net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92
  93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  94
  95 #define IPV6_MAX_STRLEN \
  96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  97
  98 static inline u32 cstamp_delta(unsigned long cstamp)
  99 {
 100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 101 }
 102
 103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 104 {
 105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 107         do_div(tmp, 1000000);
 108         return (s32)tmp;
 109 }
 110
 111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 112 {
 113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 119                 do_div(tmp, 1000000);
 120         }
 121         return (s32)tmp;
 122 }
 123
 124 #ifdef CONFIG_SYSCTL
 125 static int addrconf_sysctl_register(struct inet6_dev *idev);
 126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 127 #else
 128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 129 {
 130         return 0;
 131 }
 132
 133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 134 {
 135 }
 136 #endif
 137
 138 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 139
 140 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 141 static int ipv6_count_addresses(const struct inet6_dev *idev);
 142 static int ipv6_generate_stable_address(struct in6_addr *addr,
 143                                         u8 dad_count,
 144                                         const struct inet6_dev *idev);
 145
 146 #define IN6_ADDR_HSIZE_SHIFT    8
 147 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 148 /*
 149  *      Configured unicast address hash table
 150  */
 151 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 152 static DEFINE_SPINLOCK(addrconf_hash_lock);
 153
 154 static void addrconf_verify(void);
 155 static void addrconf_verify_rtnl(void);
 156 static void addrconf_verify_work(struct work_struct *);
 157
 158 static struct workqueue_struct *addrconf_wq;
 159 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 160
 161 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 162 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 163
 164 static void addrconf_type_change(struct net_device *dev,
 165                                  unsigned long event);
 166 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 167
 168 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 169                                                   int plen,
 170                                                   const struct net_device *dev,
 171                                                   u32 flags, u32 noflags,
 172                                                   bool no_gw);
 173
 174 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 175 static void addrconf_dad_work(struct work_struct *w);
 176 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 177                                    bool send_na);
 178 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 179 static void addrconf_rs_timer(struct timer_list *t);
 180 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 181 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182
 183 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 184                                 struct prefix_info *pinfo);
 185
 186 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 187         .forwarding             = 0,
 188         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 189         .mtu6                   = IPV6_MIN_MTU,
 190         .accept_ra              = 1,
 191         .accept_redirects       = 1,
 192         .autoconf               = 1,
 193         .force_mld_version      = 0,
 194         .mldv1_unsolicited_report_interval = 10 * HZ,
 195         .mldv2_unsolicited_report_interval = HZ,
 196         .dad_transmits          = 1,
 197         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 198         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 199         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 201         .use_tempaddr           = 0,
 202         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 203         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 204         .regen_max_retry        = REGEN_MAX_RETRY,
 205         .max_desync_factor      = MAX_DESYNC_FACTOR,
 206         .max_addresses          = IPV6_MAX_ADDRESSES,
 207         .accept_ra_defrtr       = 1,
 208         .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 /**
1987  * ipv6_dev_find - find the first device with a given source address.
1988  * @net: the net namespace
1989  * @addr: the source address
1990  *
1991  * The caller should be protected by RCU, or RTNL.
1992  */
1993 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr)
1994 {
1995         unsigned int hash = inet6_addr_hash(net, addr);
1996         struct inet6_ifaddr *ifp, *result = NULL;
1997         struct net_device *dev = NULL;
1998
1999         rcu_read_lock();
2000         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2001                 if (net_eq(dev_net(ifp->idev->dev), net) &&
2002                     ipv6_addr_equal(&ifp->addr, addr)) {
2003                         result = ifp;
2004                         break;
2005                 }
2006         }
2007
2008         if (!result) {
2009                 struct rt6_info *rt;
2010
2011                 rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
2012                 if (rt) {
2013                         dev = rt->dst.dev;
2014                         ip6_rt_put(rt);
2015                 }
2016         } else {
2017                 dev = result->idev->dev;
2018         }
2019         rcu_read_unlock();
2020
2021         return dev;
2022 }
2023 EXPORT_SYMBOL(ipv6_dev_find);
2024
2025 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2026                                      struct net_device *dev, int strict)
2027 {
2028         unsigned int hash = inet6_addr_hash(net, addr);
2029         struct inet6_ifaddr *ifp, *result = NULL;
2030
2031         rcu_read_lock();
2032         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2033                 if (!net_eq(dev_net(ifp->idev->dev), net))
2034                         continue;
2035                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2036                         if (!dev || ifp->idev->dev == dev ||
2037                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2038                                 result = ifp;
2039                                 in6_ifa_hold(ifp);
2040                                 break;
2041                         }
2042                 }
2043         }
2044         rcu_read_unlock();
2045
2046         return result;
2047 }
2048
2049 /* Gets referenced address, destroys ifaddr */
2050
2051 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2052 {
2053         if (dad_failed)
2054                 ifp->flags |= IFA_F_DADFAILED;
2055
2056         if (ifp->flags&IFA_F_TEMPORARY) {
2057                 struct inet6_ifaddr *ifpub;
2058                 spin_lock_bh(&ifp->lock);
2059                 ifpub = ifp->ifpub;
2060                 if (ifpub) {
2061                         in6_ifa_hold(ifpub);
2062                         spin_unlock_bh(&ifp->lock);
2063                         ipv6_create_tempaddr(ifpub, true);
2064                         in6_ifa_put(ifpub);
2065                 } else {
2066                         spin_unlock_bh(&ifp->lock);
2067                 }
2068                 ipv6_del_addr(ifp);
2069         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2070                 spin_lock_bh(&ifp->lock);
2071                 addrconf_del_dad_work(ifp);
2072                 ifp->flags |= IFA_F_TENTATIVE;
2073                 if (dad_failed)
2074                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2075                 spin_unlock_bh(&ifp->lock);
2076                 if (dad_failed)
2077                         ipv6_ifa_notify(0, ifp);
2078                 in6_ifa_put(ifp);
2079         } else {
2080                 ipv6_del_addr(ifp);
2081         }
2082 }
2083
2084 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2085 {
2086         int err = -ENOENT;
2087
2088         spin_lock_bh(&ifp->lock);
2089         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2090                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2091                 err = 0;
2092         }
2093         spin_unlock_bh(&ifp->lock);
2094
2095         return err;
2096 }
2097
2098 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2099 {
2100         struct inet6_dev *idev = ifp->idev;
2101         struct net *net = dev_net(ifp->idev->dev);
2102
2103         if (addrconf_dad_end(ifp)) {
2104                 in6_ifa_put(ifp);
2105                 return;
2106         }
2107
2108         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2109                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2110
2111         spin_lock_bh(&ifp->lock);
2112
2113         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2114                 struct in6_addr new_addr;
2115                 struct inet6_ifaddr *ifp2;
2116                 int retries = ifp->stable_privacy_retry + 1;
2117                 struct ifa6_config cfg = {
2118                         .pfx = &new_addr,
2119                         .plen = ifp->prefix_len,
2120                         .ifa_flags = ifp->flags,
2121                         .valid_lft = ifp->valid_lft,
2122                         .preferred_lft = ifp->prefered_lft,
2123                         .scope = ifp->scope,
2124                 };
2125
2126                 if (retries > net->ipv6.sysctl.idgen_retries) {
2127                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2128                                              ifp->idev->dev->name);
2129                         goto errdad;
2130                 }
2131
2132                 new_addr = ifp->addr;
2133                 if (ipv6_generate_stable_address(&new_addr, retries,
2134                                                  idev))
2135                         goto errdad;
2136
2137                 spin_unlock_bh(&ifp->lock);
2138
2139                 if (idev->cnf.max_addresses &&
2140                     ipv6_count_addresses(idev) >=
2141                     idev->cnf.max_addresses)
2142                         goto lock_errdad;
2143
2144                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2145                                      ifp->idev->dev->name);
2146
2147                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2148                 if (IS_ERR(ifp2))
2149                         goto lock_errdad;
2150
2151                 spin_lock_bh(&ifp2->lock);
2152                 ifp2->stable_privacy_retry = retries;
2153                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2154                 spin_unlock_bh(&ifp2->lock);
2155
2156                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2157                 in6_ifa_put(ifp2);
2158 lock_errdad:
2159                 spin_lock_bh(&ifp->lock);
2160         }
2161
2162 errdad:
2163         /* transition from _POSTDAD to _ERRDAD */
2164         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2165         spin_unlock_bh(&ifp->lock);
2166
2167         addrconf_mod_dad_work(ifp, 0);
2168         in6_ifa_put(ifp);
2169 }
2170
2171 /* Join to solicited addr multicast group.
2172  * caller must hold RTNL */
2173 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2174 {
2175         struct in6_addr maddr;
2176
2177         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2178                 return;
2179
2180         addrconf_addr_solict_mult(addr, &maddr);
2181         ipv6_dev_mc_inc(dev, &maddr);
2182 }
2183
2184 /* caller must hold RTNL */
2185 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2186 {
2187         struct in6_addr maddr;
2188
2189         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2190                 return;
2191
2192         addrconf_addr_solict_mult(addr, &maddr);
2193         __ipv6_dev_mc_dec(idev, &maddr);
2194 }
2195
2196 /* caller must hold RTNL */
2197 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2198 {
2199         struct in6_addr addr;
2200
2201         if (ifp->prefix_len >= 127) /* RFC 6164 */
2202                 return;
2203         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2204         if (ipv6_addr_any(&addr))
2205                 return;
2206         __ipv6_dev_ac_inc(ifp->idev, &addr);
2207 }
2208
2209 /* caller must hold RTNL */
2210 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2211 {
2212         struct in6_addr addr;
2213
2214         if (ifp->prefix_len >= 127) /* RFC 6164 */
2215                 return;
2216         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2217         if (ipv6_addr_any(&addr))
2218                 return;
2219         __ipv6_dev_ac_dec(ifp->idev, &addr);
2220 }
2221
2222 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2223 {
2224         switch (dev->addr_len) {
2225         case ETH_ALEN:
2226                 memcpy(eui, dev->dev_addr, 3);
2227                 eui[3] = 0xFF;
2228                 eui[4] = 0xFE;
2229                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2230                 break;
2231         case EUI64_ADDR_LEN:
2232                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2233                 eui[0] ^= 2;
2234                 break;
2235         default:
2236                 return -1;
2237         }
2238
2239         return 0;
2240 }
2241
2242 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2243 {
2244         union fwnet_hwaddr *ha;
2245
2246         if (dev->addr_len != FWNET_ALEN)
2247                 return -1;
2248
2249         ha = (union fwnet_hwaddr *)dev->dev_addr;
2250
2251         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2252         eui[0] ^= 2;
2253         return 0;
2254 }
2255
2256 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2257 {
2258         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2259         if (dev->addr_len != ARCNET_ALEN)
2260                 return -1;
2261         memset(eui, 0, 7);
2262         eui[7] = *(u8 *)dev->dev_addr;
2263         return 0;
2264 }
2265
2266 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2267 {
2268         if (dev->addr_len != INFINIBAND_ALEN)
2269                 return -1;
2270         memcpy(eui, dev->dev_addr + 12, 8);
2271         eui[0] |= 2;
2272         return 0;
2273 }
2274
2275 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2276 {
2277         if (addr == 0)
2278                 return -1;
2279         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2280                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2281                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2282                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2283                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2284                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2285         eui[1] = 0;
2286         eui[2] = 0x5E;
2287         eui[3] = 0xFE;
2288         memcpy(eui + 4, &addr, 4);
2289         return 0;
2290 }
2291
2292 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2293 {
2294         if (dev->priv_flags & IFF_ISATAP)
2295                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2296         return -1;
2297 }
2298
2299 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2300 {
2301         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2302 }
2303
2304 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2305 {
2306         memcpy(eui, dev->perm_addr, 3);
2307         memcpy(eui + 5, dev->perm_addr + 3, 3);
2308         eui[3] = 0xFF;
2309         eui[4] = 0xFE;
2310         eui[0] ^= 2;
2311         return 0;
2312 }
2313
2314 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2315 {
2316         switch (dev->type) {
2317         case ARPHRD_ETHER:
2318         case ARPHRD_FDDI:
2319                 return addrconf_ifid_eui48(eui, dev);
2320         case ARPHRD_ARCNET:
2321                 return addrconf_ifid_arcnet(eui, dev);
2322         case ARPHRD_INFINIBAND:
2323                 return addrconf_ifid_infiniband(eui, dev);
2324         case ARPHRD_SIT:
2325                 return addrconf_ifid_sit(eui, dev);
2326         case ARPHRD_IPGRE:
2327         case ARPHRD_TUNNEL:
2328                 return addrconf_ifid_gre(eui, dev);
2329         case ARPHRD_6LOWPAN:
2330                 return addrconf_ifid_6lowpan(eui, dev);
2331         case ARPHRD_IEEE1394:
2332                 return addrconf_ifid_ieee1394(eui, dev);
2333         case ARPHRD_TUNNEL6:
2334         case ARPHRD_IP6GRE:
2335         case ARPHRD_RAWIP:
2336                 return addrconf_ifid_ip6tnl(eui, dev);
2337         }
2338         return -1;
2339 }
2340
2341 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2342 {
2343         int err = -1;
2344         struct inet6_ifaddr *ifp;
2345
2346         read_lock_bh(&idev->lock);
2347         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2348                 if (ifp->scope > IFA_LINK)
2349                         break;
2350                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2351                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2352                         err = 0;
2353                         break;
2354                 }
2355         }
2356         read_unlock_bh(&idev->lock);
2357         return err;
2358 }
2359
2360 /* Generation of a randomized Interface Identifier
2361  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2362  */
2363
2364 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2365 {
2366 regen:
2367         get_random_bytes(&addr->s6_addr[8], 8);
2368
2369         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2370          * check if generated address is not inappropriate:
2371          *
2372          * - Reserved IPv6 Interface Identifers
2373          * - XXX: already assigned to an address on the device
2374          */
2375
2376         /* Subnet-router anycast: 0000:0000:0000:0000 */
2377         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2378                 goto regen;
2379
2380         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2381          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2382          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2383          */
2384         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2385             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2386                 goto regen;
2387
2388         /* Reserved subnet anycast addresses */
2389         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2390             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2391                 goto regen;
2392 }
2393
2394 /*
2395  *      Add prefix route.
2396  */
2397
2398 static void
2399 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2400                       struct net_device *dev, unsigned long expires,
2401                       u32 flags, gfp_t gfp_flags)
2402 {
2403         struct fib6_config cfg = {
2404                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2405                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2406                 .fc_ifindex = dev->ifindex,
2407                 .fc_expires = expires,
2408                 .fc_dst_len = plen,
2409                 .fc_flags = RTF_UP | flags,
2410                 .fc_nlinfo.nl_net = dev_net(dev),
2411                 .fc_protocol = RTPROT_KERNEL,
2412                 .fc_type = RTN_UNICAST,
2413         };
2414
2415         cfg.fc_dst = *pfx;
2416
2417         /* Prevent useless cloning on PtP SIT.
2418            This thing is done here expecting that the whole
2419            class of non-broadcast devices need not cloning.
2420          */
2421 #if IS_ENABLED(CONFIG_IPV6_SIT)
2422         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2423                 cfg.fc_flags |= RTF_NONEXTHOP;
2424 #endif
2425
2426         ip6_route_add(&cfg, gfp_flags, NULL);
2427 }
2428
2429
2430 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2431                                                   int plen,
2432                                                   const struct net_device *dev,
2433                                                   u32 flags, u32 noflags,
2434                                                   bool no_gw)
2435 {
2436         struct fib6_node *fn;
2437         struct fib6_info *rt = NULL;
2438         struct fib6_table *table;
2439         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2440
2441         table = fib6_get_table(dev_net(dev), tb_id);
2442         if (!table)
2443                 return NULL;
2444
2445         rcu_read_lock();
2446         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2447         if (!fn)
2448                 goto out;
2449
2450         for_each_fib6_node_rt_rcu(fn) {
2451                 /* prefix routes only use builtin fib6_nh */
2452                 if (rt->nh)
2453                         continue;
2454
2455                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2456                         continue;
2457                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2458                         continue;
2459                 if ((rt->fib6_flags & flags) != flags)
2460                         continue;
2461                 if ((rt->fib6_flags & noflags) != 0)
2462                         continue;
2463                 if (!fib6_info_hold_safe(rt))
2464                         continue;
2465                 break;
2466         }
2467 out:
2468         rcu_read_unlock();
2469         return rt;
2470 }
2471
2472
2473 /* Create "default" multicast route to the interface */
2474
2475 static void addrconf_add_mroute(struct net_device *dev)
2476 {
2477         struct fib6_config cfg = {
2478                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2479                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2480                 .fc_ifindex = dev->ifindex,
2481                 .fc_dst_len = 8,
2482                 .fc_flags = RTF_UP,
2483                 .fc_type = RTN_UNICAST,
2484                 .fc_nlinfo.nl_net = dev_net(dev),
2485         };
2486
2487         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2488
2489         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2490 }
2491
2492 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2493 {
2494         struct inet6_dev *idev;
2495
2496         ASSERT_RTNL();
2497
2498         idev = ipv6_find_idev(dev);
2499         if (IS_ERR(idev))
2500                 return idev;
2501
2502         if (idev->cnf.disable_ipv6)
2503                 return ERR_PTR(-EACCES);
2504
2505         /* Add default multicast route */
2506         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2507                 addrconf_add_mroute(dev);
2508
2509         return idev;
2510 }
2511
2512 static void manage_tempaddrs(struct inet6_dev *idev,
2513                              struct inet6_ifaddr *ifp,
2514                              __u32 valid_lft, __u32 prefered_lft,
2515                              bool create, unsigned long now)
2516 {
2517         u32 flags;
2518         struct inet6_ifaddr *ift;
2519
2520         read_lock_bh(&idev->lock);
2521         /* update all temporary addresses in the list */
2522         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2523                 int age, max_valid, max_prefered;
2524
2525                 if (ifp != ift->ifpub)
2526                         continue;
2527
2528                 /* RFC 4941 section 3.3:
2529                  * If a received option will extend the lifetime of a public
2530                  * address, the lifetimes of temporary addresses should
2531                  * be extended, subject to the overall constraint that no
2532                  * temporary addresses should ever remain "valid" or "preferred"
2533                  * for a time longer than (TEMP_VALID_LIFETIME) or
2534                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2535                  */
2536                 age = (now - ift->cstamp) / HZ;
2537                 max_valid = idev->cnf.temp_valid_lft - age;
2538                 if (max_valid < 0)
2539                         max_valid = 0;
2540
2541                 max_prefered = idev->cnf.temp_prefered_lft -
2542                                idev->desync_factor - age;
2543                 if (max_prefered < 0)
2544                         max_prefered = 0;
2545
2546                 if (valid_lft > max_valid)
2547                         valid_lft = max_valid;
2548
2549                 if (prefered_lft > max_prefered)
2550                         prefered_lft = max_prefered;
2551
2552                 spin_lock(&ift->lock);
2553                 flags = ift->flags;
2554                 ift->valid_lft = valid_lft;
2555                 ift->prefered_lft = prefered_lft;
2556                 ift->tstamp = now;
2557                 if (prefered_lft > 0)
2558                         ift->flags &= ~IFA_F_DEPRECATED;
2559
2560                 spin_unlock(&ift->lock);
2561                 if (!(flags&IFA_F_TENTATIVE))
2562                         ipv6_ifa_notify(0, ift);
2563         }
2564
2565         if ((create || list_empty(&idev->tempaddr_list)) &&
2566             idev->cnf.use_tempaddr > 0) {
2567                 /* When a new public address is created as described
2568                  * in [ADDRCONF], also create a new temporary address.
2569                  * Also create a temporary address if it's enabled but
2570                  * no temporary address currently exists.
2571                  */
2572                 read_unlock_bh(&idev->lock);
2573                 ipv6_create_tempaddr(ifp, false);
2574         } else {
2575                 read_unlock_bh(&idev->lock);
2576         }
2577 }
2578
2579 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2580 {
2581         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2582                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2583 }
2584
2585 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2586                                  const struct prefix_info *pinfo,
2587                                  struct inet6_dev *in6_dev,
2588                                  const struct in6_addr *addr, int addr_type,
2589                                  u32 addr_flags, bool sllao, bool tokenized,
2590                                  __u32 valid_lft, u32 prefered_lft)
2591 {
2592         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2593         int create = 0;
2594
2595         if (!ifp && valid_lft) {
2596                 int max_addresses = in6_dev->cnf.max_addresses;
2597                 struct ifa6_config cfg = {
2598                         .pfx = addr,
2599                         .plen = pinfo->prefix_len,
2600                         .ifa_flags = addr_flags,
2601                         .valid_lft = valid_lft,
2602                         .preferred_lft = prefered_lft,
2603                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2604                 };
2605
2606 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2607                 if ((net->ipv6.devconf_all->optimistic_dad ||
2608                      in6_dev->cnf.optimistic_dad) &&
2609                     !net->ipv6.devconf_all->forwarding && sllao)
2610                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2611 #endif
2612
2613                 /* Do not allow to create too much of autoconfigured
2614                  * addresses; this would be too easy way to crash kernel.
2615                  */
2616                 if (!max_addresses ||
2617                     ipv6_count_addresses(in6_dev) < max_addresses)
2618                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2619
2620                 if (IS_ERR_OR_NULL(ifp))
2621                         return -1;
2622
2623                 create = 1;
2624                 spin_lock_bh(&ifp->lock);
2625                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2626                 ifp->cstamp = jiffies;
2627                 ifp->tokenized = tokenized;
2628                 spin_unlock_bh(&ifp->lock);
2629                 addrconf_dad_start(ifp);
2630         }
2631
2632         if (ifp) {
2633                 u32 flags;
2634                 unsigned long now;
2635                 u32 stored_lft;
2636
2637                 /* Update lifetime (RFC4862 5.5.3 e)
2638                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2639                  * improve the reaction of SLAAC to renumbering events
2640                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2641                  */
2642                 spin_lock_bh(&ifp->lock);
2643                 now = jiffies;
2644                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2645                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2646                 else
2647                         stored_lft = 0;
2648
2649                 if (!create && stored_lft) {
2650                         ifp->valid_lft = valid_lft;
2651                         ifp->prefered_lft = prefered_lft;
2652                         ifp->tstamp = now;
2653                         flags = ifp->flags;
2654                         ifp->flags &= ~IFA_F_DEPRECATED;
2655                         spin_unlock_bh(&ifp->lock);
2656
2657                         if (!(flags&IFA_F_TENTATIVE))
2658                                 ipv6_ifa_notify(0, ifp);
2659                 } else
2660                         spin_unlock_bh(&ifp->lock);
2661
2662                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2663                                  create, now);
2664
2665                 in6_ifa_put(ifp);
2666                 addrconf_verify();
2667         }
2668
2669         return 0;
2670 }
2671 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2672
2673 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2674 {
2675         struct prefix_info *pinfo;
2676         __u32 valid_lft;
2677         __u32 prefered_lft;
2678         int addr_type, err;
2679         u32 addr_flags = 0;
2680         struct inet6_dev *in6_dev;
2681         struct net *net = dev_net(dev);
2682
2683         pinfo = (struct prefix_info *) opt;
2684
2685         if (len < sizeof(struct prefix_info)) {
2686                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2687                 return;
2688         }
2689
2690         /*
2691          *      Validation checks ([ADDRCONF], page 19)
2692          */
2693
2694         addr_type = ipv6_addr_type(&pinfo->prefix);
2695
2696         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2697                 return;
2698
2699         valid_lft = ntohl(pinfo->valid);
2700         prefered_lft = ntohl(pinfo->prefered);
2701
2702         if (prefered_lft > valid_lft) {
2703                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2704                 return;
2705         }
2706
2707         in6_dev = in6_dev_get(dev);
2708
2709         if (!in6_dev) {
2710                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2711                                     dev->name);
2712                 return;
2713         }
2714
2715         /*
2716          *      Two things going on here:
2717          *      1) Add routes for on-link prefixes
2718          *      2) Configure prefixes with the auto flag set
2719          */
2720
2721         if (pinfo->onlink) {
2722                 struct fib6_info *rt;
2723                 unsigned long rt_expires;
2724
2725                 /* Avoid arithmetic overflow. Really, we could
2726                  * save rt_expires in seconds, likely valid_lft,
2727                  * but it would require division in fib gc, that it
2728                  * not good.
2729                  */
2730                 if (HZ > USER_HZ)
2731                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2732                 else
2733                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2734
2735                 if (addrconf_finite_timeout(rt_expires))
2736                         rt_expires *= HZ;
2737
2738                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2739                                                pinfo->prefix_len,
2740                                                dev,
2741                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2742                                                RTF_DEFAULT, true);
2743
2744                 if (rt) {
2745                         /* Autoconf prefix route */
2746                         if (valid_lft == 0) {
2747                                 ip6_del_rt(net, rt, false);
2748                                 rt = NULL;
2749                         } else if (addrconf_finite_timeout(rt_expires)) {
2750                                 /* not infinity */
2751                                 fib6_set_expires(rt, jiffies + rt_expires);
2752                         } else {
2753                                 fib6_clean_expires(rt);
2754                         }
2755                 } else if (valid_lft) {
2756                         clock_t expires = 0;
2757                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2758                         if (addrconf_finite_timeout(rt_expires)) {
2759                                 /* not infinity */
2760                                 flags |= RTF_EXPIRES;
2761                                 expires = jiffies_to_clock_t(rt_expires);
2762                         }
2763                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2764                                               0, dev, expires, flags,
2765                                               GFP_ATOMIC);
2766                 }
2767                 fib6_info_release(rt);
2768         }
2769
2770         /* Try to figure out our local address for this prefix */
2771
2772         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2773                 struct in6_addr addr;
2774                 bool tokenized = false, dev_addr_generated = false;
2775
2776                 if (pinfo->prefix_len == 64) {
2777                         memcpy(&addr, &pinfo->prefix, 8);
2778
2779                         if (!ipv6_addr_any(&in6_dev->token)) {
2780                                 read_lock_bh(&in6_dev->lock);
2781                                 memcpy(addr.s6_addr + 8,
2782                                        in6_dev->token.s6_addr + 8, 8);
2783                                 read_unlock_bh(&in6_dev->lock);
2784                                 tokenized = true;
2785                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2786                                    !ipv6_generate_stable_address(&addr, 0,
2787                                                                  in6_dev)) {
2788                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2789                                 goto ok;
2790                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2791                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2792                                 goto put;
2793                         } else {
2794                                 dev_addr_generated = true;
2795                         }
2796                         goto ok;
2797                 }
2798                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2799                                     pinfo->prefix_len);
2800                 goto put;
2801
2802 ok:
2803                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2804                                                    &addr, addr_type,
2805                                                    addr_flags, sllao,
2806                                                    tokenized, valid_lft,
2807                                                    prefered_lft);
2808                 if (err)
2809                         goto put;
2810
2811                 /* Ignore error case here because previous prefix add addr was
2812                  * successful which will be notified.
2813                  */
2814                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2815                                               addr_type, addr_flags, sllao,
2816                                               tokenized, valid_lft,
2817                                               prefered_lft,
2818                                               dev_addr_generated);
2819         }
2820         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2821 put:
2822         in6_dev_put(in6_dev);
2823 }
2824
2825 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2826                 struct in6_ifreq *ireq)
2827 {
2828         struct ip_tunnel_parm p = { };
2829         int err;
2830
2831         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2832                 return -EADDRNOTAVAIL;
2833
2834         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2835         p.iph.version = 4;
2836         p.iph.ihl = 5;
2837         p.iph.protocol = IPPROTO_IPV6;
2838         p.iph.ttl = 64;
2839
2840         if (!dev->netdev_ops->ndo_tunnel_ctl)
2841                 return -EOPNOTSUPP;
2842         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2843         if (err)
2844                 return err;
2845
2846         dev = __dev_get_by_name(net, p.name);
2847         if (!dev)
2848                 return -ENOBUFS;
2849         return dev_open(dev, NULL);
2850 }
2851
2852 /*
2853  *      Set destination address.
2854  *      Special case for SIT interfaces where we create a new "virtual"
2855  *      device.
2856  */
2857 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2858 {
2859         struct net_device *dev;
2860         struct in6_ifreq ireq;
2861         int err = -ENODEV;
2862
2863         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2864                 return -ENODEV;
2865         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2866                 return -EFAULT;
2867
2868         rtnl_lock();
2869         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2870         if (dev && dev->type == ARPHRD_SIT)
2871                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2872         rtnl_unlock();
2873         return err;
2874 }
2875
2876 static int ipv6_mc_config(struct sock *sk, bool join,
2877                           const struct in6_addr *addr, int ifindex)
2878 {
2879         int ret;
2880
2881         ASSERT_RTNL();
2882
2883         lock_sock(sk);
2884         if (join)
2885                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2886         else
2887                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2888         release_sock(sk);
2889
2890         return ret;
2891 }
2892
2893 /*
2894  *      Manual configuration of address on an interface
2895  */
2896 static int inet6_addr_add(struct net *net, int ifindex,
2897                           struct ifa6_config *cfg,
2898                           struct netlink_ext_ack *extack)
2899 {
2900         struct inet6_ifaddr *ifp;
2901         struct inet6_dev *idev;
2902         struct net_device *dev;
2903         unsigned long timeout;
2904         clock_t expires;
2905         u32 flags;
2906
2907         ASSERT_RTNL();
2908
2909         if (cfg->plen > 128)
2910                 return -EINVAL;
2911
2912         /* check the lifetime */
2913         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2914                 return -EINVAL;
2915
2916         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2917                 return -EINVAL;
2918
2919         dev = __dev_get_by_index(net, ifindex);
2920         if (!dev)
2921                 return -ENODEV;
2922
2923         idev = addrconf_add_dev(dev);
2924         if (IS_ERR(idev))
2925                 return PTR_ERR(idev);
2926
2927         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2928                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2929                                          true, cfg->pfx, ifindex);
2930
2931                 if (ret < 0)
2932                         return ret;
2933         }
2934
2935         cfg->scope = ipv6_addr_scope(cfg->pfx);
2936
2937         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2938         if (addrconf_finite_timeout(timeout)) {
2939                 expires = jiffies_to_clock_t(timeout * HZ);
2940                 cfg->valid_lft = timeout;
2941                 flags = RTF_EXPIRES;
2942         } else {
2943                 expires = 0;
2944                 flags = 0;
2945                 cfg->ifa_flags |= IFA_F_PERMANENT;
2946         }
2947
2948         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2949         if (addrconf_finite_timeout(timeout)) {
2950                 if (timeout == 0)
2951                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2952                 cfg->preferred_lft = timeout;
2953         }
2954
2955         ifp = ipv6_add_addr(idev, cfg, true, extack);
2956         if (!IS_ERR(ifp)) {
2957                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2958                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2959                                               ifp->rt_priority, dev, expires,
2960                                               flags, GFP_KERNEL);
2961                 }
2962
2963                 /* Send a netlink notification if DAD is enabled and
2964                  * optimistic flag is not set
2965                  */
2966                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2967                         ipv6_ifa_notify(0, ifp);
2968                 /*
2969                  * Note that section 3.1 of RFC 4429 indicates
2970                  * that the Optimistic flag should not be set for
2971                  * manually configured addresses
2972                  */
2973                 addrconf_dad_start(ifp);
2974                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2975                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2976                                          cfg->preferred_lft, true, jiffies);
2977                 in6_ifa_put(ifp);
2978                 addrconf_verify_rtnl();
2979                 return 0;
2980         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2981                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2982                                cfg->pfx, ifindex);
2983         }
2984
2985         return PTR_ERR(ifp);
2986 }
2987
2988 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2989                           const struct in6_addr *pfx, unsigned int plen)
2990 {
2991         struct inet6_ifaddr *ifp;
2992         struct inet6_dev *idev;
2993         struct net_device *dev;
2994
2995         if (plen > 128)
2996                 return -EINVAL;
2997
2998         dev = __dev_get_by_index(net, ifindex);
2999         if (!dev)
3000                 return -ENODEV;
3001
3002         idev = __in6_dev_get(dev);
3003         if (!idev)
3004                 return -ENXIO;
3005
3006         read_lock_bh(&idev->lock);
3007         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3008                 if (ifp->prefix_len == plen &&
3009                     ipv6_addr_equal(pfx, &ifp->addr)) {
3010                         in6_ifa_hold(ifp);
3011                         read_unlock_bh(&idev->lock);
3012
3013                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3014                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3015                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3016                                                  jiffies);
3017                         ipv6_del_addr(ifp);
3018                         addrconf_verify_rtnl();
3019                         if (ipv6_addr_is_multicast(pfx)) {
3020                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3021                                                false, pfx, dev->ifindex);
3022                         }
3023                         return 0;
3024                 }
3025         }
3026         read_unlock_bh(&idev->lock);
3027         return -EADDRNOTAVAIL;
3028 }
3029
3030
3031 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3032 {
3033         struct ifa6_config cfg = {
3034                 .ifa_flags = IFA_F_PERMANENT,
3035                 .preferred_lft = INFINITY_LIFE_TIME,
3036                 .valid_lft = INFINITY_LIFE_TIME,
3037         };
3038         struct in6_ifreq ireq;
3039         int err;
3040
3041         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3042                 return -EPERM;
3043
3044         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3045                 return -EFAULT;
3046
3047         cfg.pfx = &ireq.ifr6_addr;
3048         cfg.plen = ireq.ifr6_prefixlen;
3049
3050         rtnl_lock();
3051         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3052         rtnl_unlock();
3053         return err;
3054 }
3055
3056 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3057 {
3058         struct in6_ifreq ireq;
3059         int err;
3060
3061         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3062                 return -EPERM;
3063
3064         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3065                 return -EFAULT;
3066
3067         rtnl_lock();
3068         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3069                              ireq.ifr6_prefixlen);
3070         rtnl_unlock();
3071         return err;
3072 }
3073
3074 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3075                      int plen, int scope)
3076 {
3077         struct inet6_ifaddr *ifp;
3078         struct ifa6_config cfg = {
3079                 .pfx = addr,
3080                 .plen = plen,
3081                 .ifa_flags = IFA_F_PERMANENT,
3082                 .valid_lft = INFINITY_LIFE_TIME,
3083                 .preferred_lft = INFINITY_LIFE_TIME,
3084                 .scope = scope
3085         };
3086
3087         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3088         if (!IS_ERR(ifp)) {
3089                 spin_lock_bh(&ifp->lock);
3090                 ifp->flags &= ~IFA_F_TENTATIVE;
3091                 spin_unlock_bh(&ifp->lock);
3092                 rt_genid_bump_ipv6(dev_net(idev->dev));
3093                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3094                 in6_ifa_put(ifp);
3095         }
3096 }
3097
3098 #if IS_ENABLED(CONFIG_IPV6_SIT)
3099 static void sit_add_v4_addrs(struct inet6_dev *idev)
3100 {
3101         struct in6_addr addr;
3102         struct net_device *dev;
3103         struct net *net = dev_net(idev->dev);
3104         int scope, plen;
3105         u32 pflags = 0;
3106
3107         ASSERT_RTNL();
3108
3109         memset(&addr, 0, sizeof(struct in6_addr));
3110         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3111
3112         if (idev->dev->flags&IFF_POINTOPOINT) {
3113                 addr.s6_addr32[0] = htonl(0xfe800000);
3114                 scope = IFA_LINK;
3115                 plen = 64;
3116         } else {
3117                 scope = IPV6_ADDR_COMPATv4;
3118                 plen = 96;
3119                 pflags |= RTF_NONEXTHOP;
3120         }
3121
3122         if (addr.s6_addr32[3]) {
3123                 add_addr(idev, &addr, plen, scope);
3124                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3125                                       GFP_KERNEL);
3126                 return;
3127         }
3128
3129         for_each_netdev(net, dev) {
3130                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3131                 if (in_dev && (dev->flags & IFF_UP)) {
3132                         struct in_ifaddr *ifa;
3133                         int flag = scope;
3134
3135                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3136                                 addr.s6_addr32[3] = ifa->ifa_local;
3137
3138                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3139                                         continue;
3140                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3141                                         if (idev->dev->flags&IFF_POINTOPOINT)
3142                                                 continue;
3143                                         flag |= IFA_HOST;
3144                                 }
3145
3146                                 add_addr(idev, &addr, plen, flag);
3147                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3148                                                       0, pflags, GFP_KERNEL);
3149                         }
3150                 }
3151         }
3152 }
3153 #endif
3154
3155 static void init_loopback(struct net_device *dev)
3156 {
3157         struct inet6_dev  *idev;
3158
3159         /* ::1 */
3160
3161         ASSERT_RTNL();
3162
3163         idev = ipv6_find_idev(dev);
3164         if (IS_ERR(idev)) {
3165                 pr_debug("%s: add_dev failed\n", __func__);
3166                 return;
3167         }
3168
3169         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3170 }
3171
3172 void addrconf_add_linklocal(struct inet6_dev *idev,
3173                             const struct in6_addr *addr, u32 flags)
3174 {
3175         struct ifa6_config cfg = {
3176                 .pfx = addr,
3177                 .plen = 64,
3178                 .ifa_flags = flags | IFA_F_PERMANENT,
3179                 .valid_lft = INFINITY_LIFE_TIME,
3180                 .preferred_lft = INFINITY_LIFE_TIME,
3181                 .scope = IFA_LINK
3182         };
3183         struct inet6_ifaddr *ifp;
3184
3185 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3186         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3187              idev->cnf.optimistic_dad) &&
3188             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3189                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3190 #endif
3191
3192         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3193         if (!IS_ERR(ifp)) {
3194                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3195                                       0, 0, GFP_ATOMIC);
3196                 addrconf_dad_start(ifp);
3197                 in6_ifa_put(ifp);
3198         }
3199 }
3200 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3201
3202 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3203 {
3204         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3205                 return true;
3206
3207         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3208             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3209                 return true;
3210
3211         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3212             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3213                 return true;
3214
3215         return false;
3216 }
3217
3218 static int ipv6_generate_stable_address(struct in6_addr *address,
3219                                         u8 dad_count,
3220                                         const struct inet6_dev *idev)
3221 {
3222         static DEFINE_SPINLOCK(lock);
3223         static __u32 digest[SHA1_DIGEST_WORDS];
3224         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3225
3226         static union {
3227                 char __data[SHA1_BLOCK_SIZE];
3228                 struct {
3229                         struct in6_addr secret;
3230                         __be32 prefix[2];
3231                         unsigned char hwaddr[MAX_ADDR_LEN];
3232                         u8 dad_count;
3233                 } __packed;
3234         } data;
3235
3236         struct in6_addr secret;
3237         struct in6_addr temp;
3238         struct net *net = dev_net(idev->dev);
3239
3240         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3241
3242         if (idev->cnf.stable_secret.initialized)
3243                 secret = idev->cnf.stable_secret.secret;
3244         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3245                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3246         else
3247                 return -1;
3248
3249 retry:
3250         spin_lock_bh(&lock);
3251
3252         sha1_init(digest);
3253         memset(&data, 0, sizeof(data));
3254         memset(workspace, 0, sizeof(workspace));
3255         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3256         data.prefix[0] = address->s6_addr32[0];
3257         data.prefix[1] = address->s6_addr32[1];
3258         data.secret = secret;
3259         data.dad_count = dad_count;
3260
3261         sha1_transform(digest, data.__data, workspace);
3262
3263         temp = *address;
3264         temp.s6_addr32[2] = (__force __be32)digest[0];
3265         temp.s6_addr32[3] = (__force __be32)digest[1];
3266
3267         spin_unlock_bh(&lock);
3268
3269         if (ipv6_reserved_interfaceid(temp)) {
3270                 dad_count++;
3271                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3272                         return -1;
3273                 goto retry;
3274         }
3275
3276         *address = temp;
3277         return 0;
3278 }
3279
3280 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3281 {
3282         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3283
3284         if (s->initialized)
3285                 return;
3286         s = &idev->cnf.stable_secret;
3287         get_random_bytes(&s->secret, sizeof(s->secret));
3288         s->initialized = true;
3289 }
3290
3291 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3292 {
3293         struct in6_addr addr;
3294
3295         /* no link local addresses on L3 master devices */
3296         if (netif_is_l3_master(idev->dev))
3297                 return;
3298
3299         /* no link local addresses on devices flagged as slaves */
3300         if (idev->dev->flags & IFF_SLAVE)
3301                 return;
3302
3303         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3304
3305         switch (idev->cnf.addr_gen_mode) {
3306         case IN6_ADDR_GEN_MODE_RANDOM:
3307                 ipv6_gen_mode_random_init(idev);
3308                 fallthrough;
3309         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3310                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3311                         addrconf_add_linklocal(idev, &addr,
3312                                                IFA_F_STABLE_PRIVACY);
3313                 else if (prefix_route)
3314                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3315                                               0, 0, GFP_KERNEL);
3316                 break;
3317         case IN6_ADDR_GEN_MODE_EUI64:
3318                 /* addrconf_add_linklocal also adds a prefix_route and we
3319                  * only need to care about prefix routes if ipv6_generate_eui64
3320                  * couldn't generate one.
3321                  */
3322                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3323                         addrconf_add_linklocal(idev, &addr, 0);
3324                 else if (prefix_route)
3325                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3326                                               0, 0, GFP_KERNEL);
3327                 break;
3328         case IN6_ADDR_GEN_MODE_NONE:
3329         default:
3330                 /* will not add any link local address */
3331                 break;
3332         }
3333 }
3334
3335 static void addrconf_dev_config(struct net_device *dev)
3336 {
3337         struct inet6_dev *idev;
3338
3339         ASSERT_RTNL();
3340
3341         if ((dev->type != ARPHRD_ETHER) &&
3342             (dev->type != ARPHRD_FDDI) &&
3343             (dev->type != ARPHRD_ARCNET) &&
3344             (dev->type != ARPHRD_INFINIBAND) &&
3345             (dev->type != ARPHRD_IEEE1394) &&
3346             (dev->type != ARPHRD_TUNNEL6) &&
3347             (dev->type != ARPHRD_6LOWPAN) &&
3348             (dev->type != ARPHRD_IP6GRE) &&
3349             (dev->type != ARPHRD_IPGRE) &&
3350             (dev->type != ARPHRD_TUNNEL) &&
3351             (dev->type != ARPHRD_NONE) &&
3352             (dev->type != ARPHRD_RAWIP)) {
3353                 /* Alas, we support only Ethernet autoconfiguration. */
3354                 idev = __in6_dev_get(dev);
3355                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3356                     dev->flags & IFF_MULTICAST)
3357                         ipv6_mc_up(idev);
3358                 return;
3359         }
3360
3361         idev = addrconf_add_dev(dev);
3362         if (IS_ERR(idev))
3363                 return;
3364
3365         /* this device type has no EUI support */
3366         if (dev->type == ARPHRD_NONE &&
3367             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3368                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3369
3370         addrconf_addr_gen(idev, false);
3371 }
3372
3373 #if IS_ENABLED(CONFIG_IPV6_SIT)
3374 static void addrconf_sit_config(struct net_device *dev)
3375 {
3376         struct inet6_dev *idev;
3377
3378         ASSERT_RTNL();
3379
3380         /*
3381          * Configure the tunnel with one of our IPv4
3382          * addresses... we should configure all of
3383          * our v4 addrs in the tunnel
3384          */
3385
3386         idev = ipv6_find_idev(dev);
3387         if (IS_ERR(idev)) {
3388                 pr_debug("%s: add_dev failed\n", __func__);
3389                 return;
3390         }
3391
3392         if (dev->priv_flags & IFF_ISATAP) {
3393                 addrconf_addr_gen(idev, false);
3394                 return;
3395         }
3396
3397         sit_add_v4_addrs(idev);
3398
3399         if (dev->flags&IFF_POINTOPOINT)
3400                 addrconf_add_mroute(dev);
3401 }
3402 #endif
3403
3404 #if IS_ENABLED(CONFIG_NET_IPGRE)
3405 static void addrconf_gre_config(struct net_device *dev)
3406 {
3407         struct inet6_dev *idev;
3408
3409         ASSERT_RTNL();
3410
3411         idev = ipv6_find_idev(dev);
3412         if (IS_ERR(idev)) {
3413                 pr_debug("%s: add_dev failed\n", __func__);
3414                 return;
3415         }
3416
3417         addrconf_addr_gen(idev, true);
3418         if (dev->flags & IFF_POINTOPOINT)
3419                 addrconf_add_mroute(dev);
3420 }
3421 #endif
3422
3423 static int fixup_permanent_addr(struct net *net,
3424                                 struct inet6_dev *idev,
3425                                 struct inet6_ifaddr *ifp)
3426 {
3427         /* !fib6_node means the host route was removed from the
3428          * FIB, for example, if 'lo' device is taken down. In that
3429          * case regenerate the host route.
3430          */
3431         if (!ifp->rt || !ifp->rt->fib6_node) {
3432                 struct fib6_info *f6i, *prev;
3433
3434                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3435                                          GFP_ATOMIC);
3436                 if (IS_ERR(f6i))
3437                         return PTR_ERR(f6i);
3438
3439                 /* ifp->rt can be accessed outside of rtnl */
3440                 spin_lock(&ifp->lock);
3441                 prev = ifp->rt;
3442                 ifp->rt = f6i;
3443                 spin_unlock(&ifp->lock);
3444
3445                 fib6_info_release(prev);
3446         }
3447
3448         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3449                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3450                                       ifp->rt_priority, idev->dev, 0, 0,
3451                                       GFP_ATOMIC);
3452         }
3453
3454         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3455                 addrconf_dad_start(ifp);
3456
3457         return 0;
3458 }
3459
3460 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3461 {
3462         struct inet6_ifaddr *ifp, *tmp;
3463         struct inet6_dev *idev;
3464
3465         idev = __in6_dev_get(dev);
3466         if (!idev)
3467                 return;
3468
3469         write_lock_bh(&idev->lock);
3470
3471         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3472                 if ((ifp->flags & IFA_F_PERMANENT) &&
3473                     fixup_permanent_addr(net, idev, ifp) < 0) {
3474                         write_unlock_bh(&idev->lock);
3475                         in6_ifa_hold(ifp);
3476                         ipv6_del_addr(ifp);
3477                         write_lock_bh(&idev->lock);
3478
3479                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3480                                              idev->dev->name, &ifp->addr);
3481                 }
3482         }
3483
3484         write_unlock_bh(&idev->lock);
3485 }
3486
3487 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3488                            void *ptr)
3489 {
3490         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3491         struct netdev_notifier_change_info *change_info;
3492         struct netdev_notifier_changeupper_info *info;
3493         struct inet6_dev *idev = __in6_dev_get(dev);
3494         struct net *net = dev_net(dev);
3495         int run_pending = 0;
3496         int err;
3497
3498         switch (event) {
3499         case NETDEV_REGISTER:
3500                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3501                         idev = ipv6_add_dev(dev);
3502                         if (IS_ERR(idev))
3503                                 return notifier_from_errno(PTR_ERR(idev));
3504                 }
3505                 break;
3506
3507         case NETDEV_CHANGEMTU:
3508                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3509                 if (dev->mtu < IPV6_MIN_MTU) {
3510                         addrconf_ifdown(dev, dev != net->loopback_dev);
3511                         break;
3512                 }
3513
3514                 if (idev) {
3515                         rt6_mtu_change(dev, dev->mtu);
3516                         idev->cnf.mtu6 = dev->mtu;
3517                         break;
3518                 }
3519
3520                 /* allocate new idev */
3521                 idev = ipv6_add_dev(dev);
3522                 if (IS_ERR(idev))
3523                         break;
3524
3525                 /* device is still not ready */
3526                 if (!(idev->if_flags & IF_READY))
3527                         break;
3528
3529                 run_pending = 1;
3530                 fallthrough;
3531         case NETDEV_UP:
3532         case NETDEV_CHANGE:
3533                 if (dev->flags & IFF_SLAVE)
3534                         break;
3535
3536                 if (idev && idev->cnf.disable_ipv6)
3537                         break;
3538
3539                 if (event == NETDEV_UP) {
3540                         /* restore routes for permanent addresses */
3541                         addrconf_permanent_addr(net, dev);
3542
3543                         if (!addrconf_link_ready(dev)) {
3544                                 /* device is not ready yet. */
3545                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3546                                          dev->name);
3547                                 break;
3548                         }
3549
3550                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3551                                 idev = ipv6_add_dev(dev);
3552
3553                         if (!IS_ERR_OR_NULL(idev)) {
3554                                 idev->if_flags |= IF_READY;
3555                                 run_pending = 1;
3556                         }
3557                 } else if (event == NETDEV_CHANGE) {
3558                         if (!addrconf_link_ready(dev)) {
3559                                 /* device is still not ready. */
3560                                 rt6_sync_down_dev(dev, event);
3561                                 break;
3562                         }
3563
3564                         if (!IS_ERR_OR_NULL(idev)) {
3565                                 if (idev->if_flags & IF_READY) {
3566                                         /* device is already configured -
3567                                          * but resend MLD reports, we might
3568                                          * have roamed and need to update
3569                                          * multicast snooping switches
3570                                          */
3571                                         ipv6_mc_up(idev);
3572                                         change_info = ptr;
3573                                         if (change_info->flags_changed & IFF_NOARP)
3574                                                 addrconf_dad_run(idev, true);
3575                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3576                                         break;
3577                                 }
3578                                 idev->if_flags |= IF_READY;
3579                         }
3580
3581                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3582                                 dev->name);
3583
3584                         run_pending = 1;
3585                 }
3586
3587                 switch (dev->type) {
3588 #if IS_ENABLED(CONFIG_IPV6_SIT)
3589                 case ARPHRD_SIT:
3590                         addrconf_sit_config(dev);
3591                         break;
3592 #endif
3593 #if IS_ENABLED(CONFIG_NET_IPGRE)
3594                 case ARPHRD_IPGRE:
3595                         addrconf_gre_config(dev);
3596                         break;
3597 #endif
3598                 case ARPHRD_LOOPBACK:
3599                         init_loopback(dev);
3600                         break;
3601
3602                 default:
3603                         addrconf_dev_config(dev);
3604                         break;
3605                 }
3606
3607                 if (!IS_ERR_OR_NULL(idev)) {
3608                         if (run_pending)
3609                                 addrconf_dad_run(idev, false);
3610
3611                         /* Device has an address by now */
3612                         rt6_sync_up(dev, RTNH_F_DEAD);
3613
3614                         /*
3615                          * If the MTU changed during the interface down,
3616                          * when the interface up, the changed MTU must be
3617                          * reflected in the idev as well as routers.
3618                          */
3619                         if (idev->cnf.mtu6 != dev->mtu &&
3620                             dev->mtu >= IPV6_MIN_MTU) {
3621                                 rt6_mtu_change(dev, dev->mtu);
3622                                 idev->cnf.mtu6 = dev->mtu;
3623                         }
3624                         idev->tstamp = jiffies;
3625                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3626
3627                         /*
3628                          * If the changed mtu during down is lower than
3629                          * IPV6_MIN_MTU stop IPv6 on this interface.
3630                          */
3631                         if (dev->mtu < IPV6_MIN_MTU)
3632                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3633                 }
3634                 break;
3635
3636         case NETDEV_DOWN:
3637         case NETDEV_UNREGISTER:
3638                 /*
3639                  *      Remove all addresses from this interface.
3640                  */
3641                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3642                 break;
3643
3644         case NETDEV_CHANGENAME:
3645                 if (idev) {
3646                         snmp6_unregister_dev(idev);
3647                         addrconf_sysctl_unregister(idev);
3648                         err = addrconf_sysctl_register(idev);
3649                         if (err)
3650                                 return notifier_from_errno(err);
3651                         err = snmp6_register_dev(idev);
3652                         if (err) {
3653                                 addrconf_sysctl_unregister(idev);
3654                                 return notifier_from_errno(err);
3655                         }
3656                 }
3657                 break;
3658
3659         case NETDEV_PRE_TYPE_CHANGE:
3660         case NETDEV_POST_TYPE_CHANGE:
3661                 if (idev)
3662                         addrconf_type_change(dev, event);
3663                 break;
3664
3665         case NETDEV_CHANGEUPPER:
3666                 info = ptr;
3667
3668                 /* flush all routes if dev is linked to or unlinked from
3669                  * an L3 master device (e.g., VRF)
3670                  */
3671                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3672                         addrconf_ifdown(dev, false);
3673         }
3674
3675         return NOTIFY_OK;
3676 }
3677
3678 /*
3679  *      addrconf module should be notified of a device going up
3680  */
3681 static struct notifier_block ipv6_dev_notf = {
3682         .notifier_call = addrconf_notify,
3683         .priority = ADDRCONF_NOTIFY_PRIORITY,
3684 };
3685
3686 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3687 {
3688         struct inet6_dev *idev;
3689         ASSERT_RTNL();
3690
3691         idev = __in6_dev_get(dev);
3692
3693         if (event == NETDEV_POST_TYPE_CHANGE)
3694                 ipv6_mc_remap(idev);
3695         else if (event == NETDEV_PRE_TYPE_CHANGE)
3696                 ipv6_mc_unmap(idev);
3697 }
3698
3699 static bool addr_is_local(const struct in6_addr *addr)
3700 {
3701         return ipv6_addr_type(addr) &
3702                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3703 }
3704
3705 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3706 {
3707         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3708         struct net *net = dev_net(dev);
3709         struct inet6_dev *idev;
3710         struct inet6_ifaddr *ifa, *tmp;
3711         bool keep_addr = false;
3712         int state, i;
3713
3714         ASSERT_RTNL();
3715
3716         rt6_disable_ip(dev, event);
3717
3718         idev = __in6_dev_get(dev);
3719         if (!idev)
3720                 return -ENODEV;
3721
3722         /*
3723          * Step 1: remove reference to ipv6 device from parent device.
3724          *         Do not dev_put!
3725          */
3726         if (unregister) {
3727                 idev->dead = 1;
3728
3729                 /* protected by rtnl_lock */
3730                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3731
3732                 /* Step 1.5: remove snmp6 entry */
3733                 snmp6_unregister_dev(idev);
3734
3735         }
3736
3737         /* combine the user config with event to determine if permanent
3738          * addresses are to be removed from address hash table
3739          */
3740         if (!unregister && !idev->cnf.disable_ipv6) {
3741                 /* aggregate the system setting and interface setting */
3742                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3743
3744                 if (!_keep_addr)
3745                         _keep_addr = idev->cnf.keep_addr_on_down;
3746
3747                 keep_addr = (_keep_addr > 0);
3748         }
3749
3750         /* Step 2: clear hash table */
3751         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3752                 struct hlist_head *h = &inet6_addr_lst[i];
3753
3754                 spin_lock_bh(&addrconf_hash_lock);
3755 restart:
3756                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3757                         if (ifa->idev == idev) {
3758                                 addrconf_del_dad_work(ifa);
3759                                 /* combined flag + permanent flag decide if
3760                                  * address is retained on a down event
3761                                  */
3762                                 if (!keep_addr ||
3763                                     !(ifa->flags & IFA_F_PERMANENT) ||
3764                                     addr_is_local(&ifa->addr)) {
3765                                         hlist_del_init_rcu(&ifa->addr_lst);
3766                                         goto restart;
3767                                 }
3768                         }
3769                 }
3770                 spin_unlock_bh(&addrconf_hash_lock);
3771         }
3772
3773         write_lock_bh(&idev->lock);
3774
3775         addrconf_del_rs_timer(idev);
3776
3777         /* Step 2: clear flags for stateless addrconf */
3778         if (!unregister)
3779                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3780
3781         /* Step 3: clear tempaddr list */
3782         while (!list_empty(&idev->tempaddr_list)) {
3783                 ifa = list_first_entry(&idev->tempaddr_list,
3784                                        struct inet6_ifaddr, tmp_list);
3785                 list_del(&ifa->tmp_list);
3786                 write_unlock_bh(&idev->lock);
3787                 spin_lock_bh(&ifa->lock);
3788
3789                 if (ifa->ifpub) {
3790                         in6_ifa_put(ifa->ifpub);
3791                         ifa->ifpub = NULL;
3792                 }
3793                 spin_unlock_bh(&ifa->lock);
3794                 in6_ifa_put(ifa);
3795                 write_lock_bh(&idev->lock);
3796         }
3797
3798         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3799                 struct fib6_info *rt = NULL;
3800                 bool keep;
3801
3802                 addrconf_del_dad_work(ifa);
3803
3804                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3805                         !addr_is_local(&ifa->addr);
3806
3807                 write_unlock_bh(&idev->lock);
3808                 spin_lock_bh(&ifa->lock);
3809
3810                 if (keep) {
3811                         /* set state to skip the notifier below */
3812                         state = INET6_IFADDR_STATE_DEAD;
3813                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3814                         if (!(ifa->flags & IFA_F_NODAD))
3815                                 ifa->flags |= IFA_F_TENTATIVE;
3816
3817                         rt = ifa->rt;
3818                         ifa->rt = NULL;
3819                 } else {
3820                         state = ifa->state;
3821                         ifa->state = INET6_IFADDR_STATE_DEAD;
3822                 }
3823
3824                 spin_unlock_bh(&ifa->lock);
3825
3826                 if (rt)
3827                         ip6_del_rt(net, rt, false);
3828
3829                 if (state != INET6_IFADDR_STATE_DEAD) {
3830                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3831                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3832                 } else {
3833                         if (idev->cnf.forwarding)
3834                                 addrconf_leave_anycast(ifa);
3835                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3836                 }
3837
3838                 write_lock_bh(&idev->lock);
3839                 if (!keep) {
3840                         list_del_rcu(&ifa->if_list);
3841                         in6_ifa_put(ifa);
3842                 }
3843         }
3844
3845         write_unlock_bh(&idev->lock);
3846
3847         /* Step 5: Discard anycast and multicast list */
3848         if (unregister) {
3849                 ipv6_ac_destroy_dev(idev);
3850                 ipv6_mc_destroy_dev(idev);
3851         } else {
3852                 ipv6_mc_down(idev);
3853         }
3854
3855         idev->tstamp = jiffies;
3856
3857         /* Last: Shot the device (if unregistered) */
3858         if (unregister) {
3859                 addrconf_sysctl_unregister(idev);
3860                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3861                 neigh_ifdown(&nd_tbl, dev);
3862                 in6_dev_put(idev);
3863         }
3864         return 0;
3865 }
3866
3867 static void addrconf_rs_timer(struct timer_list *t)
3868 {
3869         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3870         struct net_device *dev = idev->dev;
3871         struct in6_addr lladdr;
3872
3873         write_lock(&idev->lock);
3874         if (idev->dead || !(idev->if_flags & IF_READY))
3875                 goto out;
3876
3877         if (!ipv6_accept_ra(idev))
3878                 goto out;
3879
3880         /* Announcement received after solicitation was sent */
3881         if (idev->if_flags & IF_RA_RCVD)
3882                 goto out;
3883
3884         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3885                 write_unlock(&idev->lock);
3886                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3887                         ndisc_send_rs(dev, &lladdr,
3888                                       &in6addr_linklocal_allrouters);
3889                 else
3890                         goto put;
3891
3892                 write_lock(&idev->lock);
3893                 idev->rs_interval = rfc3315_s14_backoff_update(
3894                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3895                 /* The wait after the last probe can be shorter */
3896                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3897                                              idev->cnf.rtr_solicits) ?
3898                                       idev->cnf.rtr_solicit_delay :
3899                                       idev->rs_interval);
3900         } else {
3901                 /*
3902                  * Note: we do not support deprecated "all on-link"
3903                  * assumption any longer.
3904                  */
3905                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3906         }
3907
3908 out:
3909         write_unlock(&idev->lock);
3910 put:
3911         in6_dev_put(idev);
3912 }
3913
3914 /*
3915  *      Duplicate Address Detection
3916  */
3917 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3918 {
3919         unsigned long rand_num;
3920         struct inet6_dev *idev = ifp->idev;
3921         u64 nonce;
3922
3923         if (ifp->flags & IFA_F_OPTIMISTIC)
3924                 rand_num = 0;
3925         else
3926                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3927
3928         nonce = 0;
3929         if (idev->cnf.enhanced_dad ||
3930             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3931                 do
3932                         get_random_bytes(&nonce, 6);
3933                 while (nonce == 0);
3934         }
3935         ifp->dad_nonce = nonce;
3936         ifp->dad_probes = idev->cnf.dad_transmits;
3937         addrconf_mod_dad_work(ifp, rand_num);
3938 }
3939
3940 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3941 {
3942         struct inet6_dev *idev = ifp->idev;
3943         struct net_device *dev = idev->dev;
3944         bool bump_id, notify = false;
3945         struct net *net;
3946
3947         addrconf_join_solict(dev, &ifp->addr);
3948
3949         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3950
3951         read_lock_bh(&idev->lock);
3952         spin_lock(&ifp->lock);
3953         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3954                 goto out;
3955
3956         net = dev_net(dev);
3957         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3958             (net->ipv6.devconf_all->accept_dad < 1 &&
3959              idev->cnf.accept_dad < 1) ||
3960             !(ifp->flags&IFA_F_TENTATIVE) ||
3961             ifp->flags & IFA_F_NODAD) {
3962                 bool send_na = false;
3963
3964                 if (ifp->flags & IFA_F_TENTATIVE &&
3965                     !(ifp->flags & IFA_F_OPTIMISTIC))
3966                         send_na = true;
3967                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3968                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3969                 spin_unlock(&ifp->lock);
3970                 read_unlock_bh(&idev->lock);
3971
3972                 addrconf_dad_completed(ifp, bump_id, send_na);
3973                 return;
3974         }
3975
3976         if (!(idev->if_flags & IF_READY)) {
3977                 spin_unlock(&ifp->lock);
3978                 read_unlock_bh(&idev->lock);
3979                 /*
3980                  * If the device is not ready:
3981                  * - keep it tentative if it is a permanent address.
3982                  * - otherwise, kill it.
3983                  */
3984                 in6_ifa_hold(ifp);
3985                 addrconf_dad_stop(ifp, 0);
3986                 return;
3987         }
3988
3989         /*
3990          * Optimistic nodes can start receiving
3991          * Frames right away
3992          */
3993         if (ifp->flags & IFA_F_OPTIMISTIC) {
3994                 ip6_ins_rt(net, ifp->rt);
3995                 if (ipv6_use_optimistic_addr(net, idev)) {
3996                         /* Because optimistic nodes can use this address,
3997                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3998                          */
3999                         notify = true;
4000                 }
4001         }
4002
4003         addrconf_dad_kick(ifp);
4004 out:
4005         spin_unlock(&ifp->lock);
4006         read_unlock_bh(&idev->lock);
4007         if (notify)
4008                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4009 }
4010
4011 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4012 {
4013         bool begin_dad = false;
4014
4015         spin_lock_bh(&ifp->lock);
4016         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4017                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4018                 begin_dad = true;
4019         }
4020         spin_unlock_bh(&ifp->lock);
4021
4022         if (begin_dad)
4023                 addrconf_mod_dad_work(ifp, 0);
4024 }
4025
4026 static void addrconf_dad_work(struct work_struct *w)
4027 {
4028         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4029                                                 struct inet6_ifaddr,
4030                                                 dad_work);
4031         struct inet6_dev *idev = ifp->idev;
4032         bool bump_id, disable_ipv6 = false;
4033         struct in6_addr mcaddr;
4034
4035         enum {
4036                 DAD_PROCESS,
4037                 DAD_BEGIN,
4038                 DAD_ABORT,
4039         } action = DAD_PROCESS;
4040
4041         rtnl_lock();
4042
4043         spin_lock_bh(&ifp->lock);
4044         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4045                 action = DAD_BEGIN;
4046                 ifp->state = INET6_IFADDR_STATE_DAD;
4047         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4048                 action = DAD_ABORT;
4049                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4050
4051                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4052                      idev->cnf.accept_dad > 1) &&
4053                     !idev->cnf.disable_ipv6 &&
4054                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4055                         struct in6_addr addr;
4056
4057                         addr.s6_addr32[0] = htonl(0xfe800000);
4058                         addr.s6_addr32[1] = 0;
4059
4060                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4061                             ipv6_addr_equal(&ifp->addr, &addr)) {
4062                                 /* DAD failed for link-local based on MAC */
4063                                 idev->cnf.disable_ipv6 = 1;
4064
4065                                 pr_info("%s: IPv6 being disabled!\n",
4066                                         ifp->idev->dev->name);
4067                                 disable_ipv6 = true;
4068                         }
4069                 }
4070         }
4071         spin_unlock_bh(&ifp->lock);
4072
4073         if (action == DAD_BEGIN) {
4074                 addrconf_dad_begin(ifp);
4075                 goto out;
4076         } else if (action == DAD_ABORT) {
4077                 in6_ifa_hold(ifp);
4078                 addrconf_dad_stop(ifp, 1);
4079                 if (disable_ipv6)
4080                         addrconf_ifdown(idev->dev, false);
4081                 goto out;
4082         }
4083
4084         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4085                 goto out;
4086
4087         write_lock_bh(&idev->lock);
4088         if (idev->dead || !(idev->if_flags & IF_READY)) {
4089                 write_unlock_bh(&idev->lock);
4090                 goto out;
4091         }
4092
4093         spin_lock(&ifp->lock);
4094         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4095                 spin_unlock(&ifp->lock);
4096                 write_unlock_bh(&idev->lock);
4097                 goto out;
4098         }
4099
4100         if (ifp->dad_probes == 0) {
4101                 bool send_na = false;
4102
4103                 /*
4104                  * DAD was successful
4105                  */
4106
4107                 if (ifp->flags & IFA_F_TENTATIVE &&
4108                     !(ifp->flags & IFA_F_OPTIMISTIC))
4109                         send_na = true;
4110                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4111                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4112                 spin_unlock(&ifp->lock);
4113                 write_unlock_bh(&idev->lock);
4114
4115                 addrconf_dad_completed(ifp, bump_id, send_na);
4116
4117                 goto out;
4118         }
4119
4120         ifp->dad_probes--;
4121         addrconf_mod_dad_work(ifp,
4122                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4123                                   HZ/100));
4124         spin_unlock(&ifp->lock);
4125         write_unlock_bh(&idev->lock);
4126
4127         /* send a neighbour solicitation for our addr */
4128         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4129         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4130                       ifp->dad_nonce);
4131 out:
4132         in6_ifa_put(ifp);
4133         rtnl_unlock();
4134 }
4135
4136 /* ifp->idev must be at least read locked */
4137 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4138 {
4139         struct inet6_ifaddr *ifpiter;
4140         struct inet6_dev *idev = ifp->idev;
4141
4142         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4143                 if (ifpiter->scope > IFA_LINK)
4144                         break;
4145                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4146                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4147                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4148                     IFA_F_PERMANENT)
4149                         return false;
4150         }
4151         return true;
4152 }
4153
4154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4155                                    bool send_na)
4156 {
4157         struct net_device *dev = ifp->idev->dev;
4158         struct in6_addr lladdr;
4159         bool send_rs, send_mld;
4160
4161         addrconf_del_dad_work(ifp);
4162
4163         /*
4164          *      Configure the address for reception. Now it is valid.
4165          */
4166
4167         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4168
4169         /* If added prefix is link local and we are prepared to process
4170            router advertisements, start sending router solicitations.
4171          */
4172
4173         read_lock_bh(&ifp->idev->lock);
4174         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4175         send_rs = send_mld &&
4176                   ipv6_accept_ra(ifp->idev) &&
4177                   ifp->idev->cnf.rtr_solicits != 0 &&
4178                   (dev->flags&IFF_LOOPBACK) == 0;
4179         read_unlock_bh(&ifp->idev->lock);
4180
4181         /* While dad is in progress mld report's source address is in6_addrany.
4182          * Resend with proper ll now.
4183          */
4184         if (send_mld)
4185                 ipv6_mc_dad_complete(ifp->idev);
4186
4187         /* send unsolicited NA if enabled */
4188         if (send_na &&
4189             (ifp->idev->cnf.ndisc_notify ||
4190              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4191                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4192                               /*router=*/ !!ifp->idev->cnf.forwarding,
4193                               /*solicited=*/ false, /*override=*/ true,
4194                               /*inc_opt=*/ true);
4195         }
4196
4197         if (send_rs) {
4198                 /*
4199                  *      If a host as already performed a random delay
4200                  *      [...] as part of DAD [...] there is no need
4201                  *      to delay again before sending the first RS
4202                  */
4203                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4204                         return;
4205                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4206
4207                 write_lock_bh(&ifp->idev->lock);
4208                 spin_lock(&ifp->lock);
4209                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4210                         ifp->idev->cnf.rtr_solicit_interval);
4211                 ifp->idev->rs_probes = 1;
4212                 ifp->idev->if_flags |= IF_RS_SENT;
4213                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4214                 spin_unlock(&ifp->lock);
4215                 write_unlock_bh(&ifp->idev->lock);
4216         }
4217
4218         if (bump_id)
4219                 rt_genid_bump_ipv6(dev_net(dev));
4220
4221         /* Make sure that a new temporary address will be created
4222          * before this temporary address becomes deprecated.
4223          */
4224         if (ifp->flags & IFA_F_TEMPORARY)
4225                 addrconf_verify_rtnl();
4226 }
4227
4228 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4229 {
4230         struct inet6_ifaddr *ifp;
4231
4232         read_lock_bh(&idev->lock);
4233         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4234                 spin_lock(&ifp->lock);
4235                 if ((ifp->flags & IFA_F_TENTATIVE &&
4236                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4237                         if (restart)
4238                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4239                         addrconf_dad_kick(ifp);
4240                 }
4241                 spin_unlock(&ifp->lock);
4242         }
4243         read_unlock_bh(&idev->lock);
4244 }
4245
4246 #ifdef CONFIG_PROC_FS
4247 struct if6_iter_state {
4248         struct seq_net_private p;
4249         int bucket;
4250         int offset;
4251 };
4252
4253 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4254 {
4255         struct if6_iter_state *state = seq->private;
4256         struct net *net = seq_file_net(seq);
4257         struct inet6_ifaddr *ifa = NULL;
4258         int p = 0;
4259
4260         /* initial bucket if pos is 0 */
4261         if (pos == 0) {
4262                 state->bucket = 0;
4263                 state->offset = 0;
4264         }
4265
4266         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4267                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4268                                          addr_lst) {
4269                         if (!net_eq(dev_net(ifa->idev->dev), net))
4270                                 continue;
4271                         /* sync with offset */
4272                         if (p < state->offset) {
4273                                 p++;
4274                                 continue;
4275                         }
4276                         return ifa;
4277                 }
4278
4279                 /* prepare for next bucket */
4280                 state->offset = 0;
4281                 p = 0;
4282         }
4283         return NULL;
4284 }
4285
4286 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4287                                          struct inet6_ifaddr *ifa)
4288 {
4289         struct if6_iter_state *state = seq->private;
4290         struct net *net = seq_file_net(seq);
4291
4292         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4293                 if (!net_eq(dev_net(ifa->idev->dev), net))
4294                         continue;
4295                 state->offset++;
4296                 return ifa;
4297         }
4298
4299         state->offset = 0;
4300         while (++state->bucket < IN6_ADDR_HSIZE) {
4301                 hlist_for_each_entry_rcu(ifa,
4302                                      &inet6_addr_lst[state->bucket], addr_lst) {
4303                         if (!net_eq(dev_net(ifa->idev->dev), net))
4304                                 continue;
4305                         return ifa;
4306                 }
4307         }
4308
4309         return NULL;
4310 }
4311
4312 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4313         __acquires(rcu)
4314 {
4315         rcu_read_lock();
4316         return if6_get_first(seq, *pos);
4317 }
4318
4319 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4320 {
4321         struct inet6_ifaddr *ifa;
4322
4323         ifa = if6_get_next(seq, v);
4324         ++*pos;
4325         return ifa;
4326 }
4327
4328 static void if6_seq_stop(struct seq_file *seq, void *v)
4329         __releases(rcu)
4330 {
4331         rcu_read_unlock();
4332 }
4333
4334 static int if6_seq_show(struct seq_file *seq, void *v)
4335 {
4336         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4337         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4338                    &ifp->addr,
4339                    ifp->idev->dev->ifindex,
4340                    ifp->prefix_len,
4341                    ifp->scope,
4342                    (u8) ifp->flags,
4343                    ifp->idev->dev->name);
4344         return 0;
4345 }
4346
4347 static const struct seq_operations if6_seq_ops = {
4348         .start  = if6_seq_start,
4349         .next   = if6_seq_next,
4350         .show   = if6_seq_show,
4351         .stop   = if6_seq_stop,
4352 };
4353
4354 static int __net_init if6_proc_net_init(struct net *net)
4355 {
4356         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4357                         sizeof(struct if6_iter_state)))
4358                 return -ENOMEM;
4359         return 0;
4360 }
4361
4362 static void __net_exit if6_proc_net_exit(struct net *net)
4363 {
4364         remove_proc_entry("if_inet6", net->proc_net);
4365 }
4366
4367 static struct pernet_operations if6_proc_net_ops = {
4368         .init = if6_proc_net_init,
4369         .exit = if6_proc_net_exit,
4370 };
4371
4372 int __init if6_proc_init(void)
4373 {
4374         return register_pernet_subsys(&if6_proc_net_ops);
4375 }
4376
4377 void if6_proc_exit(void)
4378 {
4379         unregister_pernet_subsys(&if6_proc_net_ops);
4380 }
4381 #endif  /* CONFIG_PROC_FS */
4382
4383 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4384 /* Check if address is a home address configured on any interface. */
4385 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4386 {
4387         unsigned int hash = inet6_addr_hash(net, addr);
4388         struct inet6_ifaddr *ifp = NULL;
4389         int ret = 0;
4390
4391         rcu_read_lock();
4392         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4393                 if (!net_eq(dev_net(ifp->idev->dev), net))
4394                         continue;
4395                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4396                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4397                         ret = 1;
4398                         break;
4399                 }
4400         }
4401         rcu_read_unlock();
4402         return ret;
4403 }
4404 #endif
4405
4406 /* RFC6554 has some algorithm to avoid loops in segment routing by
4407  * checking if the segments contains any of a local interface address.
4408  *
4409  * Quote:
4410  *
4411  * To detect loops in the SRH, a router MUST determine if the SRH
4412  * includes multiple addresses assigned to any interface on that router.
4413  * If such addresses appear more than once and are separated by at least
4414  * one address not assigned to that router.
4415  */
4416 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4417                           unsigned char nsegs)
4418 {
4419         const struct in6_addr *addr;
4420         int i, ret = 0, found = 0;
4421         struct inet6_ifaddr *ifp;
4422         bool separated = false;
4423         unsigned int hash;
4424         bool hash_found;
4425
4426         rcu_read_lock();
4427         for (i = 0; i < nsegs; i++) {
4428                 addr = &segs[i];
4429                 hash = inet6_addr_hash(net, addr);
4430
4431                 hash_found = false;
4432                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4433                         if (!net_eq(dev_net(ifp->idev->dev), net))
4434                                 continue;
4435
4436                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4437                                 hash_found = true;
4438                                 break;
4439                         }
4440                 }
4441
4442                 if (hash_found) {
4443                         if (found > 1 && separated) {
4444                                 ret = 1;
4445                                 break;
4446                         }
4447
4448                         separated = false;
4449                         found++;
4450                 } else {
4451                         separated = true;
4452                 }
4453         }
4454         rcu_read_unlock();
4455
4456         return ret;
4457 }
4458
4459 /*
4460  *      Periodic address status verification
4461  */
4462
4463 static void addrconf_verify_rtnl(void)
4464 {
4465         unsigned long now, next, next_sec, next_sched;
4466         struct inet6_ifaddr *ifp;
4467         int i;
4468
4469         ASSERT_RTNL();
4470
4471         rcu_read_lock_bh();
4472         now = jiffies;
4473         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4474
4475         cancel_delayed_work(&addr_chk_work);
4476
4477         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4478 restart:
4479                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4480                         unsigned long age;
4481
4482                         /* When setting preferred_lft to a value not zero or
4483                          * infinity, while valid_lft is infinity
4484                          * IFA_F_PERMANENT has a non-infinity life time.
4485                          */
4486                         if ((ifp->flags & IFA_F_PERMANENT) &&
4487                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4488                                 continue;
4489
4490                         spin_lock(&ifp->lock);
4491                         /* We try to batch several events at once. */
4492                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4493
4494                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4495                             age >= ifp->valid_lft) {
4496                                 spin_unlock(&ifp->lock);
4497                                 in6_ifa_hold(ifp);
4498                                 ipv6_del_addr(ifp);
4499                                 goto restart;
4500                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4501                                 spin_unlock(&ifp->lock);
4502                                 continue;
4503                         } else if (age >= ifp->prefered_lft) {
4504                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4505                                 int deprecate = 0;
4506
4507                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4508                                         deprecate = 1;
4509                                         ifp->flags |= IFA_F_DEPRECATED;
4510                                 }
4511
4512                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4513                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4514                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4515
4516                                 spin_unlock(&ifp->lock);
4517
4518                                 if (deprecate) {
4519                                         in6_ifa_hold(ifp);
4520
4521                                         ipv6_ifa_notify(0, ifp);
4522                                         in6_ifa_put(ifp);
4523                                         goto restart;
4524                                 }
4525                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4526                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4527                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4528                                         ifp->idev->cnf.dad_transmits *
4529                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4530
4531                                 if (age >= ifp->prefered_lft - regen_advance) {
4532                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4533                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4534                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4535                                         if (!ifp->regen_count && ifpub) {
4536                                                 ifp->regen_count++;
4537                                                 in6_ifa_hold(ifp);
4538                                                 in6_ifa_hold(ifpub);
4539                                                 spin_unlock(&ifp->lock);
4540
4541                                                 spin_lock(&ifpub->lock);
4542                                                 ifpub->regen_count = 0;
4543                                                 spin_unlock(&ifpub->lock);
4544                                                 rcu_read_unlock_bh();
4545                                                 ipv6_create_tempaddr(ifpub, true);
4546                                                 in6_ifa_put(ifpub);
4547                                                 in6_ifa_put(ifp);
4548                                                 rcu_read_lock_bh();
4549                                                 goto restart;
4550                                         }
4551                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4552                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4553                                 spin_unlock(&ifp->lock);
4554                         } else {
4555                                 /* ifp->prefered_lft <= ifp->valid_lft */
4556                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4557                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4558                                 spin_unlock(&ifp->lock);
4559                         }
4560                 }
4561         }
4562
4563         next_sec = round_jiffies_up(next);
4564         next_sched = next;
4565
4566         /* If rounded timeout is accurate enough, accept it. */
4567         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4568                 next_sched = next_sec;
4569
4570         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4571         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4572                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4573
4574         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4575                  now, next, next_sec, next_sched);
4576         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4577         rcu_read_unlock_bh();
4578 }
4579
4580 static void addrconf_verify_work(struct work_struct *w)
4581 {
4582         rtnl_lock();
4583         addrconf_verify_rtnl();
4584         rtnl_unlock();
4585 }
4586
4587 static void addrconf_verify(void)
4588 {
4589         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4590 }
4591
4592 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4593                                      struct in6_addr **peer_pfx)
4594 {
4595         struct in6_addr *pfx = NULL;
4596
4597         *peer_pfx = NULL;
4598
4599         if (addr)
4600                 pfx = nla_data(addr);
4601
4602         if (local) {
4603                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4604                         *peer_pfx = pfx;
4605                 pfx = nla_data(local);
4606         }
4607
4608         return pfx;
4609 }
4610
4611 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4612         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4613         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4614         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4615         [IFA_FLAGS]             = { .len = sizeof(u32) },
4616         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4617         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4618 };
4619
4620 static int
4621 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4622                   struct netlink_ext_ack *extack)
4623 {
4624         struct net *net = sock_net(skb->sk);
4625         struct ifaddrmsg *ifm;
4626         struct nlattr *tb[IFA_MAX+1];
4627         struct in6_addr *pfx, *peer_pfx;
4628         u32 ifa_flags;
4629         int err;
4630
4631         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4632                                      ifa_ipv6_policy, extack);
4633         if (err < 0)
4634                 return err;
4635
4636         ifm = nlmsg_data(nlh);
4637         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4638         if (!pfx)
4639                 return -EINVAL;
4640
4641         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4642
4643         /* We ignore other flags so far. */
4644         ifa_flags &= IFA_F_MANAGETEMPADDR;
4645
4646         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4647                               ifm->ifa_prefixlen);
4648 }
4649
4650 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4651                                unsigned long expires, u32 flags,
4652                                bool modify_peer)
4653 {
4654         struct fib6_info *f6i;
4655         u32 prio;
4656
4657         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4658                                         ifp->prefix_len,
4659                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4660         if (!f6i)
4661                 return -ENOENT;
4662
4663         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4664         if (f6i->fib6_metric != prio) {
4665                 /* delete old one */
4666                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4667
4668                 /* add new one */
4669                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4670                                       ifp->prefix_len,
4671                                       ifp->rt_priority, ifp->idev->dev,
4672                                       expires, flags, GFP_KERNEL);
4673         } else {
4674                 if (!expires)
4675                         fib6_clean_expires(f6i);
4676                 else
4677                         fib6_set_expires(f6i, expires);
4678
4679                 fib6_info_release(f6i);
4680         }
4681
4682         return 0;
4683 }
4684
4685 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4686 {
4687         u32 flags;
4688         clock_t expires;
4689         unsigned long timeout;
4690         bool was_managetempaddr;
4691         bool had_prefixroute;
4692         bool new_peer = false;
4693
4694         ASSERT_RTNL();
4695
4696         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4697                 return -EINVAL;
4698
4699         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4700             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4701                 return -EINVAL;
4702
4703         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4704                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4705
4706         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4707         if (addrconf_finite_timeout(timeout)) {
4708                 expires = jiffies_to_clock_t(timeout * HZ);
4709                 cfg->valid_lft = timeout;
4710                 flags = RTF_EXPIRES;
4711         } else {
4712                 expires = 0;
4713                 flags = 0;
4714                 cfg->ifa_flags |= IFA_F_PERMANENT;
4715         }
4716
4717         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4718         if (addrconf_finite_timeout(timeout)) {
4719                 if (timeout == 0)
4720                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4721                 cfg->preferred_lft = timeout;
4722         }
4723
4724         if (cfg->peer_pfx &&
4725             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4726                 if (!ipv6_addr_any(&ifp->peer_addr))
4727                         cleanup_prefix_route(ifp, expires, true, true);
4728                 new_peer = true;
4729         }
4730
4731         spin_lock_bh(&ifp->lock);
4732         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4733         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4734                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4735         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4736                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4737                         IFA_F_NOPREFIXROUTE);
4738         ifp->flags |= cfg->ifa_flags;
4739         ifp->tstamp = jiffies;
4740         ifp->valid_lft = cfg->valid_lft;
4741         ifp->prefered_lft = cfg->preferred_lft;
4742
4743         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4744                 ifp->rt_priority = cfg->rt_priority;
4745
4746         if (new_peer)
4747                 ifp->peer_addr = *cfg->peer_pfx;
4748
4749         spin_unlock_bh(&ifp->lock);
4750         if (!(ifp->flags&IFA_F_TENTATIVE))
4751                 ipv6_ifa_notify(0, ifp);
4752
4753         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4754                 int rc = -ENOENT;
4755
4756                 if (had_prefixroute)
4757                         rc = modify_prefix_route(ifp, expires, flags, false);
4758
4759                 /* prefix route could have been deleted; if so restore it */
4760                 if (rc == -ENOENT) {
4761                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4762                                               ifp->rt_priority, ifp->idev->dev,
4763                                               expires, flags, GFP_KERNEL);
4764                 }
4765
4766                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4767                         rc = modify_prefix_route(ifp, expires, flags, true);
4768
4769                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4770                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4771                                               ifp->rt_priority, ifp->idev->dev,
4772                                               expires, flags, GFP_KERNEL);
4773                 }
4774         } else if (had_prefixroute) {
4775                 enum cleanup_prefix_rt_t action;
4776                 unsigned long rt_expires;
4777
4778                 write_lock_bh(&ifp->idev->lock);
4779                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4780                 write_unlock_bh(&ifp->idev->lock);
4781
4782                 if (action != CLEANUP_PREFIX_RT_NOP) {
4783                         cleanup_prefix_route(ifp, rt_expires,
4784                                 action == CLEANUP_PREFIX_RT_DEL, false);
4785                 }
4786         }
4787
4788         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4789                 if (was_managetempaddr &&
4790                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4791                         cfg->valid_lft = 0;
4792                         cfg->preferred_lft = 0;
4793                 }
4794                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4795                                  cfg->preferred_lft, !was_managetempaddr,
4796                                  jiffies);
4797         }
4798
4799         addrconf_verify_rtnl();
4800
4801         return 0;
4802 }
4803
4804 static int
4805 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4806                   struct netlink_ext_ack *extack)
4807 {
4808         struct net *net = sock_net(skb->sk);
4809         struct ifaddrmsg *ifm;
4810         struct nlattr *tb[IFA_MAX+1];
4811         struct in6_addr *peer_pfx;
4812         struct inet6_ifaddr *ifa;
4813         struct net_device *dev;
4814         struct inet6_dev *idev;
4815         struct ifa6_config cfg;
4816         int err;
4817
4818         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4819                                      ifa_ipv6_policy, extack);
4820         if (err < 0)
4821                 return err;
4822
4823         memset(&cfg, 0, sizeof(cfg));
4824
4825         ifm = nlmsg_data(nlh);
4826         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4827         if (!cfg.pfx)
4828                 return -EINVAL;
4829
4830         cfg.peer_pfx = peer_pfx;
4831         cfg.plen = ifm->ifa_prefixlen;
4832         if (tb[IFA_RT_PRIORITY])
4833                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4834
4835         cfg.valid_lft = INFINITY_LIFE_TIME;
4836         cfg.preferred_lft = INFINITY_LIFE_TIME;
4837
4838         if (tb[IFA_CACHEINFO]) {
4839                 struct ifa_cacheinfo *ci;
4840
4841                 ci = nla_data(tb[IFA_CACHEINFO]);
4842                 cfg.valid_lft = ci->ifa_valid;
4843                 cfg.preferred_lft = ci->ifa_prefered;
4844         }
4845
4846         dev =  __dev_get_by_index(net, ifm->ifa_index);
4847         if (!dev)
4848                 return -ENODEV;
4849
4850         if (tb[IFA_FLAGS])
4851                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4852         else
4853                 cfg.ifa_flags = ifm->ifa_flags;
4854
4855         /* We ignore other flags so far. */
4856         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4857                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4858                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4859
4860         idev = ipv6_find_idev(dev);
4861         if (IS_ERR(idev))
4862                 return PTR_ERR(idev);
4863
4864         if (!ipv6_allow_optimistic_dad(net, idev))
4865                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4866
4867         if (cfg.ifa_flags & IFA_F_NODAD &&
4868             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4869                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4870                 return -EINVAL;
4871         }
4872
4873         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4874         if (!ifa) {
4875                 /*
4876                  * It would be best to check for !NLM_F_CREATE here but
4877                  * userspace already relies on not having to provide this.
4878                  */
4879                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4880         }
4881
4882         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4883             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4884                 err = -EEXIST;
4885         else
4886                 err = inet6_addr_modify(ifa, &cfg);
4887
4888         in6_ifa_put(ifa);
4889
4890         return err;
4891 }
4892
4893 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4894                           u8 scope, int ifindex)
4895 {
4896         struct ifaddrmsg *ifm;
4897
4898         ifm = nlmsg_data(nlh);
4899         ifm->ifa_family = AF_INET6;
4900         ifm->ifa_prefixlen = prefixlen;
4901         ifm->ifa_flags = flags;
4902         ifm->ifa_scope = scope;
4903         ifm->ifa_index = ifindex;
4904 }
4905
4906 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4907                          unsigned long tstamp, u32 preferred, u32 valid)
4908 {
4909         struct ifa_cacheinfo ci;
4910
4911         ci.cstamp = cstamp_delta(cstamp);
4912         ci.tstamp = cstamp_delta(tstamp);
4913         ci.ifa_prefered = preferred;
4914         ci.ifa_valid = valid;
4915
4916         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4917 }
4918
4919 static inline int rt_scope(int ifa_scope)
4920 {
4921         if (ifa_scope & IFA_HOST)
4922                 return RT_SCOPE_HOST;
4923         else if (ifa_scope & IFA_LINK)
4924                 return RT_SCOPE_LINK;
4925         else if (ifa_scope & IFA_SITE)
4926                 return RT_SCOPE_SITE;
4927         else
4928                 return RT_SCOPE_UNIVERSE;
4929 }
4930
4931 static inline int inet6_ifaddr_msgsize(void)
4932 {
4933         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4934                + nla_total_size(16) /* IFA_LOCAL */
4935                + nla_total_size(16) /* IFA_ADDRESS */
4936                + nla_total_size(sizeof(struct ifa_cacheinfo))
4937                + nla_total_size(4)  /* IFA_FLAGS */
4938                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4939 }
4940
4941 enum addr_type_t {
4942         UNICAST_ADDR,
4943         MULTICAST_ADDR,
4944         ANYCAST_ADDR,
4945 };
4946
4947 struct inet6_fill_args {
4948         u32 portid;
4949         u32 seq;
4950         int event;
4951         unsigned int flags;
4952         int netnsid;
4953         int ifindex;
4954         enum addr_type_t type;
4955 };
4956
4957 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4958                              struct inet6_fill_args *args)
4959 {
4960         struct nlmsghdr  *nlh;
4961         u32 preferred, valid;
4962
4963         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4964                         sizeof(struct ifaddrmsg), args->flags);
4965         if (!nlh)
4966                 return -EMSGSIZE;
4967
4968         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4969                       ifa->idev->dev->ifindex);
4970
4971         if (args->netnsid >= 0 &&
4972             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4973                 goto error;
4974
4975         if (!((ifa->flags&IFA_F_PERMANENT) &&
4976               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4977                 preferred = ifa->prefered_lft;
4978                 valid = ifa->valid_lft;
4979                 if (preferred != INFINITY_LIFE_TIME) {
4980                         long tval = (jiffies - ifa->tstamp)/HZ;
4981                         if (preferred > tval)
4982                                 preferred -= tval;
4983                         else
4984                                 preferred = 0;
4985                         if (valid != INFINITY_LIFE_TIME) {
4986                                 if (valid > tval)
4987                                         valid -= tval;
4988                                 else
4989                                         valid = 0;
4990                         }
4991                 }
4992         } else {
4993                 preferred = INFINITY_LIFE_TIME;
4994                 valid = INFINITY_LIFE_TIME;
4995         }
4996
4997         if (!ipv6_addr_any(&ifa->peer_addr)) {
4998                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4999                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5000                         goto error;
5001         } else
5002                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5003                         goto error;
5004
5005         if (ifa->rt_priority &&
5006             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5007                 goto error;
5008
5009         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5010                 goto error;
5011
5012         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5013                 goto error;
5014
5015         nlmsg_end(skb, nlh);
5016         return 0;
5017
5018 error:
5019         nlmsg_cancel(skb, nlh);
5020         return -EMSGSIZE;
5021 }
5022
5023 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5024                                struct inet6_fill_args *args)
5025 {
5026         struct nlmsghdr  *nlh;
5027         u8 scope = RT_SCOPE_UNIVERSE;
5028         int ifindex = ifmca->idev->dev->ifindex;
5029
5030         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5031                 scope = RT_SCOPE_SITE;
5032
5033         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5034                         sizeof(struct ifaddrmsg), args->flags);
5035         if (!nlh)
5036                 return -EMSGSIZE;
5037
5038         if (args->netnsid >= 0 &&
5039             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5040                 return -EMSGSIZE;
5041
5042         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5043         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5044             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5045                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5046                 nlmsg_cancel(skb, nlh);
5047                 return -EMSGSIZE;
5048         }
5049
5050         nlmsg_end(skb, nlh);
5051         return 0;
5052 }
5053
5054 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5055                                struct inet6_fill_args *args)
5056 {
5057         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5058         int ifindex = dev ? dev->ifindex : 1;
5059         struct nlmsghdr  *nlh;
5060         u8 scope = RT_SCOPE_UNIVERSE;
5061
5062         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5063                 scope = RT_SCOPE_SITE;
5064
5065         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5066                         sizeof(struct ifaddrmsg), args->flags);
5067         if (!nlh)
5068                 return -EMSGSIZE;
5069
5070         if (args->netnsid >= 0 &&
5071             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5072                 return -EMSGSIZE;
5073
5074         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5075         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5076             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5077                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5078                 nlmsg_cancel(skb, nlh);
5079                 return -EMSGSIZE;
5080         }
5081
5082         nlmsg_end(skb, nlh);
5083         return 0;
5084 }
5085
5086 /* called with rcu_read_lock() */
5087 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5088                           struct netlink_callback *cb, int s_ip_idx,
5089                           struct inet6_fill_args *fillargs)
5090 {
5091         struct ifmcaddr6 *ifmca;
5092         struct ifacaddr6 *ifaca;
5093         int ip_idx = 0;
5094         int err = 1;
5095
5096         read_lock_bh(&idev->lock);
5097         switch (fillargs->type) {
5098         case UNICAST_ADDR: {
5099                 struct inet6_ifaddr *ifa;
5100                 fillargs->event = RTM_NEWADDR;
5101
5102                 /* unicast address incl. temp addr */
5103                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5104                         if (ip_idx < s_ip_idx)
5105                                 goto next;
5106                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5107                         if (err < 0)
5108                                 break;
5109                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5110 next:
5111                         ip_idx++;
5112                 }
5113                 break;
5114         }
5115         case MULTICAST_ADDR:
5116                 fillargs->event = RTM_GETMULTICAST;
5117
5118                 /* multicast address */
5119                 for (ifmca = idev->mc_list; ifmca;
5120                      ifmca = ifmca->next, ip_idx++) {
5121                         if (ip_idx < s_ip_idx)
5122                                 continue;
5123                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5124                         if (err < 0)
5125                                 break;
5126                 }
5127                 break;
5128         case ANYCAST_ADDR:
5129                 fillargs->event = RTM_GETANYCAST;
5130                 /* anycast address */
5131                 for (ifaca = idev->ac_list; ifaca;
5132                      ifaca = ifaca->aca_next, ip_idx++) {
5133                         if (ip_idx < s_ip_idx)
5134                                 continue;
5135                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5136                         if (err < 0)
5137                                 break;
5138                 }
5139                 break;
5140         default:
5141                 break;
5142         }
5143         read_unlock_bh(&idev->lock);
5144         cb->args[2] = ip_idx;
5145         return err;
5146 }
5147
5148 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5149                                        struct inet6_fill_args *fillargs,
5150                                        struct net **tgt_net, struct sock *sk,
5151                                        struct netlink_callback *cb)
5152 {
5153         struct netlink_ext_ack *extack = cb->extack;
5154         struct nlattr *tb[IFA_MAX+1];
5155         struct ifaddrmsg *ifm;
5156         int err, i;
5157
5158         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5159                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5160                 return -EINVAL;
5161         }
5162
5163         ifm = nlmsg_data(nlh);
5164         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5165                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5166                 return -EINVAL;
5167         }
5168
5169         fillargs->ifindex = ifm->ifa_index;
5170         if (fillargs->ifindex) {
5171                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5172                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5173         }
5174
5175         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5176                                             ifa_ipv6_policy, extack);
5177         if (err < 0)
5178                 return err;
5179
5180         for (i = 0; i <= IFA_MAX; ++i) {
5181                 if (!tb[i])
5182                         continue;
5183
5184                 if (i == IFA_TARGET_NETNSID) {
5185                         struct net *net;
5186
5187                         fillargs->netnsid = nla_get_s32(tb[i]);
5188                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5189                         if (IS_ERR(net)) {
5190                                 fillargs->netnsid = -1;
5191                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5192                                 return PTR_ERR(net);
5193                         }
5194                         *tgt_net = net;
5195                 } else {
5196                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5197                         return -EINVAL;
5198                 }
5199         }
5200
5201         return 0;
5202 }
5203
5204 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5205                            enum addr_type_t type)
5206 {
5207         const struct nlmsghdr *nlh = cb->nlh;
5208         struct inet6_fill_args fillargs = {
5209                 .portid = NETLINK_CB(cb->skb).portid,
5210                 .seq = cb->nlh->nlmsg_seq,
5211                 .flags = NLM_F_MULTI,
5212                 .netnsid = -1,
5213                 .type = type,
5214         };
5215         struct net *net = sock_net(skb->sk);
5216         struct net *tgt_net = net;
5217         int idx, s_idx, s_ip_idx;
5218         int h, s_h;
5219         struct net_device *dev;
5220         struct inet6_dev *idev;
5221         struct hlist_head *head;
5222         int err = 0;
5223
5224         s_h = cb->args[0];
5225         s_idx = idx = cb->args[1];
5226         s_ip_idx = cb->args[2];
5227
5228         if (cb->strict_check) {
5229                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5230                                                   skb->sk, cb);
5231                 if (err < 0)
5232                         goto put_tgt_net;
5233
5234                 err = 0;
5235                 if (fillargs.ifindex) {
5236                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5237                         if (!dev) {
5238                                 err = -ENODEV;
5239                                 goto put_tgt_net;
5240                         }
5241                         idev = __in6_dev_get(dev);
5242                         if (idev) {
5243                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5244                                                      &fillargs);
5245                                 if (err > 0)
5246                                         err = 0;
5247                         }
5248                         goto put_tgt_net;
5249                 }
5250         }
5251
5252         rcu_read_lock();
5253         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5254         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5255                 idx = 0;
5256                 head = &tgt_net->dev_index_head[h];
5257                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5258                         if (idx < s_idx)
5259                                 goto cont;
5260                         if (h > s_h || idx > s_idx)
5261                                 s_ip_idx = 0;
5262                         idev = __in6_dev_get(dev);
5263                         if (!idev)
5264                                 goto cont;
5265
5266                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5267                                            &fillargs) < 0)
5268                                 goto done;
5269 cont:
5270                         idx++;
5271                 }
5272         }
5273 done:
5274         rcu_read_unlock();
5275         cb->args[0] = h;
5276         cb->args[1] = idx;
5277 put_tgt_net:
5278         if (fillargs.netnsid >= 0)
5279                 put_net(tgt_net);
5280
5281         return skb->len ? : err;
5282 }
5283
5284 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5285 {
5286         enum addr_type_t type = UNICAST_ADDR;
5287
5288         return inet6_dump_addr(skb, cb, type);
5289 }
5290
5291 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5292 {
5293         enum addr_type_t type = MULTICAST_ADDR;
5294
5295         return inet6_dump_addr(skb, cb, type);
5296 }
5297
5298
5299 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5300 {
5301         enum addr_type_t type = ANYCAST_ADDR;
5302
5303         return inet6_dump_addr(skb, cb, type);
5304 }
5305
5306 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5307                                        const struct nlmsghdr *nlh,
5308                                        struct nlattr **tb,
5309                                        struct netlink_ext_ack *extack)
5310 {
5311         struct ifaddrmsg *ifm;
5312         int i, err;
5313
5314         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5315                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5316                 return -EINVAL;
5317         }
5318
5319         if (!netlink_strict_get_check(skb))
5320                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5321                                               ifa_ipv6_policy, extack);
5322
5323         ifm = nlmsg_data(nlh);
5324         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5325                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5326                 return -EINVAL;
5327         }
5328
5329         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5330                                             ifa_ipv6_policy, extack);
5331         if (err)
5332                 return err;
5333
5334         for (i = 0; i <= IFA_MAX; i++) {
5335                 if (!tb[i])
5336                         continue;
5337
5338                 switch (i) {
5339                 case IFA_TARGET_NETNSID:
5340                 case IFA_ADDRESS:
5341                 case IFA_LOCAL:
5342                         break;
5343                 default:
5344                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5345                         return -EINVAL;
5346                 }
5347         }
5348
5349         return 0;
5350 }
5351
5352 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5353                              struct netlink_ext_ack *extack)
5354 {
5355         struct net *net = sock_net(in_skb->sk);
5356         struct inet6_fill_args fillargs = {
5357                 .portid = NETLINK_CB(in_skb).portid,
5358                 .seq = nlh->nlmsg_seq,
5359                 .event = RTM_NEWADDR,
5360                 .flags = 0,
5361                 .netnsid = -1,
5362         };
5363         struct net *tgt_net = net;
5364         struct ifaddrmsg *ifm;
5365         struct nlattr *tb[IFA_MAX+1];
5366         struct in6_addr *addr = NULL, *peer;
5367         struct net_device *dev = NULL;
5368         struct inet6_ifaddr *ifa;
5369         struct sk_buff *skb;
5370         int err;
5371
5372         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5373         if (err < 0)
5374                 return err;
5375
5376         if (tb[IFA_TARGET_NETNSID]) {
5377                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5378
5379                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5380                                                   fillargs.netnsid);
5381                 if (IS_ERR(tgt_net))
5382                         return PTR_ERR(tgt_net);
5383         }
5384
5385         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5386         if (!addr)
5387                 return -EINVAL;
5388
5389         ifm = nlmsg_data(nlh);
5390         if (ifm->ifa_index)
5391                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5392
5393         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5394         if (!ifa) {
5395                 err = -EADDRNOTAVAIL;
5396                 goto errout;
5397         }
5398
5399         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5400         if (!skb) {
5401                 err = -ENOBUFS;
5402                 goto errout_ifa;
5403         }
5404
5405         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5406         if (err < 0) {
5407                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5408                 WARN_ON(err == -EMSGSIZE);
5409                 kfree_skb(skb);
5410                 goto errout_ifa;
5411         }
5412         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5413 errout_ifa:
5414         in6_ifa_put(ifa);
5415 errout:
5416         if (dev)
5417                 dev_put(dev);
5418         if (fillargs.netnsid >= 0)
5419                 put_net(tgt_net);
5420
5421         return err;
5422 }
5423
5424 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5425 {
5426         struct sk_buff *skb;
5427         struct net *net = dev_net(ifa->idev->dev);
5428         struct inet6_fill_args fillargs = {
5429                 .portid = 0,
5430                 .seq = 0,
5431                 .event = event,
5432                 .flags = 0,
5433                 .netnsid = -1,
5434         };
5435         int err = -ENOBUFS;
5436
5437         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5438         if (!skb)
5439                 goto errout;
5440
5441         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5442         if (err < 0) {
5443                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5444                 WARN_ON(err == -EMSGSIZE);
5445                 kfree_skb(skb);
5446                 goto errout;
5447         }
5448         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5449         return;
5450 errout:
5451         if (err < 0)
5452                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5453 }
5454
5455 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5456                                 __s32 *array, int bytes)
5457 {
5458         BUG_ON(bytes < (DEVCONF_MAX * 4));
5459
5460         memset(array, 0, bytes);
5461         array[DEVCONF_FORWARDING] = cnf->forwarding;
5462         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5463         array[DEVCONF_MTU6] = cnf->mtu6;
5464         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5465         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5466         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5467         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5468         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5469         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5470                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5471         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5472                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5473         array[DEVCONF_RTR_SOLICIT_DELAY] =
5474                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5475         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5476         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5477                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5478         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5479                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5480         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5481         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5482         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5483         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5484         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5485         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5486         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5487         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5488         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5489 #ifdef CONFIG_IPV6_ROUTER_PREF
5490         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5491         array[DEVCONF_RTR_PROBE_INTERVAL] =
5492                 jiffies_to_msecs(cnf->rtr_probe_interval);
5493 #ifdef CONFIG_IPV6_ROUTE_INFO
5494         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5495         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5496 #endif
5497 #endif
5498         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5499         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5500 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5501         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5502         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5503 #endif
5504 #ifdef CONFIG_IPV6_MROUTE
5505         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5506 #endif
5507         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5508         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5509         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5510         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5511         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5512         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5513         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5514         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5515         /* we omit DEVCONF_STABLE_SECRET for now */
5516         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5517         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5518         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5519         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5520         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5521 #ifdef CONFIG_IPV6_SEG6_HMAC
5522         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5523 #endif
5524         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5525         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5526         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5527         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5528         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5529 }
5530
5531 static inline size_t inet6_ifla6_size(void)
5532 {
5533         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5534              + nla_total_size(sizeof(struct ifla_cacheinfo))
5535              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5536              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5537              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5538              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5539              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5540              + 0;
5541 }
5542
5543 static inline size_t inet6_if_nlmsg_size(void)
5544 {
5545         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5546                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5547                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5548                + nla_total_size(4) /* IFLA_MTU */
5549                + nla_total_size(4) /* IFLA_LINK */
5550                + nla_total_size(1) /* IFLA_OPERSTATE */
5551                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5552 }
5553
5554 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5555                                         int bytes)
5556 {
5557         int i;
5558         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5559         BUG_ON(pad < 0);
5560
5561         /* Use put_unaligned() because stats may not be aligned for u64. */
5562         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5563         for (i = 1; i < ICMP6_MIB_MAX; i++)
5564                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5565
5566         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5567 }
5568
5569 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5570                                         int bytes, size_t syncpoff)
5571 {
5572         int i, c;
5573         u64 buff[IPSTATS_MIB_MAX];
5574         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5575
5576         BUG_ON(pad < 0);
5577
5578         memset(buff, 0, sizeof(buff));
5579         buff[0] = IPSTATS_MIB_MAX;
5580
5581         for_each_possible_cpu(c) {
5582                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5583                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5584         }
5585
5586         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5587         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5588 }
5589
5590 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5591                              int bytes)
5592 {
5593         switch (attrtype) {
5594         case IFLA_INET6_STATS:
5595                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5596                                      offsetof(struct ipstats_mib, syncp));
5597                 break;
5598         case IFLA_INET6_ICMP6STATS:
5599                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5600                 break;
5601         }
5602 }
5603
5604 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5605                                   u32 ext_filter_mask)
5606 {
5607         struct nlattr *nla;
5608         struct ifla_cacheinfo ci;
5609
5610         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5611                 goto nla_put_failure;
5612         ci.max_reasm_len = IPV6_MAXPLEN;
5613         ci.tstamp = cstamp_delta(idev->tstamp);
5614         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5615         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5616         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5617                 goto nla_put_failure;
5618         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5619         if (!nla)
5620                 goto nla_put_failure;
5621         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5622
5623         /* XXX - MC not implemented */
5624
5625         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5626                 return 0;
5627
5628         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5629         if (!nla)
5630                 goto nla_put_failure;
5631         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5632
5633         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5634         if (!nla)
5635                 goto nla_put_failure;
5636         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5637
5638         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5639         if (!nla)
5640                 goto nla_put_failure;
5641         read_lock_bh(&idev->lock);
5642         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5643         read_unlock_bh(&idev->lock);
5644
5645         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5646                 goto nla_put_failure;
5647
5648         return 0;
5649
5650 nla_put_failure:
5651         return -EMSGSIZE;
5652 }
5653
5654 static size_t inet6_get_link_af_size(const struct net_device *dev,
5655                                      u32 ext_filter_mask)
5656 {
5657         if (!__in6_dev_get(dev))
5658                 return 0;
5659
5660         return inet6_ifla6_size();
5661 }
5662
5663 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5664                               u32 ext_filter_mask)
5665 {
5666         struct inet6_dev *idev = __in6_dev_get(dev);
5667
5668         if (!idev)
5669                 return -ENODATA;
5670
5671         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5672                 return -EMSGSIZE;
5673
5674         return 0;
5675 }
5676
5677 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5678 {
5679         struct inet6_ifaddr *ifp;
5680         struct net_device *dev = idev->dev;
5681         bool clear_token, update_rs = false;
5682         struct in6_addr ll_addr;
5683
5684         ASSERT_RTNL();
5685
5686         if (!token)
5687                 return -EINVAL;
5688         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5689                 return -EINVAL;
5690         if (!ipv6_accept_ra(idev))
5691                 return -EINVAL;
5692         if (idev->cnf.rtr_solicits == 0)
5693                 return -EINVAL;
5694
5695         write_lock_bh(&idev->lock);
5696
5697         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5698         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5699
5700         write_unlock_bh(&idev->lock);
5701
5702         clear_token = ipv6_addr_any(token);
5703         if (clear_token)
5704                 goto update_lft;
5705
5706         if (!idev->dead && (idev->if_flags & IF_READY) &&
5707             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5708                              IFA_F_OPTIMISTIC)) {
5709                 /* If we're not ready, then normal ifup will take care
5710                  * of this. Otherwise, we need to request our rs here.
5711                  */
5712                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5713                 update_rs = true;
5714         }
5715
5716 update_lft:
5717         write_lock_bh(&idev->lock);
5718
5719         if (update_rs) {
5720                 idev->if_flags |= IF_RS_SENT;
5721                 idev->rs_interval = rfc3315_s14_backoff_init(
5722                         idev->cnf.rtr_solicit_interval);
5723                 idev->rs_probes = 1;
5724                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5725         }
5726
5727         /* Well, that's kinda nasty ... */
5728         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5729                 spin_lock(&ifp->lock);
5730                 if (ifp->tokenized) {
5731                         ifp->valid_lft = 0;
5732                         ifp->prefered_lft = 0;
5733                 }
5734                 spin_unlock(&ifp->lock);
5735         }
5736
5737         write_unlock_bh(&idev->lock);
5738         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5739         addrconf_verify_rtnl();
5740         return 0;
5741 }
5742
5743 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5744         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5745         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5746 };
5747
5748 static int check_addr_gen_mode(int mode)
5749 {
5750         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5751             mode != IN6_ADDR_GEN_MODE_NONE &&
5752             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5753             mode != IN6_ADDR_GEN_MODE_RANDOM)
5754                 return -EINVAL;
5755         return 1;
5756 }
5757
5758 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5759                                 int mode)
5760 {
5761         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5762             !idev->cnf.stable_secret.initialized &&
5763             !net->ipv6.devconf_dflt->stable_secret.initialized)
5764                 return -EINVAL;
5765         return 1;
5766 }
5767
5768 static int inet6_validate_link_af(const struct net_device *dev,
5769                                   const struct nlattr *nla)
5770 {
5771         struct nlattr *tb[IFLA_INET6_MAX + 1];
5772         struct inet6_dev *idev = NULL;
5773         int err;
5774
5775         if (dev) {
5776                 idev = __in6_dev_get(dev);
5777                 if (!idev)
5778                         return -EAFNOSUPPORT;
5779         }
5780
5781         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5782                                           inet6_af_policy, NULL);
5783         if (err)
5784                 return err;
5785
5786         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5787                 return -EINVAL;
5788
5789         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5790                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5791
5792                 if (check_addr_gen_mode(mode) < 0)
5793                         return -EINVAL;
5794                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5795                         return -EINVAL;
5796         }
5797
5798         return 0;
5799 }
5800
5801 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5802 {
5803         struct inet6_dev *idev = __in6_dev_get(dev);
5804         struct nlattr *tb[IFLA_INET6_MAX + 1];
5805         int err;
5806
5807         if (!idev)
5808                 return -EAFNOSUPPORT;
5809
5810         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5811                 BUG();
5812
5813         if (tb[IFLA_INET6_TOKEN]) {
5814                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5815                 if (err)
5816                         return err;
5817         }
5818
5819         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5820                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5821
5822                 idev->cnf.addr_gen_mode = mode;
5823         }
5824
5825         return 0;
5826 }
5827
5828 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5829                              u32 portid, u32 seq, int event, unsigned int flags)
5830 {
5831         struct net_device *dev = idev->dev;
5832         struct ifinfomsg *hdr;
5833         struct nlmsghdr *nlh;
5834         void *protoinfo;
5835
5836         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5837         if (!nlh)
5838                 return -EMSGSIZE;
5839
5840         hdr = nlmsg_data(nlh);
5841         hdr->ifi_family = AF_INET6;
5842         hdr->__ifi_pad = 0;
5843         hdr->ifi_type = dev->type;
5844         hdr->ifi_index = dev->ifindex;
5845         hdr->ifi_flags = dev_get_flags(dev);
5846         hdr->ifi_change = 0;
5847
5848         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5849             (dev->addr_len &&
5850              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5851             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5852             (dev->ifindex != dev_get_iflink(dev) &&
5853              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5854             nla_put_u8(skb, IFLA_OPERSTATE,
5855                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5856                 goto nla_put_failure;
5857         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5858         if (!protoinfo)
5859                 goto nla_put_failure;
5860
5861         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5862                 goto nla_put_failure;
5863
5864         nla_nest_end(skb, protoinfo);
5865         nlmsg_end(skb, nlh);
5866         return 0;
5867
5868 nla_put_failure:
5869         nlmsg_cancel(skb, nlh);
5870         return -EMSGSIZE;
5871 }
5872
5873 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5874                                    struct netlink_ext_ack *extack)
5875 {
5876         struct ifinfomsg *ifm;
5877
5878         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5879                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5880                 return -EINVAL;
5881         }
5882
5883         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5884                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5885                 return -EINVAL;
5886         }
5887
5888         ifm = nlmsg_data(nlh);
5889         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5890             ifm->ifi_change || ifm->ifi_index) {
5891                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5892                 return -EINVAL;
5893         }
5894
5895         return 0;
5896 }
5897
5898 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5899 {
5900         struct net *net = sock_net(skb->sk);
5901         int h, s_h;
5902         int idx = 0, s_idx;
5903         struct net_device *dev;
5904         struct inet6_dev *idev;
5905         struct hlist_head *head;
5906
5907         /* only requests using strict checking can pass data to
5908          * influence the dump
5909          */
5910         if (cb->strict_check) {
5911                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5912
5913                 if (err < 0)
5914                         return err;
5915         }
5916
5917         s_h = cb->args[0];
5918         s_idx = cb->args[1];
5919
5920         rcu_read_lock();
5921         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5922                 idx = 0;
5923                 head = &net->dev_index_head[h];
5924                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5925                         if (idx < s_idx)
5926                                 goto cont;
5927                         idev = __in6_dev_get(dev);
5928                         if (!idev)
5929                                 goto cont;
5930                         if (inet6_fill_ifinfo(skb, idev,
5931                                               NETLINK_CB(cb->skb).portid,
5932                                               cb->nlh->nlmsg_seq,
5933                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5934                                 goto out;
5935 cont:
5936                         idx++;
5937                 }
5938         }
5939 out:
5940         rcu_read_unlock();
5941         cb->args[1] = idx;
5942         cb->args[0] = h;
5943
5944         return skb->len;
5945 }
5946
5947 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5948 {
5949         struct sk_buff *skb;
5950         struct net *net = dev_net(idev->dev);
5951         int err = -ENOBUFS;
5952
5953         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5954         if (!skb)
5955                 goto errout;
5956
5957         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5958         if (err < 0) {
5959                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5960                 WARN_ON(err == -EMSGSIZE);
5961                 kfree_skb(skb);
5962                 goto errout;
5963         }
5964         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5965         return;
5966 errout:
5967         if (err < 0)
5968                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5969 }
5970
5971 static inline size_t inet6_prefix_nlmsg_size(void)
5972 {
5973         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5974                + nla_total_size(sizeof(struct in6_addr))
5975                + nla_total_size(sizeof(struct prefix_cacheinfo));
5976 }
5977
5978 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5979                              struct prefix_info *pinfo, u32 portid, u32 seq,
5980                              int event, unsigned int flags)
5981 {
5982         struct prefixmsg *pmsg;
5983         struct nlmsghdr *nlh;
5984         struct prefix_cacheinfo ci;
5985
5986         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5987         if (!nlh)
5988                 return -EMSGSIZE;
5989
5990         pmsg = nlmsg_data(nlh);
5991         pmsg->prefix_family = AF_INET6;
5992         pmsg->prefix_pad1 = 0;
5993         pmsg->prefix_pad2 = 0;
5994         pmsg->prefix_ifindex = idev->dev->ifindex;
5995         pmsg->prefix_len = pinfo->prefix_len;
5996         pmsg->prefix_type = pinfo->type;
5997         pmsg->prefix_pad3 = 0;
5998         pmsg->prefix_flags = 0;
5999         if (pinfo->onlink)
6000                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6001         if (pinfo->autoconf)
6002                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6003
6004         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6005                 goto nla_put_failure;
6006         ci.preferred_time = ntohl(pinfo->prefered);
6007         ci.valid_time = ntohl(pinfo->valid);
6008         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6009                 goto nla_put_failure;
6010         nlmsg_end(skb, nlh);
6011         return 0;
6012
6013 nla_put_failure:
6014         nlmsg_cancel(skb, nlh);
6015         return -EMSGSIZE;
6016 }
6017
6018 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6019                          struct prefix_info *pinfo)
6020 {
6021         struct sk_buff *skb;
6022         struct net *net = dev_net(idev->dev);
6023         int err = -ENOBUFS;
6024
6025         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6026         if (!skb)
6027                 goto errout;
6028
6029         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6030         if (err < 0) {
6031                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6032                 WARN_ON(err == -EMSGSIZE);
6033                 kfree_skb(skb);
6034                 goto errout;
6035         }
6036         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6037         return;
6038 errout:
6039         if (err < 0)
6040                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6041 }
6042
6043 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6044 {
6045         struct net *net = dev_net(ifp->idev->dev);
6046
6047         if (event)
6048                 ASSERT_RTNL();
6049
6050         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6051
6052         switch (event) {
6053         case RTM_NEWADDR:
6054                 /*
6055                  * If the address was optimistic we inserted the route at the
6056                  * start of our DAD process, so we don't need to do it again.
6057                  * If the device was taken down in the middle of the DAD
6058                  * cycle there is a race where we could get here without a
6059                  * host route, so nothing to insert. That will be fixed when
6060                  * the device is brought up.
6061                  */
6062                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6063                         ip6_ins_rt(net, ifp->rt);
6064                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6065                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6066                                 &ifp->addr, ifp->idev->dev->name);
6067                 }
6068
6069                 if (ifp->idev->cnf.forwarding)
6070                         addrconf_join_anycast(ifp);
6071                 if (!ipv6_addr_any(&ifp->peer_addr))
6072                         addrconf_prefix_route(&ifp->peer_addr, 128,
6073                                               ifp->rt_priority, ifp->idev->dev,
6074                                               0, 0, GFP_ATOMIC);
6075                 break;
6076         case RTM_DELADDR:
6077                 if (ifp->idev->cnf.forwarding)
6078                         addrconf_leave_anycast(ifp);
6079                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6080                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6081                         struct fib6_info *rt;
6082
6083                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6084                                                        ifp->idev->dev, 0, 0,
6085                                                        false);
6086                         if (rt)
6087                                 ip6_del_rt(net, rt, false);
6088                 }
6089                 if (ifp->rt) {
6090                         ip6_del_rt(net, ifp->rt, false);
6091                         ifp->rt = NULL;
6092                 }
6093                 rt_genid_bump_ipv6(net);
6094                 break;
6095         }
6096         atomic_inc(&net->ipv6.dev_addr_genid);
6097 }
6098
6099 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6100 {
6101         rcu_read_lock_bh();
6102         if (likely(ifp->idev->dead == 0))
6103                 __ipv6_ifa_notify(event, ifp);
6104         rcu_read_unlock_bh();
6105 }
6106
6107 #ifdef CONFIG_SYSCTL
6108
6109 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6110                 void *buffer, size_t *lenp, loff_t *ppos)
6111 {
6112         int *valp = ctl->data;
6113         int val = *valp;
6114         loff_t pos = *ppos;
6115         struct ctl_table lctl;
6116         int ret;
6117
6118         /*
6119          * ctl->data points to idev->cnf.forwarding, we should
6120          * not modify it until we get the rtnl lock.
6121          */
6122         lctl = *ctl;
6123         lctl.data = &val;
6124
6125         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6126
6127         if (write)
6128                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6129         if (ret)
6130                 *ppos = pos;
6131         return ret;
6132 }
6133
6134 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6135                 void *buffer, size_t *lenp, loff_t *ppos)
6136 {
6137         struct inet6_dev *idev = ctl->extra1;
6138         int min_mtu = IPV6_MIN_MTU;
6139         struct ctl_table lctl;
6140
6141         lctl = *ctl;
6142         lctl.extra1 = &min_mtu;
6143         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6144
6145         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6146 }
6147
6148 static void dev_disable_change(struct inet6_dev *idev)
6149 {
6150         struct netdev_notifier_info info;
6151
6152         if (!idev || !idev->dev)
6153                 return;
6154
6155         netdev_notifier_info_init(&info, idev->dev);
6156         if (idev->cnf.disable_ipv6)
6157                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6158         else
6159                 addrconf_notify(NULL, NETDEV_UP, &info);
6160 }
6161
6162 static void addrconf_disable_change(struct net *net, __s32 newf)
6163 {
6164         struct net_device *dev;
6165         struct inet6_dev *idev;
6166
6167         for_each_netdev(net, dev) {
6168                 idev = __in6_dev_get(dev);
6169                 if (idev) {
6170                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6171                         idev->cnf.disable_ipv6 = newf;
6172                         if (changed)
6173                                 dev_disable_change(idev);
6174                 }
6175         }
6176 }
6177
6178 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6179 {
6180         struct net *net;
6181         int old;
6182
6183         if (!rtnl_trylock())
6184                 return restart_syscall();
6185
6186         net = (struct net *)table->extra2;
6187         old = *p;
6188         *p = newf;
6189
6190         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6191                 rtnl_unlock();
6192                 return 0;
6193         }
6194
6195         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6196                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6197                 addrconf_disable_change(net, newf);
6198         } else if ((!newf) ^ (!old))
6199                 dev_disable_change((struct inet6_dev *)table->extra1);
6200
6201         rtnl_unlock();
6202         return 0;
6203 }
6204
6205 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6206                 void *buffer, size_t *lenp, loff_t *ppos)
6207 {
6208         int *valp = ctl->data;
6209         int val = *valp;
6210         loff_t pos = *ppos;
6211         struct ctl_table lctl;
6212         int ret;
6213
6214         /*
6215          * ctl->data points to idev->cnf.disable_ipv6, we should
6216          * not modify it until we get the rtnl lock.
6217          */
6218         lctl = *ctl;
6219         lctl.data = &val;
6220
6221         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6222
6223         if (write)
6224                 ret = addrconf_disable_ipv6(ctl, valp, val);
6225         if (ret)
6226                 *ppos = pos;
6227         return ret;
6228 }
6229
6230 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6231                 void *buffer, size_t *lenp, loff_t *ppos)
6232 {
6233         int *valp = ctl->data;
6234         int ret;
6235         int old, new;
6236
6237         old = *valp;
6238         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6239         new = *valp;
6240
6241         if (write && old != new) {
6242                 struct net *net = ctl->extra2;
6243
6244                 if (!rtnl_trylock())
6245                         return restart_syscall();
6246
6247                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6248                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6249                                                      NETCONFA_PROXY_NEIGH,
6250                                                      NETCONFA_IFINDEX_DEFAULT,
6251                                                      net->ipv6.devconf_dflt);
6252                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6253                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6254                                                      NETCONFA_PROXY_NEIGH,
6255                                                      NETCONFA_IFINDEX_ALL,
6256                                                      net->ipv6.devconf_all);
6257                 else {
6258                         struct inet6_dev *idev = ctl->extra1;
6259
6260                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6261                                                      NETCONFA_PROXY_NEIGH,
6262                                                      idev->dev->ifindex,
6263                                                      &idev->cnf);
6264                 }
6265                 rtnl_unlock();
6266         }
6267
6268         return ret;
6269 }
6270
6271 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6272                                          void *buffer, size_t *lenp,
6273                                          loff_t *ppos)
6274 {
6275         int ret = 0;
6276         u32 new_val;
6277         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6278         struct net *net = (struct net *)ctl->extra2;
6279         struct ctl_table tmp = {
6280                 .data = &new_val,
6281                 .maxlen = sizeof(new_val),
6282                 .mode = ctl->mode,
6283         };
6284
6285         if (!rtnl_trylock())
6286                 return restart_syscall();
6287
6288         new_val = *((u32 *)ctl->data);
6289
6290         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6291         if (ret != 0)
6292                 goto out;
6293
6294         if (write) {
6295                 if (check_addr_gen_mode(new_val) < 0) {
6296                         ret = -EINVAL;
6297                         goto out;
6298                 }
6299
6300                 if (idev) {
6301                         if (check_stable_privacy(idev, net, new_val) < 0) {
6302                                 ret = -EINVAL;
6303                                 goto out;
6304                         }
6305
6306                         if (idev->cnf.addr_gen_mode != new_val) {
6307                                 idev->cnf.addr_gen_mode = new_val;
6308                                 addrconf_dev_config(idev->dev);
6309                         }
6310                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6311                         struct net_device *dev;
6312
6313                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6314                         for_each_netdev(net, dev) {
6315                                 idev = __in6_dev_get(dev);
6316                                 if (idev &&
6317                                     idev->cnf.addr_gen_mode != new_val) {
6318                                         idev->cnf.addr_gen_mode = new_val;
6319                                         addrconf_dev_config(idev->dev);
6320                                 }
6321                         }
6322                 }
6323
6324                 *((u32 *)ctl->data) = new_val;
6325         }
6326
6327 out:
6328         rtnl_unlock();
6329
6330         return ret;
6331 }
6332
6333 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6334                                          void *buffer, size_t *lenp,
6335                                          loff_t *ppos)
6336 {
6337         int err;
6338         struct in6_addr addr;
6339         char str[IPV6_MAX_STRLEN];
6340         struct ctl_table lctl = *ctl;
6341         struct net *net = ctl->extra2;
6342         struct ipv6_stable_secret *secret = ctl->data;
6343
6344         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6345                 return -EIO;
6346
6347         lctl.maxlen = IPV6_MAX_STRLEN;
6348         lctl.data = str;
6349
6350         if (!rtnl_trylock())
6351                 return restart_syscall();
6352
6353         if (!write && !secret->initialized) {
6354                 err = -EIO;
6355                 goto out;
6356         }
6357
6358         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6359         if (err >= sizeof(str)) {
6360                 err = -EIO;
6361                 goto out;
6362         }
6363
6364         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6365         if (err || !write)
6366                 goto out;
6367
6368         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6369                 err = -EIO;
6370                 goto out;
6371         }
6372
6373         secret->initialized = true;
6374         secret->secret = addr;
6375
6376         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6377                 struct net_device *dev;
6378
6379                 for_each_netdev(net, dev) {
6380                         struct inet6_dev *idev = __in6_dev_get(dev);
6381
6382                         if (idev) {
6383                                 idev->cnf.addr_gen_mode =
6384                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6385                         }
6386                 }
6387         } else {
6388                 struct inet6_dev *idev = ctl->extra1;
6389
6390                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6391         }
6392
6393 out:
6394         rtnl_unlock();
6395
6396         return err;
6397 }
6398
6399 static
6400 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6401                                                 int write, void *buffer,
6402                                                 size_t *lenp,
6403                                                 loff_t *ppos)
6404 {
6405         int *valp = ctl->data;
6406         int val = *valp;
6407         loff_t pos = *ppos;
6408         struct ctl_table lctl;
6409         int ret;
6410
6411         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6412          * we should not modify it until we get the rtnl lock.
6413          */
6414         lctl = *ctl;
6415         lctl.data = &val;
6416
6417         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6418
6419         if (write)
6420                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6421         if (ret)
6422                 *ppos = pos;
6423         return ret;
6424 }
6425
6426 static
6427 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6428 {
6429         if (rt) {
6430                 if (action)
6431                         rt->dst.flags |= DST_NOPOLICY;
6432                 else
6433                         rt->dst.flags &= ~DST_NOPOLICY;
6434         }
6435 }
6436
6437 static
6438 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6439 {
6440         struct inet6_ifaddr *ifa;
6441
6442         read_lock_bh(&idev->lock);
6443         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6444                 spin_lock(&ifa->lock);
6445                 if (ifa->rt) {
6446                         /* host routes only use builtin fib6_nh */
6447                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6448                         int cpu;
6449
6450                         rcu_read_lock();
6451                         ifa->rt->dst_nopolicy = val ? true : false;
6452                         if (nh->rt6i_pcpu) {
6453                                 for_each_possible_cpu(cpu) {
6454                                         struct rt6_info **rtp;
6455
6456                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6457                                         addrconf_set_nopolicy(*rtp, val);
6458                                 }
6459                         }
6460                         rcu_read_unlock();
6461                 }
6462                 spin_unlock(&ifa->lock);
6463         }
6464         read_unlock_bh(&idev->lock);
6465 }
6466
6467 static
6468 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6469 {
6470         struct inet6_dev *idev;
6471         struct net *net;
6472
6473         if (!rtnl_trylock())
6474                 return restart_syscall();
6475
6476         *valp = val;
6477
6478         net = (struct net *)ctl->extra2;
6479         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6480                 rtnl_unlock();
6481                 return 0;
6482         }
6483
6484         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6485                 struct net_device *dev;
6486
6487                 for_each_netdev(net, dev) {
6488                         idev = __in6_dev_get(dev);
6489                         if (idev)
6490                                 addrconf_disable_policy_idev(idev, val);
6491                 }
6492         } else {
6493                 idev = (struct inet6_dev *)ctl->extra1;
6494                 addrconf_disable_policy_idev(idev, val);
6495         }
6496
6497         rtnl_unlock();
6498         return 0;
6499 }
6500
6501 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6502                                    void *buffer, size_t *lenp, loff_t *ppos)
6503 {
6504         int *valp = ctl->data;
6505         int val = *valp;
6506         loff_t pos = *ppos;
6507         struct ctl_table lctl;
6508         int ret;
6509
6510         lctl = *ctl;
6511         lctl.data = &val;
6512         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6513
6514         if (write && (*valp != val))
6515                 ret = addrconf_disable_policy(ctl, valp, val);
6516
6517         if (ret)
6518                 *ppos = pos;
6519
6520         return ret;
6521 }
6522
6523 static int minus_one = -1;
6524 static const int two_five_five = 255;
6525
6526 static const struct ctl_table addrconf_sysctl[] = {
6527         {
6528                 .procname       = "forwarding",
6529                 .data           = &ipv6_devconf.forwarding,
6530                 .maxlen         = sizeof(int),
6531                 .mode           = 0644,
6532                 .proc_handler   = addrconf_sysctl_forward,
6533         },
6534         {
6535                 .procname       = "hop_limit",
6536                 .data           = &ipv6_devconf.hop_limit,
6537                 .maxlen         = sizeof(int),
6538                 .mode           = 0644,
6539                 .proc_handler   = proc_dointvec_minmax,
6540                 .extra1         = (void *)SYSCTL_ONE,
6541                 .extra2         = (void *)&two_five_five,
6542         },
6543         {
6544                 .procname       = "mtu",
6545                 .data           = &ipv6_devconf.mtu6,
6546                 .maxlen         = sizeof(int),
6547                 .mode           = 0644,
6548                 .proc_handler   = addrconf_sysctl_mtu,
6549         },
6550         {
6551                 .procname       = "accept_ra",
6552                 .data           = &ipv6_devconf.accept_ra,
6553                 .maxlen         = sizeof(int),
6554                 .mode           = 0644,
6555                 .proc_handler   = proc_dointvec,
6556         },
6557         {
6558                 .procname       = "accept_redirects",
6559                 .data           = &ipv6_devconf.accept_redirects,
6560                 .maxlen         = sizeof(int),
6561                 .mode           = 0644,
6562                 .proc_handler   = proc_dointvec,
6563         },
6564         {
6565                 .procname       = "autoconf",
6566                 .data           = &ipv6_devconf.autoconf,
6567                 .maxlen         = sizeof(int),
6568                 .mode           = 0644,
6569                 .proc_handler   = proc_dointvec,
6570         },
6571         {
6572                 .procname       = "dad_transmits",
6573                 .data           = &ipv6_devconf.dad_transmits,
6574                 .maxlen         = sizeof(int),
6575                 .mode           = 0644,
6576                 .proc_handler   = proc_dointvec,
6577         },
6578         {
6579                 .procname       = "router_solicitations",
6580                 .data           = &ipv6_devconf.rtr_solicits,
6581                 .maxlen         = sizeof(int),
6582                 .mode           = 0644,
6583                 .proc_handler   = proc_dointvec_minmax,
6584                 .extra1         = &minus_one,
6585         },
6586         {
6587                 .procname       = "router_solicitation_interval",
6588                 .data           = &ipv6_devconf.rtr_solicit_interval,
6589                 .maxlen         = sizeof(int),
6590                 .mode           = 0644,
6591                 .proc_handler   = proc_dointvec_jiffies,
6592         },
6593         {
6594                 .procname       = "router_solicitation_max_interval",
6595                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6596                 .maxlen         = sizeof(int),
6597                 .mode           = 0644,
6598                 .proc_handler   = proc_dointvec_jiffies,
6599         },
6600         {
6601                 .procname       = "router_solicitation_delay",
6602                 .data           = &ipv6_devconf.rtr_solicit_delay,
6603                 .maxlen         = sizeof(int),
6604                 .mode           = 0644,
6605                 .proc_handler   = proc_dointvec_jiffies,
6606         },
6607         {
6608                 .procname       = "force_mld_version",
6609                 .data           = &ipv6_devconf.force_mld_version,
6610                 .maxlen         = sizeof(int),
6611                 .mode           = 0644,
6612                 .proc_handler   = proc_dointvec,
6613         },
6614         {
6615                 .procname       = "mldv1_unsolicited_report_interval",
6616                 .data           =
6617                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6618                 .maxlen         = sizeof(int),
6619                 .mode           = 0644,
6620                 .proc_handler   = proc_dointvec_ms_jiffies,
6621         },
6622         {
6623                 .procname       = "mldv2_unsolicited_report_interval",
6624                 .data           =
6625                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6626                 .maxlen         = sizeof(int),
6627                 .mode           = 0644,
6628                 .proc_handler   = proc_dointvec_ms_jiffies,
6629         },
6630         {
6631                 .procname       = "use_tempaddr",
6632                 .data           = &ipv6_devconf.use_tempaddr,
6633                 .maxlen         = sizeof(int),
6634                 .mode           = 0644,
6635                 .proc_handler   = proc_dointvec,
6636         },
6637         {
6638                 .procname       = "temp_valid_lft",
6639                 .data           = &ipv6_devconf.temp_valid_lft,
6640                 .maxlen         = sizeof(int),
6641                 .mode           = 0644,
6642                 .proc_handler   = proc_dointvec,
6643         },
6644         {
6645                 .procname       = "temp_prefered_lft",
6646                 .data           = &ipv6_devconf.temp_prefered_lft,
6647                 .maxlen         = sizeof(int),
6648                 .mode           = 0644,
6649                 .proc_handler   = proc_dointvec,
6650         },
6651         {
6652                 .procname       = "regen_max_retry",
6653                 .data           = &ipv6_devconf.regen_max_retry,
6654                 .maxlen         = sizeof(int),
6655                 .mode           = 0644,
6656                 .proc_handler   = proc_dointvec,
6657         },
6658         {
6659                 .procname       = "max_desync_factor",
6660                 .data           = &ipv6_devconf.max_desync_factor,
6661                 .maxlen         = sizeof(int),
6662                 .mode           = 0644,
6663                 .proc_handler   = proc_dointvec,
6664         },
6665         {
6666                 .procname       = "max_addresses",
6667                 .data           = &ipv6_devconf.max_addresses,
6668                 .maxlen         = sizeof(int),
6669                 .mode           = 0644,
6670                 .proc_handler   = proc_dointvec,
6671         },
6672         {
6673                 .procname       = "accept_ra_defrtr",
6674                 .data           = &ipv6_devconf.accept_ra_defrtr,
6675                 .maxlen         = sizeof(int),
6676                 .mode           = 0644,
6677                 .proc_handler   = proc_dointvec,
6678         },
6679         {
6680                 .procname       = "accept_ra_min_hop_limit",
6681                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6682                 .maxlen         = sizeof(int),
6683                 .mode           = 0644,
6684                 .proc_handler   = proc_dointvec,
6685         },
6686         {
6687                 .procname       = "accept_ra_pinfo",
6688                 .data           = &ipv6_devconf.accept_ra_pinfo,
6689                 .maxlen         = sizeof(int),
6690                 .mode           = 0644,
6691                 .proc_handler   = proc_dointvec,
6692         },
6693 #ifdef CONFIG_IPV6_ROUTER_PREF
6694         {
6695                 .procname       = "accept_ra_rtr_pref",
6696                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6697                 .maxlen         = sizeof(int),
6698                 .mode           = 0644,
6699                 .proc_handler   = proc_dointvec,
6700         },
6701         {
6702                 .procname       = "router_probe_interval",
6703                 .data           = &ipv6_devconf.rtr_probe_interval,
6704                 .maxlen         = sizeof(int),
6705                 .mode           = 0644,
6706                 .proc_handler   = proc_dointvec_jiffies,
6707         },
6708 #ifdef CONFIG_IPV6_ROUTE_INFO
6709         {
6710                 .procname       = "accept_ra_rt_info_min_plen",
6711                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6712                 .maxlen         = sizeof(int),
6713                 .mode           = 0644,
6714                 .proc_handler   = proc_dointvec,
6715         },
6716         {
6717                 .procname       = "accept_ra_rt_info_max_plen",
6718                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6719                 .maxlen         = sizeof(int),
6720                 .mode           = 0644,
6721                 .proc_handler   = proc_dointvec,
6722         },
6723 #endif
6724 #endif
6725         {
6726                 .procname       = "proxy_ndp",
6727                 .data           = &ipv6_devconf.proxy_ndp,
6728                 .maxlen         = sizeof(int),
6729                 .mode           = 0644,
6730                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6731         },
6732         {
6733                 .procname       = "accept_source_route",
6734                 .data           = &ipv6_devconf.accept_source_route,
6735                 .maxlen         = sizeof(int),
6736                 .mode           = 0644,
6737                 .proc_handler   = proc_dointvec,
6738         },
6739 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6740         {
6741                 .procname       = "optimistic_dad",
6742                 .data           = &ipv6_devconf.optimistic_dad,
6743                 .maxlen         = sizeof(int),
6744                 .mode           = 0644,
6745                 .proc_handler   = proc_dointvec,
6746         },
6747         {
6748                 .procname       = "use_optimistic",
6749                 .data           = &ipv6_devconf.use_optimistic,
6750                 .maxlen         = sizeof(int),
6751                 .mode           = 0644,
6752                 .proc_handler   = proc_dointvec,
6753         },
6754 #endif
6755 #ifdef CONFIG_IPV6_MROUTE
6756         {
6757                 .procname       = "mc_forwarding",
6758                 .data           = &ipv6_devconf.mc_forwarding,
6759                 .maxlen         = sizeof(int),
6760                 .mode           = 0444,
6761                 .proc_handler   = proc_dointvec,
6762         },
6763 #endif
6764         {
6765                 .procname       = "disable_ipv6",
6766                 .data           = &ipv6_devconf.disable_ipv6,
6767                 .maxlen         = sizeof(int),
6768                 .mode           = 0644,
6769                 .proc_handler   = addrconf_sysctl_disable,
6770         },
6771         {
6772                 .procname       = "accept_dad",
6773                 .data           = &ipv6_devconf.accept_dad,
6774                 .maxlen         = sizeof(int),
6775                 .mode           = 0644,
6776                 .proc_handler   = proc_dointvec,
6777         },
6778         {
6779                 .procname       = "force_tllao",
6780                 .data           = &ipv6_devconf.force_tllao,
6781                 .maxlen         = sizeof(int),
6782                 .mode           = 0644,
6783                 .proc_handler   = proc_dointvec
6784         },
6785         {
6786                 .procname       = "ndisc_notify",
6787                 .data           = &ipv6_devconf.ndisc_notify,
6788                 .maxlen         = sizeof(int),
6789                 .mode           = 0644,
6790                 .proc_handler   = proc_dointvec
6791         },
6792         {
6793                 .procname       = "suppress_frag_ndisc",
6794                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6795                 .maxlen         = sizeof(int),
6796                 .mode           = 0644,
6797                 .proc_handler   = proc_dointvec
6798         },
6799         {
6800                 .procname       = "accept_ra_from_local",
6801                 .data           = &ipv6_devconf.accept_ra_from_local,
6802                 .maxlen         = sizeof(int),
6803                 .mode           = 0644,
6804                 .proc_handler   = proc_dointvec,
6805         },
6806         {
6807                 .procname       = "accept_ra_mtu",
6808                 .data           = &ipv6_devconf.accept_ra_mtu,
6809                 .maxlen         = sizeof(int),
6810                 .mode           = 0644,
6811                 .proc_handler   = proc_dointvec,
6812         },
6813         {
6814                 .procname       = "stable_secret",
6815                 .data           = &ipv6_devconf.stable_secret,
6816                 .maxlen         = IPV6_MAX_STRLEN,
6817                 .mode           = 0600,
6818                 .proc_handler   = addrconf_sysctl_stable_secret,
6819         },
6820         {
6821                 .procname       = "use_oif_addrs_only",
6822                 .data           = &ipv6_devconf.use_oif_addrs_only,
6823                 .maxlen         = sizeof(int),
6824                 .mode           = 0644,
6825                 .proc_handler   = proc_dointvec,
6826         },
6827         {
6828                 .procname       = "ignore_routes_with_linkdown",
6829                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6830                 .maxlen         = sizeof(int),
6831                 .mode           = 0644,
6832                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6833         },
6834         {
6835                 .procname       = "drop_unicast_in_l2_multicast",
6836                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6837                 .maxlen         = sizeof(int),
6838                 .mode           = 0644,
6839                 .proc_handler   = proc_dointvec,
6840         },
6841         {
6842                 .procname       = "drop_unsolicited_na",
6843                 .data           = &ipv6_devconf.drop_unsolicited_na,
6844                 .maxlen         = sizeof(int),
6845                 .mode           = 0644,
6846                 .proc_handler   = proc_dointvec,
6847         },
6848         {
6849                 .procname       = "keep_addr_on_down",
6850                 .data           = &ipv6_devconf.keep_addr_on_down,
6851                 .maxlen         = sizeof(int),
6852                 .mode           = 0644,
6853                 .proc_handler   = proc_dointvec,
6854
6855         },
6856         {
6857                 .procname       = "seg6_enabled",
6858                 .data           = &ipv6_devconf.seg6_enabled,
6859                 .maxlen         = sizeof(int),
6860                 .mode           = 0644,
6861                 .proc_handler   = proc_dointvec,
6862         },
6863 #ifdef CONFIG_IPV6_SEG6_HMAC
6864         {
6865                 .procname       = "seg6_require_hmac",
6866                 .data           = &ipv6_devconf.seg6_require_hmac,
6867                 .maxlen         = sizeof(int),
6868                 .mode           = 0644,
6869                 .proc_handler   = proc_dointvec,
6870         },
6871 #endif
6872         {
6873                 .procname       = "enhanced_dad",
6874                 .data           = &ipv6_devconf.enhanced_dad,
6875                 .maxlen         = sizeof(int),
6876                 .mode           = 0644,
6877                 .proc_handler   = proc_dointvec,
6878         },
6879         {
6880                 .procname               = "addr_gen_mode",
6881                 .data                   = &ipv6_devconf.addr_gen_mode,
6882                 .maxlen                 = sizeof(int),
6883                 .mode                   = 0644,
6884                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6885         },
6886         {
6887                 .procname       = "disable_policy",
6888                 .data           = &ipv6_devconf.disable_policy,
6889                 .maxlen         = sizeof(int),
6890                 .mode           = 0644,
6891                 .proc_handler   = addrconf_sysctl_disable_policy,
6892         },
6893         {
6894                 .procname       = "ndisc_tclass",
6895                 .data           = &ipv6_devconf.ndisc_tclass,
6896                 .maxlen         = sizeof(int),
6897                 .mode           = 0644,
6898                 .proc_handler   = proc_dointvec_minmax,
6899                 .extra1         = (void *)SYSCTL_ZERO,
6900                 .extra2         = (void *)&two_five_five,
6901         },
6902         {
6903                 .procname       = "rpl_seg_enabled",
6904                 .data           = &ipv6_devconf.rpl_seg_enabled,
6905                 .maxlen         = sizeof(int),
6906                 .mode           = 0644,
6907                 .proc_handler   = proc_dointvec,
6908         },
6909         {
6910                 /* sentinel */
6911         }
6912 };
6913
6914 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6915                 struct inet6_dev *idev, struct ipv6_devconf *p)
6916 {
6917         int i, ifindex;
6918         struct ctl_table *table;
6919         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6920
6921         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6922         if (!table)
6923                 goto out;
6924
6925         for (i = 0; table[i].data; i++) {
6926                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6927                 /* If one of these is already set, then it is not safe to
6928                  * overwrite either of them: this makes proc_dointvec_minmax
6929                  * usable.
6930                  */
6931                 if (!table[i].extra1 && !table[i].extra2) {
6932                         table[i].extra1 = idev; /* embedded; no ref */
6933                         table[i].extra2 = net;
6934                 }
6935         }
6936
6937         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6938
6939         p->sysctl_header = register_net_sysctl(net, path, table);
6940         if (!p->sysctl_header)
6941                 goto free;
6942
6943         if (!strcmp(dev_name, "all"))
6944                 ifindex = NETCONFA_IFINDEX_ALL;
6945         else if (!strcmp(dev_name, "default"))
6946                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6947         else
6948                 ifindex = idev->dev->ifindex;
6949         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6950                                      ifindex, p);
6951         return 0;
6952
6953 free:
6954         kfree(table);
6955 out:
6956         return -ENOBUFS;
6957 }
6958
6959 static void __addrconf_sysctl_unregister(struct net *net,
6960                                          struct ipv6_devconf *p, int ifindex)
6961 {
6962         struct ctl_table *table;
6963
6964         if (!p->sysctl_header)
6965                 return;
6966
6967         table = p->sysctl_header->ctl_table_arg;
6968         unregister_net_sysctl_table(p->sysctl_header);
6969         p->sysctl_header = NULL;
6970         kfree(table);
6971
6972         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6973 }
6974
6975 static int addrconf_sysctl_register(struct inet6_dev *idev)
6976 {
6977         int err;
6978
6979         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6980                 return -EINVAL;
6981
6982         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6983                                     &ndisc_ifinfo_sysctl_change);
6984         if (err)
6985                 return err;
6986         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6987                                          idev, &idev->cnf);
6988         if (err)
6989                 neigh_sysctl_unregister(idev->nd_parms);
6990
6991         return err;
6992 }
6993
6994 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6995 {
6996         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6997                                      idev->dev->ifindex);
6998         neigh_sysctl_unregister(idev->nd_parms);
6999 }
7000
7001
7002 #endif
7003
7004 static int __net_init addrconf_init_net(struct net *net)
7005 {
7006         int err = -ENOMEM;
7007         struct ipv6_devconf *all, *dflt;
7008
7009         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7010         if (!all)
7011                 goto err_alloc_all;
7012
7013         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7014         if (!dflt)
7015                 goto err_alloc_dflt;
7016
7017         if (IS_ENABLED(CONFIG_SYSCTL) &&
7018             !net_eq(net, &init_net)) {
7019                 switch (sysctl_devconf_inherit_init_net) {
7020                 case 1:  /* copy from init_net */
7021                         memcpy(all, init_net.ipv6.devconf_all,
7022                                sizeof(ipv6_devconf));
7023                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7024                                sizeof(ipv6_devconf_dflt));
7025                         break;
7026                 case 3: /* copy from the current netns */
7027                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7028                                sizeof(ipv6_devconf));
7029                         memcpy(dflt,
7030                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7031                                sizeof(ipv6_devconf_dflt));
7032                         break;
7033                 case 0:
7034                 case 2:
7035                         /* use compiled values */
7036                         break;
7037                 }
7038         }
7039
7040         /* these will be inherited by all namespaces */
7041         dflt->autoconf = ipv6_defaults.autoconf;
7042         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7043
7044         dflt->stable_secret.initialized = false;
7045         all->stable_secret.initialized = false;
7046
7047         net->ipv6.devconf_all = all;
7048         net->ipv6.devconf_dflt = dflt;
7049
7050 #ifdef CONFIG_SYSCTL
7051         err = __addrconf_sysctl_register(net, "all", NULL, all);
7052         if (err < 0)
7053                 goto err_reg_all;
7054
7055         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7056         if (err < 0)
7057                 goto err_reg_dflt;
7058 #endif
7059         return 0;
7060
7061 #ifdef CONFIG_SYSCTL
7062 err_reg_dflt:
7063         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7064 err_reg_all:
7065         kfree(dflt);
7066 #endif
7067 err_alloc_dflt:
7068         kfree(all);
7069 err_alloc_all:
7070         return err;
7071 }
7072
7073 static void __net_exit addrconf_exit_net(struct net *net)
7074 {
7075 #ifdef CONFIG_SYSCTL
7076         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7077                                      NETCONFA_IFINDEX_DEFAULT);
7078         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7079                                      NETCONFA_IFINDEX_ALL);
7080 #endif
7081         kfree(net->ipv6.devconf_dflt);
7082         kfree(net->ipv6.devconf_all);
7083 }
7084
7085 static struct pernet_operations addrconf_ops = {
7086         .init = addrconf_init_net,
7087         .exit = addrconf_exit_net,
7088 };
7089
7090 static struct rtnl_af_ops inet6_ops __read_mostly = {
7091         .family           = AF_INET6,
7092         .fill_link_af     = inet6_fill_link_af,
7093         .get_link_af_size = inet6_get_link_af_size,
7094         .validate_link_af = inet6_validate_link_af,
7095         .set_link_af      = inet6_set_link_af,
7096 };
7097
7098 /*
7099  *      Init / cleanup code
7100  */
7101
7102 int __init addrconf_init(void)
7103 {
7104         struct inet6_dev *idev;
7105         int i, err;
7106
7107         err = ipv6_addr_label_init();
7108         if (err < 0) {
7109                 pr_crit("%s: cannot initialize default policy table: %d\n",
7110                         __func__, err);
7111                 goto out;
7112         }
7113
7114         err = register_pernet_subsys(&addrconf_ops);
7115         if (err < 0)
7116                 goto out_addrlabel;
7117
7118         addrconf_wq = create_workqueue("ipv6_addrconf");
7119         if (!addrconf_wq) {
7120                 err = -ENOMEM;
7121                 goto out_nowq;
7122         }
7123
7124         /* The addrconf netdev notifier requires that loopback_dev
7125          * has it's ipv6 private information allocated and setup
7126          * before it can bring up and give link-local addresses
7127          * to other devices which are up.
7128          *
7129          * Unfortunately, loopback_dev is not necessarily the first
7130          * entry in the global dev_base list of net devices.  In fact,
7131          * it is likely to be the very last entry on that list.
7132          * So this causes the notifier registry below to try and
7133          * give link-local addresses to all devices besides loopback_dev
7134          * first, then loopback_dev, which cases all the non-loopback_dev
7135          * devices to fail to get a link-local address.
7136          *
7137          * So, as a temporary fix, allocate the ipv6 structure for
7138          * loopback_dev first by hand.
7139          * Longer term, all of the dependencies ipv6 has upon the loopback
7140          * device and it being up should be removed.
7141          */
7142         rtnl_lock();
7143         idev = ipv6_add_dev(init_net.loopback_dev);
7144         rtnl_unlock();
7145         if (IS_ERR(idev)) {
7146                 err = PTR_ERR(idev);
7147                 goto errlo;
7148         }
7149
7150         ip6_route_init_special_entries();
7151
7152         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7153                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7154
7155         register_netdevice_notifier(&ipv6_dev_notf);
7156
7157         addrconf_verify();
7158
7159         rtnl_af_register(&inet6_ops);
7160
7161         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7162                                    NULL, inet6_dump_ifinfo, 0);
7163         if (err < 0)
7164                 goto errout;
7165
7166         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7167                                    inet6_rtm_newaddr, NULL, 0);
7168         if (err < 0)
7169                 goto errout;
7170         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7171                                    inet6_rtm_deladdr, NULL, 0);
7172         if (err < 0)
7173                 goto errout;
7174         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7175                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7176                                    RTNL_FLAG_DOIT_UNLOCKED);
7177         if (err < 0)
7178                 goto errout;
7179         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7180                                    NULL, inet6_dump_ifmcaddr, 0);
7181         if (err < 0)
7182                 goto errout;
7183         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7184                                    NULL, inet6_dump_ifacaddr, 0);
7185         if (err < 0)
7186                 goto errout;
7187         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7188                                    inet6_netconf_get_devconf,
7189                                    inet6_netconf_dump_devconf,
7190                                    RTNL_FLAG_DOIT_UNLOCKED);
7191         if (err < 0)
7192                 goto errout;
7193         err = ipv6_addr_label_rtnl_register();
7194         if (err < 0)
7195                 goto errout;
7196
7197         return 0;
7198 errout:
7199         rtnl_unregister_all(PF_INET6);
7200         rtnl_af_unregister(&inet6_ops);
7201         unregister_netdevice_notifier(&ipv6_dev_notf);
7202 errlo:
7203         destroy_workqueue(addrconf_wq);
7204 out_nowq:
7205         unregister_pernet_subsys(&addrconf_ops);
7206 out_addrlabel:
7207         ipv6_addr_label_cleanup();
7208 out:
7209         return err;
7210 }
7211
7212 void addrconf_cleanup(void)
7213 {
7214         struct net_device *dev;
7215         int i;
7216
7217         unregister_netdevice_notifier(&ipv6_dev_notf);
7218         unregister_pernet_subsys(&addrconf_ops);
7219         ipv6_addr_label_cleanup();
7220
7221         rtnl_af_unregister(&inet6_ops);
7222
7223         rtnl_lock();
7224
7225         /* clean dev list */
7226         for_each_netdev(&init_net, dev) {
7227                 if (__in6_dev_get(dev) == NULL)
7228                         continue;
7229                 addrconf_ifdown(dev, true);
7230         }
7231         addrconf_ifdown(init_net.loopback_dev, true);
7232
7233         /*
7234          *      Check hash table.
7235          */
7236         spin_lock_bh(&addrconf_hash_lock);
7237         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7238                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7239         spin_unlock_bh(&addrconf_hash_lock);
7240         cancel_delayed_work(&addr_chk_work);
7241         rtnl_unlock();
7242
7243         destroy_workqueue(addrconf_wq);
7244 }