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             <[email protected]>
   8  *      Alexey Kuznetsov        <[email protected]>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <[email protected]>
  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/ip_tunnels.h>
  67 #include <net/net_namespace.h>
  68 #include <net/sock.h>
  69 #include <net/snmp.h>
  70
  71 #include <net/6lowpan.h>
  72 #include <net/firewire.h>
  73 #include <net/ipv6.h>
  74 #include <net/protocol.h>
  75 #include <net/ndisc.h>
  76 #include <net/ip6_route.h>
  77 #include <net/addrconf.h>
  78 #include <net/tcp.h>
  79 #include <net/ip.h>
  80 #include <net/netlink.h>
  81 #include <net/pkt_sched.h>
  82 #include <net/l3mdev.h>
  83 #include <linux/if_tunnel.h>
  84 #include <linux/rtnetlink.h>
  85 #include <linux/netconf.h>
  86 #include <linux/random.h>
  87 #include <linux/uaccess.h>
  88 #include <asm/unaligned.h>
  89
  90 #include <linux/proc_fs.h>
  91 #include <linux/seq_file.h>
  92 #include <linux/export.h>
  93 #include <linux/ioam6.h>
  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 = get_random_u32_inclusive(900000, 1100000) * (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 = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = get_random_u32_inclusive(900000, 1100000) * (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 static void addrconf_verify(struct net *net);
 150 static void addrconf_verify_rtnl(struct net *net);
 151
 152 static struct workqueue_struct *addrconf_wq;
 153
 154 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 155 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 156
 157 static void addrconf_type_change(struct net_device *dev,
 158                                  unsigned long event);
 159 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 160
 161 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 162                                                   int plen,
 163                                                   const struct net_device *dev,
 164                                                   u32 flags, u32 noflags,
 165                                                   bool no_gw);
 166
 167 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 168 static void addrconf_dad_work(struct work_struct *w);
 169 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 170                                    bool send_na);
 171 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 172 static void addrconf_rs_timer(struct timer_list *t);
 173 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 174 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 175
 176 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 177                                 struct prefix_info *pinfo);
 178
 179 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 180         .forwarding             = 0,
 181         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 182         .mtu6                   = IPV6_MIN_MTU,
 183         .accept_ra              = 1,
 184         .accept_redirects       = 1,
 185         .autoconf               = 1,
 186         .force_mld_version      = 0,
 187         .mldv1_unsolicited_report_interval = 10 * HZ,
 188         .mldv2_unsolicited_report_interval = HZ,
 189         .dad_transmits          = 1,
 190         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 191         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 192         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 193         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 194         .use_tempaddr           = 0,
 195         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 196         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 197         .regen_min_advance      = REGEN_MIN_ADVANCE,
 198         .regen_max_retry        = REGEN_MAX_RETRY,
 199         .max_desync_factor      = MAX_DESYNC_FACTOR,
 200         .max_addresses          = IPV6_MAX_ADDRESSES,
 201         .accept_ra_defrtr       = 1,
 202         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 203         .accept_ra_from_local   = 0,
 204         .accept_ra_min_hop_limit= 1,
 205         .accept_ra_min_lft      = 0,
 206         .accept_ra_pinfo        = 1,
 207 #ifdef CONFIG_IPV6_ROUTER_PREF
 208         .accept_ra_rtr_pref     = 1,
 209         .rtr_probe_interval     = 60 * HZ,
 210 #ifdef CONFIG_IPV6_ROUTE_INFO
 211         .accept_ra_rt_info_min_plen = 0,
 212         .accept_ra_rt_info_max_plen = 0,
 213 #endif
 214 #endif
 215         .proxy_ndp              = 0,
 216         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 217         .disable_ipv6           = 0,
 218         .accept_dad             = 0,
 219         .suppress_frag_ndisc    = 1,
 220         .accept_ra_mtu          = 1,
 221         .stable_secret          = {
 222                 .initialized = false,
 223         },
 224         .use_oif_addrs_only     = 0,
 225         .ignore_routes_with_linkdown = 0,
 226         .keep_addr_on_down      = 0,
 227         .seg6_enabled           = 0,
 228 #ifdef CONFIG_IPV6_SEG6_HMAC
 229         .seg6_require_hmac      = 0,
 230 #endif
 231         .enhanced_dad           = 1,
 232         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 233         .disable_policy         = 0,
 234         .rpl_seg_enabled        = 0,
 235         .ioam6_enabled          = 0,
 236         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 237         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 238         .ndisc_evict_nocarrier  = 1,
 239         .ra_honor_pio_life      = 0,
 240         .ra_honor_pio_pflag     = 0,
 241 };
 242
 243 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 244         .forwarding             = 0,
 245         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 246         .mtu6                   = IPV6_MIN_MTU,
 247         .accept_ra              = 1,
 248         .accept_redirects       = 1,
 249         .autoconf               = 1,
 250         .force_mld_version      = 0,
 251         .mldv1_unsolicited_report_interval = 10 * HZ,
 252         .mldv2_unsolicited_report_interval = HZ,
 253         .dad_transmits          = 1,
 254         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 255         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 256         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 257         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 258         .use_tempaddr           = 0,
 259         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 260         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 261         .regen_min_advance      = REGEN_MIN_ADVANCE,
 262         .regen_max_retry        = REGEN_MAX_RETRY,
 263         .max_desync_factor      = MAX_DESYNC_FACTOR,
 264         .max_addresses          = IPV6_MAX_ADDRESSES,
 265         .accept_ra_defrtr       = 1,
 266         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 267         .accept_ra_from_local   = 0,
 268         .accept_ra_min_hop_limit= 1,
 269         .accept_ra_min_lft      = 0,
 270         .accept_ra_pinfo        = 1,
 271 #ifdef CONFIG_IPV6_ROUTER_PREF
 272         .accept_ra_rtr_pref     = 1,
 273         .rtr_probe_interval     = 60 * HZ,
 274 #ifdef CONFIG_IPV6_ROUTE_INFO
 275         .accept_ra_rt_info_min_plen = 0,
 276         .accept_ra_rt_info_max_plen = 0,
 277 #endif
 278 #endif
 279         .proxy_ndp              = 0,
 280         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 281         .disable_ipv6           = 0,
 282         .accept_dad             = 1,
 283         .suppress_frag_ndisc    = 1,
 284         .accept_ra_mtu          = 1,
 285         .stable_secret          = {
 286                 .initialized = false,
 287         },
 288         .use_oif_addrs_only     = 0,
 289         .ignore_routes_with_linkdown = 0,
 290         .keep_addr_on_down      = 0,
 291         .seg6_enabled           = 0,
 292 #ifdef CONFIG_IPV6_SEG6_HMAC
 293         .seg6_require_hmac      = 0,
 294 #endif
 295         .enhanced_dad           = 1,
 296         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 297         .disable_policy         = 0,
 298         .rpl_seg_enabled        = 0,
 299         .ioam6_enabled          = 0,
 300         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 301         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 302         .ndisc_evict_nocarrier  = 1,
 303         .ra_honor_pio_life      = 0,
 304         .ra_honor_pio_pflag     = 0,
 305 };
 306
 307 /* Check if link is ready: is it up and is a valid qdisc available */
 308 static inline bool addrconf_link_ready(const struct net_device *dev)
 309 {
 310         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 311 }
 312
 313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 314 {
 315         if (del_timer(&idev->rs_timer))
 316                 __in6_dev_put(idev);
 317 }
 318
 319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 320 {
 321         if (cancel_delayed_work(&ifp->dad_work))
 322                 __in6_ifa_put(ifp);
 323 }
 324
 325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 326                                   unsigned long when)
 327 {
 328         if (!mod_timer(&idev->rs_timer, jiffies + when))
 329                 in6_dev_hold(idev);
 330 }
 331
 332 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 333                                    unsigned long delay)
 334 {
 335         in6_ifa_hold(ifp);
 336         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 337                 in6_ifa_put(ifp);
 338 }
 339
 340 static int snmp6_alloc_dev(struct inet6_dev *idev)
 341 {
 342         int i;
 343
 344         idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
 345         if (!idev->stats.ipv6)
 346                 goto err_ip;
 347
 348         for_each_possible_cpu(i) {
 349                 struct ipstats_mib *addrconf_stats;
 350                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 351                 u64_stats_init(&addrconf_stats->syncp);
 352         }
 353
 354
 355         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 356                                         GFP_KERNEL);
 357         if (!idev->stats.icmpv6dev)
 358                 goto err_icmp;
 359         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 360                                            GFP_KERNEL_ACCOUNT);
 361         if (!idev->stats.icmpv6msgdev)
 362                 goto err_icmpmsg;
 363
 364         return 0;
 365
 366 err_icmpmsg:
 367         kfree(idev->stats.icmpv6dev);
 368 err_icmp:
 369         free_percpu(idev->stats.ipv6);
 370 err_ip:
 371         return -ENOMEM;
 372 }
 373
 374 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 375 {
 376         struct inet6_dev *ndev;
 377         int err = -ENOMEM;
 378
 379         ASSERT_RTNL();
 380
 381         if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
 382                 return ERR_PTR(-EINVAL);
 383
 384         ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
 385         if (!ndev)
 386                 return ERR_PTR(err);
 387
 388         rwlock_init(&ndev->lock);
 389         ndev->dev = dev;
 390         INIT_LIST_HEAD(&ndev->addr_list);
 391         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 392         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 393
 394         if (ndev->cnf.stable_secret.initialized)
 395                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 396
 397         ndev->cnf.mtu6 = dev->mtu;
 398         ndev->ra_mtu = 0;
 399         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 400         if (!ndev->nd_parms) {
 401                 kfree(ndev);
 402                 return ERR_PTR(err);
 403         }
 404         if (ndev->cnf.forwarding)
 405                 dev_disable_lro(dev);
 406         /* We refer to the device */
 407         netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
 408
 409         if (snmp6_alloc_dev(ndev) < 0) {
 410                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 411                            __func__);
 412                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 413                 netdev_put(dev, &ndev->dev_tracker);
 414                 kfree(ndev);
 415                 return ERR_PTR(err);
 416         }
 417
 418         if (dev != blackhole_netdev) {
 419                 if (snmp6_register_dev(ndev) < 0) {
 420                         netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 421                                    __func__, dev->name);
 422                         goto err_release;
 423                 }
 424         }
 425         /* One reference from device. */
 426         refcount_set(&ndev->refcnt, 1);
 427
 428         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 429                 ndev->cnf.accept_dad = -1;
 430
 431 #if IS_ENABLED(CONFIG_IPV6_SIT)
 432         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 433                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 434                 ndev->cnf.rtr_solicits = 0;
 435         }
 436 #endif
 437
 438         INIT_LIST_HEAD(&ndev->tempaddr_list);
 439         ndev->desync_factor = U32_MAX;
 440         if ((dev->flags&IFF_LOOPBACK) ||
 441             dev->type == ARPHRD_TUNNEL ||
 442             dev->type == ARPHRD_TUNNEL6 ||
 443             dev->type == ARPHRD_SIT ||
 444             dev->type == ARPHRD_NONE) {
 445                 ndev->cnf.use_tempaddr = -1;
 446         }
 447
 448         ndev->token = in6addr_any;
 449
 450         if (netif_running(dev) && addrconf_link_ready(dev))
 451                 ndev->if_flags |= IF_READY;
 452
 453         ipv6_mc_init_dev(ndev);
 454         ndev->tstamp = jiffies;
 455         if (dev != blackhole_netdev) {
 456                 err = addrconf_sysctl_register(ndev);
 457                 if (err) {
 458                         ipv6_mc_destroy_dev(ndev);
 459                         snmp6_unregister_dev(ndev);
 460                         goto err_release;
 461                 }
 462         }
 463         /* protected by rtnl_lock */
 464         rcu_assign_pointer(dev->ip6_ptr, ndev);
 465
 466         if (dev != blackhole_netdev) {
 467                 /* Join interface-local all-node multicast group */
 468                 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 469
 470                 /* Join all-node multicast group */
 471                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 472
 473                 /* Join all-router multicast group if forwarding is set */
 474                 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 476         }
 477         return ndev;
 478
 479 err_release:
 480         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 481         ndev->dead = 1;
 482         in6_dev_finish_destroy(ndev);
 483         return ERR_PTR(err);
 484 }
 485
 486 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 487 {
 488         struct inet6_dev *idev;
 489
 490         ASSERT_RTNL();
 491
 492         idev = __in6_dev_get(dev);
 493         if (!idev) {
 494                 idev = ipv6_add_dev(dev);
 495                 if (IS_ERR(idev))
 496                         return idev;
 497         }
 498
 499         if (dev->flags&IFF_UP)
 500                 ipv6_mc_up(idev);
 501         return idev;
 502 }
 503
 504 static int inet6_netconf_msgsize_devconf(int type)
 505 {
 506         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 507                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 508         bool all = false;
 509
 510         if (type == NETCONFA_ALL)
 511                 all = true;
 512
 513         if (all || type == NETCONFA_FORWARDING)
 514                 size += nla_total_size(4);
 515 #ifdef CONFIG_IPV6_MROUTE
 516         if (all || type == NETCONFA_MC_FORWARDING)
 517                 size += nla_total_size(4);
 518 #endif
 519         if (all || type == NETCONFA_PROXY_NEIGH)
 520                 size += nla_total_size(4);
 521
 522         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 523                 size += nla_total_size(4);
 524
 525         return size;
 526 }
 527
 528 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 529                                       struct ipv6_devconf *devconf, u32 portid,
 530                                       u32 seq, int event, unsigned int flags,
 531                                       int type)
 532 {
 533         struct nlmsghdr  *nlh;
 534         struct netconfmsg *ncm;
 535         bool all = false;
 536
 537         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 538                         flags);
 539         if (!nlh)
 540                 return -EMSGSIZE;
 541
 542         if (type == NETCONFA_ALL)
 543                 all = true;
 544
 545         ncm = nlmsg_data(nlh);
 546         ncm->ncm_family = AF_INET6;
 547
 548         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 549                 goto nla_put_failure;
 550
 551         if (!devconf)
 552                 goto out;
 553
 554         if ((all || type == NETCONFA_FORWARDING) &&
 555             nla_put_s32(skb, NETCONFA_FORWARDING,
 556                         READ_ONCE(devconf->forwarding)) < 0)
 557                 goto nla_put_failure;
 558 #ifdef CONFIG_IPV6_MROUTE
 559         if ((all || type == NETCONFA_MC_FORWARDING) &&
 560             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 561                         atomic_read(&devconf->mc_forwarding)) < 0)
 562                 goto nla_put_failure;
 563 #endif
 564         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 565             nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
 566                         READ_ONCE(devconf->proxy_ndp)) < 0)
 567                 goto nla_put_failure;
 568
 569         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 570             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 571                         READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0)
 572                 goto nla_put_failure;
 573
 574 out:
 575         nlmsg_end(skb, nlh);
 576         return 0;
 577
 578 nla_put_failure:
 579         nlmsg_cancel(skb, nlh);
 580         return -EMSGSIZE;
 581 }
 582
 583 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 584                                   int ifindex, struct ipv6_devconf *devconf)
 585 {
 586         struct sk_buff *skb;
 587         int err = -ENOBUFS;
 588
 589         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 590         if (!skb)
 591                 goto errout;
 592
 593         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 594                                          event, 0, type);
 595         if (err < 0) {
 596                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 597                 WARN_ON(err == -EMSGSIZE);
 598                 kfree_skb(skb);
 599                 goto errout;
 600         }
 601         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 602         return;
 603 errout:
 604         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 605 }
 606
 607 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 608         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 609         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 610         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 611         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 612 };
 613
 614 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 615                                        const struct nlmsghdr *nlh,
 616                                        struct nlattr **tb,
 617                                        struct netlink_ext_ack *extack)
 618 {
 619         int i, err;
 620
 621         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 622                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 623                 return -EINVAL;
 624         }
 625
 626         if (!netlink_strict_get_check(skb))
 627                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 628                                               tb, NETCONFA_MAX,
 629                                               devconf_ipv6_policy, extack);
 630
 631         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 632                                             tb, NETCONFA_MAX,
 633                                             devconf_ipv6_policy, extack);
 634         if (err)
 635                 return err;
 636
 637         for (i = 0; i <= NETCONFA_MAX; i++) {
 638                 if (!tb[i])
 639                         continue;
 640
 641                 switch (i) {
 642                 case NETCONFA_IFINDEX:
 643                         break;
 644                 default:
 645                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 646                         return -EINVAL;
 647                 }
 648         }
 649
 650         return 0;
 651 }
 652
 653 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 654                                      struct nlmsghdr *nlh,
 655                                      struct netlink_ext_ack *extack)
 656 {
 657         struct net *net = sock_net(in_skb->sk);
 658         struct nlattr *tb[NETCONFA_MAX+1];
 659         struct inet6_dev *in6_dev = NULL;
 660         struct net_device *dev = NULL;
 661         struct sk_buff *skb;
 662         struct ipv6_devconf *devconf;
 663         int ifindex;
 664         int err;
 665
 666         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 667         if (err < 0)
 668                 return err;
 669
 670         if (!tb[NETCONFA_IFINDEX])
 671                 return -EINVAL;
 672
 673         err = -EINVAL;
 674         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 675         switch (ifindex) {
 676         case NETCONFA_IFINDEX_ALL:
 677                 devconf = net->ipv6.devconf_all;
 678                 break;
 679         case NETCONFA_IFINDEX_DEFAULT:
 680                 devconf = net->ipv6.devconf_dflt;
 681                 break;
 682         default:
 683                 dev = dev_get_by_index(net, ifindex);
 684                 if (!dev)
 685                         return -EINVAL;
 686                 in6_dev = in6_dev_get(dev);
 687                 if (!in6_dev)
 688                         goto errout;
 689                 devconf = &in6_dev->cnf;
 690                 break;
 691         }
 692
 693         err = -ENOBUFS;
 694         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 695         if (!skb)
 696                 goto errout;
 697
 698         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 699                                          NETLINK_CB(in_skb).portid,
 700                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 701                                          NETCONFA_ALL);
 702         if (err < 0) {
 703                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 704                 WARN_ON(err == -EMSGSIZE);
 705                 kfree_skb(skb);
 706                 goto errout;
 707         }
 708         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 709 errout:
 710         if (in6_dev)
 711                 in6_dev_put(in6_dev);
 712         dev_put(dev);
 713         return err;
 714 }
 715
 716 /* Combine dev_addr_genid and dev_base_seq to detect changes.
 717  */
 718 static u32 inet6_base_seq(const struct net *net)
 719 {
 720         u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
 721                   READ_ONCE(net->dev_base_seq);
 722
 723         /* Must not return 0 (see nl_dump_check_consistent()).
 724          * Chose a value far away from 0.
 725          */
 726         if (!res)
 727                 res = 0x80000000;
 728         return res;
 729 }
 730
 731 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 732                                       struct netlink_callback *cb)
 733 {
 734         const struct nlmsghdr *nlh = cb->nlh;
 735         struct net *net = sock_net(skb->sk);
 736         struct {
 737                 unsigned long ifindex;
 738                 unsigned int all_default;
 739         } *ctx = (void *)cb->ctx;
 740         struct net_device *dev;
 741         struct inet6_dev *idev;
 742         int err = 0;
 743
 744         if (cb->strict_check) {
 745                 struct netlink_ext_ack *extack = cb->extack;
 746                 struct netconfmsg *ncm;
 747
 748                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 749                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 750                         return -EINVAL;
 751                 }
 752
 753                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 754                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 755                         return -EINVAL;
 756                 }
 757         }
 758
 759         rcu_read_lock();
 760         for_each_netdev_dump(net, dev, ctx->ifindex) {
 761                 idev = __in6_dev_get(dev);
 762                 if (!idev)
 763                         continue;
 764                 err = inet6_netconf_fill_devconf(skb, dev->ifindex,
 765                                                  &idev->cnf,
 766                                                  NETLINK_CB(cb->skb).portid,
 767                                                  nlh->nlmsg_seq,
 768                                                  RTM_NEWNETCONF,
 769                                                  NLM_F_MULTI,
 770                                                  NETCONFA_ALL);
 771                 if (err < 0)
 772                         goto done;
 773         }
 774         if (ctx->all_default == 0) {
 775                 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 776                                                  net->ipv6.devconf_all,
 777                                                  NETLINK_CB(cb->skb).portid,
 778                                                  nlh->nlmsg_seq,
 779                                                  RTM_NEWNETCONF, NLM_F_MULTI,
 780                                                  NETCONFA_ALL);
 781                 if (err < 0)
 782                         goto done;
 783                 ctx->all_default++;
 784         }
 785         if (ctx->all_default == 1) {
 786                 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 787                                                  net->ipv6.devconf_dflt,
 788                                                  NETLINK_CB(cb->skb).portid,
 789                                                  nlh->nlmsg_seq,
 790                                                  RTM_NEWNETCONF, NLM_F_MULTI,
 791                                                  NETCONFA_ALL);
 792                 if (err < 0)
 793                         goto done;
 794                 ctx->all_default++;
 795         }
 796 done:
 797         rcu_read_unlock();
 798         return err;
 799 }
 800
 801 #ifdef CONFIG_SYSCTL
 802 static void dev_forward_change(struct inet6_dev *idev)
 803 {
 804         struct net_device *dev;
 805         struct inet6_ifaddr *ifa;
 806         LIST_HEAD(tmp_addr_list);
 807
 808         if (!idev)
 809                 return;
 810         dev = idev->dev;
 811         if (idev->cnf.forwarding)
 812                 dev_disable_lro(dev);
 813         if (dev->flags & IFF_MULTICAST) {
 814                 if (idev->cnf.forwarding) {
 815                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 816                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 817                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 818                 } else {
 819                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 820                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 821                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 822                 }
 823         }
 824
 825         read_lock_bh(&idev->lock);
 826         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 827                 if (ifa->flags&IFA_F_TENTATIVE)
 828                         continue;
 829                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
 830         }
 831         read_unlock_bh(&idev->lock);
 832
 833         while (!list_empty(&tmp_addr_list)) {
 834                 ifa = list_first_entry(&tmp_addr_list,
 835                                        struct inet6_ifaddr, if_list_aux);
 836                 list_del(&ifa->if_list_aux);
 837                 if (idev->cnf.forwarding)
 838                         addrconf_join_anycast(ifa);
 839                 else
 840                         addrconf_leave_anycast(ifa);
 841         }
 842
 843         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 844                                      NETCONFA_FORWARDING,
 845                                      dev->ifindex, &idev->cnf);
 846 }
 847
 848
 849 static void addrconf_forward_change(struct net *net, __s32 newf)
 850 {
 851         struct net_device *dev;
 852         struct inet6_dev *idev;
 853
 854         for_each_netdev(net, dev) {
 855                 idev = __in6_dev_get(dev);
 856                 if (idev) {
 857                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 858
 859                         WRITE_ONCE(idev->cnf.forwarding, newf);
 860                         if (changed)
 861                                 dev_forward_change(idev);
 862                 }
 863         }
 864 }
 865
 866 static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf)
 867 {
 868         struct net *net;
 869         int old;
 870
 871         if (!rtnl_trylock())
 872                 return restart_syscall();
 873
 874         net = (struct net *)table->extra2;
 875         old = *p;
 876         WRITE_ONCE(*p, newf);
 877
 878         if (p == &net->ipv6.devconf_dflt->forwarding) {
 879                 if ((!newf) ^ (!old))
 880                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 881                                                      NETCONFA_FORWARDING,
 882                                                      NETCONFA_IFINDEX_DEFAULT,
 883                                                      net->ipv6.devconf_dflt);
 884                 rtnl_unlock();
 885                 return 0;
 886         }
 887
 888         if (p == &net->ipv6.devconf_all->forwarding) {
 889                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 890
 891                 WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf);
 892                 if ((!newf) ^ (!old_dflt))
 893                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 894                                                      NETCONFA_FORWARDING,
 895                                                      NETCONFA_IFINDEX_DEFAULT,
 896                                                      net->ipv6.devconf_dflt);
 897
 898                 addrconf_forward_change(net, newf);
 899                 if ((!newf) ^ (!old))
 900                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 901                                                      NETCONFA_FORWARDING,
 902                                                      NETCONFA_IFINDEX_ALL,
 903                                                      net->ipv6.devconf_all);
 904         } else if ((!newf) ^ (!old))
 905                 dev_forward_change((struct inet6_dev *)table->extra1);
 906         rtnl_unlock();
 907
 908         if (newf)
 909                 rt6_purge_dflt_routers(net);
 910         return 1;
 911 }
 912
 913 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 914 {
 915         struct net_device *dev;
 916         struct inet6_dev *idev;
 917
 918         for_each_netdev(net, dev) {
 919                 idev = __in6_dev_get(dev);
 920                 if (idev) {
 921                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 922
 923                         WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf);
 924                         if (changed)
 925                                 inet6_netconf_notify_devconf(dev_net(dev),
 926                                                              RTM_NEWNETCONF,
 927                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 928                                                              dev->ifindex,
 929                                                              &idev->cnf);
 930                 }
 931         }
 932 }
 933
 934 static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf)
 935 {
 936         struct net *net;
 937         int old;
 938
 939         if (!rtnl_trylock())
 940                 return restart_syscall();
 941
 942         net = (struct net *)table->extra2;
 943         old = *p;
 944         WRITE_ONCE(*p, newf);
 945
 946         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 947                 if ((!newf) ^ (!old))
 948                         inet6_netconf_notify_devconf(net,
 949                                                      RTM_NEWNETCONF,
 950                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 951                                                      NETCONFA_IFINDEX_DEFAULT,
 952                                                      net->ipv6.devconf_dflt);
 953                 rtnl_unlock();
 954                 return 0;
 955         }
 956
 957         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 958                 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf);
 959                 addrconf_linkdown_change(net, newf);
 960                 if ((!newf) ^ (!old))
 961                         inet6_netconf_notify_devconf(net,
 962                                                      RTM_NEWNETCONF,
 963                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 964                                                      NETCONFA_IFINDEX_ALL,
 965                                                      net->ipv6.devconf_all);
 966         }
 967         rtnl_unlock();
 968
 969         return 1;
 970 }
 971
 972 #endif
 973
 974 /* Nobody refers to this ifaddr, destroy it */
 975 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 976 {
 977         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 978
 979 #ifdef NET_REFCNT_DEBUG
 980         pr_debug("%s\n", __func__);
 981 #endif
 982
 983         in6_dev_put(ifp->idev);
 984
 985         if (cancel_delayed_work(&ifp->dad_work))
 986                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 987                           ifp);
 988
 989         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 990                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 991                 return;
 992         }
 993
 994         kfree_rcu(ifp, rcu);
 995 }
 996
 997 static void
 998 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 999 {
1000         struct list_head *p;
1001         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1002
1003         /*
1004          * Each device address list is sorted in order of scope -
1005          * global before linklocal.
1006          */
1007         list_for_each(p, &idev->addr_list) {
1008                 struct inet6_ifaddr *ifa
1009                         = list_entry(p, struct inet6_ifaddr, if_list);
1010                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1011                         break;
1012         }
1013
1014         list_add_tail_rcu(&ifp->if_list, p);
1015 }
1016
1017 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1018 {
1019         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1020
1021         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1022 }
1023
1024 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1025                                struct net_device *dev, unsigned int hash)
1026 {
1027         struct inet6_ifaddr *ifp;
1028
1029         hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1030                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1031                         if (!dev || ifp->idev->dev == dev)
1032                                 return true;
1033                 }
1034         }
1035         return false;
1036 }
1037
1038 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1039 {
1040         struct net *net = dev_net(dev);
1041         unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1042         int err = 0;
1043
1044         spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1045
1046         /* Ignore adding duplicate addresses on an interface */
1047         if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1048                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1049                 err = -EEXIST;
1050         } else {
1051                 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1052         }
1053
1054         spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1055
1056         return err;
1057 }
1058
1059 /* On success it returns ifp with increased reference count */
1060
1061 static struct inet6_ifaddr *
1062 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1063               bool can_block, struct netlink_ext_ack *extack)
1064 {
1065         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1066         int addr_type = ipv6_addr_type(cfg->pfx);
1067         struct net *net = dev_net(idev->dev);
1068         struct inet6_ifaddr *ifa = NULL;
1069         struct fib6_info *f6i = NULL;
1070         int err = 0;
1071
1072         if (addr_type == IPV6_ADDR_ANY) {
1073                 NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1074                 return ERR_PTR(-EADDRNOTAVAIL);
1075         } else if (addr_type & IPV6_ADDR_MULTICAST &&
1076                    !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1077                 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1078                 return ERR_PTR(-EADDRNOTAVAIL);
1079         } else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1080                    !netif_is_l3_master(idev->dev) &&
1081                    addr_type & IPV6_ADDR_LOOPBACK) {
1082                 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1083                 return ERR_PTR(-EADDRNOTAVAIL);
1084         }
1085
1086         if (idev->dead) {
1087                 NL_SET_ERR_MSG_MOD(extack, "device is going away");
1088                 err = -ENODEV;
1089                 goto out;
1090         }
1091
1092         if (idev->cnf.disable_ipv6) {
1093                 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1094                 err = -EACCES;
1095                 goto out;
1096         }
1097
1098         /* validator notifier needs to be blocking;
1099          * do not call in atomic context
1100          */
1101         if (can_block) {
1102                 struct in6_validator_info i6vi = {
1103                         .i6vi_addr = *cfg->pfx,
1104                         .i6vi_dev = idev,
1105                         .extack = extack,
1106                 };
1107
1108                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1109                 err = notifier_to_errno(err);
1110                 if (err < 0)
1111                         goto out;
1112         }
1113
1114         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1115         if (!ifa) {
1116                 err = -ENOBUFS;
1117                 goto out;
1118         }
1119
1120         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1121         if (IS_ERR(f6i)) {
1122                 err = PTR_ERR(f6i);
1123                 f6i = NULL;
1124                 goto out;
1125         }
1126
1127         neigh_parms_data_state_setall(idev->nd_parms);
1128
1129         ifa->addr = *cfg->pfx;
1130         if (cfg->peer_pfx)
1131                 ifa->peer_addr = *cfg->peer_pfx;
1132
1133         spin_lock_init(&ifa->lock);
1134         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1135         INIT_HLIST_NODE(&ifa->addr_lst);
1136         ifa->scope = cfg->scope;
1137         ifa->prefix_len = cfg->plen;
1138         ifa->rt_priority = cfg->rt_priority;
1139         ifa->flags = cfg->ifa_flags;
1140         ifa->ifa_proto = cfg->ifa_proto;
1141         /* No need to add the TENTATIVE flag for addresses with NODAD */
1142         if (!(cfg->ifa_flags & IFA_F_NODAD))
1143                 ifa->flags |= IFA_F_TENTATIVE;
1144         ifa->valid_lft = cfg->valid_lft;
1145         ifa->prefered_lft = cfg->preferred_lft;
1146         ifa->cstamp = ifa->tstamp = jiffies;
1147         ifa->tokenized = false;
1148
1149         ifa->rt = f6i;
1150
1151         ifa->idev = idev;
1152         in6_dev_hold(idev);
1153
1154         /* For caller */
1155         refcount_set(&ifa->refcnt, 1);
1156
1157         rcu_read_lock();
1158
1159         err = ipv6_add_addr_hash(idev->dev, ifa);
1160         if (err < 0) {
1161                 rcu_read_unlock();
1162                 goto out;
1163         }
1164
1165         write_lock_bh(&idev->lock);
1166
1167         /* Add to inet6_dev unicast addr list. */
1168         ipv6_link_dev_addr(idev, ifa);
1169
1170         if (ifa->flags&IFA_F_TEMPORARY) {
1171                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1172                 in6_ifa_hold(ifa);
1173         }
1174
1175         in6_ifa_hold(ifa);
1176         write_unlock_bh(&idev->lock);
1177
1178         rcu_read_unlock();
1179
1180         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1181 out:
1182         if (unlikely(err < 0)) {
1183                 fib6_info_release(f6i);
1184
1185                 if (ifa) {
1186                         if (ifa->idev)
1187                                 in6_dev_put(ifa->idev);
1188                         kfree(ifa);
1189                 }
1190                 ifa = ERR_PTR(err);
1191         }
1192
1193         return ifa;
1194 }
1195
1196 enum cleanup_prefix_rt_t {
1197         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1198         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1199         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1200 };
1201
1202 /*
1203  * Check, whether the prefix for ifp would still need a prefix route
1204  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1205  * constants.
1206  *
1207  * 1) we don't purge prefix if address was not permanent.
1208  *    prefix is managed by its own lifetime.
1209  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1210  * 3) if there are no addresses, delete prefix.
1211  * 4) if there are still other permanent address(es),
1212  *    corresponding prefix is still permanent.
1213  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1214  *    don't purge the prefix, assume user space is managing it.
1215  * 6) otherwise, update prefix lifetime to the
1216  *    longest valid lifetime among the corresponding
1217  *    addresses on the device.
1218  *    Note: subsequent RA will update lifetime.
1219  **/
1220 static enum cleanup_prefix_rt_t
1221 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1222 {
1223         struct inet6_ifaddr *ifa;
1224         struct inet6_dev *idev = ifp->idev;
1225         unsigned long lifetime;
1226         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1227
1228         *expires = jiffies;
1229
1230         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1231                 if (ifa == ifp)
1232                         continue;
1233                 if (ifa->prefix_len != ifp->prefix_len ||
1234                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1235                                        ifp->prefix_len))
1236                         continue;
1237                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1238                         return CLEANUP_PREFIX_RT_NOP;
1239
1240                 action = CLEANUP_PREFIX_RT_EXPIRE;
1241
1242                 spin_lock(&ifa->lock);
1243
1244                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1245                 /*
1246                  * Note: Because this address is
1247                  * not permanent, lifetime <
1248                  * LONG_MAX / HZ here.
1249                  */
1250                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1251                         *expires = ifa->tstamp + lifetime * HZ;
1252                 spin_unlock(&ifa->lock);
1253         }
1254
1255         return action;
1256 }
1257
1258 static void
1259 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1260                      bool del_rt, bool del_peer)
1261 {
1262         struct fib6_table *table;
1263         struct fib6_info *f6i;
1264
1265         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1266                                         ifp->prefix_len,
1267                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1268         if (f6i) {
1269                 if (del_rt)
1270                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1271                 else {
1272                         if (!(f6i->fib6_flags & RTF_EXPIRES)) {
1273                                 table = f6i->fib6_table;
1274                                 spin_lock_bh(&table->tb6_lock);
1275
1276                                 fib6_set_expires(f6i, expires);
1277                                 fib6_add_gc_list(f6i);
1278
1279                                 spin_unlock_bh(&table->tb6_lock);
1280                         }
1281                         fib6_info_release(f6i);
1282                 }
1283         }
1284 }
1285
1286
1287 /* This function wants to get referenced ifp and releases it before return */
1288
1289 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1290 {
1291         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1292         struct net *net = dev_net(ifp->idev->dev);
1293         unsigned long expires;
1294         int state;
1295
1296         ASSERT_RTNL();
1297
1298         spin_lock_bh(&ifp->lock);
1299         state = ifp->state;
1300         ifp->state = INET6_IFADDR_STATE_DEAD;
1301         spin_unlock_bh(&ifp->lock);
1302
1303         if (state == INET6_IFADDR_STATE_DEAD)
1304                 goto out;
1305
1306         spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1307         hlist_del_init_rcu(&ifp->addr_lst);
1308         spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1309
1310         write_lock_bh(&ifp->idev->lock);
1311
1312         if (ifp->flags&IFA_F_TEMPORARY) {
1313                 list_del(&ifp->tmp_list);
1314                 if (ifp->ifpub) {
1315                         in6_ifa_put(ifp->ifpub);
1316                         ifp->ifpub = NULL;
1317                 }
1318                 __in6_ifa_put(ifp);
1319         }
1320
1321         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1322                 action = check_cleanup_prefix_route(ifp, &expires);
1323
1324         list_del_rcu(&ifp->if_list);
1325         __in6_ifa_put(ifp);
1326
1327         write_unlock_bh(&ifp->idev->lock);
1328
1329         addrconf_del_dad_work(ifp);
1330
1331         ipv6_ifa_notify(RTM_DELADDR, ifp);
1332
1333         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1334
1335         if (action != CLEANUP_PREFIX_RT_NOP) {
1336                 cleanup_prefix_route(ifp, expires,
1337                         action == CLEANUP_PREFIX_RT_DEL, false);
1338         }
1339
1340         /* clean up prefsrc entries */
1341         rt6_remove_prefsrc(ifp);
1342 out:
1343         in6_ifa_put(ifp);
1344 }
1345
1346 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1347 {
1348         return READ_ONCE(idev->cnf.regen_min_advance) +
1349                 READ_ONCE(idev->cnf.regen_max_retry) *
1350                 READ_ONCE(idev->cnf.dad_transmits) *
1351                 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1352 }
1353
1354 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1355 {
1356         struct inet6_dev *idev = ifp->idev;
1357         unsigned long tmp_tstamp, age;
1358         unsigned long regen_advance;
1359         unsigned long now = jiffies;
1360         u32 if_public_preferred_lft;
1361         s32 cnf_temp_preferred_lft;
1362         struct inet6_ifaddr *ift;
1363         struct ifa6_config cfg;
1364         long max_desync_factor;
1365         struct in6_addr addr;
1366         int ret = 0;
1367
1368         write_lock_bh(&idev->lock);
1369
1370 retry:
1371         in6_dev_hold(idev);
1372         if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1373                 write_unlock_bh(&idev->lock);
1374                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1375                 in6_dev_put(idev);
1376                 ret = -1;
1377                 goto out;
1378         }
1379         spin_lock_bh(&ifp->lock);
1380         if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1381                 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/
1382                 spin_unlock_bh(&ifp->lock);
1383                 write_unlock_bh(&idev->lock);
1384                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1385                         __func__);
1386                 in6_dev_put(idev);
1387                 ret = -1;
1388                 goto out;
1389         }
1390         in6_ifa_hold(ifp);
1391         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1392         ipv6_gen_rnd_iid(&addr);
1393
1394         age = (now - ifp->tstamp) / HZ;
1395
1396         regen_advance = ipv6_get_regen_advance(idev);
1397
1398         /* recalculate max_desync_factor each time and update
1399          * idev->desync_factor if it's larger
1400          */
1401         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1402         max_desync_factor = min_t(long,
1403                                   READ_ONCE(idev->cnf.max_desync_factor),
1404                                   cnf_temp_preferred_lft - regen_advance);
1405
1406         if (unlikely(idev->desync_factor > max_desync_factor)) {
1407                 if (max_desync_factor > 0) {
1408                         get_random_bytes(&idev->desync_factor,
1409                                          sizeof(idev->desync_factor));
1410                         idev->desync_factor %= max_desync_factor;
1411                 } else {
1412                         idev->desync_factor = 0;
1413                 }
1414         }
1415
1416         if_public_preferred_lft = ifp->prefered_lft;
1417
1418         memset(&cfg, 0, sizeof(cfg));
1419         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1420                               READ_ONCE(idev->cnf.temp_valid_lft) + age);
1421         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1422         cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1423         cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1424
1425         cfg.plen = ifp->prefix_len;
1426         tmp_tstamp = ifp->tstamp;
1427         spin_unlock_bh(&ifp->lock);
1428
1429         write_unlock_bh(&idev->lock);
1430
1431         /* From RFC 4941:
1432          *
1433          *     A temporary address is created only if this calculated Preferred
1434          *     Lifetime is greater than REGEN_ADVANCE time units.  In
1435          *     particular, an implementation must not create a temporary address
1436          *     with a zero Preferred Lifetime.
1437          *
1438          *     ...
1439          *
1440          *     When creating a temporary address, the lifetime values MUST be
1441          *     derived from the corresponding prefix as follows:
1442          *
1443          *     ...
1444          *
1445          *     *  Its Preferred Lifetime is the lower of the Preferred Lifetime
1446          *        of the public address or TEMP_PREFERRED_LIFETIME -
1447          *        DESYNC_FACTOR.
1448          *
1449          * To comply with the RFC's requirements, clamp the preferred lifetime
1450          * to a minimum of regen_advance, unless that would exceed valid_lft or
1451          * ifp->prefered_lft.
1452          *
1453          * Use age calculation as in addrconf_verify to avoid unnecessary
1454          * temporary addresses being generated.
1455          */
1456         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1457         if (cfg.preferred_lft <= regen_advance + age) {
1458                 cfg.preferred_lft = regen_advance + age + 1;
1459                 if (cfg.preferred_lft > cfg.valid_lft ||
1460                     cfg.preferred_lft > if_public_preferred_lft) {
1461                         in6_ifa_put(ifp);
1462                         in6_dev_put(idev);
1463                         ret = -1;
1464                         goto out;
1465                 }
1466         }
1467
1468         cfg.ifa_flags = IFA_F_TEMPORARY;
1469         /* set in addrconf_prefix_rcv() */
1470         if (ifp->flags & IFA_F_OPTIMISTIC)
1471                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1472
1473         cfg.pfx = &addr;
1474         cfg.scope = ipv6_addr_scope(cfg.pfx);
1475
1476         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1477         if (IS_ERR(ift)) {
1478                 in6_ifa_put(ifp);
1479                 in6_dev_put(idev);
1480                 pr_info("%s: retry temporary address regeneration\n", __func__);
1481                 write_lock_bh(&idev->lock);
1482                 goto retry;
1483         }
1484
1485         spin_lock_bh(&ift->lock);
1486         ift->ifpub = ifp;
1487         ift->cstamp = now;
1488         ift->tstamp = tmp_tstamp;
1489         spin_unlock_bh(&ift->lock);
1490
1491         addrconf_dad_start(ift);
1492         in6_ifa_put(ift);
1493         in6_dev_put(idev);
1494 out:
1495         return ret;
1496 }
1497
1498 /*
1499  *      Choose an appropriate source address (RFC3484)
1500  */
1501 enum {
1502         IPV6_SADDR_RULE_INIT = 0,
1503         IPV6_SADDR_RULE_LOCAL,
1504         IPV6_SADDR_RULE_SCOPE,
1505         IPV6_SADDR_RULE_PREFERRED,
1506 #ifdef CONFIG_IPV6_MIP6
1507         IPV6_SADDR_RULE_HOA,
1508 #endif
1509         IPV6_SADDR_RULE_OIF,
1510         IPV6_SADDR_RULE_LABEL,
1511         IPV6_SADDR_RULE_PRIVACY,
1512         IPV6_SADDR_RULE_ORCHID,
1513         IPV6_SADDR_RULE_PREFIX,
1514 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1515         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1516 #endif
1517         IPV6_SADDR_RULE_MAX
1518 };
1519
1520 struct ipv6_saddr_score {
1521         int                     rule;
1522         int                     addr_type;
1523         struct inet6_ifaddr     *ifa;
1524         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1525         int                     scopedist;
1526         int                     matchlen;
1527 };
1528
1529 struct ipv6_saddr_dst {
1530         const struct in6_addr *addr;
1531         int ifindex;
1532         int scope;
1533         int label;
1534         unsigned int prefs;
1535 };
1536
1537 static inline int ipv6_saddr_preferred(int type)
1538 {
1539         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1540                 return 1;
1541         return 0;
1542 }
1543
1544 static bool ipv6_use_optimistic_addr(const struct net *net,
1545                                      const struct inet6_dev *idev)
1546 {
1547 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1548         if (!idev)
1549                 return false;
1550         if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1551             !READ_ONCE(idev->cnf.optimistic_dad))
1552                 return false;
1553         if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1554             !READ_ONCE(idev->cnf.use_optimistic))
1555                 return false;
1556
1557         return true;
1558 #else
1559         return false;
1560 #endif
1561 }
1562
1563 static bool ipv6_allow_optimistic_dad(const struct net *net,
1564                                       const struct inet6_dev *idev)
1565 {
1566 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1567         if (!idev)
1568                 return false;
1569         if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1570             !READ_ONCE(idev->cnf.optimistic_dad))
1571                 return false;
1572
1573         return true;
1574 #else
1575         return false;
1576 #endif
1577 }
1578
1579 static int ipv6_get_saddr_eval(struct net *net,
1580                                struct ipv6_saddr_score *score,
1581                                struct ipv6_saddr_dst *dst,
1582                                int i)
1583 {
1584         int ret;
1585
1586         if (i <= score->rule) {
1587                 switch (i) {
1588                 case IPV6_SADDR_RULE_SCOPE:
1589                         ret = score->scopedist;
1590                         break;
1591                 case IPV6_SADDR_RULE_PREFIX:
1592                         ret = score->matchlen;
1593                         break;
1594                 default:
1595                         ret = !!test_bit(i, score->scorebits);
1596                 }
1597                 goto out;
1598         }
1599
1600         switch (i) {
1601         case IPV6_SADDR_RULE_INIT:
1602                 /* Rule 0: remember if hiscore is not ready yet */
1603                 ret = !!score->ifa;
1604                 break;
1605         case IPV6_SADDR_RULE_LOCAL:
1606                 /* Rule 1: Prefer same address */
1607                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1608                 break;
1609         case IPV6_SADDR_RULE_SCOPE:
1610                 /* Rule 2: Prefer appropriate scope
1611                  *
1612                  *      ret
1613                  *       ^
1614                  *    -1 |  d 15
1615                  *    ---+--+-+---> scope
1616                  *       |
1617                  *       |             d is scope of the destination.
1618                  *  B-d  |  \
1619                  *       |   \      <- smaller scope is better if
1620                  *  B-15 |    \        if scope is enough for destination.
1621                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1622                  * d-C-1 | /
1623                  *       |/         <- greater is better
1624                  *   -C  /             if scope is not enough for destination.
1625                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1626                  *
1627                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1628                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1629                  * Assume B = 0 and we get C > 29.
1630                  */
1631                 ret = __ipv6_addr_src_scope(score->addr_type);
1632                 if (ret >= dst->scope)
1633                         ret = -ret;
1634                 else
1635                         ret -= 128;     /* 30 is enough */
1636                 score->scopedist = ret;
1637                 break;
1638         case IPV6_SADDR_RULE_PREFERRED:
1639             {
1640                 /* Rule 3: Avoid deprecated and optimistic addresses */
1641                 u8 avoid = IFA_F_DEPRECATED;
1642
1643                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1644                         avoid |= IFA_F_OPTIMISTIC;
1645                 ret = ipv6_saddr_preferred(score->addr_type) ||
1646                       !(score->ifa->flags & avoid);
1647                 break;
1648             }
1649 #ifdef CONFIG_IPV6_MIP6
1650         case IPV6_SADDR_RULE_HOA:
1651             {
1652                 /* Rule 4: Prefer home address */
1653                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1654                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1655                 break;
1656             }
1657 #endif
1658         case IPV6_SADDR_RULE_OIF:
1659                 /* Rule 5: Prefer outgoing interface */
1660                 ret = (!dst->ifindex ||
1661                        dst->ifindex == score->ifa->idev->dev->ifindex);
1662                 break;
1663         case IPV6_SADDR_RULE_LABEL:
1664                 /* Rule 6: Prefer matching label */
1665                 ret = ipv6_addr_label(net,
1666                                       &score->ifa->addr, score->addr_type,
1667                                       score->ifa->idev->dev->ifindex) == dst->label;
1668                 break;
1669         case IPV6_SADDR_RULE_PRIVACY:
1670             {
1671                 /* Rule 7: Prefer public address
1672                  * Note: prefer temporary address if use_tempaddr >= 2
1673                  */
1674                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1675                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1676                                 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1677                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1678                 break;
1679             }
1680         case IPV6_SADDR_RULE_ORCHID:
1681                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1682                  *          non-ORCHID vs non-ORCHID
1683                  */
1684                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1685                         ipv6_addr_orchid(dst->addr));
1686                 break;
1687         case IPV6_SADDR_RULE_PREFIX:
1688                 /* Rule 8: Use longest matching prefix */
1689                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1690                 if (ret > score->ifa->prefix_len)
1691                         ret = score->ifa->prefix_len;
1692                 score->matchlen = ret;
1693                 break;
1694 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1695         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1696                 /* Optimistic addresses still have lower precedence than other
1697                  * preferred addresses.
1698                  */
1699                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1700                 break;
1701 #endif
1702         default:
1703                 ret = 0;
1704         }
1705
1706         if (ret)
1707                 __set_bit(i, score->scorebits);
1708         score->rule = i;
1709 out:
1710         return ret;
1711 }
1712
1713 static int __ipv6_dev_get_saddr(struct net *net,
1714                                 struct ipv6_saddr_dst *dst,
1715                                 struct inet6_dev *idev,
1716                                 struct ipv6_saddr_score *scores,
1717                                 int hiscore_idx)
1718 {
1719         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1720
1721         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1722                 int i;
1723
1724                 /*
1725                  * - Tentative Address (RFC2462 section 5.4)
1726                  *  - A tentative address is not considered
1727                  *    "assigned to an interface" in the traditional
1728                  *    sense, unless it is also flagged as optimistic.
1729                  * - Candidate Source Address (section 4)
1730                  *  - In any case, anycast addresses, multicast
1731                  *    addresses, and the unspecified address MUST
1732                  *    NOT be included in a candidate set.
1733                  */
1734                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1735                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1736                         continue;
1737
1738                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1739
1740                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1741                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1742                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1743                                             idev->dev->name);
1744                         continue;
1745                 }
1746
1747                 score->rule = -1;
1748                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1749
1750                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1751                         int minihiscore, miniscore;
1752
1753                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1754                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1755
1756                         if (minihiscore > miniscore) {
1757                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1758                                     score->scopedist > 0) {
1759                                         /*
1760                                          * special case:
1761                                          * each remaining entry
1762                                          * has too small (not enough)
1763                                          * scope, because ifa entries
1764                                          * are sorted by their scope
1765                                          * values.
1766                                          */
1767                                         goto out;
1768                                 }
1769                                 break;
1770                         } else if (minihiscore < miniscore) {
1771                                 swap(hiscore, score);
1772                                 hiscore_idx = 1 - hiscore_idx;
1773
1774                                 /* restore our iterator */
1775                                 score->ifa = hiscore->ifa;
1776
1777                                 break;
1778                         }
1779                 }
1780         }
1781 out:
1782         return hiscore_idx;
1783 }
1784
1785 static int ipv6_get_saddr_master(struct net *net,
1786                                  const struct net_device *dst_dev,
1787                                  const struct net_device *master,
1788                                  struct ipv6_saddr_dst *dst,
1789                                  struct ipv6_saddr_score *scores,
1790                                  int hiscore_idx)
1791 {
1792         struct inet6_dev *idev;
1793
1794         idev = __in6_dev_get(dst_dev);
1795         if (idev)
1796                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1797                                                    scores, hiscore_idx);
1798
1799         idev = __in6_dev_get(master);
1800         if (idev)
1801                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1802                                                    scores, hiscore_idx);
1803
1804         return hiscore_idx;
1805 }
1806
1807 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1808                        const struct in6_addr *daddr, unsigned int prefs,
1809                        struct in6_addr *saddr)
1810 {
1811         struct ipv6_saddr_score scores[2], *hiscore;
1812         struct ipv6_saddr_dst dst;
1813         struct inet6_dev *idev;
1814         struct net_device *dev;
1815         int dst_type;
1816         bool use_oif_addr = false;
1817         int hiscore_idx = 0;
1818         int ret = 0;
1819
1820         dst_type = __ipv6_addr_type(daddr);
1821         dst.addr = daddr;
1822         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1823         dst.scope = __ipv6_addr_src_scope(dst_type);
1824         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1825         dst.prefs = prefs;
1826
1827         scores[hiscore_idx].rule = -1;
1828         scores[hiscore_idx].ifa = NULL;
1829
1830         rcu_read_lock();
1831
1832         /* Candidate Source Address (section 4)
1833          *  - multicast and link-local destination address,
1834          *    the set of candidate source address MUST only
1835          *    include addresses assigned to interfaces
1836          *    belonging to the same link as the outgoing
1837          *    interface.
1838          * (- For site-local destination addresses, the
1839          *    set of candidate source addresses MUST only
1840          *    include addresses assigned to interfaces
1841          *    belonging to the same site as the outgoing
1842          *    interface.)
1843          *  - "It is RECOMMENDED that the candidate source addresses
1844          *    be the set of unicast addresses assigned to the
1845          *    interface that will be used to send to the destination
1846          *    (the 'outgoing' interface)." (RFC 6724)
1847          */
1848         if (dst_dev) {
1849                 idev = __in6_dev_get(dst_dev);
1850                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1851                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1852                     (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1853                         use_oif_addr = true;
1854                 }
1855         }
1856
1857         if (use_oif_addr) {
1858                 if (idev)
1859                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1860         } else {
1861                 const struct net_device *master;
1862                 int master_idx = 0;
1863
1864                 /* if dst_dev exists and is enslaved to an L3 device, then
1865                  * prefer addresses from dst_dev and then the master over
1866                  * any other enslaved devices in the L3 domain.
1867                  */
1868                 master = l3mdev_master_dev_rcu(dst_dev);
1869                 if (master) {
1870                         master_idx = master->ifindex;
1871
1872                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1873                                                             master, &dst,
1874                                                             scores, hiscore_idx);
1875
1876                         if (scores[hiscore_idx].ifa &&
1877                             scores[hiscore_idx].scopedist >= 0)
1878                                 goto out;
1879                 }
1880
1881                 for_each_netdev_rcu(net, dev) {
1882                         /* only consider addresses on devices in the
1883                          * same L3 domain
1884                          */
1885                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1886                                 continue;
1887                         idev = __in6_dev_get(dev);
1888                         if (!idev)
1889                                 continue;
1890                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1891                 }
1892         }
1893
1894 out:
1895         hiscore = &scores[hiscore_idx];
1896         if (!hiscore->ifa)
1897                 ret = -EADDRNOTAVAIL;
1898         else
1899                 *saddr = hiscore->ifa->addr;
1900
1901         rcu_read_unlock();
1902         return ret;
1903 }
1904 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1905
1906 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1907                               u32 banned_flags)
1908 {
1909         struct inet6_ifaddr *ifp;
1910         int err = -EADDRNOTAVAIL;
1911
1912         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1913                 if (ifp->scope > IFA_LINK)
1914                         break;
1915                 if (ifp->scope == IFA_LINK &&
1916                     !(ifp->flags & banned_flags)) {
1917                         *addr = ifp->addr;
1918                         err = 0;
1919                         break;
1920                 }
1921         }
1922         return err;
1923 }
1924
1925 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1926                     u32 banned_flags)
1927 {
1928         struct inet6_dev *idev;
1929         int err = -EADDRNOTAVAIL;
1930
1931         rcu_read_lock();
1932         idev = __in6_dev_get(dev);
1933         if (idev) {
1934                 read_lock_bh(&idev->lock);
1935                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1936                 read_unlock_bh(&idev->lock);
1937         }
1938         rcu_read_unlock();
1939         return err;
1940 }
1941
1942 static int ipv6_count_addresses(const struct inet6_dev *idev)
1943 {
1944         const struct inet6_ifaddr *ifp;
1945         int cnt = 0;
1946
1947         rcu_read_lock();
1948         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1949                 cnt++;
1950         rcu_read_unlock();
1951         return cnt;
1952 }
1953
1954 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1955                   const struct net_device *dev, int strict)
1956 {
1957         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1958                                        strict, IFA_F_TENTATIVE);
1959 }
1960 EXPORT_SYMBOL(ipv6_chk_addr);
1961
1962 /* device argument is used to find the L3 domain of interest. If
1963  * skip_dev_check is set, then the ifp device is not checked against
1964  * the passed in dev argument. So the 2 cases for addresses checks are:
1965  *   1. does the address exist in the L3 domain that dev is part of
1966  *      (skip_dev_check = true), or
1967  *
1968  *   2. does the address exist on the specific device
1969  *      (skip_dev_check = false)
1970  */
1971 static struct net_device *
1972 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1973                           const struct net_device *dev, bool skip_dev_check,
1974                           int strict, u32 banned_flags)
1975 {
1976         unsigned int hash = inet6_addr_hash(net, addr);
1977         struct net_device *l3mdev, *ndev;
1978         struct inet6_ifaddr *ifp;
1979         u32 ifp_flags;
1980
1981         rcu_read_lock();
1982
1983         l3mdev = l3mdev_master_dev_rcu(dev);
1984         if (skip_dev_check)
1985                 dev = NULL;
1986
1987         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1988                 ndev = ifp->idev->dev;
1989
1990                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1991                         continue;
1992
1993                 /* Decouple optimistic from tentative for evaluation here.
1994                  * Ban optimistic addresses explicitly, when required.
1995                  */
1996                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1997                             ? (ifp->flags&~IFA_F_TENTATIVE)
1998                             : ifp->flags;
1999                 if (ipv6_addr_equal(&ifp->addr, addr) &&
2000                     !(ifp_flags&banned_flags) &&
2001                     (!dev || ndev == dev ||
2002                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2003                         rcu_read_unlock();
2004                         return ndev;
2005                 }
2006         }
2007
2008         rcu_read_unlock();
2009         return NULL;
2010 }
2011
2012 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2013                             const struct net_device *dev, bool skip_dev_check,
2014                             int strict, u32 banned_flags)
2015 {
2016         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2017                                          strict, banned_flags) ? 1 : 0;
2018 }
2019 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2020
2021
2022 /* Compares an address/prefix_len with addresses on device @dev.
2023  * If one is found it returns true.
2024  */
2025 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2026         const unsigned int prefix_len, struct net_device *dev)
2027 {
2028         const struct inet6_ifaddr *ifa;
2029         const struct inet6_dev *idev;
2030         bool ret = false;
2031
2032         rcu_read_lock();
2033         idev = __in6_dev_get(dev);
2034         if (idev) {
2035                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2036                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2037                         if (ret)
2038                                 break;
2039                 }
2040         }
2041         rcu_read_unlock();
2042
2043         return ret;
2044 }
2045 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2046
2047 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2048 {
2049         const struct inet6_ifaddr *ifa;
2050         const struct inet6_dev *idev;
2051         int     onlink;
2052
2053         onlink = 0;
2054         rcu_read_lock();
2055         idev = __in6_dev_get(dev);
2056         if (idev) {
2057                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2058                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
2059                                                    ifa->prefix_len);
2060                         if (onlink)
2061                                 break;
2062                 }
2063         }
2064         rcu_read_unlock();
2065         return onlink;
2066 }
2067 EXPORT_SYMBOL(ipv6_chk_prefix);
2068
2069 /**
2070  * ipv6_dev_find - find the first device with a given source address.
2071  * @net: the net namespace
2072  * @addr: the source address
2073  * @dev: used to find the L3 domain of interest
2074  *
2075  * The caller should be protected by RCU, or RTNL.
2076  */
2077 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2078                                  struct net_device *dev)
2079 {
2080         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2081                                          IFA_F_TENTATIVE);
2082 }
2083 EXPORT_SYMBOL(ipv6_dev_find);
2084
2085 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2086                                      struct net_device *dev, int strict)
2087 {
2088         unsigned int hash = inet6_addr_hash(net, addr);
2089         struct inet6_ifaddr *ifp, *result = NULL;
2090
2091         rcu_read_lock();
2092         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2093                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2094                         if (!dev || ifp->idev->dev == dev ||
2095                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2096                                 if (in6_ifa_hold_safe(ifp)) {
2097                                         result = ifp;
2098                                         break;
2099                                 }
2100                         }
2101                 }
2102         }
2103         rcu_read_unlock();
2104
2105         return result;
2106 }
2107
2108 /* Gets referenced address, destroys ifaddr */
2109
2110 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2111 {
2112         if (dad_failed)
2113                 ifp->flags |= IFA_F_DADFAILED;
2114
2115         if (ifp->flags&IFA_F_TEMPORARY) {
2116                 struct inet6_ifaddr *ifpub;
2117                 spin_lock_bh(&ifp->lock);
2118                 ifpub = ifp->ifpub;
2119                 if (ifpub) {
2120                         in6_ifa_hold(ifpub);
2121                         spin_unlock_bh(&ifp->lock);
2122                         ipv6_create_tempaddr(ifpub, true);
2123                         in6_ifa_put(ifpub);
2124                 } else {
2125                         spin_unlock_bh(&ifp->lock);
2126                 }
2127                 ipv6_del_addr(ifp);
2128         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2129                 spin_lock_bh(&ifp->lock);
2130                 addrconf_del_dad_work(ifp);
2131                 ifp->flags |= IFA_F_TENTATIVE;
2132                 if (dad_failed)
2133                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2134                 spin_unlock_bh(&ifp->lock);
2135                 if (dad_failed)
2136                         ipv6_ifa_notify(0, ifp);
2137                 in6_ifa_put(ifp);
2138         } else {
2139                 ipv6_del_addr(ifp);
2140         }
2141 }
2142
2143 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2144 {
2145         int err = -ENOENT;
2146
2147         spin_lock_bh(&ifp->lock);
2148         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2149                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2150                 err = 0;
2151         }
2152         spin_unlock_bh(&ifp->lock);
2153
2154         return err;
2155 }
2156
2157 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2158 {
2159         struct inet6_dev *idev = ifp->idev;
2160         struct net *net = dev_net(idev->dev);
2161         int max_addresses;
2162
2163         if (addrconf_dad_end(ifp)) {
2164                 in6_ifa_put(ifp);
2165                 return;
2166         }
2167
2168         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2169                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2170
2171         spin_lock_bh(&ifp->lock);
2172
2173         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2174                 struct in6_addr new_addr;
2175                 struct inet6_ifaddr *ifp2;
2176                 int retries = ifp->stable_privacy_retry + 1;
2177                 struct ifa6_config cfg = {
2178                         .pfx = &new_addr,
2179                         .plen = ifp->prefix_len,
2180                         .ifa_flags = ifp->flags,
2181                         .valid_lft = ifp->valid_lft,
2182                         .preferred_lft = ifp->prefered_lft,
2183                         .scope = ifp->scope,
2184                 };
2185
2186                 if (retries > net->ipv6.sysctl.idgen_retries) {
2187                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2188                                              ifp->idev->dev->name);
2189                         goto errdad;
2190                 }
2191
2192                 new_addr = ifp->addr;
2193                 if (ipv6_generate_stable_address(&new_addr, retries,
2194                                                  idev))
2195                         goto errdad;
2196
2197                 spin_unlock_bh(&ifp->lock);
2198
2199                 max_addresses = READ_ONCE(idev->cnf.max_addresses);
2200                 if (max_addresses &&
2201                     ipv6_count_addresses(idev) >= max_addresses)
2202                         goto lock_errdad;
2203
2204                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2205                                      ifp->idev->dev->name);
2206
2207                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2208                 if (IS_ERR(ifp2))
2209                         goto lock_errdad;
2210
2211                 spin_lock_bh(&ifp2->lock);
2212                 ifp2->stable_privacy_retry = retries;
2213                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2214                 spin_unlock_bh(&ifp2->lock);
2215
2216                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2217                 in6_ifa_put(ifp2);
2218 lock_errdad:
2219                 spin_lock_bh(&ifp->lock);
2220         }
2221
2222 errdad:
2223         /* transition from _POSTDAD to _ERRDAD */
2224         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2225         spin_unlock_bh(&ifp->lock);
2226
2227         addrconf_mod_dad_work(ifp, 0);
2228         in6_ifa_put(ifp);
2229 }
2230
2231 /* Join to solicited addr multicast group.
2232  * caller must hold RTNL */
2233 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2234 {
2235         struct in6_addr maddr;
2236
2237         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2238                 return;
2239
2240         addrconf_addr_solict_mult(addr, &maddr);
2241         ipv6_dev_mc_inc(dev, &maddr);
2242 }
2243
2244 /* caller must hold RTNL */
2245 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2246 {
2247         struct in6_addr maddr;
2248
2249         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2250                 return;
2251
2252         addrconf_addr_solict_mult(addr, &maddr);
2253         __ipv6_dev_mc_dec(idev, &maddr);
2254 }
2255
2256 /* caller must hold RTNL */
2257 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2258 {
2259         struct in6_addr addr;
2260
2261         if (ifp->prefix_len >= 127) /* RFC 6164 */
2262                 return;
2263         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2264         if (ipv6_addr_any(&addr))
2265                 return;
2266         __ipv6_dev_ac_inc(ifp->idev, &addr);
2267 }
2268
2269 /* caller must hold RTNL */
2270 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2271 {
2272         struct in6_addr addr;
2273
2274         if (ifp->prefix_len >= 127) /* RFC 6164 */
2275                 return;
2276         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2277         if (ipv6_addr_any(&addr))
2278                 return;
2279         __ipv6_dev_ac_dec(ifp->idev, &addr);
2280 }
2281
2282 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2283 {
2284         switch (dev->addr_len) {
2285         case ETH_ALEN:
2286                 memcpy(eui, dev->dev_addr, 3);
2287                 eui[3] = 0xFF;
2288                 eui[4] = 0xFE;
2289                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2290                 break;
2291         case EUI64_ADDR_LEN:
2292                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2293                 eui[0] ^= 2;
2294                 break;
2295         default:
2296                 return -1;
2297         }
2298
2299         return 0;
2300 }
2301
2302 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2303 {
2304         const union fwnet_hwaddr *ha;
2305
2306         if (dev->addr_len != FWNET_ALEN)
2307                 return -1;
2308
2309         ha = (const union fwnet_hwaddr *)dev->dev_addr;
2310
2311         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2312         eui[0] ^= 2;
2313         return 0;
2314 }
2315
2316 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2317 {
2318         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2319         if (dev->addr_len != ARCNET_ALEN)
2320                 return -1;
2321         memset(eui, 0, 7);
2322         eui[7] = *(u8 *)dev->dev_addr;
2323         return 0;
2324 }
2325
2326 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2327 {
2328         if (dev->addr_len != INFINIBAND_ALEN)
2329                 return -1;
2330         memcpy(eui, dev->dev_addr + 12, 8);
2331         eui[0] |= 2;
2332         return 0;
2333 }
2334
2335 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2336 {
2337         if (addr == 0)
2338                 return -1;
2339         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2340                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2341                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2342                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2343                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2344                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2345         eui[1] = 0;
2346         eui[2] = 0x5E;
2347         eui[3] = 0xFE;
2348         memcpy(eui + 4, &addr, 4);
2349         return 0;
2350 }
2351
2352 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2353 {
2354         if (dev->priv_flags & IFF_ISATAP)
2355                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2356         return -1;
2357 }
2358
2359 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2360 {
2361         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2362 }
2363
2364 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2365 {
2366         memcpy(eui, dev->perm_addr, 3);
2367         memcpy(eui + 5, dev->perm_addr + 3, 3);
2368         eui[3] = 0xFF;
2369         eui[4] = 0xFE;
2370         eui[0] ^= 2;
2371         return 0;
2372 }
2373
2374 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2375 {
2376         switch (dev->type) {
2377         case ARPHRD_ETHER:
2378         case ARPHRD_FDDI:
2379                 return addrconf_ifid_eui48(eui, dev);
2380         case ARPHRD_ARCNET:
2381                 return addrconf_ifid_arcnet(eui, dev);
2382         case ARPHRD_INFINIBAND:
2383                 return addrconf_ifid_infiniband(eui, dev);
2384         case ARPHRD_SIT:
2385                 return addrconf_ifid_sit(eui, dev);
2386         case ARPHRD_IPGRE:
2387         case ARPHRD_TUNNEL:
2388                 return addrconf_ifid_gre(eui, dev);
2389         case ARPHRD_6LOWPAN:
2390                 return addrconf_ifid_6lowpan(eui, dev);
2391         case ARPHRD_IEEE1394:
2392                 return addrconf_ifid_ieee1394(eui, dev);
2393         case ARPHRD_TUNNEL6:
2394         case ARPHRD_IP6GRE:
2395         case ARPHRD_RAWIP:
2396                 return addrconf_ifid_ip6tnl(eui, dev);
2397         }
2398         return -1;
2399 }
2400
2401 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2402 {
2403         int err = -1;
2404         struct inet6_ifaddr *ifp;
2405
2406         read_lock_bh(&idev->lock);
2407         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2408                 if (ifp->scope > IFA_LINK)
2409                         break;
2410                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2411                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2412                         err = 0;
2413                         break;
2414                 }
2415         }
2416         read_unlock_bh(&idev->lock);
2417         return err;
2418 }
2419
2420 /* Generation of a randomized Interface Identifier
2421  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2422  */
2423
2424 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2425 {
2426 regen:
2427         get_random_bytes(&addr->s6_addr[8], 8);
2428
2429         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2430          * check if generated address is not inappropriate:
2431          *
2432          * - Reserved IPv6 Interface Identifiers
2433          * - XXX: already assigned to an address on the device
2434          */
2435
2436         /* Subnet-router anycast: 0000:0000:0000:0000 */
2437         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2438                 goto regen;
2439
2440         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2441          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2442          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2443          */
2444         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2445             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2446                 goto regen;
2447
2448         /* Reserved subnet anycast addresses */
2449         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2450             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2451                 goto regen;
2452 }
2453
2454 /*
2455  *      Add prefix route.
2456  */
2457
2458 static void
2459 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2460                       struct net_device *dev, unsigned long expires,
2461                       u32 flags, gfp_t gfp_flags)
2462 {
2463         struct fib6_config cfg = {
2464                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2465                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2466                 .fc_ifindex = dev->ifindex,
2467                 .fc_expires = expires,
2468                 .fc_dst_len = plen,
2469                 .fc_flags = RTF_UP | flags,
2470                 .fc_nlinfo.nl_net = dev_net(dev),
2471                 .fc_protocol = RTPROT_KERNEL,
2472                 .fc_type = RTN_UNICAST,
2473         };
2474
2475         cfg.fc_dst = *pfx;
2476
2477         /* Prevent useless cloning on PtP SIT.
2478            This thing is done here expecting that the whole
2479            class of non-broadcast devices need not cloning.
2480          */
2481 #if IS_ENABLED(CONFIG_IPV6_SIT)
2482         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2483                 cfg.fc_flags |= RTF_NONEXTHOP;
2484 #endif
2485
2486         ip6_route_add(&cfg, gfp_flags, NULL);
2487 }
2488
2489
2490 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2491                                                   int plen,
2492                                                   const struct net_device *dev,
2493                                                   u32 flags, u32 noflags,
2494                                                   bool no_gw)
2495 {
2496         struct fib6_node *fn;
2497         struct fib6_info *rt = NULL;
2498         struct fib6_table *table;
2499         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2500
2501         table = fib6_get_table(dev_net(dev), tb_id);
2502         if (!table)
2503                 return NULL;
2504
2505         rcu_read_lock();
2506         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2507         if (!fn)
2508                 goto out;
2509
2510         for_each_fib6_node_rt_rcu(fn) {
2511                 /* prefix routes only use builtin fib6_nh */
2512                 if (rt->nh)
2513                         continue;
2514
2515                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2516                         continue;
2517                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2518                         continue;
2519                 if ((rt->fib6_flags & flags) != flags)
2520                         continue;
2521                 if ((rt->fib6_flags & noflags) != 0)
2522                         continue;
2523                 if (!fib6_info_hold_safe(rt))
2524                         continue;
2525                 break;
2526         }
2527 out:
2528         rcu_read_unlock();
2529         return rt;
2530 }
2531
2532
2533 /* Create "default" multicast route to the interface */
2534
2535 static void addrconf_add_mroute(struct net_device *dev)
2536 {
2537         struct fib6_config cfg = {
2538                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2539                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2540                 .fc_ifindex = dev->ifindex,
2541                 .fc_dst_len = 8,
2542                 .fc_flags = RTF_UP,
2543                 .fc_type = RTN_MULTICAST,
2544                 .fc_nlinfo.nl_net = dev_net(dev),
2545                 .fc_protocol = RTPROT_KERNEL,
2546         };
2547
2548         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2549
2550         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2551 }
2552
2553 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2554 {
2555         struct inet6_dev *idev;
2556
2557         ASSERT_RTNL();
2558
2559         idev = ipv6_find_idev(dev);
2560         if (IS_ERR(idev))
2561                 return idev;
2562
2563         if (idev->cnf.disable_ipv6)
2564                 return ERR_PTR(-EACCES);
2565
2566         /* Add default multicast route */
2567         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2568                 addrconf_add_mroute(dev);
2569
2570         return idev;
2571 }
2572
2573 static void manage_tempaddrs(struct inet6_dev *idev,
2574                              struct inet6_ifaddr *ifp,
2575                              __u32 valid_lft, __u32 prefered_lft,
2576                              bool create, unsigned long now)
2577 {
2578         u32 flags;
2579         struct inet6_ifaddr *ift;
2580
2581         read_lock_bh(&idev->lock);
2582         /* update all temporary addresses in the list */
2583         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2584                 int age, max_valid, max_prefered;
2585
2586                 if (ifp != ift->ifpub)
2587                         continue;
2588
2589                 /* RFC 4941 section 3.3:
2590                  * If a received option will extend the lifetime of a public
2591                  * address, the lifetimes of temporary addresses should
2592                  * be extended, subject to the overall constraint that no
2593                  * temporary addresses should ever remain "valid" or "preferred"
2594                  * for a time longer than (TEMP_VALID_LIFETIME) or
2595                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2596                  */
2597                 age = (now - ift->cstamp) / HZ;
2598                 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2599                 if (max_valid < 0)
2600                         max_valid = 0;
2601
2602                 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2603                                idev->desync_factor - age;
2604                 if (max_prefered < 0)
2605                         max_prefered = 0;
2606
2607                 if (valid_lft > max_valid)
2608                         valid_lft = max_valid;
2609
2610                 if (prefered_lft > max_prefered)
2611                         prefered_lft = max_prefered;
2612
2613                 spin_lock(&ift->lock);
2614                 flags = ift->flags;
2615                 ift->valid_lft = valid_lft;
2616                 ift->prefered_lft = prefered_lft;
2617                 ift->tstamp = now;
2618                 if (prefered_lft > 0)
2619                         ift->flags &= ~IFA_F_DEPRECATED;
2620
2621                 spin_unlock(&ift->lock);
2622                 if (!(flags&IFA_F_TENTATIVE))
2623                         ipv6_ifa_notify(0, ift);
2624         }
2625
2626         /* Also create a temporary address if it's enabled but no temporary
2627          * address currently exists.
2628          * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2629          * as part of cleanup (ie. deleting the mngtmpaddr).
2630          * We don't want that to result in creating a new temporary ip address.
2631          */
2632         if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2633                 create = true;
2634
2635         if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2636                 /* When a new public address is created as described
2637                  * in [ADDRCONF], also create a new temporary address.
2638                  */
2639                 read_unlock_bh(&idev->lock);
2640                 ipv6_create_tempaddr(ifp, false);
2641         } else {
2642                 read_unlock_bh(&idev->lock);
2643         }
2644 }
2645
2646 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2647 {
2648         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2649                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2650 }
2651
2652 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2653                                  const struct prefix_info *pinfo,
2654                                  struct inet6_dev *in6_dev,
2655                                  const struct in6_addr *addr, int addr_type,
2656                                  u32 addr_flags, bool sllao, bool tokenized,
2657                                  __u32 valid_lft, u32 prefered_lft)
2658 {
2659         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2660         int create = 0, update_lft = 0;
2661
2662         if (!ifp && valid_lft) {
2663                 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2664                 struct ifa6_config cfg = {
2665                         .pfx = addr,
2666                         .plen = pinfo->prefix_len,
2667                         .ifa_flags = addr_flags,
2668                         .valid_lft = valid_lft,
2669                         .preferred_lft = prefered_lft,
2670                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2671                         .ifa_proto = IFAPROT_KERNEL_RA
2672                 };
2673
2674 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2675                 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2676                      READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2677                     !net->ipv6.devconf_all->forwarding && sllao)
2678                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2679 #endif
2680
2681                 /* Do not allow to create too much of autoconfigured
2682                  * addresses; this would be too easy way to crash kernel.
2683                  */
2684                 if (!max_addresses ||
2685                     ipv6_count_addresses(in6_dev) < max_addresses)
2686                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2687
2688                 if (IS_ERR_OR_NULL(ifp))
2689                         return -1;
2690
2691                 create = 1;
2692                 spin_lock_bh(&ifp->lock);
2693                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2694                 ifp->cstamp = jiffies;
2695                 ifp->tokenized = tokenized;
2696                 spin_unlock_bh(&ifp->lock);
2697                 addrconf_dad_start(ifp);
2698         }
2699
2700         if (ifp) {
2701                 u32 flags;
2702                 unsigned long now;
2703                 u32 stored_lft;
2704
2705                 /* update lifetime (RFC2462 5.5.3 e) */
2706                 spin_lock_bh(&ifp->lock);
2707                 now = jiffies;
2708                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2709                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2710                 else
2711                         stored_lft = 0;
2712
2713                 /* RFC4862 Section 5.5.3e:
2714                  * "Note that the preferred lifetime of the
2715                  *  corresponding address is always reset to
2716                  *  the Preferred Lifetime in the received
2717                  *  Prefix Information option, regardless of
2718                  *  whether the valid lifetime is also reset or
2719                  *  ignored."
2720                  *
2721                  * So we should always update prefered_lft here.
2722                  */
2723                 update_lft = !create && stored_lft;
2724
2725                 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2726                         const u32 minimum_lft = min_t(u32,
2727                                 stored_lft, MIN_VALID_LIFETIME);
2728                         valid_lft = max(valid_lft, minimum_lft);
2729                 }
2730
2731                 if (update_lft) {
2732                         ifp->valid_lft = valid_lft;
2733                         ifp->prefered_lft = prefered_lft;
2734                         WRITE_ONCE(ifp->tstamp, now);
2735                         flags = ifp->flags;
2736                         ifp->flags &= ~IFA_F_DEPRECATED;
2737                         spin_unlock_bh(&ifp->lock);
2738
2739                         if (!(flags&IFA_F_TENTATIVE))
2740                                 ipv6_ifa_notify(0, ifp);
2741                 } else
2742                         spin_unlock_bh(&ifp->lock);
2743
2744                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2745                                  create, now);
2746
2747                 in6_ifa_put(ifp);
2748                 addrconf_verify(net);
2749         }
2750
2751         return 0;
2752 }
2753 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2754
2755 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2756 {
2757         struct prefix_info *pinfo;
2758         struct fib6_table *table;
2759         __u32 valid_lft;
2760         __u32 prefered_lft;
2761         int addr_type, err;
2762         u32 addr_flags = 0;
2763         struct inet6_dev *in6_dev;
2764         struct net *net = dev_net(dev);
2765         bool ignore_autoconf = false;
2766
2767         pinfo = (struct prefix_info *) opt;
2768
2769         if (len < sizeof(struct prefix_info)) {
2770                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2771                 return;
2772         }
2773
2774         /*
2775          *      Validation checks ([ADDRCONF], page 19)
2776          */
2777
2778         addr_type = ipv6_addr_type(&pinfo->prefix);
2779
2780         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2781                 return;
2782
2783         valid_lft = ntohl(pinfo->valid);
2784         prefered_lft = ntohl(pinfo->prefered);
2785
2786         if (prefered_lft > valid_lft) {
2787                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2788                 return;
2789         }
2790
2791         in6_dev = in6_dev_get(dev);
2792
2793         if (!in6_dev) {
2794                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2795                                     dev->name);
2796                 return;
2797         }
2798
2799         if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2800                 goto put;
2801
2802         /*
2803          *      Two things going on here:
2804          *      1) Add routes for on-link prefixes
2805          *      2) Configure prefixes with the auto flag set
2806          */
2807
2808         if (pinfo->onlink) {
2809                 struct fib6_info *rt;
2810                 unsigned long rt_expires;
2811
2812                 /* Avoid arithmetic overflow. Really, we could
2813                  * save rt_expires in seconds, likely valid_lft,
2814                  * but it would require division in fib gc, that it
2815                  * not good.
2816                  */
2817                 if (HZ > USER_HZ)
2818                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2819                 else
2820                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2821
2822                 if (addrconf_finite_timeout(rt_expires))
2823                         rt_expires *= HZ;
2824
2825                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2826                                                pinfo->prefix_len,
2827                                                dev,
2828                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2829                                                RTF_DEFAULT, true);
2830
2831                 if (rt) {
2832                         /* Autoconf prefix route */
2833                         if (valid_lft == 0) {
2834                                 ip6_del_rt(net, rt, false);
2835                                 rt = NULL;
2836                         } else {
2837                                 table = rt->fib6_table;
2838                                 spin_lock_bh(&table->tb6_lock);
2839
2840                                 if (addrconf_finite_timeout(rt_expires)) {
2841                                         /* not infinity */
2842                                         fib6_set_expires(rt, jiffies + rt_expires);
2843                                         fib6_add_gc_list(rt);
2844                                 } else {
2845                                         fib6_clean_expires(rt);
2846                                         fib6_remove_gc_list(rt);
2847                                 }
2848
2849                                 spin_unlock_bh(&table->tb6_lock);
2850                         }
2851                 } else if (valid_lft) {
2852                         clock_t expires = 0;
2853                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2854                         if (addrconf_finite_timeout(rt_expires)) {
2855                                 /* not infinity */
2856                                 flags |= RTF_EXPIRES;
2857                                 expires = jiffies_to_clock_t(rt_expires);
2858                         }
2859                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2860                                               0, dev, expires, flags,
2861                                               GFP_ATOMIC);
2862                 }
2863                 fib6_info_release(rt);
2864         }
2865
2866         /* Try to figure out our local address for this prefix */
2867
2868         ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd;
2869         if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) {
2870                 struct in6_addr addr;
2871                 bool tokenized = false, dev_addr_generated = false;
2872
2873                 if (pinfo->prefix_len == 64) {
2874                         memcpy(&addr, &pinfo->prefix, 8);
2875
2876                         if (!ipv6_addr_any(&in6_dev->token)) {
2877                                 read_lock_bh(&in6_dev->lock);
2878                                 memcpy(addr.s6_addr + 8,
2879                                        in6_dev->token.s6_addr + 8, 8);
2880                                 read_unlock_bh(&in6_dev->lock);
2881                                 tokenized = true;
2882                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2883                                    !ipv6_generate_stable_address(&addr, 0,
2884                                                                  in6_dev)) {
2885                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2886                                 goto ok;
2887                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2888                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2889                                 goto put;
2890                         } else {
2891                                 dev_addr_generated = true;
2892                         }
2893                         goto ok;
2894                 }
2895                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2896                                     pinfo->prefix_len);
2897                 goto put;
2898
2899 ok:
2900                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2901                                                    &addr, addr_type,
2902                                                    addr_flags, sllao,
2903                                                    tokenized, valid_lft,
2904                                                    prefered_lft);
2905                 if (err)
2906                         goto put;
2907
2908                 /* Ignore error case here because previous prefix add addr was
2909                  * successful which will be notified.
2910                  */
2911                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2912                                               addr_type, addr_flags, sllao,
2913                                               tokenized, valid_lft,
2914                                               prefered_lft,
2915                                               dev_addr_generated);
2916         }
2917         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2918 put:
2919         in6_dev_put(in6_dev);
2920 }
2921
2922 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2923                 struct in6_ifreq *ireq)
2924 {
2925         struct ip_tunnel_parm_kern p = { };
2926         int err;
2927
2928         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2929                 return -EADDRNOTAVAIL;
2930
2931         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2932         p.iph.version = 4;
2933         p.iph.ihl = 5;
2934         p.iph.protocol = IPPROTO_IPV6;
2935         p.iph.ttl = 64;
2936
2937         if (!dev->netdev_ops->ndo_tunnel_ctl)
2938                 return -EOPNOTSUPP;
2939         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2940         if (err)
2941                 return err;
2942
2943         dev = __dev_get_by_name(net, p.name);
2944         if (!dev)
2945                 return -ENOBUFS;
2946         return dev_open(dev, NULL);
2947 }
2948
2949 /*
2950  *      Set destination address.
2951  *      Special case for SIT interfaces where we create a new "virtual"
2952  *      device.
2953  */
2954 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2955 {
2956         struct net_device *dev;
2957         struct in6_ifreq ireq;
2958         int err = -ENODEV;
2959
2960         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2961                 return -ENODEV;
2962         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2963                 return -EFAULT;
2964
2965         rtnl_lock();
2966         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2967         if (dev && dev->type == ARPHRD_SIT)
2968                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2969         rtnl_unlock();
2970         return err;
2971 }
2972
2973 static int ipv6_mc_config(struct sock *sk, bool join,
2974                           const struct in6_addr *addr, int ifindex)
2975 {
2976         int ret;
2977
2978         ASSERT_RTNL();
2979
2980         lock_sock(sk);
2981         if (join)
2982                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2983         else
2984                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2985         release_sock(sk);
2986
2987         return ret;
2988 }
2989
2990 /*
2991  *      Manual configuration of address on an interface
2992  */
2993 static int inet6_addr_add(struct net *net, int ifindex,
2994                           struct ifa6_config *cfg,
2995                           struct netlink_ext_ack *extack)
2996 {
2997         struct inet6_ifaddr *ifp;
2998         struct inet6_dev *idev;
2999         struct net_device *dev;
3000         unsigned long timeout;
3001         clock_t expires;
3002         u32 flags;
3003
3004         ASSERT_RTNL();
3005
3006         if (cfg->plen > 128) {
3007                 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3008                 return -EINVAL;
3009         }
3010
3011         /* check the lifetime */
3012         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
3013                 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
3014                 return -EINVAL;
3015         }
3016
3017         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3018                 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3019                 return -EINVAL;
3020         }
3021
3022         dev = __dev_get_by_index(net, ifindex);
3023         if (!dev)
3024                 return -ENODEV;
3025
3026         idev = addrconf_add_dev(dev);
3027         if (IS_ERR(idev)) {
3028                 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3029                 return PTR_ERR(idev);
3030         }
3031
3032         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3033                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3034                                          true, cfg->pfx, ifindex);
3035
3036                 if (ret < 0) {
3037                         NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3038                         return ret;
3039                 }
3040         }
3041
3042         cfg->scope = ipv6_addr_scope(cfg->pfx);
3043
3044         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
3045         if (addrconf_finite_timeout(timeout)) {
3046                 expires = jiffies_to_clock_t(timeout * HZ);
3047                 cfg->valid_lft = timeout;
3048                 flags = RTF_EXPIRES;
3049         } else {
3050                 expires = 0;
3051                 flags = 0;
3052                 cfg->ifa_flags |= IFA_F_PERMANENT;
3053         }
3054
3055         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3056         if (addrconf_finite_timeout(timeout)) {
3057                 if (timeout == 0)
3058                         cfg->ifa_flags |= IFA_F_DEPRECATED;
3059                 cfg->preferred_lft = timeout;
3060         }
3061
3062         ifp = ipv6_add_addr(idev, cfg, true, extack);
3063         if (!IS_ERR(ifp)) {
3064                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3065                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3066                                               ifp->rt_priority, dev, expires,
3067                                               flags, GFP_KERNEL);
3068                 }
3069
3070                 /* Send a netlink notification if DAD is enabled and
3071                  * optimistic flag is not set
3072                  */
3073                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3074                         ipv6_ifa_notify(0, ifp);
3075                 /*
3076                  * Note that section 3.1 of RFC 4429 indicates
3077                  * that the Optimistic flag should not be set for
3078                  * manually configured addresses
3079                  */
3080                 addrconf_dad_start(ifp);
3081                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3082                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
3083                                          cfg->preferred_lft, true, jiffies);
3084                 in6_ifa_put(ifp);
3085                 addrconf_verify_rtnl(net);
3086                 return 0;
3087         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3088                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3089                                cfg->pfx, ifindex);
3090         }
3091
3092         return PTR_ERR(ifp);
3093 }
3094
3095 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3096                           const struct in6_addr *pfx, unsigned int plen,
3097                           struct netlink_ext_ack *extack)
3098 {
3099         struct inet6_ifaddr *ifp;
3100         struct inet6_dev *idev;
3101         struct net_device *dev;
3102
3103         if (plen > 128) {
3104                 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3105                 return -EINVAL;
3106         }
3107
3108         dev = __dev_get_by_index(net, ifindex);
3109         if (!dev) {
3110                 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3111                 return -ENODEV;
3112         }
3113
3114         idev = __in6_dev_get(dev);
3115         if (!idev) {
3116                 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3117                 return -ENXIO;
3118         }
3119
3120         read_lock_bh(&idev->lock);
3121         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3122                 if (ifp->prefix_len == plen &&
3123                     ipv6_addr_equal(pfx, &ifp->addr)) {
3124                         in6_ifa_hold(ifp);
3125                         read_unlock_bh(&idev->lock);
3126
3127                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3128                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3129                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3130                                                  jiffies);
3131                         ipv6_del_addr(ifp);
3132                         addrconf_verify_rtnl(net);
3133                         if (ipv6_addr_is_multicast(pfx)) {
3134                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3135                                                false, pfx, dev->ifindex);
3136                         }
3137                         return 0;
3138                 }
3139         }
3140         read_unlock_bh(&idev->lock);
3141
3142         NL_SET_ERR_MSG_MOD(extack, "address not found");
3143         return -EADDRNOTAVAIL;
3144 }
3145
3146
3147 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3148 {
3149         struct ifa6_config cfg = {
3150                 .ifa_flags = IFA_F_PERMANENT,
3151                 .preferred_lft = INFINITY_LIFE_TIME,
3152                 .valid_lft = INFINITY_LIFE_TIME,
3153         };
3154         struct in6_ifreq ireq;
3155         int err;
3156
3157         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3158                 return -EPERM;
3159
3160         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3161                 return -EFAULT;
3162
3163         cfg.pfx = &ireq.ifr6_addr;
3164         cfg.plen = ireq.ifr6_prefixlen;
3165
3166         rtnl_lock();
3167         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3168         rtnl_unlock();
3169         return err;
3170 }
3171
3172 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3173 {
3174         struct in6_ifreq ireq;
3175         int err;
3176
3177         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3178                 return -EPERM;
3179
3180         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3181                 return -EFAULT;
3182
3183         rtnl_lock();
3184         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3185                              ireq.ifr6_prefixlen, NULL);
3186         rtnl_unlock();
3187         return err;
3188 }
3189
3190 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3191                      int plen, int scope, u8 proto)
3192 {
3193         struct inet6_ifaddr *ifp;
3194         struct ifa6_config cfg = {
3195                 .pfx = addr,
3196                 .plen = plen,
3197                 .ifa_flags = IFA_F_PERMANENT,
3198                 .valid_lft = INFINITY_LIFE_TIME,
3199                 .preferred_lft = INFINITY_LIFE_TIME,
3200                 .scope = scope,
3201                 .ifa_proto = proto
3202         };
3203
3204         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3205         if (!IS_ERR(ifp)) {
3206                 spin_lock_bh(&ifp->lock);
3207                 ifp->flags &= ~IFA_F_TENTATIVE;
3208                 spin_unlock_bh(&ifp->lock);
3209                 rt_genid_bump_ipv6(dev_net(idev->dev));
3210                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3211                 in6_ifa_put(ifp);
3212         }
3213 }
3214
3215 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3216 static void add_v4_addrs(struct inet6_dev *idev)
3217 {
3218         struct in6_addr addr;
3219         struct net_device *dev;
3220         struct net *net = dev_net(idev->dev);
3221         int scope, plen, offset = 0;
3222         u32 pflags = 0;
3223
3224         ASSERT_RTNL();
3225
3226         memset(&addr, 0, sizeof(struct in6_addr));
3227         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3228         if (idev->dev->addr_len == sizeof(struct in6_addr))
3229                 offset = sizeof(struct in6_addr) - 4;
3230         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3231
3232         if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3233                 scope = IPV6_ADDR_COMPATv4;
3234                 plen = 96;
3235                 pflags |= RTF_NONEXTHOP;
3236         } else {
3237                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3238                         return;
3239
3240                 addr.s6_addr32[0] = htonl(0xfe800000);
3241                 scope = IFA_LINK;
3242                 plen = 64;
3243         }
3244
3245         if (addr.s6_addr32[3]) {
3246                 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3247                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3248                                       GFP_KERNEL);
3249                 return;
3250         }
3251
3252         for_each_netdev(net, dev) {
3253                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3254                 if (in_dev && (dev->flags & IFF_UP)) {
3255                         struct in_ifaddr *ifa;
3256                         int flag = scope;
3257
3258                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3259                                 addr.s6_addr32[3] = ifa->ifa_local;
3260
3261                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3262                                         continue;
3263                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3264                                         if (idev->dev->flags&IFF_POINTOPOINT)
3265                                                 continue;
3266                                         flag |= IFA_HOST;
3267                                 }
3268
3269                                 add_addr(idev, &addr, plen, flag,
3270                                          IFAPROT_UNSPEC);
3271                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3272                                                       0, pflags, GFP_KERNEL);
3273                         }
3274                 }
3275         }
3276 }
3277 #endif
3278
3279 static void init_loopback(struct net_device *dev)
3280 {
3281         struct inet6_dev  *idev;
3282
3283         /* ::1 */
3284
3285         ASSERT_RTNL();
3286
3287         idev = ipv6_find_idev(dev);
3288         if (IS_ERR(idev)) {
3289                 pr_debug("%s: add_dev failed\n", __func__);
3290                 return;
3291         }
3292
3293         add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3294 }
3295
3296 void addrconf_add_linklocal(struct inet6_dev *idev,
3297                             const struct in6_addr *addr, u32 flags)
3298 {
3299         struct ifa6_config cfg = {
3300                 .pfx = addr,
3301                 .plen = 64,
3302                 .ifa_flags = flags | IFA_F_PERMANENT,
3303                 .valid_lft = INFINITY_LIFE_TIME,
3304                 .preferred_lft = INFINITY_LIFE_TIME,
3305                 .scope = IFA_LINK,
3306                 .ifa_proto = IFAPROT_KERNEL_LL
3307         };
3308         struct inet6_ifaddr *ifp;
3309
3310 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3311         if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3312              READ_ONCE(idev->cnf.optimistic_dad)) &&
3313             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3314                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3315 #endif
3316
3317         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3318         if (!IS_ERR(ifp)) {
3319                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3320                                       0, 0, GFP_ATOMIC);
3321                 addrconf_dad_start(ifp);
3322                 in6_ifa_put(ifp);
3323         }
3324 }
3325 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3326
3327 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3328 {
3329         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3330                 return true;
3331
3332         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3333             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3334                 return true;
3335
3336         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3337             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3338                 return true;
3339
3340         return false;
3341 }
3342
3343 static int ipv6_generate_stable_address(struct in6_addr *address,
3344                                         u8 dad_count,
3345                                         const struct inet6_dev *idev)
3346 {
3347         static DEFINE_SPINLOCK(lock);
3348         static __u32 digest[SHA1_DIGEST_WORDS];
3349         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3350
3351         static union {
3352                 char __data[SHA1_BLOCK_SIZE];
3353                 struct {
3354                         struct in6_addr secret;
3355                         __be32 prefix[2];
3356                         unsigned char hwaddr[MAX_ADDR_LEN];
3357                         u8 dad_count;
3358                 } __packed;
3359         } data;
3360
3361         struct in6_addr secret;
3362         struct in6_addr temp;
3363         struct net *net = dev_net(idev->dev);
3364
3365         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3366
3367         if (idev->cnf.stable_secret.initialized)
3368                 secret = idev->cnf.stable_secret.secret;
3369         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3370                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3371         else
3372                 return -1;
3373
3374 retry:
3375         spin_lock_bh(&lock);
3376
3377         sha1_init(digest);
3378         memset(&data, 0, sizeof(data));
3379         memset(workspace, 0, sizeof(workspace));
3380         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3381         data.prefix[0] = address->s6_addr32[0];
3382         data.prefix[1] = address->s6_addr32[1];
3383         data.secret = secret;
3384         data.dad_count = dad_count;
3385
3386         sha1_transform(digest, data.__data, workspace);
3387
3388         temp = *address;
3389         temp.s6_addr32[2] = (__force __be32)digest[0];
3390         temp.s6_addr32[3] = (__force __be32)digest[1];
3391
3392         spin_unlock_bh(&lock);
3393
3394         if (ipv6_reserved_interfaceid(temp)) {
3395                 dad_count++;
3396                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3397                         return -1;
3398                 goto retry;
3399         }
3400
3401         *address = temp;
3402         return 0;
3403 }
3404
3405 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3406 {
3407         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3408
3409         if (s->initialized)
3410                 return;
3411         s = &idev->cnf.stable_secret;
3412         get_random_bytes(&s->secret, sizeof(s->secret));
3413         s->initialized = true;
3414 }
3415
3416 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3417 {
3418         struct in6_addr addr;
3419
3420         /* no link local addresses on L3 master devices */
3421         if (netif_is_l3_master(idev->dev))
3422                 return;
3423
3424         /* no link local addresses on devices flagged as slaves */
3425         if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3426                 return;
3427
3428         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3429
3430         switch (idev->cnf.addr_gen_mode) {
3431         case IN6_ADDR_GEN_MODE_RANDOM:
3432                 ipv6_gen_mode_random_init(idev);
3433                 fallthrough;
3434         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3435                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3436                         addrconf_add_linklocal(idev, &addr,
3437                                                IFA_F_STABLE_PRIVACY);
3438                 else if (prefix_route)
3439                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3440                                               0, 0, GFP_KERNEL);
3441                 break;
3442         case IN6_ADDR_GEN_MODE_EUI64:
3443                 /* addrconf_add_linklocal also adds a prefix_route and we
3444                  * only need to care about prefix routes if ipv6_generate_eui64
3445                  * couldn't generate one.
3446                  */
3447                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3448                         addrconf_add_linklocal(idev, &addr, 0);
3449                 else if (prefix_route)
3450                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3451                                               0, 0, GFP_KERNEL);
3452                 break;
3453         case IN6_ADDR_GEN_MODE_NONE:
3454         default:
3455                 /* will not add any link local address */
3456                 break;
3457         }
3458 }
3459
3460 static void addrconf_dev_config(struct net_device *dev)
3461 {
3462         struct inet6_dev *idev;
3463
3464         ASSERT_RTNL();
3465
3466         if ((dev->type != ARPHRD_ETHER) &&
3467             (dev->type != ARPHRD_FDDI) &&
3468             (dev->type != ARPHRD_ARCNET) &&
3469             (dev->type != ARPHRD_INFINIBAND) &&
3470             (dev->type != ARPHRD_IEEE1394) &&
3471             (dev->type != ARPHRD_TUNNEL6) &&
3472             (dev->type != ARPHRD_6LOWPAN) &&
3473             (dev->type != ARPHRD_TUNNEL) &&
3474             (dev->type != ARPHRD_NONE) &&
3475             (dev->type != ARPHRD_RAWIP)) {
3476                 /* Alas, we support only Ethernet autoconfiguration. */
3477                 idev = __in6_dev_get(dev);
3478                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3479                     dev->flags & IFF_MULTICAST)
3480                         ipv6_mc_up(idev);
3481                 return;
3482         }
3483
3484         idev = addrconf_add_dev(dev);
3485         if (IS_ERR(idev))
3486                 return;
3487
3488         /* this device type has no EUI support */
3489         if (dev->type == ARPHRD_NONE &&
3490             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3491                 WRITE_ONCE(idev->cnf.addr_gen_mode,
3492                            IN6_ADDR_GEN_MODE_RANDOM);
3493
3494         addrconf_addr_gen(idev, false);
3495 }
3496
3497 #if IS_ENABLED(CONFIG_IPV6_SIT)
3498 static void addrconf_sit_config(struct net_device *dev)
3499 {
3500         struct inet6_dev *idev;
3501
3502         ASSERT_RTNL();
3503
3504         /*
3505          * Configure the tunnel with one of our IPv4
3506          * addresses... we should configure all of
3507          * our v4 addrs in the tunnel
3508          */
3509
3510         idev = ipv6_find_idev(dev);
3511         if (IS_ERR(idev)) {
3512                 pr_debug("%s: add_dev failed\n", __func__);
3513                 return;
3514         }
3515
3516         if (dev->priv_flags & IFF_ISATAP) {
3517                 addrconf_addr_gen(idev, false);
3518                 return;
3519         }
3520
3521         add_v4_addrs(idev);
3522
3523         if (dev->flags&IFF_POINTOPOINT)
3524                 addrconf_add_mroute(dev);
3525 }
3526 #endif
3527
3528 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3529 static void addrconf_gre_config(struct net_device *dev)
3530 {
3531         struct inet6_dev *idev;
3532
3533         ASSERT_RTNL();
3534
3535         idev = ipv6_find_idev(dev);
3536         if (IS_ERR(idev)) {
3537                 pr_debug("%s: add_dev failed\n", __func__);
3538                 return;
3539         }
3540
3541         if (dev->type == ARPHRD_ETHER) {
3542                 addrconf_addr_gen(idev, true);
3543                 return;
3544         }
3545
3546         add_v4_addrs(idev);
3547
3548         if (dev->flags & IFF_POINTOPOINT)
3549                 addrconf_add_mroute(dev);
3550 }
3551 #endif
3552
3553 static void addrconf_init_auto_addrs(struct net_device *dev)
3554 {
3555         switch (dev->type) {
3556 #if IS_ENABLED(CONFIG_IPV6_SIT)
3557         case ARPHRD_SIT:
3558                 addrconf_sit_config(dev);
3559                 break;
3560 #endif
3561 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3562         case ARPHRD_IP6GRE:
3563         case ARPHRD_IPGRE:
3564                 addrconf_gre_config(dev);
3565                 break;
3566 #endif
3567         case ARPHRD_LOOPBACK:
3568                 init_loopback(dev);
3569                 break;
3570
3571         default:
3572                 addrconf_dev_config(dev);
3573                 break;
3574         }
3575 }
3576
3577 static int fixup_permanent_addr(struct net *net,
3578                                 struct inet6_dev *idev,
3579                                 struct inet6_ifaddr *ifp)
3580 {
3581         /* !fib6_node means the host route was removed from the
3582          * FIB, for example, if 'lo' device is taken down. In that
3583          * case regenerate the host route.
3584          */
3585         if (!ifp->rt || !ifp->rt->fib6_node) {
3586                 struct fib6_info *f6i, *prev;
3587
3588                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3589                                          GFP_ATOMIC, NULL);
3590                 if (IS_ERR(f6i))
3591                         return PTR_ERR(f6i);
3592
3593                 /* ifp->rt can be accessed outside of rtnl */
3594                 spin_lock(&ifp->lock);
3595                 prev = ifp->rt;
3596                 ifp->rt = f6i;
3597                 spin_unlock(&ifp->lock);
3598
3599                 fib6_info_release(prev);
3600         }
3601
3602         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3603                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3604                                       ifp->rt_priority, idev->dev, 0, 0,
3605                                       GFP_ATOMIC);
3606         }
3607
3608         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3609                 addrconf_dad_start(ifp);
3610
3611         return 0;
3612 }
3613
3614 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3615 {
3616         struct inet6_ifaddr *ifp, *tmp;
3617         struct inet6_dev *idev;
3618
3619         idev = __in6_dev_get(dev);
3620         if (!idev)
3621                 return;
3622
3623         write_lock_bh(&idev->lock);
3624
3625         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3626                 if ((ifp->flags & IFA_F_PERMANENT) &&
3627                     fixup_permanent_addr(net, idev, ifp) < 0) {
3628                         write_unlock_bh(&idev->lock);
3629                         in6_ifa_hold(ifp);
3630                         ipv6_del_addr(ifp);
3631                         write_lock_bh(&idev->lock);
3632
3633                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3634                                              idev->dev->name, &ifp->addr);
3635                 }
3636         }
3637
3638         write_unlock_bh(&idev->lock);
3639 }
3640
3641 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3642                            void *ptr)
3643 {
3644         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3645         struct netdev_notifier_change_info *change_info;
3646         struct netdev_notifier_changeupper_info *info;
3647         struct inet6_dev *idev = __in6_dev_get(dev);
3648         struct net *net = dev_net(dev);
3649         int run_pending = 0;
3650         int err;
3651
3652         switch (event) {
3653         case NETDEV_REGISTER:
3654                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3655                         idev = ipv6_add_dev(dev);
3656                         if (IS_ERR(idev))
3657                                 return notifier_from_errno(PTR_ERR(idev));
3658                 }
3659                 break;
3660
3661         case NETDEV_CHANGEMTU:
3662                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3663                 if (dev->mtu < IPV6_MIN_MTU) {
3664                         addrconf_ifdown(dev, dev != net->loopback_dev);
3665                         break;
3666                 }
3667
3668                 if (idev) {
3669                         rt6_mtu_change(dev, dev->mtu);
3670                         WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3671                         break;
3672                 }
3673
3674                 /* allocate new idev */
3675                 idev = ipv6_add_dev(dev);
3676                 if (IS_ERR(idev))
3677                         break;
3678
3679                 /* device is still not ready */
3680                 if (!(idev->if_flags & IF_READY))
3681                         break;
3682
3683                 run_pending = 1;
3684                 fallthrough;
3685         case NETDEV_UP:
3686         case NETDEV_CHANGE:
3687                 if (idev && idev->cnf.disable_ipv6)
3688                         break;
3689
3690                 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3691                         if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3692                             dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3693                                 ipv6_mc_up(idev);
3694                         break;
3695                 }
3696
3697                 if (event == NETDEV_UP) {
3698                         /* restore routes for permanent addresses */
3699                         addrconf_permanent_addr(net, dev);
3700
3701                         if (!addrconf_link_ready(dev)) {
3702                                 /* device is not ready yet. */
3703                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3704                                          dev->name);
3705                                 break;
3706                         }
3707
3708                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3709                                 idev = ipv6_add_dev(dev);
3710
3711                         if (!IS_ERR_OR_NULL(idev)) {
3712                                 idev->if_flags |= IF_READY;
3713                                 run_pending = 1;
3714                         }
3715                 } else if (event == NETDEV_CHANGE) {
3716                         if (!addrconf_link_ready(dev)) {
3717                                 /* device is still not ready. */
3718                                 rt6_sync_down_dev(dev, event);
3719                                 break;
3720                         }
3721
3722                         if (!IS_ERR_OR_NULL(idev)) {
3723                                 if (idev->if_flags & IF_READY) {
3724                                         /* device is already configured -
3725                                          * but resend MLD reports, we might
3726                                          * have roamed and need to update
3727                                          * multicast snooping switches
3728                                          */
3729                                         ipv6_mc_up(idev);
3730                                         change_info = ptr;
3731                                         if (change_info->flags_changed & IFF_NOARP)
3732                                                 addrconf_dad_run(idev, true);
3733                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3734                                         break;
3735                                 }
3736                                 idev->if_flags |= IF_READY;
3737                         }
3738
3739                         pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3740                                  dev->name);
3741
3742                         run_pending = 1;
3743                 }
3744
3745                 addrconf_init_auto_addrs(dev);
3746
3747                 if (!IS_ERR_OR_NULL(idev)) {
3748                         if (run_pending)
3749                                 addrconf_dad_run(idev, false);
3750
3751                         /* Device has an address by now */
3752                         rt6_sync_up(dev, RTNH_F_DEAD);
3753
3754                         /*
3755                          * If the MTU changed during the interface down,
3756                          * when the interface up, the changed MTU must be
3757                          * reflected in the idev as well as routers.
3758                          */
3759                         if (idev->cnf.mtu6 != dev->mtu &&
3760                             dev->mtu >= IPV6_MIN_MTU) {
3761                                 rt6_mtu_change(dev, dev->mtu);
3762                                 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3763                         }
3764                         WRITE_ONCE(idev->tstamp, jiffies);
3765                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3766
3767                         /*
3768                          * If the changed mtu during down is lower than
3769                          * IPV6_MIN_MTU stop IPv6 on this interface.
3770                          */
3771                         if (dev->mtu < IPV6_MIN_MTU)
3772                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3773                 }
3774                 break;
3775
3776         case NETDEV_DOWN:
3777         case NETDEV_UNREGISTER:
3778                 /*
3779                  *      Remove all addresses from this interface.
3780                  */
3781                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3782                 break;
3783
3784         case NETDEV_CHANGENAME:
3785                 if (idev) {
3786                         snmp6_unregister_dev(idev);
3787                         addrconf_sysctl_unregister(idev);
3788                         err = addrconf_sysctl_register(idev);
3789                         if (err)
3790                                 return notifier_from_errno(err);
3791                         err = snmp6_register_dev(idev);
3792                         if (err) {
3793                                 addrconf_sysctl_unregister(idev);
3794                                 return notifier_from_errno(err);
3795                         }
3796                 }
3797                 break;
3798
3799         case NETDEV_PRE_TYPE_CHANGE:
3800         case NETDEV_POST_TYPE_CHANGE:
3801                 if (idev)
3802                         addrconf_type_change(dev, event);
3803                 break;
3804
3805         case NETDEV_CHANGEUPPER:
3806                 info = ptr;
3807
3808                 /* flush all routes if dev is linked to or unlinked from
3809                  * an L3 master device (e.g., VRF)
3810                  */
3811                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3812                         addrconf_ifdown(dev, false);
3813         }
3814
3815         return NOTIFY_OK;
3816 }
3817
3818 /*
3819  *      addrconf module should be notified of a device going up
3820  */
3821 static struct notifier_block ipv6_dev_notf = {
3822         .notifier_call = addrconf_notify,
3823         .priority = ADDRCONF_NOTIFY_PRIORITY,
3824 };
3825
3826 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3827 {
3828         struct inet6_dev *idev;
3829         ASSERT_RTNL();
3830
3831         idev = __in6_dev_get(dev);
3832
3833         if (event == NETDEV_POST_TYPE_CHANGE)
3834                 ipv6_mc_remap(idev);
3835         else if (event == NETDEV_PRE_TYPE_CHANGE)
3836                 ipv6_mc_unmap(idev);
3837 }
3838
3839 static bool addr_is_local(const struct in6_addr *addr)
3840 {
3841         return ipv6_addr_type(addr) &
3842                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3843 }
3844
3845 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3846 {
3847         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3848         struct net *net = dev_net(dev);
3849         struct inet6_dev *idev;
3850         struct inet6_ifaddr *ifa;
3851         LIST_HEAD(tmp_addr_list);
3852         bool keep_addr = false;
3853         bool was_ready;
3854         int state, i;
3855
3856         ASSERT_RTNL();
3857
3858         rt6_disable_ip(dev, event);
3859
3860         idev = __in6_dev_get(dev);
3861         if (!idev)
3862                 return -ENODEV;
3863
3864         /*
3865          * Step 1: remove reference to ipv6 device from parent device.
3866          *         Do not dev_put!
3867          */
3868         if (unregister) {
3869                 idev->dead = 1;
3870
3871                 /* protected by rtnl_lock */
3872                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3873
3874                 /* Step 1.5: remove snmp6 entry */
3875                 snmp6_unregister_dev(idev);
3876
3877         }
3878
3879         /* combine the user config with event to determine if permanent
3880          * addresses are to be removed from address hash table
3881          */
3882         if (!unregister && !idev->cnf.disable_ipv6) {
3883                 /* aggregate the system setting and interface setting */
3884                 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3885
3886                 if (!_keep_addr)
3887                         _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3888
3889                 keep_addr = (_keep_addr > 0);
3890         }
3891
3892         /* Step 2: clear hash table */
3893         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3894                 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3895
3896                 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3897 restart:
3898                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3899                         if (ifa->idev == idev) {
3900                                 addrconf_del_dad_work(ifa);
3901                                 /* combined flag + permanent flag decide if
3902                                  * address is retained on a down event
3903                                  */
3904                                 if (!keep_addr ||
3905                                     !(ifa->flags & IFA_F_PERMANENT) ||
3906                                     addr_is_local(&ifa->addr)) {
3907                                         hlist_del_init_rcu(&ifa->addr_lst);
3908                                         goto restart;
3909                                 }
3910                         }
3911                 }
3912                 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3913         }
3914
3915         write_lock_bh(&idev->lock);
3916
3917         addrconf_del_rs_timer(idev);
3918
3919         /* Step 2: clear flags for stateless addrconf, repeated down
3920          *         detection
3921          */
3922         was_ready = idev->if_flags & IF_READY;
3923         if (!unregister)
3924                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3925
3926         /* Step 3: clear tempaddr list */
3927         while (!list_empty(&idev->tempaddr_list)) {
3928                 ifa = list_first_entry(&idev->tempaddr_list,
3929                                        struct inet6_ifaddr, tmp_list);
3930                 list_del(&ifa->tmp_list);
3931                 write_unlock_bh(&idev->lock);
3932                 spin_lock_bh(&ifa->lock);
3933
3934                 if (ifa->ifpub) {
3935                         in6_ifa_put(ifa->ifpub);
3936                         ifa->ifpub = NULL;
3937                 }
3938                 spin_unlock_bh(&ifa->lock);
3939                 in6_ifa_put(ifa);
3940                 write_lock_bh(&idev->lock);
3941         }
3942
3943         list_for_each_entry(ifa, &idev->addr_list, if_list)
3944                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3945         write_unlock_bh(&idev->lock);
3946
3947         while (!list_empty(&tmp_addr_list)) {
3948                 struct fib6_info *rt = NULL;
3949                 bool keep;
3950
3951                 ifa = list_first_entry(&tmp_addr_list,
3952                                        struct inet6_ifaddr, if_list_aux);
3953                 list_del(&ifa->if_list_aux);
3954
3955                 addrconf_del_dad_work(ifa);
3956
3957                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3958                         !addr_is_local(&ifa->addr);
3959
3960                 spin_lock_bh(&ifa->lock);
3961
3962                 if (keep) {
3963                         /* set state to skip the notifier below */
3964                         state = INET6_IFADDR_STATE_DEAD;
3965                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3966                         if (!(ifa->flags & IFA_F_NODAD))
3967                                 ifa->flags |= IFA_F_TENTATIVE;
3968
3969                         rt = ifa->rt;
3970                         ifa->rt = NULL;
3971                 } else {
3972                         state = ifa->state;
3973                         ifa->state = INET6_IFADDR_STATE_DEAD;
3974                 }
3975
3976                 spin_unlock_bh(&ifa->lock);
3977
3978                 if (rt)
3979                         ip6_del_rt(net, rt, false);
3980
3981                 if (state != INET6_IFADDR_STATE_DEAD) {
3982                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3983                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3984                 } else {
3985                         if (idev->cnf.forwarding)
3986                                 addrconf_leave_anycast(ifa);
3987                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3988                 }
3989
3990                 if (!keep) {
3991                         write_lock_bh(&idev->lock);
3992                         list_del_rcu(&ifa->if_list);
3993                         write_unlock_bh(&idev->lock);
3994                         in6_ifa_put(ifa);
3995                 }
3996         }
3997
3998         /* Step 5: Discard anycast and multicast list */
3999         if (unregister) {
4000                 ipv6_ac_destroy_dev(idev);
4001                 ipv6_mc_destroy_dev(idev);
4002         } else if (was_ready) {
4003                 ipv6_mc_down(idev);
4004         }
4005
4006         WRITE_ONCE(idev->tstamp, jiffies);
4007         idev->ra_mtu = 0;
4008
4009         /* Last: Shot the device (if unregistered) */
4010         if (unregister) {
4011                 addrconf_sysctl_unregister(idev);
4012                 neigh_parms_release(&nd_tbl, idev->nd_parms);
4013                 neigh_ifdown(&nd_tbl, dev);
4014                 in6_dev_put(idev);
4015         }
4016         return 0;
4017 }
4018
4019 static void addrconf_rs_timer(struct timer_list *t)
4020 {
4021         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
4022         struct net_device *dev = idev->dev;
4023         struct in6_addr lladdr;
4024         int rtr_solicits;
4025
4026         write_lock(&idev->lock);
4027         if (idev->dead || !(idev->if_flags & IF_READY))
4028                 goto out;
4029
4030         if (!ipv6_accept_ra(idev))
4031                 goto out;
4032
4033         /* Announcement received after solicitation was sent */
4034         if (idev->if_flags & IF_RA_RCVD)
4035                 goto out;
4036
4037         rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4038
4039         if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4040                 write_unlock(&idev->lock);
4041                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4042                         ndisc_send_rs(dev, &lladdr,
4043                                       &in6addr_linklocal_allrouters);
4044                 else
4045                         goto put;
4046
4047                 write_lock(&idev->lock);
4048                 idev->rs_interval = rfc3315_s14_backoff_update(
4049                                 idev->rs_interval,
4050                                 READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4051                 /* The wait after the last probe can be shorter */
4052                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4053                                              READ_ONCE(idev->cnf.rtr_solicits)) ?
4054                                       READ_ONCE(idev->cnf.rtr_solicit_delay) :
4055                                       idev->rs_interval);
4056         } else {
4057                 /*
4058                  * Note: we do not support deprecated "all on-link"
4059                  * assumption any longer.
4060                  */
4061                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4062         }
4063
4064 out:
4065         write_unlock(&idev->lock);
4066 put:
4067         in6_dev_put(idev);
4068 }
4069
4070 /*
4071  *      Duplicate Address Detection
4072  */
4073 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4074 {
4075         struct inet6_dev *idev = ifp->idev;
4076         unsigned long rand_num;
4077         u64 nonce;
4078
4079         if (ifp->flags & IFA_F_OPTIMISTIC)
4080                 rand_num = 0;
4081         else
4082                 rand_num = get_random_u32_below(
4083                                 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4084
4085         nonce = 0;
4086         if (READ_ONCE(idev->cnf.enhanced_dad) ||
4087             READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4088                 do
4089                         get_random_bytes(&nonce, 6);
4090                 while (nonce == 0);
4091         }
4092         ifp->dad_nonce = nonce;
4093         ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4094         addrconf_mod_dad_work(ifp, rand_num);
4095 }
4096
4097 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4098 {
4099         struct inet6_dev *idev = ifp->idev;
4100         struct net_device *dev = idev->dev;
4101         bool bump_id, notify = false;
4102         struct net *net;
4103
4104         addrconf_join_solict(dev, &ifp->addr);
4105
4106         read_lock_bh(&idev->lock);
4107         spin_lock(&ifp->lock);
4108         if (ifp->state == INET6_IFADDR_STATE_DEAD)
4109                 goto out;
4110
4111         net = dev_net(dev);
4112         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4113             (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4114              READ_ONCE(idev->cnf.accept_dad) < 1) ||
4115             !(ifp->flags&IFA_F_TENTATIVE) ||
4116             ifp->flags & IFA_F_NODAD) {
4117                 bool send_na = false;
4118
4119                 if (ifp->flags & IFA_F_TENTATIVE &&
4120                     !(ifp->flags & IFA_F_OPTIMISTIC))
4121                         send_na = true;
4122                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4123                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4124                 spin_unlock(&ifp->lock);
4125                 read_unlock_bh(&idev->lock);
4126
4127                 addrconf_dad_completed(ifp, bump_id, send_na);
4128                 return;
4129         }
4130
4131         if (!(idev->if_flags & IF_READY)) {
4132                 spin_unlock(&ifp->lock);
4133                 read_unlock_bh(&idev->lock);
4134                 /*
4135                  * If the device is not ready:
4136                  * - keep it tentative if it is a permanent address.
4137                  * - otherwise, kill it.
4138                  */
4139                 in6_ifa_hold(ifp);
4140                 addrconf_dad_stop(ifp, 0);
4141                 return;
4142         }
4143
4144         /*
4145          * Optimistic nodes can start receiving
4146          * Frames right away
4147          */
4148         if (ifp->flags & IFA_F_OPTIMISTIC) {
4149                 ip6_ins_rt(net, ifp->rt);
4150                 if (ipv6_use_optimistic_addr(net, idev)) {
4151                         /* Because optimistic nodes can use this address,
4152                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4153                          */
4154                         notify = true;
4155                 }
4156         }
4157
4158         addrconf_dad_kick(ifp);
4159 out:
4160         spin_unlock(&ifp->lock);
4161         read_unlock_bh(&idev->lock);
4162         if (notify)
4163                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4164 }
4165
4166 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4167 {
4168         bool begin_dad = false;
4169
4170         spin_lock_bh(&ifp->lock);
4171         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4172                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4173                 begin_dad = true;
4174         }
4175         spin_unlock_bh(&ifp->lock);
4176
4177         if (begin_dad)
4178                 addrconf_mod_dad_work(ifp, 0);
4179 }
4180
4181 static void addrconf_dad_work(struct work_struct *w)
4182 {
4183         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4184                                                 struct inet6_ifaddr,
4185                                                 dad_work);
4186         struct inet6_dev *idev = ifp->idev;
4187         bool bump_id, disable_ipv6 = false;
4188         struct in6_addr mcaddr;
4189
4190         enum {
4191                 DAD_PROCESS,
4192                 DAD_BEGIN,
4193                 DAD_ABORT,
4194         } action = DAD_PROCESS;
4195
4196         rtnl_lock();
4197
4198         spin_lock_bh(&ifp->lock);
4199         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4200                 action = DAD_BEGIN;
4201                 ifp->state = INET6_IFADDR_STATE_DAD;
4202         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4203                 action = DAD_ABORT;
4204                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4205
4206                 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->accept_dad) > 1 ||
4207                      READ_ONCE(idev->cnf.accept_dad) > 1) &&
4208                     !idev->cnf.disable_ipv6 &&
4209                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4210                         struct in6_addr addr;
4211
4212                         addr.s6_addr32[0] = htonl(0xfe800000);
4213                         addr.s6_addr32[1] = 0;
4214
4215                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4216                             ipv6_addr_equal(&ifp->addr, &addr)) {
4217                                 /* DAD failed for link-local based on MAC */
4218                                 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4219
4220                                 pr_info("%s: IPv6 being disabled!\n",
4221                                         ifp->idev->dev->name);
4222                                 disable_ipv6 = true;
4223                         }
4224                 }
4225         }
4226         spin_unlock_bh(&ifp->lock);
4227
4228         if (action == DAD_BEGIN) {
4229                 addrconf_dad_begin(ifp);
4230                 goto out;
4231         } else if (action == DAD_ABORT) {
4232                 in6_ifa_hold(ifp);
4233                 addrconf_dad_stop(ifp, 1);
4234                 if (disable_ipv6)
4235                         addrconf_ifdown(idev->dev, false);
4236                 goto out;
4237         }
4238
4239         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4240                 goto out;
4241
4242         write_lock_bh(&idev->lock);
4243         if (idev->dead || !(idev->if_flags & IF_READY)) {
4244                 write_unlock_bh(&idev->lock);
4245                 goto out;
4246         }
4247
4248         spin_lock(&ifp->lock);
4249         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4250                 spin_unlock(&ifp->lock);
4251                 write_unlock_bh(&idev->lock);
4252                 goto out;
4253         }
4254
4255         if (ifp->dad_probes == 0) {
4256                 bool send_na = false;
4257
4258                 /*
4259                  * DAD was successful
4260                  */
4261
4262                 if (ifp->flags & IFA_F_TENTATIVE &&
4263                     !(ifp->flags & IFA_F_OPTIMISTIC))
4264                         send_na = true;
4265                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4266                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4267                 spin_unlock(&ifp->lock);
4268                 write_unlock_bh(&idev->lock);
4269
4270                 addrconf_dad_completed(ifp, bump_id, send_na);
4271
4272                 goto out;
4273         }
4274
4275         ifp->dad_probes--;
4276         addrconf_mod_dad_work(ifp,
4277                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4278                                   HZ/100));
4279         spin_unlock(&ifp->lock);
4280         write_unlock_bh(&idev->lock);
4281
4282         /* send a neighbour solicitation for our addr */
4283         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4284         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4285                       ifp->dad_nonce);
4286 out:
4287         in6_ifa_put(ifp);
4288         rtnl_unlock();
4289 }
4290
4291 /* ifp->idev must be at least read locked */
4292 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4293 {
4294         struct inet6_ifaddr *ifpiter;
4295         struct inet6_dev *idev = ifp->idev;
4296
4297         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4298                 if (ifpiter->scope > IFA_LINK)
4299                         break;
4300                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4301                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4302                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4303                     IFA_F_PERMANENT)
4304                         return false;
4305         }
4306         return true;
4307 }
4308
4309 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4310                                    bool send_na)
4311 {
4312         struct net_device *dev = ifp->idev->dev;
4313         struct in6_addr lladdr;
4314         bool send_rs, send_mld;
4315
4316         addrconf_del_dad_work(ifp);
4317
4318         /*
4319          *      Configure the address for reception. Now it is valid.
4320          */
4321
4322         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4323
4324         /* If added prefix is link local and we are prepared to process
4325            router advertisements, start sending router solicitations.
4326          */
4327
4328         read_lock_bh(&ifp->idev->lock);
4329         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4330         send_rs = send_mld &&
4331                   ipv6_accept_ra(ifp->idev) &&
4332                   READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4333                   (dev->flags & IFF_LOOPBACK) == 0 &&
4334                   (dev->type != ARPHRD_TUNNEL) &&
4335                   !netif_is_team_port(dev);
4336         read_unlock_bh(&ifp->idev->lock);
4337
4338         /* While dad is in progress mld report's source address is in6_addrany.
4339          * Resend with proper ll now.
4340          */
4341         if (send_mld)
4342                 ipv6_mc_dad_complete(ifp->idev);
4343
4344         /* send unsolicited NA if enabled */
4345         if (send_na &&
4346             (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4347              READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4348                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4349                               /*router=*/ !!ifp->idev->cnf.forwarding,
4350                               /*solicited=*/ false, /*override=*/ true,
4351                               /*inc_opt=*/ true);
4352         }
4353
4354         if (send_rs) {
4355                 /*
4356                  *      If a host as already performed a random delay
4357                  *      [...] as part of DAD [...] there is no need
4358                  *      to delay again before sending the first RS
4359                  */
4360                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4361                         return;
4362                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4363
4364                 write_lock_bh(&ifp->idev->lock);
4365                 spin_lock(&ifp->lock);
4366                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4367                         READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4368                 ifp->idev->rs_probes = 1;
4369                 ifp->idev->if_flags |= IF_RS_SENT;
4370                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4371                 spin_unlock(&ifp->lock);
4372                 write_unlock_bh(&ifp->idev->lock);
4373         }
4374
4375         if (bump_id)
4376                 rt_genid_bump_ipv6(dev_net(dev));
4377
4378         /* Make sure that a new temporary address will be created
4379          * before this temporary address becomes deprecated.
4380          */
4381         if (ifp->flags & IFA_F_TEMPORARY)
4382                 addrconf_verify_rtnl(dev_net(dev));
4383 }
4384
4385 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4386 {
4387         struct inet6_ifaddr *ifp;
4388
4389         read_lock_bh(&idev->lock);
4390         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4391                 spin_lock(&ifp->lock);
4392                 if ((ifp->flags & IFA_F_TENTATIVE &&
4393                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4394                         if (restart)
4395                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4396                         addrconf_dad_kick(ifp);
4397                 }
4398                 spin_unlock(&ifp->lock);
4399         }
4400         read_unlock_bh(&idev->lock);
4401 }
4402
4403 #ifdef CONFIG_PROC_FS
4404 struct if6_iter_state {
4405         struct seq_net_private p;
4406         int bucket;
4407         int offset;
4408 };
4409
4410 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4411 {
4412         struct if6_iter_state *state = seq->private;
4413         struct net *net = seq_file_net(seq);
4414         struct inet6_ifaddr *ifa = NULL;
4415         int p = 0;
4416
4417         /* initial bucket if pos is 0 */
4418         if (pos == 0) {
4419                 state->bucket = 0;
4420                 state->offset = 0;
4421         }
4422
4423         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4424                 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4425                                          addr_lst) {
4426                         /* sync with offset */
4427                         if (p < state->offset) {
4428                                 p++;
4429                                 continue;
4430                         }
4431                         return ifa;
4432                 }
4433
4434                 /* prepare for next bucket */
4435                 state->offset = 0;
4436                 p = 0;
4437         }
4438         return NULL;
4439 }
4440
4441 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4442                                          struct inet6_ifaddr *ifa)
4443 {
4444         struct if6_iter_state *state = seq->private;
4445         struct net *net = seq_file_net(seq);
4446
4447         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4448                 state->offset++;
4449                 return ifa;
4450         }
4451
4452         state->offset = 0;
4453         while (++state->bucket < IN6_ADDR_HSIZE) {
4454                 hlist_for_each_entry_rcu(ifa,
4455                                      &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4456                         return ifa;
4457                 }
4458         }
4459
4460         return NULL;
4461 }
4462
4463 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4464         __acquires(rcu)
4465 {
4466         rcu_read_lock();
4467         return if6_get_first(seq, *pos);
4468 }
4469
4470 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4471 {
4472         struct inet6_ifaddr *ifa;
4473
4474         ifa = if6_get_next(seq, v);
4475         ++*pos;
4476         return ifa;
4477 }
4478
4479 static void if6_seq_stop(struct seq_file *seq, void *v)
4480         __releases(rcu)
4481 {
4482         rcu_read_unlock();
4483 }
4484
4485 static int if6_seq_show(struct seq_file *seq, void *v)
4486 {
4487         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4488         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4489                    &ifp->addr,
4490                    ifp->idev->dev->ifindex,
4491                    ifp->prefix_len,
4492                    ifp->scope,
4493                    (u8) ifp->flags,
4494                    ifp->idev->dev->name);
4495         return 0;
4496 }
4497
4498 static const struct seq_operations if6_seq_ops = {
4499         .start  = if6_seq_start,
4500         .next   = if6_seq_next,
4501         .show   = if6_seq_show,
4502         .stop   = if6_seq_stop,
4503 };
4504
4505 static int __net_init if6_proc_net_init(struct net *net)
4506 {
4507         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4508                         sizeof(struct if6_iter_state)))
4509                 return -ENOMEM;
4510         return 0;
4511 }
4512
4513 static void __net_exit if6_proc_net_exit(struct net *net)
4514 {
4515         remove_proc_entry("if_inet6", net->proc_net);
4516 }
4517
4518 static struct pernet_operations if6_proc_net_ops = {
4519         .init = if6_proc_net_init,
4520         .exit = if6_proc_net_exit,
4521 };
4522
4523 int __init if6_proc_init(void)
4524 {
4525         return register_pernet_subsys(&if6_proc_net_ops);
4526 }
4527
4528 void if6_proc_exit(void)
4529 {
4530         unregister_pernet_subsys(&if6_proc_net_ops);
4531 }
4532 #endif  /* CONFIG_PROC_FS */
4533
4534 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4535 /* Check if address is a home address configured on any interface. */
4536 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4537 {
4538         unsigned int hash = inet6_addr_hash(net, addr);
4539         struct inet6_ifaddr *ifp = NULL;
4540         int ret = 0;
4541
4542         rcu_read_lock();
4543         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4544                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4545                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4546                         ret = 1;
4547                         break;
4548                 }
4549         }
4550         rcu_read_unlock();
4551         return ret;
4552 }
4553 #endif
4554
4555 /* RFC6554 has some algorithm to avoid loops in segment routing by
4556  * checking if the segments contains any of a local interface address.
4557  *
4558  * Quote:
4559  *
4560  * To detect loops in the SRH, a router MUST determine if the SRH
4561  * includes multiple addresses assigned to any interface on that router.
4562  * If such addresses appear more than once and are separated by at least
4563  * one address not assigned to that router.
4564  */
4565 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4566                           unsigned char nsegs)
4567 {
4568         const struct in6_addr *addr;
4569         int i, ret = 0, found = 0;
4570         struct inet6_ifaddr *ifp;
4571         bool separated = false;
4572         unsigned int hash;
4573         bool hash_found;
4574
4575         rcu_read_lock();
4576         for (i = 0; i < nsegs; i++) {
4577                 addr = &segs[i];
4578                 hash = inet6_addr_hash(net, addr);
4579
4580                 hash_found = false;
4581                 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4582
4583                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4584                                 hash_found = true;
4585                                 break;
4586                         }
4587                 }
4588
4589                 if (hash_found) {
4590                         if (found > 1 && separated) {
4591                                 ret = 1;
4592                                 break;
4593                         }
4594
4595                         separated = false;
4596                         found++;
4597                 } else {
4598                         separated = true;
4599                 }
4600         }
4601         rcu_read_unlock();
4602
4603         return ret;
4604 }
4605
4606 /*
4607  *      Periodic address status verification
4608  */
4609
4610 static void addrconf_verify_rtnl(struct net *net)
4611 {
4612         unsigned long now, next, next_sec, next_sched;
4613         struct inet6_ifaddr *ifp;
4614         int i;
4615
4616         ASSERT_RTNL();
4617
4618         rcu_read_lock_bh();
4619         now = jiffies;
4620         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4621
4622         cancel_delayed_work(&net->ipv6.addr_chk_work);
4623
4624         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4625 restart:
4626                 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4627                         unsigned long age;
4628
4629                         /* When setting preferred_lft to a value not zero or
4630                          * infinity, while valid_lft is infinity
4631                          * IFA_F_PERMANENT has a non-infinity life time.
4632                          */
4633                         if ((ifp->flags & IFA_F_PERMANENT) &&
4634                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4635                                 continue;
4636
4637                         spin_lock(&ifp->lock);
4638                         /* We try to batch several events at once. */
4639                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4640
4641                         if ((ifp->flags&IFA_F_TEMPORARY) &&
4642                             !(ifp->flags&IFA_F_TENTATIVE) &&
4643                             ifp->prefered_lft != INFINITY_LIFE_TIME &&
4644                             !ifp->regen_count && ifp->ifpub) {
4645                                 /* This is a non-regenerated temporary addr. */
4646
4647                                 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4648
4649                                 if (age + regen_advance >= ifp->prefered_lft) {
4650                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4651                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4652                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4653
4654                                         ifp->regen_count++;
4655                                         in6_ifa_hold(ifp);
4656                                         in6_ifa_hold(ifpub);
4657                                         spin_unlock(&ifp->lock);
4658
4659                                         spin_lock(&ifpub->lock);
4660                                         ifpub->regen_count = 0;
4661                                         spin_unlock(&ifpub->lock);
4662                                         rcu_read_unlock_bh();
4663                                         ipv6_create_tempaddr(ifpub, true);
4664                                         in6_ifa_put(ifpub);
4665                                         in6_ifa_put(ifp);
4666                                         rcu_read_lock_bh();
4667                                         goto restart;
4668                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4669                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4670                         }
4671
4672                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4673                             age >= ifp->valid_lft) {
4674                                 spin_unlock(&ifp->lock);
4675                                 in6_ifa_hold(ifp);
4676                                 rcu_read_unlock_bh();
4677                                 ipv6_del_addr(ifp);
4678                                 rcu_read_lock_bh();
4679                                 goto restart;
4680                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4681                                 spin_unlock(&ifp->lock);
4682                                 continue;
4683                         } else if (age >= ifp->prefered_lft) {
4684                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4685                                 int deprecate = 0;
4686
4687                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4688                                         deprecate = 1;
4689                                         ifp->flags |= IFA_F_DEPRECATED;
4690                                 }
4691
4692                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4693                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4694                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4695
4696                                 spin_unlock(&ifp->lock);
4697
4698                                 if (deprecate) {
4699                                         in6_ifa_hold(ifp);
4700
4701                                         ipv6_ifa_notify(0, ifp);
4702                                         in6_ifa_put(ifp);
4703                                         goto restart;
4704                                 }
4705                         } else {
4706                                 /* ifp->prefered_lft <= ifp->valid_lft */
4707                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4708                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4709                                 spin_unlock(&ifp->lock);
4710                         }
4711                 }
4712         }
4713
4714         next_sec = round_jiffies_up(next);
4715         next_sched = next;
4716
4717         /* If rounded timeout is accurate enough, accept it. */
4718         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4719                 next_sched = next_sec;
4720
4721         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4722         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4723                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4724
4725         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4726                  now, next, next_sec, next_sched);
4727         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4728         rcu_read_unlock_bh();
4729 }
4730
4731 static void addrconf_verify_work(struct work_struct *w)
4732 {
4733         struct net *net = container_of(to_delayed_work(w), struct net,
4734                                        ipv6.addr_chk_work);
4735
4736         rtnl_lock();
4737         addrconf_verify_rtnl(net);
4738         rtnl_unlock();
4739 }
4740
4741 static void addrconf_verify(struct net *net)
4742 {
4743         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4744 }
4745
4746 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4747                                      struct in6_addr **peer_pfx)
4748 {
4749         struct in6_addr *pfx = NULL;
4750
4751         *peer_pfx = NULL;
4752
4753         if (addr)
4754                 pfx = nla_data(addr);
4755
4756         if (local) {
4757                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4758                         *peer_pfx = pfx;
4759                 pfx = nla_data(local);
4760         }
4761
4762         return pfx;
4763 }
4764
4765 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4766         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4767         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4768         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4769         [IFA_FLAGS]             = { .len = sizeof(u32) },
4770         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4771         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4772         [IFA_PROTO]             = { .type = NLA_U8 },
4773 };
4774
4775 static int
4776 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4777                   struct netlink_ext_ack *extack)
4778 {
4779         struct net *net = sock_net(skb->sk);
4780         struct ifaddrmsg *ifm;
4781         struct nlattr *tb[IFA_MAX+1];
4782         struct in6_addr *pfx, *peer_pfx;
4783         u32 ifa_flags;
4784         int err;
4785
4786         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4787                                      ifa_ipv6_policy, extack);
4788         if (err < 0)
4789                 return err;
4790
4791         ifm = nlmsg_data(nlh);
4792         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4793         if (!pfx)
4794                 return -EINVAL;
4795
4796         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4797
4798         /* We ignore other flags so far. */
4799         ifa_flags &= IFA_F_MANAGETEMPADDR;
4800
4801         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4802                               ifm->ifa_prefixlen, extack);
4803 }
4804
4805 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4806                                unsigned long expires, u32 flags,
4807                                bool modify_peer)
4808 {
4809         struct fib6_table *table;
4810         struct fib6_info *f6i;
4811         u32 prio;
4812
4813         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4814                                         ifp->prefix_len,
4815                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4816         if (!f6i)
4817                 return -ENOENT;
4818
4819         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4820         if (f6i->fib6_metric != prio) {
4821                 /* delete old one */
4822                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4823
4824                 /* add new one */
4825                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4826                                       ifp->prefix_len,
4827                                       ifp->rt_priority, ifp->idev->dev,
4828                                       expires, flags, GFP_KERNEL);
4829         } else {
4830                 table = f6i->fib6_table;
4831                 spin_lock_bh(&table->tb6_lock);
4832
4833                 if (!(flags & RTF_EXPIRES)) {
4834                         fib6_clean_expires(f6i);
4835                         fib6_remove_gc_list(f6i);
4836                 } else {
4837                         fib6_set_expires(f6i, expires);
4838                         fib6_add_gc_list(f6i);
4839                 }
4840
4841                 spin_unlock_bh(&table->tb6_lock);
4842
4843                 fib6_info_release(f6i);
4844         }
4845
4846         return 0;
4847 }
4848
4849 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4850                              struct ifa6_config *cfg)
4851 {
4852         u32 flags;
4853         clock_t expires;
4854         unsigned long timeout;
4855         bool was_managetempaddr;
4856         bool had_prefixroute;
4857         bool new_peer = false;
4858
4859         ASSERT_RTNL();
4860
4861         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4862                 return -EINVAL;
4863
4864         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4865             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4866                 return -EINVAL;
4867
4868         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4869                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4870
4871         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4872         if (addrconf_finite_timeout(timeout)) {
4873                 expires = jiffies_to_clock_t(timeout * HZ);
4874                 cfg->valid_lft = timeout;
4875                 flags = RTF_EXPIRES;
4876         } else {
4877                 expires = 0;
4878                 flags = 0;
4879                 cfg->ifa_flags |= IFA_F_PERMANENT;
4880         }
4881
4882         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4883         if (addrconf_finite_timeout(timeout)) {
4884                 if (timeout == 0)
4885                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4886                 cfg->preferred_lft = timeout;
4887         }
4888
4889         if (cfg->peer_pfx &&
4890             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4891                 if (!ipv6_addr_any(&ifp->peer_addr))
4892                         cleanup_prefix_route(ifp, expires, true, true);
4893                 new_peer = true;
4894         }
4895
4896         spin_lock_bh(&ifp->lock);
4897         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4898         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4899                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4900         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4901                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4902                         IFA_F_NOPREFIXROUTE);
4903         ifp->flags |= cfg->ifa_flags;
4904         WRITE_ONCE(ifp->tstamp, jiffies);
4905         WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4906         WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4907         WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4908
4909         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4910                 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4911
4912         if (new_peer)
4913                 ifp->peer_addr = *cfg->peer_pfx;
4914
4915         spin_unlock_bh(&ifp->lock);
4916         if (!(ifp->flags&IFA_F_TENTATIVE))
4917                 ipv6_ifa_notify(0, ifp);
4918
4919         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4920                 int rc = -ENOENT;
4921
4922                 if (had_prefixroute)
4923                         rc = modify_prefix_route(ifp, expires, flags, false);
4924
4925                 /* prefix route could have been deleted; if so restore it */
4926                 if (rc == -ENOENT) {
4927                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4928                                               ifp->rt_priority, ifp->idev->dev,
4929                                               expires, flags, GFP_KERNEL);
4930                 }
4931
4932                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4933                         rc = modify_prefix_route(ifp, expires, flags, true);
4934
4935                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4936                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4937                                               ifp->rt_priority, ifp->idev->dev,
4938                                               expires, flags, GFP_KERNEL);
4939                 }
4940         } else if (had_prefixroute) {
4941                 enum cleanup_prefix_rt_t action;
4942                 unsigned long rt_expires;
4943
4944                 write_lock_bh(&ifp->idev->lock);
4945                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4946                 write_unlock_bh(&ifp->idev->lock);
4947
4948                 if (action != CLEANUP_PREFIX_RT_NOP) {
4949                         cleanup_prefix_route(ifp, rt_expires,
4950                                 action == CLEANUP_PREFIX_RT_DEL, false);
4951                 }
4952         }
4953
4954         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4955                 if (was_managetempaddr &&
4956                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4957                         cfg->valid_lft = 0;
4958                         cfg->preferred_lft = 0;
4959                 }
4960                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4961                                  cfg->preferred_lft, !was_managetempaddr,
4962                                  jiffies);
4963         }
4964
4965         addrconf_verify_rtnl(net);
4966
4967         return 0;
4968 }
4969
4970 static int
4971 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4972                   struct netlink_ext_ack *extack)
4973 {
4974         struct net *net = sock_net(skb->sk);
4975         struct ifaddrmsg *ifm;
4976         struct nlattr *tb[IFA_MAX+1];
4977         struct in6_addr *peer_pfx;
4978         struct inet6_ifaddr *ifa;
4979         struct net_device *dev;
4980         struct inet6_dev *idev;
4981         struct ifa6_config cfg;
4982         int err;
4983
4984         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4985                                      ifa_ipv6_policy, extack);
4986         if (err < 0)
4987                 return err;
4988
4989         memset(&cfg, 0, sizeof(cfg));
4990
4991         ifm = nlmsg_data(nlh);
4992         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4993         if (!cfg.pfx)
4994                 return -EINVAL;
4995
4996         cfg.peer_pfx = peer_pfx;
4997         cfg.plen = ifm->ifa_prefixlen;
4998         if (tb[IFA_RT_PRIORITY])
4999                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
5000
5001         if (tb[IFA_PROTO])
5002                 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
5003
5004         cfg.valid_lft = INFINITY_LIFE_TIME;
5005         cfg.preferred_lft = INFINITY_LIFE_TIME;
5006
5007         if (tb[IFA_CACHEINFO]) {
5008                 struct ifa_cacheinfo *ci;
5009
5010                 ci = nla_data(tb[IFA_CACHEINFO]);
5011                 cfg.valid_lft = ci->ifa_valid;
5012                 cfg.preferred_lft = ci->ifa_prefered;
5013         }
5014
5015         dev =  __dev_get_by_index(net, ifm->ifa_index);
5016         if (!dev) {
5017                 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5018                 return -ENODEV;
5019         }
5020
5021         if (tb[IFA_FLAGS])
5022                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
5023         else
5024                 cfg.ifa_flags = ifm->ifa_flags;
5025
5026         /* We ignore other flags so far. */
5027         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
5028                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
5029                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
5030
5031         idev = ipv6_find_idev(dev);
5032         if (IS_ERR(idev))
5033                 return PTR_ERR(idev);
5034
5035         if (!ipv6_allow_optimistic_dad(net, idev))
5036                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5037
5038         if (cfg.ifa_flags & IFA_F_NODAD &&
5039             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5040                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5041                 return -EINVAL;
5042         }
5043
5044         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5045         if (!ifa) {
5046                 /*
5047                  * It would be best to check for !NLM_F_CREATE here but
5048                  * userspace already relies on not having to provide this.
5049                  */
5050                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
5051         }
5052
5053         if (nlh->nlmsg_flags & NLM_F_EXCL ||
5054             !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5055                 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5056                 err = -EEXIST;
5057         } else {
5058                 err = inet6_addr_modify(net, ifa, &cfg);
5059         }
5060
5061         in6_ifa_put(ifa);
5062
5063         return err;
5064 }
5065
5066 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5067                           u8 scope, int ifindex)
5068 {
5069         struct ifaddrmsg *ifm;
5070
5071         ifm = nlmsg_data(nlh);
5072         ifm->ifa_family = AF_INET6;
5073         ifm->ifa_prefixlen = prefixlen;
5074         ifm->ifa_flags = flags;
5075         ifm->ifa_scope = scope;
5076         ifm->ifa_index = ifindex;
5077 }
5078
5079 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5080                          unsigned long tstamp, u32 preferred, u32 valid)
5081 {
5082         struct ifa_cacheinfo ci;
5083
5084         ci.cstamp = cstamp_delta(cstamp);
5085         ci.tstamp = cstamp_delta(tstamp);
5086         ci.ifa_prefered = preferred;
5087         ci.ifa_valid = valid;
5088
5089         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5090 }
5091
5092 static inline int rt_scope(int ifa_scope)
5093 {
5094         if (ifa_scope & IFA_HOST)
5095                 return RT_SCOPE_HOST;
5096         else if (ifa_scope & IFA_LINK)
5097                 return RT_SCOPE_LINK;
5098         else if (ifa_scope & IFA_SITE)
5099                 return RT_SCOPE_SITE;
5100         else
5101                 return RT_SCOPE_UNIVERSE;
5102 }
5103
5104 static inline int inet6_ifaddr_msgsize(void)
5105 {
5106         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5107                + nla_total_size(16) /* IFA_LOCAL */
5108                + nla_total_size(16) /* IFA_ADDRESS */
5109                + nla_total_size(sizeof(struct ifa_cacheinfo))
5110                + nla_total_size(4)  /* IFA_FLAGS */
5111                + nla_total_size(1)  /* IFA_PROTO */
5112                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
5113 }
5114
5115 enum addr_type_t {
5116         UNICAST_ADDR,
5117         MULTICAST_ADDR,
5118         ANYCAST_ADDR,
5119 };
5120
5121 struct inet6_fill_args {
5122         u32 portid;
5123         u32 seq;
5124         int event;
5125         unsigned int flags;
5126         int netnsid;
5127         int ifindex;
5128         enum addr_type_t type;
5129 };
5130
5131 static int inet6_fill_ifaddr(struct sk_buff *skb,
5132                              const struct inet6_ifaddr *ifa,
5133                              struct inet6_fill_args *args)
5134 {
5135         struct nlmsghdr *nlh;
5136         u32 preferred, valid;
5137         u32 flags, priority;
5138         u8 proto;
5139
5140         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5141                         sizeof(struct ifaddrmsg), args->flags);
5142         if (!nlh)
5143                 return -EMSGSIZE;
5144
5145         flags = READ_ONCE(ifa->flags);
5146         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5147                       ifa->idev->dev->ifindex);
5148
5149         if (args->netnsid >= 0 &&
5150             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5151                 goto error;
5152
5153         preferred = READ_ONCE(ifa->prefered_lft);
5154         valid = READ_ONCE(ifa->valid_lft);
5155
5156         if (!((flags & IFA_F_PERMANENT) &&
5157               (preferred == INFINITY_LIFE_TIME))) {
5158                 if (preferred != INFINITY_LIFE_TIME) {
5159                         long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5160
5161                         if (preferred > tval)
5162                                 preferred -= tval;
5163                         else
5164                                 preferred = 0;
5165                         if (valid != INFINITY_LIFE_TIME) {
5166                                 if (valid > tval)
5167                                         valid -= tval;
5168                                 else
5169                                         valid = 0;
5170                         }
5171                 }
5172         } else {
5173                 preferred = INFINITY_LIFE_TIME;
5174                 valid = INFINITY_LIFE_TIME;
5175         }
5176
5177         if (!ipv6_addr_any(&ifa->peer_addr)) {
5178                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5179                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5180                         goto error;
5181         } else {
5182                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5183                         goto error;
5184         }
5185
5186         priority = READ_ONCE(ifa->rt_priority);
5187         if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5188                 goto error;
5189
5190         if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5191                           preferred, valid) < 0)
5192                 goto error;
5193
5194         if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5195                 goto error;
5196
5197         proto = READ_ONCE(ifa->ifa_proto);
5198         if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5199                 goto error;
5200
5201         nlmsg_end(skb, nlh);
5202         return 0;
5203
5204 error:
5205         nlmsg_cancel(skb, nlh);
5206         return -EMSGSIZE;
5207 }
5208
5209 static int inet6_fill_ifmcaddr(struct sk_buff *skb,
5210                                const struct ifmcaddr6 *ifmca,
5211                                struct inet6_fill_args *args)
5212 {
5213         int ifindex = ifmca->idev->dev->ifindex;
5214         u8 scope = RT_SCOPE_UNIVERSE;
5215         struct nlmsghdr *nlh;
5216
5217         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5218                 scope = RT_SCOPE_SITE;
5219
5220         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5221                         sizeof(struct ifaddrmsg), args->flags);
5222         if (!nlh)
5223                 return -EMSGSIZE;
5224
5225         if (args->netnsid >= 0 &&
5226             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5227                 nlmsg_cancel(skb, nlh);
5228                 return -EMSGSIZE;
5229         }
5230
5231         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5232         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5233             put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5234                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5235                 nlmsg_cancel(skb, nlh);
5236                 return -EMSGSIZE;
5237         }
5238
5239         nlmsg_end(skb, nlh);
5240         return 0;
5241 }
5242
5243 static int inet6_fill_ifacaddr(struct sk_buff *skb,
5244                                const struct ifacaddr6 *ifaca,
5245                                struct inet6_fill_args *args)
5246 {
5247         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5248         int ifindex = dev ? dev->ifindex : 1;
5249         u8 scope = RT_SCOPE_UNIVERSE;
5250         struct nlmsghdr *nlh;
5251
5252         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5253                 scope = RT_SCOPE_SITE;
5254
5255         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5256                         sizeof(struct ifaddrmsg), args->flags);
5257         if (!nlh)
5258                 return -EMSGSIZE;
5259
5260         if (args->netnsid >= 0 &&
5261             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5262                 nlmsg_cancel(skb, nlh);
5263                 return -EMSGSIZE;
5264         }
5265
5266         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5267         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5268             put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5269                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5270                 nlmsg_cancel(skb, nlh);
5271                 return -EMSGSIZE;
5272         }
5273
5274         nlmsg_end(skb, nlh);
5275         return 0;
5276 }
5277
5278 /* called with rcu_read_lock() */
5279 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5280                           struct netlink_callback *cb, int *s_ip_idx,
5281                           struct inet6_fill_args *fillargs)
5282 {
5283         const struct ifmcaddr6 *ifmca;
5284         const struct ifacaddr6 *ifaca;
5285         int ip_idx = 0;
5286         int err = 0;
5287
5288         switch (fillargs->type) {
5289         case UNICAST_ADDR: {
5290                 const struct inet6_ifaddr *ifa;
5291                 fillargs->event = RTM_NEWADDR;
5292
5293                 /* unicast address incl. temp addr */
5294                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5295                         if (ip_idx < *s_ip_idx)
5296                                 goto next;
5297                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5298                         if (err < 0)
5299                                 break;
5300                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5301 next:
5302                         ip_idx++;
5303                 }
5304                 break;
5305         }
5306         case MULTICAST_ADDR:
5307                 fillargs->event = RTM_GETMULTICAST;
5308
5309                 /* multicast address */
5310                 for (ifmca = rcu_dereference(idev->mc_list);
5311                      ifmca;
5312                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5313                         if (ip_idx < *s_ip_idx)
5314                                 continue;
5315                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5316                         if (err < 0)
5317                                 break;
5318                 }
5319                 break;
5320         case ANYCAST_ADDR:
5321                 fillargs->event = RTM_GETANYCAST;
5322                 /* anycast address */
5323                 for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5324                      ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5325                         if (ip_idx < *s_ip_idx)
5326                                 continue;
5327                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5328                         if (err < 0)
5329                                 break;
5330                 }
5331                 break;
5332         default:
5333                 break;
5334         }
5335         *s_ip_idx = err ? ip_idx : 0;
5336         return err;
5337 }
5338
5339 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5340                                        struct inet6_fill_args *fillargs,
5341                                        struct net **tgt_net, struct sock *sk,
5342                                        struct netlink_callback *cb)
5343 {
5344         struct netlink_ext_ack *extack = cb->extack;
5345         struct nlattr *tb[IFA_MAX+1];
5346         struct ifaddrmsg *ifm;
5347         int err, i;
5348
5349         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5350                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5351                 return -EINVAL;
5352         }
5353
5354         ifm = nlmsg_data(nlh);
5355         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5356                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5357                 return -EINVAL;
5358         }
5359
5360         fillargs->ifindex = ifm->ifa_index;
5361         if (fillargs->ifindex) {
5362                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5363                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5364         }
5365
5366         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5367                                             ifa_ipv6_policy, extack);
5368         if (err < 0)
5369                 return err;
5370
5371         for (i = 0; i <= IFA_MAX; ++i) {
5372                 if (!tb[i])
5373                         continue;
5374
5375                 if (i == IFA_TARGET_NETNSID) {
5376                         struct net *net;
5377
5378                         fillargs->netnsid = nla_get_s32(tb[i]);
5379                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5380                         if (IS_ERR(net)) {
5381                                 fillargs->netnsid = -1;
5382                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5383                                 return PTR_ERR(net);
5384                         }
5385                         *tgt_net = net;
5386                 } else {
5387                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5388                         return -EINVAL;
5389                 }
5390         }
5391
5392         return 0;
5393 }
5394
5395 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5396                            enum addr_type_t type)
5397 {
5398         struct net *tgt_net = sock_net(skb->sk);
5399         const struct nlmsghdr *nlh = cb->nlh;
5400         struct inet6_fill_args fillargs = {
5401                 .portid = NETLINK_CB(cb->skb).portid,
5402                 .seq = cb->nlh->nlmsg_seq,
5403                 .flags = NLM_F_MULTI,
5404                 .netnsid = -1,
5405                 .type = type,
5406         };
5407         struct {
5408                 unsigned long ifindex;
5409                 int ip_idx;
5410         } *ctx = (void *)cb->ctx;
5411         struct net_device *dev;
5412         struct inet6_dev *idev;
5413         int err = 0;
5414
5415         rcu_read_lock();
5416         if (cb->strict_check) {
5417                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5418                                                   skb->sk, cb);
5419                 if (err < 0)
5420                         goto done;
5421
5422                 err = 0;
5423                 if (fillargs.ifindex) {
5424                         dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5425                         if (!dev) {
5426                                 err = -ENODEV;
5427                                 goto done;
5428                         }
5429                         idev = __in6_dev_get(dev);
5430                         if (idev)
5431                                 err = in6_dump_addrs(idev, skb, cb,
5432                                                      &ctx->ip_idx,
5433                                                      &fillargs);
5434                         goto done;
5435                 }
5436         }
5437
5438         cb->seq = inet6_base_seq(tgt_net);
5439         for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5440                 idev = __in6_dev_get(dev);
5441                 if (!idev)
5442                         continue;
5443                 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5444                                      &fillargs);
5445                 if (err < 0)
5446                         goto done;
5447         }
5448 done:
5449         rcu_read_unlock();
5450         if (fillargs.netnsid >= 0)
5451                 put_net(tgt_net);
5452
5453         return err;
5454 }
5455
5456 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5457 {
5458         enum addr_type_t type = UNICAST_ADDR;
5459
5460         return inet6_dump_addr(skb, cb, type);
5461 }
5462
5463 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5464 {
5465         enum addr_type_t type = MULTICAST_ADDR;
5466
5467         return inet6_dump_addr(skb, cb, type);
5468 }
5469
5470
5471 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5472 {
5473         enum addr_type_t type = ANYCAST_ADDR;
5474
5475         return inet6_dump_addr(skb, cb, type);
5476 }
5477
5478 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5479                                        const struct nlmsghdr *nlh,
5480                                        struct nlattr **tb,
5481                                        struct netlink_ext_ack *extack)
5482 {
5483         struct ifaddrmsg *ifm;
5484         int i, err;
5485
5486         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5487                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5488                 return -EINVAL;
5489         }
5490
5491         if (!netlink_strict_get_check(skb))
5492                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5493                                               ifa_ipv6_policy, extack);
5494
5495         ifm = nlmsg_data(nlh);
5496         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5497                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5498                 return -EINVAL;
5499         }
5500
5501         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5502                                             ifa_ipv6_policy, extack);
5503         if (err)
5504                 return err;
5505
5506         for (i = 0; i <= IFA_MAX; i++) {
5507                 if (!tb[i])
5508                         continue;
5509
5510                 switch (i) {
5511                 case IFA_TARGET_NETNSID:
5512                 case IFA_ADDRESS:
5513                 case IFA_LOCAL:
5514                         break;
5515                 default:
5516                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5517                         return -EINVAL;
5518                 }
5519         }
5520
5521         return 0;
5522 }
5523
5524 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5525                              struct netlink_ext_ack *extack)
5526 {
5527         struct net *tgt_net = sock_net(in_skb->sk);
5528         struct inet6_fill_args fillargs = {
5529                 .portid = NETLINK_CB(in_skb).portid,
5530                 .seq = nlh->nlmsg_seq,
5531                 .event = RTM_NEWADDR,
5532                 .flags = 0,
5533                 .netnsid = -1,
5534         };
5535         struct ifaddrmsg *ifm;
5536         struct nlattr *tb[IFA_MAX+1];
5537         struct in6_addr *addr = NULL, *peer;
5538         struct net_device *dev = NULL;
5539         struct inet6_ifaddr *ifa;
5540         struct sk_buff *skb;
5541         int err;
5542
5543         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5544         if (err < 0)
5545                 return err;
5546
5547         if (tb[IFA_TARGET_NETNSID]) {
5548                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5549
5550                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5551                                                   fillargs.netnsid);
5552                 if (IS_ERR(tgt_net))
5553                         return PTR_ERR(tgt_net);
5554         }
5555
5556         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5557         if (!addr) {
5558                 err = -EINVAL;
5559                 goto errout;
5560         }
5561         ifm = nlmsg_data(nlh);
5562         if (ifm->ifa_index)
5563                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5564
5565         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5566         if (!ifa) {
5567                 err = -EADDRNOTAVAIL;
5568                 goto errout;
5569         }
5570
5571         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5572         if (!skb) {
5573                 err = -ENOBUFS;
5574                 goto errout_ifa;
5575         }
5576
5577         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5578         if (err < 0) {
5579                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5580                 WARN_ON(err == -EMSGSIZE);
5581                 kfree_skb(skb);
5582                 goto errout_ifa;
5583         }
5584         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5585 errout_ifa:
5586         in6_ifa_put(ifa);
5587 errout:
5588         dev_put(dev);
5589         if (fillargs.netnsid >= 0)
5590                 put_net(tgt_net);
5591
5592         return err;
5593 }
5594
5595 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5596 {
5597         struct sk_buff *skb;
5598         struct net *net = dev_net(ifa->idev->dev);
5599         struct inet6_fill_args fillargs = {
5600                 .portid = 0,
5601                 .seq = 0,
5602                 .event = event,
5603                 .flags = 0,
5604                 .netnsid = -1,
5605         };
5606         int err = -ENOBUFS;
5607
5608         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5609         if (!skb)
5610                 goto errout;
5611
5612         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5613         if (err < 0) {
5614                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5615                 WARN_ON(err == -EMSGSIZE);
5616                 kfree_skb(skb);
5617                 goto errout;
5618         }
5619         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5620         return;
5621 errout:
5622         rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5623 }
5624
5625 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5626                                __s32 *array, int bytes)
5627 {
5628         BUG_ON(bytes < (DEVCONF_MAX * 4));
5629
5630         memset(array, 0, bytes);
5631         array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5632         array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5633         array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5634         array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5635         array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5636         array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5637         array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5638         array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5639         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5640                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5641         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5642                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5643         array[DEVCONF_RTR_SOLICIT_DELAY] =
5644                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5645         array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5646         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5647                 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5648         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5649                 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5650         array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5651         array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5652         array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5653         array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5654         array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5655         array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5656         array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5657         array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5658         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5659                 READ_ONCE(cnf->accept_ra_min_hop_limit);
5660         array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5661 #ifdef CONFIG_IPV6_ROUTER_PREF
5662         array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5663         array[DEVCONF_RTR_PROBE_INTERVAL] =
5664                 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5665 #ifdef CONFIG_IPV6_ROUTE_INFO
5666         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5667                 READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5668         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5669                 READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5670 #endif
5671 #endif
5672         array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5673         array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5674                 READ_ONCE(cnf->accept_source_route);
5675 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5676         array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5677         array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5678 #endif
5679 #ifdef CONFIG_IPV6_MROUTE
5680         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5681 #endif
5682         array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5683         array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5684         array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5685         array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5686         array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5687                 READ_ONCE(cnf->suppress_frag_ndisc);
5688         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5689                 READ_ONCE(cnf->accept_ra_from_local);
5690         array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5691         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5692                 READ_ONCE(cnf->ignore_routes_with_linkdown);
5693         /* we omit DEVCONF_STABLE_SECRET for now */
5694         array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5695         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5696                 READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5697         array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5698         array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5699         array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5700 #ifdef CONFIG_IPV6_SEG6_HMAC
5701         array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5702 #endif
5703         array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5704         array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5705         array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5706         array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5707         array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5708         array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5709         array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5710         array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5711         array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5712                 READ_ONCE(cnf->ndisc_evict_nocarrier);
5713         array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5714                 READ_ONCE(cnf->accept_untracked_na);
5715         array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5716 }
5717
5718 static inline size_t inet6_ifla6_size(void)
5719 {
5720         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5721              + nla_total_size(sizeof(struct ifla_cacheinfo))
5722              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5723              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5724              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5725              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5726              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5727              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5728              + 0;
5729 }
5730
5731 static inline size_t inet6_if_nlmsg_size(void)
5732 {
5733         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5734                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5735                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5736                + nla_total_size(4) /* IFLA_MTU */
5737                + nla_total_size(4) /* IFLA_LINK */
5738                + nla_total_size(1) /* IFLA_OPERSTATE */
5739                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5740 }
5741
5742 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5743                                         int bytes)
5744 {
5745         int i;
5746         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5747         BUG_ON(pad < 0);
5748
5749         /* Use put_unaligned() because stats may not be aligned for u64. */
5750         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5751         for (i = 1; i < ICMP6_MIB_MAX; i++)
5752                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5753
5754         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5755 }
5756
5757 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5758                                         int bytes, size_t syncpoff)
5759 {
5760         int i, c;
5761         u64 buff[IPSTATS_MIB_MAX];
5762         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5763
5764         BUG_ON(pad < 0);
5765
5766         memset(buff, 0, sizeof(buff));
5767         buff[0] = IPSTATS_MIB_MAX;
5768
5769         for_each_possible_cpu(c) {
5770                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5771                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5772         }
5773
5774         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5775         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5776 }
5777
5778 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5779                              int bytes)
5780 {
5781         switch (attrtype) {
5782         case IFLA_INET6_STATS:
5783                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5784                                      offsetof(struct ipstats_mib, syncp));
5785                 break;
5786         case IFLA_INET6_ICMP6STATS:
5787                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5788                 break;
5789         }
5790 }
5791
5792 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5793                                   u32 ext_filter_mask)
5794 {
5795         struct ifla_cacheinfo ci;
5796         struct nlattr *nla;
5797         u32 ra_mtu;
5798
5799         if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5800                 goto nla_put_failure;
5801         ci.max_reasm_len = IPV6_MAXPLEN;
5802         ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5803         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5804         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5805         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5806                 goto nla_put_failure;
5807         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5808         if (!nla)
5809                 goto nla_put_failure;
5810         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5811
5812         /* XXX - MC not implemented */
5813
5814         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5815                 return 0;
5816
5817         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5818         if (!nla)
5819                 goto nla_put_failure;
5820         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5821
5822         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5823         if (!nla)
5824                 goto nla_put_failure;
5825         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5826
5827         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5828         if (!nla)
5829                 goto nla_put_failure;
5830         read_lock_bh(&idev->lock);
5831         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5832         read_unlock_bh(&idev->lock);
5833
5834         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5835                        READ_ONCE(idev->cnf.addr_gen_mode)))
5836                 goto nla_put_failure;
5837
5838         ra_mtu = READ_ONCE(idev->ra_mtu);
5839         if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5840                 goto nla_put_failure;
5841
5842         return 0;
5843
5844 nla_put_failure:
5845         return -EMSGSIZE;
5846 }
5847
5848 static size_t inet6_get_link_af_size(const struct net_device *dev,
5849                                      u32 ext_filter_mask)
5850 {
5851         if (!__in6_dev_get(dev))
5852                 return 0;
5853
5854         return inet6_ifla6_size();
5855 }
5856
5857 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5858                               u32 ext_filter_mask)
5859 {
5860         struct inet6_dev *idev = __in6_dev_get(dev);
5861
5862         if (!idev)
5863                 return -ENODATA;
5864
5865         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5866                 return -EMSGSIZE;
5867
5868         return 0;
5869 }
5870
5871 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5872                              struct netlink_ext_ack *extack)
5873 {
5874         struct inet6_ifaddr *ifp;
5875         struct net_device *dev = idev->dev;
5876         bool clear_token, update_rs = false;
5877         struct in6_addr ll_addr;
5878
5879         ASSERT_RTNL();
5880
5881         if (!token)
5882                 return -EINVAL;
5883
5884         if (dev->flags & IFF_LOOPBACK) {
5885                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5886                 return -EINVAL;
5887         }
5888
5889         if (dev->flags & IFF_NOARP) {
5890                 NL_SET_ERR_MSG_MOD(extack,
5891                                    "Device does not do neighbour discovery");
5892                 return -EINVAL;
5893         }
5894
5895         if (!ipv6_accept_ra(idev)) {
5896                 NL_SET_ERR_MSG_MOD(extack,
5897                                    "Router advertisement is disabled on device");
5898                 return -EINVAL;
5899         }
5900
5901         if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5902                 NL_SET_ERR_MSG(extack,
5903                                "Router solicitation is disabled on device");
5904                 return -EINVAL;
5905         }
5906
5907         write_lock_bh(&idev->lock);
5908
5909         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5910         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5911
5912         write_unlock_bh(&idev->lock);
5913
5914         clear_token = ipv6_addr_any(token);
5915         if (clear_token)
5916                 goto update_lft;
5917
5918         if (!idev->dead && (idev->if_flags & IF_READY) &&
5919             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5920                              IFA_F_OPTIMISTIC)) {
5921                 /* If we're not ready, then normal ifup will take care
5922                  * of this. Otherwise, we need to request our rs here.
5923                  */
5924                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5925                 update_rs = true;
5926         }
5927
5928 update_lft:
5929         write_lock_bh(&idev->lock);
5930
5931         if (update_rs) {
5932                 idev->if_flags |= IF_RS_SENT;
5933                 idev->rs_interval = rfc3315_s14_backoff_init(
5934                         READ_ONCE(idev->cnf.rtr_solicit_interval));
5935                 idev->rs_probes = 1;
5936                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5937         }
5938
5939         /* Well, that's kinda nasty ... */
5940         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5941                 spin_lock(&ifp->lock);
5942                 if (ifp->tokenized) {
5943                         ifp->valid_lft = 0;
5944                         ifp->prefered_lft = 0;
5945                 }
5946                 spin_unlock(&ifp->lock);
5947         }
5948
5949         write_unlock_bh(&idev->lock);
5950         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5951         addrconf_verify_rtnl(dev_net(dev));
5952         return 0;
5953 }
5954
5955 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5956         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5957         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5958         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5959                                             .reject_message =
5960                                                 "IFLA_INET6_RA_MTU can not be set" },
5961 };
5962
5963 static int check_addr_gen_mode(int mode)
5964 {
5965         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5966             mode != IN6_ADDR_GEN_MODE_NONE &&
5967             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5968             mode != IN6_ADDR_GEN_MODE_RANDOM)
5969                 return -EINVAL;
5970         return 1;
5971 }
5972
5973 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5974                                 int mode)
5975 {
5976         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5977             !idev->cnf.stable_secret.initialized &&
5978             !net->ipv6.devconf_dflt->stable_secret.initialized)
5979                 return -EINVAL;
5980         return 1;
5981 }
5982
5983 static int inet6_validate_link_af(const struct net_device *dev,
5984                                   const struct nlattr *nla,
5985                                   struct netlink_ext_ack *extack)
5986 {
5987         struct nlattr *tb[IFLA_INET6_MAX + 1];
5988         struct inet6_dev *idev = NULL;
5989         int err;
5990
5991         if (dev) {
5992                 idev = __in6_dev_get(dev);
5993                 if (!idev)
5994                         return -EAFNOSUPPORT;
5995         }
5996
5997         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5998                                           inet6_af_policy, extack);
5999         if (err)
6000                 return err;
6001
6002         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6003                 return -EINVAL;
6004
6005         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6006                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6007
6008                 if (check_addr_gen_mode(mode) < 0)
6009                         return -EINVAL;
6010                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6011                         return -EINVAL;
6012         }
6013
6014         return 0;
6015 }
6016
6017 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6018                              struct netlink_ext_ack *extack)
6019 {
6020         struct inet6_dev *idev = __in6_dev_get(dev);
6021         struct nlattr *tb[IFLA_INET6_MAX + 1];
6022         int err;
6023
6024         if (!idev)
6025                 return -EAFNOSUPPORT;
6026
6027         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6028                 return -EINVAL;
6029
6030         if (tb[IFLA_INET6_TOKEN]) {
6031                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6032                                         extack);
6033                 if (err)
6034                         return err;
6035         }
6036
6037         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6038                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6039
6040                 WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6041         }
6042
6043         return 0;
6044 }
6045
6046 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6047                              u32 portid, u32 seq, int event, unsigned int flags)
6048 {
6049         struct net_device *dev = idev->dev;
6050         struct ifinfomsg *hdr;
6051         struct nlmsghdr *nlh;
6052         int ifindex, iflink;
6053         void *protoinfo;
6054
6055         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6056         if (!nlh)
6057                 return -EMSGSIZE;
6058
6059         hdr = nlmsg_data(nlh);
6060         hdr->ifi_family = AF_INET6;
6061         hdr->__ifi_pad = 0;
6062         hdr->ifi_type = dev->type;
6063         ifindex = READ_ONCE(dev->ifindex);
6064         hdr->ifi_index = ifindex;
6065         hdr->ifi_flags = dev_get_flags(dev);
6066         hdr->ifi_change = 0;
6067
6068         iflink = dev_get_iflink(dev);
6069         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6070             (dev->addr_len &&
6071              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6072             nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6073             (ifindex != iflink &&
6074              nla_put_u32(skb, IFLA_LINK, iflink)) ||
6075             nla_put_u8(skb, IFLA_OPERSTATE,
6076                        netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6077                 goto nla_put_failure;
6078         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6079         if (!protoinfo)
6080                 goto nla_put_failure;
6081
6082         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6083                 goto nla_put_failure;
6084
6085         nla_nest_end(skb, protoinfo);
6086         nlmsg_end(skb, nlh);
6087         return 0;
6088
6089 nla_put_failure:
6090         nlmsg_cancel(skb, nlh);
6091         return -EMSGSIZE;
6092 }
6093
6094 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6095                                    struct netlink_ext_ack *extack)
6096 {
6097         struct ifinfomsg *ifm;
6098
6099         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6100                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6101                 return -EINVAL;
6102         }
6103
6104         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6105                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6106                 return -EINVAL;
6107         }
6108
6109         ifm = nlmsg_data(nlh);
6110         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6111             ifm->ifi_change || ifm->ifi_index) {
6112                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6113                 return -EINVAL;
6114         }
6115
6116         return 0;
6117 }
6118
6119 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6120 {
6121         struct net *net = sock_net(skb->sk);
6122         struct {
6123                 unsigned long ifindex;
6124         } *ctx = (void *)cb->ctx;
6125         struct net_device *dev;
6126         struct inet6_dev *idev;
6127         int err;
6128
6129         /* only requests using strict checking can pass data to
6130          * influence the dump
6131          */
6132         if (cb->strict_check) {
6133                 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6134
6135                 if (err < 0)
6136                         return err;
6137         }
6138
6139         err = 0;
6140         rcu_read_lock();
6141         for_each_netdev_dump(net, dev, ctx->ifindex) {
6142                 idev = __in6_dev_get(dev);
6143                 if (!idev)
6144                         continue;
6145                 err = inet6_fill_ifinfo(skb, idev,
6146                                         NETLINK_CB(cb->skb).portid,
6147                                         cb->nlh->nlmsg_seq,
6148                                         RTM_NEWLINK, NLM_F_MULTI);
6149                 if (err < 0)
6150                         break;
6151         }
6152         rcu_read_unlock();
6153
6154         return err;
6155 }
6156
6157 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6158 {
6159         struct sk_buff *skb;
6160         struct net *net = dev_net(idev->dev);
6161         int err = -ENOBUFS;
6162
6163         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6164         if (!skb)
6165                 goto errout;
6166
6167         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6168         if (err < 0) {
6169                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6170                 WARN_ON(err == -EMSGSIZE);
6171                 kfree_skb(skb);
6172                 goto errout;
6173         }
6174         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6175         return;
6176 errout:
6177         rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6178 }
6179
6180 static inline size_t inet6_prefix_nlmsg_size(void)
6181 {
6182         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6183                + nla_total_size(sizeof(struct in6_addr))
6184                + nla_total_size(sizeof(struct prefix_cacheinfo));
6185 }
6186
6187 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6188                              struct prefix_info *pinfo, u32 portid, u32 seq,
6189                              int event, unsigned int flags)
6190 {
6191         struct prefixmsg *pmsg;
6192         struct nlmsghdr *nlh;
6193         struct prefix_cacheinfo ci;
6194
6195         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6196         if (!nlh)
6197                 return -EMSGSIZE;
6198
6199         pmsg = nlmsg_data(nlh);
6200         pmsg->prefix_family = AF_INET6;
6201         pmsg->prefix_pad1 = 0;
6202         pmsg->prefix_pad2 = 0;
6203         pmsg->prefix_ifindex = idev->dev->ifindex;
6204         pmsg->prefix_len = pinfo->prefix_len;
6205         pmsg->prefix_type = pinfo->type;
6206         pmsg->prefix_pad3 = 0;
6207         pmsg->prefix_flags = pinfo->flags;
6208
6209         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6210                 goto nla_put_failure;
6211         ci.preferred_time = ntohl(pinfo->prefered);
6212         ci.valid_time = ntohl(pinfo->valid);
6213         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6214                 goto nla_put_failure;
6215         nlmsg_end(skb, nlh);
6216         return 0;
6217
6218 nla_put_failure:
6219         nlmsg_cancel(skb, nlh);
6220         return -EMSGSIZE;
6221 }
6222
6223 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6224                          struct prefix_info *pinfo)
6225 {
6226         struct sk_buff *skb;
6227         struct net *net = dev_net(idev->dev);
6228         int err = -ENOBUFS;
6229
6230         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6231         if (!skb)
6232                 goto errout;
6233
6234         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6235         if (err < 0) {
6236                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6237                 WARN_ON(err == -EMSGSIZE);
6238                 kfree_skb(skb);
6239                 goto errout;
6240         }
6241         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6242         return;
6243 errout:
6244         rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6245 }
6246
6247 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6248 {
6249         struct net *net = dev_net(ifp->idev->dev);
6250
6251         if (event)
6252                 ASSERT_RTNL();
6253
6254         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6255
6256         switch (event) {
6257         case RTM_NEWADDR:
6258                 /*
6259                  * If the address was optimistic we inserted the route at the
6260                  * start of our DAD process, so we don't need to do it again.
6261                  * If the device was taken down in the middle of the DAD
6262                  * cycle there is a race where we could get here without a
6263                  * host route, so nothing to insert. That will be fixed when
6264                  * the device is brought up.
6265                  */
6266                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6267                         ip6_ins_rt(net, ifp->rt);
6268                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6269                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6270                                 &ifp->addr, ifp->idev->dev->name);
6271                 }
6272
6273                 if (ifp->idev->cnf.forwarding)
6274                         addrconf_join_anycast(ifp);
6275                 if (!ipv6_addr_any(&ifp->peer_addr))
6276                         addrconf_prefix_route(&ifp->peer_addr, 128,
6277                                               ifp->rt_priority, ifp->idev->dev,
6278                                               0, 0, GFP_ATOMIC);
6279                 break;
6280         case RTM_DELADDR:
6281                 if (ifp->idev->cnf.forwarding)
6282                         addrconf_leave_anycast(ifp);
6283                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6284                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6285                         struct fib6_info *rt;
6286
6287                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6288                                                        ifp->idev->dev, 0, 0,
6289                                                        false);
6290                         if (rt)
6291                                 ip6_del_rt(net, rt, false);
6292                 }
6293                 if (ifp->rt) {
6294                         ip6_del_rt(net, ifp->rt, false);
6295                         ifp->rt = NULL;
6296                 }
6297                 rt_genid_bump_ipv6(net);
6298                 break;
6299         }
6300         atomic_inc(&net->ipv6.dev_addr_genid);
6301 }
6302
6303 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6304 {
6305         if (likely(ifp->idev->dead == 0))
6306                 __ipv6_ifa_notify(event, ifp);
6307 }
6308
6309 #ifdef CONFIG_SYSCTL
6310
6311 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6312                 void *buffer, size_t *lenp, loff_t *ppos)
6313 {
6314         int *valp = ctl->data;
6315         int val = *valp;
6316         loff_t pos = *ppos;
6317         struct ctl_table lctl;
6318         int ret;
6319
6320         /*
6321          * ctl->data points to idev->cnf.forwarding, we should
6322          * not modify it until we get the rtnl lock.
6323          */
6324         lctl = *ctl;
6325         lctl.data = &val;
6326
6327         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6328
6329         if (write)
6330                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6331         if (ret)
6332                 *ppos = pos;
6333         return ret;
6334 }
6335
6336 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6337                 void *buffer, size_t *lenp, loff_t *ppos)
6338 {
6339         struct inet6_dev *idev = ctl->extra1;
6340         int min_mtu = IPV6_MIN_MTU;
6341         struct ctl_table lctl;
6342
6343         lctl = *ctl;
6344         lctl.extra1 = &min_mtu;
6345         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6346
6347         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6348 }
6349
6350 static void dev_disable_change(struct inet6_dev *idev)
6351 {
6352         struct netdev_notifier_info info;
6353
6354         if (!idev || !idev->dev)
6355                 return;
6356
6357         netdev_notifier_info_init(&info, idev->dev);
6358         if (idev->cnf.disable_ipv6)
6359                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6360         else
6361                 addrconf_notify(NULL, NETDEV_UP, &info);
6362 }
6363
6364 static void addrconf_disable_change(struct net *net, __s32 newf)
6365 {
6366         struct net_device *dev;
6367         struct inet6_dev *idev;
6368
6369         for_each_netdev(net, dev) {
6370                 idev = __in6_dev_get(dev);
6371                 if (idev) {
6372                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6373
6374                         WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6375                         if (changed)
6376                                 dev_disable_change(idev);
6377                 }
6378         }
6379 }
6380
6381 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6382 {
6383         struct net *net = (struct net *)table->extra2;
6384         int old;
6385
6386         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6387                 WRITE_ONCE(*p, newf);
6388                 return 0;
6389         }
6390
6391         if (!rtnl_trylock())
6392                 return restart_syscall();
6393
6394         old = *p;
6395         WRITE_ONCE(*p, newf);
6396
6397         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6398                 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6399                 addrconf_disable_change(net, newf);
6400         } else if ((!newf) ^ (!old))
6401                 dev_disable_change((struct inet6_dev *)table->extra1);
6402
6403         rtnl_unlock();
6404         return 0;
6405 }
6406
6407 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6408                 void *buffer, size_t *lenp, loff_t *ppos)
6409 {
6410         int *valp = ctl->data;
6411         int val = *valp;
6412         loff_t pos = *ppos;
6413         struct ctl_table lctl;
6414         int ret;
6415
6416         /*
6417          * ctl->data points to idev->cnf.disable_ipv6, we should
6418          * not modify it until we get the rtnl lock.
6419          */
6420         lctl = *ctl;
6421         lctl.data = &val;
6422
6423         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6424
6425         if (write)
6426                 ret = addrconf_disable_ipv6(ctl, valp, val);
6427         if (ret)
6428                 *ppos = pos;
6429         return ret;
6430 }
6431
6432 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6433                 void *buffer, size_t *lenp, loff_t *ppos)
6434 {
6435         int *valp = ctl->data;
6436         int ret;
6437         int old, new;
6438
6439         old = *valp;
6440         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6441         new = *valp;
6442
6443         if (write && old != new) {
6444                 struct net *net = ctl->extra2;
6445
6446                 if (!rtnl_trylock())
6447                         return restart_syscall();
6448
6449                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6450                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6451                                                      NETCONFA_PROXY_NEIGH,
6452                                                      NETCONFA_IFINDEX_DEFAULT,
6453                                                      net->ipv6.devconf_dflt);
6454                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6455                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6456                                                      NETCONFA_PROXY_NEIGH,
6457                                                      NETCONFA_IFINDEX_ALL,
6458                                                      net->ipv6.devconf_all);
6459                 else {
6460                         struct inet6_dev *idev = ctl->extra1;
6461
6462                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6463                                                      NETCONFA_PROXY_NEIGH,
6464                                                      idev->dev->ifindex,
6465                                                      &idev->cnf);
6466                 }
6467                 rtnl_unlock();
6468         }
6469
6470         return ret;
6471 }
6472
6473 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6474                                          void *buffer, size_t *lenp,
6475                                          loff_t *ppos)
6476 {
6477         int ret = 0;
6478         u32 new_val;
6479         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6480         struct net *net = (struct net *)ctl->extra2;
6481         struct ctl_table tmp = {
6482                 .data = &new_val,
6483                 .maxlen = sizeof(new_val),
6484                 .mode = ctl->mode,
6485         };
6486
6487         if (!rtnl_trylock())
6488                 return restart_syscall();
6489
6490         new_val = *((u32 *)ctl->data);
6491
6492         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6493         if (ret != 0)
6494                 goto out;
6495
6496         if (write) {
6497                 if (check_addr_gen_mode(new_val) < 0) {
6498                         ret = -EINVAL;
6499                         goto out;
6500                 }
6501
6502                 if (idev) {
6503                         if (check_stable_privacy(idev, net, new_val) < 0) {
6504                                 ret = -EINVAL;
6505                                 goto out;
6506                         }
6507
6508                         if (idev->cnf.addr_gen_mode != new_val) {
6509                                 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6510                                 addrconf_init_auto_addrs(idev->dev);
6511                         }
6512                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6513                         struct net_device *dev;
6514
6515                         WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6516                         for_each_netdev(net, dev) {
6517                                 idev = __in6_dev_get(dev);
6518                                 if (idev &&
6519                                     idev->cnf.addr_gen_mode != new_val) {
6520                                         WRITE_ONCE(idev->cnf.addr_gen_mode,
6521                                                   new_val);
6522                                         addrconf_init_auto_addrs(idev->dev);
6523                                 }
6524                         }
6525                 }
6526
6527                 WRITE_ONCE(*((u32 *)ctl->data), new_val);
6528         }
6529
6530 out:
6531         rtnl_unlock();
6532
6533         return ret;
6534 }
6535
6536 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6537                                          void *buffer, size_t *lenp,
6538                                          loff_t *ppos)
6539 {
6540         int err;
6541         struct in6_addr addr;
6542         char str[IPV6_MAX_STRLEN];
6543         struct ctl_table lctl = *ctl;
6544         struct net *net = ctl->extra2;
6545         struct ipv6_stable_secret *secret = ctl->data;
6546
6547         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6548                 return -EIO;
6549
6550         lctl.maxlen = IPV6_MAX_STRLEN;
6551         lctl.data = str;
6552
6553         if (!rtnl_trylock())
6554                 return restart_syscall();
6555
6556         if (!write && !secret->initialized) {
6557                 err = -EIO;
6558                 goto out;
6559         }
6560
6561         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6562         if (err >= sizeof(str)) {
6563                 err = -EIO;
6564                 goto out;
6565         }
6566
6567         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6568         if (err || !write)
6569                 goto out;
6570
6571         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6572                 err = -EIO;
6573                 goto out;
6574         }
6575
6576         secret->initialized = true;
6577         secret->secret = addr;
6578
6579         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6580                 struct net_device *dev;
6581
6582                 for_each_netdev(net, dev) {
6583                         struct inet6_dev *idev = __in6_dev_get(dev);
6584
6585                         if (idev) {
6586                                 WRITE_ONCE(idev->cnf.addr_gen_mode,
6587                                            IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6588                         }
6589                 }
6590         } else {
6591                 struct inet6_dev *idev = ctl->extra1;
6592
6593                 WRITE_ONCE(idev->cnf.addr_gen_mode,
6594                            IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6595         }
6596
6597 out:
6598         rtnl_unlock();
6599
6600         return err;
6601 }
6602
6603 static
6604 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6605                                                 int write, void *buffer,
6606                                                 size_t *lenp,
6607                                                 loff_t *ppos)
6608 {
6609         int *valp = ctl->data;
6610         int val = *valp;
6611         loff_t pos = *ppos;
6612         struct ctl_table lctl;
6613         int ret;
6614
6615         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6616          * we should not modify it until we get the rtnl lock.
6617          */
6618         lctl = *ctl;
6619         lctl.data = &val;
6620
6621         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6622
6623         if (write)
6624                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6625         if (ret)
6626                 *ppos = pos;
6627         return ret;
6628 }
6629
6630 static
6631 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6632 {
6633         if (rt) {
6634                 if (action)
6635                         rt->dst.flags |= DST_NOPOLICY;
6636                 else
6637                         rt->dst.flags &= ~DST_NOPOLICY;
6638         }
6639 }
6640
6641 static
6642 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6643 {
6644         struct inet6_ifaddr *ifa;
6645
6646         read_lock_bh(&idev->lock);
6647         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6648                 spin_lock(&ifa->lock);
6649                 if (ifa->rt) {
6650                         /* host routes only use builtin fib6_nh */
6651                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6652                         int cpu;
6653
6654                         rcu_read_lock();
6655                         ifa->rt->dst_nopolicy = val ? true : false;
6656                         if (nh->rt6i_pcpu) {
6657                                 for_each_possible_cpu(cpu) {
6658                                         struct rt6_info **rtp;
6659
6660                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6661                                         addrconf_set_nopolicy(*rtp, val);
6662                                 }
6663                         }
6664                         rcu_read_unlock();
6665                 }
6666                 spin_unlock(&ifa->lock);
6667         }
6668         read_unlock_bh(&idev->lock);
6669 }
6670
6671 static
6672 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6673 {
6674         struct net *net = (struct net *)ctl->extra2;
6675         struct inet6_dev *idev;
6676
6677         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6678                 WRITE_ONCE(*valp, val);
6679                 return 0;
6680         }
6681
6682         if (!rtnl_trylock())
6683                 return restart_syscall();
6684
6685         WRITE_ONCE(*valp, val);
6686
6687         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6688                 struct net_device *dev;
6689
6690                 for_each_netdev(net, dev) {
6691                         idev = __in6_dev_get(dev);
6692                         if (idev)
6693                                 addrconf_disable_policy_idev(idev, val);
6694                 }
6695         } else {
6696                 idev = (struct inet6_dev *)ctl->extra1;
6697                 addrconf_disable_policy_idev(idev, val);
6698         }
6699
6700         rtnl_unlock();
6701         return 0;
6702 }
6703
6704 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6705                                    void *buffer, size_t *lenp, loff_t *ppos)
6706 {
6707         int *valp = ctl->data;
6708         int val = *valp;
6709         loff_t pos = *ppos;
6710         struct ctl_table lctl;
6711         int ret;
6712
6713         lctl = *ctl;
6714         lctl.data = &val;
6715         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6716
6717         if (write && (*valp != val))
6718                 ret = addrconf_disable_policy(ctl, valp, val);
6719
6720         if (ret)
6721                 *ppos = pos;
6722
6723         return ret;
6724 }
6725
6726 static int minus_one = -1;
6727 static const int two_five_five = 255;
6728 static u32 ioam6_if_id_max = U16_MAX;
6729
6730 static const struct ctl_table addrconf_sysctl[] = {
6731         {
6732                 .procname       = "forwarding",
6733                 .data           = &ipv6_devconf.forwarding,
6734                 .maxlen         = sizeof(int),
6735                 .mode           = 0644,
6736                 .proc_handler   = addrconf_sysctl_forward,
6737         },
6738         {
6739                 .procname       = "hop_limit",
6740                 .data           = &ipv6_devconf.hop_limit,
6741                 .maxlen         = sizeof(int),
6742                 .mode           = 0644,
6743                 .proc_handler   = proc_dointvec_minmax,
6744                 .extra1         = (void *)SYSCTL_ONE,
6745                 .extra2         = (void *)&two_five_five,
6746         },
6747         {
6748                 .procname       = "mtu",
6749                 .data           = &ipv6_devconf.mtu6,
6750                 .maxlen         = sizeof(int),
6751                 .mode           = 0644,
6752                 .proc_handler   = addrconf_sysctl_mtu,
6753         },
6754         {
6755                 .procname       = "accept_ra",
6756                 .data           = &ipv6_devconf.accept_ra,
6757                 .maxlen         = sizeof(int),
6758                 .mode           = 0644,
6759                 .proc_handler   = proc_dointvec,
6760         },
6761         {
6762                 .procname       = "accept_redirects",
6763                 .data           = &ipv6_devconf.accept_redirects,
6764                 .maxlen         = sizeof(int),
6765                 .mode           = 0644,
6766                 .proc_handler   = proc_dointvec,
6767         },
6768         {
6769                 .procname       = "autoconf",
6770                 .data           = &ipv6_devconf.autoconf,
6771                 .maxlen         = sizeof(int),
6772                 .mode           = 0644,
6773                 .proc_handler   = proc_dointvec,
6774         },
6775         {
6776                 .procname       = "dad_transmits",
6777                 .data           = &ipv6_devconf.dad_transmits,
6778                 .maxlen         = sizeof(int),
6779                 .mode           = 0644,
6780                 .proc_handler   = proc_dointvec,
6781         },
6782         {
6783                 .procname       = "router_solicitations",
6784                 .data           = &ipv6_devconf.rtr_solicits,
6785                 .maxlen         = sizeof(int),
6786                 .mode           = 0644,
6787                 .proc_handler   = proc_dointvec_minmax,
6788                 .extra1         = &minus_one,
6789         },
6790         {
6791                 .procname       = "router_solicitation_interval",
6792                 .data           = &ipv6_devconf.rtr_solicit_interval,
6793                 .maxlen         = sizeof(int),
6794                 .mode           = 0644,
6795                 .proc_handler   = proc_dointvec_jiffies,
6796         },
6797         {
6798                 .procname       = "router_solicitation_max_interval",
6799                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6800                 .maxlen         = sizeof(int),
6801                 .mode           = 0644,
6802                 .proc_handler   = proc_dointvec_jiffies,
6803         },
6804         {
6805                 .procname       = "router_solicitation_delay",
6806                 .data           = &ipv6_devconf.rtr_solicit_delay,
6807                 .maxlen         = sizeof(int),
6808                 .mode           = 0644,
6809                 .proc_handler   = proc_dointvec_jiffies,
6810         },
6811         {
6812                 .procname       = "force_mld_version",
6813                 .data           = &ipv6_devconf.force_mld_version,
6814                 .maxlen         = sizeof(int),
6815                 .mode           = 0644,
6816                 .proc_handler   = proc_dointvec,
6817         },
6818         {
6819                 .procname       = "mldv1_unsolicited_report_interval",
6820                 .data           =
6821                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6822                 .maxlen         = sizeof(int),
6823                 .mode           = 0644,
6824                 .proc_handler   = proc_dointvec_ms_jiffies,
6825         },
6826         {
6827                 .procname       = "mldv2_unsolicited_report_interval",
6828                 .data           =
6829                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6830                 .maxlen         = sizeof(int),
6831                 .mode           = 0644,
6832                 .proc_handler   = proc_dointvec_ms_jiffies,
6833         },
6834         {
6835                 .procname       = "use_tempaddr",
6836                 .data           = &ipv6_devconf.use_tempaddr,
6837                 .maxlen         = sizeof(int),
6838                 .mode           = 0644,
6839                 .proc_handler   = proc_dointvec,
6840         },
6841         {
6842                 .procname       = "temp_valid_lft",
6843                 .data           = &ipv6_devconf.temp_valid_lft,
6844                 .maxlen         = sizeof(int),
6845                 .mode           = 0644,
6846                 .proc_handler   = proc_dointvec,
6847         },
6848         {
6849                 .procname       = "temp_prefered_lft",
6850                 .data           = &ipv6_devconf.temp_prefered_lft,
6851                 .maxlen         = sizeof(int),
6852                 .mode           = 0644,
6853                 .proc_handler   = proc_dointvec,
6854         },
6855         {
6856                 .procname       = "regen_min_advance",
6857                 .data           = &ipv6_devconf.regen_min_advance,
6858                 .maxlen         = sizeof(int),
6859                 .mode           = 0644,
6860                 .proc_handler   = proc_dointvec,
6861         },
6862         {
6863                 .procname       = "regen_max_retry",
6864                 .data           = &ipv6_devconf.regen_max_retry,
6865                 .maxlen         = sizeof(int),
6866                 .mode           = 0644,
6867                 .proc_handler   = proc_dointvec,
6868         },
6869         {
6870                 .procname       = "max_desync_factor",
6871                 .data           = &ipv6_devconf.max_desync_factor,
6872                 .maxlen         = sizeof(int),
6873                 .mode           = 0644,
6874                 .proc_handler   = proc_dointvec,
6875         },
6876         {
6877                 .procname       = "max_addresses",
6878                 .data           = &ipv6_devconf.max_addresses,
6879                 .maxlen         = sizeof(int),
6880                 .mode           = 0644,
6881                 .proc_handler   = proc_dointvec,
6882         },
6883         {
6884                 .procname       = "accept_ra_defrtr",
6885                 .data           = &ipv6_devconf.accept_ra_defrtr,
6886                 .maxlen         = sizeof(int),
6887                 .mode           = 0644,
6888                 .proc_handler   = proc_dointvec,
6889         },
6890         {
6891                 .procname       = "ra_defrtr_metric",
6892                 .data           = &ipv6_devconf.ra_defrtr_metric,
6893                 .maxlen         = sizeof(u32),
6894                 .mode           = 0644,
6895                 .proc_handler   = proc_douintvec_minmax,
6896                 .extra1         = (void *)SYSCTL_ONE,
6897         },
6898         {
6899                 .procname       = "accept_ra_min_hop_limit",
6900                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6901                 .maxlen         = sizeof(int),
6902                 .mode           = 0644,
6903                 .proc_handler   = proc_dointvec,
6904         },
6905         {
6906                 .procname       = "accept_ra_min_lft",
6907                 .data           = &ipv6_devconf.accept_ra_min_lft,
6908                 .maxlen         = sizeof(int),
6909                 .mode           = 0644,
6910                 .proc_handler   = proc_dointvec,
6911         },
6912         {
6913                 .procname       = "accept_ra_pinfo",
6914                 .data           = &ipv6_devconf.accept_ra_pinfo,
6915                 .maxlen         = sizeof(int),
6916                 .mode           = 0644,
6917                 .proc_handler   = proc_dointvec,
6918         },
6919         {
6920                 .procname       = "ra_honor_pio_life",
6921                 .data           = &ipv6_devconf.ra_honor_pio_life,
6922                 .maxlen         = sizeof(u8),
6923                 .mode           = 0644,
6924                 .proc_handler   = proc_dou8vec_minmax,
6925                 .extra1         = SYSCTL_ZERO,
6926                 .extra2         = SYSCTL_ONE,
6927         },
6928         {
6929                 .procname       = "ra_honor_pio_pflag",
6930                 .data           = &ipv6_devconf.ra_honor_pio_pflag,
6931                 .maxlen         = sizeof(u8),
6932                 .mode           = 0644,
6933                 .proc_handler   = proc_dou8vec_minmax,
6934                 .extra1         = SYSCTL_ZERO,
6935                 .extra2         = SYSCTL_ONE,
6936         },
6937 #ifdef CONFIG_IPV6_ROUTER_PREF
6938         {
6939                 .procname       = "accept_ra_rtr_pref",
6940                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6941                 .maxlen         = sizeof(int),
6942                 .mode           = 0644,
6943                 .proc_handler   = proc_dointvec,
6944         },
6945         {
6946                 .procname       = "router_probe_interval",
6947                 .data           = &ipv6_devconf.rtr_probe_interval,
6948                 .maxlen         = sizeof(int),
6949                 .mode           = 0644,
6950                 .proc_handler   = proc_dointvec_jiffies,
6951         },
6952 #ifdef CONFIG_IPV6_ROUTE_INFO
6953         {
6954                 .procname       = "accept_ra_rt_info_min_plen",
6955                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6956                 .maxlen         = sizeof(int),
6957                 .mode           = 0644,
6958                 .proc_handler   = proc_dointvec,
6959         },
6960         {
6961                 .procname       = "accept_ra_rt_info_max_plen",
6962                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6963                 .maxlen         = sizeof(int),
6964                 .mode           = 0644,
6965                 .proc_handler   = proc_dointvec,
6966         },
6967 #endif
6968 #endif
6969         {
6970                 .procname       = "proxy_ndp",
6971                 .data           = &ipv6_devconf.proxy_ndp,
6972                 .maxlen         = sizeof(int),
6973                 .mode           = 0644,
6974                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6975         },
6976         {
6977                 .procname       = "accept_source_route",
6978                 .data           = &ipv6_devconf.accept_source_route,
6979                 .maxlen         = sizeof(int),
6980                 .mode           = 0644,
6981                 .proc_handler   = proc_dointvec,
6982         },
6983 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6984         {
6985                 .procname       = "optimistic_dad",
6986                 .data           = &ipv6_devconf.optimistic_dad,
6987                 .maxlen         = sizeof(int),
6988                 .mode           = 0644,
6989                 .proc_handler   = proc_dointvec,
6990         },
6991         {
6992                 .procname       = "use_optimistic",
6993                 .data           = &ipv6_devconf.use_optimistic,
6994                 .maxlen         = sizeof(int),
6995                 .mode           = 0644,
6996                 .proc_handler   = proc_dointvec,
6997         },
6998 #endif
6999 #ifdef CONFIG_IPV6_MROUTE
7000         {
7001                 .procname       = "mc_forwarding",
7002                 .data           = &ipv6_devconf.mc_forwarding,
7003                 .maxlen         = sizeof(int),
7004                 .mode           = 0444,
7005                 .proc_handler   = proc_dointvec,
7006         },
7007 #endif
7008         {
7009                 .procname       = "disable_ipv6",
7010                 .data           = &ipv6_devconf.disable_ipv6,
7011                 .maxlen         = sizeof(int),
7012                 .mode           = 0644,
7013                 .proc_handler   = addrconf_sysctl_disable,
7014         },
7015         {
7016                 .procname       = "accept_dad",
7017                 .data           = &ipv6_devconf.accept_dad,
7018                 .maxlen         = sizeof(int),
7019                 .mode           = 0644,
7020                 .proc_handler   = proc_dointvec,
7021         },
7022         {
7023                 .procname       = "force_tllao",
7024                 .data           = &ipv6_devconf.force_tllao,
7025                 .maxlen         = sizeof(int),
7026                 .mode           = 0644,
7027                 .proc_handler   = proc_dointvec
7028         },
7029         {
7030                 .procname       = "ndisc_notify",
7031                 .data           = &ipv6_devconf.ndisc_notify,
7032                 .maxlen         = sizeof(int),
7033                 .mode           = 0644,
7034                 .proc_handler   = proc_dointvec
7035         },
7036         {
7037                 .procname       = "suppress_frag_ndisc",
7038                 .data           = &ipv6_devconf.suppress_frag_ndisc,
7039                 .maxlen         = sizeof(int),
7040                 .mode           = 0644,
7041                 .proc_handler   = proc_dointvec
7042         },
7043         {
7044                 .procname       = "accept_ra_from_local",
7045                 .data           = &ipv6_devconf.accept_ra_from_local,
7046                 .maxlen         = sizeof(int),
7047                 .mode           = 0644,
7048                 .proc_handler   = proc_dointvec,
7049         },
7050         {
7051                 .procname       = "accept_ra_mtu",
7052                 .data           = &ipv6_devconf.accept_ra_mtu,
7053                 .maxlen         = sizeof(int),
7054                 .mode           = 0644,
7055                 .proc_handler   = proc_dointvec,
7056         },
7057         {
7058                 .procname       = "stable_secret",
7059                 .data           = &ipv6_devconf.stable_secret,
7060                 .maxlen         = IPV6_MAX_STRLEN,
7061                 .mode           = 0600,
7062                 .proc_handler   = addrconf_sysctl_stable_secret,
7063         },
7064         {
7065                 .procname       = "use_oif_addrs_only",
7066                 .data           = &ipv6_devconf.use_oif_addrs_only,
7067                 .maxlen         = sizeof(int),
7068                 .mode           = 0644,
7069                 .proc_handler   = proc_dointvec,
7070         },
7071         {
7072                 .procname       = "ignore_routes_with_linkdown",
7073                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
7074                 .maxlen         = sizeof(int),
7075                 .mode           = 0644,
7076                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
7077         },
7078         {
7079                 .procname       = "drop_unicast_in_l2_multicast",
7080                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
7081                 .maxlen         = sizeof(int),
7082                 .mode           = 0644,
7083                 .proc_handler   = proc_dointvec,
7084         },
7085         {
7086                 .procname       = "drop_unsolicited_na",
7087                 .data           = &ipv6_devconf.drop_unsolicited_na,
7088                 .maxlen         = sizeof(int),
7089                 .mode           = 0644,
7090                 .proc_handler   = proc_dointvec,
7091         },
7092         {
7093                 .procname       = "keep_addr_on_down",
7094                 .data           = &ipv6_devconf.keep_addr_on_down,
7095                 .maxlen         = sizeof(int),
7096                 .mode           = 0644,
7097                 .proc_handler   = proc_dointvec,
7098
7099         },
7100         {
7101                 .procname       = "seg6_enabled",
7102                 .data           = &ipv6_devconf.seg6_enabled,
7103                 .maxlen         = sizeof(int),
7104                 .mode           = 0644,
7105                 .proc_handler   = proc_dointvec,
7106         },
7107 #ifdef CONFIG_IPV6_SEG6_HMAC
7108         {
7109                 .procname       = "seg6_require_hmac",
7110                 .data           = &ipv6_devconf.seg6_require_hmac,
7111                 .maxlen         = sizeof(int),
7112                 .mode           = 0644,
7113                 .proc_handler   = proc_dointvec,
7114         },
7115 #endif
7116         {
7117                 .procname       = "enhanced_dad",
7118                 .data           = &ipv6_devconf.enhanced_dad,
7119                 .maxlen         = sizeof(int),
7120                 .mode           = 0644,
7121                 .proc_handler   = proc_dointvec,
7122         },
7123         {
7124                 .procname       = "addr_gen_mode",
7125                 .data           = &ipv6_devconf.addr_gen_mode,
7126                 .maxlen         = sizeof(int),
7127                 .mode           = 0644,
7128                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
7129         },
7130         {
7131                 .procname       = "disable_policy",
7132                 .data           = &ipv6_devconf.disable_policy,
7133                 .maxlen         = sizeof(int),
7134                 .mode           = 0644,
7135                 .proc_handler   = addrconf_sysctl_disable_policy,
7136         },
7137         {
7138                 .procname       = "ndisc_tclass",
7139                 .data           = &ipv6_devconf.ndisc_tclass,
7140                 .maxlen         = sizeof(int),
7141                 .mode           = 0644,
7142                 .proc_handler   = proc_dointvec_minmax,
7143                 .extra1         = (void *)SYSCTL_ZERO,
7144                 .extra2         = (void *)&two_five_five,
7145         },
7146         {
7147                 .procname       = "rpl_seg_enabled",
7148                 .data           = &ipv6_devconf.rpl_seg_enabled,
7149                 .maxlen         = sizeof(int),
7150                 .mode           = 0644,
7151                 .proc_handler   = proc_dointvec,
7152         },
7153         {
7154                 .procname       = "ioam6_enabled",
7155                 .data           = &ipv6_devconf.ioam6_enabled,
7156                 .maxlen         = sizeof(u8),
7157                 .mode           = 0644,
7158                 .proc_handler   = proc_dou8vec_minmax,
7159                 .extra1         = (void *)SYSCTL_ZERO,
7160                 .extra2         = (void *)SYSCTL_ONE,
7161         },
7162         {
7163                 .procname       = "ioam6_id",
7164                 .data           = &ipv6_devconf.ioam6_id,
7165                 .maxlen         = sizeof(u32),
7166                 .mode           = 0644,
7167                 .proc_handler   = proc_douintvec_minmax,
7168                 .extra1         = (void *)SYSCTL_ZERO,
7169                 .extra2         = (void *)&ioam6_if_id_max,
7170         },
7171         {
7172                 .procname       = "ioam6_id_wide",
7173                 .data           = &ipv6_devconf.ioam6_id_wide,
7174                 .maxlen         = sizeof(u32),
7175                 .mode           = 0644,
7176                 .proc_handler   = proc_douintvec,
7177         },
7178         {
7179                 .procname       = "ndisc_evict_nocarrier",
7180                 .data           = &ipv6_devconf.ndisc_evict_nocarrier,
7181                 .maxlen         = sizeof(u8),
7182                 .mode           = 0644,
7183                 .proc_handler   = proc_dou8vec_minmax,
7184                 .extra1         = (void *)SYSCTL_ZERO,
7185                 .extra2         = (void *)SYSCTL_ONE,
7186         },
7187         {
7188                 .procname       = "accept_untracked_na",
7189                 .data           = &ipv6_devconf.accept_untracked_na,
7190                 .maxlen         = sizeof(int),
7191                 .mode           = 0644,
7192                 .proc_handler   = proc_dointvec_minmax,
7193                 .extra1         = SYSCTL_ZERO,
7194                 .extra2         = SYSCTL_TWO,
7195         },
7196 };
7197
7198 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7199                 struct inet6_dev *idev, struct ipv6_devconf *p)
7200 {
7201         size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7202         int i, ifindex;
7203         struct ctl_table *table;
7204         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7205
7206         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7207         if (!table)
7208                 goto out;
7209
7210         for (i = 0; i < table_size; i++) {
7211                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7212                 /* If one of these is already set, then it is not safe to
7213                  * overwrite either of them: this makes proc_dointvec_minmax
7214                  * usable.
7215                  */
7216                 if (!table[i].extra1 && !table[i].extra2) {
7217                         table[i].extra1 = idev; /* embedded; no ref */
7218                         table[i].extra2 = net;
7219                 }
7220         }
7221
7222         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7223
7224         p->sysctl_header = register_net_sysctl_sz(net, path, table,
7225                                                   table_size);
7226         if (!p->sysctl_header)
7227                 goto free;
7228
7229         if (!strcmp(dev_name, "all"))
7230                 ifindex = NETCONFA_IFINDEX_ALL;
7231         else if (!strcmp(dev_name, "default"))
7232                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7233         else
7234                 ifindex = idev->dev->ifindex;
7235         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7236                                      ifindex, p);
7237         return 0;
7238
7239 free:
7240         kfree(table);
7241 out:
7242         return -ENOBUFS;
7243 }
7244
7245 static void __addrconf_sysctl_unregister(struct net *net,
7246                                          struct ipv6_devconf *p, int ifindex)
7247 {
7248         const struct ctl_table *table;
7249
7250         if (!p->sysctl_header)
7251                 return;
7252
7253         table = p->sysctl_header->ctl_table_arg;
7254         unregister_net_sysctl_table(p->sysctl_header);
7255         p->sysctl_header = NULL;
7256         kfree(table);
7257
7258         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7259 }
7260
7261 static int addrconf_sysctl_register(struct inet6_dev *idev)
7262 {
7263         int err;
7264
7265         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7266                 return -EINVAL;
7267
7268         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7269                                     &ndisc_ifinfo_sysctl_change);
7270         if (err)
7271                 return err;
7272         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7273                                          idev, &idev->cnf);
7274         if (err)
7275                 neigh_sysctl_unregister(idev->nd_parms);
7276
7277         return err;
7278 }
7279
7280 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7281 {
7282         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7283                                      idev->dev->ifindex);
7284         neigh_sysctl_unregister(idev->nd_parms);
7285 }
7286
7287
7288 #endif
7289
7290 static int __net_init addrconf_init_net(struct net *net)
7291 {
7292         int err = -ENOMEM;
7293         struct ipv6_devconf *all, *dflt;
7294
7295         spin_lock_init(&net->ipv6.addrconf_hash_lock);
7296         INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7297         net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7298                                            sizeof(struct hlist_head),
7299                                            GFP_KERNEL);
7300         if (!net->ipv6.inet6_addr_lst)
7301                 goto err_alloc_addr;
7302
7303         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7304         if (!all)
7305                 goto err_alloc_all;
7306
7307         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7308         if (!dflt)
7309                 goto err_alloc_dflt;
7310
7311         if (!net_eq(net, &init_net)) {
7312                 switch (net_inherit_devconf()) {
7313                 case 1:  /* copy from init_net */
7314                         memcpy(all, init_net.ipv6.devconf_all,
7315                                sizeof(ipv6_devconf));
7316                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7317                                sizeof(ipv6_devconf_dflt));
7318                         break;
7319                 case 3: /* copy from the current netns */
7320                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7321                                sizeof(ipv6_devconf));
7322                         memcpy(dflt,
7323                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7324                                sizeof(ipv6_devconf_dflt));
7325                         break;
7326                 case 0:
7327                 case 2:
7328                         /* use compiled values */
7329                         break;
7330                 }
7331         }
7332
7333         /* these will be inherited by all namespaces */
7334         dflt->autoconf = ipv6_defaults.autoconf;
7335         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7336
7337         dflt->stable_secret.initialized = false;
7338         all->stable_secret.initialized = false;
7339
7340         net->ipv6.devconf_all = all;
7341         net->ipv6.devconf_dflt = dflt;
7342
7343 #ifdef CONFIG_SYSCTL
7344         err = __addrconf_sysctl_register(net, "all", NULL, all);
7345         if (err < 0)
7346                 goto err_reg_all;
7347
7348         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7349         if (err < 0)
7350                 goto err_reg_dflt;
7351 #endif
7352         return 0;
7353
7354 #ifdef CONFIG_SYSCTL
7355 err_reg_dflt:
7356         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7357 err_reg_all:
7358         kfree(dflt);
7359         net->ipv6.devconf_dflt = NULL;
7360 #endif
7361 err_alloc_dflt:
7362         kfree(all);
7363         net->ipv6.devconf_all = NULL;
7364 err_alloc_all:
7365         kfree(net->ipv6.inet6_addr_lst);
7366 err_alloc_addr:
7367         return err;
7368 }
7369
7370 static void __net_exit addrconf_exit_net(struct net *net)
7371 {
7372         int i;
7373
7374 #ifdef CONFIG_SYSCTL
7375         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7376                                      NETCONFA_IFINDEX_DEFAULT);
7377         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7378                                      NETCONFA_IFINDEX_ALL);
7379 #endif
7380         kfree(net->ipv6.devconf_dflt);
7381         net->ipv6.devconf_dflt = NULL;
7382         kfree(net->ipv6.devconf_all);
7383         net->ipv6.devconf_all = NULL;
7384
7385         cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7386         /*
7387          *      Check hash table, then free it.
7388          */
7389         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7390                 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7391
7392         kfree(net->ipv6.inet6_addr_lst);
7393         net->ipv6.inet6_addr_lst = NULL;
7394 }
7395
7396 static struct pernet_operations addrconf_ops = {
7397         .init = addrconf_init_net,
7398         .exit = addrconf_exit_net,
7399 };
7400
7401 static struct rtnl_af_ops inet6_ops __read_mostly = {
7402         .family           = AF_INET6,
7403         .fill_link_af     = inet6_fill_link_af,
7404         .get_link_af_size = inet6_get_link_af_size,
7405         .validate_link_af = inet6_validate_link_af,
7406         .set_link_af      = inet6_set_link_af,
7407 };
7408
7409 /*
7410  *      Init / cleanup code
7411  */
7412
7413 int __init addrconf_init(void)
7414 {
7415         struct inet6_dev *idev;
7416         int err;
7417
7418         err = ipv6_addr_label_init();
7419         if (err < 0) {
7420                 pr_crit("%s: cannot initialize default policy table: %d\n",
7421                         __func__, err);
7422                 goto out;
7423         }
7424
7425         err = register_pernet_subsys(&addrconf_ops);
7426         if (err < 0)
7427                 goto out_addrlabel;
7428
7429         /* All works using addrconf_wq need to lock rtnl. */
7430         addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7431         if (!addrconf_wq) {
7432                 err = -ENOMEM;
7433                 goto out_nowq;
7434         }
7435
7436         rtnl_lock();
7437         idev = ipv6_add_dev(blackhole_netdev);
7438         rtnl_unlock();
7439         if (IS_ERR(idev)) {
7440                 err = PTR_ERR(idev);
7441                 goto errlo;
7442         }
7443
7444         ip6_route_init_special_entries();
7445
7446         register_netdevice_notifier(&ipv6_dev_notf);
7447
7448         addrconf_verify(&init_net);
7449
7450         rtnl_af_register(&inet6_ops);
7451
7452         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7453                                    NULL, inet6_dump_ifinfo, RTNL_FLAG_DUMP_UNLOCKED);
7454         if (err < 0)
7455                 goto errout;
7456
7457         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7458                                    inet6_rtm_newaddr, NULL, 0);
7459         if (err < 0)
7460                 goto errout;
7461         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7462                                    inet6_rtm_deladdr, NULL, 0);
7463         if (err < 0)
7464                 goto errout;
7465         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7466                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7467                                    RTNL_FLAG_DOIT_UNLOCKED |
7468                                    RTNL_FLAG_DUMP_UNLOCKED);
7469         if (err < 0)
7470                 goto errout;
7471         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7472                                    NULL, inet6_dump_ifmcaddr,
7473                                    RTNL_FLAG_DUMP_UNLOCKED);
7474         if (err < 0)
7475                 goto errout;
7476         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7477                                    NULL, inet6_dump_ifacaddr,
7478                                    RTNL_FLAG_DUMP_UNLOCKED);
7479         if (err < 0)
7480                 goto errout;
7481         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7482                                    inet6_netconf_get_devconf,
7483                                    inet6_netconf_dump_devconf,
7484                                    RTNL_FLAG_DOIT_UNLOCKED |
7485                                    RTNL_FLAG_DUMP_UNLOCKED);
7486         if (err < 0)
7487                 goto errout;
7488         err = ipv6_addr_label_rtnl_register();
7489         if (err < 0)
7490                 goto errout;
7491
7492         return 0;
7493 errout:
7494         rtnl_unregister_all(PF_INET6);
7495         rtnl_af_unregister(&inet6_ops);
7496         unregister_netdevice_notifier(&ipv6_dev_notf);
7497 errlo:
7498         destroy_workqueue(addrconf_wq);
7499 out_nowq:
7500         unregister_pernet_subsys(&addrconf_ops);
7501 out_addrlabel:
7502         ipv6_addr_label_cleanup();
7503 out:
7504         return err;
7505 }
7506
7507 void addrconf_cleanup(void)
7508 {
7509         struct net_device *dev;
7510
7511         unregister_netdevice_notifier(&ipv6_dev_notf);
7512         unregister_pernet_subsys(&addrconf_ops);
7513         ipv6_addr_label_cleanup();
7514
7515         rtnl_af_unregister(&inet6_ops);
7516
7517         rtnl_lock();
7518
7519         /* clean dev list */
7520         for_each_netdev(&init_net, dev) {
7521                 if (__in6_dev_get(dev) == NULL)
7522                         continue;
7523                 addrconf_ifdown(dev, true);
7524         }
7525         addrconf_ifdown(init_net.loopback_dev, true);
7526
7527         rtnl_unlock();
7528
7529         destroy_workqueue(addrconf_wq);
7530 }