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 static struct net_device *
1897 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1898                           const struct net_device *dev, bool skip_dev_check,
1899                           int strict, u32 banned_flags)
1900 {
1901         unsigned int hash = inet6_addr_hash(net, addr);
1902         struct net_device *l3mdev, *ndev;
1903         struct inet6_ifaddr *ifp;
1904         u32 ifp_flags;
1905
1906         rcu_read_lock();
1907
1908         l3mdev = l3mdev_master_dev_rcu(dev);
1909         if (skip_dev_check)
1910                 dev = NULL;
1911
1912         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1913                 ndev = ifp->idev->dev;
1914                 if (!net_eq(dev_net(ndev), net))
1915                         continue;
1916
1917                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1918                         continue;
1919
1920                 /* Decouple optimistic from tentative for evaluation here.
1921                  * Ban optimistic addresses explicitly, when required.
1922                  */
1923                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1924                             ? (ifp->flags&~IFA_F_TENTATIVE)
1925                             : ifp->flags;
1926                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1927                     !(ifp_flags&banned_flags) &&
1928                     (!dev || ndev == dev ||
1929                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1930                         rcu_read_unlock();
1931                         return ndev;
1932                 }
1933         }
1934
1935         rcu_read_unlock();
1936         return NULL;
1937 }
1938
1939 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1940                             const struct net_device *dev, bool skip_dev_check,
1941                             int strict, u32 banned_flags)
1942 {
1943         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1944                                          strict, banned_flags) ? 1 : 0;
1945 }
1946 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1947
1948
1949 /* Compares an address/prefix_len with addresses on device @dev.
1950  * If one is found it returns true.
1951  */
1952 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1953         const unsigned int prefix_len, struct net_device *dev)
1954 {
1955         const struct inet6_ifaddr *ifa;
1956         const struct inet6_dev *idev;
1957         bool ret = false;
1958
1959         rcu_read_lock();
1960         idev = __in6_dev_get(dev);
1961         if (idev) {
1962                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1963                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1964                         if (ret)
1965                                 break;
1966                 }
1967         }
1968         rcu_read_unlock();
1969
1970         return ret;
1971 }
1972 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1973
1974 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1975 {
1976         const struct inet6_ifaddr *ifa;
1977         const struct inet6_dev *idev;
1978         int     onlink;
1979
1980         onlink = 0;
1981         rcu_read_lock();
1982         idev = __in6_dev_get(dev);
1983         if (idev) {
1984                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1985                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1986                                                    ifa->prefix_len);
1987                         if (onlink)
1988                                 break;
1989                 }
1990         }
1991         rcu_read_unlock();
1992         return onlink;
1993 }
1994 EXPORT_SYMBOL(ipv6_chk_prefix);
1995
1996 /**
1997  * ipv6_dev_find - find the first device with a given source address.
1998  * @net: the net namespace
1999  * @addr: the source address
2000  * @dev: used to find the L3 domain of interest
2001  *
2002  * The caller should be protected by RCU, or RTNL.
2003  */
2004 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2005                                  struct net_device *dev)
2006 {
2007         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2008                                          IFA_F_TENTATIVE);
2009 }
2010 EXPORT_SYMBOL(ipv6_dev_find);
2011
2012 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2013                                      struct net_device *dev, int strict)
2014 {
2015         unsigned int hash = inet6_addr_hash(net, addr);
2016         struct inet6_ifaddr *ifp, *result = NULL;
2017
2018         rcu_read_lock();
2019         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2020                 if (!net_eq(dev_net(ifp->idev->dev), net))
2021                         continue;
2022                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2023                         if (!dev || ifp->idev->dev == dev ||
2024                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2025                                 result = ifp;
2026                                 in6_ifa_hold(ifp);
2027                                 break;
2028                         }
2029                 }
2030         }
2031         rcu_read_unlock();
2032
2033         return result;
2034 }
2035
2036 /* Gets referenced address, destroys ifaddr */
2037
2038 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2039 {
2040         if (dad_failed)
2041                 ifp->flags |= IFA_F_DADFAILED;
2042
2043         if (ifp->flags&IFA_F_TEMPORARY) {
2044                 struct inet6_ifaddr *ifpub;
2045                 spin_lock_bh(&ifp->lock);
2046                 ifpub = ifp->ifpub;
2047                 if (ifpub) {
2048                         in6_ifa_hold(ifpub);
2049                         spin_unlock_bh(&ifp->lock);
2050                         ipv6_create_tempaddr(ifpub, true);
2051                         in6_ifa_put(ifpub);
2052                 } else {
2053                         spin_unlock_bh(&ifp->lock);
2054                 }
2055                 ipv6_del_addr(ifp);
2056         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2057                 spin_lock_bh(&ifp->lock);
2058                 addrconf_del_dad_work(ifp);
2059                 ifp->flags |= IFA_F_TENTATIVE;
2060                 if (dad_failed)
2061                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2062                 spin_unlock_bh(&ifp->lock);
2063                 if (dad_failed)
2064                         ipv6_ifa_notify(0, ifp);
2065                 in6_ifa_put(ifp);
2066         } else {
2067                 ipv6_del_addr(ifp);
2068         }
2069 }
2070
2071 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2072 {
2073         int err = -ENOENT;
2074
2075         spin_lock_bh(&ifp->lock);
2076         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2077                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2078                 err = 0;
2079         }
2080         spin_unlock_bh(&ifp->lock);
2081
2082         return err;
2083 }
2084
2085 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2086 {
2087         struct inet6_dev *idev = ifp->idev;
2088         struct net *net = dev_net(ifp->idev->dev);
2089
2090         if (addrconf_dad_end(ifp)) {
2091                 in6_ifa_put(ifp);
2092                 return;
2093         }
2094
2095         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2096                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2097
2098         spin_lock_bh(&ifp->lock);
2099
2100         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2101                 struct in6_addr new_addr;
2102                 struct inet6_ifaddr *ifp2;
2103                 int retries = ifp->stable_privacy_retry + 1;
2104                 struct ifa6_config cfg = {
2105                         .pfx = &new_addr,
2106                         .plen = ifp->prefix_len,
2107                         .ifa_flags = ifp->flags,
2108                         .valid_lft = ifp->valid_lft,
2109                         .preferred_lft = ifp->prefered_lft,
2110                         .scope = ifp->scope,
2111                 };
2112
2113                 if (retries > net->ipv6.sysctl.idgen_retries) {
2114                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2115                                              ifp->idev->dev->name);
2116                         goto errdad;
2117                 }
2118
2119                 new_addr = ifp->addr;
2120                 if (ipv6_generate_stable_address(&new_addr, retries,
2121                                                  idev))
2122                         goto errdad;
2123
2124                 spin_unlock_bh(&ifp->lock);
2125
2126                 if (idev->cnf.max_addresses &&
2127                     ipv6_count_addresses(idev) >=
2128                     idev->cnf.max_addresses)
2129                         goto lock_errdad;
2130
2131                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2132                                      ifp->idev->dev->name);
2133
2134                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2135                 if (IS_ERR(ifp2))
2136                         goto lock_errdad;
2137
2138                 spin_lock_bh(&ifp2->lock);
2139                 ifp2->stable_privacy_retry = retries;
2140                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2141                 spin_unlock_bh(&ifp2->lock);
2142
2143                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2144                 in6_ifa_put(ifp2);
2145 lock_errdad:
2146                 spin_lock_bh(&ifp->lock);
2147         }
2148
2149 errdad:
2150         /* transition from _POSTDAD to _ERRDAD */
2151         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2152         spin_unlock_bh(&ifp->lock);
2153
2154         addrconf_mod_dad_work(ifp, 0);
2155         in6_ifa_put(ifp);
2156 }
2157
2158 /* Join to solicited addr multicast group.
2159  * caller must hold RTNL */
2160 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2161 {
2162         struct in6_addr maddr;
2163
2164         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2165                 return;
2166
2167         addrconf_addr_solict_mult(addr, &maddr);
2168         ipv6_dev_mc_inc(dev, &maddr);
2169 }
2170
2171 /* caller must hold RTNL */
2172 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2173 {
2174         struct in6_addr maddr;
2175
2176         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2177                 return;
2178
2179         addrconf_addr_solict_mult(addr, &maddr);
2180         __ipv6_dev_mc_dec(idev, &maddr);
2181 }
2182
2183 /* caller must hold RTNL */
2184 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2185 {
2186         struct in6_addr addr;
2187
2188         if (ifp->prefix_len >= 127) /* RFC 6164 */
2189                 return;
2190         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2191         if (ipv6_addr_any(&addr))
2192                 return;
2193         __ipv6_dev_ac_inc(ifp->idev, &addr);
2194 }
2195
2196 /* caller must hold RTNL */
2197 static void addrconf_leave_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_dec(ifp->idev, &addr);
2207 }
2208
2209 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2210 {
2211         switch (dev->addr_len) {
2212         case ETH_ALEN:
2213                 memcpy(eui, dev->dev_addr, 3);
2214                 eui[3] = 0xFF;
2215                 eui[4] = 0xFE;
2216                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2217                 break;
2218         case EUI64_ADDR_LEN:
2219                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2220                 eui[0] ^= 2;
2221                 break;
2222         default:
2223                 return -1;
2224         }
2225
2226         return 0;
2227 }
2228
2229 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2230 {
2231         union fwnet_hwaddr *ha;
2232
2233         if (dev->addr_len != FWNET_ALEN)
2234                 return -1;
2235
2236         ha = (union fwnet_hwaddr *)dev->dev_addr;
2237
2238         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2239         eui[0] ^= 2;
2240         return 0;
2241 }
2242
2243 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2244 {
2245         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2246         if (dev->addr_len != ARCNET_ALEN)
2247                 return -1;
2248         memset(eui, 0, 7);
2249         eui[7] = *(u8 *)dev->dev_addr;
2250         return 0;
2251 }
2252
2253 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2254 {
2255         if (dev->addr_len != INFINIBAND_ALEN)
2256                 return -1;
2257         memcpy(eui, dev->dev_addr + 12, 8);
2258         eui[0] |= 2;
2259         return 0;
2260 }
2261
2262 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2263 {
2264         if (addr == 0)
2265                 return -1;
2266         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2267                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2268                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2269                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2270                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2271                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2272         eui[1] = 0;
2273         eui[2] = 0x5E;
2274         eui[3] = 0xFE;
2275         memcpy(eui + 4, &addr, 4);
2276         return 0;
2277 }
2278
2279 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2280 {
2281         if (dev->priv_flags & IFF_ISATAP)
2282                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2283         return -1;
2284 }
2285
2286 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2287 {
2288         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2289 }
2290
2291 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2292 {
2293         memcpy(eui, dev->perm_addr, 3);
2294         memcpy(eui + 5, dev->perm_addr + 3, 3);
2295         eui[3] = 0xFF;
2296         eui[4] = 0xFE;
2297         eui[0] ^= 2;
2298         return 0;
2299 }
2300
2301 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2302 {
2303         switch (dev->type) {
2304         case ARPHRD_ETHER:
2305         case ARPHRD_FDDI:
2306                 return addrconf_ifid_eui48(eui, dev);
2307         case ARPHRD_ARCNET:
2308                 return addrconf_ifid_arcnet(eui, dev);
2309         case ARPHRD_INFINIBAND:
2310                 return addrconf_ifid_infiniband(eui, dev);
2311         case ARPHRD_SIT:
2312                 return addrconf_ifid_sit(eui, dev);
2313         case ARPHRD_IPGRE:
2314         case ARPHRD_TUNNEL:
2315                 return addrconf_ifid_gre(eui, dev);
2316         case ARPHRD_6LOWPAN:
2317                 return addrconf_ifid_6lowpan(eui, dev);
2318         case ARPHRD_IEEE1394:
2319                 return addrconf_ifid_ieee1394(eui, dev);
2320         case ARPHRD_TUNNEL6:
2321         case ARPHRD_IP6GRE:
2322         case ARPHRD_RAWIP:
2323                 return addrconf_ifid_ip6tnl(eui, dev);
2324         }
2325         return -1;
2326 }
2327
2328 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2329 {
2330         int err = -1;
2331         struct inet6_ifaddr *ifp;
2332
2333         read_lock_bh(&idev->lock);
2334         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2335                 if (ifp->scope > IFA_LINK)
2336                         break;
2337                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2338                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2339                         err = 0;
2340                         break;
2341                 }
2342         }
2343         read_unlock_bh(&idev->lock);
2344         return err;
2345 }
2346
2347 /* Generation of a randomized Interface Identifier
2348  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2349  */
2350
2351 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2352 {
2353 regen:
2354         get_random_bytes(&addr->s6_addr[8], 8);
2355
2356         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2357          * check if generated address is not inappropriate:
2358          *
2359          * - Reserved IPv6 Interface Identifers
2360          * - XXX: already assigned to an address on the device
2361          */
2362
2363         /* Subnet-router anycast: 0000:0000:0000:0000 */
2364         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2365                 goto regen;
2366
2367         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2368          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2369          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2370          */
2371         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2372             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2373                 goto regen;
2374
2375         /* Reserved subnet anycast addresses */
2376         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2377             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2378                 goto regen;
2379 }
2380
2381 /*
2382  *      Add prefix route.
2383  */
2384
2385 static void
2386 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2387                       struct net_device *dev, unsigned long expires,
2388                       u32 flags, gfp_t gfp_flags)
2389 {
2390         struct fib6_config cfg = {
2391                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2392                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2393                 .fc_ifindex = dev->ifindex,
2394                 .fc_expires = expires,
2395                 .fc_dst_len = plen,
2396                 .fc_flags = RTF_UP | flags,
2397                 .fc_nlinfo.nl_net = dev_net(dev),
2398                 .fc_protocol = RTPROT_KERNEL,
2399                 .fc_type = RTN_UNICAST,
2400         };
2401
2402         cfg.fc_dst = *pfx;
2403
2404         /* Prevent useless cloning on PtP SIT.
2405            This thing is done here expecting that the whole
2406            class of non-broadcast devices need not cloning.
2407          */
2408 #if IS_ENABLED(CONFIG_IPV6_SIT)
2409         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2410                 cfg.fc_flags |= RTF_NONEXTHOP;
2411 #endif
2412
2413         ip6_route_add(&cfg, gfp_flags, NULL);
2414 }
2415
2416
2417 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2418                                                   int plen,
2419                                                   const struct net_device *dev,
2420                                                   u32 flags, u32 noflags,
2421                                                   bool no_gw)
2422 {
2423         struct fib6_node *fn;
2424         struct fib6_info *rt = NULL;
2425         struct fib6_table *table;
2426         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2427
2428         table = fib6_get_table(dev_net(dev), tb_id);
2429         if (!table)
2430                 return NULL;
2431
2432         rcu_read_lock();
2433         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2434         if (!fn)
2435                 goto out;
2436
2437         for_each_fib6_node_rt_rcu(fn) {
2438                 /* prefix routes only use builtin fib6_nh */
2439                 if (rt->nh)
2440                         continue;
2441
2442                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2443                         continue;
2444                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2445                         continue;
2446                 if ((rt->fib6_flags & flags) != flags)
2447                         continue;
2448                 if ((rt->fib6_flags & noflags) != 0)
2449                         continue;
2450                 if (!fib6_info_hold_safe(rt))
2451                         continue;
2452                 break;
2453         }
2454 out:
2455         rcu_read_unlock();
2456         return rt;
2457 }
2458
2459
2460 /* Create "default" multicast route to the interface */
2461
2462 static void addrconf_add_mroute(struct net_device *dev)
2463 {
2464         struct fib6_config cfg = {
2465                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2466                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2467                 .fc_ifindex = dev->ifindex,
2468                 .fc_dst_len = 8,
2469                 .fc_flags = RTF_UP,
2470                 .fc_type = RTN_MULTICAST,
2471                 .fc_nlinfo.nl_net = dev_net(dev),
2472                 .fc_protocol = RTPROT_KERNEL,
2473         };
2474
2475         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2476
2477         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2478 }
2479
2480 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2481 {
2482         struct inet6_dev *idev;
2483
2484         ASSERT_RTNL();
2485
2486         idev = ipv6_find_idev(dev);
2487         if (IS_ERR(idev))
2488                 return idev;
2489
2490         if (idev->cnf.disable_ipv6)
2491                 return ERR_PTR(-EACCES);
2492
2493         /* Add default multicast route */
2494         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2495                 addrconf_add_mroute(dev);
2496
2497         return idev;
2498 }
2499
2500 static void manage_tempaddrs(struct inet6_dev *idev,
2501                              struct inet6_ifaddr *ifp,
2502                              __u32 valid_lft, __u32 prefered_lft,
2503                              bool create, unsigned long now)
2504 {
2505         u32 flags;
2506         struct inet6_ifaddr *ift;
2507
2508         read_lock_bh(&idev->lock);
2509         /* update all temporary addresses in the list */
2510         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2511                 int age, max_valid, max_prefered;
2512
2513                 if (ifp != ift->ifpub)
2514                         continue;
2515
2516                 /* RFC 4941 section 3.3:
2517                  * If a received option will extend the lifetime of a public
2518                  * address, the lifetimes of temporary addresses should
2519                  * be extended, subject to the overall constraint that no
2520                  * temporary addresses should ever remain "valid" or "preferred"
2521                  * for a time longer than (TEMP_VALID_LIFETIME) or
2522                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2523                  */
2524                 age = (now - ift->cstamp) / HZ;
2525                 max_valid = idev->cnf.temp_valid_lft - age;
2526                 if (max_valid < 0)
2527                         max_valid = 0;
2528
2529                 max_prefered = idev->cnf.temp_prefered_lft -
2530                                idev->desync_factor - age;
2531                 if (max_prefered < 0)
2532                         max_prefered = 0;
2533
2534                 if (valid_lft > max_valid)
2535                         valid_lft = max_valid;
2536
2537                 if (prefered_lft > max_prefered)
2538                         prefered_lft = max_prefered;
2539
2540                 spin_lock(&ift->lock);
2541                 flags = ift->flags;
2542                 ift->valid_lft = valid_lft;
2543                 ift->prefered_lft = prefered_lft;
2544                 ift->tstamp = now;
2545                 if (prefered_lft > 0)
2546                         ift->flags &= ~IFA_F_DEPRECATED;
2547
2548                 spin_unlock(&ift->lock);
2549                 if (!(flags&IFA_F_TENTATIVE))
2550                         ipv6_ifa_notify(0, ift);
2551         }
2552
2553         if ((create || list_empty(&idev->tempaddr_list)) &&
2554             idev->cnf.use_tempaddr > 0) {
2555                 /* When a new public address is created as described
2556                  * in [ADDRCONF], also create a new temporary address.
2557                  * Also create a temporary address if it's enabled but
2558                  * no temporary address currently exists.
2559                  */
2560                 read_unlock_bh(&idev->lock);
2561                 ipv6_create_tempaddr(ifp, false);
2562         } else {
2563                 read_unlock_bh(&idev->lock);
2564         }
2565 }
2566
2567 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2568 {
2569         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2570                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2571 }
2572
2573 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2574                                  const struct prefix_info *pinfo,
2575                                  struct inet6_dev *in6_dev,
2576                                  const struct in6_addr *addr, int addr_type,
2577                                  u32 addr_flags, bool sllao, bool tokenized,
2578                                  __u32 valid_lft, u32 prefered_lft)
2579 {
2580         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2581         int create = 0;
2582
2583         if (!ifp && valid_lft) {
2584                 int max_addresses = in6_dev->cnf.max_addresses;
2585                 struct ifa6_config cfg = {
2586                         .pfx = addr,
2587                         .plen = pinfo->prefix_len,
2588                         .ifa_flags = addr_flags,
2589                         .valid_lft = valid_lft,
2590                         .preferred_lft = prefered_lft,
2591                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2592                 };
2593
2594 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2595                 if ((net->ipv6.devconf_all->optimistic_dad ||
2596                      in6_dev->cnf.optimistic_dad) &&
2597                     !net->ipv6.devconf_all->forwarding && sllao)
2598                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2599 #endif
2600
2601                 /* Do not allow to create too much of autoconfigured
2602                  * addresses; this would be too easy way to crash kernel.
2603                  */
2604                 if (!max_addresses ||
2605                     ipv6_count_addresses(in6_dev) < max_addresses)
2606                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2607
2608                 if (IS_ERR_OR_NULL(ifp))
2609                         return -1;
2610
2611                 create = 1;
2612                 spin_lock_bh(&ifp->lock);
2613                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2614                 ifp->cstamp = jiffies;
2615                 ifp->tokenized = tokenized;
2616                 spin_unlock_bh(&ifp->lock);
2617                 addrconf_dad_start(ifp);
2618         }
2619
2620         if (ifp) {
2621                 u32 flags;
2622                 unsigned long now;
2623                 u32 stored_lft;
2624
2625                 /* Update lifetime (RFC4862 5.5.3 e)
2626                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2627                  * improve the reaction of SLAAC to renumbering events
2628                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2629                  */
2630                 spin_lock_bh(&ifp->lock);
2631                 now = jiffies;
2632                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2633                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2634                 else
2635                         stored_lft = 0;
2636
2637                 if (!create && stored_lft) {
2638                         ifp->valid_lft = valid_lft;
2639                         ifp->prefered_lft = prefered_lft;
2640                         ifp->tstamp = now;
2641                         flags = ifp->flags;
2642                         ifp->flags &= ~IFA_F_DEPRECATED;
2643                         spin_unlock_bh(&ifp->lock);
2644
2645                         if (!(flags&IFA_F_TENTATIVE))
2646                                 ipv6_ifa_notify(0, ifp);
2647                 } else
2648                         spin_unlock_bh(&ifp->lock);
2649
2650                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2651                                  create, now);
2652
2653                 in6_ifa_put(ifp);
2654                 addrconf_verify();
2655         }
2656
2657         return 0;
2658 }
2659 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2660
2661 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2662 {
2663         struct prefix_info *pinfo;
2664         __u32 valid_lft;
2665         __u32 prefered_lft;
2666         int addr_type, err;
2667         u32 addr_flags = 0;
2668         struct inet6_dev *in6_dev;
2669         struct net *net = dev_net(dev);
2670
2671         pinfo = (struct prefix_info *) opt;
2672
2673         if (len < sizeof(struct prefix_info)) {
2674                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2675                 return;
2676         }
2677
2678         /*
2679          *      Validation checks ([ADDRCONF], page 19)
2680          */
2681
2682         addr_type = ipv6_addr_type(&pinfo->prefix);
2683
2684         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2685                 return;
2686
2687         valid_lft = ntohl(pinfo->valid);
2688         prefered_lft = ntohl(pinfo->prefered);
2689
2690         if (prefered_lft > valid_lft) {
2691                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2692                 return;
2693         }
2694
2695         in6_dev = in6_dev_get(dev);
2696
2697         if (!in6_dev) {
2698                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2699                                     dev->name);
2700                 return;
2701         }
2702
2703         /*
2704          *      Two things going on here:
2705          *      1) Add routes for on-link prefixes
2706          *      2) Configure prefixes with the auto flag set
2707          */
2708
2709         if (pinfo->onlink) {
2710                 struct fib6_info *rt;
2711                 unsigned long rt_expires;
2712
2713                 /* Avoid arithmetic overflow. Really, we could
2714                  * save rt_expires in seconds, likely valid_lft,
2715                  * but it would require division in fib gc, that it
2716                  * not good.
2717                  */
2718                 if (HZ > USER_HZ)
2719                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2720                 else
2721                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2722
2723                 if (addrconf_finite_timeout(rt_expires))
2724                         rt_expires *= HZ;
2725
2726                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2727                                                pinfo->prefix_len,
2728                                                dev,
2729                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2730                                                RTF_DEFAULT, true);
2731
2732                 if (rt) {
2733                         /* Autoconf prefix route */
2734                         if (valid_lft == 0) {
2735                                 ip6_del_rt(net, rt, false);
2736                                 rt = NULL;
2737                         } else if (addrconf_finite_timeout(rt_expires)) {
2738                                 /* not infinity */
2739                                 fib6_set_expires(rt, jiffies + rt_expires);
2740                         } else {
2741                                 fib6_clean_expires(rt);
2742                         }
2743                 } else if (valid_lft) {
2744                         clock_t expires = 0;
2745                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2746                         if (addrconf_finite_timeout(rt_expires)) {
2747                                 /* not infinity */
2748                                 flags |= RTF_EXPIRES;
2749                                 expires = jiffies_to_clock_t(rt_expires);
2750                         }
2751                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2752                                               0, dev, expires, flags,
2753                                               GFP_ATOMIC);
2754                 }
2755                 fib6_info_release(rt);
2756         }
2757
2758         /* Try to figure out our local address for this prefix */
2759
2760         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2761                 struct in6_addr addr;
2762                 bool tokenized = false, dev_addr_generated = false;
2763
2764                 if (pinfo->prefix_len == 64) {
2765                         memcpy(&addr, &pinfo->prefix, 8);
2766
2767                         if (!ipv6_addr_any(&in6_dev->token)) {
2768                                 read_lock_bh(&in6_dev->lock);
2769                                 memcpy(addr.s6_addr + 8,
2770                                        in6_dev->token.s6_addr + 8, 8);
2771                                 read_unlock_bh(&in6_dev->lock);
2772                                 tokenized = true;
2773                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2774                                    !ipv6_generate_stable_address(&addr, 0,
2775                                                                  in6_dev)) {
2776                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2777                                 goto ok;
2778                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2779                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2780                                 goto put;
2781                         } else {
2782                                 dev_addr_generated = true;
2783                         }
2784                         goto ok;
2785                 }
2786                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2787                                     pinfo->prefix_len);
2788                 goto put;
2789
2790 ok:
2791                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2792                                                    &addr, addr_type,
2793                                                    addr_flags, sllao,
2794                                                    tokenized, valid_lft,
2795                                                    prefered_lft);
2796                 if (err)
2797                         goto put;
2798
2799                 /* Ignore error case here because previous prefix add addr was
2800                  * successful which will be notified.
2801                  */
2802                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2803                                               addr_type, addr_flags, sllao,
2804                                               tokenized, valid_lft,
2805                                               prefered_lft,
2806                                               dev_addr_generated);
2807         }
2808         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2809 put:
2810         in6_dev_put(in6_dev);
2811 }
2812
2813 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2814                 struct in6_ifreq *ireq)
2815 {
2816         struct ip_tunnel_parm p = { };
2817         int err;
2818
2819         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2820                 return -EADDRNOTAVAIL;
2821
2822         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2823         p.iph.version = 4;
2824         p.iph.ihl = 5;
2825         p.iph.protocol = IPPROTO_IPV6;
2826         p.iph.ttl = 64;
2827
2828         if (!dev->netdev_ops->ndo_tunnel_ctl)
2829                 return -EOPNOTSUPP;
2830         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2831         if (err)
2832                 return err;
2833
2834         dev = __dev_get_by_name(net, p.name);
2835         if (!dev)
2836                 return -ENOBUFS;
2837         return dev_open(dev, NULL);
2838 }
2839
2840 /*
2841  *      Set destination address.
2842  *      Special case for SIT interfaces where we create a new "virtual"
2843  *      device.
2844  */
2845 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2846 {
2847         struct net_device *dev;
2848         struct in6_ifreq ireq;
2849         int err = -ENODEV;
2850
2851         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2852                 return -ENODEV;
2853         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2854                 return -EFAULT;
2855
2856         rtnl_lock();
2857         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2858         if (dev && dev->type == ARPHRD_SIT)
2859                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2860         rtnl_unlock();
2861         return err;
2862 }
2863
2864 static int ipv6_mc_config(struct sock *sk, bool join,
2865                           const struct in6_addr *addr, int ifindex)
2866 {
2867         int ret;
2868
2869         ASSERT_RTNL();
2870
2871         lock_sock(sk);
2872         if (join)
2873                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2874         else
2875                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2876         release_sock(sk);
2877
2878         return ret;
2879 }
2880
2881 /*
2882  *      Manual configuration of address on an interface
2883  */
2884 static int inet6_addr_add(struct net *net, int ifindex,
2885                           struct ifa6_config *cfg,
2886                           struct netlink_ext_ack *extack)
2887 {
2888         struct inet6_ifaddr *ifp;
2889         struct inet6_dev *idev;
2890         struct net_device *dev;
2891         unsigned long timeout;
2892         clock_t expires;
2893         u32 flags;
2894
2895         ASSERT_RTNL();
2896
2897         if (cfg->plen > 128)
2898                 return -EINVAL;
2899
2900         /* check the lifetime */
2901         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2902                 return -EINVAL;
2903
2904         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2905                 return -EINVAL;
2906
2907         dev = __dev_get_by_index(net, ifindex);
2908         if (!dev)
2909                 return -ENODEV;
2910
2911         idev = addrconf_add_dev(dev);
2912         if (IS_ERR(idev))
2913                 return PTR_ERR(idev);
2914
2915         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2916                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2917                                          true, cfg->pfx, ifindex);
2918
2919                 if (ret < 0)
2920                         return ret;
2921         }
2922
2923         cfg->scope = ipv6_addr_scope(cfg->pfx);
2924
2925         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2926         if (addrconf_finite_timeout(timeout)) {
2927                 expires = jiffies_to_clock_t(timeout * HZ);
2928                 cfg->valid_lft = timeout;
2929                 flags = RTF_EXPIRES;
2930         } else {
2931                 expires = 0;
2932                 flags = 0;
2933                 cfg->ifa_flags |= IFA_F_PERMANENT;
2934         }
2935
2936         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2937         if (addrconf_finite_timeout(timeout)) {
2938                 if (timeout == 0)
2939                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2940                 cfg->preferred_lft = timeout;
2941         }
2942
2943         ifp = ipv6_add_addr(idev, cfg, true, extack);
2944         if (!IS_ERR(ifp)) {
2945                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2946                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2947                                               ifp->rt_priority, dev, expires,
2948                                               flags, GFP_KERNEL);
2949                 }
2950
2951                 /* Send a netlink notification if DAD is enabled and
2952                  * optimistic flag is not set
2953                  */
2954                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2955                         ipv6_ifa_notify(0, ifp);
2956                 /*
2957                  * Note that section 3.1 of RFC 4429 indicates
2958                  * that the Optimistic flag should not be set for
2959                  * manually configured addresses
2960                  */
2961                 addrconf_dad_start(ifp);
2962                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2963                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2964                                          cfg->preferred_lft, true, jiffies);
2965                 in6_ifa_put(ifp);
2966                 addrconf_verify_rtnl();
2967                 return 0;
2968         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2969                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2970                                cfg->pfx, ifindex);
2971         }
2972
2973         return PTR_ERR(ifp);
2974 }
2975
2976 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2977                           const struct in6_addr *pfx, unsigned int plen)
2978 {
2979         struct inet6_ifaddr *ifp;
2980         struct inet6_dev *idev;
2981         struct net_device *dev;
2982
2983         if (plen > 128)
2984                 return -EINVAL;
2985
2986         dev = __dev_get_by_index(net, ifindex);
2987         if (!dev)
2988                 return -ENODEV;
2989
2990         idev = __in6_dev_get(dev);
2991         if (!idev)
2992                 return -ENXIO;
2993
2994         read_lock_bh(&idev->lock);
2995         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2996                 if (ifp->prefix_len == plen &&
2997                     ipv6_addr_equal(pfx, &ifp->addr)) {
2998                         in6_ifa_hold(ifp);
2999                         read_unlock_bh(&idev->lock);
3000
3001                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3002                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3003                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3004                                                  jiffies);
3005                         ipv6_del_addr(ifp);
3006                         addrconf_verify_rtnl();
3007                         if (ipv6_addr_is_multicast(pfx)) {
3008                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3009                                                false, pfx, dev->ifindex);
3010                         }
3011                         return 0;
3012                 }
3013         }
3014         read_unlock_bh(&idev->lock);
3015         return -EADDRNOTAVAIL;
3016 }
3017
3018
3019 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3020 {
3021         struct ifa6_config cfg = {
3022                 .ifa_flags = IFA_F_PERMANENT,
3023                 .preferred_lft = INFINITY_LIFE_TIME,
3024                 .valid_lft = INFINITY_LIFE_TIME,
3025         };
3026         struct in6_ifreq ireq;
3027         int err;
3028
3029         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3030                 return -EPERM;
3031
3032         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3033                 return -EFAULT;
3034
3035         cfg.pfx = &ireq.ifr6_addr;
3036         cfg.plen = ireq.ifr6_prefixlen;
3037
3038         rtnl_lock();
3039         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3040         rtnl_unlock();
3041         return err;
3042 }
3043
3044 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3045 {
3046         struct in6_ifreq ireq;
3047         int err;
3048
3049         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3050                 return -EPERM;
3051
3052         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3053                 return -EFAULT;
3054
3055         rtnl_lock();
3056         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3057                              ireq.ifr6_prefixlen);
3058         rtnl_unlock();
3059         return err;
3060 }
3061
3062 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3063                      int plen, int scope)
3064 {
3065         struct inet6_ifaddr *ifp;
3066         struct ifa6_config cfg = {
3067                 .pfx = addr,
3068                 .plen = plen,
3069                 .ifa_flags = IFA_F_PERMANENT,
3070                 .valid_lft = INFINITY_LIFE_TIME,
3071                 .preferred_lft = INFINITY_LIFE_TIME,
3072                 .scope = scope
3073         };
3074
3075         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3076         if (!IS_ERR(ifp)) {
3077                 spin_lock_bh(&ifp->lock);
3078                 ifp->flags &= ~IFA_F_TENTATIVE;
3079                 spin_unlock_bh(&ifp->lock);
3080                 rt_genid_bump_ipv6(dev_net(idev->dev));
3081                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3082                 in6_ifa_put(ifp);
3083         }
3084 }
3085
3086 #if IS_ENABLED(CONFIG_IPV6_SIT)
3087 static void sit_add_v4_addrs(struct inet6_dev *idev)
3088 {
3089         struct in6_addr addr;
3090         struct net_device *dev;
3091         struct net *net = dev_net(idev->dev);
3092         int scope, plen;
3093         u32 pflags = 0;
3094
3095         ASSERT_RTNL();
3096
3097         memset(&addr, 0, sizeof(struct in6_addr));
3098         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3099
3100         if (idev->dev->flags&IFF_POINTOPOINT) {
3101                 addr.s6_addr32[0] = htonl(0xfe800000);
3102                 scope = IFA_LINK;
3103                 plen = 64;
3104         } else {
3105                 scope = IPV6_ADDR_COMPATv4;
3106                 plen = 96;
3107                 pflags |= RTF_NONEXTHOP;
3108         }
3109
3110         if (addr.s6_addr32[3]) {
3111                 add_addr(idev, &addr, plen, scope);
3112                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3113                                       GFP_KERNEL);
3114                 return;
3115         }
3116
3117         for_each_netdev(net, dev) {
3118                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3119                 if (in_dev && (dev->flags & IFF_UP)) {
3120                         struct in_ifaddr *ifa;
3121                         int flag = scope;
3122
3123                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3124                                 addr.s6_addr32[3] = ifa->ifa_local;
3125
3126                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3127                                         continue;
3128                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3129                                         if (idev->dev->flags&IFF_POINTOPOINT)
3130                                                 continue;
3131                                         flag |= IFA_HOST;
3132                                 }
3133
3134                                 add_addr(idev, &addr, plen, flag);
3135                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3136                                                       0, pflags, GFP_KERNEL);
3137                         }
3138                 }
3139         }
3140 }
3141 #endif
3142
3143 static void init_loopback(struct net_device *dev)
3144 {
3145         struct inet6_dev  *idev;
3146
3147         /* ::1 */
3148
3149         ASSERT_RTNL();
3150
3151         idev = ipv6_find_idev(dev);
3152         if (IS_ERR(idev)) {
3153                 pr_debug("%s: add_dev failed\n", __func__);
3154                 return;
3155         }
3156
3157         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3158 }
3159
3160 void addrconf_add_linklocal(struct inet6_dev *idev,
3161                             const struct in6_addr *addr, u32 flags)
3162 {
3163         struct ifa6_config cfg = {
3164                 .pfx = addr,
3165                 .plen = 64,
3166                 .ifa_flags = flags | IFA_F_PERMANENT,
3167                 .valid_lft = INFINITY_LIFE_TIME,
3168                 .preferred_lft = INFINITY_LIFE_TIME,
3169                 .scope = IFA_LINK
3170         };
3171         struct inet6_ifaddr *ifp;
3172
3173 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3174         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3175              idev->cnf.optimistic_dad) &&
3176             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3177                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3178 #endif
3179
3180         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3181         if (!IS_ERR(ifp)) {
3182                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3183                                       0, 0, GFP_ATOMIC);
3184                 addrconf_dad_start(ifp);
3185                 in6_ifa_put(ifp);
3186         }
3187 }
3188 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3189
3190 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3191 {
3192         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3193                 return true;
3194
3195         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3196             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3197                 return true;
3198
3199         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3200             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3201                 return true;
3202
3203         return false;
3204 }
3205
3206 static int ipv6_generate_stable_address(struct in6_addr *address,
3207                                         u8 dad_count,
3208                                         const struct inet6_dev *idev)
3209 {
3210         static DEFINE_SPINLOCK(lock);
3211         static __u32 digest[SHA1_DIGEST_WORDS];
3212         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3213
3214         static union {
3215                 char __data[SHA1_BLOCK_SIZE];
3216                 struct {
3217                         struct in6_addr secret;
3218                         __be32 prefix[2];
3219                         unsigned char hwaddr[MAX_ADDR_LEN];
3220                         u8 dad_count;
3221                 } __packed;
3222         } data;
3223
3224         struct in6_addr secret;
3225         struct in6_addr temp;
3226         struct net *net = dev_net(idev->dev);
3227
3228         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3229
3230         if (idev->cnf.stable_secret.initialized)
3231                 secret = idev->cnf.stable_secret.secret;
3232         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3233                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3234         else
3235                 return -1;
3236
3237 retry:
3238         spin_lock_bh(&lock);
3239
3240         sha1_init(digest);
3241         memset(&data, 0, sizeof(data));
3242         memset(workspace, 0, sizeof(workspace));
3243         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3244         data.prefix[0] = address->s6_addr32[0];
3245         data.prefix[1] = address->s6_addr32[1];
3246         data.secret = secret;
3247         data.dad_count = dad_count;
3248
3249         sha1_transform(digest, data.__data, workspace);
3250
3251         temp = *address;
3252         temp.s6_addr32[2] = (__force __be32)digest[0];
3253         temp.s6_addr32[3] = (__force __be32)digest[1];
3254
3255         spin_unlock_bh(&lock);
3256
3257         if (ipv6_reserved_interfaceid(temp)) {
3258                 dad_count++;
3259                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3260                         return -1;
3261                 goto retry;
3262         }
3263
3264         *address = temp;
3265         return 0;
3266 }
3267
3268 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3269 {
3270         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3271
3272         if (s->initialized)
3273                 return;
3274         s = &idev->cnf.stable_secret;
3275         get_random_bytes(&s->secret, sizeof(s->secret));
3276         s->initialized = true;
3277 }
3278
3279 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3280 {
3281         struct in6_addr addr;
3282
3283         /* no link local addresses on L3 master devices */
3284         if (netif_is_l3_master(idev->dev))
3285                 return;
3286
3287         /* no link local addresses on devices flagged as slaves */
3288         if (idev->dev->flags & IFF_SLAVE)
3289                 return;
3290
3291         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3292
3293         switch (idev->cnf.addr_gen_mode) {
3294         case IN6_ADDR_GEN_MODE_RANDOM:
3295                 ipv6_gen_mode_random_init(idev);
3296                 fallthrough;
3297         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3298                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3299                         addrconf_add_linklocal(idev, &addr,
3300                                                IFA_F_STABLE_PRIVACY);
3301                 else if (prefix_route)
3302                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3303                                               0, 0, GFP_KERNEL);
3304                 break;
3305         case IN6_ADDR_GEN_MODE_EUI64:
3306                 /* addrconf_add_linklocal also adds a prefix_route and we
3307                  * only need to care about prefix routes if ipv6_generate_eui64
3308                  * couldn't generate one.
3309                  */
3310                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3311                         addrconf_add_linklocal(idev, &addr, 0);
3312                 else if (prefix_route)
3313                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3314                                               0, 0, GFP_KERNEL);
3315                 break;
3316         case IN6_ADDR_GEN_MODE_NONE:
3317         default:
3318                 /* will not add any link local address */
3319                 break;
3320         }
3321 }
3322
3323 static void addrconf_dev_config(struct net_device *dev)
3324 {
3325         struct inet6_dev *idev;
3326
3327         ASSERT_RTNL();
3328
3329         if ((dev->type != ARPHRD_ETHER) &&
3330             (dev->type != ARPHRD_FDDI) &&
3331             (dev->type != ARPHRD_ARCNET) &&
3332             (dev->type != ARPHRD_INFINIBAND) &&
3333             (dev->type != ARPHRD_IEEE1394) &&
3334             (dev->type != ARPHRD_TUNNEL6) &&
3335             (dev->type != ARPHRD_6LOWPAN) &&
3336             (dev->type != ARPHRD_IP6GRE) &&
3337             (dev->type != ARPHRD_IPGRE) &&
3338             (dev->type != ARPHRD_TUNNEL) &&
3339             (dev->type != ARPHRD_NONE) &&
3340             (dev->type != ARPHRD_RAWIP)) {
3341                 /* Alas, we support only Ethernet autoconfiguration. */
3342                 idev = __in6_dev_get(dev);
3343                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3344                     dev->flags & IFF_MULTICAST)
3345                         ipv6_mc_up(idev);
3346                 return;
3347         }
3348
3349         idev = addrconf_add_dev(dev);
3350         if (IS_ERR(idev))
3351                 return;
3352
3353         /* this device type has no EUI support */
3354         if (dev->type == ARPHRD_NONE &&
3355             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3356                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3357
3358         addrconf_addr_gen(idev, false);
3359 }
3360
3361 #if IS_ENABLED(CONFIG_IPV6_SIT)
3362 static void addrconf_sit_config(struct net_device *dev)
3363 {
3364         struct inet6_dev *idev;
3365
3366         ASSERT_RTNL();
3367
3368         /*
3369          * Configure the tunnel with one of our IPv4
3370          * addresses... we should configure all of
3371          * our v4 addrs in the tunnel
3372          */
3373
3374         idev = ipv6_find_idev(dev);
3375         if (IS_ERR(idev)) {
3376                 pr_debug("%s: add_dev failed\n", __func__);
3377                 return;
3378         }
3379
3380         if (dev->priv_flags & IFF_ISATAP) {
3381                 addrconf_addr_gen(idev, false);
3382                 return;
3383         }
3384
3385         sit_add_v4_addrs(idev);
3386
3387         if (dev->flags&IFF_POINTOPOINT)
3388                 addrconf_add_mroute(dev);
3389 }
3390 #endif
3391
3392 #if IS_ENABLED(CONFIG_NET_IPGRE)
3393 static void addrconf_gre_config(struct net_device *dev)
3394 {
3395         struct inet6_dev *idev;
3396
3397         ASSERT_RTNL();
3398
3399         idev = ipv6_find_idev(dev);
3400         if (IS_ERR(idev)) {
3401                 pr_debug("%s: add_dev failed\n", __func__);
3402                 return;
3403         }
3404
3405         addrconf_addr_gen(idev, true);
3406         if (dev->flags & IFF_POINTOPOINT)
3407                 addrconf_add_mroute(dev);
3408 }
3409 #endif
3410
3411 static int fixup_permanent_addr(struct net *net,
3412                                 struct inet6_dev *idev,
3413                                 struct inet6_ifaddr *ifp)
3414 {
3415         /* !fib6_node means the host route was removed from the
3416          * FIB, for example, if 'lo' device is taken down. In that
3417          * case regenerate the host route.
3418          */
3419         if (!ifp->rt || !ifp->rt->fib6_node) {
3420                 struct fib6_info *f6i, *prev;
3421
3422                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3423                                          GFP_ATOMIC);
3424                 if (IS_ERR(f6i))
3425                         return PTR_ERR(f6i);
3426
3427                 /* ifp->rt can be accessed outside of rtnl */
3428                 spin_lock(&ifp->lock);
3429                 prev = ifp->rt;
3430                 ifp->rt = f6i;
3431                 spin_unlock(&ifp->lock);
3432
3433                 fib6_info_release(prev);
3434         }
3435
3436         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3437                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3438                                       ifp->rt_priority, idev->dev, 0, 0,
3439                                       GFP_ATOMIC);
3440         }
3441
3442         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3443                 addrconf_dad_start(ifp);
3444
3445         return 0;
3446 }
3447
3448 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3449 {
3450         struct inet6_ifaddr *ifp, *tmp;
3451         struct inet6_dev *idev;
3452
3453         idev = __in6_dev_get(dev);
3454         if (!idev)
3455                 return;
3456
3457         write_lock_bh(&idev->lock);
3458
3459         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3460                 if ((ifp->flags & IFA_F_PERMANENT) &&
3461                     fixup_permanent_addr(net, idev, ifp) < 0) {
3462                         write_unlock_bh(&idev->lock);
3463                         in6_ifa_hold(ifp);
3464                         ipv6_del_addr(ifp);
3465                         write_lock_bh(&idev->lock);
3466
3467                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3468                                              idev->dev->name, &ifp->addr);
3469                 }
3470         }
3471
3472         write_unlock_bh(&idev->lock);
3473 }
3474
3475 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3476                            void *ptr)
3477 {
3478         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3479         struct netdev_notifier_change_info *change_info;
3480         struct netdev_notifier_changeupper_info *info;
3481         struct inet6_dev *idev = __in6_dev_get(dev);
3482         struct net *net = dev_net(dev);
3483         int run_pending = 0;
3484         int err;
3485
3486         switch (event) {
3487         case NETDEV_REGISTER:
3488                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3489                         idev = ipv6_add_dev(dev);
3490                         if (IS_ERR(idev))
3491                                 return notifier_from_errno(PTR_ERR(idev));
3492                 }
3493                 break;
3494
3495         case NETDEV_CHANGEMTU:
3496                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3497                 if (dev->mtu < IPV6_MIN_MTU) {
3498                         addrconf_ifdown(dev, dev != net->loopback_dev);
3499                         break;
3500                 }
3501
3502                 if (idev) {
3503                         rt6_mtu_change(dev, dev->mtu);
3504                         idev->cnf.mtu6 = dev->mtu;
3505                         break;
3506                 }
3507
3508                 /* allocate new idev */
3509                 idev = ipv6_add_dev(dev);
3510                 if (IS_ERR(idev))
3511                         break;
3512
3513                 /* device is still not ready */
3514                 if (!(idev->if_flags & IF_READY))
3515                         break;
3516
3517                 run_pending = 1;
3518                 fallthrough;
3519         case NETDEV_UP:
3520         case NETDEV_CHANGE:
3521                 if (dev->flags & IFF_SLAVE)
3522                         break;
3523
3524                 if (idev && idev->cnf.disable_ipv6)
3525                         break;
3526
3527                 if (event == NETDEV_UP) {
3528                         /* restore routes for permanent addresses */
3529                         addrconf_permanent_addr(net, dev);
3530
3531                         if (!addrconf_link_ready(dev)) {
3532                                 /* device is not ready yet. */
3533                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3534                                          dev->name);
3535                                 break;
3536                         }
3537
3538                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3539                                 idev = ipv6_add_dev(dev);
3540
3541                         if (!IS_ERR_OR_NULL(idev)) {
3542                                 idev->if_flags |= IF_READY;
3543                                 run_pending = 1;
3544                         }
3545                 } else if (event == NETDEV_CHANGE) {
3546                         if (!addrconf_link_ready(dev)) {
3547                                 /* device is still not ready. */
3548                                 rt6_sync_down_dev(dev, event);
3549                                 break;
3550                         }
3551
3552                         if (!IS_ERR_OR_NULL(idev)) {
3553                                 if (idev->if_flags & IF_READY) {
3554                                         /* device is already configured -
3555                                          * but resend MLD reports, we might
3556                                          * have roamed and need to update
3557                                          * multicast snooping switches
3558                                          */
3559                                         ipv6_mc_up(idev);
3560                                         change_info = ptr;
3561                                         if (change_info->flags_changed & IFF_NOARP)
3562                                                 addrconf_dad_run(idev, true);
3563                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3564                                         break;
3565                                 }
3566                                 idev->if_flags |= IF_READY;
3567                         }
3568
3569                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3570                                 dev->name);
3571
3572                         run_pending = 1;
3573                 }
3574
3575                 switch (dev->type) {
3576 #if IS_ENABLED(CONFIG_IPV6_SIT)
3577                 case ARPHRD_SIT:
3578                         addrconf_sit_config(dev);
3579                         break;
3580 #endif
3581 #if IS_ENABLED(CONFIG_NET_IPGRE)
3582                 case ARPHRD_IPGRE:
3583                         addrconf_gre_config(dev);
3584                         break;
3585 #endif
3586                 case ARPHRD_LOOPBACK:
3587                         init_loopback(dev);
3588                         break;
3589
3590                 default:
3591                         addrconf_dev_config(dev);
3592                         break;
3593                 }
3594
3595                 if (!IS_ERR_OR_NULL(idev)) {
3596                         if (run_pending)
3597                                 addrconf_dad_run(idev, false);
3598
3599                         /* Device has an address by now */
3600                         rt6_sync_up(dev, RTNH_F_DEAD);
3601
3602                         /*
3603                          * If the MTU changed during the interface down,
3604                          * when the interface up, the changed MTU must be
3605                          * reflected in the idev as well as routers.
3606                          */
3607                         if (idev->cnf.mtu6 != dev->mtu &&
3608                             dev->mtu >= IPV6_MIN_MTU) {
3609                                 rt6_mtu_change(dev, dev->mtu);
3610                                 idev->cnf.mtu6 = dev->mtu;
3611                         }
3612                         idev->tstamp = jiffies;
3613                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3614
3615                         /*
3616                          * If the changed mtu during down is lower than
3617                          * IPV6_MIN_MTU stop IPv6 on this interface.
3618                          */
3619                         if (dev->mtu < IPV6_MIN_MTU)
3620                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3621                 }
3622                 break;
3623
3624         case NETDEV_DOWN:
3625         case NETDEV_UNREGISTER:
3626                 /*
3627                  *      Remove all addresses from this interface.
3628                  */
3629                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3630                 break;
3631
3632         case NETDEV_CHANGENAME:
3633                 if (idev) {
3634                         snmp6_unregister_dev(idev);
3635                         addrconf_sysctl_unregister(idev);
3636                         err = addrconf_sysctl_register(idev);
3637                         if (err)
3638                                 return notifier_from_errno(err);
3639                         err = snmp6_register_dev(idev);
3640                         if (err) {
3641                                 addrconf_sysctl_unregister(idev);
3642                                 return notifier_from_errno(err);
3643                         }
3644                 }
3645                 break;
3646
3647         case NETDEV_PRE_TYPE_CHANGE:
3648         case NETDEV_POST_TYPE_CHANGE:
3649                 if (idev)
3650                         addrconf_type_change(dev, event);
3651                 break;
3652
3653         case NETDEV_CHANGEUPPER:
3654                 info = ptr;
3655
3656                 /* flush all routes if dev is linked to or unlinked from
3657                  * an L3 master device (e.g., VRF)
3658                  */
3659                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3660                         addrconf_ifdown(dev, false);
3661         }
3662
3663         return NOTIFY_OK;
3664 }
3665
3666 /*
3667  *      addrconf module should be notified of a device going up
3668  */
3669 static struct notifier_block ipv6_dev_notf = {
3670         .notifier_call = addrconf_notify,
3671         .priority = ADDRCONF_NOTIFY_PRIORITY,
3672 };
3673
3674 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3675 {
3676         struct inet6_dev *idev;
3677         ASSERT_RTNL();
3678
3679         idev = __in6_dev_get(dev);
3680
3681         if (event == NETDEV_POST_TYPE_CHANGE)
3682                 ipv6_mc_remap(idev);
3683         else if (event == NETDEV_PRE_TYPE_CHANGE)
3684                 ipv6_mc_unmap(idev);
3685 }
3686
3687 static bool addr_is_local(const struct in6_addr *addr)
3688 {
3689         return ipv6_addr_type(addr) &
3690                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3691 }
3692
3693 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3694 {
3695         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3696         struct net *net = dev_net(dev);
3697         struct inet6_dev *idev;
3698         struct inet6_ifaddr *ifa, *tmp;
3699         bool keep_addr = false;
3700         int state, i;
3701
3702         ASSERT_RTNL();
3703
3704         rt6_disable_ip(dev, event);
3705
3706         idev = __in6_dev_get(dev);
3707         if (!idev)
3708                 return -ENODEV;
3709
3710         /*
3711          * Step 1: remove reference to ipv6 device from parent device.
3712          *         Do not dev_put!
3713          */
3714         if (unregister) {
3715                 idev->dead = 1;
3716
3717                 /* protected by rtnl_lock */
3718                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3719
3720                 /* Step 1.5: remove snmp6 entry */
3721                 snmp6_unregister_dev(idev);
3722
3723         }
3724
3725         /* combine the user config with event to determine if permanent
3726          * addresses are to be removed from address hash table
3727          */
3728         if (!unregister && !idev->cnf.disable_ipv6) {
3729                 /* aggregate the system setting and interface setting */
3730                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3731
3732                 if (!_keep_addr)
3733                         _keep_addr = idev->cnf.keep_addr_on_down;
3734
3735                 keep_addr = (_keep_addr > 0);
3736         }
3737
3738         /* Step 2: clear hash table */
3739         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3740                 struct hlist_head *h = &inet6_addr_lst[i];
3741
3742                 spin_lock_bh(&addrconf_hash_lock);
3743 restart:
3744                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3745                         if (ifa->idev == idev) {
3746                                 addrconf_del_dad_work(ifa);
3747                                 /* combined flag + permanent flag decide if
3748                                  * address is retained on a down event
3749                                  */
3750                                 if (!keep_addr ||
3751                                     !(ifa->flags & IFA_F_PERMANENT) ||
3752                                     addr_is_local(&ifa->addr)) {
3753                                         hlist_del_init_rcu(&ifa->addr_lst);
3754                                         goto restart;
3755                                 }
3756                         }
3757                 }
3758                 spin_unlock_bh(&addrconf_hash_lock);
3759         }
3760
3761         write_lock_bh(&idev->lock);
3762
3763         addrconf_del_rs_timer(idev);
3764
3765         /* Step 2: clear flags for stateless addrconf */
3766         if (!unregister)
3767                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3768
3769         /* Step 3: clear tempaddr list */
3770         while (!list_empty(&idev->tempaddr_list)) {
3771                 ifa = list_first_entry(&idev->tempaddr_list,
3772                                        struct inet6_ifaddr, tmp_list);
3773                 list_del(&ifa->tmp_list);
3774                 write_unlock_bh(&idev->lock);
3775                 spin_lock_bh(&ifa->lock);
3776
3777                 if (ifa->ifpub) {
3778                         in6_ifa_put(ifa->ifpub);
3779                         ifa->ifpub = NULL;
3780                 }
3781                 spin_unlock_bh(&ifa->lock);
3782                 in6_ifa_put(ifa);
3783                 write_lock_bh(&idev->lock);
3784         }
3785
3786         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3787                 struct fib6_info *rt = NULL;
3788                 bool keep;
3789
3790                 addrconf_del_dad_work(ifa);
3791
3792                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3793                         !addr_is_local(&ifa->addr);
3794
3795                 write_unlock_bh(&idev->lock);
3796                 spin_lock_bh(&ifa->lock);
3797
3798                 if (keep) {
3799                         /* set state to skip the notifier below */
3800                         state = INET6_IFADDR_STATE_DEAD;
3801                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3802                         if (!(ifa->flags & IFA_F_NODAD))
3803                                 ifa->flags |= IFA_F_TENTATIVE;
3804
3805                         rt = ifa->rt;
3806                         ifa->rt = NULL;
3807                 } else {
3808                         state = ifa->state;
3809                         ifa->state = INET6_IFADDR_STATE_DEAD;
3810                 }
3811
3812                 spin_unlock_bh(&ifa->lock);
3813
3814                 if (rt)
3815                         ip6_del_rt(net, rt, false);
3816
3817                 if (state != INET6_IFADDR_STATE_DEAD) {
3818                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3819                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3820                 } else {
3821                         if (idev->cnf.forwarding)
3822                                 addrconf_leave_anycast(ifa);
3823                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3824                 }
3825
3826                 write_lock_bh(&idev->lock);
3827                 if (!keep) {
3828                         list_del_rcu(&ifa->if_list);
3829                         in6_ifa_put(ifa);
3830                 }
3831         }
3832
3833         write_unlock_bh(&idev->lock);
3834
3835         /* Step 5: Discard anycast and multicast list */
3836         if (unregister) {
3837                 ipv6_ac_destroy_dev(idev);
3838                 ipv6_mc_destroy_dev(idev);
3839         } else {
3840                 ipv6_mc_down(idev);
3841         }
3842
3843         idev->tstamp = jiffies;
3844
3845         /* Last: Shot the device (if unregistered) */
3846         if (unregister) {
3847                 addrconf_sysctl_unregister(idev);
3848                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3849                 neigh_ifdown(&nd_tbl, dev);
3850                 in6_dev_put(idev);
3851         }
3852         return 0;
3853 }
3854
3855 static void addrconf_rs_timer(struct timer_list *t)
3856 {
3857         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3858         struct net_device *dev = idev->dev;
3859         struct in6_addr lladdr;
3860
3861         write_lock(&idev->lock);
3862         if (idev->dead || !(idev->if_flags & IF_READY))
3863                 goto out;
3864
3865         if (!ipv6_accept_ra(idev))
3866                 goto out;
3867
3868         /* Announcement received after solicitation was sent */
3869         if (idev->if_flags & IF_RA_RCVD)
3870                 goto out;
3871
3872         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3873                 write_unlock(&idev->lock);
3874                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3875                         ndisc_send_rs(dev, &lladdr,
3876                                       &in6addr_linklocal_allrouters);
3877                 else
3878                         goto put;
3879
3880                 write_lock(&idev->lock);
3881                 idev->rs_interval = rfc3315_s14_backoff_update(
3882                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3883                 /* The wait after the last probe can be shorter */
3884                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3885                                              idev->cnf.rtr_solicits) ?
3886                                       idev->cnf.rtr_solicit_delay :
3887                                       idev->rs_interval);
3888         } else {
3889                 /*
3890                  * Note: we do not support deprecated "all on-link"
3891                  * assumption any longer.
3892                  */
3893                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3894         }
3895
3896 out:
3897         write_unlock(&idev->lock);
3898 put:
3899         in6_dev_put(idev);
3900 }
3901
3902 /*
3903  *      Duplicate Address Detection
3904  */
3905 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3906 {
3907         unsigned long rand_num;
3908         struct inet6_dev *idev = ifp->idev;
3909         u64 nonce;
3910
3911         if (ifp->flags & IFA_F_OPTIMISTIC)
3912                 rand_num = 0;
3913         else
3914                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3915
3916         nonce = 0;
3917         if (idev->cnf.enhanced_dad ||
3918             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3919                 do
3920                         get_random_bytes(&nonce, 6);
3921                 while (nonce == 0);
3922         }
3923         ifp->dad_nonce = nonce;
3924         ifp->dad_probes = idev->cnf.dad_transmits;
3925         addrconf_mod_dad_work(ifp, rand_num);
3926 }
3927
3928 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3929 {
3930         struct inet6_dev *idev = ifp->idev;
3931         struct net_device *dev = idev->dev;
3932         bool bump_id, notify = false;
3933         struct net *net;
3934
3935         addrconf_join_solict(dev, &ifp->addr);
3936
3937         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3938
3939         read_lock_bh(&idev->lock);
3940         spin_lock(&ifp->lock);
3941         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3942                 goto out;
3943
3944         net = dev_net(dev);
3945         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3946             (net->ipv6.devconf_all->accept_dad < 1 &&
3947              idev->cnf.accept_dad < 1) ||
3948             !(ifp->flags&IFA_F_TENTATIVE) ||
3949             ifp->flags & IFA_F_NODAD) {
3950                 bool send_na = false;
3951
3952                 if (ifp->flags & IFA_F_TENTATIVE &&
3953                     !(ifp->flags & IFA_F_OPTIMISTIC))
3954                         send_na = true;
3955                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3956                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3957                 spin_unlock(&ifp->lock);
3958                 read_unlock_bh(&idev->lock);
3959
3960                 addrconf_dad_completed(ifp, bump_id, send_na);
3961                 return;
3962         }
3963
3964         if (!(idev->if_flags & IF_READY)) {
3965                 spin_unlock(&ifp->lock);
3966                 read_unlock_bh(&idev->lock);
3967                 /*
3968                  * If the device is not ready:
3969                  * - keep it tentative if it is a permanent address.
3970                  * - otherwise, kill it.
3971                  */
3972                 in6_ifa_hold(ifp);
3973                 addrconf_dad_stop(ifp, 0);
3974                 return;
3975         }
3976
3977         /*
3978          * Optimistic nodes can start receiving
3979          * Frames right away
3980          */
3981         if (ifp->flags & IFA_F_OPTIMISTIC) {
3982                 ip6_ins_rt(net, ifp->rt);
3983                 if (ipv6_use_optimistic_addr(net, idev)) {
3984                         /* Because optimistic nodes can use this address,
3985                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3986                          */
3987                         notify = true;
3988                 }
3989         }
3990
3991         addrconf_dad_kick(ifp);
3992 out:
3993         spin_unlock(&ifp->lock);
3994         read_unlock_bh(&idev->lock);
3995         if (notify)
3996                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3997 }
3998
3999 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4000 {
4001         bool begin_dad = false;
4002
4003         spin_lock_bh(&ifp->lock);
4004         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4005                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4006                 begin_dad = true;
4007         }
4008         spin_unlock_bh(&ifp->lock);
4009
4010         if (begin_dad)
4011                 addrconf_mod_dad_work(ifp, 0);
4012 }
4013
4014 static void addrconf_dad_work(struct work_struct *w)
4015 {
4016         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4017                                                 struct inet6_ifaddr,
4018                                                 dad_work);
4019         struct inet6_dev *idev = ifp->idev;
4020         bool bump_id, disable_ipv6 = false;
4021         struct in6_addr mcaddr;
4022
4023         enum {
4024                 DAD_PROCESS,
4025                 DAD_BEGIN,
4026                 DAD_ABORT,
4027         } action = DAD_PROCESS;
4028
4029         rtnl_lock();
4030
4031         spin_lock_bh(&ifp->lock);
4032         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4033                 action = DAD_BEGIN;
4034                 ifp->state = INET6_IFADDR_STATE_DAD;
4035         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4036                 action = DAD_ABORT;
4037                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4038
4039                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4040                      idev->cnf.accept_dad > 1) &&
4041                     !idev->cnf.disable_ipv6 &&
4042                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4043                         struct in6_addr addr;
4044
4045                         addr.s6_addr32[0] = htonl(0xfe800000);
4046                         addr.s6_addr32[1] = 0;
4047
4048                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4049                             ipv6_addr_equal(&ifp->addr, &addr)) {
4050                                 /* DAD failed for link-local based on MAC */
4051                                 idev->cnf.disable_ipv6 = 1;
4052
4053                                 pr_info("%s: IPv6 being disabled!\n",
4054                                         ifp->idev->dev->name);
4055                                 disable_ipv6 = true;
4056                         }
4057                 }
4058         }
4059         spin_unlock_bh(&ifp->lock);
4060
4061         if (action == DAD_BEGIN) {
4062                 addrconf_dad_begin(ifp);
4063                 goto out;
4064         } else if (action == DAD_ABORT) {
4065                 in6_ifa_hold(ifp);
4066                 addrconf_dad_stop(ifp, 1);
4067                 if (disable_ipv6)
4068                         addrconf_ifdown(idev->dev, false);
4069                 goto out;
4070         }
4071
4072         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4073                 goto out;
4074
4075         write_lock_bh(&idev->lock);
4076         if (idev->dead || !(idev->if_flags & IF_READY)) {
4077                 write_unlock_bh(&idev->lock);
4078                 goto out;
4079         }
4080
4081         spin_lock(&ifp->lock);
4082         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4083                 spin_unlock(&ifp->lock);
4084                 write_unlock_bh(&idev->lock);
4085                 goto out;
4086         }
4087
4088         if (ifp->dad_probes == 0) {
4089                 bool send_na = false;
4090
4091                 /*
4092                  * DAD was successful
4093                  */
4094
4095                 if (ifp->flags & IFA_F_TENTATIVE &&
4096                     !(ifp->flags & IFA_F_OPTIMISTIC))
4097                         send_na = true;
4098                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4099                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4100                 spin_unlock(&ifp->lock);
4101                 write_unlock_bh(&idev->lock);
4102
4103                 addrconf_dad_completed(ifp, bump_id, send_na);
4104
4105                 goto out;
4106         }
4107
4108         ifp->dad_probes--;
4109         addrconf_mod_dad_work(ifp,
4110                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4111                                   HZ/100));
4112         spin_unlock(&ifp->lock);
4113         write_unlock_bh(&idev->lock);
4114
4115         /* send a neighbour solicitation for our addr */
4116         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4117         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4118                       ifp->dad_nonce);
4119 out:
4120         in6_ifa_put(ifp);
4121         rtnl_unlock();
4122 }
4123
4124 /* ifp->idev must be at least read locked */
4125 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4126 {
4127         struct inet6_ifaddr *ifpiter;
4128         struct inet6_dev *idev = ifp->idev;
4129
4130         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4131                 if (ifpiter->scope > IFA_LINK)
4132                         break;
4133                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4134                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4135                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4136                     IFA_F_PERMANENT)
4137                         return false;
4138         }
4139         return true;
4140 }
4141
4142 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4143                                    bool send_na)
4144 {
4145         struct net_device *dev = ifp->idev->dev;
4146         struct in6_addr lladdr;
4147         bool send_rs, send_mld;
4148
4149         addrconf_del_dad_work(ifp);
4150
4151         /*
4152          *      Configure the address for reception. Now it is valid.
4153          */
4154
4155         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4156
4157         /* If added prefix is link local and we are prepared to process
4158            router advertisements, start sending router solicitations.
4159          */
4160
4161         read_lock_bh(&ifp->idev->lock);
4162         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4163         send_rs = send_mld &&
4164                   ipv6_accept_ra(ifp->idev) &&
4165                   ifp->idev->cnf.rtr_solicits != 0 &&
4166                   (dev->flags&IFF_LOOPBACK) == 0;
4167         read_unlock_bh(&ifp->idev->lock);
4168
4169         /* While dad is in progress mld report's source address is in6_addrany.
4170          * Resend with proper ll now.
4171          */
4172         if (send_mld)
4173                 ipv6_mc_dad_complete(ifp->idev);
4174
4175         /* send unsolicited NA if enabled */
4176         if (send_na &&
4177             (ifp->idev->cnf.ndisc_notify ||
4178              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4179                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4180                               /*router=*/ !!ifp->idev->cnf.forwarding,
4181                               /*solicited=*/ false, /*override=*/ true,
4182                               /*inc_opt=*/ true);
4183         }
4184
4185         if (send_rs) {
4186                 /*
4187                  *      If a host as already performed a random delay
4188                  *      [...] as part of DAD [...] there is no need
4189                  *      to delay again before sending the first RS
4190                  */
4191                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4192                         return;
4193                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4194
4195                 write_lock_bh(&ifp->idev->lock);
4196                 spin_lock(&ifp->lock);
4197                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4198                         ifp->idev->cnf.rtr_solicit_interval);
4199                 ifp->idev->rs_probes = 1;
4200                 ifp->idev->if_flags |= IF_RS_SENT;
4201                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4202                 spin_unlock(&ifp->lock);
4203                 write_unlock_bh(&ifp->idev->lock);
4204         }
4205
4206         if (bump_id)
4207                 rt_genid_bump_ipv6(dev_net(dev));
4208
4209         /* Make sure that a new temporary address will be created
4210          * before this temporary address becomes deprecated.
4211          */
4212         if (ifp->flags & IFA_F_TEMPORARY)
4213                 addrconf_verify_rtnl();
4214 }
4215
4216 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4217 {
4218         struct inet6_ifaddr *ifp;
4219
4220         read_lock_bh(&idev->lock);
4221         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4222                 spin_lock(&ifp->lock);
4223                 if ((ifp->flags & IFA_F_TENTATIVE &&
4224                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4225                         if (restart)
4226                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4227                         addrconf_dad_kick(ifp);
4228                 }
4229                 spin_unlock(&ifp->lock);
4230         }
4231         read_unlock_bh(&idev->lock);
4232 }
4233
4234 #ifdef CONFIG_PROC_FS
4235 struct if6_iter_state {
4236         struct seq_net_private p;
4237         int bucket;
4238         int offset;
4239 };
4240
4241 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4242 {
4243         struct if6_iter_state *state = seq->private;
4244         struct net *net = seq_file_net(seq);
4245         struct inet6_ifaddr *ifa = NULL;
4246         int p = 0;
4247
4248         /* initial bucket if pos is 0 */
4249         if (pos == 0) {
4250                 state->bucket = 0;
4251                 state->offset = 0;
4252         }
4253
4254         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4255                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4256                                          addr_lst) {
4257                         if (!net_eq(dev_net(ifa->idev->dev), net))
4258                                 continue;
4259                         /* sync with offset */
4260                         if (p < state->offset) {
4261                                 p++;
4262                                 continue;
4263                         }
4264                         return ifa;
4265                 }
4266
4267                 /* prepare for next bucket */
4268                 state->offset = 0;
4269                 p = 0;
4270         }
4271         return NULL;
4272 }
4273
4274 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4275                                          struct inet6_ifaddr *ifa)
4276 {
4277         struct if6_iter_state *state = seq->private;
4278         struct net *net = seq_file_net(seq);
4279
4280         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4281                 if (!net_eq(dev_net(ifa->idev->dev), net))
4282                         continue;
4283                 state->offset++;
4284                 return ifa;
4285         }
4286
4287         state->offset = 0;
4288         while (++state->bucket < IN6_ADDR_HSIZE) {
4289                 hlist_for_each_entry_rcu(ifa,
4290                                      &inet6_addr_lst[state->bucket], addr_lst) {
4291                         if (!net_eq(dev_net(ifa->idev->dev), net))
4292                                 continue;
4293                         return ifa;
4294                 }
4295         }
4296
4297         return NULL;
4298 }
4299
4300 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4301         __acquires(rcu)
4302 {
4303         rcu_read_lock();
4304         return if6_get_first(seq, *pos);
4305 }
4306
4307 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4308 {
4309         struct inet6_ifaddr *ifa;
4310
4311         ifa = if6_get_next(seq, v);
4312         ++*pos;
4313         return ifa;
4314 }
4315
4316 static void if6_seq_stop(struct seq_file *seq, void *v)
4317         __releases(rcu)
4318 {
4319         rcu_read_unlock();
4320 }
4321
4322 static int if6_seq_show(struct seq_file *seq, void *v)
4323 {
4324         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4325         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4326                    &ifp->addr,
4327                    ifp->idev->dev->ifindex,
4328                    ifp->prefix_len,
4329                    ifp->scope,
4330                    (u8) ifp->flags,
4331                    ifp->idev->dev->name);
4332         return 0;
4333 }
4334
4335 static const struct seq_operations if6_seq_ops = {
4336         .start  = if6_seq_start,
4337         .next   = if6_seq_next,
4338         .show   = if6_seq_show,
4339         .stop   = if6_seq_stop,
4340 };
4341
4342 static int __net_init if6_proc_net_init(struct net *net)
4343 {
4344         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4345                         sizeof(struct if6_iter_state)))
4346                 return -ENOMEM;
4347         return 0;
4348 }
4349
4350 static void __net_exit if6_proc_net_exit(struct net *net)
4351 {
4352         remove_proc_entry("if_inet6", net->proc_net);
4353 }
4354
4355 static struct pernet_operations if6_proc_net_ops = {
4356         .init = if6_proc_net_init,
4357         .exit = if6_proc_net_exit,
4358 };
4359
4360 int __init if6_proc_init(void)
4361 {
4362         return register_pernet_subsys(&if6_proc_net_ops);
4363 }
4364
4365 void if6_proc_exit(void)
4366 {
4367         unregister_pernet_subsys(&if6_proc_net_ops);
4368 }
4369 #endif  /* CONFIG_PROC_FS */
4370
4371 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4372 /* Check if address is a home address configured on any interface. */
4373 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4374 {
4375         unsigned int hash = inet6_addr_hash(net, addr);
4376         struct inet6_ifaddr *ifp = NULL;
4377         int ret = 0;
4378
4379         rcu_read_lock();
4380         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4381                 if (!net_eq(dev_net(ifp->idev->dev), net))
4382                         continue;
4383                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4384                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4385                         ret = 1;
4386                         break;
4387                 }
4388         }
4389         rcu_read_unlock();
4390         return ret;
4391 }
4392 #endif
4393
4394 /* RFC6554 has some algorithm to avoid loops in segment routing by
4395  * checking if the segments contains any of a local interface address.
4396  *
4397  * Quote:
4398  *
4399  * To detect loops in the SRH, a router MUST determine if the SRH
4400  * includes multiple addresses assigned to any interface on that router.
4401  * If such addresses appear more than once and are separated by at least
4402  * one address not assigned to that router.
4403  */
4404 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4405                           unsigned char nsegs)
4406 {
4407         const struct in6_addr *addr;
4408         int i, ret = 0, found = 0;
4409         struct inet6_ifaddr *ifp;
4410         bool separated = false;
4411         unsigned int hash;
4412         bool hash_found;
4413
4414         rcu_read_lock();
4415         for (i = 0; i < nsegs; i++) {
4416                 addr = &segs[i];
4417                 hash = inet6_addr_hash(net, addr);
4418
4419                 hash_found = false;
4420                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4421                         if (!net_eq(dev_net(ifp->idev->dev), net))
4422                                 continue;
4423
4424                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4425                                 hash_found = true;
4426                                 break;
4427                         }
4428                 }
4429
4430                 if (hash_found) {
4431                         if (found > 1 && separated) {
4432                                 ret = 1;
4433                                 break;
4434                         }
4435
4436                         separated = false;
4437                         found++;
4438                 } else {
4439                         separated = true;
4440                 }
4441         }
4442         rcu_read_unlock();
4443
4444         return ret;
4445 }
4446
4447 /*
4448  *      Periodic address status verification
4449  */
4450
4451 static void addrconf_verify_rtnl(void)
4452 {
4453         unsigned long now, next, next_sec, next_sched;
4454         struct inet6_ifaddr *ifp;
4455         int i;
4456
4457         ASSERT_RTNL();
4458
4459         rcu_read_lock_bh();
4460         now = jiffies;
4461         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4462
4463         cancel_delayed_work(&addr_chk_work);
4464
4465         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4466 restart:
4467                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4468                         unsigned long age;
4469
4470                         /* When setting preferred_lft to a value not zero or
4471                          * infinity, while valid_lft is infinity
4472                          * IFA_F_PERMANENT has a non-infinity life time.
4473                          */
4474                         if ((ifp->flags & IFA_F_PERMANENT) &&
4475                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4476                                 continue;
4477
4478                         spin_lock(&ifp->lock);
4479                         /* We try to batch several events at once. */
4480                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4481
4482                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4483                             age >= ifp->valid_lft) {
4484                                 spin_unlock(&ifp->lock);
4485                                 in6_ifa_hold(ifp);
4486                                 ipv6_del_addr(ifp);
4487                                 goto restart;
4488                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4489                                 spin_unlock(&ifp->lock);
4490                                 continue;
4491                         } else if (age >= ifp->prefered_lft) {
4492                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4493                                 int deprecate = 0;
4494
4495                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4496                                         deprecate = 1;
4497                                         ifp->flags |= IFA_F_DEPRECATED;
4498                                 }
4499
4500                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4501                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4502                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4503
4504                                 spin_unlock(&ifp->lock);
4505
4506                                 if (deprecate) {
4507                                         in6_ifa_hold(ifp);
4508
4509                                         ipv6_ifa_notify(0, ifp);
4510                                         in6_ifa_put(ifp);
4511                                         goto restart;
4512                                 }
4513                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4514                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4515                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4516                                         ifp->idev->cnf.dad_transmits *
4517                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4518
4519                                 if (age >= ifp->prefered_lft - regen_advance) {
4520                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4521                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4522                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4523                                         if (!ifp->regen_count && ifpub) {
4524                                                 ifp->regen_count++;
4525                                                 in6_ifa_hold(ifp);
4526                                                 in6_ifa_hold(ifpub);
4527                                                 spin_unlock(&ifp->lock);
4528
4529                                                 spin_lock(&ifpub->lock);
4530                                                 ifpub->regen_count = 0;
4531                                                 spin_unlock(&ifpub->lock);
4532                                                 rcu_read_unlock_bh();
4533                                                 ipv6_create_tempaddr(ifpub, true);
4534                                                 in6_ifa_put(ifpub);
4535                                                 in6_ifa_put(ifp);
4536                                                 rcu_read_lock_bh();
4537                                                 goto restart;
4538                                         }
4539                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4540                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4541                                 spin_unlock(&ifp->lock);
4542                         } else {
4543                                 /* ifp->prefered_lft <= ifp->valid_lft */
4544                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4545                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4546                                 spin_unlock(&ifp->lock);
4547                         }
4548                 }
4549         }
4550
4551         next_sec = round_jiffies_up(next);
4552         next_sched = next;
4553
4554         /* If rounded timeout is accurate enough, accept it. */
4555         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4556                 next_sched = next_sec;
4557
4558         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4559         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4560                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4561
4562         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4563                  now, next, next_sec, next_sched);
4564         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4565         rcu_read_unlock_bh();
4566 }
4567
4568 static void addrconf_verify_work(struct work_struct *w)
4569 {
4570         rtnl_lock();
4571         addrconf_verify_rtnl();
4572         rtnl_unlock();
4573 }
4574
4575 static void addrconf_verify(void)
4576 {
4577         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4578 }
4579
4580 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4581                                      struct in6_addr **peer_pfx)
4582 {
4583         struct in6_addr *pfx = NULL;
4584
4585         *peer_pfx = NULL;
4586
4587         if (addr)
4588                 pfx = nla_data(addr);
4589
4590         if (local) {
4591                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4592                         *peer_pfx = pfx;
4593                 pfx = nla_data(local);
4594         }
4595
4596         return pfx;
4597 }
4598
4599 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4600         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4601         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4602         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4603         [IFA_FLAGS]             = { .len = sizeof(u32) },
4604         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4605         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4606 };
4607
4608 static int
4609 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4610                   struct netlink_ext_ack *extack)
4611 {
4612         struct net *net = sock_net(skb->sk);
4613         struct ifaddrmsg *ifm;
4614         struct nlattr *tb[IFA_MAX+1];
4615         struct in6_addr *pfx, *peer_pfx;
4616         u32 ifa_flags;
4617         int err;
4618
4619         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4620                                      ifa_ipv6_policy, extack);
4621         if (err < 0)
4622                 return err;
4623
4624         ifm = nlmsg_data(nlh);
4625         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4626         if (!pfx)
4627                 return -EINVAL;
4628
4629         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4630
4631         /* We ignore other flags so far. */
4632         ifa_flags &= IFA_F_MANAGETEMPADDR;
4633
4634         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4635                               ifm->ifa_prefixlen);
4636 }
4637
4638 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4639                                unsigned long expires, u32 flags,
4640                                bool modify_peer)
4641 {
4642         struct fib6_info *f6i;
4643         u32 prio;
4644
4645         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4646                                         ifp->prefix_len,
4647                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4648         if (!f6i)
4649                 return -ENOENT;
4650
4651         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4652         if (f6i->fib6_metric != prio) {
4653                 /* delete old one */
4654                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4655
4656                 /* add new one */
4657                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4658                                       ifp->prefix_len,
4659                                       ifp->rt_priority, ifp->idev->dev,
4660                                       expires, flags, GFP_KERNEL);
4661         } else {
4662                 if (!expires)
4663                         fib6_clean_expires(f6i);
4664                 else
4665                         fib6_set_expires(f6i, expires);
4666
4667                 fib6_info_release(f6i);
4668         }
4669
4670         return 0;
4671 }
4672
4673 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4674 {
4675         u32 flags;
4676         clock_t expires;
4677         unsigned long timeout;
4678         bool was_managetempaddr;
4679         bool had_prefixroute;
4680         bool new_peer = false;
4681
4682         ASSERT_RTNL();
4683
4684         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4685                 return -EINVAL;
4686
4687         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4688             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4689                 return -EINVAL;
4690
4691         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4692                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4693
4694         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4695         if (addrconf_finite_timeout(timeout)) {
4696                 expires = jiffies_to_clock_t(timeout * HZ);
4697                 cfg->valid_lft = timeout;
4698                 flags = RTF_EXPIRES;
4699         } else {
4700                 expires = 0;
4701                 flags = 0;
4702                 cfg->ifa_flags |= IFA_F_PERMANENT;
4703         }
4704
4705         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4706         if (addrconf_finite_timeout(timeout)) {
4707                 if (timeout == 0)
4708                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4709                 cfg->preferred_lft = timeout;
4710         }
4711
4712         if (cfg->peer_pfx &&
4713             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4714                 if (!ipv6_addr_any(&ifp->peer_addr))
4715                         cleanup_prefix_route(ifp, expires, true, true);
4716                 new_peer = true;
4717         }
4718
4719         spin_lock_bh(&ifp->lock);
4720         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4721         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4722                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4723         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4724                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4725                         IFA_F_NOPREFIXROUTE);
4726         ifp->flags |= cfg->ifa_flags;
4727         ifp->tstamp = jiffies;
4728         ifp->valid_lft = cfg->valid_lft;
4729         ifp->prefered_lft = cfg->preferred_lft;
4730
4731         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4732                 ifp->rt_priority = cfg->rt_priority;
4733
4734         if (new_peer)
4735                 ifp->peer_addr = *cfg->peer_pfx;
4736
4737         spin_unlock_bh(&ifp->lock);
4738         if (!(ifp->flags&IFA_F_TENTATIVE))
4739                 ipv6_ifa_notify(0, ifp);
4740
4741         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4742                 int rc = -ENOENT;
4743
4744                 if (had_prefixroute)
4745                         rc = modify_prefix_route(ifp, expires, flags, false);
4746
4747                 /* prefix route could have been deleted; if so restore it */
4748                 if (rc == -ENOENT) {
4749                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4750                                               ifp->rt_priority, ifp->idev->dev,
4751                                               expires, flags, GFP_KERNEL);
4752                 }
4753
4754                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4755                         rc = modify_prefix_route(ifp, expires, flags, true);
4756
4757                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4758                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4759                                               ifp->rt_priority, ifp->idev->dev,
4760                                               expires, flags, GFP_KERNEL);
4761                 }
4762         } else if (had_prefixroute) {
4763                 enum cleanup_prefix_rt_t action;
4764                 unsigned long rt_expires;
4765
4766                 write_lock_bh(&ifp->idev->lock);
4767                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4768                 write_unlock_bh(&ifp->idev->lock);
4769
4770                 if (action != CLEANUP_PREFIX_RT_NOP) {
4771                         cleanup_prefix_route(ifp, rt_expires,
4772                                 action == CLEANUP_PREFIX_RT_DEL, false);
4773                 }
4774         }
4775
4776         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4777                 if (was_managetempaddr &&
4778                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4779                         cfg->valid_lft = 0;
4780                         cfg->preferred_lft = 0;
4781                 }
4782                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4783                                  cfg->preferred_lft, !was_managetempaddr,
4784                                  jiffies);
4785         }
4786
4787         addrconf_verify_rtnl();
4788
4789         return 0;
4790 }
4791
4792 static int
4793 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4794                   struct netlink_ext_ack *extack)
4795 {
4796         struct net *net = sock_net(skb->sk);
4797         struct ifaddrmsg *ifm;
4798         struct nlattr *tb[IFA_MAX+1];
4799         struct in6_addr *peer_pfx;
4800         struct inet6_ifaddr *ifa;
4801         struct net_device *dev;
4802         struct inet6_dev *idev;
4803         struct ifa6_config cfg;
4804         int err;
4805
4806         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4807                                      ifa_ipv6_policy, extack);
4808         if (err < 0)
4809                 return err;
4810
4811         memset(&cfg, 0, sizeof(cfg));
4812
4813         ifm = nlmsg_data(nlh);
4814         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4815         if (!cfg.pfx)
4816                 return -EINVAL;
4817
4818         cfg.peer_pfx = peer_pfx;
4819         cfg.plen = ifm->ifa_prefixlen;
4820         if (tb[IFA_RT_PRIORITY])
4821                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4822
4823         cfg.valid_lft = INFINITY_LIFE_TIME;
4824         cfg.preferred_lft = INFINITY_LIFE_TIME;
4825
4826         if (tb[IFA_CACHEINFO]) {
4827                 struct ifa_cacheinfo *ci;
4828
4829                 ci = nla_data(tb[IFA_CACHEINFO]);
4830                 cfg.valid_lft = ci->ifa_valid;
4831                 cfg.preferred_lft = ci->ifa_prefered;
4832         }
4833
4834         dev =  __dev_get_by_index(net, ifm->ifa_index);
4835         if (!dev)
4836                 return -ENODEV;
4837
4838         if (tb[IFA_FLAGS])
4839                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4840         else
4841                 cfg.ifa_flags = ifm->ifa_flags;
4842
4843         /* We ignore other flags so far. */
4844         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4845                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4846                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4847
4848         idev = ipv6_find_idev(dev);
4849         if (IS_ERR(idev))
4850                 return PTR_ERR(idev);
4851
4852         if (!ipv6_allow_optimistic_dad(net, idev))
4853                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4854
4855         if (cfg.ifa_flags & IFA_F_NODAD &&
4856             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4857                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4858                 return -EINVAL;
4859         }
4860
4861         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4862         if (!ifa) {
4863                 /*
4864                  * It would be best to check for !NLM_F_CREATE here but
4865                  * userspace already relies on not having to provide this.
4866                  */
4867                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4868         }
4869
4870         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4871             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4872                 err = -EEXIST;
4873         else
4874                 err = inet6_addr_modify(ifa, &cfg);
4875
4876         in6_ifa_put(ifa);
4877
4878         return err;
4879 }
4880
4881 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4882                           u8 scope, int ifindex)
4883 {
4884         struct ifaddrmsg *ifm;
4885
4886         ifm = nlmsg_data(nlh);
4887         ifm->ifa_family = AF_INET6;
4888         ifm->ifa_prefixlen = prefixlen;
4889         ifm->ifa_flags = flags;
4890         ifm->ifa_scope = scope;
4891         ifm->ifa_index = ifindex;
4892 }
4893
4894 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4895                          unsigned long tstamp, u32 preferred, u32 valid)
4896 {
4897         struct ifa_cacheinfo ci;
4898
4899         ci.cstamp = cstamp_delta(cstamp);
4900         ci.tstamp = cstamp_delta(tstamp);
4901         ci.ifa_prefered = preferred;
4902         ci.ifa_valid = valid;
4903
4904         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4905 }
4906
4907 static inline int rt_scope(int ifa_scope)
4908 {
4909         if (ifa_scope & IFA_HOST)
4910                 return RT_SCOPE_HOST;
4911         else if (ifa_scope & IFA_LINK)
4912                 return RT_SCOPE_LINK;
4913         else if (ifa_scope & IFA_SITE)
4914                 return RT_SCOPE_SITE;
4915         else
4916                 return RT_SCOPE_UNIVERSE;
4917 }
4918
4919 static inline int inet6_ifaddr_msgsize(void)
4920 {
4921         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4922                + nla_total_size(16) /* IFA_LOCAL */
4923                + nla_total_size(16) /* IFA_ADDRESS */
4924                + nla_total_size(sizeof(struct ifa_cacheinfo))
4925                + nla_total_size(4)  /* IFA_FLAGS */
4926                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4927 }
4928
4929 enum addr_type_t {
4930         UNICAST_ADDR,
4931         MULTICAST_ADDR,
4932         ANYCAST_ADDR,
4933 };
4934
4935 struct inet6_fill_args {
4936         u32 portid;
4937         u32 seq;
4938         int event;
4939         unsigned int flags;
4940         int netnsid;
4941         int ifindex;
4942         enum addr_type_t type;
4943 };
4944
4945 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4946                              struct inet6_fill_args *args)
4947 {
4948         struct nlmsghdr  *nlh;
4949         u32 preferred, valid;
4950
4951         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4952                         sizeof(struct ifaddrmsg), args->flags);
4953         if (!nlh)
4954                 return -EMSGSIZE;
4955
4956         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4957                       ifa->idev->dev->ifindex);
4958
4959         if (args->netnsid >= 0 &&
4960             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4961                 goto error;
4962
4963         if (!((ifa->flags&IFA_F_PERMANENT) &&
4964               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4965                 preferred = ifa->prefered_lft;
4966                 valid = ifa->valid_lft;
4967                 if (preferred != INFINITY_LIFE_TIME) {
4968                         long tval = (jiffies - ifa->tstamp)/HZ;
4969                         if (preferred > tval)
4970                                 preferred -= tval;
4971                         else
4972                                 preferred = 0;
4973                         if (valid != INFINITY_LIFE_TIME) {
4974                                 if (valid > tval)
4975                                         valid -= tval;
4976                                 else
4977                                         valid = 0;
4978                         }
4979                 }
4980         } else {
4981                 preferred = INFINITY_LIFE_TIME;
4982                 valid = INFINITY_LIFE_TIME;
4983         }
4984
4985         if (!ipv6_addr_any(&ifa->peer_addr)) {
4986                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4987                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4988                         goto error;
4989         } else
4990                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4991                         goto error;
4992
4993         if (ifa->rt_priority &&
4994             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4995                 goto error;
4996
4997         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4998                 goto error;
4999
5000         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5001                 goto error;
5002
5003         nlmsg_end(skb, nlh);
5004         return 0;
5005
5006 error:
5007         nlmsg_cancel(skb, nlh);
5008         return -EMSGSIZE;
5009 }
5010
5011 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5012                                struct inet6_fill_args *args)
5013 {
5014         struct nlmsghdr  *nlh;
5015         u8 scope = RT_SCOPE_UNIVERSE;
5016         int ifindex = ifmca->idev->dev->ifindex;
5017
5018         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5019                 scope = RT_SCOPE_SITE;
5020
5021         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5022                         sizeof(struct ifaddrmsg), args->flags);
5023         if (!nlh)
5024                 return -EMSGSIZE;
5025
5026         if (args->netnsid >= 0 &&
5027             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5028                 nlmsg_cancel(skb, nlh);
5029                 return -EMSGSIZE;
5030         }
5031
5032         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5033         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5034             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5035                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5036                 nlmsg_cancel(skb, nlh);
5037                 return -EMSGSIZE;
5038         }
5039
5040         nlmsg_end(skb, nlh);
5041         return 0;
5042 }
5043
5044 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5045                                struct inet6_fill_args *args)
5046 {
5047         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5048         int ifindex = dev ? dev->ifindex : 1;
5049         struct nlmsghdr  *nlh;
5050         u8 scope = RT_SCOPE_UNIVERSE;
5051
5052         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5053                 scope = RT_SCOPE_SITE;
5054
5055         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5056                         sizeof(struct ifaddrmsg), args->flags);
5057         if (!nlh)
5058                 return -EMSGSIZE;
5059
5060         if (args->netnsid >= 0 &&
5061             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5062                 nlmsg_cancel(skb, nlh);
5063                 return -EMSGSIZE;
5064         }
5065
5066         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5067         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5068             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5069                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5070                 nlmsg_cancel(skb, nlh);
5071                 return -EMSGSIZE;
5072         }
5073
5074         nlmsg_end(skb, nlh);
5075         return 0;
5076 }
5077
5078 /* called with rcu_read_lock() */
5079 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5080                           struct netlink_callback *cb, int s_ip_idx,
5081                           struct inet6_fill_args *fillargs)
5082 {
5083         struct ifmcaddr6 *ifmca;
5084         struct ifacaddr6 *ifaca;
5085         int ip_idx = 0;
5086         int err = 1;
5087
5088         read_lock_bh(&idev->lock);
5089         switch (fillargs->type) {
5090         case UNICAST_ADDR: {
5091                 struct inet6_ifaddr *ifa;
5092                 fillargs->event = RTM_NEWADDR;
5093
5094                 /* unicast address incl. temp addr */
5095                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5096                         if (ip_idx < s_ip_idx)
5097                                 goto next;
5098                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5099                         if (err < 0)
5100                                 break;
5101                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5102 next:
5103                         ip_idx++;
5104                 }
5105                 break;
5106         }
5107         case MULTICAST_ADDR:
5108                 fillargs->event = RTM_GETMULTICAST;
5109
5110                 /* multicast address */
5111                 for (ifmca = idev->mc_list; ifmca;
5112                      ifmca = ifmca->next, ip_idx++) {
5113                         if (ip_idx < s_ip_idx)
5114                                 continue;
5115                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5116                         if (err < 0)
5117                                 break;
5118                 }
5119                 break;
5120         case ANYCAST_ADDR:
5121                 fillargs->event = RTM_GETANYCAST;
5122                 /* anycast address */
5123                 for (ifaca = idev->ac_list; ifaca;
5124                      ifaca = ifaca->aca_next, ip_idx++) {
5125                         if (ip_idx < s_ip_idx)
5126                                 continue;
5127                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5128                         if (err < 0)
5129                                 break;
5130                 }
5131                 break;
5132         default:
5133                 break;
5134         }
5135         read_unlock_bh(&idev->lock);
5136         cb->args[2] = ip_idx;
5137         return err;
5138 }
5139
5140 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5141                                        struct inet6_fill_args *fillargs,
5142                                        struct net **tgt_net, struct sock *sk,
5143                                        struct netlink_callback *cb)
5144 {
5145         struct netlink_ext_ack *extack = cb->extack;
5146         struct nlattr *tb[IFA_MAX+1];
5147         struct ifaddrmsg *ifm;
5148         int err, i;
5149
5150         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5151                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5152                 return -EINVAL;
5153         }
5154
5155         ifm = nlmsg_data(nlh);
5156         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5157                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5158                 return -EINVAL;
5159         }
5160
5161         fillargs->ifindex = ifm->ifa_index;
5162         if (fillargs->ifindex) {
5163                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5164                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5165         }
5166
5167         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5168                                             ifa_ipv6_policy, extack);
5169         if (err < 0)
5170                 return err;
5171
5172         for (i = 0; i <= IFA_MAX; ++i) {
5173                 if (!tb[i])
5174                         continue;
5175
5176                 if (i == IFA_TARGET_NETNSID) {
5177                         struct net *net;
5178
5179                         fillargs->netnsid = nla_get_s32(tb[i]);
5180                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5181                         if (IS_ERR(net)) {
5182                                 fillargs->netnsid = -1;
5183                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5184                                 return PTR_ERR(net);
5185                         }
5186                         *tgt_net = net;
5187                 } else {
5188                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5189                         return -EINVAL;
5190                 }
5191         }
5192
5193         return 0;
5194 }
5195
5196 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5197                            enum addr_type_t type)
5198 {
5199         const struct nlmsghdr *nlh = cb->nlh;
5200         struct inet6_fill_args fillargs = {
5201                 .portid = NETLINK_CB(cb->skb).portid,
5202                 .seq = cb->nlh->nlmsg_seq,
5203                 .flags = NLM_F_MULTI,
5204                 .netnsid = -1,
5205                 .type = type,
5206         };
5207         struct net *net = sock_net(skb->sk);
5208         struct net *tgt_net = net;
5209         int idx, s_idx, s_ip_idx;
5210         int h, s_h;
5211         struct net_device *dev;
5212         struct inet6_dev *idev;
5213         struct hlist_head *head;
5214         int err = 0;
5215
5216         s_h = cb->args[0];
5217         s_idx = idx = cb->args[1];
5218         s_ip_idx = cb->args[2];
5219
5220         if (cb->strict_check) {
5221                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5222                                                   skb->sk, cb);
5223                 if (err < 0)
5224                         goto put_tgt_net;
5225
5226                 err = 0;
5227                 if (fillargs.ifindex) {
5228                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5229                         if (!dev) {
5230                                 err = -ENODEV;
5231                                 goto put_tgt_net;
5232                         }
5233                         idev = __in6_dev_get(dev);
5234                         if (idev) {
5235                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5236                                                      &fillargs);
5237                                 if (err > 0)
5238                                         err = 0;
5239                         }
5240                         goto put_tgt_net;
5241                 }
5242         }
5243
5244         rcu_read_lock();
5245         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5246         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5247                 idx = 0;
5248                 head = &tgt_net->dev_index_head[h];
5249                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5250                         if (idx < s_idx)
5251                                 goto cont;
5252                         if (h > s_h || idx > s_idx)
5253                                 s_ip_idx = 0;
5254                         idev = __in6_dev_get(dev);
5255                         if (!idev)
5256                                 goto cont;
5257
5258                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5259                                            &fillargs) < 0)
5260                                 goto done;
5261 cont:
5262                         idx++;
5263                 }
5264         }
5265 done:
5266         rcu_read_unlock();
5267         cb->args[0] = h;
5268         cb->args[1] = idx;
5269 put_tgt_net:
5270         if (fillargs.netnsid >= 0)
5271                 put_net(tgt_net);
5272
5273         return skb->len ? : err;
5274 }
5275
5276 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5277 {
5278         enum addr_type_t type = UNICAST_ADDR;
5279
5280         return inet6_dump_addr(skb, cb, type);
5281 }
5282
5283 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5284 {
5285         enum addr_type_t type = MULTICAST_ADDR;
5286
5287         return inet6_dump_addr(skb, cb, type);
5288 }
5289
5290
5291 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5292 {
5293         enum addr_type_t type = ANYCAST_ADDR;
5294
5295         return inet6_dump_addr(skb, cb, type);
5296 }
5297
5298 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5299                                        const struct nlmsghdr *nlh,
5300                                        struct nlattr **tb,
5301                                        struct netlink_ext_ack *extack)
5302 {
5303         struct ifaddrmsg *ifm;
5304         int i, err;
5305
5306         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5307                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5308                 return -EINVAL;
5309         }
5310
5311         if (!netlink_strict_get_check(skb))
5312                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5313                                               ifa_ipv6_policy, extack);
5314
5315         ifm = nlmsg_data(nlh);
5316         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5317                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5318                 return -EINVAL;
5319         }
5320
5321         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5322                                             ifa_ipv6_policy, extack);
5323         if (err)
5324                 return err;
5325
5326         for (i = 0; i <= IFA_MAX; i++) {
5327                 if (!tb[i])
5328                         continue;
5329
5330                 switch (i) {
5331                 case IFA_TARGET_NETNSID:
5332                 case IFA_ADDRESS:
5333                 case IFA_LOCAL:
5334                         break;
5335                 default:
5336                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5337                         return -EINVAL;
5338                 }
5339         }
5340
5341         return 0;
5342 }
5343
5344 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5345                              struct netlink_ext_ack *extack)
5346 {
5347         struct net *net = sock_net(in_skb->sk);
5348         struct inet6_fill_args fillargs = {
5349                 .portid = NETLINK_CB(in_skb).portid,
5350                 .seq = nlh->nlmsg_seq,
5351                 .event = RTM_NEWADDR,
5352                 .flags = 0,
5353                 .netnsid = -1,
5354         };
5355         struct net *tgt_net = net;
5356         struct ifaddrmsg *ifm;
5357         struct nlattr *tb[IFA_MAX+1];
5358         struct in6_addr *addr = NULL, *peer;
5359         struct net_device *dev = NULL;
5360         struct inet6_ifaddr *ifa;
5361         struct sk_buff *skb;
5362         int err;
5363
5364         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5365         if (err < 0)
5366                 return err;
5367
5368         if (tb[IFA_TARGET_NETNSID]) {
5369                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5370
5371                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5372                                                   fillargs.netnsid);
5373                 if (IS_ERR(tgt_net))
5374                         return PTR_ERR(tgt_net);
5375         }
5376
5377         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5378         if (!addr)
5379                 return -EINVAL;
5380
5381         ifm = nlmsg_data(nlh);
5382         if (ifm->ifa_index)
5383                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5384
5385         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5386         if (!ifa) {
5387                 err = -EADDRNOTAVAIL;
5388                 goto errout;
5389         }
5390
5391         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5392         if (!skb) {
5393                 err = -ENOBUFS;
5394                 goto errout_ifa;
5395         }
5396
5397         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5398         if (err < 0) {
5399                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5400                 WARN_ON(err == -EMSGSIZE);
5401                 kfree_skb(skb);
5402                 goto errout_ifa;
5403         }
5404         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5405 errout_ifa:
5406         in6_ifa_put(ifa);
5407 errout:
5408         if (dev)
5409                 dev_put(dev);
5410         if (fillargs.netnsid >= 0)
5411                 put_net(tgt_net);
5412
5413         return err;
5414 }
5415
5416 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5417 {
5418         struct sk_buff *skb;
5419         struct net *net = dev_net(ifa->idev->dev);
5420         struct inet6_fill_args fillargs = {
5421                 .portid = 0,
5422                 .seq = 0,
5423                 .event = event,
5424                 .flags = 0,
5425                 .netnsid = -1,
5426         };
5427         int err = -ENOBUFS;
5428
5429         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5430         if (!skb)
5431                 goto errout;
5432
5433         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5434         if (err < 0) {
5435                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5436                 WARN_ON(err == -EMSGSIZE);
5437                 kfree_skb(skb);
5438                 goto errout;
5439         }
5440         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5441         return;
5442 errout:
5443         if (err < 0)
5444                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5445 }
5446
5447 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5448                                 __s32 *array, int bytes)
5449 {
5450         BUG_ON(bytes < (DEVCONF_MAX * 4));
5451
5452         memset(array, 0, bytes);
5453         array[DEVCONF_FORWARDING] = cnf->forwarding;
5454         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5455         array[DEVCONF_MTU6] = cnf->mtu6;
5456         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5457         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5458         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5459         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5460         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5461         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5462                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5463         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5464                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5465         array[DEVCONF_RTR_SOLICIT_DELAY] =
5466                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5467         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5468         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5469                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5470         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5471                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5472         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5473         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5474         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5475         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5476         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5477         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5478         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5479         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5480         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5481 #ifdef CONFIG_IPV6_ROUTER_PREF
5482         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5483         array[DEVCONF_RTR_PROBE_INTERVAL] =
5484                 jiffies_to_msecs(cnf->rtr_probe_interval);
5485 #ifdef CONFIG_IPV6_ROUTE_INFO
5486         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5487         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5488 #endif
5489 #endif
5490         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5491         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5492 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5493         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5494         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5495 #endif
5496 #ifdef CONFIG_IPV6_MROUTE
5497         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5498 #endif
5499         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5500         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5501         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5502         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5503         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5504         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5505         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5506         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5507         /* we omit DEVCONF_STABLE_SECRET for now */
5508         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5509         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5510         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5511         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5512         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5513 #ifdef CONFIG_IPV6_SEG6_HMAC
5514         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5515 #endif
5516         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5517         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5518         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5519         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5520         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5521 }
5522
5523 static inline size_t inet6_ifla6_size(void)
5524 {
5525         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5526              + nla_total_size(sizeof(struct ifla_cacheinfo))
5527              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5528              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5529              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5530              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5531              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5532              + 0;
5533 }
5534
5535 static inline size_t inet6_if_nlmsg_size(void)
5536 {
5537         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5538                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5539                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5540                + nla_total_size(4) /* IFLA_MTU */
5541                + nla_total_size(4) /* IFLA_LINK */
5542                + nla_total_size(1) /* IFLA_OPERSTATE */
5543                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5544 }
5545
5546 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5547                                         int bytes)
5548 {
5549         int i;
5550         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5551         BUG_ON(pad < 0);
5552
5553         /* Use put_unaligned() because stats may not be aligned for u64. */
5554         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5555         for (i = 1; i < ICMP6_MIB_MAX; i++)
5556                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5557
5558         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5559 }
5560
5561 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5562                                         int bytes, size_t syncpoff)
5563 {
5564         int i, c;
5565         u64 buff[IPSTATS_MIB_MAX];
5566         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5567
5568         BUG_ON(pad < 0);
5569
5570         memset(buff, 0, sizeof(buff));
5571         buff[0] = IPSTATS_MIB_MAX;
5572
5573         for_each_possible_cpu(c) {
5574                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5575                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5576         }
5577
5578         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5579         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5580 }
5581
5582 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5583                              int bytes)
5584 {
5585         switch (attrtype) {
5586         case IFLA_INET6_STATS:
5587                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5588                                      offsetof(struct ipstats_mib, syncp));
5589                 break;
5590         case IFLA_INET6_ICMP6STATS:
5591                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5592                 break;
5593         }
5594 }
5595
5596 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5597                                   u32 ext_filter_mask)
5598 {
5599         struct nlattr *nla;
5600         struct ifla_cacheinfo ci;
5601
5602         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5603                 goto nla_put_failure;
5604         ci.max_reasm_len = IPV6_MAXPLEN;
5605         ci.tstamp = cstamp_delta(idev->tstamp);
5606         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5607         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5608         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5609                 goto nla_put_failure;
5610         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5611         if (!nla)
5612                 goto nla_put_failure;
5613         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5614
5615         /* XXX - MC not implemented */
5616
5617         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5618                 return 0;
5619
5620         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5621         if (!nla)
5622                 goto nla_put_failure;
5623         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5624
5625         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5626         if (!nla)
5627                 goto nla_put_failure;
5628         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5629
5630         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5631         if (!nla)
5632                 goto nla_put_failure;
5633         read_lock_bh(&idev->lock);
5634         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5635         read_unlock_bh(&idev->lock);
5636
5637         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5638                 goto nla_put_failure;
5639
5640         return 0;
5641
5642 nla_put_failure:
5643         return -EMSGSIZE;
5644 }
5645
5646 static size_t inet6_get_link_af_size(const struct net_device *dev,
5647                                      u32 ext_filter_mask)
5648 {
5649         if (!__in6_dev_get(dev))
5650                 return 0;
5651
5652         return inet6_ifla6_size();
5653 }
5654
5655 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5656                               u32 ext_filter_mask)
5657 {
5658         struct inet6_dev *idev = __in6_dev_get(dev);
5659
5660         if (!idev)
5661                 return -ENODATA;
5662
5663         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5664                 return -EMSGSIZE;
5665
5666         return 0;
5667 }
5668
5669 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5670 {
5671         struct inet6_ifaddr *ifp;
5672         struct net_device *dev = idev->dev;
5673         bool clear_token, update_rs = false;
5674         struct in6_addr ll_addr;
5675
5676         ASSERT_RTNL();
5677
5678         if (!token)
5679                 return -EINVAL;
5680         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5681                 return -EINVAL;
5682         if (!ipv6_accept_ra(idev))
5683                 return -EINVAL;
5684         if (idev->cnf.rtr_solicits == 0)
5685                 return -EINVAL;
5686
5687         write_lock_bh(&idev->lock);
5688
5689         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5690         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5691
5692         write_unlock_bh(&idev->lock);
5693
5694         clear_token = ipv6_addr_any(token);
5695         if (clear_token)
5696                 goto update_lft;
5697
5698         if (!idev->dead && (idev->if_flags & IF_READY) &&
5699             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5700                              IFA_F_OPTIMISTIC)) {
5701                 /* If we're not ready, then normal ifup will take care
5702                  * of this. Otherwise, we need to request our rs here.
5703                  */
5704                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5705                 update_rs = true;
5706         }
5707
5708 update_lft:
5709         write_lock_bh(&idev->lock);
5710
5711         if (update_rs) {
5712                 idev->if_flags |= IF_RS_SENT;
5713                 idev->rs_interval = rfc3315_s14_backoff_init(
5714                         idev->cnf.rtr_solicit_interval);
5715                 idev->rs_probes = 1;
5716                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5717         }
5718
5719         /* Well, that's kinda nasty ... */
5720         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5721                 spin_lock(&ifp->lock);
5722                 if (ifp->tokenized) {
5723                         ifp->valid_lft = 0;
5724                         ifp->prefered_lft = 0;
5725                 }
5726                 spin_unlock(&ifp->lock);
5727         }
5728
5729         write_unlock_bh(&idev->lock);
5730         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5731         addrconf_verify_rtnl();
5732         return 0;
5733 }
5734
5735 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5736         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5737         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5738 };
5739
5740 static int check_addr_gen_mode(int mode)
5741 {
5742         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5743             mode != IN6_ADDR_GEN_MODE_NONE &&
5744             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5745             mode != IN6_ADDR_GEN_MODE_RANDOM)
5746                 return -EINVAL;
5747         return 1;
5748 }
5749
5750 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5751                                 int mode)
5752 {
5753         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5754             !idev->cnf.stable_secret.initialized &&
5755             !net->ipv6.devconf_dflt->stable_secret.initialized)
5756                 return -EINVAL;
5757         return 1;
5758 }
5759
5760 static int inet6_validate_link_af(const struct net_device *dev,
5761                                   const struct nlattr *nla)
5762 {
5763         struct nlattr *tb[IFLA_INET6_MAX + 1];
5764         struct inet6_dev *idev = NULL;
5765         int err;
5766
5767         if (dev) {
5768                 idev = __in6_dev_get(dev);
5769                 if (!idev)
5770                         return -EAFNOSUPPORT;
5771         }
5772
5773         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5774                                           inet6_af_policy, NULL);
5775         if (err)
5776                 return err;
5777
5778         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5779                 return -EINVAL;
5780
5781         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5782                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5783
5784                 if (check_addr_gen_mode(mode) < 0)
5785                         return -EINVAL;
5786                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5787                         return -EINVAL;
5788         }
5789
5790         return 0;
5791 }
5792
5793 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5794 {
5795         struct inet6_dev *idev = __in6_dev_get(dev);
5796         struct nlattr *tb[IFLA_INET6_MAX + 1];
5797         int err;
5798
5799         if (!idev)
5800                 return -EAFNOSUPPORT;
5801
5802         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5803                 BUG();
5804
5805         if (tb[IFLA_INET6_TOKEN]) {
5806                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5807                 if (err)
5808                         return err;
5809         }
5810
5811         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5812                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5813
5814                 idev->cnf.addr_gen_mode = mode;
5815         }
5816
5817         return 0;
5818 }
5819
5820 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5821                              u32 portid, u32 seq, int event, unsigned int flags)
5822 {
5823         struct net_device *dev = idev->dev;
5824         struct ifinfomsg *hdr;
5825         struct nlmsghdr *nlh;
5826         void *protoinfo;
5827
5828         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5829         if (!nlh)
5830                 return -EMSGSIZE;
5831
5832         hdr = nlmsg_data(nlh);
5833         hdr->ifi_family = AF_INET6;
5834         hdr->__ifi_pad = 0;
5835         hdr->ifi_type = dev->type;
5836         hdr->ifi_index = dev->ifindex;
5837         hdr->ifi_flags = dev_get_flags(dev);
5838         hdr->ifi_change = 0;
5839
5840         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5841             (dev->addr_len &&
5842              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5843             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5844             (dev->ifindex != dev_get_iflink(dev) &&
5845              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5846             nla_put_u8(skb, IFLA_OPERSTATE,
5847                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5848                 goto nla_put_failure;
5849         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5850         if (!protoinfo)
5851                 goto nla_put_failure;
5852
5853         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5854                 goto nla_put_failure;
5855
5856         nla_nest_end(skb, protoinfo);
5857         nlmsg_end(skb, nlh);
5858         return 0;
5859
5860 nla_put_failure:
5861         nlmsg_cancel(skb, nlh);
5862         return -EMSGSIZE;
5863 }
5864
5865 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5866                                    struct netlink_ext_ack *extack)
5867 {
5868         struct ifinfomsg *ifm;
5869
5870         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5871                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5872                 return -EINVAL;
5873         }
5874
5875         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5876                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5877                 return -EINVAL;
5878         }
5879
5880         ifm = nlmsg_data(nlh);
5881         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5882             ifm->ifi_change || ifm->ifi_index) {
5883                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5884                 return -EINVAL;
5885         }
5886
5887         return 0;
5888 }
5889
5890 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5891 {
5892         struct net *net = sock_net(skb->sk);
5893         int h, s_h;
5894         int idx = 0, s_idx;
5895         struct net_device *dev;
5896         struct inet6_dev *idev;
5897         struct hlist_head *head;
5898
5899         /* only requests using strict checking can pass data to
5900          * influence the dump
5901          */
5902         if (cb->strict_check) {
5903                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5904
5905                 if (err < 0)
5906                         return err;
5907         }
5908
5909         s_h = cb->args[0];
5910         s_idx = cb->args[1];
5911
5912         rcu_read_lock();
5913         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5914                 idx = 0;
5915                 head = &net->dev_index_head[h];
5916                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5917                         if (idx < s_idx)
5918                                 goto cont;
5919                         idev = __in6_dev_get(dev);
5920                         if (!idev)
5921                                 goto cont;
5922                         if (inet6_fill_ifinfo(skb, idev,
5923                                               NETLINK_CB(cb->skb).portid,
5924                                               cb->nlh->nlmsg_seq,
5925                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5926                                 goto out;
5927 cont:
5928                         idx++;
5929                 }
5930         }
5931 out:
5932         rcu_read_unlock();
5933         cb->args[1] = idx;
5934         cb->args[0] = h;
5935
5936         return skb->len;
5937 }
5938
5939 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5940 {
5941         struct sk_buff *skb;
5942         struct net *net = dev_net(idev->dev);
5943         int err = -ENOBUFS;
5944
5945         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5946         if (!skb)
5947                 goto errout;
5948
5949         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5950         if (err < 0) {
5951                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5952                 WARN_ON(err == -EMSGSIZE);
5953                 kfree_skb(skb);
5954                 goto errout;
5955         }
5956         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5957         return;
5958 errout:
5959         if (err < 0)
5960                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5961 }
5962
5963 static inline size_t inet6_prefix_nlmsg_size(void)
5964 {
5965         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5966                + nla_total_size(sizeof(struct in6_addr))
5967                + nla_total_size(sizeof(struct prefix_cacheinfo));
5968 }
5969
5970 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5971                              struct prefix_info *pinfo, u32 portid, u32 seq,
5972                              int event, unsigned int flags)
5973 {
5974         struct prefixmsg *pmsg;
5975         struct nlmsghdr *nlh;
5976         struct prefix_cacheinfo ci;
5977
5978         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5979         if (!nlh)
5980                 return -EMSGSIZE;
5981
5982         pmsg = nlmsg_data(nlh);
5983         pmsg->prefix_family = AF_INET6;
5984         pmsg->prefix_pad1 = 0;
5985         pmsg->prefix_pad2 = 0;
5986         pmsg->prefix_ifindex = idev->dev->ifindex;
5987         pmsg->prefix_len = pinfo->prefix_len;
5988         pmsg->prefix_type = pinfo->type;
5989         pmsg->prefix_pad3 = 0;
5990         pmsg->prefix_flags = 0;
5991         if (pinfo->onlink)
5992                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5993         if (pinfo->autoconf)
5994                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5995
5996         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5997                 goto nla_put_failure;
5998         ci.preferred_time = ntohl(pinfo->prefered);
5999         ci.valid_time = ntohl(pinfo->valid);
6000         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6001                 goto nla_put_failure;
6002         nlmsg_end(skb, nlh);
6003         return 0;
6004
6005 nla_put_failure:
6006         nlmsg_cancel(skb, nlh);
6007         return -EMSGSIZE;
6008 }
6009
6010 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6011                          struct prefix_info *pinfo)
6012 {
6013         struct sk_buff *skb;
6014         struct net *net = dev_net(idev->dev);
6015         int err = -ENOBUFS;
6016
6017         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6018         if (!skb)
6019                 goto errout;
6020
6021         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6022         if (err < 0) {
6023                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6024                 WARN_ON(err == -EMSGSIZE);
6025                 kfree_skb(skb);
6026                 goto errout;
6027         }
6028         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6029         return;
6030 errout:
6031         if (err < 0)
6032                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6033 }
6034
6035 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6036 {
6037         struct net *net = dev_net(ifp->idev->dev);
6038
6039         if (event)
6040                 ASSERT_RTNL();
6041
6042         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6043
6044         switch (event) {
6045         case RTM_NEWADDR:
6046                 /*
6047                  * If the address was optimistic we inserted the route at the
6048                  * start of our DAD process, so we don't need to do it again.
6049                  * If the device was taken down in the middle of the DAD
6050                  * cycle there is a race where we could get here without a
6051                  * host route, so nothing to insert. That will be fixed when
6052                  * the device is brought up.
6053                  */
6054                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6055                         ip6_ins_rt(net, ifp->rt);
6056                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6057                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6058                                 &ifp->addr, ifp->idev->dev->name);
6059                 }
6060
6061                 if (ifp->idev->cnf.forwarding)
6062                         addrconf_join_anycast(ifp);
6063                 if (!ipv6_addr_any(&ifp->peer_addr))
6064                         addrconf_prefix_route(&ifp->peer_addr, 128,
6065                                               ifp->rt_priority, ifp->idev->dev,
6066                                               0, 0, GFP_ATOMIC);
6067                 break;
6068         case RTM_DELADDR:
6069                 if (ifp->idev->cnf.forwarding)
6070                         addrconf_leave_anycast(ifp);
6071                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6072                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6073                         struct fib6_info *rt;
6074
6075                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6076                                                        ifp->idev->dev, 0, 0,
6077                                                        false);
6078                         if (rt)
6079                                 ip6_del_rt(net, rt, false);
6080                 }
6081                 if (ifp->rt) {
6082                         ip6_del_rt(net, ifp->rt, false);
6083                         ifp->rt = NULL;
6084                 }
6085                 rt_genid_bump_ipv6(net);
6086                 break;
6087         }
6088         atomic_inc(&net->ipv6.dev_addr_genid);
6089 }
6090
6091 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6092 {
6093         rcu_read_lock_bh();
6094         if (likely(ifp->idev->dead == 0))
6095                 __ipv6_ifa_notify(event, ifp);
6096         rcu_read_unlock_bh();
6097 }
6098
6099 #ifdef CONFIG_SYSCTL
6100
6101 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6102                 void *buffer, size_t *lenp, loff_t *ppos)
6103 {
6104         int *valp = ctl->data;
6105         int val = *valp;
6106         loff_t pos = *ppos;
6107         struct ctl_table lctl;
6108         int ret;
6109
6110         /*
6111          * ctl->data points to idev->cnf.forwarding, we should
6112          * not modify it until we get the rtnl lock.
6113          */
6114         lctl = *ctl;
6115         lctl.data = &val;
6116
6117         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6118
6119         if (write)
6120                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6121         if (ret)
6122                 *ppos = pos;
6123         return ret;
6124 }
6125
6126 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6127                 void *buffer, size_t *lenp, loff_t *ppos)
6128 {
6129         struct inet6_dev *idev = ctl->extra1;
6130         int min_mtu = IPV6_MIN_MTU;
6131         struct ctl_table lctl;
6132
6133         lctl = *ctl;
6134         lctl.extra1 = &min_mtu;
6135         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6136
6137         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6138 }
6139
6140 static void dev_disable_change(struct inet6_dev *idev)
6141 {
6142         struct netdev_notifier_info info;
6143
6144         if (!idev || !idev->dev)
6145                 return;
6146
6147         netdev_notifier_info_init(&info, idev->dev);
6148         if (idev->cnf.disable_ipv6)
6149                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6150         else
6151                 addrconf_notify(NULL, NETDEV_UP, &info);
6152 }
6153
6154 static void addrconf_disable_change(struct net *net, __s32 newf)
6155 {
6156         struct net_device *dev;
6157         struct inet6_dev *idev;
6158
6159         for_each_netdev(net, dev) {
6160                 idev = __in6_dev_get(dev);
6161                 if (idev) {
6162                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6163                         idev->cnf.disable_ipv6 = newf;
6164                         if (changed)
6165                                 dev_disable_change(idev);
6166                 }
6167         }
6168 }
6169
6170 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6171 {
6172         struct net *net;
6173         int old;
6174
6175         if (!rtnl_trylock())
6176                 return restart_syscall();
6177
6178         net = (struct net *)table->extra2;
6179         old = *p;
6180         *p = newf;
6181
6182         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6183                 rtnl_unlock();
6184                 return 0;
6185         }
6186
6187         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6188                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6189                 addrconf_disable_change(net, newf);
6190         } else if ((!newf) ^ (!old))
6191                 dev_disable_change((struct inet6_dev *)table->extra1);
6192
6193         rtnl_unlock();
6194         return 0;
6195 }
6196
6197 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6198                 void *buffer, size_t *lenp, loff_t *ppos)
6199 {
6200         int *valp = ctl->data;
6201         int val = *valp;
6202         loff_t pos = *ppos;
6203         struct ctl_table lctl;
6204         int ret;
6205
6206         /*
6207          * ctl->data points to idev->cnf.disable_ipv6, we should
6208          * not modify it until we get the rtnl lock.
6209          */
6210         lctl = *ctl;
6211         lctl.data = &val;
6212
6213         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6214
6215         if (write)
6216                 ret = addrconf_disable_ipv6(ctl, valp, val);
6217         if (ret)
6218                 *ppos = pos;
6219         return ret;
6220 }
6221
6222 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6223                 void *buffer, size_t *lenp, loff_t *ppos)
6224 {
6225         int *valp = ctl->data;
6226         int ret;
6227         int old, new;
6228
6229         old = *valp;
6230         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6231         new = *valp;
6232
6233         if (write && old != new) {
6234                 struct net *net = ctl->extra2;
6235
6236                 if (!rtnl_trylock())
6237                         return restart_syscall();
6238
6239                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6240                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6241                                                      NETCONFA_PROXY_NEIGH,
6242                                                      NETCONFA_IFINDEX_DEFAULT,
6243                                                      net->ipv6.devconf_dflt);
6244                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6245                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6246                                                      NETCONFA_PROXY_NEIGH,
6247                                                      NETCONFA_IFINDEX_ALL,
6248                                                      net->ipv6.devconf_all);
6249                 else {
6250                         struct inet6_dev *idev = ctl->extra1;
6251
6252                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6253                                                      NETCONFA_PROXY_NEIGH,
6254                                                      idev->dev->ifindex,
6255                                                      &idev->cnf);
6256                 }
6257                 rtnl_unlock();
6258         }
6259
6260         return ret;
6261 }
6262
6263 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6264                                          void *buffer, size_t *lenp,
6265                                          loff_t *ppos)
6266 {
6267         int ret = 0;
6268         u32 new_val;
6269         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6270         struct net *net = (struct net *)ctl->extra2;
6271         struct ctl_table tmp = {
6272                 .data = &new_val,
6273                 .maxlen = sizeof(new_val),
6274                 .mode = ctl->mode,
6275         };
6276
6277         if (!rtnl_trylock())
6278                 return restart_syscall();
6279
6280         new_val = *((u32 *)ctl->data);
6281
6282         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6283         if (ret != 0)
6284                 goto out;
6285
6286         if (write) {
6287                 if (check_addr_gen_mode(new_val) < 0) {
6288                         ret = -EINVAL;
6289                         goto out;
6290                 }
6291
6292                 if (idev) {
6293                         if (check_stable_privacy(idev, net, new_val) < 0) {
6294                                 ret = -EINVAL;
6295                                 goto out;
6296                         }
6297
6298                         if (idev->cnf.addr_gen_mode != new_val) {
6299                                 idev->cnf.addr_gen_mode = new_val;
6300                                 addrconf_dev_config(idev->dev);
6301                         }
6302                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6303                         struct net_device *dev;
6304
6305                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6306                         for_each_netdev(net, dev) {
6307                                 idev = __in6_dev_get(dev);
6308                                 if (idev &&
6309                                     idev->cnf.addr_gen_mode != new_val) {
6310                                         idev->cnf.addr_gen_mode = new_val;
6311                                         addrconf_dev_config(idev->dev);
6312                                 }
6313                         }
6314                 }
6315
6316                 *((u32 *)ctl->data) = new_val;
6317         }
6318
6319 out:
6320         rtnl_unlock();
6321
6322         return ret;
6323 }
6324
6325 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6326                                          void *buffer, size_t *lenp,
6327                                          loff_t *ppos)
6328 {
6329         int err;
6330         struct in6_addr addr;
6331         char str[IPV6_MAX_STRLEN];
6332         struct ctl_table lctl = *ctl;
6333         struct net *net = ctl->extra2;
6334         struct ipv6_stable_secret *secret = ctl->data;
6335
6336         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6337                 return -EIO;
6338
6339         lctl.maxlen = IPV6_MAX_STRLEN;
6340         lctl.data = str;
6341
6342         if (!rtnl_trylock())
6343                 return restart_syscall();
6344
6345         if (!write && !secret->initialized) {
6346                 err = -EIO;
6347                 goto out;
6348         }
6349
6350         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6351         if (err >= sizeof(str)) {
6352                 err = -EIO;
6353                 goto out;
6354         }
6355
6356         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6357         if (err || !write)
6358                 goto out;
6359
6360         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6361                 err = -EIO;
6362                 goto out;
6363         }
6364
6365         secret->initialized = true;
6366         secret->secret = addr;
6367
6368         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6369                 struct net_device *dev;
6370
6371                 for_each_netdev(net, dev) {
6372                         struct inet6_dev *idev = __in6_dev_get(dev);
6373
6374                         if (idev) {
6375                                 idev->cnf.addr_gen_mode =
6376                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6377                         }
6378                 }
6379         } else {
6380                 struct inet6_dev *idev = ctl->extra1;
6381
6382                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6383         }
6384
6385 out:
6386         rtnl_unlock();
6387
6388         return err;
6389 }
6390
6391 static
6392 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6393                                                 int write, void *buffer,
6394                                                 size_t *lenp,
6395                                                 loff_t *ppos)
6396 {
6397         int *valp = ctl->data;
6398         int val = *valp;
6399         loff_t pos = *ppos;
6400         struct ctl_table lctl;
6401         int ret;
6402
6403         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6404          * we should not modify it until we get the rtnl lock.
6405          */
6406         lctl = *ctl;
6407         lctl.data = &val;
6408
6409         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6410
6411         if (write)
6412                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6413         if (ret)
6414                 *ppos = pos;
6415         return ret;
6416 }
6417
6418 static
6419 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6420 {
6421         if (rt) {
6422                 if (action)
6423                         rt->dst.flags |= DST_NOPOLICY;
6424                 else
6425                         rt->dst.flags &= ~DST_NOPOLICY;
6426         }
6427 }
6428
6429 static
6430 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6431 {
6432         struct inet6_ifaddr *ifa;
6433
6434         read_lock_bh(&idev->lock);
6435         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6436                 spin_lock(&ifa->lock);
6437                 if (ifa->rt) {
6438                         /* host routes only use builtin fib6_nh */
6439                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6440                         int cpu;
6441
6442                         rcu_read_lock();
6443                         ifa->rt->dst_nopolicy = val ? true : false;
6444                         if (nh->rt6i_pcpu) {
6445                                 for_each_possible_cpu(cpu) {
6446                                         struct rt6_info **rtp;
6447
6448                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6449                                         addrconf_set_nopolicy(*rtp, val);
6450                                 }
6451                         }
6452                         rcu_read_unlock();
6453                 }
6454                 spin_unlock(&ifa->lock);
6455         }
6456         read_unlock_bh(&idev->lock);
6457 }
6458
6459 static
6460 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6461 {
6462         struct inet6_dev *idev;
6463         struct net *net;
6464
6465         if (!rtnl_trylock())
6466                 return restart_syscall();
6467
6468         *valp = val;
6469
6470         net = (struct net *)ctl->extra2;
6471         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6472                 rtnl_unlock();
6473                 return 0;
6474         }
6475
6476         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6477                 struct net_device *dev;
6478
6479                 for_each_netdev(net, dev) {
6480                         idev = __in6_dev_get(dev);
6481                         if (idev)
6482                                 addrconf_disable_policy_idev(idev, val);
6483                 }
6484         } else {
6485                 idev = (struct inet6_dev *)ctl->extra1;
6486                 addrconf_disable_policy_idev(idev, val);
6487         }
6488
6489         rtnl_unlock();
6490         return 0;
6491 }
6492
6493 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6494                                    void *buffer, size_t *lenp, loff_t *ppos)
6495 {
6496         int *valp = ctl->data;
6497         int val = *valp;
6498         loff_t pos = *ppos;
6499         struct ctl_table lctl;
6500         int ret;
6501
6502         lctl = *ctl;
6503         lctl.data = &val;
6504         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6505
6506         if (write && (*valp != val))
6507                 ret = addrconf_disable_policy(ctl, valp, val);
6508
6509         if (ret)
6510                 *ppos = pos;
6511
6512         return ret;
6513 }
6514
6515 static int minus_one = -1;
6516 static const int two_five_five = 255;
6517
6518 static const struct ctl_table addrconf_sysctl[] = {
6519         {
6520                 .procname       = "forwarding",
6521                 .data           = &ipv6_devconf.forwarding,
6522                 .maxlen         = sizeof(int),
6523                 .mode           = 0644,
6524                 .proc_handler   = addrconf_sysctl_forward,
6525         },
6526         {
6527                 .procname       = "hop_limit",
6528                 .data           = &ipv6_devconf.hop_limit,
6529                 .maxlen         = sizeof(int),
6530                 .mode           = 0644,
6531                 .proc_handler   = proc_dointvec_minmax,
6532                 .extra1         = (void *)SYSCTL_ONE,
6533                 .extra2         = (void *)&two_five_five,
6534         },
6535         {
6536                 .procname       = "mtu",
6537                 .data           = &ipv6_devconf.mtu6,
6538                 .maxlen         = sizeof(int),
6539                 .mode           = 0644,
6540                 .proc_handler   = addrconf_sysctl_mtu,
6541         },
6542         {
6543                 .procname       = "accept_ra",
6544                 .data           = &ipv6_devconf.accept_ra,
6545                 .maxlen         = sizeof(int),
6546                 .mode           = 0644,
6547                 .proc_handler   = proc_dointvec,
6548         },
6549         {
6550                 .procname       = "accept_redirects",
6551                 .data           = &ipv6_devconf.accept_redirects,
6552                 .maxlen         = sizeof(int),
6553                 .mode           = 0644,
6554                 .proc_handler   = proc_dointvec,
6555         },
6556         {
6557                 .procname       = "autoconf",
6558                 .data           = &ipv6_devconf.autoconf,
6559                 .maxlen         = sizeof(int),
6560                 .mode           = 0644,
6561                 .proc_handler   = proc_dointvec,
6562         },
6563         {
6564                 .procname       = "dad_transmits",
6565                 .data           = &ipv6_devconf.dad_transmits,
6566                 .maxlen         = sizeof(int),
6567                 .mode           = 0644,
6568                 .proc_handler   = proc_dointvec,
6569         },
6570         {
6571                 .procname       = "router_solicitations",
6572                 .data           = &ipv6_devconf.rtr_solicits,
6573                 .maxlen         = sizeof(int),
6574                 .mode           = 0644,
6575                 .proc_handler   = proc_dointvec_minmax,
6576                 .extra1         = &minus_one,
6577         },
6578         {
6579                 .procname       = "router_solicitation_interval",
6580                 .data           = &ipv6_devconf.rtr_solicit_interval,
6581                 .maxlen         = sizeof(int),
6582                 .mode           = 0644,
6583                 .proc_handler   = proc_dointvec_jiffies,
6584         },
6585         {
6586                 .procname       = "router_solicitation_max_interval",
6587                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6588                 .maxlen         = sizeof(int),
6589                 .mode           = 0644,
6590                 .proc_handler   = proc_dointvec_jiffies,
6591         },
6592         {
6593                 .procname       = "router_solicitation_delay",
6594                 .data           = &ipv6_devconf.rtr_solicit_delay,
6595                 .maxlen         = sizeof(int),
6596                 .mode           = 0644,
6597                 .proc_handler   = proc_dointvec_jiffies,
6598         },
6599         {
6600                 .procname       = "force_mld_version",
6601                 .data           = &ipv6_devconf.force_mld_version,
6602                 .maxlen         = sizeof(int),
6603                 .mode           = 0644,
6604                 .proc_handler   = proc_dointvec,
6605         },
6606         {
6607                 .procname       = "mldv1_unsolicited_report_interval",
6608                 .data           =
6609                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6610                 .maxlen         = sizeof(int),
6611                 .mode           = 0644,
6612                 .proc_handler   = proc_dointvec_ms_jiffies,
6613         },
6614         {
6615                 .procname       = "mldv2_unsolicited_report_interval",
6616                 .data           =
6617                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6618                 .maxlen         = sizeof(int),
6619                 .mode           = 0644,
6620                 .proc_handler   = proc_dointvec_ms_jiffies,
6621         },
6622         {
6623                 .procname       = "use_tempaddr",
6624                 .data           = &ipv6_devconf.use_tempaddr,
6625                 .maxlen         = sizeof(int),
6626                 .mode           = 0644,
6627                 .proc_handler   = proc_dointvec,
6628         },
6629         {
6630                 .procname       = "temp_valid_lft",
6631                 .data           = &ipv6_devconf.temp_valid_lft,
6632                 .maxlen         = sizeof(int),
6633                 .mode           = 0644,
6634                 .proc_handler   = proc_dointvec,
6635         },
6636         {
6637                 .procname       = "temp_prefered_lft",
6638                 .data           = &ipv6_devconf.temp_prefered_lft,
6639                 .maxlen         = sizeof(int),
6640                 .mode           = 0644,
6641                 .proc_handler   = proc_dointvec,
6642         },
6643         {
6644                 .procname       = "regen_max_retry",
6645                 .data           = &ipv6_devconf.regen_max_retry,
6646                 .maxlen         = sizeof(int),
6647                 .mode           = 0644,
6648                 .proc_handler   = proc_dointvec,
6649         },
6650         {
6651                 .procname       = "max_desync_factor",
6652                 .data           = &ipv6_devconf.max_desync_factor,
6653                 .maxlen         = sizeof(int),
6654                 .mode           = 0644,
6655                 .proc_handler   = proc_dointvec,
6656         },
6657         {
6658                 .procname       = "max_addresses",
6659                 .data           = &ipv6_devconf.max_addresses,
6660                 .maxlen         = sizeof(int),
6661                 .mode           = 0644,
6662                 .proc_handler   = proc_dointvec,
6663         },
6664         {
6665                 .procname       = "accept_ra_defrtr",
6666                 .data           = &ipv6_devconf.accept_ra_defrtr,
6667                 .maxlen         = sizeof(int),
6668                 .mode           = 0644,
6669                 .proc_handler   = proc_dointvec,
6670         },
6671         {
6672                 .procname       = "accept_ra_min_hop_limit",
6673                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6674                 .maxlen         = sizeof(int),
6675                 .mode           = 0644,
6676                 .proc_handler   = proc_dointvec,
6677         },
6678         {
6679                 .procname       = "accept_ra_pinfo",
6680                 .data           = &ipv6_devconf.accept_ra_pinfo,
6681                 .maxlen         = sizeof(int),
6682                 .mode           = 0644,
6683                 .proc_handler   = proc_dointvec,
6684         },
6685 #ifdef CONFIG_IPV6_ROUTER_PREF
6686         {
6687                 .procname       = "accept_ra_rtr_pref",
6688                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6689                 .maxlen         = sizeof(int),
6690                 .mode           = 0644,
6691                 .proc_handler   = proc_dointvec,
6692         },
6693         {
6694                 .procname       = "router_probe_interval",
6695                 .data           = &ipv6_devconf.rtr_probe_interval,
6696                 .maxlen         = sizeof(int),
6697                 .mode           = 0644,
6698                 .proc_handler   = proc_dointvec_jiffies,
6699         },
6700 #ifdef CONFIG_IPV6_ROUTE_INFO
6701         {
6702                 .procname       = "accept_ra_rt_info_min_plen",
6703                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6704                 .maxlen         = sizeof(int),
6705                 .mode           = 0644,
6706                 .proc_handler   = proc_dointvec,
6707         },
6708         {
6709                 .procname       = "accept_ra_rt_info_max_plen",
6710                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6711                 .maxlen         = sizeof(int),
6712                 .mode           = 0644,
6713                 .proc_handler   = proc_dointvec,
6714         },
6715 #endif
6716 #endif
6717         {
6718                 .procname       = "proxy_ndp",
6719                 .data           = &ipv6_devconf.proxy_ndp,
6720                 .maxlen         = sizeof(int),
6721                 .mode           = 0644,
6722                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6723         },
6724         {
6725                 .procname       = "accept_source_route",
6726                 .data           = &ipv6_devconf.accept_source_route,
6727                 .maxlen         = sizeof(int),
6728                 .mode           = 0644,
6729                 .proc_handler   = proc_dointvec,
6730         },
6731 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6732         {
6733                 .procname       = "optimistic_dad",
6734                 .data           = &ipv6_devconf.optimistic_dad,
6735                 .maxlen         = sizeof(int),
6736                 .mode           = 0644,
6737                 .proc_handler   = proc_dointvec,
6738         },
6739         {
6740                 .procname       = "use_optimistic",
6741                 .data           = &ipv6_devconf.use_optimistic,
6742                 .maxlen         = sizeof(int),
6743                 .mode           = 0644,
6744                 .proc_handler   = proc_dointvec,
6745         },
6746 #endif
6747 #ifdef CONFIG_IPV6_MROUTE
6748         {
6749                 .procname       = "mc_forwarding",
6750                 .data           = &ipv6_devconf.mc_forwarding,
6751                 .maxlen         = sizeof(int),
6752                 .mode           = 0444,
6753                 .proc_handler   = proc_dointvec,
6754         },
6755 #endif
6756         {
6757                 .procname       = "disable_ipv6",
6758                 .data           = &ipv6_devconf.disable_ipv6,
6759                 .maxlen         = sizeof(int),
6760                 .mode           = 0644,
6761                 .proc_handler   = addrconf_sysctl_disable,
6762         },
6763         {
6764                 .procname       = "accept_dad",
6765                 .data           = &ipv6_devconf.accept_dad,
6766                 .maxlen         = sizeof(int),
6767                 .mode           = 0644,
6768                 .proc_handler   = proc_dointvec,
6769         },
6770         {
6771                 .procname       = "force_tllao",
6772                 .data           = &ipv6_devconf.force_tllao,
6773                 .maxlen         = sizeof(int),
6774                 .mode           = 0644,
6775                 .proc_handler   = proc_dointvec
6776         },
6777         {
6778                 .procname       = "ndisc_notify",
6779                 .data           = &ipv6_devconf.ndisc_notify,
6780                 .maxlen         = sizeof(int),
6781                 .mode           = 0644,
6782                 .proc_handler   = proc_dointvec
6783         },
6784         {
6785                 .procname       = "suppress_frag_ndisc",
6786                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6787                 .maxlen         = sizeof(int),
6788                 .mode           = 0644,
6789                 .proc_handler   = proc_dointvec
6790         },
6791         {
6792                 .procname       = "accept_ra_from_local",
6793                 .data           = &ipv6_devconf.accept_ra_from_local,
6794                 .maxlen         = sizeof(int),
6795                 .mode           = 0644,
6796                 .proc_handler   = proc_dointvec,
6797         },
6798         {
6799                 .procname       = "accept_ra_mtu",
6800                 .data           = &ipv6_devconf.accept_ra_mtu,
6801                 .maxlen         = sizeof(int),
6802                 .mode           = 0644,
6803                 .proc_handler   = proc_dointvec,
6804         },
6805         {
6806                 .procname       = "stable_secret",
6807                 .data           = &ipv6_devconf.stable_secret,
6808                 .maxlen         = IPV6_MAX_STRLEN,
6809                 .mode           = 0600,
6810                 .proc_handler   = addrconf_sysctl_stable_secret,
6811         },
6812         {
6813                 .procname       = "use_oif_addrs_only",
6814                 .data           = &ipv6_devconf.use_oif_addrs_only,
6815                 .maxlen         = sizeof(int),
6816                 .mode           = 0644,
6817                 .proc_handler   = proc_dointvec,
6818         },
6819         {
6820                 .procname       = "ignore_routes_with_linkdown",
6821                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6822                 .maxlen         = sizeof(int),
6823                 .mode           = 0644,
6824                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6825         },
6826         {
6827                 .procname       = "drop_unicast_in_l2_multicast",
6828                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6829                 .maxlen         = sizeof(int),
6830                 .mode           = 0644,
6831                 .proc_handler   = proc_dointvec,
6832         },
6833         {
6834                 .procname       = "drop_unsolicited_na",
6835                 .data           = &ipv6_devconf.drop_unsolicited_na,
6836                 .maxlen         = sizeof(int),
6837                 .mode           = 0644,
6838                 .proc_handler   = proc_dointvec,
6839         },
6840         {
6841                 .procname       = "keep_addr_on_down",
6842                 .data           = &ipv6_devconf.keep_addr_on_down,
6843                 .maxlen         = sizeof(int),
6844                 .mode           = 0644,
6845                 .proc_handler   = proc_dointvec,
6846
6847         },
6848         {
6849                 .procname       = "seg6_enabled",
6850                 .data           = &ipv6_devconf.seg6_enabled,
6851                 .maxlen         = sizeof(int),
6852                 .mode           = 0644,
6853                 .proc_handler   = proc_dointvec,
6854         },
6855 #ifdef CONFIG_IPV6_SEG6_HMAC
6856         {
6857                 .procname       = "seg6_require_hmac",
6858                 .data           = &ipv6_devconf.seg6_require_hmac,
6859                 .maxlen         = sizeof(int),
6860                 .mode           = 0644,
6861                 .proc_handler   = proc_dointvec,
6862         },
6863 #endif
6864         {
6865                 .procname       = "enhanced_dad",
6866                 .data           = &ipv6_devconf.enhanced_dad,
6867                 .maxlen         = sizeof(int),
6868                 .mode           = 0644,
6869                 .proc_handler   = proc_dointvec,
6870         },
6871         {
6872                 .procname               = "addr_gen_mode",
6873                 .data                   = &ipv6_devconf.addr_gen_mode,
6874                 .maxlen                 = sizeof(int),
6875                 .mode                   = 0644,
6876                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6877         },
6878         {
6879                 .procname       = "disable_policy",
6880                 .data           = &ipv6_devconf.disable_policy,
6881                 .maxlen         = sizeof(int),
6882                 .mode           = 0644,
6883                 .proc_handler   = addrconf_sysctl_disable_policy,
6884         },
6885         {
6886                 .procname       = "ndisc_tclass",
6887                 .data           = &ipv6_devconf.ndisc_tclass,
6888                 .maxlen         = sizeof(int),
6889                 .mode           = 0644,
6890                 .proc_handler   = proc_dointvec_minmax,
6891                 .extra1         = (void *)SYSCTL_ZERO,
6892                 .extra2         = (void *)&two_five_five,
6893         },
6894         {
6895                 .procname       = "rpl_seg_enabled",
6896                 .data           = &ipv6_devconf.rpl_seg_enabled,
6897                 .maxlen         = sizeof(int),
6898                 .mode           = 0644,
6899                 .proc_handler   = proc_dointvec,
6900         },
6901         {
6902                 /* sentinel */
6903         }
6904 };
6905
6906 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6907                 struct inet6_dev *idev, struct ipv6_devconf *p)
6908 {
6909         int i, ifindex;
6910         struct ctl_table *table;
6911         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6912
6913         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6914         if (!table)
6915                 goto out;
6916
6917         for (i = 0; table[i].data; i++) {
6918                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6919                 /* If one of these is already set, then it is not safe to
6920                  * overwrite either of them: this makes proc_dointvec_minmax
6921                  * usable.
6922                  */
6923                 if (!table[i].extra1 && !table[i].extra2) {
6924                         table[i].extra1 = idev; /* embedded; no ref */
6925                         table[i].extra2 = net;
6926                 }
6927         }
6928
6929         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6930
6931         p->sysctl_header = register_net_sysctl(net, path, table);
6932         if (!p->sysctl_header)
6933                 goto free;
6934
6935         if (!strcmp(dev_name, "all"))
6936                 ifindex = NETCONFA_IFINDEX_ALL;
6937         else if (!strcmp(dev_name, "default"))
6938                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6939         else
6940                 ifindex = idev->dev->ifindex;
6941         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6942                                      ifindex, p);
6943         return 0;
6944
6945 free:
6946         kfree(table);
6947 out:
6948         return -ENOBUFS;
6949 }
6950
6951 static void __addrconf_sysctl_unregister(struct net *net,
6952                                          struct ipv6_devconf *p, int ifindex)
6953 {
6954         struct ctl_table *table;
6955
6956         if (!p->sysctl_header)
6957                 return;
6958
6959         table = p->sysctl_header->ctl_table_arg;
6960         unregister_net_sysctl_table(p->sysctl_header);
6961         p->sysctl_header = NULL;
6962         kfree(table);
6963
6964         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6965 }
6966
6967 static int addrconf_sysctl_register(struct inet6_dev *idev)
6968 {
6969         int err;
6970
6971         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6972                 return -EINVAL;
6973
6974         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6975                                     &ndisc_ifinfo_sysctl_change);
6976         if (err)
6977                 return err;
6978         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6979                                          idev, &idev->cnf);
6980         if (err)
6981                 neigh_sysctl_unregister(idev->nd_parms);
6982
6983         return err;
6984 }
6985
6986 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6987 {
6988         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6989                                      idev->dev->ifindex);
6990         neigh_sysctl_unregister(idev->nd_parms);
6991 }
6992
6993
6994 #endif
6995
6996 static int __net_init addrconf_init_net(struct net *net)
6997 {
6998         int err = -ENOMEM;
6999         struct ipv6_devconf *all, *dflt;
7000
7001         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7002         if (!all)
7003                 goto err_alloc_all;
7004
7005         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7006         if (!dflt)
7007                 goto err_alloc_dflt;
7008
7009         if (IS_ENABLED(CONFIG_SYSCTL) &&
7010             !net_eq(net, &init_net)) {
7011                 switch (sysctl_devconf_inherit_init_net) {
7012                 case 1:  /* copy from init_net */
7013                         memcpy(all, init_net.ipv6.devconf_all,
7014                                sizeof(ipv6_devconf));
7015                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7016                                sizeof(ipv6_devconf_dflt));
7017                         break;
7018                 case 3: /* copy from the current netns */
7019                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7020                                sizeof(ipv6_devconf));
7021                         memcpy(dflt,
7022                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7023                                sizeof(ipv6_devconf_dflt));
7024                         break;
7025                 case 0:
7026                 case 2:
7027                         /* use compiled values */
7028                         break;
7029                 }
7030         }
7031
7032         /* these will be inherited by all namespaces */
7033         dflt->autoconf = ipv6_defaults.autoconf;
7034         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7035
7036         dflt->stable_secret.initialized = false;
7037         all->stable_secret.initialized = false;
7038
7039         net->ipv6.devconf_all = all;
7040         net->ipv6.devconf_dflt = dflt;
7041
7042 #ifdef CONFIG_SYSCTL
7043         err = __addrconf_sysctl_register(net, "all", NULL, all);
7044         if (err < 0)
7045                 goto err_reg_all;
7046
7047         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7048         if (err < 0)
7049                 goto err_reg_dflt;
7050 #endif
7051         return 0;
7052
7053 #ifdef CONFIG_SYSCTL
7054 err_reg_dflt:
7055         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7056 err_reg_all:
7057         kfree(dflt);
7058 #endif
7059 err_alloc_dflt:
7060         kfree(all);
7061 err_alloc_all:
7062         return err;
7063 }
7064
7065 static void __net_exit addrconf_exit_net(struct net *net)
7066 {
7067 #ifdef CONFIG_SYSCTL
7068         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7069                                      NETCONFA_IFINDEX_DEFAULT);
7070         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7071                                      NETCONFA_IFINDEX_ALL);
7072 #endif
7073         kfree(net->ipv6.devconf_dflt);
7074         kfree(net->ipv6.devconf_all);
7075 }
7076
7077 static struct pernet_operations addrconf_ops = {
7078         .init = addrconf_init_net,
7079         .exit = addrconf_exit_net,
7080 };
7081
7082 static struct rtnl_af_ops inet6_ops __read_mostly = {
7083         .family           = AF_INET6,
7084         .fill_link_af     = inet6_fill_link_af,
7085         .get_link_af_size = inet6_get_link_af_size,
7086         .validate_link_af = inet6_validate_link_af,
7087         .set_link_af      = inet6_set_link_af,
7088 };
7089
7090 /*
7091  *      Init / cleanup code
7092  */
7093
7094 int __init addrconf_init(void)
7095 {
7096         struct inet6_dev *idev;
7097         int i, err;
7098
7099         err = ipv6_addr_label_init();
7100         if (err < 0) {
7101                 pr_crit("%s: cannot initialize default policy table: %d\n",
7102                         __func__, err);
7103                 goto out;
7104         }
7105
7106         err = register_pernet_subsys(&addrconf_ops);
7107         if (err < 0)
7108                 goto out_addrlabel;
7109
7110         addrconf_wq = create_workqueue("ipv6_addrconf");
7111         if (!addrconf_wq) {
7112                 err = -ENOMEM;
7113                 goto out_nowq;
7114         }
7115
7116         /* The addrconf netdev notifier requires that loopback_dev
7117          * has it's ipv6 private information allocated and setup
7118          * before it can bring up and give link-local addresses
7119          * to other devices which are up.
7120          *
7121          * Unfortunately, loopback_dev is not necessarily the first
7122          * entry in the global dev_base list of net devices.  In fact,
7123          * it is likely to be the very last entry on that list.
7124          * So this causes the notifier registry below to try and
7125          * give link-local addresses to all devices besides loopback_dev
7126          * first, then loopback_dev, which cases all the non-loopback_dev
7127          * devices to fail to get a link-local address.
7128          *
7129          * So, as a temporary fix, allocate the ipv6 structure for
7130          * loopback_dev first by hand.
7131          * Longer term, all of the dependencies ipv6 has upon the loopback
7132          * device and it being up should be removed.
7133          */
7134         rtnl_lock();
7135         idev = ipv6_add_dev(init_net.loopback_dev);
7136         rtnl_unlock();
7137         if (IS_ERR(idev)) {
7138                 err = PTR_ERR(idev);
7139                 goto errlo;
7140         }
7141
7142         ip6_route_init_special_entries();
7143
7144         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7145                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7146
7147         register_netdevice_notifier(&ipv6_dev_notf);
7148
7149         addrconf_verify();
7150
7151         rtnl_af_register(&inet6_ops);
7152
7153         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7154                                    NULL, inet6_dump_ifinfo, 0);
7155         if (err < 0)
7156                 goto errout;
7157
7158         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7159                                    inet6_rtm_newaddr, NULL, 0);
7160         if (err < 0)
7161                 goto errout;
7162         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7163                                    inet6_rtm_deladdr, NULL, 0);
7164         if (err < 0)
7165                 goto errout;
7166         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7167                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7168                                    RTNL_FLAG_DOIT_UNLOCKED);
7169         if (err < 0)
7170                 goto errout;
7171         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7172                                    NULL, inet6_dump_ifmcaddr, 0);
7173         if (err < 0)
7174                 goto errout;
7175         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7176                                    NULL, inet6_dump_ifacaddr, 0);
7177         if (err < 0)
7178                 goto errout;
7179         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7180                                    inet6_netconf_get_devconf,
7181                                    inet6_netconf_dump_devconf,
7182                                    RTNL_FLAG_DOIT_UNLOCKED);
7183         if (err < 0)
7184                 goto errout;
7185         err = ipv6_addr_label_rtnl_register();
7186         if (err < 0)
7187                 goto errout;
7188
7189         return 0;
7190 errout:
7191         rtnl_unregister_all(PF_INET6);
7192         rtnl_af_unregister(&inet6_ops);
7193         unregister_netdevice_notifier(&ipv6_dev_notf);
7194 errlo:
7195         destroy_workqueue(addrconf_wq);
7196 out_nowq:
7197         unregister_pernet_subsys(&addrconf_ops);
7198 out_addrlabel:
7199         ipv6_addr_label_cleanup();
7200 out:
7201         return err;
7202 }
7203
7204 void addrconf_cleanup(void)
7205 {
7206         struct net_device *dev;
7207         int i;
7208
7209         unregister_netdevice_notifier(&ipv6_dev_notf);
7210         unregister_pernet_subsys(&addrconf_ops);
7211         ipv6_addr_label_cleanup();
7212
7213         rtnl_af_unregister(&inet6_ops);
7214
7215         rtnl_lock();
7216
7217         /* clean dev list */
7218         for_each_netdev(&init_net, dev) {
7219                 if (__in6_dev_get(dev) == NULL)
7220                         continue;
7221                 addrconf_ifdown(dev, true);
7222         }
7223         addrconf_ifdown(init_net.loopback_dev, true);
7224
7225         /*
7226          *      Check hash table.
7227          */
7228         spin_lock_bh(&addrconf_hash_lock);
7229         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7230                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7231         spin_unlock_bh(&addrconf_hash_lock);
7232         cancel_delayed_work(&addr_chk_work);
7233         rtnl_unlock();
7234
7235         destroy_workqueue(addrconf_wq);
7236 }