
 264         .accept_ra_min_hop_limit= 1,
 265         .accept_ra_pinfo        = 1,
 266 #ifdef CONFIG_IPV6_ROUTER_PREF
 267         .accept_ra_rtr_pref     = 1,
 268         .rtr_probe_interval     = 60 * HZ,
 269 #ifdef CONFIG_IPV6_ROUTE_INFO
 270         .accept_ra_rt_info_min_plen = 0,
 271         .accept_ra_rt_info_max_plen = 0,
 272 #endif
 273 #endif
 274         .proxy_ndp              = 0,
 275         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 276         .disable_ipv6           = 0,
 277         .accept_dad             = 1,
 278         .suppress_frag_ndisc    = 1,
 279         .accept_ra_mtu          = 1,
 280         .stable_secret          = {
 281                 .initialized = false,
 282         },
 283         .use_oif_addrs_only     = 0,
 284         .ignore_routes_with_linkdown = 0,
 285         .keep_addr_on_down      = 0,
 286         .seg6_enabled           = 0,
 287 #ifdef CONFIG_IPV6_SEG6_HMAC
 288         .seg6_require_hmac      = 0,
 289 #endif
 290         .enhanced_dad           = 1,
 291         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 292         .disable_policy         = 0,
 293         .rpl_seg_enabled        = 0,
 294 };
 295
 296 /* Check if link is ready: is it up and is a valid qdisc available */
 297 static inline bool addrconf_link_ready(const struct net_device *dev)
 298 {
 299         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 300 }
 301
 302 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 303 {
 304         if (del_timer(&idev->rs_timer))
 305                 __in6_dev_put(idev);
 306 }
 307
 308 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 309 {
 310         if (cancel_delayed_work(&ifp->dad_work))
 311                 __in6_ifa_put(ifp);
 312 }
 313
 314 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 315                                   unsigned long when)
 316 {
 317         if (!timer_pending(&idev->rs_timer))
 318                 in6_dev_hold(idev);
 319         mod_timer(&idev->rs_timer, jiffies + when);
 320 }
 321
 322 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 323                                    unsigned long delay)
 324 {
 325         in6_ifa_hold(ifp);
 326         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 327                 in6_ifa_put(ifp);
 328 }
 329
 330 static int snmp6_alloc_dev(struct inet6_dev *idev)
 331 {
 332         int i;
 333
 334         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 335         if (!idev->stats.ipv6)
 336                 goto err_ip;
 337
 338         for_each_possible_cpu(i) {
 339                 struct ipstats_mib *addrconf_stats;
 340                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 341                 u64_stats_init(&addrconf_stats->syncp);
 342         }
 343
 344
 345         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 346                                         GFP_KERNEL);
 347         if (!idev->stats.icmpv6dev)
 348                 goto err_icmp;
 349         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 350                                            GFP_KERNEL);
 351         if (!idev->stats.icmpv6msgdev)
 352                 goto err_icmpmsg;
 353
 354         return 0;
 355
 356 err_icmpmsg:
 357         kfree(idev->stats.icmpv6dev);
 358 err_icmp:
 359         free_percpu(idev->stats.ipv6);
 360 err_ip:
 361         return -ENOMEM;
 362 }
 363
 364 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 365 {
 366         struct inet6_dev *ndev;
 367         int err = -ENOMEM;
 368
 369         ASSERT_RTNL();
 370
 371         if (dev->mtu < IPV6_MIN_MTU)
 372                 return ERR_PTR(-EINVAL);
 373
 374         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 375         if (!ndev)
 376                 return ERR_PTR(err);
 377
 378         rwlock_init(&ndev->lock);
 379         ndev->dev = dev;
 380         INIT_LIST_HEAD(&ndev->addr_list);
 381         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 382         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 383
 384         if (ndev->cnf.stable_secret.initialized)
 385                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 386
 387         ndev->cnf.mtu6 = dev->mtu;
 388         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 389         if (!ndev->nd_parms) {
 390                 kfree(ndev);
 391                 return ERR_PTR(err);
 392         }
 393         if (ndev->cnf.forwarding)
 394                 dev_disable_lro(dev);
 395         /* We refer to the device */
 396         dev_hold(dev);
 397
 398         if (snmp6_alloc_dev(ndev) < 0) {
 399                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 400                            __func__);
 401                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 402                 dev_put(dev);
 403                 kfree(ndev);
 404                 return ERR_PTR(err);
 405         }
 406
 407         if (snmp6_register_dev(ndev) < 0) {
 408                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 409                            __func__, dev->name);
 410                 goto err_release;
 411         }
 412
 413         /* One reference from device. */
 414         refcount_set(&ndev->refcnt, 1);
 415
 416         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 417                 ndev->cnf.accept_dad = -1;
 418
 419 #if IS_ENABLED(CONFIG_IPV6_SIT)
 420         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 421                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 422                 ndev->cnf.rtr_solicits = 0;
 423         }
 424 #endif
 425
 426         INIT_LIST_HEAD(&ndev->tempaddr_list);
 427         ndev->desync_factor = U32_MAX;
 428         if ((dev->flags&IFF_LOOPBACK) ||
 429             dev->type == ARPHRD_TUNNEL ||
 430             dev->type == ARPHRD_TUNNEL6 ||
 431             dev->type == ARPHRD_SIT ||
 432             dev->type == ARPHRD_NONE) {
 433                 ndev->cnf.use_tempaddr = -1;
 434         }
 435
 436         ndev->token = in6addr_any;
 437
 438         if (netif_running(dev) && addrconf_link_ready(dev))
 439                 ndev->if_flags |= IF_READY;
 440
 441         ipv6_mc_init_dev(ndev);
 442         ndev->tstamp = jiffies;
 443         err = addrconf_sysctl_register(ndev);
 444         if (err) {
 445                 ipv6_mc_destroy_dev(ndev);
 446                 snmp6_unregister_dev(ndev);
 447                 goto err_release;
 448         }
 449         /* protected by rtnl_lock */
 450         rcu_assign_pointer(dev->ip6_ptr, ndev);
 451
 452         /* Join interface-local all-node multicast group */
 453         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 454
 455         /* Join all-node multicast group */
 456         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 457
 458         /* Join all-router multicast group if forwarding is set */
 459         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 460                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 461
 462         return ndev;
 463
 464 err_release:
 465         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 466         ndev->dead = 1;
 467         in6_dev_finish_destroy(ndev);
 468         return ERR_PTR(err);
 469 }
 470
 471 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 472 {
 473         struct inet6_dev *idev;
 474
 475         ASSERT_RTNL();
 476
 477         idev = __in6_dev_get(dev);
 478         if (!idev) {
 479                 idev = ipv6_add_dev(dev);
 480                 if (IS_ERR(idev))
 481                         return idev;
 482         }
 483
 484         if (dev->flags&IFF_UP)
 485                 ipv6_mc_up(idev);
 486         return idev;
 487 }
 488
 489 static int inet6_netconf_msgsize_devconf(int type)
 490 {
 491         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 492                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 493         bool all = false;
 494
 495         if (type == NETCONFA_ALL)
 496                 all = true;
 497
 498         if (all || type == NETCONFA_FORWARDING)
 499                 size += nla_total_size(4);
 500 #ifdef CONFIG_IPV6_MROUTE
 501         if (all || type == NETCONFA_MC_FORWARDING)
 502                 size += nla_total_size(4);
 503 #endif
 504         if (all || type == NETCONFA_PROXY_NEIGH)
 505                 size += nla_total_size(4);
 506
 507         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 508                 size += nla_total_size(4);
 509
 510         return size;
 511 }
 512
 513 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 514                                       struct ipv6_devconf *devconf, u32 portid,
 515                                       u32 seq, int event, unsigned int flags,
 516                                       int type)
 517 {
 518         struct nlmsghdr  *nlh;
 519         struct netconfmsg *ncm;
 520         bool all = false;
 521
 522         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 523                         flags);
 524         if (!nlh)
 525                 return -EMSGSIZE;
 526
 527         if (type == NETCONFA_ALL)
 528                 all = true;
 529
 530         ncm = nlmsg_data(nlh);
 531         ncm->ncm_family = AF_INET6;
 532
 533         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 534                 goto nla_put_failure;
 535
 536         if (!devconf)
 537                 goto out;
 538
 539         if ((all || type == NETCONFA_FORWARDING) &&
 540             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 541                 goto nla_put_failure;
 542 #ifdef CONFIG_IPV6_MROUTE
 543         if ((all || type == NETCONFA_MC_FORWARDING) &&
 544             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 545                         devconf->mc_forwarding) < 0)
 546                 goto nla_put_failure;
 547 #endif
 548         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 549             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 550                 goto nla_put_failure;
 551
 552         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 553             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 554                         devconf->ignore_routes_with_linkdown) < 0)
 555                 goto nla_put_failure;
 556
 557 out:
 558         nlmsg_end(skb, nlh);
 559         return 0;
 560
 561 nla_put_failure:
 562         nlmsg_cancel(skb, nlh);
 563         return -EMSGSIZE;
 564 }
 565
 566 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 567                                   int ifindex, struct ipv6_devconf *devconf)
 568 {
 569         struct sk_buff *skb;
 570         int err = -ENOBUFS;
 571
 572         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 573         if (!skb)
 574                 goto errout;
 575
 576         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 577                                          event, 0, type);
 578         if (err < 0) {
 579                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 580                 WARN_ON(err == -EMSGSIZE);
 581                 kfree_skb(skb);
 582                 goto errout;
 583         }
 584         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 585         return;
 586 errout:
 587         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 588 }
 589
 590 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 591         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 592         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 593         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 594         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 595 };
 596
 597 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 598                                        const struct nlmsghdr *nlh,
 599                                        struct nlattr **tb,
 600                                        struct netlink_ext_ack *extack)
 601 {
 602         int i, err;
 603
 604         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 605                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 606                 return -EINVAL;
 607         }
 608
 609         if (!netlink_strict_get_check(skb))
 610                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 611                                               tb, NETCONFA_MAX,
 612                                               devconf_ipv6_policy, extack);
 613
 614         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 615                                             tb, NETCONFA_MAX,
 616                                             devconf_ipv6_policy, extack);
 617         if (err)
 618                 return err;
 619
 620         for (i = 0; i <= NETCONFA_MAX; i++) {
 621                 if (!tb[i])
 622                         continue;
 623
 624                 switch (i) {
 625                 case NETCONFA_IFINDEX:
 626                         break;
 627                 default:
 628                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 629                         return -EINVAL;
 630                 }
 631         }
 632
 633         return 0;
 634 }
 635
 636 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 637                                      struct nlmsghdr *nlh,
 638                                      struct netlink_ext_ack *extack)
 639 {
 640         struct net *net = sock_net(in_skb->sk);
 641         struct nlattr *tb[NETCONFA_MAX+1];
 642         struct inet6_dev *in6_dev = NULL;
 643         struct net_device *dev = NULL;
 644         struct sk_buff *skb;
 645         struct ipv6_devconf *devconf;
 646         int ifindex;
 647         int err;
 648
 649         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 650         if (err < 0)
 651                 return err;
 652
 653         if (!tb[NETCONFA_IFINDEX])
 654                 return -EINVAL;
 655
 656         err = -EINVAL;
 657         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 658         switch (ifindex) {
 659         case NETCONFA_IFINDEX_ALL:
 660                 devconf = net->ipv6.devconf_all;
 661                 break;
 662         case NETCONFA_IFINDEX_DEFAULT:
 663                 devconf = net->ipv6.devconf_dflt;
 664                 break;
 665         default:
 666                 dev = dev_get_by_index(net, ifindex);
 667                 if (!dev)
 668                         return -EINVAL;
 669                 in6_dev = in6_dev_get(dev);
 670                 if (!in6_dev)
 671                         goto errout;
 672                 devconf = &in6_dev->cnf;
 673                 break;
 674         }
 675
 676         err = -ENOBUFS;
 677         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 678         if (!skb)
 679                 goto errout;
 680
 681         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 682                                          NETLINK_CB(in_skb).portid,
 683                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 684                                          NETCONFA_ALL);
 685         if (err < 0) {
 686                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 687                 WARN_ON(err == -EMSGSIZE);
 688                 kfree_skb(skb);
 689                 goto errout;
 690         }
 691         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 692 errout:
 693         if (in6_dev)
 694                 in6_dev_put(in6_dev);
 695         if (dev)
 696                 dev_put(dev);
 697         return err;
 698 }
 699
 700 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 701                                       struct netlink_callback *cb)
 702 {
 703         const struct nlmsghdr *nlh = cb->nlh;
 704         struct net *net = sock_net(skb->sk);
 705         int h, s_h;
 706         int idx, s_idx;
 707         struct net_device *dev;
 708         struct inet6_dev *idev;
 709         struct hlist_head *head;
 710
 711         if (cb->strict_check) {
 712                 struct netlink_ext_ack *extack = cb->extack;
 713                 struct netconfmsg *ncm;
 714
 715                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 716                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 717                         return -EINVAL;
 718                 }
 719
 720                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 721                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 722                         return -EINVAL;
 723                 }
 724         }
 725
 726         s_h = cb->args[0];
 727         s_idx = idx = cb->args[1];
 728
 729         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 730                 idx = 0;
 731                 head = &net->dev_index_head[h];
 732                 rcu_read_lock();
 733                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 734                           net->dev_base_seq;
 735                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 736                         if (idx < s_idx)
 737                                 goto cont;
 738                         idev = __in6_dev_get(dev);
 739                         if (!idev)
 740                                 goto cont;
 741
 742                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 743                                                        &idev->cnf,
 744                                                        NETLINK_CB(cb->skb).portid,
 745                                                        nlh->nlmsg_seq,
 746                                                        RTM_NEWNETCONF,
 747                                                        NLM_F_MULTI,
 748                                                        NETCONFA_ALL) < 0) {
 749                                 rcu_read_unlock();
 750                                 goto done;
 751                         }
 752                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 753 cont:
 754                         idx++;
 755                 }
 756                 rcu_read_unlock();
 757         }
 758         if (h == NETDEV_HASHENTRIES) {
 759                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 760                                                net->ipv6.devconf_all,
 761                                                NETLINK_CB(cb->skb).portid,
 762                                                nlh->nlmsg_seq,
 763                                                RTM_NEWNETCONF, NLM_F_MULTI,
 764                                                NETCONFA_ALL) < 0)
 765                         goto done;
 766                 else
 767                         h++;
 768         }
 769         if (h == NETDEV_HASHENTRIES + 1) {
 770                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 771                                                net->ipv6.devconf_dflt,
 772                                                NETLINK_CB(cb->skb).portid,
 773                                                nlh->nlmsg_seq,
 774                                                RTM_NEWNETCONF, NLM_F_MULTI,
 775                                                NETCONFA_ALL) < 0)
 776                         goto done;
 777                 else
 778                         h++;
 779         }
 780 done:
 781         cb->args[0] = h;
 782         cb->args[1] = idx;
 783
 784         return skb->len;
 785 }
 786
 787 #ifdef CONFIG_SYSCTL
 788 static void dev_forward_change(struct inet6_dev *idev)
 789 {
 790         struct net_device *dev;
 791         struct inet6_ifaddr *ifa;
 792
 793         if (!idev)
 794                 return;
 795         dev = idev->dev;
 796         if (idev->cnf.forwarding)
 797                 dev_disable_lro(dev);
 798         if (dev->flags & IFF_MULTICAST) {
 799                 if (idev->cnf.forwarding) {
 800                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 801                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 802                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 803                 } else {
 804                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 805                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 806                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 807                 }
 808         }
 809
 810         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 811                 if (ifa->flags&IFA_F_TENTATIVE)
 812                         continue;
 813                 if (idev->cnf.forwarding)
 814                         addrconf_join_anycast(ifa);
 815                 else
 816                         addrconf_leave_anycast(ifa);
 817         }
 818         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 819                                      NETCONFA_FORWARDING,
 820                                      dev->ifindex, &idev->cnf);
 821 }
 822
 823
 824 static void addrconf_forward_change(struct net *net, __s32 newf)
 825 {
 826         struct net_device *dev;
 827         struct inet6_dev *idev;
 828
 829         for_each_netdev(net, dev) {
 830                 idev = __in6_dev_get(dev);
 831                 if (idev) {
 832                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 833                         idev->cnf.forwarding = newf;
 834                         if (changed)
 835                                 dev_forward_change(idev);
 836                 }
 837         }
 838 }
 839
 840 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 841 {
 842         struct net *net;
 843         int old;
 844
 845         if (!rtnl_trylock())
 846                 return restart_syscall();
 847
 848         net = (struct net *)table->extra2;
 849         old = *p;
 850         *p = newf;
 851
 852         if (p == &net->ipv6.devconf_dflt->forwarding) {
 853                 if ((!newf) ^ (!old))
 854                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 855                                                      NETCONFA_FORWARDING,
 856                                                      NETCONFA_IFINDEX_DEFAULT,
 857                                                      net->ipv6.devconf_dflt);
 858                 rtnl_unlock();
 859                 return 0;
 860         }
 861
 862         if (p == &net->ipv6.devconf_all->forwarding) {
 863                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 864
 865                 net->ipv6.devconf_dflt->forwarding = newf;
 866                 if ((!newf) ^ (!old_dflt))
 867                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 868                                                      NETCONFA_FORWARDING,
 869                                                      NETCONFA_IFINDEX_DEFAULT,
 870                                                      net->ipv6.devconf_dflt);
 871
 872                 addrconf_forward_change(net, newf);
 873                 if ((!newf) ^ (!old))
 874                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 875                                                      NETCONFA_FORWARDING,
 876                                                      NETCONFA_IFINDEX_ALL,
 877                                                      net->ipv6.devconf_all);
 878         } else if ((!newf) ^ (!old))
 879                 dev_forward_change((struct inet6_dev *)table->extra1);
 880         rtnl_unlock();
 881
 882         if (newf)
 883                 rt6_purge_dflt_routers(net);
 884         return 1;
 885 }
 886
 887 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 888 {
 889         struct net_device *dev;
 890         struct inet6_dev *idev;
 891
 892         for_each_netdev(net, dev) {
 893                 idev = __in6_dev_get(dev);
 894                 if (idev) {
 895                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 896
 897                         idev->cnf.ignore_routes_with_linkdown = newf;
 898                         if (changed)
 899                                 inet6_netconf_notify_devconf(dev_net(dev),
 900                                                              RTM_NEWNETCONF,
 901                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 902                                                              dev->ifindex,
 903                                                              &idev->cnf);
 904                 }
 905         }
 906 }
 907
 908 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 909 {
 910         struct net *net;
 911         int old;
 912
 913         if (!rtnl_trylock())
 914                 return restart_syscall();
 915
 916         net = (struct net *)table->extra2;
 917         old = *p;
 918         *p = newf;
 919
 920         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 921                 if ((!newf) ^ (!old))
 922                         inet6_netconf_notify_devconf(net,
 923                                                      RTM_NEWNETCONF,
 924                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 925                                                      NETCONFA_IFINDEX_DEFAULT,
 926                                                      net->ipv6.devconf_dflt);
 927                 rtnl_unlock();
 928                 return 0;
 929         }
 930
 931         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 932                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 933                 addrconf_linkdown_change(net, newf);
 934                 if ((!newf) ^ (!old))
 935                         inet6_netconf_notify_devconf(net,
 936                                                      RTM_NEWNETCONF,
 937                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 938                                                      NETCONFA_IFINDEX_ALL,
 939                                                      net->ipv6.devconf_all);
 940         }
 941         rtnl_unlock();
 942
 943         return 1;
 944 }
 945
 946 #endif
 947
 948 /* Nobody refers to this ifaddr, destroy it */
 949 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 950 {
 951         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 952
 953 #ifdef NET_REFCNT_DEBUG
 954         pr_debug("%s\n", __func__);
 955 #endif
 956
 957         in6_dev_put(ifp->idev);
 958
 959         if (cancel_delayed_work(&ifp->dad_work))
 960                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 961                           ifp);
 962
 963         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 964                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 965                 return;
 966         }
 967
 968         kfree_rcu(ifp, rcu);
 969 }
 970
 971 static void
 972 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 973 {
 974         struct list_head *p;
 975         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 976
 977         /*
 978          * Each device address list is sorted in order of scope -
 979          * global before linklocal.
 980          */
 981         list_for_each(p, &idev->addr_list) {
 982                 struct inet6_ifaddr *ifa
 983                         = list_entry(p, struct inet6_ifaddr, if_list);
 984                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 985                         break;
 986         }
 987
 988         list_add_tail_rcu(&ifp->if_list, p);
 989 }
 990
 991 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 992 {
 993         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 994
 995         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 996 }
 997
 998 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 999                                struct net_device *dev, unsigned int hash)
1000 {
1001         struct inet6_ifaddr *ifp;
1002
1003         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1004                 if (!net_eq(dev_net(ifp->idev->dev), net))
1005                         continue;
1006                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1007                         if (!dev || ifp->idev->dev == dev)
1008                                 return true;
1009                 }
1010         }
1011         return false;
1012 }
1013
1014 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1015 {
1016         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1017         int err = 0;
1018
1019         spin_lock(&addrconf_hash_lock);
1020
1021         /* Ignore adding duplicate addresses on an interface */
1022         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1023                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1024                 err = -EEXIST;
1025         } else {
1026                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1027         }
1028
1029         spin_unlock(&addrconf_hash_lock);
1030
1031         return err;
1032 }
1033
1034 /* On success it returns ifp with increased reference count */
1035
1036 static struct inet6_ifaddr *
1037 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1038               bool can_block, struct netlink_ext_ack *extack)
1039 {
1040         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1041         int addr_type = ipv6_addr_type(cfg->pfx);
1042         struct net *net = dev_net(idev->dev);
1043         struct inet6_ifaddr *ifa = NULL;
1044         struct fib6_info *f6i = NULL;
1045         int err = 0;
1046
1047         if (addr_type == IPV6_ADDR_ANY ||
1048             (addr_type & IPV6_ADDR_MULTICAST &&
1049              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1050             (!(idev->dev->flags & IFF_LOOPBACK) &&
1051              !netif_is_l3_master(idev->dev) &&
1052              addr_type & IPV6_ADDR_LOOPBACK))
1053                 return ERR_PTR(-EADDRNOTAVAIL);
1054
1055         if (idev->dead) {
1056                 err = -ENODEV;                  /*XXX*/
1057                 goto out;
1058         }
1059
1060         if (idev->cnf.disable_ipv6) {
1061                 err = -EACCES;
1062                 goto out;
1063         }
1064
1065         /* validator notifier needs to be blocking;
1066          * do not call in atomic context
1067          */
1068         if (can_block) {
1069                 struct in6_validator_info i6vi = {
1070                         .i6vi_addr = *cfg->pfx,
1071                         .i6vi_dev = idev,
1072                         .extack = extack,
1073                 };
1074
1075                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1076                 err = notifier_to_errno(err);
1077                 if (err < 0)
1078                         goto out;
1079         }
1080
1081         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1082         if (!ifa) {
1083                 err = -ENOBUFS;
1084                 goto out;
1085         }
1086
1087         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1088         if (IS_ERR(f6i)) {
1089                 err = PTR_ERR(f6i);
1090                 f6i = NULL;
1091                 goto out;
1092         }
1093
1094         if (net->ipv6.devconf_all->disable_policy ||
1095             idev->cnf.disable_policy)
1096                 f6i->dst_nopolicy = true;
1097
1098         neigh_parms_data_state_setall(idev->nd_parms);
1099
1100         ifa->addr = *cfg->pfx;
1101         if (cfg->peer_pfx)
1102                 ifa->peer_addr = *cfg->peer_pfx;
1103
1104         spin_lock_init(&ifa->lock);
1105         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1106         INIT_HLIST_NODE(&ifa->addr_lst);
1107         ifa->scope = cfg->scope;
1108         ifa->prefix_len = cfg->plen;
1109         ifa->rt_priority = cfg->rt_priority;
1110         ifa->flags = cfg->ifa_flags;
1111         /* No need to add the TENTATIVE flag for addresses with NODAD */
1112         if (!(cfg->ifa_flags & IFA_F_NODAD))
1113                 ifa->flags |= IFA_F_TENTATIVE;
1114         ifa->valid_lft = cfg->valid_lft;
1115         ifa->prefered_lft = cfg->preferred_lft;
1116         ifa->cstamp = ifa->tstamp = jiffies;
1117         ifa->tokenized = false;
1118
1119         ifa->rt = f6i;
1120
1121         ifa->idev = idev;
1122         in6_dev_hold(idev);
1123
1124         /* For caller */
1125         refcount_set(&ifa->refcnt, 1);
1126
1127         rcu_read_lock_bh();
1128
1129         err = ipv6_add_addr_hash(idev->dev, ifa);
1130         if (err < 0) {
1131                 rcu_read_unlock_bh();
1132                 goto out;
1133         }
1134
1135         write_lock(&idev->lock);
1136
1137         /* Add to inet6_dev unicast addr list. */
1138         ipv6_link_dev_addr(idev, ifa);
1139
1140         if (ifa->flags&IFA_F_TEMPORARY) {
1141                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1142                 in6_ifa_hold(ifa);
1143         }
1144
1145         in6_ifa_hold(ifa);
1146         write_unlock(&idev->lock);
1147
1148         rcu_read_unlock_bh();
1149
1150         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1151 out:
1152         if (unlikely(err < 0)) {
1153                 fib6_info_release(f6i);
1154
1155                 if (ifa) {
1156                         if (ifa->idev)
1157                                 in6_dev_put(ifa->idev);
1158                         kfree(ifa);
1159                 }
1160                 ifa = ERR_PTR(err);
1161         }
1162
1163         return ifa;
1164 }
1165
1166 enum cleanup_prefix_rt_t {
1167         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1168         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1169         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1170 };
1171
1172 /*
1173  * Check, whether the prefix for ifp would still need a prefix route
1174  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1175  * constants.
1176  *
1177  * 1) we don't purge prefix if address was not permanent.
1178  *    prefix is managed by its own lifetime.
1179  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1180  * 3) if there are no addresses, delete prefix.
1181  * 4) if there are still other permanent address(es),
1182  *    corresponding prefix is still permanent.
1183  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1184  *    don't purge the prefix, assume user space is managing it.
1185  * 6) otherwise, update prefix lifetime to the
1186  *    longest valid lifetime among the corresponding
1187  *    addresses on the device.
1188  *    Note: subsequent RA will update lifetime.
1189  **/
1190 static enum cleanup_prefix_rt_t
1191 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1192 {
1193         struct inet6_ifaddr *ifa;
1194         struct inet6_dev *idev = ifp->idev;
1195         unsigned long lifetime;
1196         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1197
1198         *expires = jiffies;
1199
1200         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1201                 if (ifa == ifp)
1202                         continue;
1203                 if (ifa->prefix_len != ifp->prefix_len ||
1204                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1205                                        ifp->prefix_len))
1206                         continue;
1207                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1208                         return CLEANUP_PREFIX_RT_NOP;
1209
1210                 action = CLEANUP_PREFIX_RT_EXPIRE;
1211
1212                 spin_lock(&ifa->lock);
1213
1214                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1215                 /*
1216                  * Note: Because this address is
1217                  * not permanent, lifetime <
1218                  * LONG_MAX / HZ here.
1219                  */
1220                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1221                         *expires = ifa->tstamp + lifetime * HZ;
1222                 spin_unlock(&ifa->lock);
1223         }
1224
1225         return action;
1226 }
1227
1228 static void
1229 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1230                      bool del_rt, bool del_peer)
1231 {
1232         struct fib6_info *f6i;
1233
1234         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1235                                         ifp->prefix_len,
1236                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1237         if (f6i) {
1238                 if (del_rt)
1239                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1240                 else {
1241                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1242                                 fib6_set_expires(f6i, expires);
1243                         fib6_info_release(f6i);
1244                 }
1245         }
1246 }
1247
1248
1249 /* This function wants to get referenced ifp and releases it before return */
1250
1251 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1252 {
1253         int state;
1254         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1255         unsigned long expires;
1256
1257         ASSERT_RTNL();
1258
1259         spin_lock_bh(&ifp->lock);
1260         state = ifp->state;
1261         ifp->state = INET6_IFADDR_STATE_DEAD;
1262         spin_unlock_bh(&ifp->lock);
1263
1264         if (state == INET6_IFADDR_STATE_DEAD)
1265                 goto out;
1266
1267         spin_lock_bh(&addrconf_hash_lock);
1268         hlist_del_init_rcu(&ifp->addr_lst);
1269         spin_unlock_bh(&addrconf_hash_lock);
1270
1271         write_lock_bh(&ifp->idev->lock);
1272
1273         if (ifp->flags&IFA_F_TEMPORARY) {
1274                 list_del(&ifp->tmp_list);
1275                 if (ifp->ifpub) {
1276                         in6_ifa_put(ifp->ifpub);
1277                         ifp->ifpub = NULL;
1278                 }
1279                 __in6_ifa_put(ifp);
1280         }
1281
1282         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1283                 action = check_cleanup_prefix_route(ifp, &expires);
1284
1285         list_del_rcu(&ifp->if_list);
1286         __in6_ifa_put(ifp);
1287
1288         write_unlock_bh(&ifp->idev->lock);
1289
1290         addrconf_del_dad_work(ifp);
1291
1292         ipv6_ifa_notify(RTM_DELADDR, ifp);
1293
1294         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1295
1296         if (action != CLEANUP_PREFIX_RT_NOP) {
1297                 cleanup_prefix_route(ifp, expires,
1298                         action == CLEANUP_PREFIX_RT_DEL, false);
1299         }
1300
1301         /* clean up prefsrc entries */
1302         rt6_remove_prefsrc(ifp);
1303 out:
1304         in6_ifa_put(ifp);
1305 }
1306
1307 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1308 {
1309         struct inet6_dev *idev = ifp->idev;
1310         unsigned long tmp_tstamp, age;
1311         unsigned long regen_advance;
1312         unsigned long now = jiffies;
1313         s32 cnf_temp_preferred_lft;
1314         struct inet6_ifaddr *ift;
1315         struct ifa6_config cfg;
1316         long max_desync_factor;
1317         struct in6_addr addr;
1318         int ret = 0;
1319
1320         write_lock_bh(&idev->lock);
1321
1322 retry:
1323         in6_dev_hold(idev);
1324         if (idev->cnf.use_tempaddr <= 0) {
1325                 write_unlock_bh(&idev->lock);
1326                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1327                 in6_dev_put(idev);
1328                 ret = -1;
1329                 goto out;
1330         }
1331         spin_lock_bh(&ifp->lock);
1332         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1333                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1334                 spin_unlock_bh(&ifp->lock);
1335                 write_unlock_bh(&idev->lock);
1336                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1337                         __func__);
1338                 in6_dev_put(idev);
1339                 ret = -1;
1340                 goto out;
1341         }
1342         in6_ifa_hold(ifp);
1343         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1344         ipv6_gen_rnd_iid(&addr);
1345
1346         age = (now - ifp->tstamp) / HZ;
1347
1348         regen_advance = idev->cnf.regen_max_retry *
1349                         idev->cnf.dad_transmits *
1350                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1351
1352         /* recalculate max_desync_factor each time and update
1353          * idev->desync_factor if it's larger
1354          */
1355         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1356         max_desync_factor = min_t(__u32,
1357                                   idev->cnf.max_desync_factor,
1358                                   cnf_temp_preferred_lft - regen_advance);
1359
1360         if (unlikely(idev->desync_factor > max_desync_factor)) {
1361                 if (max_desync_factor > 0) {
1362                         get_random_bytes(&idev->desync_factor,
1363                                          sizeof(idev->desync_factor));
1364                         idev->desync_factor %= max_desync_factor;
1365                 } else {
1366                         idev->desync_factor = 0;
1367                 }
1368         }
1369
1370         memset(&cfg, 0, sizeof(cfg));
1371         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1372                               idev->cnf.temp_valid_lft + age);
1373         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1374         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1375
1376         cfg.plen = ifp->prefix_len;
1377         tmp_tstamp = ifp->tstamp;
1378         spin_unlock_bh(&ifp->lock);
1379
1380         write_unlock_bh(&idev->lock);
1381
1382         /* A temporary address is created only if this calculated Preferred
1383          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1384          * an implementation must not create a temporary address with a zero
1385          * Preferred Lifetime.
1386          * Use age calculation as in addrconf_verify to avoid unnecessary
1387          * temporary addresses being generated.
1388          */
1389         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1390         if (cfg.preferred_lft <= regen_advance + age) {
1391                 in6_ifa_put(ifp);
1392                 in6_dev_put(idev);
1393                 ret = -1;
1394                 goto out;
1395         }
1396
1397         cfg.ifa_flags = IFA_F_TEMPORARY;
1398         /* set in addrconf_prefix_rcv() */
1399         if (ifp->flags & IFA_F_OPTIMISTIC)
1400                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1401
1402         cfg.pfx = &addr;
1403         cfg.scope = ipv6_addr_scope(cfg.pfx);
1404
1405         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1406         if (IS_ERR(ift)) {
1407                 in6_ifa_put(ifp);
1408                 in6_dev_put(idev);
1409                 pr_info("%s: retry temporary address regeneration\n", __func__);
1410                 write_lock_bh(&idev->lock);
1411                 goto retry;
1412         }
1413
1414         spin_lock_bh(&ift->lock);
1415         ift->ifpub = ifp;
1416         ift->cstamp = now;
1417         ift->tstamp = tmp_tstamp;
1418         spin_unlock_bh(&ift->lock);
1419
1420         addrconf_dad_start(ift);
1421         in6_ifa_put(ift);
1422         in6_dev_put(idev);
1423 out:
1424         return ret;
1425 }
1426
1427 /*
1428  *      Choose an appropriate source address (RFC3484)
1429  */
1430 enum {
1431         IPV6_SADDR_RULE_INIT = 0,
1432         IPV6_SADDR_RULE_LOCAL,
1433         IPV6_SADDR_RULE_SCOPE,
1434         IPV6_SADDR_RULE_PREFERRED,
1435 #ifdef CONFIG_IPV6_MIP6
1436         IPV6_SADDR_RULE_HOA,
1437 #endif
1438         IPV6_SADDR_RULE_OIF,
1439         IPV6_SADDR_RULE_LABEL,
1440         IPV6_SADDR_RULE_PRIVACY,
1441         IPV6_SADDR_RULE_ORCHID,
1442         IPV6_SADDR_RULE_PREFIX,
1443 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1444         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1445 #endif
1446         IPV6_SADDR_RULE_MAX
1447 };
1448
1449 struct ipv6_saddr_score {
1450         int                     rule;
1451         int                     addr_type;
1452         struct inet6_ifaddr     *ifa;
1453         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1454         int                     scopedist;
1455         int                     matchlen;
1456 };
1457
1458 struct ipv6_saddr_dst {
1459         const struct in6_addr *addr;
1460         int ifindex;
1461         int scope;
1462         int label;
1463         unsigned int prefs;
1464 };
1465
1466 static inline int ipv6_saddr_preferred(int type)
1467 {
1468         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1469                 return 1;
1470         return 0;
1471 }
1472
1473 static bool ipv6_use_optimistic_addr(struct net *net,
1474                                      struct inet6_dev *idev)
1475 {
1476 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1477         if (!idev)
1478                 return false;
1479         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1480                 return false;
1481         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1482                 return false;
1483
1484         return true;
1485 #else
1486         return false;
1487 #endif
1488 }
1489
1490 static bool ipv6_allow_optimistic_dad(struct net *net,
1491                                       struct inet6_dev *idev)
1492 {
1493 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1494         if (!idev)
1495                 return false;
1496         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1497                 return false;
1498
1499         return true;
1500 #else
1501         return false;
1502 #endif
1503 }
1504
1505 static int ipv6_get_saddr_eval(struct net *net,
1506                                struct ipv6_saddr_score *score,
1507                                struct ipv6_saddr_dst *dst,
1508                                int i)
1509 {
1510         int ret;
1511
1512         if (i <= score->rule) {
1513                 switch (i) {
1514                 case IPV6_SADDR_RULE_SCOPE:
1515                         ret = score->scopedist;
1516                         break;
1517                 case IPV6_SADDR_RULE_PREFIX:
1518                         ret = score->matchlen;
1519                         break;
1520                 default:
1521                         ret = !!test_bit(i, score->scorebits);
1522                 }
1523                 goto out;
1524         }
1525
1526         switch (i) {
1527         case IPV6_SADDR_RULE_INIT:
1528                 /* Rule 0: remember if hiscore is not ready yet */
1529                 ret = !!score->ifa;
1530                 break;
1531         case IPV6_SADDR_RULE_LOCAL:
1532                 /* Rule 1: Prefer same address */
1533                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1534                 break;
1535         case IPV6_SADDR_RULE_SCOPE:
1536                 /* Rule 2: Prefer appropriate scope
1537                  *
1538                  *      ret
1539                  *       ^
1540                  *    -1 |  d 15
1541                  *    ---+--+-+---> scope
1542                  *       |
1543                  *       |             d is scope of the destination.
1544                  *  B-d  |  \
1545                  *       |   \      <- smaller scope is better if
1546                  *  B-15 |    \        if scope is enough for destination.
1547                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1548                  * d-C-1 | /
1549                  *       |/         <- greater is better
1550                  *   -C  /             if scope is not enough for destination.
1551                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1552                  *
1553                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1554                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1555                  * Assume B = 0 and we get C > 29.
1556                  */
1557                 ret = __ipv6_addr_src_scope(score->addr_type);
1558                 if (ret >= dst->scope)
1559                         ret = -ret;
1560                 else
1561                         ret -= 128;     /* 30 is enough */
1562                 score->scopedist = ret;
1563                 break;
1564         case IPV6_SADDR_RULE_PREFERRED:
1565             {
1566                 /* Rule 3: Avoid deprecated and optimistic addresses */
1567                 u8 avoid = IFA_F_DEPRECATED;
1568
1569                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1570                         avoid |= IFA_F_OPTIMISTIC;
1571                 ret = ipv6_saddr_preferred(score->addr_type) ||
1572                       !(score->ifa->flags & avoid);
1573                 break;
1574             }
1575 #ifdef CONFIG_IPV6_MIP6
1576         case IPV6_SADDR_RULE_HOA:
1577             {
1578                 /* Rule 4: Prefer home address */
1579                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1580                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1581                 break;
1582             }
1583 #endif
1584         case IPV6_SADDR_RULE_OIF:
1585                 /* Rule 5: Prefer outgoing interface */
1586                 ret = (!dst->ifindex ||
1587                        dst->ifindex == score->ifa->idev->dev->ifindex);
1588                 break;
1589         case IPV6_SADDR_RULE_LABEL:
1590                 /* Rule 6: Prefer matching label */
1591                 ret = ipv6_addr_label(net,
1592                                       &score->ifa->addr, score->addr_type,
1593                                       score->ifa->idev->dev->ifindex) == dst->label;
1594                 break;
1595         case IPV6_SADDR_RULE_PRIVACY:
1596             {
1597                 /* Rule 7: Prefer public address
1598                  * Note: prefer temporary address if use_tempaddr >= 2
1599                  */
1600                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1601                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1602                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1603                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1604                 break;
1605             }
1606         case IPV6_SADDR_RULE_ORCHID:
1607                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1608                  *          non-ORCHID vs non-ORCHID
1609                  */
1610                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1611                         ipv6_addr_orchid(dst->addr));
1612                 break;
1613         case IPV6_SADDR_RULE_PREFIX:
1614                 /* Rule 8: Use longest matching prefix */
1615                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1616                 if (ret > score->ifa->prefix_len)
1617                         ret = score->ifa->prefix_len;
1618                 score->matchlen = ret;
1619                 break;
1620 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1621         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1622                 /* Optimistic addresses still have lower precedence than other
1623                  * preferred addresses.
1624                  */
1625                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1626                 break;
1627 #endif
1628         default:
1629                 ret = 0;
1630         }
1631
1632         if (ret)
1633                 __set_bit(i, score->scorebits);
1634         score->rule = i;
1635 out:
1636         return ret;
1637 }
1638
1639 static int __ipv6_dev_get_saddr(struct net *net,
1640                                 struct ipv6_saddr_dst *dst,
1641                                 struct inet6_dev *idev,
1642                                 struct ipv6_saddr_score *scores,
1643                                 int hiscore_idx)
1644 {
1645         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1646
1647         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1648                 int i;
1649
1650                 /*
1651                  * - Tentative Address (RFC2462 section 5.4)
1652                  *  - A tentative address is not considered
1653                  *    "assigned to an interface" in the traditional
1654                  *    sense, unless it is also flagged as optimistic.
1655                  * - Candidate Source Address (section 4)
1656                  *  - In any case, anycast addresses, multicast
1657                  *    addresses, and the unspecified address MUST
1658                  *    NOT be included in a candidate set.
1659                  */
1660                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1661                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1662                         continue;
1663
1664                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1665
1666                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1667                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1668                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1669                                             idev->dev->name);
1670                         continue;
1671                 }
1672
1673                 score->rule = -1;
1674                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1675
1676                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1677                         int minihiscore, miniscore;
1678
1679                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1680                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1681
1682                         if (minihiscore > miniscore) {
1683                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1684                                     score->scopedist > 0) {
1685                                         /*
1686                                          * special case:
1687                                          * each remaining entry
1688                                          * has too small (not enough)
1689                                          * scope, because ifa entries
1690                                          * are sorted by their scope
1691                                          * values.
1692                                          */
1693                                         goto out;
1694                                 }
1695                                 break;
1696                         } else if (minihiscore < miniscore) {
1697                                 swap(hiscore, score);
1698                                 hiscore_idx = 1 - hiscore_idx;
1699
1700                                 /* restore our iterator */
1701                                 score->ifa = hiscore->ifa;
1702
1703                                 break;
1704                         }
1705                 }
1706         }
1707 out:
1708         return hiscore_idx;
1709 }
1710
1711 static int ipv6_get_saddr_master(struct net *net,
1712                                  const struct net_device *dst_dev,
1713                                  const struct net_device *master,
1714                                  struct ipv6_saddr_dst *dst,
1715                                  struct ipv6_saddr_score *scores,
1716                                  int hiscore_idx)
1717 {
1718         struct inet6_dev *idev;
1719
1720         idev = __in6_dev_get(dst_dev);
1721         if (idev)
1722                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1723                                                    scores, hiscore_idx);
1724
1725         idev = __in6_dev_get(master);
1726         if (idev)
1727                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1728                                                    scores, hiscore_idx);
1729
1730         return hiscore_idx;
1731 }
1732
1733 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1734                        const struct in6_addr *daddr, unsigned int prefs,
1735                        struct in6_addr *saddr)
1736 {
1737         struct ipv6_saddr_score scores[2], *hiscore;
1738         struct ipv6_saddr_dst dst;
1739         struct inet6_dev *idev;
1740         struct net_device *dev;
1741         int dst_type;
1742         bool use_oif_addr = false;
1743         int hiscore_idx = 0;
1744         int ret = 0;
1745
1746         dst_type = __ipv6_addr_type(daddr);
1747         dst.addr = daddr;
1748         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1749         dst.scope = __ipv6_addr_src_scope(dst_type);
1750         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1751         dst.prefs = prefs;
1752
1753         scores[hiscore_idx].rule = -1;
1754         scores[hiscore_idx].ifa = NULL;
1755
1756         rcu_read_lock();
1757
1758         /* Candidate Source Address (section 4)
1759          *  - multicast and link-local destination address,
1760          *    the set of candidate source address MUST only
1761          *    include addresses assigned to interfaces
1762          *    belonging to the same link as the outgoing
1763          *    interface.
1764          * (- For site-local destination addresses, the
1765          *    set of candidate source addresses MUST only
1766          *    include addresses assigned to interfaces
1767          *    belonging to the same site as the outgoing
1768          *    interface.)
1769          *  - "It is RECOMMENDED that the candidate source addresses
1770          *    be the set of unicast addresses assigned to the
1771          *    interface that will be used to send to the destination
1772          *    (the 'outgoing' interface)." (RFC 6724)
1773          */
1774         if (dst_dev) {
1775                 idev = __in6_dev_get(dst_dev);
1776                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1777                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1778                     (idev && idev->cnf.use_oif_addrs_only)) {
1779                         use_oif_addr = true;
1780                 }
1781         }
1782
1783         if (use_oif_addr) {
1784                 if (idev)
1785                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1786         } else {
1787                 const struct net_device *master;
1788                 int master_idx = 0;
1789
1790                 /* if dst_dev exists and is enslaved to an L3 device, then
1791                  * prefer addresses from dst_dev and then the master over
1792                  * any other enslaved devices in the L3 domain.
1793                  */
1794                 master = l3mdev_master_dev_rcu(dst_dev);
1795                 if (master) {
1796                         master_idx = master->ifindex;
1797
1798                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1799                                                             master, &dst,
1800                                                             scores, hiscore_idx);
1801
1802                         if (scores[hiscore_idx].ifa)
1803                                 goto out;
1804                 }
1805
1806                 for_each_netdev_rcu(net, dev) {
1807                         /* only consider addresses on devices in the
1808                          * same L3 domain
1809                          */
1810                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1811                                 continue;
1812                         idev = __in6_dev_get(dev);
1813                         if (!idev)
1814                                 continue;
1815                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1816                 }
1817         }
1818
1819 out:
1820         hiscore = &scores[hiscore_idx];
1821         if (!hiscore->ifa)
1822                 ret = -EADDRNOTAVAIL;
1823         else
1824                 *saddr = hiscore->ifa->addr;
1825
1826         rcu_read_unlock();
1827         return ret;
1828 }
1829 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1830
1831 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1832                       u32 banned_flags)
1833 {
1834         struct inet6_ifaddr *ifp;
1835         int err = -EADDRNOTAVAIL;
1836
1837         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1838                 if (ifp->scope > IFA_LINK)
1839                         break;
1840                 if (ifp->scope == IFA_LINK &&
1841                     !(ifp->flags & banned_flags)) {
1842                         *addr = ifp->addr;
1843                         err = 0;
1844                         break;
1845                 }
1846         }
1847         return err;
1848 }
1849
1850 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1851                     u32 banned_flags)
1852 {
1853         struct inet6_dev *idev;
1854         int err = -EADDRNOTAVAIL;
1855
1856         rcu_read_lock();
1857         idev = __in6_dev_get(dev);
1858         if (idev) {
1859                 read_lock_bh(&idev->lock);
1860                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1861                 read_unlock_bh(&idev->lock);
1862         }
1863         rcu_read_unlock();
1864         return err;
1865 }
1866
1867 static int ipv6_count_addresses(const struct inet6_dev *idev)
1868 {
1869         const struct inet6_ifaddr *ifp;
1870         int cnt = 0;
1871
1872         rcu_read_lock();
1873         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1874                 cnt++;
1875         rcu_read_unlock();
1876         return cnt;
1877 }
1878
1879 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1880                   const struct net_device *dev, int strict)
1881 {
1882         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1883                                        strict, IFA_F_TENTATIVE);
1884 }
1885 EXPORT_SYMBOL(ipv6_chk_addr);
1886
1887 /* device argument is used to find the L3 domain of interest. If
1888  * skip_dev_check is set, then the ifp device is not checked against
1889  * the passed in dev argument. So the 2 cases for addresses checks are:
1890  *   1. does the address exist in the L3 domain that dev is part of
1891  *      (skip_dev_check = true), or
1892  *
1893  *   2. does the address exist on the specific device
1894  *      (skip_dev_check = false)
1895  */
1896 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1897                             const struct net_device *dev, bool skip_dev_check,
1898                             int strict, u32 banned_flags)
1899 {
1900         unsigned int hash = inet6_addr_hash(net, addr);
1901         const struct net_device *l3mdev;
1902         struct inet6_ifaddr *ifp;
1903         u32 ifp_flags;
1904
1905         rcu_read_lock();
1906
1907         l3mdev = l3mdev_master_dev_rcu(dev);
1908         if (skip_dev_check)
1909                 dev = NULL;
1910
1911         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1912                 if (!net_eq(dev_net(ifp->idev->dev), net))
1913                         continue;
1914
1915                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1916                         continue;
1917
1918                 /* Decouple optimistic from tentative for evaluation here.
1919                  * Ban optimistic addresses explicitly, when required.
1920                  */
1921                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1922                             ? (ifp->flags&~IFA_F_TENTATIVE)
1923                             : ifp->flags;
1924                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1925                     !(ifp_flags&banned_flags) &&
1926                     (!dev || ifp->idev->dev == dev ||
1927                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1928                         rcu_read_unlock();
1929                         return 1;
1930                 }
1931         }
1932
1933         rcu_read_unlock();
1934         return 0;
1935 }
1936 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1937
1938
1939 /* Compares an address/prefix_len with addresses on device @dev.
1940  * If one is found it returns true.
1941  */
1942 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1943         const unsigned int prefix_len, struct net_device *dev)
1944 {
1945         const struct inet6_ifaddr *ifa;
1946         const struct inet6_dev *idev;
1947         bool ret = false;
1948
1949         rcu_read_lock();
1950         idev = __in6_dev_get(dev);
1951         if (idev) {
1952                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1953                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1954                         if (ret)
1955                                 break;
1956                 }
1957         }
1958         rcu_read_unlock();
1959
1960         return ret;
1961 }
1962 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1963
1964 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1965 {
1966         const struct inet6_ifaddr *ifa;
1967         const struct inet6_dev *idev;
1968         int     onlink;
1969
1970         onlink = 0;
1971         rcu_read_lock();
1972         idev = __in6_dev_get(dev);
1973         if (idev) {
1974                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1975                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1976                                                    ifa->prefix_len);
1977                         if (onlink)
1978                                 break;
1979                 }
1980         }
1981         rcu_read_unlock();
1982         return onlink;
1983 }
1984 EXPORT_SYMBOL(ipv6_chk_prefix);
1985
1986 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1987                                      struct net_device *dev, int strict)
1988 {
1989         unsigned int hash = inet6_addr_hash(net, addr);
1990         struct inet6_ifaddr *ifp, *result = NULL;
1991
1992         rcu_read_lock();
1993         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1994                 if (!net_eq(dev_net(ifp->idev->dev), net))
1995                         continue;
1996                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1997                         if (!dev || ifp->idev->dev == dev ||
1998                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1999                                 result = ifp;
2000                                 in6_ifa_hold(ifp);
2001                                 break;
2002                         }
2003                 }
2004         }
2005         rcu_read_unlock();
2006
2007         return result;
2008 }
2009
2010 /* Gets referenced address, destroys ifaddr */
2011
2012 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2013 {
2014         if (dad_failed)
2015                 ifp->flags |= IFA_F_DADFAILED;
2016
2017         if (ifp->flags&IFA_F_TEMPORARY) {
2018                 struct inet6_ifaddr *ifpub;
2019                 spin_lock_bh(&ifp->lock);
2020                 ifpub = ifp->ifpub;
2021                 if (ifpub) {
2022                         in6_ifa_hold(ifpub);
2023                         spin_unlock_bh(&ifp->lock);
2024                         ipv6_create_tempaddr(ifpub, true);
2025                         in6_ifa_put(ifpub);
2026                 } else {
2027                         spin_unlock_bh(&ifp->lock);
2028                 }
2029                 ipv6_del_addr(ifp);
2030         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2031                 spin_lock_bh(&ifp->lock);
2032                 addrconf_del_dad_work(ifp);
2033                 ifp->flags |= IFA_F_TENTATIVE;
2034                 if (dad_failed)
2035                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2036                 spin_unlock_bh(&ifp->lock);
2037                 if (dad_failed)
2038                         ipv6_ifa_notify(0, ifp);
2039                 in6_ifa_put(ifp);
2040         } else {
2041                 ipv6_del_addr(ifp);
2042         }
2043 }
2044
2045 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2046 {
2047         int err = -ENOENT;
2048
2049         spin_lock_bh(&ifp->lock);
2050         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2051                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2052                 err = 0;
2053         }
2054         spin_unlock_bh(&ifp->lock);
2055
2056         return err;
2057 }
2058
2059 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2060 {
2061         struct inet6_dev *idev = ifp->idev;
2062         struct net *net = dev_net(ifp->idev->dev);
2063
2064         if (addrconf_dad_end(ifp)) {
2065                 in6_ifa_put(ifp);
2066                 return;
2067         }
2068
2069         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2070                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2071
2072         spin_lock_bh(&ifp->lock);
2073
2074         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2075                 struct in6_addr new_addr;
2076                 struct inet6_ifaddr *ifp2;
2077                 int retries = ifp->stable_privacy_retry + 1;
2078                 struct ifa6_config cfg = {
2079                         .pfx = &new_addr,
2080                         .plen = ifp->prefix_len,
2081                         .ifa_flags = ifp->flags,
2082                         .valid_lft = ifp->valid_lft,
2083                         .preferred_lft = ifp->prefered_lft,
2084                         .scope = ifp->scope,
2085                 };
2086
2087                 if (retries > net->ipv6.sysctl.idgen_retries) {
2088                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2089                                              ifp->idev->dev->name);
2090                         goto errdad;
2091                 }
2092
2093                 new_addr = ifp->addr;
2094                 if (ipv6_generate_stable_address(&new_addr, retries,
2095                                                  idev))
2096                         goto errdad;
2097
2098                 spin_unlock_bh(&ifp->lock);
2099
2100                 if (idev->cnf.max_addresses &&
2101                     ipv6_count_addresses(idev) >=
2102                     idev->cnf.max_addresses)
2103                         goto lock_errdad;
2104
2105                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2106                                      ifp->idev->dev->name);
2107
2108                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2109                 if (IS_ERR(ifp2))
2110                         goto lock_errdad;
2111
2112                 spin_lock_bh(&ifp2->lock);
2113                 ifp2->stable_privacy_retry = retries;
2114                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2115                 spin_unlock_bh(&ifp2->lock);
2116
2117                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2118                 in6_ifa_put(ifp2);
2119 lock_errdad:
2120                 spin_lock_bh(&ifp->lock);
2121         }
2122
2123 errdad:
2124         /* transition from _POSTDAD to _ERRDAD */
2125         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2126         spin_unlock_bh(&ifp->lock);
2127
2128         addrconf_mod_dad_work(ifp, 0);
2129         in6_ifa_put(ifp);
2130 }
2131
2132 /* Join to solicited addr multicast group.
2133  * caller must hold RTNL */
2134 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2135 {
2136         struct in6_addr maddr;
2137
2138         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2139                 return;
2140
2141         addrconf_addr_solict_mult(addr, &maddr);
2142         ipv6_dev_mc_inc(dev, &maddr);
2143 }
2144
2145 /* caller must hold RTNL */
2146 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2147 {
2148         struct in6_addr maddr;
2149
2150         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2151                 return;
2152
2153         addrconf_addr_solict_mult(addr, &maddr);
2154         __ipv6_dev_mc_dec(idev, &maddr);
2155 }
2156
2157 /* caller must hold RTNL */
2158 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2159 {
2160         struct in6_addr addr;
2161
2162         if (ifp->prefix_len >= 127) /* RFC 6164 */
2163                 return;
2164         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2165         if (ipv6_addr_any(&addr))
2166                 return;
2167         __ipv6_dev_ac_inc(ifp->idev, &addr);
2168 }
2169
2170 /* caller must hold RTNL */
2171 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2172 {
2173         struct in6_addr addr;
2174
2175         if (ifp->prefix_len >= 127) /* RFC 6164 */
2176                 return;
2177         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2178         if (ipv6_addr_any(&addr))
2179                 return;
2180         __ipv6_dev_ac_dec(ifp->idev, &addr);
2181 }
2182
2183 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2184 {
2185         switch (dev->addr_len) {
2186         case ETH_ALEN:
2187                 memcpy(eui, dev->dev_addr, 3);
2188                 eui[3] = 0xFF;
2189                 eui[4] = 0xFE;
2190                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2191                 break;
2192         case EUI64_ADDR_LEN:
2193                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2194                 eui[0] ^= 2;
2195                 break;
2196         default:
2197                 return -1;
2198         }
2199
2200         return 0;
2201 }
2202
2203 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2204 {
2205         union fwnet_hwaddr *ha;
2206
2207         if (dev->addr_len != FWNET_ALEN)
2208                 return -1;
2209
2210         ha = (union fwnet_hwaddr *)dev->dev_addr;
2211
2212         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2213         eui[0] ^= 2;
2214         return 0;
2215 }
2216
2217 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2218 {
2219         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2220         if (dev->addr_len != ARCNET_ALEN)
2221                 return -1;
2222         memset(eui, 0, 7);
2223         eui[7] = *(u8 *)dev->dev_addr;
2224         return 0;
2225 }
2226
2227 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2228 {
2229         if (dev->addr_len != INFINIBAND_ALEN)
2230                 return -1;
2231         memcpy(eui, dev->dev_addr + 12, 8);
2232         eui[0] |= 2;
2233         return 0;
2234 }
2235
2236 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2237 {
2238         if (addr == 0)
2239                 return -1;
2240         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2241                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2242                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2243                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2244                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2245                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2246         eui[1] = 0;
2247         eui[2] = 0x5E;
2248         eui[3] = 0xFE;
2249         memcpy(eui + 4, &addr, 4);
2250         return 0;
2251 }
2252
2253 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2254 {
2255         if (dev->priv_flags & IFF_ISATAP)
2256                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2257         return -1;
2258 }
2259
2260 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2261 {
2262         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2263 }
2264
2265 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2266 {
2267         memcpy(eui, dev->perm_addr, 3);
2268         memcpy(eui + 5, dev->perm_addr + 3, 3);
2269         eui[3] = 0xFF;
2270         eui[4] = 0xFE;
2271         eui[0] ^= 2;
2272         return 0;
2273 }
2274
2275 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2276 {
2277         switch (dev->type) {
2278         case ARPHRD_ETHER:
2279         case ARPHRD_FDDI:
2280                 return addrconf_ifid_eui48(eui, dev);
2281         case ARPHRD_ARCNET:
2282                 return addrconf_ifid_arcnet(eui, dev);
2283         case ARPHRD_INFINIBAND:
2284                 return addrconf_ifid_infiniband(eui, dev);
2285         case ARPHRD_SIT:
2286                 return addrconf_ifid_sit(eui, dev);
2287         case ARPHRD_IPGRE:
2288         case ARPHRD_TUNNEL:
2289                 return addrconf_ifid_gre(eui, dev);
2290         case ARPHRD_6LOWPAN:
2291                 return addrconf_ifid_6lowpan(eui, dev);
2292         case ARPHRD_IEEE1394:
2293                 return addrconf_ifid_ieee1394(eui, dev);
2294         case ARPHRD_TUNNEL6:
2295         case ARPHRD_IP6GRE:
2296         case ARPHRD_RAWIP:
2297                 return addrconf_ifid_ip6tnl(eui, dev);
2298         }
2299         return -1;
2300 }
2301
2302 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2303 {
2304         int err = -1;
2305         struct inet6_ifaddr *ifp;
2306
2307         read_lock_bh(&idev->lock);
2308         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2309                 if (ifp->scope > IFA_LINK)
2310                         break;
2311                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2312                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2313                         err = 0;
2314                         break;
2315                 }
2316         }
2317         read_unlock_bh(&idev->lock);
2318         return err;
2319 }
2320
2321 /* Generation of a randomized Interface Identifier
2322  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2323  */
2324
2325 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2326 {
2327 regen:
2328         get_random_bytes(&addr->s6_addr[8], 8);
2329
2330         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2331          * check if generated address is not inappropriate:
2332          *
2333          * - Reserved IPv6 Interface Identifers
2334          * - XXX: already assigned to an address on the device
2335          */
2336
2337         /* Subnet-router anycast: 0000:0000:0000:0000 */
2338         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2339                 goto regen;
2340
2341         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2342          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2343          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2344          */
2345         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2346             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2347                 goto regen;
2348
2349         /* Reserved subnet anycast addresses */
2350         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2351             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2352                 goto regen;
2353 }
2354
2355 /*
2356  *      Add prefix route.
2357  */
2358
2359 static void
2360 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2361                       struct net_device *dev, unsigned long expires,
2362                       u32 flags, gfp_t gfp_flags)
2363 {
2364         struct fib6_config cfg = {
2365                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2366                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2367                 .fc_ifindex = dev->ifindex,
2368                 .fc_expires = expires,
2369                 .fc_dst_len = plen,
2370                 .fc_flags = RTF_UP | flags,
2371                 .fc_nlinfo.nl_net = dev_net(dev),
2372                 .fc_protocol = RTPROT_KERNEL,
2373                 .fc_type = RTN_UNICAST,
2374         };
2375
2376         cfg.fc_dst = *pfx;
2377
2378         /* Prevent useless cloning on PtP SIT.
2379            This thing is done here expecting that the whole
2380            class of non-broadcast devices need not cloning.
2381          */
2382 #if IS_ENABLED(CONFIG_IPV6_SIT)
2383         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2384                 cfg.fc_flags |= RTF_NONEXTHOP;
2385 #endif
2386
2387         ip6_route_add(&cfg, gfp_flags, NULL);
2388 }
2389
2390
2391 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2392                                                   int plen,
2393                                                   const struct net_device *dev,
2394                                                   u32 flags, u32 noflags,
2395                                                   bool no_gw)
2396 {
2397         struct fib6_node *fn;
2398         struct fib6_info *rt = NULL;
2399         struct fib6_table *table;
2400         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2401
2402         table = fib6_get_table(dev_net(dev), tb_id);
2403         if (!table)
2404                 return NULL;
2405
2406         rcu_read_lock();
2407         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2408         if (!fn)
2409                 goto out;
2410
2411         for_each_fib6_node_rt_rcu(fn) {
2412                 /* prefix routes only use builtin fib6_nh */
2413                 if (rt->nh)
2414                         continue;
2415
2416                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2417                         continue;
2418                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2419                         continue;
2420                 if ((rt->fib6_flags & flags) != flags)
2421                         continue;
2422                 if ((rt->fib6_flags & noflags) != 0)
2423                         continue;
2424                 if (!fib6_info_hold_safe(rt))
2425                         continue;
2426                 break;
2427         }
2428 out:
2429         rcu_read_unlock();
2430         return rt;
2431 }
2432
2433
2434 /* Create "default" multicast route to the interface */
2435
2436 static void addrconf_add_mroute(struct net_device *dev)
2437 {
2438         struct fib6_config cfg = {
2439                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2440                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2441                 .fc_ifindex = dev->ifindex,
2442                 .fc_dst_len = 8,
2443                 .fc_flags = RTF_UP,
2444                 .fc_type = RTN_UNICAST,
2445                 .fc_nlinfo.nl_net = dev_net(dev),
2446         };
2447
2448         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2449
2450         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2451 }
2452
2453 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2454 {
2455         struct inet6_dev *idev;
2456
2457         ASSERT_RTNL();
2458
2459         idev = ipv6_find_idev(dev);
2460         if (IS_ERR(idev))
2461                 return idev;
2462
2463         if (idev->cnf.disable_ipv6)
2464                 return ERR_PTR(-EACCES);
2465
2466         /* Add default multicast route */
2467         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2468                 addrconf_add_mroute(dev);
2469
2470         return idev;
2471 }
2472
2473 static void manage_tempaddrs(struct inet6_dev *idev,
2474                              struct inet6_ifaddr *ifp,
2475                              __u32 valid_lft, __u32 prefered_lft,
2476                              bool create, unsigned long now)
2477 {
2478         u32 flags;
2479         struct inet6_ifaddr *ift;
2480
2481         read_lock_bh(&idev->lock);
2482         /* update all temporary addresses in the list */
2483         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2484                 int age, max_valid, max_prefered;
2485
2486                 if (ifp != ift->ifpub)
2487                         continue;
2488
2489                 /* RFC 4941 section 3.3:
2490                  * If a received option will extend the lifetime of a public
2491                  * address, the lifetimes of temporary addresses should
2492                  * be extended, subject to the overall constraint that no
2493                  * temporary addresses should ever remain "valid" or "preferred"
2494                  * for a time longer than (TEMP_VALID_LIFETIME) or
2495                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2496                  */
2497                 age = (now - ift->cstamp) / HZ;
2498                 max_valid = idev->cnf.temp_valid_lft - age;
2499                 if (max_valid < 0)
2500                         max_valid = 0;
2501
2502                 max_prefered = idev->cnf.temp_prefered_lft -
2503                                idev->desync_factor - age;
2504                 if (max_prefered < 0)
2505                         max_prefered = 0;
2506
2507                 if (valid_lft > max_valid)
2508                         valid_lft = max_valid;
2509
2510                 if (prefered_lft > max_prefered)
2511                         prefered_lft = max_prefered;
2512
2513                 spin_lock(&ift->lock);
2514                 flags = ift->flags;
2515                 ift->valid_lft = valid_lft;
2516                 ift->prefered_lft = prefered_lft;
2517                 ift->tstamp = now;
2518                 if (prefered_lft > 0)
2519                         ift->flags &= ~IFA_F_DEPRECATED;
2520
2521                 spin_unlock(&ift->lock);
2522                 if (!(flags&IFA_F_TENTATIVE))
2523                         ipv6_ifa_notify(0, ift);
2524         }
2525
2526         if ((create || list_empty(&idev->tempaddr_list)) &&
2527             idev->cnf.use_tempaddr > 0) {
2528                 /* When a new public address is created as described
2529                  * in [ADDRCONF], also create a new temporary address.
2530                  * Also create a temporary address if it's enabled but
2531                  * no temporary address currently exists.
2532                  */
2533                 read_unlock_bh(&idev->lock);
2534                 ipv6_create_tempaddr(ifp, false);
2535         } else {
2536                 read_unlock_bh(&idev->lock);
2537         }
2538 }
2539
2540 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2541 {
2542         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2543                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2544 }
2545
2546 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2547                                  const struct prefix_info *pinfo,
2548                                  struct inet6_dev *in6_dev,
2549                                  const struct in6_addr *addr, int addr_type,
2550                                  u32 addr_flags, bool sllao, bool tokenized,
2551                                  __u32 valid_lft, u32 prefered_lft)
2552 {
2553         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2554         int create = 0;
2555
2556         if (!ifp && valid_lft) {
2557                 int max_addresses = in6_dev->cnf.max_addresses;
2558                 struct ifa6_config cfg = {
2559                         .pfx = addr,
2560                         .plen = pinfo->prefix_len,
2561                         .ifa_flags = addr_flags,
2562                         .valid_lft = valid_lft,
2563                         .preferred_lft = prefered_lft,
2564                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2565                 };
2566
2567 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2568                 if ((net->ipv6.devconf_all->optimistic_dad ||
2569                      in6_dev->cnf.optimistic_dad) &&
2570                     !net->ipv6.devconf_all->forwarding && sllao)
2571                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2572 #endif
2573
2574                 /* Do not allow to create too much of autoconfigured
2575                  * addresses; this would be too easy way to crash kernel.
2576                  */
2577                 if (!max_addresses ||
2578                     ipv6_count_addresses(in6_dev) < max_addresses)
2579                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2580
2581                 if (IS_ERR_OR_NULL(ifp))
2582                         return -1;
2583
2584                 create = 1;
2585                 spin_lock_bh(&ifp->lock);
2586                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2587                 ifp->cstamp = jiffies;
2588                 ifp->tokenized = tokenized;
2589                 spin_unlock_bh(&ifp->lock);
2590                 addrconf_dad_start(ifp);
2591         }
2592
2593         if (ifp) {
2594                 u32 flags;
2595                 unsigned long now;
2596                 u32 stored_lft;
2597
2598                 /* Update lifetime (RFC4862 5.5.3 e)
2599                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2600                  * improve the reaction of SLAAC to renumbering events
2601                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2602                  */
2603                 spin_lock_bh(&ifp->lock);
2604                 now = jiffies;
2605                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2606                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2607                 else
2608                         stored_lft = 0;
2609
2610                 if (!create && stored_lft) {
2611                         ifp->valid_lft = valid_lft;
2612                         ifp->prefered_lft = prefered_lft;
2613                         ifp->tstamp = now;
2614                         flags = ifp->flags;
2615                         ifp->flags &= ~IFA_F_DEPRECATED;
2616                         spin_unlock_bh(&ifp->lock);
2617
2618                         if (!(flags&IFA_F_TENTATIVE))
2619                                 ipv6_ifa_notify(0, ifp);
2620                 } else
2621                         spin_unlock_bh(&ifp->lock);
2622
2623                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2624                                  create, now);
2625
2626                 in6_ifa_put(ifp);
2627                 addrconf_verify();
2628         }
2629
2630         return 0;
2631 }
2632 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2633
2634 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2635 {
2636         struct prefix_info *pinfo;
2637         __u32 valid_lft;
2638         __u32 prefered_lft;
2639         int addr_type, err;
2640         u32 addr_flags = 0;
2641         struct inet6_dev *in6_dev;
2642         struct net *net = dev_net(dev);
2643
2644         pinfo = (struct prefix_info *) opt;
2645
2646         if (len < sizeof(struct prefix_info)) {
2647                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2648                 return;
2649         }
2650
2651         /*
2652          *      Validation checks ([ADDRCONF], page 19)
2653          */
2654
2655         addr_type = ipv6_addr_type(&pinfo->prefix);
2656
2657         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2658                 return;
2659
2660         valid_lft = ntohl(pinfo->valid);
2661         prefered_lft = ntohl(pinfo->prefered);
2662
2663         if (prefered_lft > valid_lft) {
2664                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2665                 return;
2666         }
2667
2668         in6_dev = in6_dev_get(dev);
2669
2670         if (!in6_dev) {
2671                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2672                                     dev->name);
2673                 return;
2674         }
2675
2676         /*
2677          *      Two things going on here:
2678          *      1) Add routes for on-link prefixes
2679          *      2) Configure prefixes with the auto flag set
2680          */
2681
2682         if (pinfo->onlink) {
2683                 struct fib6_info *rt;
2684                 unsigned long rt_expires;
2685
2686                 /* Avoid arithmetic overflow. Really, we could
2687                  * save rt_expires in seconds, likely valid_lft,
2688                  * but it would require division in fib gc, that it
2689                  * not good.
2690                  */
2691                 if (HZ > USER_HZ)
2692                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2693                 else
2694                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2695
2696                 if (addrconf_finite_timeout(rt_expires))
2697                         rt_expires *= HZ;
2698
2699                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2700                                                pinfo->prefix_len,
2701                                                dev,
2702                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2703                                                RTF_DEFAULT, true);
2704
2705                 if (rt) {
2706                         /* Autoconf prefix route */
2707                         if (valid_lft == 0) {
2708                                 ip6_del_rt(net, rt, false);
2709                                 rt = NULL;
2710                         } else if (addrconf_finite_timeout(rt_expires)) {
2711                                 /* not infinity */
2712                                 fib6_set_expires(rt, jiffies + rt_expires);
2713                         } else {
2714                                 fib6_clean_expires(rt);
2715                         }
2716                 } else if (valid_lft) {
2717                         clock_t expires = 0;
2718                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2719                         if (addrconf_finite_timeout(rt_expires)) {
2720                                 /* not infinity */
2721                                 flags |= RTF_EXPIRES;
2722                                 expires = jiffies_to_clock_t(rt_expires);
2723                         }
2724                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2725                                               0, dev, expires, flags,
2726                                               GFP_ATOMIC);
2727                 }
2728                 fib6_info_release(rt);
2729         }
2730
2731         /* Try to figure out our local address for this prefix */
2732
2733         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2734                 struct in6_addr addr;
2735                 bool tokenized = false, dev_addr_generated = false;
2736
2737                 if (pinfo->prefix_len == 64) {
2738                         memcpy(&addr, &pinfo->prefix, 8);
2739
2740                         if (!ipv6_addr_any(&in6_dev->token)) {
2741                                 read_lock_bh(&in6_dev->lock);
2742                                 memcpy(addr.s6_addr + 8,
2743                                        in6_dev->token.s6_addr + 8, 8);
2744                                 read_unlock_bh(&in6_dev->lock);
2745                                 tokenized = true;
2746                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2747                                    !ipv6_generate_stable_address(&addr, 0,
2748                                                                  in6_dev)) {
2749                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2750                                 goto ok;
2751                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2752                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2753                                 goto put;
2754                         } else {
2755                                 dev_addr_generated = true;
2756                         }
2757                         goto ok;
2758                 }
2759                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2760                                     pinfo->prefix_len);
2761                 goto put;
2762
2763 ok:
2764                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2765                                                    &addr, addr_type,
2766                                                    addr_flags, sllao,
2767                                                    tokenized, valid_lft,
2768                                                    prefered_lft);
2769                 if (err)
2770                         goto put;
2771
2772                 /* Ignore error case here because previous prefix add addr was
2773                  * successful which will be notified.
2774                  */
2775                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2776                                               addr_type, addr_flags, sllao,
2777                                               tokenized, valid_lft,
2778                                               prefered_lft,
2779                                               dev_addr_generated);
2780         }
2781         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2782 put:
2783         in6_dev_put(in6_dev);
2784 }
2785
2786 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2787                 struct in6_ifreq *ireq)
2788 {
2789         struct ip_tunnel_parm p = { };
2790         int err;
2791
2792         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2793                 return -EADDRNOTAVAIL;
2794
2795         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2796         p.iph.version = 4;
2797         p.iph.ihl = 5;
2798         p.iph.protocol = IPPROTO_IPV6;
2799         p.iph.ttl = 64;
2800
2801         if (!dev->netdev_ops->ndo_tunnel_ctl)
2802                 return -EOPNOTSUPP;
2803         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2804         if (err)
2805                 return err;
2806
2807         dev = __dev_get_by_name(net, p.name);
2808         if (!dev)
2809                 return -ENOBUFS;
2810         return dev_open(dev, NULL);
2811 }
2812
2813 /*
2814  *      Set destination address.
2815  *      Special case for SIT interfaces where we create a new "virtual"
2816  *      device.
2817  */
2818 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2819 {
2820         struct net_device *dev;
2821         struct in6_ifreq ireq;
2822         int err = -ENODEV;
2823
2824         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2825                 return -ENODEV;
2826         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2827                 return -EFAULT;
2828
2829         rtnl_lock();
2830         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2831         if (dev && dev->type == ARPHRD_SIT)
2832                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2833         rtnl_unlock();
2834         return err;
2835 }
2836
2837 static int ipv6_mc_config(struct sock *sk, bool join,
2838                           const struct in6_addr *addr, int ifindex)
2839 {
2840         int ret;
2841
2842         ASSERT_RTNL();
2843
2844         lock_sock(sk);
2845         if (join)
2846                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2847         else
2848                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2849         release_sock(sk);
2850
2851         return ret;
2852 }
2853
2854 /*
2855  *      Manual configuration of address on an interface
2856  */
2857 static int inet6_addr_add(struct net *net, int ifindex,
2858                           struct ifa6_config *cfg,
2859                           struct netlink_ext_ack *extack)
2860 {
2861         struct inet6_ifaddr *ifp;
2862         struct inet6_dev *idev;
2863         struct net_device *dev;
2864         unsigned long timeout;
2865         clock_t expires;
2866         u32 flags;
2867
2868         ASSERT_RTNL();
2869
2870         if (cfg->plen > 128)
2871                 return -EINVAL;
2872
2873         /* check the lifetime */
2874         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2875                 return -EINVAL;
2876
2877         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2878                 return -EINVAL;
2879
2880         dev = __dev_get_by_index(net, ifindex);
2881         if (!dev)
2882                 return -ENODEV;
2883
2884         idev = addrconf_add_dev(dev);
2885         if (IS_ERR(idev))
2886                 return PTR_ERR(idev);
2887
2888         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2889                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2890                                          true, cfg->pfx, ifindex);
2891
2892                 if (ret < 0)
2893                         return ret;
2894         }
2895
2896         cfg->scope = ipv6_addr_scope(cfg->pfx);
2897
2898         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2899         if (addrconf_finite_timeout(timeout)) {
2900                 expires = jiffies_to_clock_t(timeout * HZ);
2901                 cfg->valid_lft = timeout;
2902                 flags = RTF_EXPIRES;
2903         } else {
2904                 expires = 0;
2905                 flags = 0;
2906                 cfg->ifa_flags |= IFA_F_PERMANENT;
2907         }
2908
2909         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2910         if (addrconf_finite_timeout(timeout)) {
2911                 if (timeout == 0)
2912                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2913                 cfg->preferred_lft = timeout;
2914         }
2915
2916         ifp = ipv6_add_addr(idev, cfg, true, extack);
2917         if (!IS_ERR(ifp)) {
2918                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2919                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2920                                               ifp->rt_priority, dev, expires,
2921                                               flags, GFP_KERNEL);
2922                 }
2923
2924                 /* Send a netlink notification if DAD is enabled and
2925                  * optimistic flag is not set
2926                  */
2927                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2928                         ipv6_ifa_notify(0, ifp);
2929                 /*
2930                  * Note that section 3.1 of RFC 4429 indicates
2931                  * that the Optimistic flag should not be set for
2932                  * manually configured addresses
2933                  */
2934                 addrconf_dad_start(ifp);
2935                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2936                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2937                                          cfg->preferred_lft, true, jiffies);
2938                 in6_ifa_put(ifp);
2939                 addrconf_verify_rtnl();
2940                 return 0;
2941         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2942                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2943                                cfg->pfx, ifindex);
2944         }
2945
2946         return PTR_ERR(ifp);
2947 }
2948
2949 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2950                           const struct in6_addr *pfx, unsigned int plen)
2951 {
2952         struct inet6_ifaddr *ifp;
2953         struct inet6_dev *idev;
2954         struct net_device *dev;
2955
2956         if (plen > 128)
2957                 return -EINVAL;
2958
2959         dev = __dev_get_by_index(net, ifindex);
2960         if (!dev)
2961                 return -ENODEV;
2962
2963         idev = __in6_dev_get(dev);
2964         if (!idev)
2965                 return -ENXIO;
2966
2967         read_lock_bh(&idev->lock);
2968         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2969                 if (ifp->prefix_len == plen &&
2970                     ipv6_addr_equal(pfx, &ifp->addr)) {
2971                         in6_ifa_hold(ifp);
2972                         read_unlock_bh(&idev->lock);
2973
2974                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2975                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2976                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2977                                                  jiffies);
2978                         ipv6_del_addr(ifp);
2979                         addrconf_verify_rtnl();
2980                         if (ipv6_addr_is_multicast(pfx)) {
2981                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2982                                                false, pfx, dev->ifindex);
2983                         }
2984                         return 0;
2985                 }
2986         }
2987         read_unlock_bh(&idev->lock);
2988         return -EADDRNOTAVAIL;
2989 }
2990
2991
2992 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2993 {
2994         struct ifa6_config cfg = {
2995                 .ifa_flags = IFA_F_PERMANENT,
2996                 .preferred_lft = INFINITY_LIFE_TIME,
2997                 .valid_lft = INFINITY_LIFE_TIME,
2998         };
2999         struct in6_ifreq ireq;
3000         int err;
3001
3002         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3003                 return -EPERM;
3004
3005         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3006                 return -EFAULT;
3007
3008         cfg.pfx = &ireq.ifr6_addr;
3009         cfg.plen = ireq.ifr6_prefixlen;
3010
3011         rtnl_lock();
3012         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3013         rtnl_unlock();
3014         return err;
3015 }
3016
3017 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3018 {
3019         struct in6_ifreq ireq;
3020         int err;
3021
3022         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3023                 return -EPERM;
3024
3025         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3026                 return -EFAULT;
3027
3028         rtnl_lock();
3029         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3030                              ireq.ifr6_prefixlen);
3031         rtnl_unlock();
3032         return err;
3033 }
3034
3035 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3036                      int plen, int scope)
3037 {
3038         struct inet6_ifaddr *ifp;
3039         struct ifa6_config cfg = {
3040                 .pfx = addr,
3041                 .plen = plen,
3042                 .ifa_flags = IFA_F_PERMANENT,
3043                 .valid_lft = INFINITY_LIFE_TIME,
3044                 .preferred_lft = INFINITY_LIFE_TIME,
3045                 .scope = scope
3046         };
3047
3048         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3049         if (!IS_ERR(ifp)) {
3050                 spin_lock_bh(&ifp->lock);
3051                 ifp->flags &= ~IFA_F_TENTATIVE;
3052                 spin_unlock_bh(&ifp->lock);
3053                 rt_genid_bump_ipv6(dev_net(idev->dev));
3054                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3055                 in6_ifa_put(ifp);
3056         }
3057 }
3058
3059 #if IS_ENABLED(CONFIG_IPV6_SIT)
3060 static void sit_add_v4_addrs(struct inet6_dev *idev)
3061 {
3062         struct in6_addr addr;
3063         struct net_device *dev;
3064         struct net *net = dev_net(idev->dev);
3065         int scope, plen;
3066         u32 pflags = 0;
3067
3068         ASSERT_RTNL();
3069
3070         memset(&addr, 0, sizeof(struct in6_addr));
3071         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3072
3073         if (idev->dev->flags&IFF_POINTOPOINT) {
3074                 addr.s6_addr32[0] = htonl(0xfe800000);
3075                 scope = IFA_LINK;
3076                 plen = 64;
3077         } else {
3078                 scope = IPV6_ADDR_COMPATv4;
3079                 plen = 96;
3080                 pflags |= RTF_NONEXTHOP;
3081         }
3082
3083         if (addr.s6_addr32[3]) {
3084                 add_addr(idev, &addr, plen, scope);
3085                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3086                                       GFP_KERNEL);
3087                 return;
3088         }
3089
3090         for_each_netdev(net, dev) {
3091                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3092                 if (in_dev && (dev->flags & IFF_UP)) {
3093                         struct in_ifaddr *ifa;
3094                         int flag = scope;
3095
3096                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3097                                 addr.s6_addr32[3] = ifa->ifa_local;
3098
3099                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3100                                         continue;
3101                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3102                                         if (idev->dev->flags&IFF_POINTOPOINT)
3103                                                 continue;
3104                                         flag |= IFA_HOST;
3105                                 }
3106
3107                                 add_addr(idev, &addr, plen, flag);
3108                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3109                                                       0, pflags, GFP_KERNEL);
3110                         }
3111                 }
3112         }
3113 }
3114 #endif
3115
3116 static void init_loopback(struct net_device *dev)
3117 {
3118         struct inet6_dev  *idev;
3119
3120         /* ::1 */
3121
3122         ASSERT_RTNL();
3123
3124         idev = ipv6_find_idev(dev);
3125         if (IS_ERR(idev)) {
3126                 pr_debug("%s: add_dev failed\n", __func__);
3127                 return;
3128         }
3129
3130         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3131 }
3132
3133 void addrconf_add_linklocal(struct inet6_dev *idev,
3134                             const struct in6_addr *addr, u32 flags)
3135 {
3136         struct ifa6_config cfg = {
3137                 .pfx = addr,
3138                 .plen = 64,
3139                 .ifa_flags = flags | IFA_F_PERMANENT,
3140                 .valid_lft = INFINITY_LIFE_TIME,
3141                 .preferred_lft = INFINITY_LIFE_TIME,
3142                 .scope = IFA_LINK
3143         };
3144         struct inet6_ifaddr *ifp;
3145
3146 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3147         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3148              idev->cnf.optimistic_dad) &&
3149             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3150                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3151 #endif
3152
3153         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3154         if (!IS_ERR(ifp)) {
3155                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3156                                       0, 0, GFP_ATOMIC);
3157                 addrconf_dad_start(ifp);
3158                 in6_ifa_put(ifp);
3159         }
3160 }
3161 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3162
3163 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3164 {
3165         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3166                 return true;
3167
3168         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3169             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3170                 return true;
3171
3172         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3173             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3174                 return true;
3175
3176         return false;
3177 }
3178
3179 static int ipv6_generate_stable_address(struct in6_addr *address,
3180                                         u8 dad_count,
3181                                         const struct inet6_dev *idev)
3182 {
3183         static DEFINE_SPINLOCK(lock);
3184         static __u32 digest[SHA1_DIGEST_WORDS];
3185         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3186
3187         static union {
3188                 char __data[SHA1_BLOCK_SIZE];
3189                 struct {
3190                         struct in6_addr secret;
3191                         __be32 prefix[2];
3192                         unsigned char hwaddr[MAX_ADDR_LEN];
3193                         u8 dad_count;
3194                 } __packed;
3195         } data;
3196
3197         struct in6_addr secret;
3198         struct in6_addr temp;
3199         struct net *net = dev_net(idev->dev);
3200
3201         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3202
3203         if (idev->cnf.stable_secret.initialized)
3204                 secret = idev->cnf.stable_secret.secret;
3205         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3206                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3207         else
3208                 return -1;
3209
3210 retry:
3211         spin_lock_bh(&lock);
3212
3213         sha1_init(digest);
3214         memset(&data, 0, sizeof(data));
3215         memset(workspace, 0, sizeof(workspace));
3216         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3217         data.prefix[0] = address->s6_addr32[0];
3218         data.prefix[1] = address->s6_addr32[1];
3219         data.secret = secret;
3220         data.dad_count = dad_count;
3221
3222         sha1_transform(digest, data.__data, workspace);
3223
3224         temp = *address;
3225         temp.s6_addr32[2] = (__force __be32)digest[0];
3226         temp.s6_addr32[3] = (__force __be32)digest[1];
3227
3228         spin_unlock_bh(&lock);
3229
3230         if (ipv6_reserved_interfaceid(temp)) {
3231                 dad_count++;
3232                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3233                         return -1;
3234                 goto retry;
3235         }
3236
3237         *address = temp;
3238         return 0;
3239 }
3240
3241 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3242 {
3243         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3244
3245         if (s->initialized)
3246                 return;
3247         s = &idev->cnf.stable_secret;
3248         get_random_bytes(&s->secret, sizeof(s->secret));
3249         s->initialized = true;
3250 }
3251
3252 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3253 {
3254         struct in6_addr addr;
3255
3256         /* no link local addresses on L3 master devices */
3257         if (netif_is_l3_master(idev->dev))
3258                 return;
3259
3260         /* no link local addresses on devices flagged as slaves */
3261         if (idev->dev->flags & IFF_SLAVE)
3262                 return;
3263
3264         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3265
3266         switch (idev->cnf.addr_gen_mode) {
3267         case IN6_ADDR_GEN_MODE_RANDOM:
3268                 ipv6_gen_mode_random_init(idev);
3269                 fallthrough;
3270         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3271                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3272                         addrconf_add_linklocal(idev, &addr,
3273                                                IFA_F_STABLE_PRIVACY);
3274                 else if (prefix_route)
3275                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3276                                               0, 0, GFP_KERNEL);
3277                 break;
3278         case IN6_ADDR_GEN_MODE_EUI64:
3279                 /* addrconf_add_linklocal also adds a prefix_route and we
3280                  * only need to care about prefix routes if ipv6_generate_eui64
3281                  * couldn't generate one.
3282                  */
3283                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3284                         addrconf_add_linklocal(idev, &addr, 0);
3285                 else if (prefix_route)
3286                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3287                                               0, 0, GFP_KERNEL);
3288                 break;
3289         case IN6_ADDR_GEN_MODE_NONE:
3290         default:
3291                 /* will not add any link local address */
3292                 break;
3293         }
3294 }
3295
3296 static void addrconf_dev_config(struct net_device *dev)
3297 {
3298         struct inet6_dev *idev;
3299
3300         ASSERT_RTNL();
3301
3302         if ((dev->type != ARPHRD_ETHER) &&
3303             (dev->type != ARPHRD_FDDI) &&
3304             (dev->type != ARPHRD_ARCNET) &&
3305             (dev->type != ARPHRD_INFINIBAND) &&
3306             (dev->type != ARPHRD_IEEE1394) &&
3307             (dev->type != ARPHRD_TUNNEL6) &&
3308             (dev->type != ARPHRD_6LOWPAN) &&
3309             (dev->type != ARPHRD_IP6GRE) &&
3310             (dev->type != ARPHRD_IPGRE) &&
3311             (dev->type != ARPHRD_TUNNEL) &&
3312             (dev->type != ARPHRD_NONE) &&
3313             (dev->type != ARPHRD_RAWIP)) {
3314                 /* Alas, we support only Ethernet autoconfiguration. */
3315                 idev = __in6_dev_get(dev);
3316                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3317                     dev->flags & IFF_MULTICAST)
3318                         ipv6_mc_up(idev);
3319                 return;
3320         }
3321
3322         idev = addrconf_add_dev(dev);
3323         if (IS_ERR(idev))
3324                 return;
3325
3326         /* this device type has no EUI support */
3327         if (dev->type == ARPHRD_NONE &&
3328             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3329                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3330
3331         addrconf_addr_gen(idev, false);
3332 }
3333
3334 #if IS_ENABLED(CONFIG_IPV6_SIT)
3335 static void addrconf_sit_config(struct net_device *dev)
3336 {
3337         struct inet6_dev *idev;
3338
3339         ASSERT_RTNL();
3340
3341         /*
3342          * Configure the tunnel with one of our IPv4
3343          * addresses... we should configure all of
3344          * our v4 addrs in the tunnel
3345          */
3346
3347         idev = ipv6_find_idev(dev);
3348         if (IS_ERR(idev)) {
3349                 pr_debug("%s: add_dev failed\n", __func__);
3350                 return;
3351         }
3352
3353         if (dev->priv_flags & IFF_ISATAP) {
3354                 addrconf_addr_gen(idev, false);
3355                 return;
3356         }
3357
3358         sit_add_v4_addrs(idev);
3359
3360         if (dev->flags&IFF_POINTOPOINT)
3361                 addrconf_add_mroute(dev);
3362 }
3363 #endif
3364
3365 #if IS_ENABLED(CONFIG_NET_IPGRE)
3366 static void addrconf_gre_config(struct net_device *dev)
3367 {
3368         struct inet6_dev *idev;
3369
3370         ASSERT_RTNL();
3371
3372         idev = ipv6_find_idev(dev);
3373         if (IS_ERR(idev)) {
3374                 pr_debug("%s: add_dev failed\n", __func__);
3375                 return;
3376         }
3377
3378         addrconf_addr_gen(idev, true);
3379         if (dev->flags & IFF_POINTOPOINT)
3380                 addrconf_add_mroute(dev);
3381 }
3382 #endif
3383
3384 static int fixup_permanent_addr(struct net *net,
3385                                 struct inet6_dev *idev,
3386                                 struct inet6_ifaddr *ifp)
3387 {
3388         /* !fib6_node means the host route was removed from the
3389          * FIB, for example, if 'lo' device is taken down. In that
3390          * case regenerate the host route.
3391          */
3392         if (!ifp->rt || !ifp->rt->fib6_node) {
3393                 struct fib6_info *f6i, *prev;
3394
3395                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3396                                          GFP_ATOMIC);
3397                 if (IS_ERR(f6i))
3398                         return PTR_ERR(f6i);
3399
3400                 /* ifp->rt can be accessed outside of rtnl */
3401                 spin_lock(&ifp->lock);
3402                 prev = ifp->rt;
3403                 ifp->rt = f6i;
3404                 spin_unlock(&ifp->lock);
3405
3406                 fib6_info_release(prev);
3407         }
3408
3409         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3410                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3411                                       ifp->rt_priority, idev->dev, 0, 0,
3412                                       GFP_ATOMIC);
3413         }
3414
3415         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3416                 addrconf_dad_start(ifp);
3417
3418         return 0;
3419 }
3420
3421 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3422 {
3423         struct inet6_ifaddr *ifp, *tmp;
3424         struct inet6_dev *idev;
3425
3426         idev = __in6_dev_get(dev);
3427         if (!idev)
3428                 return;
3429
3430         write_lock_bh(&idev->lock);
3431
3432         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3433                 if ((ifp->flags & IFA_F_PERMANENT) &&
3434                     fixup_permanent_addr(net, idev, ifp) < 0) {
3435                         write_unlock_bh(&idev->lock);
3436                         in6_ifa_hold(ifp);
3437                         ipv6_del_addr(ifp);
3438                         write_lock_bh(&idev->lock);
3439
3440                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3441                                              idev->dev->name, &ifp->addr);
3442                 }
3443         }
3444
3445         write_unlock_bh(&idev->lock);
3446 }
3447
3448 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3449                            void *ptr)
3450 {
3451         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3452         struct netdev_notifier_change_info *change_info;
3453         struct netdev_notifier_changeupper_info *info;
3454         struct inet6_dev *idev = __in6_dev_get(dev);
3455         struct net *net = dev_net(dev);
3456         int run_pending = 0;
3457         int err;
3458
3459         switch (event) {
3460         case NETDEV_REGISTER:
3461                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3462                         idev = ipv6_add_dev(dev);
3463                         if (IS_ERR(idev))
3464                                 return notifier_from_errno(PTR_ERR(idev));
3465                 }
3466                 break;
3467
3468         case NETDEV_CHANGEMTU:
3469                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3470                 if (dev->mtu < IPV6_MIN_MTU) {
3471                         addrconf_ifdown(dev, dev != net->loopback_dev);
3472                         break;
3473                 }
3474
3475                 if (idev) {
3476                         rt6_mtu_change(dev, dev->mtu);
3477                         idev->cnf.mtu6 = dev->mtu;
3478                         break;
3479                 }
3480
3481                 /* allocate new idev */
3482                 idev = ipv6_add_dev(dev);
3483                 if (IS_ERR(idev))
3484                         break;
3485
3486                 /* device is still not ready */
3487                 if (!(idev->if_flags & IF_READY))
3488                         break;
3489
3490                 run_pending = 1;
3491                 fallthrough;
3492         case NETDEV_UP:
3493         case NETDEV_CHANGE:
3494                 if (dev->flags & IFF_SLAVE)
3495                         break;
3496
3497                 if (idev && idev->cnf.disable_ipv6)
3498                         break;
3499
3500                 if (event == NETDEV_UP) {
3501                         /* restore routes for permanent addresses */
3502                         addrconf_permanent_addr(net, dev);
3503
3504                         if (!addrconf_link_ready(dev)) {
3505                                 /* device is not ready yet. */
3506                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3507                                          dev->name);
3508                                 break;
3509                         }
3510
3511                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3512                                 idev = ipv6_add_dev(dev);
3513
3514                         if (!IS_ERR_OR_NULL(idev)) {
3515                                 idev->if_flags |= IF_READY;
3516                                 run_pending = 1;
3517                         }
3518                 } else if (event == NETDEV_CHANGE) {
3519                         if (!addrconf_link_ready(dev)) {
3520                                 /* device is still not ready. */
3521                                 rt6_sync_down_dev(dev, event);
3522                                 break;
3523                         }
3524
3525                         if (!IS_ERR_OR_NULL(idev)) {
3526                                 if (idev->if_flags & IF_READY) {
3527                                         /* device is already configured -
3528                                          * but resend MLD reports, we might
3529                                          * have roamed and need to update
3530                                          * multicast snooping switches
3531                                          */
3532                                         ipv6_mc_up(idev);
3533                                         change_info = ptr;
3534                                         if (change_info->flags_changed & IFF_NOARP)
3535                                                 addrconf_dad_run(idev, true);
3536                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3537                                         break;
3538                                 }
3539                                 idev->if_flags |= IF_READY;
3540                         }
3541
3542                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3543                                 dev->name);
3544
3545                         run_pending = 1;
3546                 }
3547
3548                 switch (dev->type) {
3549 #if IS_ENABLED(CONFIG_IPV6_SIT)
3550                 case ARPHRD_SIT:
3551                         addrconf_sit_config(dev);
3552                         break;
3553 #endif
3554 #if IS_ENABLED(CONFIG_NET_IPGRE)
3555                 case ARPHRD_IPGRE:
3556                         addrconf_gre_config(dev);
3557                         break;
3558 #endif
3559                 case ARPHRD_LOOPBACK:
3560                         init_loopback(dev);
3561                         break;
3562
3563                 default:
3564                         addrconf_dev_config(dev);
3565                         break;
3566                 }
3567
3568                 if (!IS_ERR_OR_NULL(idev)) {
3569                         if (run_pending)
3570                                 addrconf_dad_run(idev, false);
3571
3572                         /* Device has an address by now */
3573                         rt6_sync_up(dev, RTNH_F_DEAD);
3574
3575                         /*
3576                          * If the MTU changed during the interface down,
3577                          * when the interface up, the changed MTU must be
3578                          * reflected in the idev as well as routers.
3579                          */
3580                         if (idev->cnf.mtu6 != dev->mtu &&
3581                             dev->mtu >= IPV6_MIN_MTU) {
3582                                 rt6_mtu_change(dev, dev->mtu);
3583                                 idev->cnf.mtu6 = dev->mtu;
3584                         }
3585                         idev->tstamp = jiffies;
3586                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3587
3588                         /*
3589                          * If the changed mtu during down is lower than
3590                          * IPV6_MIN_MTU stop IPv6 on this interface.
3591                          */
3592                         if (dev->mtu < IPV6_MIN_MTU)
3593                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3594                 }
3595                 break;
3596
3597         case NETDEV_DOWN:
3598         case NETDEV_UNREGISTER:
3599                 /*
3600                  *      Remove all addresses from this interface.
3601                  */
3602                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3603                 break;
3604
3605         case NETDEV_CHANGENAME:
3606                 if (idev) {
3607                         snmp6_unregister_dev(idev);
3608                         addrconf_sysctl_unregister(idev);
3609                         err = addrconf_sysctl_register(idev);
3610                         if (err)
3611                                 return notifier_from_errno(err);
3612                         err = snmp6_register_dev(idev);
3613                         if (err) {
3614                                 addrconf_sysctl_unregister(idev);
3615                                 return notifier_from_errno(err);
3616                         }
3617                 }
3618                 break;
3619
3620         case NETDEV_PRE_TYPE_CHANGE:
3621         case NETDEV_POST_TYPE_CHANGE:
3622                 if (idev)
3623                         addrconf_type_change(dev, event);
3624                 break;
3625
3626         case NETDEV_CHANGEUPPER:
3627                 info = ptr;
3628
3629                 /* flush all routes if dev is linked to or unlinked from
3630                  * an L3 master device (e.g., VRF)
3631                  */
3632                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3633                         addrconf_ifdown(dev, 0);
3634         }
3635
3636         return NOTIFY_OK;
3637 }
3638
3639 /*
3640  *      addrconf module should be notified of a device going up
3641  */
3642 static struct notifier_block ipv6_dev_notf = {
3643         .notifier_call = addrconf_notify,
3644         .priority = ADDRCONF_NOTIFY_PRIORITY,
3645 };
3646
3647 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3648 {
3649         struct inet6_dev *idev;
3650         ASSERT_RTNL();
3651
3652         idev = __in6_dev_get(dev);
3653
3654         if (event == NETDEV_POST_TYPE_CHANGE)
3655                 ipv6_mc_remap(idev);
3656         else if (event == NETDEV_PRE_TYPE_CHANGE)
3657                 ipv6_mc_unmap(idev);
3658 }
3659
3660 static bool addr_is_local(const struct in6_addr *addr)
3661 {
3662         return ipv6_addr_type(addr) &
3663                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3664 }
3665
3666 static int addrconf_ifdown(struct net_device *dev, int how)
3667 {
3668         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3669         struct net *net = dev_net(dev);
3670         struct inet6_dev *idev;
3671         struct inet6_ifaddr *ifa, *tmp;
3672         bool keep_addr = false;
3673         int state, i;
3674
3675         ASSERT_RTNL();
3676
3677         rt6_disable_ip(dev, event);
3678
3679         idev = __in6_dev_get(dev);
3680         if (!idev)
3681                 return -ENODEV;
3682
3683         /*
3684          * Step 1: remove reference to ipv6 device from parent device.
3685          *         Do not dev_put!
3686          */
3687         if (how) {
3688                 idev->dead = 1;
3689
3690                 /* protected by rtnl_lock */
3691                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3692
3693                 /* Step 1.5: remove snmp6 entry */
3694                 snmp6_unregister_dev(idev);
3695
3696         }
3697
3698         /* combine the user config with event to determine if permanent
3699          * addresses are to be removed from address hash table
3700          */
3701         if (!how && !idev->cnf.disable_ipv6) {
3702                 /* aggregate the system setting and interface setting */
3703                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3704
3705                 if (!_keep_addr)
3706                         _keep_addr = idev->cnf.keep_addr_on_down;
3707
3708                 keep_addr = (_keep_addr > 0);
3709         }
3710
3711         /* Step 2: clear hash table */
3712         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3713                 struct hlist_head *h = &inet6_addr_lst[i];
3714
3715                 spin_lock_bh(&addrconf_hash_lock);
3716 restart:
3717                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3718                         if (ifa->idev == idev) {
3719                                 addrconf_del_dad_work(ifa);
3720                                 /* combined flag + permanent flag decide if
3721                                  * address is retained on a down event
3722                                  */
3723                                 if (!keep_addr ||
3724                                     !(ifa->flags & IFA_F_PERMANENT) ||
3725                                     addr_is_local(&ifa->addr)) {
3726                                         hlist_del_init_rcu(&ifa->addr_lst);
3727                                         goto restart;
3728                                 }
3729                         }
3730                 }
3731                 spin_unlock_bh(&addrconf_hash_lock);
3732         }
3733
3734         write_lock_bh(&idev->lock);
3735
3736         addrconf_del_rs_timer(idev);
3737
3738         /* Step 2: clear flags for stateless addrconf */
3739         if (!how)
3740                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3741
3742         /* Step 3: clear tempaddr list */
3743         while (!list_empty(&idev->tempaddr_list)) {
3744                 ifa = list_first_entry(&idev->tempaddr_list,
3745                                        struct inet6_ifaddr, tmp_list);
3746                 list_del(&ifa->tmp_list);
3747                 write_unlock_bh(&idev->lock);
3748                 spin_lock_bh(&ifa->lock);
3749
3750                 if (ifa->ifpub) {
3751                         in6_ifa_put(ifa->ifpub);
3752                         ifa->ifpub = NULL;
3753                 }
3754                 spin_unlock_bh(&ifa->lock);
3755                 in6_ifa_put(ifa);
3756                 write_lock_bh(&idev->lock);
3757         }
3758
3759         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3760                 struct fib6_info *rt = NULL;
3761                 bool keep;
3762
3763                 addrconf_del_dad_work(ifa);
3764
3765                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3766                         !addr_is_local(&ifa->addr);
3767
3768                 write_unlock_bh(&idev->lock);
3769                 spin_lock_bh(&ifa->lock);
3770
3771                 if (keep) {
3772                         /* set state to skip the notifier below */
3773                         state = INET6_IFADDR_STATE_DEAD;
3774                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3775                         if (!(ifa->flags & IFA_F_NODAD))
3776                                 ifa->flags |= IFA_F_TENTATIVE;
3777
3778                         rt = ifa->rt;
3779                         ifa->rt = NULL;
3780                 } else {
3781                         state = ifa->state;
3782                         ifa->state = INET6_IFADDR_STATE_DEAD;
3783                 }
3784
3785                 spin_unlock_bh(&ifa->lock);
3786
3787                 if (rt)
3788                         ip6_del_rt(net, rt, false);
3789
3790                 if (state != INET6_IFADDR_STATE_DEAD) {
3791                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3792                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3793                 } else {
3794                         if (idev->cnf.forwarding)
3795                                 addrconf_leave_anycast(ifa);
3796                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3797                 }
3798
3799                 write_lock_bh(&idev->lock);
3800                 if (!keep) {
3801                         list_del_rcu(&ifa->if_list);
3802                         in6_ifa_put(ifa);
3803                 }
3804         }
3805
3806         write_unlock_bh(&idev->lock);
3807
3808         /* Step 5: Discard anycast and multicast list */
3809         if (how) {
3810                 ipv6_ac_destroy_dev(idev);
3811                 ipv6_mc_destroy_dev(idev);
3812         } else {
3813                 ipv6_mc_down(idev);
3814         }
3815
3816         idev->tstamp = jiffies;
3817
3818         /* Last: Shot the device (if unregistered) */
3819         if (how) {
3820                 addrconf_sysctl_unregister(idev);
3821                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3822                 neigh_ifdown(&nd_tbl, dev);
3823                 in6_dev_put(idev);
3824         }
3825         return 0;
3826 }
3827
3828 static void addrconf_rs_timer(struct timer_list *t)
3829 {
3830         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3831         struct net_device *dev = idev->dev;
3832         struct in6_addr lladdr;
3833
3834         write_lock(&idev->lock);
3835         if (idev->dead || !(idev->if_flags & IF_READY))
3836                 goto out;
3837
3838         if (!ipv6_accept_ra(idev))
3839                 goto out;
3840
3841         /* Announcement received after solicitation was sent */
3842         if (idev->if_flags & IF_RA_RCVD)
3843                 goto out;
3844
3845         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3846                 write_unlock(&idev->lock);
3847                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3848                         ndisc_send_rs(dev, &lladdr,
3849                                       &in6addr_linklocal_allrouters);
3850                 else
3851                         goto put;
3852
3853                 write_lock(&idev->lock);
3854                 idev->rs_interval = rfc3315_s14_backoff_update(
3855                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3856                 /* The wait after the last probe can be shorter */
3857                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3858                                              idev->cnf.rtr_solicits) ?
3859                                       idev->cnf.rtr_solicit_delay :
3860                                       idev->rs_interval);
3861         } else {
3862                 /*
3863                  * Note: we do not support deprecated "all on-link"
3864                  * assumption any longer.
3865                  */
3866                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3867         }
3868
3869 out:
3870         write_unlock(&idev->lock);
3871 put:
3872         in6_dev_put(idev);
3873 }
3874
3875 /*
3876  *      Duplicate Address Detection
3877  */
3878 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3879 {
3880         unsigned long rand_num;
3881         struct inet6_dev *idev = ifp->idev;
3882         u64 nonce;
3883
3884         if (ifp->flags & IFA_F_OPTIMISTIC)
3885                 rand_num = 0;
3886         else
3887                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3888
3889         nonce = 0;
3890         if (idev->cnf.enhanced_dad ||
3891             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3892                 do
3893                         get_random_bytes(&nonce, 6);
3894                 while (nonce == 0);
3895         }
3896         ifp->dad_nonce = nonce;
3897         ifp->dad_probes = idev->cnf.dad_transmits;
3898         addrconf_mod_dad_work(ifp, rand_num);
3899 }
3900
3901 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3902 {
3903         struct inet6_dev *idev = ifp->idev;
3904         struct net_device *dev = idev->dev;
3905         bool bump_id, notify = false;
3906         struct net *net;
3907
3908         addrconf_join_solict(dev, &ifp->addr);
3909
3910         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3911
3912         read_lock_bh(&idev->lock);
3913         spin_lock(&ifp->lock);
3914         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3915                 goto out;
3916
3917         net = dev_net(dev);
3918         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3919             (net->ipv6.devconf_all->accept_dad < 1 &&
3920              idev->cnf.accept_dad < 1) ||
3921             !(ifp->flags&IFA_F_TENTATIVE) ||
3922             ifp->flags & IFA_F_NODAD) {
3923                 bool send_na = false;
3924
3925                 if (ifp->flags & IFA_F_TENTATIVE &&
3926                     !(ifp->flags & IFA_F_OPTIMISTIC))
3927                         send_na = true;
3928                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3929                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3930                 spin_unlock(&ifp->lock);
3931                 read_unlock_bh(&idev->lock);
3932
3933                 addrconf_dad_completed(ifp, bump_id, send_na);
3934                 return;
3935         }
3936
3937         if (!(idev->if_flags & IF_READY)) {
3938                 spin_unlock(&ifp->lock);
3939                 read_unlock_bh(&idev->lock);
3940                 /*
3941                  * If the device is not ready:
3942                  * - keep it tentative if it is a permanent address.
3943                  * - otherwise, kill it.
3944                  */
3945                 in6_ifa_hold(ifp);
3946                 addrconf_dad_stop(ifp, 0);
3947                 return;
3948         }
3949
3950         /*
3951          * Optimistic nodes can start receiving
3952          * Frames right away
3953          */
3954         if (ifp->flags & IFA_F_OPTIMISTIC) {
3955                 ip6_ins_rt(net, ifp->rt);
3956                 if (ipv6_use_optimistic_addr(net, idev)) {
3957                         /* Because optimistic nodes can use this address,
3958                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3959                          */
3960                         notify = true;
3961                 }
3962         }
3963
3964         addrconf_dad_kick(ifp);
3965 out:
3966         spin_unlock(&ifp->lock);
3967         read_unlock_bh(&idev->lock);
3968         if (notify)
3969                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3970 }
3971
3972 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3973 {
3974         bool begin_dad = false;
3975
3976         spin_lock_bh(&ifp->lock);
3977         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3978                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3979                 begin_dad = true;
3980         }
3981         spin_unlock_bh(&ifp->lock);
3982
3983         if (begin_dad)
3984                 addrconf_mod_dad_work(ifp, 0);
3985 }
3986
3987 static void addrconf_dad_work(struct work_struct *w)
3988 {
3989         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3990                                                 struct inet6_ifaddr,
3991                                                 dad_work);
3992         struct inet6_dev *idev = ifp->idev;
3993         bool bump_id, disable_ipv6 = false;
3994         struct in6_addr mcaddr;
3995
3996         enum {
3997                 DAD_PROCESS,
3998                 DAD_BEGIN,
3999                 DAD_ABORT,
4000         } action = DAD_PROCESS;
4001
4002         rtnl_lock();
4003
4004         spin_lock_bh(&ifp->lock);
4005         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4006                 action = DAD_BEGIN;
4007                 ifp->state = INET6_IFADDR_STATE_DAD;
4008         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4009                 action = DAD_ABORT;
4010                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4011
4012                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4013                      idev->cnf.accept_dad > 1) &&
4014                     !idev->cnf.disable_ipv6 &&
4015                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4016                         struct in6_addr addr;
4017
4018                         addr.s6_addr32[0] = htonl(0xfe800000);
4019                         addr.s6_addr32[1] = 0;
4020
4021                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4022                             ipv6_addr_equal(&ifp->addr, &addr)) {
4023                                 /* DAD failed for link-local based on MAC */
4024                                 idev->cnf.disable_ipv6 = 1;
4025
4026                                 pr_info("%s: IPv6 being disabled!\n",
4027                                         ifp->idev->dev->name);
4028                                 disable_ipv6 = true;
4029                         }
4030                 }
4031         }
4032         spin_unlock_bh(&ifp->lock);
4033
4034         if (action == DAD_BEGIN) {
4035                 addrconf_dad_begin(ifp);
4036                 goto out;
4037         } else if (action == DAD_ABORT) {
4038                 in6_ifa_hold(ifp);
4039                 addrconf_dad_stop(ifp, 1);
4040                 if (disable_ipv6)
4041                         addrconf_ifdown(idev->dev, 0);
4042                 goto out;
4043         }
4044
4045         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4046                 goto out;
4047
4048         write_lock_bh(&idev->lock);
4049         if (idev->dead || !(idev->if_flags & IF_READY)) {
4050                 write_unlock_bh(&idev->lock);
4051                 goto out;
4052         }
4053
4054         spin_lock(&ifp->lock);
4055         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4056                 spin_unlock(&ifp->lock);
4057                 write_unlock_bh(&idev->lock);
4058                 goto out;
4059         }
4060
4061         if (ifp->dad_probes == 0) {
4062                 bool send_na = false;
4063
4064                 /*
4065                  * DAD was successful
4066                  */
4067
4068                 if (ifp->flags & IFA_F_TENTATIVE &&
4069                     !(ifp->flags & IFA_F_OPTIMISTIC))
4070                         send_na = true;
4071                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4072                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4073                 spin_unlock(&ifp->lock);
4074                 write_unlock_bh(&idev->lock);
4075
4076                 addrconf_dad_completed(ifp, bump_id, send_na);
4077
4078                 goto out;
4079         }
4080
4081         ifp->dad_probes--;
4082         addrconf_mod_dad_work(ifp,
4083                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4084                                   HZ/100));
4085         spin_unlock(&ifp->lock);
4086         write_unlock_bh(&idev->lock);
4087
4088         /* send a neighbour solicitation for our addr */
4089         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4090         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4091                       ifp->dad_nonce);
4092 out:
4093         in6_ifa_put(ifp);
4094         rtnl_unlock();
4095 }
4096
4097 /* ifp->idev must be at least read locked */
4098 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4099 {
4100         struct inet6_ifaddr *ifpiter;
4101         struct inet6_dev *idev = ifp->idev;
4102
4103         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4104                 if (ifpiter->scope > IFA_LINK)
4105                         break;
4106                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4107                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4108                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4109                     IFA_F_PERMANENT)
4110                         return false;
4111         }
4112         return true;
4113 }
4114
4115 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4116                                    bool send_na)
4117 {
4118         struct net_device *dev = ifp->idev->dev;
4119         struct in6_addr lladdr;
4120         bool send_rs, send_mld;
4121
4122         addrconf_del_dad_work(ifp);
4123
4124         /*
4125          *      Configure the address for reception. Now it is valid.
4126          */
4127
4128         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4129
4130         /* If added prefix is link local and we are prepared to process
4131            router advertisements, start sending router solicitations.
4132          */
4133
4134         read_lock_bh(&ifp->idev->lock);
4135         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4136         send_rs = send_mld &&
4137                   ipv6_accept_ra(ifp->idev) &&
4138                   ifp->idev->cnf.rtr_solicits != 0 &&
4139                   (dev->flags&IFF_LOOPBACK) == 0;
4140         read_unlock_bh(&ifp->idev->lock);
4141
4142         /* While dad is in progress mld report's source address is in6_addrany.
4143          * Resend with proper ll now.
4144          */
4145         if (send_mld)
4146                 ipv6_mc_dad_complete(ifp->idev);
4147
4148         /* send unsolicited NA if enabled */
4149         if (send_na &&
4150             (ifp->idev->cnf.ndisc_notify ||
4151              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4152                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4153                               /*router=*/ !!ifp->idev->cnf.forwarding,
4154                               /*solicited=*/ false, /*override=*/ true,
4155                               /*inc_opt=*/ true);
4156         }
4157
4158         if (send_rs) {
4159                 /*
4160                  *      If a host as already performed a random delay
4161                  *      [...] as part of DAD [...] there is no need
4162                  *      to delay again before sending the first RS
4163                  */
4164                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4165                         return;
4166                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4167
4168                 write_lock_bh(&ifp->idev->lock);
4169                 spin_lock(&ifp->lock);
4170                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4171                         ifp->idev->cnf.rtr_solicit_interval);
4172                 ifp->idev->rs_probes = 1;
4173                 ifp->idev->if_flags |= IF_RS_SENT;
4174                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4175                 spin_unlock(&ifp->lock);
4176                 write_unlock_bh(&ifp->idev->lock);
4177         }
4178
4179         if (bump_id)
4180                 rt_genid_bump_ipv6(dev_net(dev));
4181
4182         /* Make sure that a new temporary address will be created
4183          * before this temporary address becomes deprecated.
4184          */
4185         if (ifp->flags & IFA_F_TEMPORARY)
4186                 addrconf_verify_rtnl();
4187 }
4188
4189 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4190 {
4191         struct inet6_ifaddr *ifp;
4192
4193         read_lock_bh(&idev->lock);
4194         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4195                 spin_lock(&ifp->lock);
4196                 if ((ifp->flags & IFA_F_TENTATIVE &&
4197                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4198                         if (restart)
4199                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4200                         addrconf_dad_kick(ifp);
4201                 }
4202                 spin_unlock(&ifp->lock);
4203         }
4204         read_unlock_bh(&idev->lock);
4205 }
4206
4207 #ifdef CONFIG_PROC_FS
4208 struct if6_iter_state {
4209         struct seq_net_private p;
4210         int bucket;
4211         int offset;
4212 };
4213
4214 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4215 {
4216         struct if6_iter_state *state = seq->private;
4217         struct net *net = seq_file_net(seq);
4218         struct inet6_ifaddr *ifa = NULL;
4219         int p = 0;
4220
4221         /* initial bucket if pos is 0 */
4222         if (pos == 0) {
4223                 state->bucket = 0;
4224                 state->offset = 0;
4225         }
4226
4227         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4228                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4229                                          addr_lst) {
4230                         if (!net_eq(dev_net(ifa->idev->dev), net))
4231                                 continue;
4232                         /* sync with offset */
4233                         if (p < state->offset) {
4234                                 p++;
4235                                 continue;
4236                         }
4237                         return ifa;
4238                 }
4239
4240                 /* prepare for next bucket */
4241                 state->offset = 0;
4242                 p = 0;
4243         }
4244         return NULL;
4245 }
4246
4247 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4248                                          struct inet6_ifaddr *ifa)
4249 {
4250         struct if6_iter_state *state = seq->private;
4251         struct net *net = seq_file_net(seq);
4252
4253         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4254                 if (!net_eq(dev_net(ifa->idev->dev), net))
4255                         continue;
4256                 state->offset++;
4257                 return ifa;
4258         }
4259
4260         state->offset = 0;
4261         while (++state->bucket < IN6_ADDR_HSIZE) {
4262                 hlist_for_each_entry_rcu(ifa,
4263                                      &inet6_addr_lst[state->bucket], addr_lst) {
4264                         if (!net_eq(dev_net(ifa->idev->dev), net))
4265                                 continue;
4266                         return ifa;
4267                 }
4268         }
4269
4270         return NULL;
4271 }
4272
4273 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4274         __acquires(rcu)
4275 {
4276         rcu_read_lock();
4277         return if6_get_first(seq, *pos);
4278 }
4279
4280 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4281 {
4282         struct inet6_ifaddr *ifa;
4283
4284         ifa = if6_get_next(seq, v);
4285         ++*pos;
4286         return ifa;
4287 }
4288
4289 static void if6_seq_stop(struct seq_file *seq, void *v)
4290         __releases(rcu)
4291 {
4292         rcu_read_unlock();
4293 }
4294
4295 static int if6_seq_show(struct seq_file *seq, void *v)
4296 {
4297         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4298         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4299                    &ifp->addr,
4300                    ifp->idev->dev->ifindex,
4301                    ifp->prefix_len,
4302                    ifp->scope,
4303                    (u8) ifp->flags,
4304                    ifp->idev->dev->name);
4305         return 0;
4306 }
4307
4308 static const struct seq_operations if6_seq_ops = {
4309         .start  = if6_seq_start,
4310         .next   = if6_seq_next,
4311         .show   = if6_seq_show,
4312         .stop   = if6_seq_stop,
4313 };
4314
4315 static int __net_init if6_proc_net_init(struct net *net)
4316 {
4317         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4318                         sizeof(struct if6_iter_state)))
4319                 return -ENOMEM;
4320         return 0;
4321 }
4322
4323 static void __net_exit if6_proc_net_exit(struct net *net)
4324 {
4325         remove_proc_entry("if_inet6", net->proc_net);
4326 }
4327
4328 static struct pernet_operations if6_proc_net_ops = {
4329         .init = if6_proc_net_init,
4330         .exit = if6_proc_net_exit,
4331 };
4332
4333 int __init if6_proc_init(void)
4334 {
4335         return register_pernet_subsys(&if6_proc_net_ops);
4336 }
4337
4338 void if6_proc_exit(void)
4339 {
4340         unregister_pernet_subsys(&if6_proc_net_ops);
4341 }
4342 #endif  /* CONFIG_PROC_FS */
4343
4344 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4345 /* Check if address is a home address configured on any interface. */
4346 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4347 {
4348         unsigned int hash = inet6_addr_hash(net, addr);
4349         struct inet6_ifaddr *ifp = NULL;
4350         int ret = 0;
4351
4352         rcu_read_lock();
4353         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4354                 if (!net_eq(dev_net(ifp->idev->dev), net))
4355                         continue;
4356                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4357                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4358                         ret = 1;
4359                         break;
4360                 }
4361         }
4362         rcu_read_unlock();
4363         return ret;
4364 }
4365 #endif
4366
4367 /* RFC6554 has some algorithm to avoid loops in segment routing by
4368  * checking if the segments contains any of a local interface address.
4369  *
4370  * Quote:
4371  *
4372  * To detect loops in the SRH, a router MUST determine if the SRH
4373  * includes multiple addresses assigned to any interface on that router.
4374  * If such addresses appear more than once and are separated by at least
4375  * one address not assigned to that router.
4376  */
4377 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4378                           unsigned char nsegs)
4379 {
4380         const struct in6_addr *addr;
4381         int i, ret = 0, found = 0;
4382         struct inet6_ifaddr *ifp;
4383         bool separated = false;
4384         unsigned int hash;
4385         bool hash_found;
4386
4387         rcu_read_lock();
4388         for (i = 0; i < nsegs; i++) {
4389                 addr = &segs[i];
4390                 hash = inet6_addr_hash(net, addr);
4391
4392                 hash_found = false;
4393                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4394                         if (!net_eq(dev_net(ifp->idev->dev), net))
4395                                 continue;
4396
4397                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4398                                 hash_found = true;
4399                                 break;
4400                         }
4401                 }
4402
4403                 if (hash_found) {
4404                         if (found > 1 && separated) {
4405                                 ret = 1;
4406                                 break;
4407                         }
4408
4409                         separated = false;
4410                         found++;
4411                 } else {
4412                         separated = true;
4413                 }
4414         }
4415         rcu_read_unlock();
4416
4417         return ret;
4418 }
4419
4420 /*
4421  *      Periodic address status verification
4422  */
4423
4424 static void addrconf_verify_rtnl(void)
4425 {
4426         unsigned long now, next, next_sec, next_sched;
4427         struct inet6_ifaddr *ifp;
4428         int i;
4429
4430         ASSERT_RTNL();
4431
4432         rcu_read_lock_bh();
4433         now = jiffies;
4434         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4435
4436         cancel_delayed_work(&addr_chk_work);
4437
4438         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4439 restart:
4440                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4441                         unsigned long age;
4442
4443                         /* When setting preferred_lft to a value not zero or
4444                          * infinity, while valid_lft is infinity
4445                          * IFA_F_PERMANENT has a non-infinity life time.
4446                          */
4447                         if ((ifp->flags & IFA_F_PERMANENT) &&
4448                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4449                                 continue;
4450
4451                         spin_lock(&ifp->lock);
4452                         /* We try to batch several events at once. */
4453                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4454
4455                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4456                             age >= ifp->valid_lft) {
4457                                 spin_unlock(&ifp->lock);
4458                                 in6_ifa_hold(ifp);
4459                                 ipv6_del_addr(ifp);
4460                                 goto restart;
4461                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4462                                 spin_unlock(&ifp->lock);
4463                                 continue;
4464                         } else if (age >= ifp->prefered_lft) {
4465                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4466                                 int deprecate = 0;
4467
4468                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4469                                         deprecate = 1;
4470                                         ifp->flags |= IFA_F_DEPRECATED;
4471                                 }
4472
4473                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4474                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4475                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4476
4477                                 spin_unlock(&ifp->lock);
4478
4479                                 if (deprecate) {
4480                                         in6_ifa_hold(ifp);
4481
4482                                         ipv6_ifa_notify(0, ifp);
4483                                         in6_ifa_put(ifp);
4484                                         goto restart;
4485                                 }
4486                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4487                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4488                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4489                                         ifp->idev->cnf.dad_transmits *
4490                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4491
4492                                 if (age >= ifp->prefered_lft - regen_advance) {
4493                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4494                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4495                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4496                                         if (!ifp->regen_count && ifpub) {
4497                                                 ifp->regen_count++;
4498                                                 in6_ifa_hold(ifp);
4499                                                 in6_ifa_hold(ifpub);
4500                                                 spin_unlock(&ifp->lock);
4501
4502                                                 spin_lock(&ifpub->lock);
4503                                                 ifpub->regen_count = 0;
4504                                                 spin_unlock(&ifpub->lock);
4505                                                 rcu_read_unlock_bh();
4506                                                 ipv6_create_tempaddr(ifpub, true);
4507                                                 in6_ifa_put(ifpub);
4508                                                 in6_ifa_put(ifp);
4509                                                 rcu_read_lock_bh();
4510                                                 goto restart;
4511                                         }
4512                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4513                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4514                                 spin_unlock(&ifp->lock);
4515                         } else {
4516                                 /* ifp->prefered_lft <= ifp->valid_lft */
4517                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4518                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4519                                 spin_unlock(&ifp->lock);
4520                         }
4521                 }
4522         }
4523
4524         next_sec = round_jiffies_up(next);
4525         next_sched = next;
4526
4527         /* If rounded timeout is accurate enough, accept it. */
4528         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4529                 next_sched = next_sec;
4530
4531         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4532         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4533                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4534
4535         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4536                  now, next, next_sec, next_sched);
4537         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4538         rcu_read_unlock_bh();
4539 }
4540
4541 static void addrconf_verify_work(struct work_struct *w)
4542 {
4543         rtnl_lock();
4544         addrconf_verify_rtnl();
4545         rtnl_unlock();
4546 }
4547
4548 static void addrconf_verify(void)
4549 {
4550         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4551 }
4552
4553 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4554                                      struct in6_addr **peer_pfx)
4555 {
4556         struct in6_addr *pfx = NULL;
4557
4558         *peer_pfx = NULL;
4559
4560         if (addr)
4561                 pfx = nla_data(addr);
4562
4563         if (local) {
4564                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4565                         *peer_pfx = pfx;
4566                 pfx = nla_data(local);
4567         }
4568
4569         return pfx;
4570 }
4571
4572 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4573         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4574         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4575         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4576         [IFA_FLAGS]             = { .len = sizeof(u32) },
4577         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4578         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4579 };
4580
4581 static int
4582 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4583                   struct netlink_ext_ack *extack)
4584 {
4585         struct net *net = sock_net(skb->sk);
4586         struct ifaddrmsg *ifm;
4587         struct nlattr *tb[IFA_MAX+1];
4588         struct in6_addr *pfx, *peer_pfx;
4589         u32 ifa_flags;
4590         int err;
4591
4592         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4593                                      ifa_ipv6_policy, extack);
4594         if (err < 0)
4595                 return err;
4596
4597         ifm = nlmsg_data(nlh);
4598         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4599         if (!pfx)
4600                 return -EINVAL;
4601
4602         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4603
4604         /* We ignore other flags so far. */
4605         ifa_flags &= IFA_F_MANAGETEMPADDR;
4606
4607         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4608                               ifm->ifa_prefixlen);
4609 }
4610
4611 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4612                                unsigned long expires, u32 flags,
4613                                bool modify_peer)
4614 {
4615         struct fib6_info *f6i;
4616         u32 prio;
4617
4618         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4619                                         ifp->prefix_len,
4620                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4621         if (!f6i)
4622                 return -ENOENT;
4623
4624         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4625         if (f6i->fib6_metric != prio) {
4626                 /* delete old one */
4627                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4628
4629                 /* add new one */
4630                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4631                                       ifp->prefix_len,
4632                                       ifp->rt_priority, ifp->idev->dev,
4633                                       expires, flags, GFP_KERNEL);
4634         } else {
4635                 if (!expires)
4636                         fib6_clean_expires(f6i);
4637                 else
4638                         fib6_set_expires(f6i, expires);
4639
4640                 fib6_info_release(f6i);
4641         }
4642
4643         return 0;
4644 }
4645
4646 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4647 {
4648         u32 flags;
4649         clock_t expires;
4650         unsigned long timeout;
4651         bool was_managetempaddr;
4652         bool had_prefixroute;
4653         bool new_peer = false;
4654
4655         ASSERT_RTNL();
4656
4657         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4658                 return -EINVAL;
4659
4660         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4661             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4662                 return -EINVAL;
4663
4664         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4665                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4666
4667         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4668         if (addrconf_finite_timeout(timeout)) {
4669                 expires = jiffies_to_clock_t(timeout * HZ);
4670                 cfg->valid_lft = timeout;
4671                 flags = RTF_EXPIRES;
4672         } else {
4673                 expires = 0;
4674                 flags = 0;
4675                 cfg->ifa_flags |= IFA_F_PERMANENT;
4676         }
4677
4678         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4679         if (addrconf_finite_timeout(timeout)) {
4680                 if (timeout == 0)
4681                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4682                 cfg->preferred_lft = timeout;
4683         }
4684
4685         if (cfg->peer_pfx &&
4686             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4687                 if (!ipv6_addr_any(&ifp->peer_addr))
4688                         cleanup_prefix_route(ifp, expires, true, true);
4689                 new_peer = true;
4690         }
4691
4692         spin_lock_bh(&ifp->lock);
4693         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4694         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4695                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4696         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4697                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4698                         IFA_F_NOPREFIXROUTE);
4699         ifp->flags |= cfg->ifa_flags;
4700         ifp->tstamp = jiffies;
4701         ifp->valid_lft = cfg->valid_lft;
4702         ifp->prefered_lft = cfg->preferred_lft;
4703
4704         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4705                 ifp->rt_priority = cfg->rt_priority;
4706
4707         if (new_peer)
4708                 ifp->peer_addr = *cfg->peer_pfx;
4709
4710         spin_unlock_bh(&ifp->lock);
4711         if (!(ifp->flags&IFA_F_TENTATIVE))
4712                 ipv6_ifa_notify(0, ifp);
4713
4714         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4715                 int rc = -ENOENT;
4716
4717                 if (had_prefixroute)
4718                         rc = modify_prefix_route(ifp, expires, flags, false);
4719
4720                 /* prefix route could have been deleted; if so restore it */
4721                 if (rc == -ENOENT) {
4722                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4723                                               ifp->rt_priority, ifp->idev->dev,
4724                                               expires, flags, GFP_KERNEL);
4725                 }
4726
4727                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4728                         rc = modify_prefix_route(ifp, expires, flags, true);
4729
4730                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4731                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4732                                               ifp->rt_priority, ifp->idev->dev,
4733                                               expires, flags, GFP_KERNEL);
4734                 }
4735         } else if (had_prefixroute) {
4736                 enum cleanup_prefix_rt_t action;
4737                 unsigned long rt_expires;
4738
4739                 write_lock_bh(&ifp->idev->lock);
4740                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4741                 write_unlock_bh(&ifp->idev->lock);
4742
4743                 if (action != CLEANUP_PREFIX_RT_NOP) {
4744                         cleanup_prefix_route(ifp, rt_expires,
4745                                 action == CLEANUP_PREFIX_RT_DEL, false);
4746                 }
4747         }
4748
4749         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4750                 if (was_managetempaddr &&
4751                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4752                         cfg->valid_lft = 0;
4753                         cfg->preferred_lft = 0;
4754                 }
4755                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4756                                  cfg->preferred_lft, !was_managetempaddr,
4757                                  jiffies);
4758         }
4759
4760         addrconf_verify_rtnl();
4761
4762         return 0;
4763 }
4764
4765 static int
4766 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4767                   struct netlink_ext_ack *extack)
4768 {
4769         struct net *net = sock_net(skb->sk);
4770         struct ifaddrmsg *ifm;
4771         struct nlattr *tb[IFA_MAX+1];
4772         struct in6_addr *peer_pfx;
4773         struct inet6_ifaddr *ifa;
4774         struct net_device *dev;
4775         struct inet6_dev *idev;
4776         struct ifa6_config cfg;
4777         int err;
4778
4779         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4780                                      ifa_ipv6_policy, extack);
4781         if (err < 0)
4782                 return err;
4783
4784         memset(&cfg, 0, sizeof(cfg));
4785
4786         ifm = nlmsg_data(nlh);
4787         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4788         if (!cfg.pfx)
4789                 return -EINVAL;
4790
4791         cfg.peer_pfx = peer_pfx;
4792         cfg.plen = ifm->ifa_prefixlen;
4793         if (tb[IFA_RT_PRIORITY])
4794                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4795
4796         cfg.valid_lft = INFINITY_LIFE_TIME;
4797         cfg.preferred_lft = INFINITY_LIFE_TIME;
4798
4799         if (tb[IFA_CACHEINFO]) {
4800                 struct ifa_cacheinfo *ci;
4801
4802                 ci = nla_data(tb[IFA_CACHEINFO]);
4803                 cfg.valid_lft = ci->ifa_valid;
4804                 cfg.preferred_lft = ci->ifa_prefered;
4805         }
4806
4807         dev =  __dev_get_by_index(net, ifm->ifa_index);
4808         if (!dev)
4809                 return -ENODEV;
4810
4811         if (tb[IFA_FLAGS])
4812                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4813         else
4814                 cfg.ifa_flags = ifm->ifa_flags;
4815
4816         /* We ignore other flags so far. */
4817         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4818                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4819                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4820
4821         idev = ipv6_find_idev(dev);
4822         if (IS_ERR(idev))
4823                 return PTR_ERR(idev);
4824
4825         if (!ipv6_allow_optimistic_dad(net, idev))
4826                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4827
4828         if (cfg.ifa_flags & IFA_F_NODAD &&
4829             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4830                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4831                 return -EINVAL;
4832         }
4833
4834         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4835         if (!ifa) {
4836                 /*
4837                  * It would be best to check for !NLM_F_CREATE here but
4838                  * userspace already relies on not having to provide this.
4839                  */
4840                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4841         }
4842
4843         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4844             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4845                 err = -EEXIST;
4846         else
4847                 err = inet6_addr_modify(ifa, &cfg);
4848
4849         in6_ifa_put(ifa);
4850
4851         return err;
4852 }
4853
4854 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4855                           u8 scope, int ifindex)
4856 {
4857         struct ifaddrmsg *ifm;
4858
4859         ifm = nlmsg_data(nlh);
4860         ifm->ifa_family = AF_INET6;
4861         ifm->ifa_prefixlen = prefixlen;
4862         ifm->ifa_flags = flags;
4863         ifm->ifa_scope = scope;
4864         ifm->ifa_index = ifindex;
4865 }
4866
4867 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4868                          unsigned long tstamp, u32 preferred, u32 valid)
4869 {
4870         struct ifa_cacheinfo ci;
4871
4872         ci.cstamp = cstamp_delta(cstamp);
4873         ci.tstamp = cstamp_delta(tstamp);
4874         ci.ifa_prefered = preferred;
4875         ci.ifa_valid = valid;
4876
4877         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4878 }
4879
4880 static inline int rt_scope(int ifa_scope)
4881 {
4882         if (ifa_scope & IFA_HOST)
4883                 return RT_SCOPE_HOST;
4884         else if (ifa_scope & IFA_LINK)
4885                 return RT_SCOPE_LINK;
4886         else if (ifa_scope & IFA_SITE)
4887                 return RT_SCOPE_SITE;
4888         else
4889                 return RT_SCOPE_UNIVERSE;
4890 }
4891
4892 static inline int inet6_ifaddr_msgsize(void)
4893 {
4894         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4895                + nla_total_size(16) /* IFA_LOCAL */
4896                + nla_total_size(16) /* IFA_ADDRESS */
4897                + nla_total_size(sizeof(struct ifa_cacheinfo))
4898                + nla_total_size(4)  /* IFA_FLAGS */
4899                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4900 }
4901
4902 enum addr_type_t {
4903         UNICAST_ADDR,
4904         MULTICAST_ADDR,
4905         ANYCAST_ADDR,
4906 };
4907
4908 struct inet6_fill_args {
4909         u32 portid;
4910         u32 seq;
4911         int event;
4912         unsigned int flags;
4913         int netnsid;
4914         int ifindex;
4915         enum addr_type_t type;
4916 };
4917
4918 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4919                              struct inet6_fill_args *args)
4920 {
4921         struct nlmsghdr  *nlh;
4922         u32 preferred, valid;
4923
4924         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4925                         sizeof(struct ifaddrmsg), args->flags);
4926         if (!nlh)
4927                 return -EMSGSIZE;
4928
4929         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4930                       ifa->idev->dev->ifindex);
4931
4932         if (args->netnsid >= 0 &&
4933             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4934                 goto error;
4935
4936         if (!((ifa->flags&IFA_F_PERMANENT) &&
4937               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4938                 preferred = ifa->prefered_lft;
4939                 valid = ifa->valid_lft;
4940                 if (preferred != INFINITY_LIFE_TIME) {
4941                         long tval = (jiffies - ifa->tstamp)/HZ;
4942                         if (preferred > tval)
4943                                 preferred -= tval;
4944                         else
4945                                 preferred = 0;
4946                         if (valid != INFINITY_LIFE_TIME) {
4947                                 if (valid > tval)
4948                                         valid -= tval;
4949                                 else
4950                                         valid = 0;
4951                         }
4952                 }
4953         } else {
4954                 preferred = INFINITY_LIFE_TIME;
4955                 valid = INFINITY_LIFE_TIME;
4956         }
4957
4958         if (!ipv6_addr_any(&ifa->peer_addr)) {
4959                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4960                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4961                         goto error;
4962         } else
4963                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4964                         goto error;
4965
4966         if (ifa->rt_priority &&
4967             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4968                 goto error;
4969
4970         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4971                 goto error;
4972
4973         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4974                 goto error;
4975
4976         nlmsg_end(skb, nlh);
4977         return 0;
4978
4979 error:
4980         nlmsg_cancel(skb, nlh);
4981         return -EMSGSIZE;
4982 }
4983
4984 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4985                                struct inet6_fill_args *args)
4986 {
4987         struct nlmsghdr  *nlh;
4988         u8 scope = RT_SCOPE_UNIVERSE;
4989         int ifindex = ifmca->idev->dev->ifindex;
4990
4991         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4992                 scope = RT_SCOPE_SITE;
4993
4994         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4995                         sizeof(struct ifaddrmsg), args->flags);
4996         if (!nlh)
4997                 return -EMSGSIZE;
4998
4999         if (args->netnsid >= 0 &&
5000             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5001                 return -EMSGSIZE;
5002
5003         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5004         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5005             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5006                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5007                 nlmsg_cancel(skb, nlh);
5008                 return -EMSGSIZE;
5009         }
5010
5011         nlmsg_end(skb, nlh);
5012         return 0;
5013 }
5014
5015 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5016                                struct inet6_fill_args *args)
5017 {
5018         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5019         int ifindex = dev ? dev->ifindex : 1;
5020         struct nlmsghdr  *nlh;
5021         u8 scope = RT_SCOPE_UNIVERSE;
5022
5023         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5024                 scope = RT_SCOPE_SITE;
5025
5026         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5027                         sizeof(struct ifaddrmsg), args->flags);
5028         if (!nlh)
5029                 return -EMSGSIZE;
5030
5031         if (args->netnsid >= 0 &&
5032             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5033                 return -EMSGSIZE;
5034
5035         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5036         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5037             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5038                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5039                 nlmsg_cancel(skb, nlh);
5040                 return -EMSGSIZE;
5041         }
5042
5043         nlmsg_end(skb, nlh);
5044         return 0;
5045 }
5046
5047 /* called with rcu_read_lock() */
5048 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5049                           struct netlink_callback *cb, int s_ip_idx,
5050                           struct inet6_fill_args *fillargs)
5051 {
5052         struct ifmcaddr6 *ifmca;
5053         struct ifacaddr6 *ifaca;
5054         int ip_idx = 0;
5055         int err = 1;
5056
5057         read_lock_bh(&idev->lock);
5058         switch (fillargs->type) {
5059         case UNICAST_ADDR: {
5060                 struct inet6_ifaddr *ifa;
5061                 fillargs->event = RTM_NEWADDR;
5062
5063                 /* unicast address incl. temp addr */
5064                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5065                         if (ip_idx < s_ip_idx)
5066                                 goto next;
5067                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5068                         if (err < 0)
5069                                 break;
5070                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5071 next:
5072                         ip_idx++;
5073                 }
5074                 break;
5075         }
5076         case MULTICAST_ADDR:
5077                 fillargs->event = RTM_GETMULTICAST;
5078
5079                 /* multicast address */
5080                 for (ifmca = idev->mc_list; ifmca;
5081                      ifmca = ifmca->next, ip_idx++) {
5082                         if (ip_idx < s_ip_idx)
5083                                 continue;
5084                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5085                         if (err < 0)
5086                                 break;
5087                 }
5088                 break;
5089         case ANYCAST_ADDR:
5090                 fillargs->event = RTM_GETANYCAST;
5091                 /* anycast address */
5092                 for (ifaca = idev->ac_list; ifaca;
5093                      ifaca = ifaca->aca_next, ip_idx++) {
5094                         if (ip_idx < s_ip_idx)
5095                                 continue;
5096                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5097                         if (err < 0)
5098                                 break;
5099                 }
5100                 break;
5101         default:
5102                 break;
5103         }
5104         read_unlock_bh(&idev->lock);
5105         cb->args[2] = ip_idx;
5106         return err;
5107 }
5108
5109 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5110                                        struct inet6_fill_args *fillargs,
5111                                        struct net **tgt_net, struct sock *sk,
5112                                        struct netlink_callback *cb)
5113 {
5114         struct netlink_ext_ack *extack = cb->extack;
5115         struct nlattr *tb[IFA_MAX+1];
5116         struct ifaddrmsg *ifm;
5117         int err, i;
5118
5119         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5120                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5121                 return -EINVAL;
5122         }
5123
5124         ifm = nlmsg_data(nlh);
5125         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5126                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5127                 return -EINVAL;
5128         }
5129
5130         fillargs->ifindex = ifm->ifa_index;
5131         if (fillargs->ifindex) {
5132                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5133                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5134         }
5135
5136         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5137                                             ifa_ipv6_policy, extack);
5138         if (err < 0)
5139                 return err;
5140
5141         for (i = 0; i <= IFA_MAX; ++i) {
5142                 if (!tb[i])
5143                         continue;
5144
5145                 if (i == IFA_TARGET_NETNSID) {
5146                         struct net *net;
5147
5148                         fillargs->netnsid = nla_get_s32(tb[i]);
5149                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5150                         if (IS_ERR(net)) {
5151                                 fillargs->netnsid = -1;
5152                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5153                                 return PTR_ERR(net);
5154                         }
5155                         *tgt_net = net;
5156                 } else {
5157                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5158                         return -EINVAL;
5159                 }
5160         }
5161
5162         return 0;
5163 }
5164
5165 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5166                            enum addr_type_t type)
5167 {
5168         const struct nlmsghdr *nlh = cb->nlh;
5169         struct inet6_fill_args fillargs = {
5170                 .portid = NETLINK_CB(cb->skb).portid,
5171                 .seq = cb->nlh->nlmsg_seq,
5172                 .flags = NLM_F_MULTI,
5173                 .netnsid = -1,
5174                 .type = type,
5175         };
5176         struct net *net = sock_net(skb->sk);
5177         struct net *tgt_net = net;
5178         int idx, s_idx, s_ip_idx;
5179         int h, s_h;
5180         struct net_device *dev;
5181         struct inet6_dev *idev;
5182         struct hlist_head *head;
5183         int err = 0;
5184
5185         s_h = cb->args[0];
5186         s_idx = idx = cb->args[1];
5187         s_ip_idx = cb->args[2];
5188
5189         if (cb->strict_check) {
5190                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5191                                                   skb->sk, cb);
5192                 if (err < 0)
5193                         goto put_tgt_net;
5194
5195                 err = 0;
5196                 if (fillargs.ifindex) {
5197                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5198                         if (!dev) {
5199                                 err = -ENODEV;
5200                                 goto put_tgt_net;
5201                         }
5202                         idev = __in6_dev_get(dev);
5203                         if (idev) {
5204                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5205                                                      &fillargs);
5206                                 if (err > 0)
5207                                         err = 0;
5208                         }
5209                         goto put_tgt_net;
5210                 }
5211         }
5212
5213         rcu_read_lock();
5214         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5215         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5216                 idx = 0;
5217                 head = &tgt_net->dev_index_head[h];
5218                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5219                         if (idx < s_idx)
5220                                 goto cont;
5221                         if (h > s_h || idx > s_idx)
5222                                 s_ip_idx = 0;
5223                         idev = __in6_dev_get(dev);
5224                         if (!idev)
5225                                 goto cont;
5226
5227                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5228                                            &fillargs) < 0)
5229                                 goto done;
5230 cont:
5231                         idx++;
5232                 }
5233         }
5234 done:
5235         rcu_read_unlock();
5236         cb->args[0] = h;
5237         cb->args[1] = idx;
5238 put_tgt_net:
5239         if (fillargs.netnsid >= 0)
5240                 put_net(tgt_net);
5241
5242         return skb->len ? : err;
5243 }
5244
5245 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5246 {
5247         enum addr_type_t type = UNICAST_ADDR;
5248
5249         return inet6_dump_addr(skb, cb, type);
5250 }
5251
5252 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5253 {
5254         enum addr_type_t type = MULTICAST_ADDR;
5255
5256         return inet6_dump_addr(skb, cb, type);
5257 }
5258
5259
5260 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5261 {
5262         enum addr_type_t type = ANYCAST_ADDR;
5263
5264         return inet6_dump_addr(skb, cb, type);
5265 }
5266
5267 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5268                                        const struct nlmsghdr *nlh,
5269                                        struct nlattr **tb,
5270                                        struct netlink_ext_ack *extack)
5271 {
5272         struct ifaddrmsg *ifm;
5273         int i, err;
5274
5275         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5276                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5277                 return -EINVAL;
5278         }
5279
5280         if (!netlink_strict_get_check(skb))
5281                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5282                                               ifa_ipv6_policy, extack);
5283
5284         ifm = nlmsg_data(nlh);
5285         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5286                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5287                 return -EINVAL;
5288         }
5289
5290         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5291                                             ifa_ipv6_policy, extack);
5292         if (err)
5293                 return err;
5294
5295         for (i = 0; i <= IFA_MAX; i++) {
5296                 if (!tb[i])
5297                         continue;
5298
5299                 switch (i) {
5300                 case IFA_TARGET_NETNSID:
5301                 case IFA_ADDRESS:
5302                 case IFA_LOCAL:
5303                         break;
5304                 default:
5305                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5306                         return -EINVAL;
5307                 }
5308         }
5309
5310         return 0;
5311 }
5312
5313 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5314                              struct netlink_ext_ack *extack)
5315 {
5316         struct net *net = sock_net(in_skb->sk);
5317         struct inet6_fill_args fillargs = {
5318                 .portid = NETLINK_CB(in_skb).portid,
5319                 .seq = nlh->nlmsg_seq,
5320                 .event = RTM_NEWADDR,
5321                 .flags = 0,
5322                 .netnsid = -1,
5323         };
5324         struct net *tgt_net = net;
5325         struct ifaddrmsg *ifm;
5326         struct nlattr *tb[IFA_MAX+1];
5327         struct in6_addr *addr = NULL, *peer;
5328         struct net_device *dev = NULL;
5329         struct inet6_ifaddr *ifa;
5330         struct sk_buff *skb;
5331         int err;
5332
5333         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5334         if (err < 0)
5335                 return err;
5336
5337         if (tb[IFA_TARGET_NETNSID]) {
5338                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5339
5340                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5341                                                   fillargs.netnsid);
5342                 if (IS_ERR(tgt_net))
5343                         return PTR_ERR(tgt_net);
5344         }
5345
5346         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5347         if (!addr)
5348                 return -EINVAL;
5349
5350         ifm = nlmsg_data(nlh);
5351         if (ifm->ifa_index)
5352                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5353
5354         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5355         if (!ifa) {
5356                 err = -EADDRNOTAVAIL;
5357                 goto errout;
5358         }
5359
5360         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5361         if (!skb) {
5362                 err = -ENOBUFS;
5363                 goto errout_ifa;
5364         }
5365
5366         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5367         if (err < 0) {
5368                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5369                 WARN_ON(err == -EMSGSIZE);
5370                 kfree_skb(skb);
5371                 goto errout_ifa;
5372         }
5373         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5374 errout_ifa:
5375         in6_ifa_put(ifa);
5376 errout:
5377         if (dev)
5378                 dev_put(dev);
5379         if (fillargs.netnsid >= 0)
5380                 put_net(tgt_net);
5381
5382         return err;
5383 }
5384
5385 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5386 {
5387         struct sk_buff *skb;
5388         struct net *net = dev_net(ifa->idev->dev);
5389         struct inet6_fill_args fillargs = {
5390                 .portid = 0,
5391                 .seq = 0,
5392                 .event = event,
5393                 .flags = 0,
5394                 .netnsid = -1,
5395         };
5396         int err = -ENOBUFS;
5397
5398         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5399         if (!skb)
5400                 goto errout;
5401
5402         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5403         if (err < 0) {
5404                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5405                 WARN_ON(err == -EMSGSIZE);
5406                 kfree_skb(skb);
5407                 goto errout;
5408         }
5409         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5410         return;
5411 errout:
5412         if (err < 0)
5413                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5414 }
5415
5416 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5417                                 __s32 *array, int bytes)
5418 {
5419         BUG_ON(bytes < (DEVCONF_MAX * 4));
5420
5421         memset(array, 0, bytes);
5422         array[DEVCONF_FORWARDING] = cnf->forwarding;
5423         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5424         array[DEVCONF_MTU6] = cnf->mtu6;
5425         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5426         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5427         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5428         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5429         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5430         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5431                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5432         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5433                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5434         array[DEVCONF_RTR_SOLICIT_DELAY] =
5435                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5436         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5437         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5438                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5439         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5440                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5441         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5442         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5443         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5444         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5445         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5446         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5447         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5448         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5449         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5450 #ifdef CONFIG_IPV6_ROUTER_PREF
5451         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5452         array[DEVCONF_RTR_PROBE_INTERVAL] =
5453                 jiffies_to_msecs(cnf->rtr_probe_interval);
5454 #ifdef CONFIG_IPV6_ROUTE_INFO
5455         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5456         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5457 #endif
5458 #endif
5459         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5460         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5461 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5462         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5463         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5464 #endif
5465 #ifdef CONFIG_IPV6_MROUTE
5466         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5467 #endif
5468         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5469         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5470         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5471         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5472         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5473         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5474         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5475         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5476         /* we omit DEVCONF_STABLE_SECRET for now */
5477         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5478         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5479         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5480         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5481         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5482 #ifdef CONFIG_IPV6_SEG6_HMAC
5483         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5484 #endif
5485         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5486         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5487         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5488         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5489         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5490 }
5491
5492 static inline size_t inet6_ifla6_size(void)
5493 {
5494         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5495              + nla_total_size(sizeof(struct ifla_cacheinfo))
5496              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5497              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5498              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5499              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5500              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5501              + 0;
5502 }
5503
5504 static inline size_t inet6_if_nlmsg_size(void)
5505 {
5506         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5507                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5508                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5509                + nla_total_size(4) /* IFLA_MTU */
5510                + nla_total_size(4) /* IFLA_LINK */
5511                + nla_total_size(1) /* IFLA_OPERSTATE */
5512                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5513 }
5514
5515 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5516                                         int bytes)
5517 {
5518         int i;
5519         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5520         BUG_ON(pad < 0);
5521
5522         /* Use put_unaligned() because stats may not be aligned for u64. */
5523         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5524         for (i = 1; i < ICMP6_MIB_MAX; i++)
5525                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5526
5527         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5528 }
5529
5530 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5531                                         int bytes, size_t syncpoff)
5532 {
5533         int i, c;
5534         u64 buff[IPSTATS_MIB_MAX];
5535         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5536
5537         BUG_ON(pad < 0);
5538
5539         memset(buff, 0, sizeof(buff));
5540         buff[0] = IPSTATS_MIB_MAX;
5541
5542         for_each_possible_cpu(c) {
5543                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5544                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5545         }
5546
5547         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5548         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5549 }
5550
5551 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5552                              int bytes)
5553 {
5554         switch (attrtype) {
5555         case IFLA_INET6_STATS:
5556                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5557                                      offsetof(struct ipstats_mib, syncp));
5558                 break;
5559         case IFLA_INET6_ICMP6STATS:
5560                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5561                 break;
5562         }
5563 }
5564
5565 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5566                                   u32 ext_filter_mask)
5567 {
5568         struct nlattr *nla;
5569         struct ifla_cacheinfo ci;
5570
5571         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5572                 goto nla_put_failure;
5573         ci.max_reasm_len = IPV6_MAXPLEN;
5574         ci.tstamp = cstamp_delta(idev->tstamp);
5575         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5576         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5577         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5578                 goto nla_put_failure;
5579         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5580         if (!nla)
5581                 goto nla_put_failure;
5582         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5583
5584         /* XXX - MC not implemented */
5585
5586         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5587                 return 0;
5588
5589         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5590         if (!nla)
5591                 goto nla_put_failure;
5592         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5593
5594         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5595         if (!nla)
5596                 goto nla_put_failure;
5597         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5598
5599         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5600         if (!nla)
5601                 goto nla_put_failure;
5602         read_lock_bh(&idev->lock);
5603         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5604         read_unlock_bh(&idev->lock);
5605
5606         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5607                 goto nla_put_failure;
5608
5609         return 0;
5610
5611 nla_put_failure:
5612         return -EMSGSIZE;
5613 }
5614
5615 static size_t inet6_get_link_af_size(const struct net_device *dev,
5616                                      u32 ext_filter_mask)
5617 {
5618         if (!__in6_dev_get(dev))
5619                 return 0;
5620
5621         return inet6_ifla6_size();
5622 }
5623
5624 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5625                               u32 ext_filter_mask)
5626 {
5627         struct inet6_dev *idev = __in6_dev_get(dev);
5628
5629         if (!idev)
5630                 return -ENODATA;
5631
5632         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5633                 return -EMSGSIZE;
5634
5635         return 0;
5636 }
5637
5638 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5639 {
5640         struct inet6_ifaddr *ifp;
5641         struct net_device *dev = idev->dev;
5642         bool clear_token, update_rs = false;
5643         struct in6_addr ll_addr;
5644
5645         ASSERT_RTNL();
5646
5647         if (!token)
5648                 return -EINVAL;
5649         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5650                 return -EINVAL;
5651         if (!ipv6_accept_ra(idev))
5652                 return -EINVAL;
5653         if (idev->cnf.rtr_solicits == 0)
5654                 return -EINVAL;
5655
5656         write_lock_bh(&idev->lock);
5657
5658         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5659         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5660
5661         write_unlock_bh(&idev->lock);
5662
5663         clear_token = ipv6_addr_any(token);
5664         if (clear_token)
5665                 goto update_lft;
5666
5667         if (!idev->dead && (idev->if_flags & IF_READY) &&
5668             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5669                              IFA_F_OPTIMISTIC)) {
5670                 /* If we're not ready, then normal ifup will take care
5671                  * of this. Otherwise, we need to request our rs here.
5672                  */
5673                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5674                 update_rs = true;
5675         }
5676
5677 update_lft:
5678         write_lock_bh(&idev->lock);
5679
5680         if (update_rs) {
5681                 idev->if_flags |= IF_RS_SENT;
5682                 idev->rs_interval = rfc3315_s14_backoff_init(
5683                         idev->cnf.rtr_solicit_interval);
5684                 idev->rs_probes = 1;
5685                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5686         }
5687
5688         /* Well, that's kinda nasty ... */
5689         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5690                 spin_lock(&ifp->lock);
5691                 if (ifp->tokenized) {
5692                         ifp->valid_lft = 0;
5693                         ifp->prefered_lft = 0;
5694                 }
5695                 spin_unlock(&ifp->lock);
5696         }
5697
5698         write_unlock_bh(&idev->lock);
5699         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5700         addrconf_verify_rtnl();
5701         return 0;
5702 }
5703
5704 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5705         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5706         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5707 };
5708
5709 static int check_addr_gen_mode(int mode)
5710 {
5711         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5712             mode != IN6_ADDR_GEN_MODE_NONE &&
5713             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5714             mode != IN6_ADDR_GEN_MODE_RANDOM)
5715                 return -EINVAL;
5716         return 1;
5717 }
5718
5719 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5720                                 int mode)
5721 {
5722         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5723             !idev->cnf.stable_secret.initialized &&
5724             !net->ipv6.devconf_dflt->stable_secret.initialized)
5725                 return -EINVAL;
5726         return 1;
5727 }
5728
5729 static int inet6_validate_link_af(const struct net_device *dev,
5730                                   const struct nlattr *nla)
5731 {
5732         struct nlattr *tb[IFLA_INET6_MAX + 1];
5733         struct inet6_dev *idev = NULL;
5734         int err;
5735
5736         if (dev) {
5737                 idev = __in6_dev_get(dev);
5738                 if (!idev)
5739                         return -EAFNOSUPPORT;
5740         }
5741
5742         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5743                                           inet6_af_policy, NULL);
5744         if (err)
5745                 return err;
5746
5747         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5748                 return -EINVAL;
5749
5750         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5751                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5752
5753                 if (check_addr_gen_mode(mode) < 0)
5754                         return -EINVAL;
5755                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5756                         return -EINVAL;
5757         }
5758
5759         return 0;
5760 }
5761
5762 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5763 {
5764         struct inet6_dev *idev = __in6_dev_get(dev);
5765         struct nlattr *tb[IFLA_INET6_MAX + 1];
5766         int err;
5767
5768         if (!idev)
5769                 return -EAFNOSUPPORT;
5770
5771         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5772                 BUG();
5773
5774         if (tb[IFLA_INET6_TOKEN]) {
5775                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5776                 if (err)
5777                         return err;
5778         }
5779
5780         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5781                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5782
5783                 idev->cnf.addr_gen_mode = mode;
5784         }
5785
5786         return 0;
5787 }
5788
5789 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5790                              u32 portid, u32 seq, int event, unsigned int flags)
5791 {
5792         struct net_device *dev = idev->dev;
5793         struct ifinfomsg *hdr;
5794         struct nlmsghdr *nlh;
5795         void *protoinfo;
5796
5797         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5798         if (!nlh)
5799                 return -EMSGSIZE;
5800
5801         hdr = nlmsg_data(nlh);
5802         hdr->ifi_family = AF_INET6;
5803         hdr->__ifi_pad = 0;
5804         hdr->ifi_type = dev->type;
5805         hdr->ifi_index = dev->ifindex;
5806         hdr->ifi_flags = dev_get_flags(dev);
5807         hdr->ifi_change = 0;
5808
5809         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5810             (dev->addr_len &&
5811              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5812             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5813             (dev->ifindex != dev_get_iflink(dev) &&
5814              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5815             nla_put_u8(skb, IFLA_OPERSTATE,
5816                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5817                 goto nla_put_failure;
5818         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5819         if (!protoinfo)
5820                 goto nla_put_failure;
5821
5822         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5823                 goto nla_put_failure;
5824
5825         nla_nest_end(skb, protoinfo);
5826         nlmsg_end(skb, nlh);
5827         return 0;
5828
5829 nla_put_failure:
5830         nlmsg_cancel(skb, nlh);
5831         return -EMSGSIZE;
5832 }
5833
5834 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5835                                    struct netlink_ext_ack *extack)
5836 {
5837         struct ifinfomsg *ifm;
5838
5839         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5840                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5841                 return -EINVAL;
5842         }
5843
5844         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5845                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5846                 return -EINVAL;
5847         }
5848
5849         ifm = nlmsg_data(nlh);
5850         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5851             ifm->ifi_change || ifm->ifi_index) {
5852                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5853                 return -EINVAL;
5854         }
5855
5856         return 0;
5857 }
5858
5859 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5860 {
5861         struct net *net = sock_net(skb->sk);
5862         int h, s_h;
5863         int idx = 0, s_idx;
5864         struct net_device *dev;
5865         struct inet6_dev *idev;
5866         struct hlist_head *head;
5867
5868         /* only requests using strict checking can pass data to
5869          * influence the dump
5870          */
5871         if (cb->strict_check) {
5872                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5873
5874                 if (err < 0)
5875                         return err;
5876         }
5877
5878         s_h = cb->args[0];
5879         s_idx = cb->args[1];
5880
5881         rcu_read_lock();
5882         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5883                 idx = 0;
5884                 head = &net->dev_index_head[h];
5885                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5886                         if (idx < s_idx)
5887                                 goto cont;
5888                         idev = __in6_dev_get(dev);
5889                         if (!idev)
5890                                 goto cont;
5891                         if (inet6_fill_ifinfo(skb, idev,
5892                                               NETLINK_CB(cb->skb).portid,
5893                                               cb->nlh->nlmsg_seq,
5894                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5895                                 goto out;
5896 cont:
5897                         idx++;
5898                 }
5899         }
5900 out:
5901         rcu_read_unlock();
5902         cb->args[1] = idx;
5903         cb->args[0] = h;
5904
5905         return skb->len;
5906 }
5907
5908 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5909 {
5910         struct sk_buff *skb;
5911         struct net *net = dev_net(idev->dev);
5912         int err = -ENOBUFS;
5913
5914         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5915         if (!skb)
5916                 goto errout;
5917
5918         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5919         if (err < 0) {
5920                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5921                 WARN_ON(err == -EMSGSIZE);
5922                 kfree_skb(skb);
5923                 goto errout;
5924         }
5925         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5926         return;
5927 errout:
5928         if (err < 0)
5929                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5930 }
5931
5932 static inline size_t inet6_prefix_nlmsg_size(void)
5933 {
5934         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5935                + nla_total_size(sizeof(struct in6_addr))
5936                + nla_total_size(sizeof(struct prefix_cacheinfo));
5937 }
5938
5939 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5940                              struct prefix_info *pinfo, u32 portid, u32 seq,
5941                              int event, unsigned int flags)
5942 {
5943         struct prefixmsg *pmsg;
5944         struct nlmsghdr *nlh;
5945         struct prefix_cacheinfo ci;
5946
5947         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5948         if (!nlh)
5949                 return -EMSGSIZE;
5950
5951         pmsg = nlmsg_data(nlh);
5952         pmsg->prefix_family = AF_INET6;
5953         pmsg->prefix_pad1 = 0;
5954         pmsg->prefix_pad2 = 0;
5955         pmsg->prefix_ifindex = idev->dev->ifindex;
5956         pmsg->prefix_len = pinfo->prefix_len;
5957         pmsg->prefix_type = pinfo->type;
5958         pmsg->prefix_pad3 = 0;
5959         pmsg->prefix_flags = 0;
5960         if (pinfo->onlink)
5961                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5962         if (pinfo->autoconf)
5963                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5964
5965         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5966                 goto nla_put_failure;
5967         ci.preferred_time = ntohl(pinfo->prefered);
5968         ci.valid_time = ntohl(pinfo->valid);
5969         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5970                 goto nla_put_failure;
5971         nlmsg_end(skb, nlh);
5972         return 0;
5973
5974 nla_put_failure:
5975         nlmsg_cancel(skb, nlh);
5976         return -EMSGSIZE;
5977 }
5978
5979 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5980                          struct prefix_info *pinfo)
5981 {
5982         struct sk_buff *skb;
5983         struct net *net = dev_net(idev->dev);
5984         int err = -ENOBUFS;
5985
5986         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5987         if (!skb)
5988                 goto errout;
5989
5990         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5991         if (err < 0) {
5992                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5993                 WARN_ON(err == -EMSGSIZE);
5994                 kfree_skb(skb);
5995                 goto errout;
5996         }
5997         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5998         return;
5999 errout:
6000         if (err < 0)
6001                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6002 }
6003
6004 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6005 {
6006         struct net *net = dev_net(ifp->idev->dev);
6007
6008         if (event)
6009                 ASSERT_RTNL();
6010
6011         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6012
6013         switch (event) {
6014         case RTM_NEWADDR:
6015                 /*
6016                  * If the address was optimistic we inserted the route at the
6017                  * start of our DAD process, so we don't need to do it again.
6018                  * If the device was taken down in the middle of the DAD
6019                  * cycle there is a race where we could get here without a
6020                  * host route, so nothing to insert. That will be fixed when
6021                  * the device is brought up.
6022                  */
6023                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6024                         ip6_ins_rt(net, ifp->rt);
6025                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6026                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6027                                 &ifp->addr, ifp->idev->dev->name);
6028                 }
6029
6030                 if (ifp->idev->cnf.forwarding)
6031                         addrconf_join_anycast(ifp);
6032                 if (!ipv6_addr_any(&ifp->peer_addr))
6033                         addrconf_prefix_route(&ifp->peer_addr, 128,
6034                                               ifp->rt_priority, ifp->idev->dev,
6035                                               0, 0, GFP_ATOMIC);
6036                 break;
6037         case RTM_DELADDR:
6038                 if (ifp->idev->cnf.forwarding)
6039                         addrconf_leave_anycast(ifp);
6040                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6041                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6042                         struct fib6_info *rt;
6043
6044                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6045                                                        ifp->idev->dev, 0, 0,
6046                                                        false);
6047                         if (rt)
6048                                 ip6_del_rt(net, rt, false);
6049                 }
6050                 if (ifp->rt) {
6051                         ip6_del_rt(net, ifp->rt, false);
6052                         ifp->rt = NULL;
6053                 }
6054                 rt_genid_bump_ipv6(net);
6055                 break;
6056         }
6057         atomic_inc(&net->ipv6.dev_addr_genid);
6058 }
6059
6060 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6061 {
6062         rcu_read_lock_bh();
6063         if (likely(ifp->idev->dead == 0))
6064                 __ipv6_ifa_notify(event, ifp);
6065         rcu_read_unlock_bh();
6066 }
6067
6068 #ifdef CONFIG_SYSCTL
6069
6070 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6071                 void *buffer, size_t *lenp, loff_t *ppos)
6072 {
6073         int *valp = ctl->data;
6074         int val = *valp;
6075         loff_t pos = *ppos;
6076         struct ctl_table lctl;
6077         int ret;
6078
6079         /*
6080          * ctl->data points to idev->cnf.forwarding, we should
6081          * not modify it until we get the rtnl lock.
6082          */
6083         lctl = *ctl;
6084         lctl.data = &val;
6085
6086         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6087
6088         if (write)
6089                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6090         if (ret)
6091                 *ppos = pos;
6092         return ret;
6093 }
6094
6095 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6096                 void *buffer, size_t *lenp, loff_t *ppos)
6097 {
6098         struct inet6_dev *idev = ctl->extra1;
6099         int min_mtu = IPV6_MIN_MTU;
6100         struct ctl_table lctl;
6101
6102         lctl = *ctl;
6103         lctl.extra1 = &min_mtu;
6104         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6105
6106         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6107 }
6108
6109 static void dev_disable_change(struct inet6_dev *idev)
6110 {
6111         struct netdev_notifier_info info;
6112
6113         if (!idev || !idev->dev)
6114                 return;
6115
6116         netdev_notifier_info_init(&info, idev->dev);
6117         if (idev->cnf.disable_ipv6)
6118                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6119         else
6120                 addrconf_notify(NULL, NETDEV_UP, &info);
6121 }
6122
6123 static void addrconf_disable_change(struct net *net, __s32 newf)
6124 {
6125         struct net_device *dev;
6126         struct inet6_dev *idev;
6127
6128         for_each_netdev(net, dev) {
6129                 idev = __in6_dev_get(dev);
6130                 if (idev) {
6131                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6132                         idev->cnf.disable_ipv6 = newf;
6133                         if (changed)
6134                                 dev_disable_change(idev);
6135                 }
6136         }
6137 }
6138
6139 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6140 {
6141         struct net *net;
6142         int old;
6143
6144         if (!rtnl_trylock())
6145                 return restart_syscall();
6146
6147         net = (struct net *)table->extra2;
6148         old = *p;
6149         *p = newf;
6150
6151         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6152                 rtnl_unlock();
6153                 return 0;
6154         }
6155
6156         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6157                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6158                 addrconf_disable_change(net, newf);
6159         } else if ((!newf) ^ (!old))
6160                 dev_disable_change((struct inet6_dev *)table->extra1);
6161
6162         rtnl_unlock();
6163         return 0;
6164 }
6165
6166 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6167                 void *buffer, size_t *lenp, loff_t *ppos)
6168 {
6169         int *valp = ctl->data;
6170         int val = *valp;
6171         loff_t pos = *ppos;
6172         struct ctl_table lctl;
6173         int ret;
6174
6175         /*
6176          * ctl->data points to idev->cnf.disable_ipv6, we should
6177          * not modify it until we get the rtnl lock.
6178          */
6179         lctl = *ctl;
6180         lctl.data = &val;
6181
6182         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6183
6184         if (write)
6185                 ret = addrconf_disable_ipv6(ctl, valp, val);
6186         if (ret)
6187                 *ppos = pos;
6188         return ret;
6189 }
6190
6191 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6192                 void *buffer, size_t *lenp, loff_t *ppos)
6193 {
6194         int *valp = ctl->data;
6195         int ret;
6196         int old, new;
6197
6198         old = *valp;
6199         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6200         new = *valp;
6201
6202         if (write && old != new) {
6203                 struct net *net = ctl->extra2;
6204
6205                 if (!rtnl_trylock())
6206                         return restart_syscall();
6207
6208                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6209                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6210                                                      NETCONFA_PROXY_NEIGH,
6211                                                      NETCONFA_IFINDEX_DEFAULT,
6212                                                      net->ipv6.devconf_dflt);
6213                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6214                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6215                                                      NETCONFA_PROXY_NEIGH,
6216                                                      NETCONFA_IFINDEX_ALL,
6217                                                      net->ipv6.devconf_all);
6218                 else {
6219                         struct inet6_dev *idev = ctl->extra1;
6220
6221                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6222                                                      NETCONFA_PROXY_NEIGH,
6223                                                      idev->dev->ifindex,
6224                                                      &idev->cnf);
6225                 }
6226                 rtnl_unlock();
6227         }
6228
6229         return ret;
6230 }
6231
6232 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6233                                          void *buffer, size_t *lenp,
6234                                          loff_t *ppos)
6235 {
6236         int ret = 0;
6237         u32 new_val;
6238         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6239         struct net *net = (struct net *)ctl->extra2;
6240         struct ctl_table tmp = {
6241                 .data = &new_val,
6242                 .maxlen = sizeof(new_val),
6243                 .mode = ctl->mode,
6244         };
6245
6246         if (!rtnl_trylock())
6247                 return restart_syscall();
6248
6249         new_val = *((u32 *)ctl->data);
6250
6251         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6252         if (ret != 0)
6253                 goto out;
6254
6255         if (write) {
6256                 if (check_addr_gen_mode(new_val) < 0) {
6257                         ret = -EINVAL;
6258                         goto out;
6259                 }
6260
6261                 if (idev) {
6262                         if (check_stable_privacy(idev, net, new_val) < 0) {
6263                                 ret = -EINVAL;
6264                                 goto out;
6265                         }
6266
6267                         if (idev->cnf.addr_gen_mode != new_val) {
6268                                 idev->cnf.addr_gen_mode = new_val;
6269                                 addrconf_dev_config(idev->dev);
6270                         }
6271                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6272                         struct net_device *dev;
6273
6274                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6275                         for_each_netdev(net, dev) {
6276                                 idev = __in6_dev_get(dev);
6277                                 if (idev &&
6278                                     idev->cnf.addr_gen_mode != new_val) {
6279                                         idev->cnf.addr_gen_mode = new_val;
6280                                         addrconf_dev_config(idev->dev);
6281                                 }
6282                         }
6283                 }
6284
6285                 *((u32 *)ctl->data) = new_val;
6286         }
6287
6288 out:
6289         rtnl_unlock();
6290
6291         return ret;
6292 }
6293
6294 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6295                                          void *buffer, size_t *lenp,
6296                                          loff_t *ppos)
6297 {
6298         int err;
6299         struct in6_addr addr;
6300         char str[IPV6_MAX_STRLEN];
6301         struct ctl_table lctl = *ctl;
6302         struct net *net = ctl->extra2;
6303         struct ipv6_stable_secret *secret = ctl->data;
6304
6305         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6306                 return -EIO;
6307
6308         lctl.maxlen = IPV6_MAX_STRLEN;
6309         lctl.data = str;
6310
6311         if (!rtnl_trylock())
6312                 return restart_syscall();
6313
6314         if (!write && !secret->initialized) {
6315                 err = -EIO;
6316                 goto out;
6317         }
6318
6319         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6320         if (err >= sizeof(str)) {
6321                 err = -EIO;
6322                 goto out;
6323         }
6324
6325         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6326         if (err || !write)
6327                 goto out;
6328
6329         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6330                 err = -EIO;
6331                 goto out;
6332         }
6333
6334         secret->initialized = true;
6335         secret->secret = addr;
6336
6337         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6338                 struct net_device *dev;
6339
6340                 for_each_netdev(net, dev) {
6341                         struct inet6_dev *idev = __in6_dev_get(dev);
6342
6343                         if (idev) {
6344                                 idev->cnf.addr_gen_mode =
6345                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6346                         }
6347                 }
6348         } else {
6349                 struct inet6_dev *idev = ctl->extra1;
6350
6351                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6352         }
6353
6354 out:
6355         rtnl_unlock();
6356
6357         return err;
6358 }
6359
6360 static
6361 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6362                                                 int write, void *buffer,
6363                                                 size_t *lenp,
6364                                                 loff_t *ppos)
6365 {
6366         int *valp = ctl->data;
6367         int val = *valp;
6368         loff_t pos = *ppos;
6369         struct ctl_table lctl;
6370         int ret;
6371
6372         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6373          * we should not modify it until we get the rtnl lock.
6374          */
6375         lctl = *ctl;
6376         lctl.data = &val;
6377
6378         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6379
6380         if (write)
6381                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6382         if (ret)
6383                 *ppos = pos;
6384         return ret;
6385 }
6386
6387 static
6388 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6389 {
6390         if (rt) {
6391                 if (action)
6392                         rt->dst.flags |= DST_NOPOLICY;
6393                 else
6394                         rt->dst.flags &= ~DST_NOPOLICY;
6395         }
6396 }
6397
6398 static
6399 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6400 {
6401         struct inet6_ifaddr *ifa;
6402
6403         read_lock_bh(&idev->lock);
6404         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6405                 spin_lock(&ifa->lock);
6406                 if (ifa->rt) {
6407                         /* host routes only use builtin fib6_nh */
6408                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6409                         int cpu;
6410
6411                         rcu_read_lock();
6412                         ifa->rt->dst_nopolicy = val ? true : false;
6413                         if (nh->rt6i_pcpu) {
6414                                 for_each_possible_cpu(cpu) {
6415                                         struct rt6_info **rtp;
6416
6417                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6418                                         addrconf_set_nopolicy(*rtp, val);
6419                                 }
6420                         }
6421                         rcu_read_unlock();
6422                 }
6423                 spin_unlock(&ifa->lock);
6424         }
6425         read_unlock_bh(&idev->lock);
6426 }
6427
6428 static
6429 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6430 {
6431         struct inet6_dev *idev;
6432         struct net *net;
6433
6434         if (!rtnl_trylock())
6435                 return restart_syscall();
6436
6437         *valp = val;
6438
6439         net = (struct net *)ctl->extra2;
6440         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6441                 rtnl_unlock();
6442                 return 0;
6443         }
6444
6445         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6446                 struct net_device *dev;
6447
6448                 for_each_netdev(net, dev) {
6449                         idev = __in6_dev_get(dev);
6450                         if (idev)
6451                                 addrconf_disable_policy_idev(idev, val);
6452                 }
6453         } else {
6454                 idev = (struct inet6_dev *)ctl->extra1;
6455                 addrconf_disable_policy_idev(idev, val);
6456         }
6457
6458         rtnl_unlock();
6459         return 0;
6460 }
6461
6462 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6463                                    void *buffer, size_t *lenp, loff_t *ppos)
6464 {
6465         int *valp = ctl->data;
6466         int val = *valp;
6467         loff_t pos = *ppos;
6468         struct ctl_table lctl;
6469         int ret;
6470
6471         lctl = *ctl;
6472         lctl.data = &val;
6473         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6474
6475         if (write && (*valp != val))
6476                 ret = addrconf_disable_policy(ctl, valp, val);
6477
6478         if (ret)
6479                 *ppos = pos;
6480
6481         return ret;
6482 }
6483
6484 static int minus_one = -1;
6485 static const int two_five_five = 255;
6486
6487 static const struct ctl_table addrconf_sysctl[] = {
6488         {
6489                 .procname       = "forwarding",
6490                 .data           = &ipv6_devconf.forwarding,
6491                 .maxlen         = sizeof(int),
6492                 .mode           = 0644,
6493                 .proc_handler   = addrconf_sysctl_forward,
6494         },
6495         {
6496                 .procname       = "hop_limit",
6497                 .data           = &ipv6_devconf.hop_limit,
6498                 .maxlen         = sizeof(int),
6499                 .mode           = 0644,
6500                 .proc_handler   = proc_dointvec_minmax,
6501                 .extra1         = (void *)SYSCTL_ONE,
6502                 .extra2         = (void *)&two_five_five,
6503         },
6504         {
6505                 .procname       = "mtu",
6506                 .data           = &ipv6_devconf.mtu6,
6507                 .maxlen         = sizeof(int),
6508                 .mode           = 0644,
6509                 .proc_handler   = addrconf_sysctl_mtu,
6510         },
6511         {
6512                 .procname       = "accept_ra",
6513                 .data           = &ipv6_devconf.accept_ra,
6514                 .maxlen         = sizeof(int),
6515                 .mode           = 0644,
6516                 .proc_handler   = proc_dointvec,
6517         },
6518         {
6519                 .procname       = "accept_redirects",
6520                 .data           = &ipv6_devconf.accept_redirects,
6521                 .maxlen         = sizeof(int),
6522                 .mode           = 0644,
6523                 .proc_handler   = proc_dointvec,
6524         },
6525         {
6526                 .procname       = "autoconf",
6527                 .data           = &ipv6_devconf.autoconf,
6528                 .maxlen         = sizeof(int),
6529                 .mode           = 0644,
6530                 .proc_handler   = proc_dointvec,
6531         },
6532         {
6533                 .procname       = "dad_transmits",
6534                 .data           = &ipv6_devconf.dad_transmits,
6535                 .maxlen         = sizeof(int),
6536                 .mode           = 0644,
6537                 .proc_handler   = proc_dointvec,
6538         },
6539         {
6540                 .procname       = "router_solicitations",
6541                 .data           = &ipv6_devconf.rtr_solicits,
6542                 .maxlen         = sizeof(int),
6543                 .mode           = 0644,
6544                 .proc_handler   = proc_dointvec_minmax,
6545                 .extra1         = &minus_one,
6546         },
6547         {
6548                 .procname       = "router_solicitation_interval",
6549                 .data           = &ipv6_devconf.rtr_solicit_interval,
6550                 .maxlen         = sizeof(int),
6551                 .mode           = 0644,
6552                 .proc_handler   = proc_dointvec_jiffies,
6553         },
6554         {
6555                 .procname       = "router_solicitation_max_interval",
6556                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6557                 .maxlen         = sizeof(int),
6558                 .mode           = 0644,
6559                 .proc_handler   = proc_dointvec_jiffies,
6560         },
6561         {
6562                 .procname       = "router_solicitation_delay",
6563                 .data           = &ipv6_devconf.rtr_solicit_delay,
6564                 .maxlen         = sizeof(int),
6565                 .mode           = 0644,
6566                 .proc_handler   = proc_dointvec_jiffies,
6567         },
6568         {
6569                 .procname       = "force_mld_version",
6570                 .data           = &ipv6_devconf.force_mld_version,
6571                 .maxlen         = sizeof(int),
6572                 .mode           = 0644,
6573                 .proc_handler   = proc_dointvec,
6574         },
6575         {
6576                 .procname       = "mldv1_unsolicited_report_interval",
6577                 .data           =
6578                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6579                 .maxlen         = sizeof(int),
6580                 .mode           = 0644,
6581                 .proc_handler   = proc_dointvec_ms_jiffies,
6582         },
6583         {
6584                 .procname       = "mldv2_unsolicited_report_interval",
6585                 .data           =
6586                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6587                 .maxlen         = sizeof(int),
6588                 .mode           = 0644,
6589                 .proc_handler   = proc_dointvec_ms_jiffies,
6590         },
6591         {
6592                 .procname       = "use_tempaddr",
6593                 .data           = &ipv6_devconf.use_tempaddr,
6594                 .maxlen         = sizeof(int),
6595                 .mode           = 0644,
6596                 .proc_handler   = proc_dointvec,
6597         },
6598         {
6599                 .procname       = "temp_valid_lft",
6600                 .data           = &ipv6_devconf.temp_valid_lft,
6601                 .maxlen         = sizeof(int),
6602                 .mode           = 0644,
6603                 .proc_handler   = proc_dointvec,
6604         },
6605         {
6606                 .procname       = "temp_prefered_lft",
6607                 .data           = &ipv6_devconf.temp_prefered_lft,
6608                 .maxlen         = sizeof(int),
6609                 .mode           = 0644,
6610                 .proc_handler   = proc_dointvec,
6611         },
6612         {
6613                 .procname       = "regen_max_retry",
6614                 .data           = &ipv6_devconf.regen_max_retry,
6615                 .maxlen         = sizeof(int),
6616                 .mode           = 0644,
6617                 .proc_handler   = proc_dointvec,
6618         },
6619         {
6620                 .procname       = "max_desync_factor",
6621                 .data           = &ipv6_devconf.max_desync_factor,
6622                 .maxlen         = sizeof(int),
6623                 .mode           = 0644,
6624                 .proc_handler   = proc_dointvec,
6625         },
6626         {
6627                 .procname       = "max_addresses",
6628                 .data           = &ipv6_devconf.max_addresses,
6629                 .maxlen         = sizeof(int),
6630                 .mode           = 0644,
6631                 .proc_handler   = proc_dointvec,
6632         },
6633         {
6634                 .procname       = "accept_ra_defrtr",
6635                 .data           = &ipv6_devconf.accept_ra_defrtr,
6636                 .maxlen         = sizeof(int),
6637                 .mode           = 0644,
6638                 .proc_handler   = proc_dointvec,
6639         },
6640         {
6641                 .procname       = "accept_ra_min_hop_limit",
6642                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6643                 .maxlen         = sizeof(int),
6644                 .mode           = 0644,
6645                 .proc_handler   = proc_dointvec,
6646         },
6647         {
6648                 .procname       = "accept_ra_pinfo",
6649                 .data           = &ipv6_devconf.accept_ra_pinfo,
6650                 .maxlen         = sizeof(int),
6651                 .mode           = 0644,
6652                 .proc_handler   = proc_dointvec,
6653         },
6654 #ifdef CONFIG_IPV6_ROUTER_PREF
6655         {
6656                 .procname       = "accept_ra_rtr_pref",
6657                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6658                 .maxlen         = sizeof(int),
6659                 .mode           = 0644,
6660                 .proc_handler   = proc_dointvec,
6661         },
6662         {
6663                 .procname       = "router_probe_interval",
6664                 .data           = &ipv6_devconf.rtr_probe_interval,
6665                 .maxlen         = sizeof(int),
6666                 .mode           = 0644,
6667                 .proc_handler   = proc_dointvec_jiffies,
6668         },
6669 #ifdef CONFIG_IPV6_ROUTE_INFO
6670         {
6671                 .procname       = "accept_ra_rt_info_min_plen",
6672                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6673                 .maxlen         = sizeof(int),
6674                 .mode           = 0644,
6675                 .proc_handler   = proc_dointvec,
6676         },
6677         {
6678                 .procname       = "accept_ra_rt_info_max_plen",
6679                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6680                 .maxlen         = sizeof(int),
6681                 .mode           = 0644,
6682                 .proc_handler   = proc_dointvec,
6683         },
6684 #endif
6685 #endif
6686         {
6687                 .procname       = "proxy_ndp",
6688                 .data           = &ipv6_devconf.proxy_ndp,
6689                 .maxlen         = sizeof(int),
6690                 .mode           = 0644,
6691                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6692         },
6693         {
6694                 .procname       = "accept_source_route",
6695                 .data           = &ipv6_devconf.accept_source_route,
6696                 .maxlen         = sizeof(int),
6697                 .mode           = 0644,
6698                 .proc_handler   = proc_dointvec,
6699         },
6700 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6701         {
6702                 .procname       = "optimistic_dad",
6703                 .data           = &ipv6_devconf.optimistic_dad,
6704                 .maxlen         = sizeof(int),
6705                 .mode           = 0644,
6706                 .proc_handler   = proc_dointvec,
6707         },
6708         {
6709                 .procname       = "use_optimistic",
6710                 .data           = &ipv6_devconf.use_optimistic,
6711                 .maxlen         = sizeof(int),
6712                 .mode           = 0644,
6713                 .proc_handler   = proc_dointvec,
6714         },
6715 #endif
6716 #ifdef CONFIG_IPV6_MROUTE
6717         {
6718                 .procname       = "mc_forwarding",
6719                 .data           = &ipv6_devconf.mc_forwarding,
6720                 .maxlen         = sizeof(int),
6721                 .mode           = 0444,
6722                 .proc_handler   = proc_dointvec,
6723         },
6724 #endif
6725         {
6726                 .procname       = "disable_ipv6",
6727                 .data           = &ipv6_devconf.disable_ipv6,
6728                 .maxlen         = sizeof(int),
6729                 .mode           = 0644,
6730                 .proc_handler   = addrconf_sysctl_disable,
6731         },
6732         {
6733                 .procname       = "accept_dad",
6734                 .data           = &ipv6_devconf.accept_dad,
6735                 .maxlen         = sizeof(int),
6736                 .mode           = 0644,
6737                 .proc_handler   = proc_dointvec,
6738         },
6739         {
6740                 .procname       = "force_tllao",
6741                 .data           = &ipv6_devconf.force_tllao,
6742                 .maxlen         = sizeof(int),
6743                 .mode           = 0644,
6744                 .proc_handler   = proc_dointvec
6745         },
6746         {
6747                 .procname       = "ndisc_notify",
6748                 .data           = &ipv6_devconf.ndisc_notify,
6749                 .maxlen         = sizeof(int),
6750                 .mode           = 0644,
6751                 .proc_handler   = proc_dointvec
6752         },
6753         {
6754                 .procname       = "suppress_frag_ndisc",
6755                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6756                 .maxlen         = sizeof(int),
6757                 .mode           = 0644,
6758                 .proc_handler   = proc_dointvec
6759         },
6760         {
6761                 .procname       = "accept_ra_from_local",
6762                 .data           = &ipv6_devconf.accept_ra_from_local,
6763                 .maxlen         = sizeof(int),
6764                 .mode           = 0644,
6765                 .proc_handler   = proc_dointvec,
6766         },
6767         {
6768                 .procname       = "accept_ra_mtu",
6769                 .data           = &ipv6_devconf.accept_ra_mtu,
6770                 .maxlen         = sizeof(int),
6771                 .mode           = 0644,
6772                 .proc_handler   = proc_dointvec,
6773         },
6774         {
6775                 .procname       = "stable_secret",
6776                 .data           = &ipv6_devconf.stable_secret,
6777                 .maxlen         = IPV6_MAX_STRLEN,
6778                 .mode           = 0600,
6779                 .proc_handler   = addrconf_sysctl_stable_secret,
6780         },
6781         {
6782                 .procname       = "use_oif_addrs_only",
6783                 .data           = &ipv6_devconf.use_oif_addrs_only,
6784                 .maxlen         = sizeof(int),
6785                 .mode           = 0644,
6786                 .proc_handler   = proc_dointvec,
6787         },
6788         {
6789                 .procname       = "ignore_routes_with_linkdown",
6790                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6791                 .maxlen         = sizeof(int),
6792                 .mode           = 0644,
6793                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6794         },
6795         {
6796                 .procname       = "drop_unicast_in_l2_multicast",
6797                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6798                 .maxlen         = sizeof(int),
6799                 .mode           = 0644,
6800                 .proc_handler   = proc_dointvec,
6801         },
6802         {
6803                 .procname       = "drop_unsolicited_na",
6804                 .data           = &ipv6_devconf.drop_unsolicited_na,
6805                 .maxlen         = sizeof(int),
6806                 .mode           = 0644,
6807                 .proc_handler   = proc_dointvec,
6808         },
6809         {
6810                 .procname       = "keep_addr_on_down",
6811                 .data           = &ipv6_devconf.keep_addr_on_down,
6812                 .maxlen         = sizeof(int),
6813                 .mode           = 0644,
6814                 .proc_handler   = proc_dointvec,
6815
6816         },
6817         {
6818                 .procname       = "seg6_enabled",
6819                 .data           = &ipv6_devconf.seg6_enabled,
6820                 .maxlen         = sizeof(int),
6821                 .mode           = 0644,
6822                 .proc_handler   = proc_dointvec,
6823         },
6824 #ifdef CONFIG_IPV6_SEG6_HMAC
6825         {
6826                 .procname       = "seg6_require_hmac",
6827                 .data           = &ipv6_devconf.seg6_require_hmac,
6828                 .maxlen         = sizeof(int),
6829                 .mode           = 0644,
6830                 .proc_handler   = proc_dointvec,
6831         },
6832 #endif
6833         {
6834                 .procname       = "enhanced_dad",
6835                 .data           = &ipv6_devconf.enhanced_dad,
6836                 .maxlen         = sizeof(int),
6837                 .mode           = 0644,
6838                 .proc_handler   = proc_dointvec,
6839         },
6840         {
6841                 .procname               = "addr_gen_mode",
6842                 .data                   = &ipv6_devconf.addr_gen_mode,
6843                 .maxlen                 = sizeof(int),
6844                 .mode                   = 0644,
6845                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6846         },
6847         {
6848                 .procname       = "disable_policy",
6849                 .data           = &ipv6_devconf.disable_policy,
6850                 .maxlen         = sizeof(int),
6851                 .mode           = 0644,
6852                 .proc_handler   = addrconf_sysctl_disable_policy,
6853         },
6854         {
6855                 .procname       = "ndisc_tclass",
6856                 .data           = &ipv6_devconf.ndisc_tclass,
6857                 .maxlen         = sizeof(int),
6858                 .mode           = 0644,
6859                 .proc_handler   = proc_dointvec_minmax,
6860                 .extra1         = (void *)SYSCTL_ZERO,
6861                 .extra2         = (void *)&two_five_five,
6862         },
6863         {
6864                 .procname       = "rpl_seg_enabled",
6865                 .data           = &ipv6_devconf.rpl_seg_enabled,
6866                 .maxlen         = sizeof(int),
6867                 .mode           = 0644,
6868                 .proc_handler   = proc_dointvec,
6869         },
6870         {
6871                 /* sentinel */
6872         }
6873 };
6874
6875 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6876                 struct inet6_dev *idev, struct ipv6_devconf *p)
6877 {
6878         int i, ifindex;
6879         struct ctl_table *table;
6880         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6881
6882         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6883         if (!table)
6884                 goto out;
6885
6886         for (i = 0; table[i].data; i++) {
6887                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6888                 /* If one of these is already set, then it is not safe to
6889                  * overwrite either of them: this makes proc_dointvec_minmax
6890                  * usable.
6891                  */
6892                 if (!table[i].extra1 && !table[i].extra2) {
6893                         table[i].extra1 = idev; /* embedded; no ref */
6894                         table[i].extra2 = net;
6895                 }
6896         }
6897
6898         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6899
6900         p->sysctl_header = register_net_sysctl(net, path, table);
6901         if (!p->sysctl_header)
6902                 goto free;
6903
6904         if (!strcmp(dev_name, "all"))
6905                 ifindex = NETCONFA_IFINDEX_ALL;
6906         else if (!strcmp(dev_name, "default"))
6907                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6908         else
6909                 ifindex = idev->dev->ifindex;
6910         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6911                                      ifindex, p);
6912         return 0;
6913
6914 free:
6915         kfree(table);
6916 out:
6917         return -ENOBUFS;
6918 }
6919
6920 static void __addrconf_sysctl_unregister(struct net *net,
6921                                          struct ipv6_devconf *p, int ifindex)
6922 {
6923         struct ctl_table *table;
6924
6925         if (!p->sysctl_header)
6926                 return;
6927
6928         table = p->sysctl_header->ctl_table_arg;
6929         unregister_net_sysctl_table(p->sysctl_header);
6930         p->sysctl_header = NULL;
6931         kfree(table);
6932
6933         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6934 }
6935
6936 static int addrconf_sysctl_register(struct inet6_dev *idev)
6937 {
6938         int err;
6939
6940         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6941                 return -EINVAL;
6942
6943         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6944                                     &ndisc_ifinfo_sysctl_change);
6945         if (err)
6946                 return err;
6947         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6948                                          idev, &idev->cnf);
6949         if (err)
6950                 neigh_sysctl_unregister(idev->nd_parms);
6951
6952         return err;
6953 }
6954
6955 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6956 {
6957         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6958                                      idev->dev->ifindex);
6959         neigh_sysctl_unregister(idev->nd_parms);
6960 }
6961
6962
6963 #endif
6964
6965 static int __net_init addrconf_init_net(struct net *net)
6966 {
6967         int err = -ENOMEM;
6968         struct ipv6_devconf *all, *dflt;
6969
6970         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6971         if (!all)
6972                 goto err_alloc_all;
6973
6974         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6975         if (!dflt)
6976                 goto err_alloc_dflt;
6977
6978         if (IS_ENABLED(CONFIG_SYSCTL) &&
6979             !net_eq(net, &init_net)) {
6980                 switch (sysctl_devconf_inherit_init_net) {
6981                 case 1:  /* copy from init_net */
6982                         memcpy(all, init_net.ipv6.devconf_all,
6983                                sizeof(ipv6_devconf));
6984                         memcpy(dflt, init_net.ipv6.devconf_dflt,
6985                                sizeof(ipv6_devconf_dflt));
6986                         break;
6987                 case 3: /* copy from the current netns */
6988                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
6989                                sizeof(ipv6_devconf));
6990                         memcpy(dflt,
6991                                current->nsproxy->net_ns->ipv6.devconf_dflt,
6992                                sizeof(ipv6_devconf_dflt));
6993                         break;
6994                 case 0:
6995                 case 2:
6996                         /* use compiled values */
6997                         break;
6998                 }
6999         }
7000
7001         /* these will be inherited by all namespaces */
7002         dflt->autoconf = ipv6_defaults.autoconf;
7003         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7004
7005         dflt->stable_secret.initialized = false;
7006         all->stable_secret.initialized = false;
7007
7008         net->ipv6.devconf_all = all;
7009         net->ipv6.devconf_dflt = dflt;
7010
7011 #ifdef CONFIG_SYSCTL
7012         err = __addrconf_sysctl_register(net, "all", NULL, all);
7013         if (err < 0)
7014                 goto err_reg_all;
7015
7016         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7017         if (err < 0)
7018                 goto err_reg_dflt;
7019 #endif
7020         return 0;
7021
7022 #ifdef CONFIG_SYSCTL
7023 err_reg_dflt:
7024         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7025 err_reg_all:
7026         kfree(dflt);
7027 #endif
7028 err_alloc_dflt:
7029         kfree(all);
7030 err_alloc_all:
7031         return err;
7032 }
7033
7034 static void __net_exit addrconf_exit_net(struct net *net)
7035 {
7036 #ifdef CONFIG_SYSCTL
7037         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7038                                      NETCONFA_IFINDEX_DEFAULT);
7039         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7040                                      NETCONFA_IFINDEX_ALL);
7041 #endif
7042         kfree(net->ipv6.devconf_dflt);
7043         kfree(net->ipv6.devconf_all);
7044 }
7045
7046 static struct pernet_operations addrconf_ops = {
7047         .init = addrconf_init_net,
7048         .exit = addrconf_exit_net,
7049 };
7050
7051 static struct rtnl_af_ops inet6_ops __read_mostly = {
7052         .family           = AF_INET6,
7053         .fill_link_af     = inet6_fill_link_af,
7054         .get_link_af_size = inet6_get_link_af_size,
7055         .validate_link_af = inet6_validate_link_af,
7056         .set_link_af      = inet6_set_link_af,
7057 };
7058
7059 /*
7060  *      Init / cleanup code
7061  */
7062
7063 int __init addrconf_init(void)
7064 {
7065         struct inet6_dev *idev;
7066         int i, err;
7067
7068         err = ipv6_addr_label_init();
7069         if (err < 0) {
7070                 pr_crit("%s: cannot initialize default policy table: %d\n",
7071                         __func__, err);
7072                 goto out;
7073         }
7074
7075         err = register_pernet_subsys(&addrconf_ops);
7076         if (err < 0)
7077                 goto out_addrlabel;
7078
7079         addrconf_wq = create_workqueue("ipv6_addrconf");
7080         if (!addrconf_wq) {
7081                 err = -ENOMEM;
7082                 goto out_nowq;
7083         }
7084
7085         /* The addrconf netdev notifier requires that loopback_dev
7086          * has it's ipv6 private information allocated and setup
7087          * before it can bring up and give link-local addresses
7088          * to other devices which are up.
7089          *
7090          * Unfortunately, loopback_dev is not necessarily the first
7091          * entry in the global dev_base list of net devices.  In fact,
7092          * it is likely to be the very last entry on that list.
7093          * So this causes the notifier registry below to try and
7094          * give link-local addresses to all devices besides loopback_dev
7095          * first, then loopback_dev, which cases all the non-loopback_dev
7096          * devices to fail to get a link-local address.
7097          *
7098          * So, as a temporary fix, allocate the ipv6 structure for
7099          * loopback_dev first by hand.
7100          * Longer term, all of the dependencies ipv6 has upon the loopback
7101          * device and it being up should be removed.
7102          */
7103         rtnl_lock();
7104         idev = ipv6_add_dev(init_net.loopback_dev);
7105         rtnl_unlock();
7106         if (IS_ERR(idev)) {
7107                 err = PTR_ERR(idev);
7108                 goto errlo;
7109         }
7110
7111         ip6_route_init_special_entries();
7112
7113         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7114                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7115
7116         register_netdevice_notifier(&ipv6_dev_notf);
7117
7118         addrconf_verify();
7119
7120         rtnl_af_register(&inet6_ops);
7121
7122         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7123                                    NULL, inet6_dump_ifinfo, 0);
7124         if (err < 0)
7125                 goto errout;
7126
7127         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7128                                    inet6_rtm_newaddr, NULL, 0);
7129         if (err < 0)
7130                 goto errout;
7131         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7132                                    inet6_rtm_deladdr, NULL, 0);
7133         if (err < 0)
7134                 goto errout;
7135         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7136                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7137                                    RTNL_FLAG_DOIT_UNLOCKED);
7138         if (err < 0)
7139                 goto errout;
7140         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7141                                    NULL, inet6_dump_ifmcaddr, 0);
7142         if (err < 0)
7143                 goto errout;
7144         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7145                                    NULL, inet6_dump_ifacaddr, 0);
7146         if (err < 0)
7147                 goto errout;
7148         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7149                                    inet6_netconf_get_devconf,
7150                                    inet6_netconf_dump_devconf,
7151                                    RTNL_FLAG_DOIT_UNLOCKED);
7152         if (err < 0)
7153                 goto errout;
7154         err = ipv6_addr_label_rtnl_register();
7155         if (err < 0)
7156                 goto errout;
7157
7158         return 0;
7159 errout:
7160         rtnl_unregister_all(PF_INET6);
7161         rtnl_af_unregister(&inet6_ops);
7162         unregister_netdevice_notifier(&ipv6_dev_notf);
7163 errlo:
7164         destroy_workqueue(addrconf_wq);
7165 out_nowq:
7166         unregister_pernet_subsys(&addrconf_ops);
7167 out_addrlabel:
7168         ipv6_addr_label_cleanup();
7169 out:
7170         return err;
7171 }
7172
7173 void addrconf_cleanup(void)
7174 {
7175         struct net_device *dev;
7176         int i;
7177
7178         unregister_netdevice_notifier(&ipv6_dev_notf);
7179         unregister_pernet_subsys(&addrconf_ops);
7180         ipv6_addr_label_cleanup();
7181
7182         rtnl_af_unregister(&inet6_ops);
7183
7184         rtnl_lock();
7185
7186         /* clean dev list */
7187         for_each_netdev(&init_net, dev) {
7188                 if (__in6_dev_get(dev) == NULL)
7189                         continue;
7190                 addrconf_ifdown(dev, 1);
7191         }
7192         addrconf_ifdown(init_net.loopback_dev, 2);
7193
7194         /*
7195          *      Check hash table.
7196          */
7197         spin_lock_bh(&addrconf_hash_lock);
7198         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7199                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7200         spin_unlock_bh(&addrconf_hash_lock);
7201         cancel_delayed_work(&addr_chk_work);
7202         rtnl_unlock();
7203
7204         destroy_workqueue(addrconf_wq);
7205 }