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 <linux/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_jhash(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 delete_tempaddrs(struct inet6_dev *idev,
2574                              struct inet6_ifaddr *ifp)
2575 {
2576         struct inet6_ifaddr *ift, *tmp;
2577
2578         write_lock_bh(&idev->lock);
2579         list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) {
2580                 if (ift->ifpub != ifp)
2581                         continue;
2582
2583                 in6_ifa_hold(ift);
2584                 write_unlock_bh(&idev->lock);
2585                 ipv6_del_addr(ift);
2586                 write_lock_bh(&idev->lock);
2587         }
2588         write_unlock_bh(&idev->lock);
2589 }
2590
2591 static void manage_tempaddrs(struct inet6_dev *idev,
2592                              struct inet6_ifaddr *ifp,
2593                              __u32 valid_lft, __u32 prefered_lft,
2594                              bool create, unsigned long now)
2595 {
2596         u32 flags;
2597         struct inet6_ifaddr *ift;
2598
2599         read_lock_bh(&idev->lock);
2600         /* update all temporary addresses in the list */
2601         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2602                 int age, max_valid, max_prefered;
2603
2604                 if (ifp != ift->ifpub)
2605                         continue;
2606
2607                 /* RFC 4941 section 3.3:
2608                  * If a received option will extend the lifetime of a public
2609                  * address, the lifetimes of temporary addresses should
2610                  * be extended, subject to the overall constraint that no
2611                  * temporary addresses should ever remain "valid" or "preferred"
2612                  * for a time longer than (TEMP_VALID_LIFETIME) or
2613                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2614                  */
2615                 age = (now - ift->cstamp) / HZ;
2616                 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2617                 if (max_valid < 0)
2618                         max_valid = 0;
2619
2620                 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2621                                idev->desync_factor - age;
2622                 if (max_prefered < 0)
2623                         max_prefered = 0;
2624
2625                 if (valid_lft > max_valid)
2626                         valid_lft = max_valid;
2627
2628                 if (prefered_lft > max_prefered)
2629                         prefered_lft = max_prefered;
2630
2631                 spin_lock(&ift->lock);
2632                 flags = ift->flags;
2633                 ift->valid_lft = valid_lft;
2634                 ift->prefered_lft = prefered_lft;
2635                 ift->tstamp = now;
2636                 if (prefered_lft > 0)
2637                         ift->flags &= ~IFA_F_DEPRECATED;
2638
2639                 spin_unlock(&ift->lock);
2640                 if (!(flags&IFA_F_TENTATIVE))
2641                         ipv6_ifa_notify(0, ift);
2642         }
2643
2644         /* Also create a temporary address if it's enabled but no temporary
2645          * address currently exists.
2646          * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2647          * as part of cleanup (ie. deleting the mngtmpaddr).
2648          * We don't want that to result in creating a new temporary ip address.
2649          */
2650         if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2651                 create = true;
2652
2653         if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2654                 /* When a new public address is created as described
2655                  * in [ADDRCONF], also create a new temporary address.
2656                  */
2657                 read_unlock_bh(&idev->lock);
2658                 ipv6_create_tempaddr(ifp, false);
2659         } else {
2660                 read_unlock_bh(&idev->lock);
2661         }
2662 }
2663
2664 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2665 {
2666         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2667                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2668 }
2669
2670 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2671                                  const struct prefix_info *pinfo,
2672                                  struct inet6_dev *in6_dev,
2673                                  const struct in6_addr *addr, int addr_type,
2674                                  u32 addr_flags, bool sllao, bool tokenized,
2675                                  __u32 valid_lft, u32 prefered_lft)
2676 {
2677         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2678         int create = 0, update_lft = 0;
2679
2680         if (!ifp && valid_lft) {
2681                 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2682                 struct ifa6_config cfg = {
2683                         .pfx = addr,
2684                         .plen = pinfo->prefix_len,
2685                         .ifa_flags = addr_flags,
2686                         .valid_lft = valid_lft,
2687                         .preferred_lft = prefered_lft,
2688                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2689                         .ifa_proto = IFAPROT_KERNEL_RA
2690                 };
2691
2692 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2693                 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2694                      READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2695                     !net->ipv6.devconf_all->forwarding && sllao)
2696                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2697 #endif
2698
2699                 /* Do not allow to create too much of autoconfigured
2700                  * addresses; this would be too easy way to crash kernel.
2701                  */
2702                 if (!max_addresses ||
2703                     ipv6_count_addresses(in6_dev) < max_addresses)
2704                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2705
2706                 if (IS_ERR_OR_NULL(ifp))
2707                         return -1;
2708
2709                 create = 1;
2710                 spin_lock_bh(&ifp->lock);
2711                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2712                 ifp->cstamp = jiffies;
2713                 ifp->tokenized = tokenized;
2714                 spin_unlock_bh(&ifp->lock);
2715                 addrconf_dad_start(ifp);
2716         }
2717
2718         if (ifp) {
2719                 u32 flags;
2720                 unsigned long now;
2721                 u32 stored_lft;
2722
2723                 /* update lifetime (RFC2462 5.5.3 e) */
2724                 spin_lock_bh(&ifp->lock);
2725                 now = jiffies;
2726                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2727                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2728                 else
2729                         stored_lft = 0;
2730
2731                 /* RFC4862 Section 5.5.3e:
2732                  * "Note that the preferred lifetime of the
2733                  *  corresponding address is always reset to
2734                  *  the Preferred Lifetime in the received
2735                  *  Prefix Information option, regardless of
2736                  *  whether the valid lifetime is also reset or
2737                  *  ignored."
2738                  *
2739                  * So we should always update prefered_lft here.
2740                  */
2741                 update_lft = !create && stored_lft;
2742
2743                 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2744                         const u32 minimum_lft = min_t(u32,
2745                                 stored_lft, MIN_VALID_LIFETIME);
2746                         valid_lft = max(valid_lft, minimum_lft);
2747                 }
2748
2749                 if (update_lft) {
2750                         ifp->valid_lft = valid_lft;
2751                         ifp->prefered_lft = prefered_lft;
2752                         WRITE_ONCE(ifp->tstamp, now);
2753                         flags = ifp->flags;
2754                         ifp->flags &= ~IFA_F_DEPRECATED;
2755                         spin_unlock_bh(&ifp->lock);
2756
2757                         if (!(flags&IFA_F_TENTATIVE))
2758                                 ipv6_ifa_notify(0, ifp);
2759                 } else
2760                         spin_unlock_bh(&ifp->lock);
2761
2762                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2763                                  create, now);
2764
2765                 in6_ifa_put(ifp);
2766                 addrconf_verify(net);
2767         }
2768
2769         return 0;
2770 }
2771 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2772
2773 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2774 {
2775         struct prefix_info *pinfo;
2776         struct fib6_table *table;
2777         __u32 valid_lft;
2778         __u32 prefered_lft;
2779         int addr_type, err;
2780         u32 addr_flags = 0;
2781         struct inet6_dev *in6_dev;
2782         struct net *net = dev_net(dev);
2783         bool ignore_autoconf = false;
2784
2785         pinfo = (struct prefix_info *) opt;
2786
2787         if (len < sizeof(struct prefix_info)) {
2788                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2789                 return;
2790         }
2791
2792         /*
2793          *      Validation checks ([ADDRCONF], page 19)
2794          */
2795
2796         addr_type = ipv6_addr_type(&pinfo->prefix);
2797
2798         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2799                 return;
2800
2801         valid_lft = ntohl(pinfo->valid);
2802         prefered_lft = ntohl(pinfo->prefered);
2803
2804         if (prefered_lft > valid_lft) {
2805                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2806                 return;
2807         }
2808
2809         in6_dev = in6_dev_get(dev);
2810
2811         if (!in6_dev) {
2812                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2813                                     dev->name);
2814                 return;
2815         }
2816
2817         if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2818                 goto put;
2819
2820         /*
2821          *      Two things going on here:
2822          *      1) Add routes for on-link prefixes
2823          *      2) Configure prefixes with the auto flag set
2824          */
2825
2826         if (pinfo->onlink) {
2827                 struct fib6_info *rt;
2828                 unsigned long rt_expires;
2829
2830                 /* Avoid arithmetic overflow. Really, we could
2831                  * save rt_expires in seconds, likely valid_lft,
2832                  * but it would require division in fib gc, that it
2833                  * not good.
2834                  */
2835                 if (HZ > USER_HZ)
2836                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2837                 else
2838                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2839
2840                 if (addrconf_finite_timeout(rt_expires))
2841                         rt_expires *= HZ;
2842
2843                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2844                                                pinfo->prefix_len,
2845                                                dev,
2846                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2847                                                RTF_DEFAULT, true);
2848
2849                 if (rt) {
2850                         /* Autoconf prefix route */
2851                         if (valid_lft == 0) {
2852                                 ip6_del_rt(net, rt, false);
2853                                 rt = NULL;
2854                         } else {
2855                                 table = rt->fib6_table;
2856                                 spin_lock_bh(&table->tb6_lock);
2857
2858                                 if (addrconf_finite_timeout(rt_expires)) {
2859                                         /* not infinity */
2860                                         fib6_set_expires(rt, jiffies + rt_expires);
2861                                         fib6_add_gc_list(rt);
2862                                 } else {
2863                                         fib6_clean_expires(rt);
2864                                         fib6_remove_gc_list(rt);
2865                                 }
2866
2867                                 spin_unlock_bh(&table->tb6_lock);
2868                         }
2869                 } else if (valid_lft) {
2870                         clock_t expires = 0;
2871                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2872                         if (addrconf_finite_timeout(rt_expires)) {
2873                                 /* not infinity */
2874                                 flags |= RTF_EXPIRES;
2875                                 expires = jiffies_to_clock_t(rt_expires);
2876                         }
2877                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2878                                               0, dev, expires, flags,
2879                                               GFP_ATOMIC);
2880                 }
2881                 fib6_info_release(rt);
2882         }
2883
2884         /* Try to figure out our local address for this prefix */
2885
2886         ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd;
2887         if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) {
2888                 struct in6_addr addr;
2889                 bool tokenized = false, dev_addr_generated = false;
2890
2891                 if (pinfo->prefix_len == 64) {
2892                         memcpy(&addr, &pinfo->prefix, 8);
2893
2894                         if (!ipv6_addr_any(&in6_dev->token)) {
2895                                 read_lock_bh(&in6_dev->lock);
2896                                 memcpy(addr.s6_addr + 8,
2897                                        in6_dev->token.s6_addr + 8, 8);
2898                                 read_unlock_bh(&in6_dev->lock);
2899                                 tokenized = true;
2900                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2901                                    !ipv6_generate_stable_address(&addr, 0,
2902                                                                  in6_dev)) {
2903                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2904                                 goto ok;
2905                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2906                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2907                                 goto put;
2908                         } else {
2909                                 dev_addr_generated = true;
2910                         }
2911                         goto ok;
2912                 }
2913                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2914                                     pinfo->prefix_len);
2915                 goto put;
2916
2917 ok:
2918                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2919                                                    &addr, addr_type,
2920                                                    addr_flags, sllao,
2921                                                    tokenized, valid_lft,
2922                                                    prefered_lft);
2923                 if (err)
2924                         goto put;
2925
2926                 /* Ignore error case here because previous prefix add addr was
2927                  * successful which will be notified.
2928                  */
2929                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2930                                               addr_type, addr_flags, sllao,
2931                                               tokenized, valid_lft,
2932                                               prefered_lft,
2933                                               dev_addr_generated);
2934         }
2935         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2936 put:
2937         in6_dev_put(in6_dev);
2938 }
2939
2940 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2941                 struct in6_ifreq *ireq)
2942 {
2943         struct ip_tunnel_parm_kern p = { };
2944         int err;
2945
2946         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2947                 return -EADDRNOTAVAIL;
2948
2949         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2950         p.iph.version = 4;
2951         p.iph.ihl = 5;
2952         p.iph.protocol = IPPROTO_IPV6;
2953         p.iph.ttl = 64;
2954
2955         if (!dev->netdev_ops->ndo_tunnel_ctl)
2956                 return -EOPNOTSUPP;
2957         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2958         if (err)
2959                 return err;
2960
2961         dev = __dev_get_by_name(net, p.name);
2962         if (!dev)
2963                 return -ENOBUFS;
2964         return dev_open(dev, NULL);
2965 }
2966
2967 /*
2968  *      Set destination address.
2969  *      Special case for SIT interfaces where we create a new "virtual"
2970  *      device.
2971  */
2972 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2973 {
2974         struct net_device *dev;
2975         struct in6_ifreq ireq;
2976         int err = -ENODEV;
2977
2978         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2979                 return -ENODEV;
2980         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2981                 return -EFAULT;
2982
2983         rtnl_lock();
2984         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2985         if (dev && dev->type == ARPHRD_SIT)
2986                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2987         rtnl_unlock();
2988         return err;
2989 }
2990
2991 static int ipv6_mc_config(struct sock *sk, bool join,
2992                           const struct in6_addr *addr, int ifindex)
2993 {
2994         int ret;
2995
2996         ASSERT_RTNL();
2997
2998         lock_sock(sk);
2999         if (join)
3000                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
3001         else
3002                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
3003         release_sock(sk);
3004
3005         return ret;
3006 }
3007
3008 /*
3009  *      Manual configuration of address on an interface
3010  */
3011 static int inet6_addr_add(struct net *net, int ifindex,
3012                           struct ifa6_config *cfg,
3013                           struct netlink_ext_ack *extack)
3014 {
3015         struct inet6_ifaddr *ifp;
3016         struct inet6_dev *idev;
3017         struct net_device *dev;
3018         unsigned long timeout;
3019         clock_t expires;
3020         u32 flags;
3021
3022         ASSERT_RTNL();
3023
3024         if (cfg->plen > 128) {
3025                 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3026                 return -EINVAL;
3027         }
3028
3029         /* check the lifetime */
3030         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
3031                 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
3032                 return -EINVAL;
3033         }
3034
3035         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3036                 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3037                 return -EINVAL;
3038         }
3039
3040         dev = __dev_get_by_index(net, ifindex);
3041         if (!dev)
3042                 return -ENODEV;
3043
3044         idev = addrconf_add_dev(dev);
3045         if (IS_ERR(idev)) {
3046                 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3047                 return PTR_ERR(idev);
3048         }
3049
3050         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3051                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3052                                          true, cfg->pfx, ifindex);
3053
3054                 if (ret < 0) {
3055                         NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3056                         return ret;
3057                 }
3058         }
3059
3060         cfg->scope = ipv6_addr_scope(cfg->pfx);
3061
3062         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
3063         if (addrconf_finite_timeout(timeout)) {
3064                 expires = jiffies_to_clock_t(timeout * HZ);
3065                 cfg->valid_lft = timeout;
3066                 flags = RTF_EXPIRES;
3067         } else {
3068                 expires = 0;
3069                 flags = 0;
3070                 cfg->ifa_flags |= IFA_F_PERMANENT;
3071         }
3072
3073         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3074         if (addrconf_finite_timeout(timeout)) {
3075                 if (timeout == 0)
3076                         cfg->ifa_flags |= IFA_F_DEPRECATED;
3077                 cfg->preferred_lft = timeout;
3078         }
3079
3080         ifp = ipv6_add_addr(idev, cfg, true, extack);
3081         if (!IS_ERR(ifp)) {
3082                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3083                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3084                                               ifp->rt_priority, dev, expires,
3085                                               flags, GFP_KERNEL);
3086                 }
3087
3088                 /* Send a netlink notification if DAD is enabled and
3089                  * optimistic flag is not set
3090                  */
3091                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3092                         ipv6_ifa_notify(0, ifp);
3093                 /*
3094                  * Note that section 3.1 of RFC 4429 indicates
3095                  * that the Optimistic flag should not be set for
3096                  * manually configured addresses
3097                  */
3098                 addrconf_dad_start(ifp);
3099                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3100                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
3101                                          cfg->preferred_lft, true, jiffies);
3102                 in6_ifa_put(ifp);
3103                 addrconf_verify_rtnl(net);
3104                 return 0;
3105         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3106                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3107                                cfg->pfx, ifindex);
3108         }
3109
3110         return PTR_ERR(ifp);
3111 }
3112
3113 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3114                           const struct in6_addr *pfx, unsigned int plen,
3115                           struct netlink_ext_ack *extack)
3116 {
3117         struct inet6_ifaddr *ifp;
3118         struct inet6_dev *idev;
3119         struct net_device *dev;
3120
3121         if (plen > 128) {
3122                 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3123                 return -EINVAL;
3124         }
3125
3126         dev = __dev_get_by_index(net, ifindex);
3127         if (!dev) {
3128                 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3129                 return -ENODEV;
3130         }
3131
3132         idev = __in6_dev_get(dev);
3133         if (!idev) {
3134                 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3135                 return -ENXIO;
3136         }
3137
3138         read_lock_bh(&idev->lock);
3139         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3140                 if (ifp->prefix_len == plen &&
3141                     ipv6_addr_equal(pfx, &ifp->addr)) {
3142                         in6_ifa_hold(ifp);
3143                         read_unlock_bh(&idev->lock);
3144
3145                         ipv6_del_addr(ifp);
3146
3147                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3148                             (ifp->flags & IFA_F_MANAGETEMPADDR))
3149                                 delete_tempaddrs(idev, ifp);
3150
3151                         addrconf_verify_rtnl(net);
3152                         if (ipv6_addr_is_multicast(pfx)) {
3153                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3154                                                false, pfx, dev->ifindex);
3155                         }
3156                         return 0;
3157                 }
3158         }
3159         read_unlock_bh(&idev->lock);
3160
3161         NL_SET_ERR_MSG_MOD(extack, "address not found");
3162         return -EADDRNOTAVAIL;
3163 }
3164
3165
3166 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3167 {
3168         struct ifa6_config cfg = {
3169                 .ifa_flags = IFA_F_PERMANENT,
3170                 .preferred_lft = INFINITY_LIFE_TIME,
3171                 .valid_lft = INFINITY_LIFE_TIME,
3172         };
3173         struct in6_ifreq ireq;
3174         int err;
3175
3176         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3177                 return -EPERM;
3178
3179         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3180                 return -EFAULT;
3181
3182         cfg.pfx = &ireq.ifr6_addr;
3183         cfg.plen = ireq.ifr6_prefixlen;
3184
3185         rtnl_lock();
3186         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3187         rtnl_unlock();
3188         return err;
3189 }
3190
3191 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3192 {
3193         struct in6_ifreq ireq;
3194         int err;
3195
3196         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3197                 return -EPERM;
3198
3199         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3200                 return -EFAULT;
3201
3202         rtnl_lock();
3203         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3204                              ireq.ifr6_prefixlen, NULL);
3205         rtnl_unlock();
3206         return err;
3207 }
3208
3209 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3210                      int plen, int scope, u8 proto)
3211 {
3212         struct inet6_ifaddr *ifp;
3213         struct ifa6_config cfg = {
3214                 .pfx = addr,
3215                 .plen = plen,
3216                 .ifa_flags = IFA_F_PERMANENT,
3217                 .valid_lft = INFINITY_LIFE_TIME,
3218                 .preferred_lft = INFINITY_LIFE_TIME,
3219                 .scope = scope,
3220                 .ifa_proto = proto
3221         };
3222
3223         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3224         if (!IS_ERR(ifp)) {
3225                 spin_lock_bh(&ifp->lock);
3226                 ifp->flags &= ~IFA_F_TENTATIVE;
3227                 spin_unlock_bh(&ifp->lock);
3228                 rt_genid_bump_ipv6(dev_net(idev->dev));
3229                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3230                 in6_ifa_put(ifp);
3231         }
3232 }
3233
3234 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3235 static void add_v4_addrs(struct inet6_dev *idev)
3236 {
3237         struct in6_addr addr;
3238         struct net_device *dev;
3239         struct net *net = dev_net(idev->dev);
3240         int scope, plen, offset = 0;
3241         u32 pflags = 0;
3242
3243         ASSERT_RTNL();
3244
3245         memset(&addr, 0, sizeof(struct in6_addr));
3246         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3247         if (idev->dev->addr_len == sizeof(struct in6_addr))
3248                 offset = sizeof(struct in6_addr) - 4;
3249         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3250
3251         if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3252                 scope = IPV6_ADDR_COMPATv4;
3253                 plen = 96;
3254                 pflags |= RTF_NONEXTHOP;
3255         } else {
3256                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3257                         return;
3258
3259                 addr.s6_addr32[0] = htonl(0xfe800000);
3260                 scope = IFA_LINK;
3261                 plen = 64;
3262         }
3263
3264         if (addr.s6_addr32[3]) {
3265                 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3266                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3267                                       GFP_KERNEL);
3268                 return;
3269         }
3270
3271         for_each_netdev(net, dev) {
3272                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3273                 if (in_dev && (dev->flags & IFF_UP)) {
3274                         struct in_ifaddr *ifa;
3275                         int flag = scope;
3276
3277                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3278                                 addr.s6_addr32[3] = ifa->ifa_local;
3279
3280                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3281                                         continue;
3282                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3283                                         if (idev->dev->flags&IFF_POINTOPOINT)
3284                                                 continue;
3285                                         flag |= IFA_HOST;
3286                                 }
3287
3288                                 add_addr(idev, &addr, plen, flag,
3289                                          IFAPROT_UNSPEC);
3290                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3291                                                       0, pflags, GFP_KERNEL);
3292                         }
3293                 }
3294         }
3295 }
3296 #endif
3297
3298 static void init_loopback(struct net_device *dev)
3299 {
3300         struct inet6_dev  *idev;
3301
3302         /* ::1 */
3303
3304         ASSERT_RTNL();
3305
3306         idev = ipv6_find_idev(dev);
3307         if (IS_ERR(idev)) {
3308                 pr_debug("%s: add_dev failed\n", __func__);
3309                 return;
3310         }
3311
3312         add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3313 }
3314
3315 void addrconf_add_linklocal(struct inet6_dev *idev,
3316                             const struct in6_addr *addr, u32 flags)
3317 {
3318         struct ifa6_config cfg = {
3319                 .pfx = addr,
3320                 .plen = 64,
3321                 .ifa_flags = flags | IFA_F_PERMANENT,
3322                 .valid_lft = INFINITY_LIFE_TIME,
3323                 .preferred_lft = INFINITY_LIFE_TIME,
3324                 .scope = IFA_LINK,
3325                 .ifa_proto = IFAPROT_KERNEL_LL
3326         };
3327         struct inet6_ifaddr *ifp;
3328
3329 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3330         if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3331              READ_ONCE(idev->cnf.optimistic_dad)) &&
3332             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3333                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3334 #endif
3335
3336         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3337         if (!IS_ERR(ifp)) {
3338                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3339                                       0, 0, GFP_ATOMIC);
3340                 addrconf_dad_start(ifp);
3341                 in6_ifa_put(ifp);
3342         }
3343 }
3344 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3345
3346 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3347 {
3348         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3349                 return true;
3350
3351         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3352             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3353                 return true;
3354
3355         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3356             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3357                 return true;
3358
3359         return false;
3360 }
3361
3362 static int ipv6_generate_stable_address(struct in6_addr *address,
3363                                         u8 dad_count,
3364                                         const struct inet6_dev *idev)
3365 {
3366         static DEFINE_SPINLOCK(lock);
3367         static __u32 digest[SHA1_DIGEST_WORDS];
3368         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3369
3370         static union {
3371                 char __data[SHA1_BLOCK_SIZE];
3372                 struct {
3373                         struct in6_addr secret;
3374                         __be32 prefix[2];
3375                         unsigned char hwaddr[MAX_ADDR_LEN];
3376                         u8 dad_count;
3377                 } __packed;
3378         } data;
3379
3380         struct in6_addr secret;
3381         struct in6_addr temp;
3382         struct net *net = dev_net(idev->dev);
3383
3384         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3385
3386         if (idev->cnf.stable_secret.initialized)
3387                 secret = idev->cnf.stable_secret.secret;
3388         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3389                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3390         else
3391                 return -1;
3392
3393 retry:
3394         spin_lock_bh(&lock);
3395
3396         sha1_init(digest);
3397         memset(&data, 0, sizeof(data));
3398         memset(workspace, 0, sizeof(workspace));
3399         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3400         data.prefix[0] = address->s6_addr32[0];
3401         data.prefix[1] = address->s6_addr32[1];
3402         data.secret = secret;
3403         data.dad_count = dad_count;
3404
3405         sha1_transform(digest, data.__data, workspace);
3406
3407         temp = *address;
3408         temp.s6_addr32[2] = (__force __be32)digest[0];
3409         temp.s6_addr32[3] = (__force __be32)digest[1];
3410
3411         spin_unlock_bh(&lock);
3412
3413         if (ipv6_reserved_interfaceid(temp)) {
3414                 dad_count++;
3415                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3416                         return -1;
3417                 goto retry;
3418         }
3419
3420         *address = temp;
3421         return 0;
3422 }
3423
3424 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3425 {
3426         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3427
3428         if (s->initialized)
3429                 return;
3430         s = &idev->cnf.stable_secret;
3431         get_random_bytes(&s->secret, sizeof(s->secret));
3432         s->initialized = true;
3433 }
3434
3435 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3436 {
3437         struct in6_addr addr;
3438
3439         /* no link local addresses on L3 master devices */
3440         if (netif_is_l3_master(idev->dev))
3441                 return;
3442
3443         /* no link local addresses on devices flagged as slaves */
3444         if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3445                 return;
3446
3447         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3448
3449         switch (idev->cnf.addr_gen_mode) {
3450         case IN6_ADDR_GEN_MODE_RANDOM:
3451                 ipv6_gen_mode_random_init(idev);
3452                 fallthrough;
3453         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3454                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3455                         addrconf_add_linklocal(idev, &addr,
3456                                                IFA_F_STABLE_PRIVACY);
3457                 else if (prefix_route)
3458                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3459                                               0, 0, GFP_KERNEL);
3460                 break;
3461         case IN6_ADDR_GEN_MODE_EUI64:
3462                 /* addrconf_add_linklocal also adds a prefix_route and we
3463                  * only need to care about prefix routes if ipv6_generate_eui64
3464                  * couldn't generate one.
3465                  */
3466                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3467                         addrconf_add_linklocal(idev, &addr, 0);
3468                 else if (prefix_route)
3469                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3470                                               0, 0, GFP_KERNEL);
3471                 break;
3472         case IN6_ADDR_GEN_MODE_NONE:
3473         default:
3474                 /* will not add any link local address */
3475                 break;
3476         }
3477 }
3478
3479 static void addrconf_dev_config(struct net_device *dev)
3480 {
3481         struct inet6_dev *idev;
3482
3483         ASSERT_RTNL();
3484
3485         if ((dev->type != ARPHRD_ETHER) &&
3486             (dev->type != ARPHRD_FDDI) &&
3487             (dev->type != ARPHRD_ARCNET) &&
3488             (dev->type != ARPHRD_INFINIBAND) &&
3489             (dev->type != ARPHRD_IEEE1394) &&
3490             (dev->type != ARPHRD_TUNNEL6) &&
3491             (dev->type != ARPHRD_6LOWPAN) &&
3492             (dev->type != ARPHRD_TUNNEL) &&
3493             (dev->type != ARPHRD_NONE) &&
3494             (dev->type != ARPHRD_RAWIP)) {
3495                 /* Alas, we support only Ethernet autoconfiguration. */
3496                 idev = __in6_dev_get(dev);
3497                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3498                     dev->flags & IFF_MULTICAST)
3499                         ipv6_mc_up(idev);
3500                 return;
3501         }
3502
3503         idev = addrconf_add_dev(dev);
3504         if (IS_ERR(idev))
3505                 return;
3506
3507         /* this device type has no EUI support */
3508         if (dev->type == ARPHRD_NONE &&
3509             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3510                 WRITE_ONCE(idev->cnf.addr_gen_mode,
3511                            IN6_ADDR_GEN_MODE_RANDOM);
3512
3513         addrconf_addr_gen(idev, false);
3514 }
3515
3516 #if IS_ENABLED(CONFIG_IPV6_SIT)
3517 static void addrconf_sit_config(struct net_device *dev)
3518 {
3519         struct inet6_dev *idev;
3520
3521         ASSERT_RTNL();
3522
3523         /*
3524          * Configure the tunnel with one of our IPv4
3525          * addresses... we should configure all of
3526          * our v4 addrs in the tunnel
3527          */
3528
3529         idev = ipv6_find_idev(dev);
3530         if (IS_ERR(idev)) {
3531                 pr_debug("%s: add_dev failed\n", __func__);
3532                 return;
3533         }
3534
3535         if (dev->priv_flags & IFF_ISATAP) {
3536                 addrconf_addr_gen(idev, false);
3537                 return;
3538         }
3539
3540         add_v4_addrs(idev);
3541
3542         if (dev->flags&IFF_POINTOPOINT)
3543                 addrconf_add_mroute(dev);
3544 }
3545 #endif
3546
3547 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3548 static void addrconf_gre_config(struct net_device *dev)
3549 {
3550         struct inet6_dev *idev;
3551
3552         ASSERT_RTNL();
3553
3554         idev = ipv6_find_idev(dev);
3555         if (IS_ERR(idev)) {
3556                 pr_debug("%s: add_dev failed\n", __func__);
3557                 return;
3558         }
3559
3560         if (dev->type == ARPHRD_ETHER) {
3561                 addrconf_addr_gen(idev, true);
3562                 return;
3563         }
3564
3565         add_v4_addrs(idev);
3566
3567         if (dev->flags & IFF_POINTOPOINT)
3568                 addrconf_add_mroute(dev);
3569 }
3570 #endif
3571
3572 static void addrconf_init_auto_addrs(struct net_device *dev)
3573 {
3574         switch (dev->type) {
3575 #if IS_ENABLED(CONFIG_IPV6_SIT)
3576         case ARPHRD_SIT:
3577                 addrconf_sit_config(dev);
3578                 break;
3579 #endif
3580 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3581         case ARPHRD_IP6GRE:
3582         case ARPHRD_IPGRE:
3583                 addrconf_gre_config(dev);
3584                 break;
3585 #endif
3586         case ARPHRD_LOOPBACK:
3587                 init_loopback(dev);
3588                 break;
3589
3590         default:
3591                 addrconf_dev_config(dev);
3592                 break;
3593         }
3594 }
3595
3596 static int fixup_permanent_addr(struct net *net,
3597                                 struct inet6_dev *idev,
3598                                 struct inet6_ifaddr *ifp)
3599 {
3600         /* !fib6_node means the host route was removed from the
3601          * FIB, for example, if 'lo' device is taken down. In that
3602          * case regenerate the host route.
3603          */
3604         if (!ifp->rt || !ifp->rt->fib6_node) {
3605                 struct fib6_info *f6i, *prev;
3606
3607                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3608                                          GFP_ATOMIC, NULL);
3609                 if (IS_ERR(f6i))
3610                         return PTR_ERR(f6i);
3611
3612                 /* ifp->rt can be accessed outside of rtnl */
3613                 spin_lock(&ifp->lock);
3614                 prev = ifp->rt;
3615                 ifp->rt = f6i;
3616                 spin_unlock(&ifp->lock);
3617
3618                 fib6_info_release(prev);
3619         }
3620
3621         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3622                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3623                                       ifp->rt_priority, idev->dev, 0, 0,
3624                                       GFP_ATOMIC);
3625         }
3626
3627         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3628                 addrconf_dad_start(ifp);
3629
3630         return 0;
3631 }
3632
3633 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3634 {
3635         struct inet6_ifaddr *ifp, *tmp;
3636         struct inet6_dev *idev;
3637
3638         idev = __in6_dev_get(dev);
3639         if (!idev)
3640                 return;
3641
3642         write_lock_bh(&idev->lock);
3643
3644         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3645                 if ((ifp->flags & IFA_F_PERMANENT) &&
3646                     fixup_permanent_addr(net, idev, ifp) < 0) {
3647                         write_unlock_bh(&idev->lock);
3648                         in6_ifa_hold(ifp);
3649                         ipv6_del_addr(ifp);
3650                         write_lock_bh(&idev->lock);
3651
3652                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3653                                              idev->dev->name, &ifp->addr);
3654                 }
3655         }
3656
3657         write_unlock_bh(&idev->lock);
3658 }
3659
3660 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3661                            void *ptr)
3662 {
3663         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3664         struct netdev_notifier_change_info *change_info;
3665         struct netdev_notifier_changeupper_info *info;
3666         struct inet6_dev *idev = __in6_dev_get(dev);
3667         struct net *net = dev_net(dev);
3668         int run_pending = 0;
3669         int err;
3670
3671         switch (event) {
3672         case NETDEV_REGISTER:
3673                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3674                         idev = ipv6_add_dev(dev);
3675                         if (IS_ERR(idev))
3676                                 return notifier_from_errno(PTR_ERR(idev));
3677                 }
3678                 break;
3679
3680         case NETDEV_CHANGEMTU:
3681                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3682                 if (dev->mtu < IPV6_MIN_MTU) {
3683                         addrconf_ifdown(dev, dev != net->loopback_dev);
3684                         break;
3685                 }
3686
3687                 if (idev) {
3688                         rt6_mtu_change(dev, dev->mtu);
3689                         WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3690                         break;
3691                 }
3692
3693                 /* allocate new idev */
3694                 idev = ipv6_add_dev(dev);
3695                 if (IS_ERR(idev))
3696                         break;
3697
3698                 /* device is still not ready */
3699                 if (!(idev->if_flags & IF_READY))
3700                         break;
3701
3702                 run_pending = 1;
3703                 fallthrough;
3704         case NETDEV_UP:
3705         case NETDEV_CHANGE:
3706                 if (idev && idev->cnf.disable_ipv6)
3707                         break;
3708
3709                 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3710                         if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3711                             dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3712                                 ipv6_mc_up(idev);
3713                         break;
3714                 }
3715
3716                 if (event == NETDEV_UP) {
3717                         /* restore routes for permanent addresses */
3718                         addrconf_permanent_addr(net, dev);
3719
3720                         if (!addrconf_link_ready(dev)) {
3721                                 /* device is not ready yet. */
3722                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3723                                          dev->name);
3724                                 break;
3725                         }
3726
3727                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3728                                 idev = ipv6_add_dev(dev);
3729
3730                         if (!IS_ERR_OR_NULL(idev)) {
3731                                 idev->if_flags |= IF_READY;
3732                                 run_pending = 1;
3733                         }
3734                 } else if (event == NETDEV_CHANGE) {
3735                         if (!addrconf_link_ready(dev)) {
3736                                 /* device is still not ready. */
3737                                 rt6_sync_down_dev(dev, event);
3738                                 break;
3739                         }
3740
3741                         if (!IS_ERR_OR_NULL(idev)) {
3742                                 if (idev->if_flags & IF_READY) {
3743                                         /* device is already configured -
3744                                          * but resend MLD reports, we might
3745                                          * have roamed and need to update
3746                                          * multicast snooping switches
3747                                          */
3748                                         ipv6_mc_up(idev);
3749                                         change_info = ptr;
3750                                         if (change_info->flags_changed & IFF_NOARP)
3751                                                 addrconf_dad_run(idev, true);
3752                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3753                                         break;
3754                                 }
3755                                 idev->if_flags |= IF_READY;
3756                         }
3757
3758                         pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3759                                  dev->name);
3760
3761                         run_pending = 1;
3762                 }
3763
3764                 addrconf_init_auto_addrs(dev);
3765
3766                 if (!IS_ERR_OR_NULL(idev)) {
3767                         if (run_pending)
3768                                 addrconf_dad_run(idev, false);
3769
3770                         /* Device has an address by now */
3771                         rt6_sync_up(dev, RTNH_F_DEAD);
3772
3773                         /*
3774                          * If the MTU changed during the interface down,
3775                          * when the interface up, the changed MTU must be
3776                          * reflected in the idev as well as routers.
3777                          */
3778                         if (idev->cnf.mtu6 != dev->mtu &&
3779                             dev->mtu >= IPV6_MIN_MTU) {
3780                                 rt6_mtu_change(dev, dev->mtu);
3781                                 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3782                         }
3783                         WRITE_ONCE(idev->tstamp, jiffies);
3784                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3785
3786                         /*
3787                          * If the changed mtu during down is lower than
3788                          * IPV6_MIN_MTU stop IPv6 on this interface.
3789                          */
3790                         if (dev->mtu < IPV6_MIN_MTU)
3791                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3792                 }
3793                 break;
3794
3795         case NETDEV_DOWN:
3796         case NETDEV_UNREGISTER:
3797                 /*
3798                  *      Remove all addresses from this interface.
3799                  */
3800                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3801                 break;
3802
3803         case NETDEV_CHANGENAME:
3804                 if (idev) {
3805                         snmp6_unregister_dev(idev);
3806                         addrconf_sysctl_unregister(idev);
3807                         err = addrconf_sysctl_register(idev);
3808                         if (err)
3809                                 return notifier_from_errno(err);
3810                         err = snmp6_register_dev(idev);
3811                         if (err) {
3812                                 addrconf_sysctl_unregister(idev);
3813                                 return notifier_from_errno(err);
3814                         }
3815                 }
3816                 break;
3817
3818         case NETDEV_PRE_TYPE_CHANGE:
3819         case NETDEV_POST_TYPE_CHANGE:
3820                 if (idev)
3821                         addrconf_type_change(dev, event);
3822                 break;
3823
3824         case NETDEV_CHANGEUPPER:
3825                 info = ptr;
3826
3827                 /* flush all routes if dev is linked to or unlinked from
3828                  * an L3 master device (e.g., VRF)
3829                  */
3830                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3831                         addrconf_ifdown(dev, false);
3832         }
3833
3834         return NOTIFY_OK;
3835 }
3836
3837 /*
3838  *      addrconf module should be notified of a device going up
3839  */
3840 static struct notifier_block ipv6_dev_notf = {
3841         .notifier_call = addrconf_notify,
3842         .priority = ADDRCONF_NOTIFY_PRIORITY,
3843 };
3844
3845 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3846 {
3847         struct inet6_dev *idev;
3848         ASSERT_RTNL();
3849
3850         idev = __in6_dev_get(dev);
3851
3852         if (event == NETDEV_POST_TYPE_CHANGE)
3853                 ipv6_mc_remap(idev);
3854         else if (event == NETDEV_PRE_TYPE_CHANGE)
3855                 ipv6_mc_unmap(idev);
3856 }
3857
3858 static bool addr_is_local(const struct in6_addr *addr)
3859 {
3860         return ipv6_addr_type(addr) &
3861                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3862 }
3863
3864 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3865 {
3866         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3867         struct net *net = dev_net(dev);
3868         struct inet6_dev *idev;
3869         struct inet6_ifaddr *ifa;
3870         LIST_HEAD(tmp_addr_list);
3871         bool keep_addr = false;
3872         bool was_ready;
3873         int state, i;
3874
3875         ASSERT_RTNL();
3876
3877         rt6_disable_ip(dev, event);
3878
3879         idev = __in6_dev_get(dev);
3880         if (!idev)
3881                 return -ENODEV;
3882
3883         /*
3884          * Step 1: remove reference to ipv6 device from parent device.
3885          *         Do not dev_put!
3886          */
3887         if (unregister) {
3888                 idev->dead = 1;
3889
3890                 /* protected by rtnl_lock */
3891                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3892
3893                 /* Step 1.5: remove snmp6 entry */
3894                 snmp6_unregister_dev(idev);
3895
3896         }
3897
3898         /* combine the user config with event to determine if permanent
3899          * addresses are to be removed from address hash table
3900          */
3901         if (!unregister && !idev->cnf.disable_ipv6) {
3902                 /* aggregate the system setting and interface setting */
3903                 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3904
3905                 if (!_keep_addr)
3906                         _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3907
3908                 keep_addr = (_keep_addr > 0);
3909         }
3910
3911         /* Step 2: clear hash table */
3912         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3913                 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3914
3915                 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3916 restart:
3917                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3918                         if (ifa->idev == idev) {
3919                                 addrconf_del_dad_work(ifa);
3920                                 /* combined flag + permanent flag decide if
3921                                  * address is retained on a down event
3922                                  */
3923                                 if (!keep_addr ||
3924                                     !(ifa->flags & IFA_F_PERMANENT) ||
3925                                     addr_is_local(&ifa->addr)) {
3926                                         hlist_del_init_rcu(&ifa->addr_lst);
3927                                         goto restart;
3928                                 }
3929                         }
3930                 }
3931                 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3932         }
3933
3934         write_lock_bh(&idev->lock);
3935
3936         addrconf_del_rs_timer(idev);
3937
3938         /* Step 2: clear flags for stateless addrconf, repeated down
3939          *         detection
3940          */
3941         was_ready = idev->if_flags & IF_READY;
3942         if (!unregister)
3943                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3944
3945         /* Step 3: clear tempaddr list */
3946         while (!list_empty(&idev->tempaddr_list)) {
3947                 ifa = list_first_entry(&idev->tempaddr_list,
3948                                        struct inet6_ifaddr, tmp_list);
3949                 list_del(&ifa->tmp_list);
3950                 write_unlock_bh(&idev->lock);
3951                 spin_lock_bh(&ifa->lock);
3952
3953                 if (ifa->ifpub) {
3954                         in6_ifa_put(ifa->ifpub);
3955                         ifa->ifpub = NULL;
3956                 }
3957                 spin_unlock_bh(&ifa->lock);
3958                 in6_ifa_put(ifa);
3959                 write_lock_bh(&idev->lock);
3960         }
3961
3962         list_for_each_entry(ifa, &idev->addr_list, if_list)
3963                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3964         write_unlock_bh(&idev->lock);
3965
3966         while (!list_empty(&tmp_addr_list)) {
3967                 struct fib6_info *rt = NULL;
3968                 bool keep;
3969
3970                 ifa = list_first_entry(&tmp_addr_list,
3971                                        struct inet6_ifaddr, if_list_aux);
3972                 list_del(&ifa->if_list_aux);
3973
3974                 addrconf_del_dad_work(ifa);
3975
3976                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3977                         !addr_is_local(&ifa->addr);
3978
3979                 spin_lock_bh(&ifa->lock);
3980
3981                 if (keep) {
3982                         /* set state to skip the notifier below */
3983                         state = INET6_IFADDR_STATE_DEAD;
3984                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3985                         if (!(ifa->flags & IFA_F_NODAD))
3986                                 ifa->flags |= IFA_F_TENTATIVE;
3987
3988                         rt = ifa->rt;
3989                         ifa->rt = NULL;
3990                 } else {
3991                         state = ifa->state;
3992                         ifa->state = INET6_IFADDR_STATE_DEAD;
3993                 }
3994
3995                 spin_unlock_bh(&ifa->lock);
3996
3997                 if (rt)
3998                         ip6_del_rt(net, rt, false);
3999
4000                 if (state != INET6_IFADDR_STATE_DEAD) {
4001                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
4002                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
4003                 } else {
4004                         if (idev->cnf.forwarding)
4005                                 addrconf_leave_anycast(ifa);
4006                         addrconf_leave_solict(ifa->idev, &ifa->addr);
4007                 }
4008
4009                 if (!keep) {
4010                         write_lock_bh(&idev->lock);
4011                         list_del_rcu(&ifa->if_list);
4012                         write_unlock_bh(&idev->lock);
4013                         in6_ifa_put(ifa);
4014                 }
4015         }
4016
4017         /* Step 5: Discard anycast and multicast list */
4018         if (unregister) {
4019                 ipv6_ac_destroy_dev(idev);
4020                 ipv6_mc_destroy_dev(idev);
4021         } else if (was_ready) {
4022                 ipv6_mc_down(idev);
4023         }
4024
4025         WRITE_ONCE(idev->tstamp, jiffies);
4026         idev->ra_mtu = 0;
4027
4028         /* Last: Shot the device (if unregistered) */
4029         if (unregister) {
4030                 addrconf_sysctl_unregister(idev);
4031                 neigh_parms_release(&nd_tbl, idev->nd_parms);
4032                 neigh_ifdown(&nd_tbl, dev);
4033                 in6_dev_put(idev);
4034         }
4035         return 0;
4036 }
4037
4038 static void addrconf_rs_timer(struct timer_list *t)
4039 {
4040         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
4041         struct net_device *dev = idev->dev;
4042         struct in6_addr lladdr;
4043         int rtr_solicits;
4044
4045         write_lock(&idev->lock);
4046         if (idev->dead || !(idev->if_flags & IF_READY))
4047                 goto out;
4048
4049         if (!ipv6_accept_ra(idev))
4050                 goto out;
4051
4052         /* Announcement received after solicitation was sent */
4053         if (idev->if_flags & IF_RA_RCVD)
4054                 goto out;
4055
4056         rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4057
4058         if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4059                 write_unlock(&idev->lock);
4060                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4061                         ndisc_send_rs(dev, &lladdr,
4062                                       &in6addr_linklocal_allrouters);
4063                 else
4064                         goto put;
4065
4066                 write_lock(&idev->lock);
4067                 idev->rs_interval = rfc3315_s14_backoff_update(
4068                                 idev->rs_interval,
4069                                 READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4070                 /* The wait after the last probe can be shorter */
4071                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4072                                              READ_ONCE(idev->cnf.rtr_solicits)) ?
4073                                       READ_ONCE(idev->cnf.rtr_solicit_delay) :
4074                                       idev->rs_interval);
4075         } else {
4076                 /*
4077                  * Note: we do not support deprecated "all on-link"
4078                  * assumption any longer.
4079                  */
4080                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4081         }
4082
4083 out:
4084         write_unlock(&idev->lock);
4085 put:
4086         in6_dev_put(idev);
4087 }
4088
4089 /*
4090  *      Duplicate Address Detection
4091  */
4092 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4093 {
4094         struct inet6_dev *idev = ifp->idev;
4095         unsigned long rand_num;
4096         u64 nonce;
4097
4098         if (ifp->flags & IFA_F_OPTIMISTIC)
4099                 rand_num = 0;
4100         else
4101                 rand_num = get_random_u32_below(
4102                                 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4103
4104         nonce = 0;
4105         if (READ_ONCE(idev->cnf.enhanced_dad) ||
4106             READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4107                 do
4108                         get_random_bytes(&nonce, 6);
4109                 while (nonce == 0);
4110         }
4111         ifp->dad_nonce = nonce;
4112         ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4113         addrconf_mod_dad_work(ifp, rand_num);
4114 }
4115
4116 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4117 {
4118         struct inet6_dev *idev = ifp->idev;
4119         struct net_device *dev = idev->dev;
4120         bool bump_id, notify = false;
4121         struct net *net;
4122
4123         addrconf_join_solict(dev, &ifp->addr);
4124
4125         read_lock_bh(&idev->lock);
4126         spin_lock(&ifp->lock);
4127         if (ifp->state == INET6_IFADDR_STATE_DEAD)
4128                 goto out;
4129
4130         net = dev_net(dev);
4131         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4132             (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4133              READ_ONCE(idev->cnf.accept_dad) < 1) ||
4134             !(ifp->flags&IFA_F_TENTATIVE) ||
4135             ifp->flags & IFA_F_NODAD) {
4136                 bool send_na = false;
4137
4138                 if (ifp->flags & IFA_F_TENTATIVE &&
4139                     !(ifp->flags & IFA_F_OPTIMISTIC))
4140                         send_na = true;
4141                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4142                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4143                 spin_unlock(&ifp->lock);
4144                 read_unlock_bh(&idev->lock);
4145
4146                 addrconf_dad_completed(ifp, bump_id, send_na);
4147                 return;
4148         }
4149
4150         if (!(idev->if_flags & IF_READY)) {
4151                 spin_unlock(&ifp->lock);
4152                 read_unlock_bh(&idev->lock);
4153                 /*
4154                  * If the device is not ready:
4155                  * - keep it tentative if it is a permanent address.
4156                  * - otherwise, kill it.
4157                  */
4158                 in6_ifa_hold(ifp);
4159                 addrconf_dad_stop(ifp, 0);
4160                 return;
4161         }
4162
4163         /*
4164          * Optimistic nodes can start receiving
4165          * Frames right away
4166          */
4167         if (ifp->flags & IFA_F_OPTIMISTIC) {
4168                 ip6_ins_rt(net, ifp->rt);
4169                 if (ipv6_use_optimistic_addr(net, idev)) {
4170                         /* Because optimistic nodes can use this address,
4171                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4172                          */
4173                         notify = true;
4174                 }
4175         }
4176
4177         addrconf_dad_kick(ifp);
4178 out:
4179         spin_unlock(&ifp->lock);
4180         read_unlock_bh(&idev->lock);
4181         if (notify)
4182                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4183 }
4184
4185 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4186 {
4187         bool begin_dad = false;
4188
4189         spin_lock_bh(&ifp->lock);
4190         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4191                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4192                 begin_dad = true;
4193         }
4194         spin_unlock_bh(&ifp->lock);
4195
4196         if (begin_dad)
4197                 addrconf_mod_dad_work(ifp, 0);
4198 }
4199
4200 static void addrconf_dad_work(struct work_struct *w)
4201 {
4202         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4203                                                 struct inet6_ifaddr,
4204                                                 dad_work);
4205         struct inet6_dev *idev = ifp->idev;
4206         bool bump_id, disable_ipv6 = false;
4207         struct in6_addr mcaddr;
4208
4209         enum {
4210                 DAD_PROCESS,
4211                 DAD_BEGIN,
4212                 DAD_ABORT,
4213         } action = DAD_PROCESS;
4214
4215         rtnl_lock();
4216
4217         spin_lock_bh(&ifp->lock);
4218         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4219                 action = DAD_BEGIN;
4220                 ifp->state = INET6_IFADDR_STATE_DAD;
4221         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4222                 action = DAD_ABORT;
4223                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4224
4225                 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->accept_dad) > 1 ||
4226                      READ_ONCE(idev->cnf.accept_dad) > 1) &&
4227                     !idev->cnf.disable_ipv6 &&
4228                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4229                         struct in6_addr addr;
4230
4231                         addr.s6_addr32[0] = htonl(0xfe800000);
4232                         addr.s6_addr32[1] = 0;
4233
4234                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4235                             ipv6_addr_equal(&ifp->addr, &addr)) {
4236                                 /* DAD failed for link-local based on MAC */
4237                                 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4238
4239                                 pr_info("%s: IPv6 being disabled!\n",
4240                                         ifp->idev->dev->name);
4241                                 disable_ipv6 = true;
4242                         }
4243                 }
4244         }
4245         spin_unlock_bh(&ifp->lock);
4246
4247         if (action == DAD_BEGIN) {
4248                 addrconf_dad_begin(ifp);
4249                 goto out;
4250         } else if (action == DAD_ABORT) {
4251                 in6_ifa_hold(ifp);
4252                 addrconf_dad_stop(ifp, 1);
4253                 if (disable_ipv6)
4254                         addrconf_ifdown(idev->dev, false);
4255                 goto out;
4256         }
4257
4258         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4259                 goto out;
4260
4261         write_lock_bh(&idev->lock);
4262         if (idev->dead || !(idev->if_flags & IF_READY)) {
4263                 write_unlock_bh(&idev->lock);
4264                 goto out;
4265         }
4266
4267         spin_lock(&ifp->lock);
4268         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4269                 spin_unlock(&ifp->lock);
4270                 write_unlock_bh(&idev->lock);
4271                 goto out;
4272         }
4273
4274         if (ifp->dad_probes == 0) {
4275                 bool send_na = false;
4276
4277                 /*
4278                  * DAD was successful
4279                  */
4280
4281                 if (ifp->flags & IFA_F_TENTATIVE &&
4282                     !(ifp->flags & IFA_F_OPTIMISTIC))
4283                         send_na = true;
4284                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4285                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4286                 spin_unlock(&ifp->lock);
4287                 write_unlock_bh(&idev->lock);
4288
4289                 addrconf_dad_completed(ifp, bump_id, send_na);
4290
4291                 goto out;
4292         }
4293
4294         ifp->dad_probes--;
4295         addrconf_mod_dad_work(ifp,
4296                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4297                                   HZ/100));
4298         spin_unlock(&ifp->lock);
4299         write_unlock_bh(&idev->lock);
4300
4301         /* send a neighbour solicitation for our addr */
4302         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4303         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4304                       ifp->dad_nonce);
4305 out:
4306         in6_ifa_put(ifp);
4307         rtnl_unlock();
4308 }
4309
4310 /* ifp->idev must be at least read locked */
4311 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4312 {
4313         struct inet6_ifaddr *ifpiter;
4314         struct inet6_dev *idev = ifp->idev;
4315
4316         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4317                 if (ifpiter->scope > IFA_LINK)
4318                         break;
4319                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4320                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4321                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4322                     IFA_F_PERMANENT)
4323                         return false;
4324         }
4325         return true;
4326 }
4327
4328 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4329                                    bool send_na)
4330 {
4331         struct net_device *dev = ifp->idev->dev;
4332         struct in6_addr lladdr;
4333         bool send_rs, send_mld;
4334
4335         addrconf_del_dad_work(ifp);
4336
4337         /*
4338          *      Configure the address for reception. Now it is valid.
4339          */
4340
4341         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4342
4343         /* If added prefix is link local and we are prepared to process
4344            router advertisements, start sending router solicitations.
4345          */
4346
4347         read_lock_bh(&ifp->idev->lock);
4348         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4349         send_rs = send_mld &&
4350                   ipv6_accept_ra(ifp->idev) &&
4351                   READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4352                   (dev->flags & IFF_LOOPBACK) == 0 &&
4353                   (dev->type != ARPHRD_TUNNEL) &&
4354                   !netif_is_team_port(dev);
4355         read_unlock_bh(&ifp->idev->lock);
4356
4357         /* While dad is in progress mld report's source address is in6_addrany.
4358          * Resend with proper ll now.
4359          */
4360         if (send_mld)
4361                 ipv6_mc_dad_complete(ifp->idev);
4362
4363         /* send unsolicited NA if enabled */
4364         if (send_na &&
4365             (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4366              READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4367                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4368                               /*router=*/ !!ifp->idev->cnf.forwarding,
4369                               /*solicited=*/ false, /*override=*/ true,
4370                               /*inc_opt=*/ true);
4371         }
4372
4373         if (send_rs) {
4374                 /*
4375                  *      If a host as already performed a random delay
4376                  *      [...] as part of DAD [...] there is no need
4377                  *      to delay again before sending the first RS
4378                  */
4379                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4380                         return;
4381                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4382
4383                 write_lock_bh(&ifp->idev->lock);
4384                 spin_lock(&ifp->lock);
4385                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4386                         READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4387                 ifp->idev->rs_probes = 1;
4388                 ifp->idev->if_flags |= IF_RS_SENT;
4389                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4390                 spin_unlock(&ifp->lock);
4391                 write_unlock_bh(&ifp->idev->lock);
4392         }
4393
4394         if (bump_id)
4395                 rt_genid_bump_ipv6(dev_net(dev));
4396
4397         /* Make sure that a new temporary address will be created
4398          * before this temporary address becomes deprecated.
4399          */
4400         if (ifp->flags & IFA_F_TEMPORARY)
4401                 addrconf_verify_rtnl(dev_net(dev));
4402 }
4403
4404 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4405 {
4406         struct inet6_ifaddr *ifp;
4407
4408         read_lock_bh(&idev->lock);
4409         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4410                 spin_lock(&ifp->lock);
4411                 if ((ifp->flags & IFA_F_TENTATIVE &&
4412                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4413                         if (restart)
4414                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4415                         addrconf_dad_kick(ifp);
4416                 }
4417                 spin_unlock(&ifp->lock);
4418         }
4419         read_unlock_bh(&idev->lock);
4420 }
4421
4422 #ifdef CONFIG_PROC_FS
4423 struct if6_iter_state {
4424         struct seq_net_private p;
4425         int bucket;
4426         int offset;
4427 };
4428
4429 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4430 {
4431         struct if6_iter_state *state = seq->private;
4432         struct net *net = seq_file_net(seq);
4433         struct inet6_ifaddr *ifa = NULL;
4434         int p = 0;
4435
4436         /* initial bucket if pos is 0 */
4437         if (pos == 0) {
4438                 state->bucket = 0;
4439                 state->offset = 0;
4440         }
4441
4442         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4443                 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4444                                          addr_lst) {
4445                         /* sync with offset */
4446                         if (p < state->offset) {
4447                                 p++;
4448                                 continue;
4449                         }
4450                         return ifa;
4451                 }
4452
4453                 /* prepare for next bucket */
4454                 state->offset = 0;
4455                 p = 0;
4456         }
4457         return NULL;
4458 }
4459
4460 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4461                                          struct inet6_ifaddr *ifa)
4462 {
4463         struct if6_iter_state *state = seq->private;
4464         struct net *net = seq_file_net(seq);
4465
4466         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4467                 state->offset++;
4468                 return ifa;
4469         }
4470
4471         state->offset = 0;
4472         while (++state->bucket < IN6_ADDR_HSIZE) {
4473                 hlist_for_each_entry_rcu(ifa,
4474                                      &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4475                         return ifa;
4476                 }
4477         }
4478
4479         return NULL;
4480 }
4481
4482 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4483         __acquires(rcu)
4484 {
4485         rcu_read_lock();
4486         return if6_get_first(seq, *pos);
4487 }
4488
4489 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4490 {
4491         struct inet6_ifaddr *ifa;
4492
4493         ifa = if6_get_next(seq, v);
4494         ++*pos;
4495         return ifa;
4496 }
4497
4498 static void if6_seq_stop(struct seq_file *seq, void *v)
4499         __releases(rcu)
4500 {
4501         rcu_read_unlock();
4502 }
4503
4504 static int if6_seq_show(struct seq_file *seq, void *v)
4505 {
4506         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4507         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4508                    &ifp->addr,
4509                    ifp->idev->dev->ifindex,
4510                    ifp->prefix_len,
4511                    ifp->scope,
4512                    (u8) ifp->flags,
4513                    ifp->idev->dev->name);
4514         return 0;
4515 }
4516
4517 static const struct seq_operations if6_seq_ops = {
4518         .start  = if6_seq_start,
4519         .next   = if6_seq_next,
4520         .show   = if6_seq_show,
4521         .stop   = if6_seq_stop,
4522 };
4523
4524 static int __net_init if6_proc_net_init(struct net *net)
4525 {
4526         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4527                         sizeof(struct if6_iter_state)))
4528                 return -ENOMEM;
4529         return 0;
4530 }
4531
4532 static void __net_exit if6_proc_net_exit(struct net *net)
4533 {
4534         remove_proc_entry("if_inet6", net->proc_net);
4535 }
4536
4537 static struct pernet_operations if6_proc_net_ops = {
4538         .init = if6_proc_net_init,
4539         .exit = if6_proc_net_exit,
4540 };
4541
4542 int __init if6_proc_init(void)
4543 {
4544         return register_pernet_subsys(&if6_proc_net_ops);
4545 }
4546
4547 void if6_proc_exit(void)
4548 {
4549         unregister_pernet_subsys(&if6_proc_net_ops);
4550 }
4551 #endif  /* CONFIG_PROC_FS */
4552
4553 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4554 /* Check if address is a home address configured on any interface. */
4555 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4556 {
4557         unsigned int hash = inet6_addr_hash(net, addr);
4558         struct inet6_ifaddr *ifp = NULL;
4559         int ret = 0;
4560
4561         rcu_read_lock();
4562         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4563                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4564                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4565                         ret = 1;
4566                         break;
4567                 }
4568         }
4569         rcu_read_unlock();
4570         return ret;
4571 }
4572 #endif
4573
4574 /* RFC6554 has some algorithm to avoid loops in segment routing by
4575  * checking if the segments contains any of a local interface address.
4576  *
4577  * Quote:
4578  *
4579  * To detect loops in the SRH, a router MUST determine if the SRH
4580  * includes multiple addresses assigned to any interface on that router.
4581  * If such addresses appear more than once and are separated by at least
4582  * one address not assigned to that router.
4583  */
4584 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4585                           unsigned char nsegs)
4586 {
4587         const struct in6_addr *addr;
4588         int i, ret = 0, found = 0;
4589         struct inet6_ifaddr *ifp;
4590         bool separated = false;
4591         unsigned int hash;
4592         bool hash_found;
4593
4594         rcu_read_lock();
4595         for (i = 0; i < nsegs; i++) {
4596                 addr = &segs[i];
4597                 hash = inet6_addr_hash(net, addr);
4598
4599                 hash_found = false;
4600                 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4601
4602                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4603                                 hash_found = true;
4604                                 break;
4605                         }
4606                 }
4607
4608                 if (hash_found) {
4609                         if (found > 1 && separated) {
4610                                 ret = 1;
4611                                 break;
4612                         }
4613
4614                         separated = false;
4615                         found++;
4616                 } else {
4617                         separated = true;
4618                 }
4619         }
4620         rcu_read_unlock();
4621
4622         return ret;
4623 }
4624
4625 /*
4626  *      Periodic address status verification
4627  */
4628
4629 static void addrconf_verify_rtnl(struct net *net)
4630 {
4631         unsigned long now, next, next_sec, next_sched;
4632         struct inet6_ifaddr *ifp;
4633         int i;
4634
4635         ASSERT_RTNL();
4636
4637         rcu_read_lock_bh();
4638         now = jiffies;
4639         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4640
4641         cancel_delayed_work(&net->ipv6.addr_chk_work);
4642
4643         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4644 restart:
4645                 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4646                         unsigned long age;
4647
4648                         /* When setting preferred_lft to a value not zero or
4649                          * infinity, while valid_lft is infinity
4650                          * IFA_F_PERMANENT has a non-infinity life time.
4651                          */
4652                         if ((ifp->flags & IFA_F_PERMANENT) &&
4653                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4654                                 continue;
4655
4656                         spin_lock(&ifp->lock);
4657                         /* We try to batch several events at once. */
4658                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4659
4660                         if ((ifp->flags&IFA_F_TEMPORARY) &&
4661                             !(ifp->flags&IFA_F_TENTATIVE) &&
4662                             ifp->prefered_lft != INFINITY_LIFE_TIME &&
4663                             !ifp->regen_count && ifp->ifpub) {
4664                                 /* This is a non-regenerated temporary addr. */
4665
4666                                 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4667
4668                                 if (age + regen_advance >= ifp->prefered_lft) {
4669                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4670                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4671                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4672
4673                                         ifp->regen_count++;
4674                                         in6_ifa_hold(ifp);
4675                                         in6_ifa_hold(ifpub);
4676                                         spin_unlock(&ifp->lock);
4677
4678                                         spin_lock(&ifpub->lock);
4679                                         ifpub->regen_count = 0;
4680                                         spin_unlock(&ifpub->lock);
4681                                         rcu_read_unlock_bh();
4682                                         ipv6_create_tempaddr(ifpub, true);
4683                                         in6_ifa_put(ifpub);
4684                                         in6_ifa_put(ifp);
4685                                         rcu_read_lock_bh();
4686                                         goto restart;
4687                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4688                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4689                         }
4690
4691                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4692                             age >= ifp->valid_lft) {
4693                                 spin_unlock(&ifp->lock);
4694                                 in6_ifa_hold(ifp);
4695                                 rcu_read_unlock_bh();
4696                                 ipv6_del_addr(ifp);
4697                                 rcu_read_lock_bh();
4698                                 goto restart;
4699                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4700                                 spin_unlock(&ifp->lock);
4701                                 continue;
4702                         } else if (age >= ifp->prefered_lft) {
4703                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4704                                 int deprecate = 0;
4705
4706                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4707                                         deprecate = 1;
4708                                         ifp->flags |= IFA_F_DEPRECATED;
4709                                 }
4710
4711                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4712                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4713                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4714
4715                                 spin_unlock(&ifp->lock);
4716
4717                                 if (deprecate) {
4718                                         in6_ifa_hold(ifp);
4719
4720                                         ipv6_ifa_notify(0, ifp);
4721                                         in6_ifa_put(ifp);
4722                                         goto restart;
4723                                 }
4724                         } else {
4725                                 /* ifp->prefered_lft <= ifp->valid_lft */
4726                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4727                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4728                                 spin_unlock(&ifp->lock);
4729                         }
4730                 }
4731         }
4732
4733         next_sec = round_jiffies_up(next);
4734         next_sched = next;
4735
4736         /* If rounded timeout is accurate enough, accept it. */
4737         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4738                 next_sched = next_sec;
4739
4740         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4741         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4742                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4743
4744         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4745                  now, next, next_sec, next_sched);
4746         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4747         rcu_read_unlock_bh();
4748 }
4749
4750 static void addrconf_verify_work(struct work_struct *w)
4751 {
4752         struct net *net = container_of(to_delayed_work(w), struct net,
4753                                        ipv6.addr_chk_work);
4754
4755         rtnl_lock();
4756         addrconf_verify_rtnl(net);
4757         rtnl_unlock();
4758 }
4759
4760 static void addrconf_verify(struct net *net)
4761 {
4762         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4763 }
4764
4765 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4766                                      struct in6_addr **peer_pfx)
4767 {
4768         struct in6_addr *pfx = NULL;
4769
4770         *peer_pfx = NULL;
4771
4772         if (addr)
4773                 pfx = nla_data(addr);
4774
4775         if (local) {
4776                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4777                         *peer_pfx = pfx;
4778                 pfx = nla_data(local);
4779         }
4780
4781         return pfx;
4782 }
4783
4784 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4785         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4786         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4787         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4788         [IFA_FLAGS]             = { .len = sizeof(u32) },
4789         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4790         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4791         [IFA_PROTO]             = { .type = NLA_U8 },
4792 };
4793
4794 static int
4795 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4796                   struct netlink_ext_ack *extack)
4797 {
4798         struct net *net = sock_net(skb->sk);
4799         struct ifaddrmsg *ifm;
4800         struct nlattr *tb[IFA_MAX+1];
4801         struct in6_addr *pfx, *peer_pfx;
4802         u32 ifa_flags;
4803         int err;
4804
4805         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4806                                      ifa_ipv6_policy, extack);
4807         if (err < 0)
4808                 return err;
4809
4810         ifm = nlmsg_data(nlh);
4811         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4812         if (!pfx)
4813                 return -EINVAL;
4814
4815         ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4816
4817         /* We ignore other flags so far. */
4818         ifa_flags &= IFA_F_MANAGETEMPADDR;
4819
4820         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4821                               ifm->ifa_prefixlen, extack);
4822 }
4823
4824 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp,
4825                                unsigned long expires, u32 flags,
4826                                bool modify_peer)
4827 {
4828         struct fib6_table *table;
4829         struct fib6_info *f6i;
4830         u32 prio;
4831
4832         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4833                                         ifp->prefix_len,
4834                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4835         if (!f6i)
4836                 return -ENOENT;
4837
4838         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4839         if (f6i->fib6_metric != prio) {
4840                 /* delete old one */
4841                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4842
4843                 /* add new one */
4844                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4845                                       ifp->prefix_len,
4846                                       ifp->rt_priority, ifp->idev->dev,
4847                                       expires, flags, GFP_KERNEL);
4848                 return 0;
4849         }
4850         if (f6i != net->ipv6.fib6_null_entry) {
4851                 table = f6i->fib6_table;
4852                 spin_lock_bh(&table->tb6_lock);
4853
4854                 if (!(flags & RTF_EXPIRES)) {
4855                         fib6_clean_expires(f6i);
4856                         fib6_remove_gc_list(f6i);
4857                 } else {
4858                         fib6_set_expires(f6i, expires);
4859                         fib6_add_gc_list(f6i);
4860                 }
4861
4862                 spin_unlock_bh(&table->tb6_lock);
4863         }
4864         fib6_info_release(f6i);
4865
4866         return 0;
4867 }
4868
4869 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4870                              struct ifa6_config *cfg)
4871 {
4872         u32 flags;
4873         clock_t expires;
4874         unsigned long timeout;
4875         bool was_managetempaddr;
4876         bool had_prefixroute;
4877         bool new_peer = false;
4878
4879         ASSERT_RTNL();
4880
4881         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4882                 return -EINVAL;
4883
4884         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4885             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4886                 return -EINVAL;
4887
4888         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4889                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4890
4891         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4892         if (addrconf_finite_timeout(timeout)) {
4893                 expires = jiffies_to_clock_t(timeout * HZ);
4894                 cfg->valid_lft = timeout;
4895                 flags = RTF_EXPIRES;
4896         } else {
4897                 expires = 0;
4898                 flags = 0;
4899                 cfg->ifa_flags |= IFA_F_PERMANENT;
4900         }
4901
4902         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4903         if (addrconf_finite_timeout(timeout)) {
4904                 if (timeout == 0)
4905                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4906                 cfg->preferred_lft = timeout;
4907         }
4908
4909         if (cfg->peer_pfx &&
4910             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4911                 if (!ipv6_addr_any(&ifp->peer_addr))
4912                         cleanup_prefix_route(ifp, expires, true, true);
4913                 new_peer = true;
4914         }
4915
4916         spin_lock_bh(&ifp->lock);
4917         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4918         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4919                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4920         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4921                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4922                         IFA_F_NOPREFIXROUTE);
4923         ifp->flags |= cfg->ifa_flags;
4924         WRITE_ONCE(ifp->tstamp, jiffies);
4925         WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4926         WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4927         WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4928
4929         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4930                 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4931
4932         if (new_peer)
4933                 ifp->peer_addr = *cfg->peer_pfx;
4934
4935         spin_unlock_bh(&ifp->lock);
4936         if (!(ifp->flags&IFA_F_TENTATIVE))
4937                 ipv6_ifa_notify(0, ifp);
4938
4939         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4940                 int rc = -ENOENT;
4941
4942                 if (had_prefixroute)
4943                         rc = modify_prefix_route(net, ifp, expires, flags, false);
4944
4945                 /* prefix route could have been deleted; if so restore it */
4946                 if (rc == -ENOENT) {
4947                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4948                                               ifp->rt_priority, ifp->idev->dev,
4949                                               expires, flags, GFP_KERNEL);
4950                 }
4951
4952                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4953                         rc = modify_prefix_route(net, ifp, expires, flags, true);
4954
4955                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4956                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4957                                               ifp->rt_priority, ifp->idev->dev,
4958                                               expires, flags, GFP_KERNEL);
4959                 }
4960         } else if (had_prefixroute) {
4961                 enum cleanup_prefix_rt_t action;
4962                 unsigned long rt_expires;
4963
4964                 write_lock_bh(&ifp->idev->lock);
4965                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4966                 write_unlock_bh(&ifp->idev->lock);
4967
4968                 if (action != CLEANUP_PREFIX_RT_NOP) {
4969                         cleanup_prefix_route(ifp, rt_expires,
4970                                 action == CLEANUP_PREFIX_RT_DEL, false);
4971                 }
4972         }
4973
4974         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4975                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4976                         delete_tempaddrs(ifp->idev, ifp);
4977                 else
4978                         manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4979                                          cfg->preferred_lft, !was_managetempaddr,
4980                                          jiffies);
4981         }
4982
4983         addrconf_verify_rtnl(net);
4984
4985         return 0;
4986 }
4987
4988 static int
4989 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4990                   struct netlink_ext_ack *extack)
4991 {
4992         struct net *net = sock_net(skb->sk);
4993         struct ifaddrmsg *ifm;
4994         struct nlattr *tb[IFA_MAX+1];
4995         struct in6_addr *peer_pfx;
4996         struct inet6_ifaddr *ifa;
4997         struct net_device *dev;
4998         struct inet6_dev *idev;
4999         struct ifa6_config cfg;
5000         int err;
5001
5002         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5003                                      ifa_ipv6_policy, extack);
5004         if (err < 0)
5005                 return err;
5006
5007         memset(&cfg, 0, sizeof(cfg));
5008
5009         ifm = nlmsg_data(nlh);
5010         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
5011         if (!cfg.pfx)
5012                 return -EINVAL;
5013
5014         cfg.peer_pfx = peer_pfx;
5015         cfg.plen = ifm->ifa_prefixlen;
5016         if (tb[IFA_RT_PRIORITY])
5017                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
5018
5019         if (tb[IFA_PROTO])
5020                 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
5021
5022         cfg.valid_lft = INFINITY_LIFE_TIME;
5023         cfg.preferred_lft = INFINITY_LIFE_TIME;
5024
5025         if (tb[IFA_CACHEINFO]) {
5026                 struct ifa_cacheinfo *ci;
5027
5028                 ci = nla_data(tb[IFA_CACHEINFO]);
5029                 cfg.valid_lft = ci->ifa_valid;
5030                 cfg.preferred_lft = ci->ifa_prefered;
5031         }
5032
5033         dev =  __dev_get_by_index(net, ifm->ifa_index);
5034         if (!dev) {
5035                 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5036                 return -ENODEV;
5037         }
5038
5039         cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
5040
5041         /* We ignore other flags so far. */
5042         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
5043                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
5044                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
5045
5046         idev = ipv6_find_idev(dev);
5047         if (IS_ERR(idev))
5048                 return PTR_ERR(idev);
5049
5050         if (!ipv6_allow_optimistic_dad(net, idev))
5051                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5052
5053         if (cfg.ifa_flags & IFA_F_NODAD &&
5054             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5055                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5056                 return -EINVAL;
5057         }
5058
5059         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5060         if (!ifa) {
5061                 /*
5062                  * It would be best to check for !NLM_F_CREATE here but
5063                  * userspace already relies on not having to provide this.
5064                  */
5065                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
5066         }
5067
5068         if (nlh->nlmsg_flags & NLM_F_EXCL ||
5069             !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5070                 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5071                 err = -EEXIST;
5072         } else {
5073                 err = inet6_addr_modify(net, ifa, &cfg);
5074         }
5075
5076         in6_ifa_put(ifa);
5077
5078         return err;
5079 }
5080
5081 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5082                           u8 scope, int ifindex)
5083 {
5084         struct ifaddrmsg *ifm;
5085
5086         ifm = nlmsg_data(nlh);
5087         ifm->ifa_family = AF_INET6;
5088         ifm->ifa_prefixlen = prefixlen;
5089         ifm->ifa_flags = flags;
5090         ifm->ifa_scope = scope;
5091         ifm->ifa_index = ifindex;
5092 }
5093
5094 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5095                          unsigned long tstamp, u32 preferred, u32 valid)
5096 {
5097         struct ifa_cacheinfo ci;
5098
5099         ci.cstamp = cstamp_delta(cstamp);
5100         ci.tstamp = cstamp_delta(tstamp);
5101         ci.ifa_prefered = preferred;
5102         ci.ifa_valid = valid;
5103
5104         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5105 }
5106
5107 static inline int rt_scope(int ifa_scope)
5108 {
5109         if (ifa_scope & IFA_HOST)
5110                 return RT_SCOPE_HOST;
5111         else if (ifa_scope & IFA_LINK)
5112                 return RT_SCOPE_LINK;
5113         else if (ifa_scope & IFA_SITE)
5114                 return RT_SCOPE_SITE;
5115         else
5116                 return RT_SCOPE_UNIVERSE;
5117 }
5118
5119 static inline int inet6_ifaddr_msgsize(void)
5120 {
5121         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5122                + nla_total_size(16) /* IFA_LOCAL */
5123                + nla_total_size(16) /* IFA_ADDRESS */
5124                + nla_total_size(sizeof(struct ifa_cacheinfo))
5125                + nla_total_size(4)  /* IFA_FLAGS */
5126                + nla_total_size(1)  /* IFA_PROTO */
5127                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
5128 }
5129
5130 enum addr_type_t {
5131         UNICAST_ADDR,
5132         MULTICAST_ADDR,
5133         ANYCAST_ADDR,
5134 };
5135
5136 struct inet6_fill_args {
5137         u32 portid;
5138         u32 seq;
5139         int event;
5140         unsigned int flags;
5141         int netnsid;
5142         int ifindex;
5143         enum addr_type_t type;
5144 };
5145
5146 static int inet6_fill_ifaddr(struct sk_buff *skb,
5147                              const struct inet6_ifaddr *ifa,
5148                              struct inet6_fill_args *args)
5149 {
5150         struct nlmsghdr *nlh;
5151         u32 preferred, valid;
5152         u32 flags, priority;
5153         u8 proto;
5154
5155         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5156                         sizeof(struct ifaddrmsg), args->flags);
5157         if (!nlh)
5158                 return -EMSGSIZE;
5159
5160         flags = READ_ONCE(ifa->flags);
5161         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5162                       ifa->idev->dev->ifindex);
5163
5164         if (args->netnsid >= 0 &&
5165             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5166                 goto error;
5167
5168         preferred = READ_ONCE(ifa->prefered_lft);
5169         valid = READ_ONCE(ifa->valid_lft);
5170
5171         if (!((flags & IFA_F_PERMANENT) &&
5172               (preferred == INFINITY_LIFE_TIME))) {
5173                 if (preferred != INFINITY_LIFE_TIME) {
5174                         long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5175
5176                         if (preferred > tval)
5177                                 preferred -= tval;
5178                         else
5179                                 preferred = 0;
5180                         if (valid != INFINITY_LIFE_TIME) {
5181                                 if (valid > tval)
5182                                         valid -= tval;
5183                                 else
5184                                         valid = 0;
5185                         }
5186                 }
5187         } else {
5188                 preferred = INFINITY_LIFE_TIME;
5189                 valid = INFINITY_LIFE_TIME;
5190         }
5191
5192         if (!ipv6_addr_any(&ifa->peer_addr)) {
5193                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5194                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5195                         goto error;
5196         } else {
5197                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5198                         goto error;
5199         }
5200
5201         priority = READ_ONCE(ifa->rt_priority);
5202         if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5203                 goto error;
5204
5205         if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5206                           preferred, valid) < 0)
5207                 goto error;
5208
5209         if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5210                 goto error;
5211
5212         proto = READ_ONCE(ifa->ifa_proto);
5213         if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5214                 goto error;
5215
5216         nlmsg_end(skb, nlh);
5217         return 0;
5218
5219 error:
5220         nlmsg_cancel(skb, nlh);
5221         return -EMSGSIZE;
5222 }
5223
5224 static int inet6_fill_ifmcaddr(struct sk_buff *skb,
5225                                const struct ifmcaddr6 *ifmca,
5226                                struct inet6_fill_args *args)
5227 {
5228         int ifindex = ifmca->idev->dev->ifindex;
5229         u8 scope = RT_SCOPE_UNIVERSE;
5230         struct nlmsghdr *nlh;
5231
5232         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5233                 scope = RT_SCOPE_SITE;
5234
5235         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5236                         sizeof(struct ifaddrmsg), args->flags);
5237         if (!nlh)
5238                 return -EMSGSIZE;
5239
5240         if (args->netnsid >= 0 &&
5241             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5242                 nlmsg_cancel(skb, nlh);
5243                 return -EMSGSIZE;
5244         }
5245
5246         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5247         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5248             put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5249                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5250                 nlmsg_cancel(skb, nlh);
5251                 return -EMSGSIZE;
5252         }
5253
5254         nlmsg_end(skb, nlh);
5255         return 0;
5256 }
5257
5258 static int inet6_fill_ifacaddr(struct sk_buff *skb,
5259                                const struct ifacaddr6 *ifaca,
5260                                struct inet6_fill_args *args)
5261 {
5262         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5263         int ifindex = dev ? dev->ifindex : 1;
5264         u8 scope = RT_SCOPE_UNIVERSE;
5265         struct nlmsghdr *nlh;
5266
5267         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5268                 scope = RT_SCOPE_SITE;
5269
5270         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5271                         sizeof(struct ifaddrmsg), args->flags);
5272         if (!nlh)
5273                 return -EMSGSIZE;
5274
5275         if (args->netnsid >= 0 &&
5276             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5277                 nlmsg_cancel(skb, nlh);
5278                 return -EMSGSIZE;
5279         }
5280
5281         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5282         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5283             put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5284                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5285                 nlmsg_cancel(skb, nlh);
5286                 return -EMSGSIZE;
5287         }
5288
5289         nlmsg_end(skb, nlh);
5290         return 0;
5291 }
5292
5293 /* called with rcu_read_lock() */
5294 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5295                           struct netlink_callback *cb, int *s_ip_idx,
5296                           struct inet6_fill_args *fillargs)
5297 {
5298         const struct ifmcaddr6 *ifmca;
5299         const struct ifacaddr6 *ifaca;
5300         int ip_idx = 0;
5301         int err = 0;
5302
5303         switch (fillargs->type) {
5304         case UNICAST_ADDR: {
5305                 const struct inet6_ifaddr *ifa;
5306                 fillargs->event = RTM_NEWADDR;
5307
5308                 /* unicast address incl. temp addr */
5309                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5310                         if (ip_idx < *s_ip_idx)
5311                                 goto next;
5312                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5313                         if (err < 0)
5314                                 break;
5315                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5316 next:
5317                         ip_idx++;
5318                 }
5319                 break;
5320         }
5321         case MULTICAST_ADDR:
5322                 fillargs->event = RTM_GETMULTICAST;
5323
5324                 /* multicast address */
5325                 for (ifmca = rcu_dereference(idev->mc_list);
5326                      ifmca;
5327                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5328                         if (ip_idx < *s_ip_idx)
5329                                 continue;
5330                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5331                         if (err < 0)
5332                                 break;
5333                 }
5334                 break;
5335         case ANYCAST_ADDR:
5336                 fillargs->event = RTM_GETANYCAST;
5337                 /* anycast address */
5338                 for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5339                      ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5340                         if (ip_idx < *s_ip_idx)
5341                                 continue;
5342                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5343                         if (err < 0)
5344                                 break;
5345                 }
5346                 break;
5347         default:
5348                 break;
5349         }
5350         *s_ip_idx = err ? ip_idx : 0;
5351         return err;
5352 }
5353
5354 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5355                                        struct inet6_fill_args *fillargs,
5356                                        struct net **tgt_net, struct sock *sk,
5357                                        struct netlink_callback *cb)
5358 {
5359         struct netlink_ext_ack *extack = cb->extack;
5360         struct nlattr *tb[IFA_MAX+1];
5361         struct ifaddrmsg *ifm;
5362         int err, i;
5363
5364         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5365                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5366                 return -EINVAL;
5367         }
5368
5369         ifm = nlmsg_data(nlh);
5370         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5371                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5372                 return -EINVAL;
5373         }
5374
5375         fillargs->ifindex = ifm->ifa_index;
5376         if (fillargs->ifindex) {
5377                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5378                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5379         }
5380
5381         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5382                                             ifa_ipv6_policy, extack);
5383         if (err < 0)
5384                 return err;
5385
5386         for (i = 0; i <= IFA_MAX; ++i) {
5387                 if (!tb[i])
5388                         continue;
5389
5390                 if (i == IFA_TARGET_NETNSID) {
5391                         struct net *net;
5392
5393                         fillargs->netnsid = nla_get_s32(tb[i]);
5394                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5395                         if (IS_ERR(net)) {
5396                                 fillargs->netnsid = -1;
5397                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5398                                 return PTR_ERR(net);
5399                         }
5400                         *tgt_net = net;
5401                 } else {
5402                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5403                         return -EINVAL;
5404                 }
5405         }
5406
5407         return 0;
5408 }
5409
5410 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5411                            enum addr_type_t type)
5412 {
5413         struct net *tgt_net = sock_net(skb->sk);
5414         const struct nlmsghdr *nlh = cb->nlh;
5415         struct inet6_fill_args fillargs = {
5416                 .portid = NETLINK_CB(cb->skb).portid,
5417                 .seq = cb->nlh->nlmsg_seq,
5418                 .flags = NLM_F_MULTI,
5419                 .netnsid = -1,
5420                 .type = type,
5421         };
5422         struct {
5423                 unsigned long ifindex;
5424                 int ip_idx;
5425         } *ctx = (void *)cb->ctx;
5426         struct net_device *dev;
5427         struct inet6_dev *idev;
5428         int err = 0;
5429
5430         rcu_read_lock();
5431         if (cb->strict_check) {
5432                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5433                                                   skb->sk, cb);
5434                 if (err < 0)
5435                         goto done;
5436
5437                 err = 0;
5438                 if (fillargs.ifindex) {
5439                         dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5440                         if (!dev) {
5441                                 err = -ENODEV;
5442                                 goto done;
5443                         }
5444                         idev = __in6_dev_get(dev);
5445                         if (idev)
5446                                 err = in6_dump_addrs(idev, skb, cb,
5447                                                      &ctx->ip_idx,
5448                                                      &fillargs);
5449                         goto done;
5450                 }
5451         }
5452
5453         cb->seq = inet6_base_seq(tgt_net);
5454         for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5455                 idev = __in6_dev_get(dev);
5456                 if (!idev)
5457                         continue;
5458                 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5459                                      &fillargs);
5460                 if (err < 0)
5461                         goto done;
5462         }
5463 done:
5464         rcu_read_unlock();
5465         if (fillargs.netnsid >= 0)
5466                 put_net(tgt_net);
5467
5468         return err;
5469 }
5470
5471 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5472 {
5473         enum addr_type_t type = UNICAST_ADDR;
5474
5475         return inet6_dump_addr(skb, cb, type);
5476 }
5477
5478 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5479 {
5480         enum addr_type_t type = MULTICAST_ADDR;
5481
5482         return inet6_dump_addr(skb, cb, type);
5483 }
5484
5485
5486 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5487 {
5488         enum addr_type_t type = ANYCAST_ADDR;
5489
5490         return inet6_dump_addr(skb, cb, type);
5491 }
5492
5493 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5494                                        const struct nlmsghdr *nlh,
5495                                        struct nlattr **tb,
5496                                        struct netlink_ext_ack *extack)
5497 {
5498         struct ifaddrmsg *ifm;
5499         int i, err;
5500
5501         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5502                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5503                 return -EINVAL;
5504         }
5505
5506         if (!netlink_strict_get_check(skb))
5507                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5508                                               ifa_ipv6_policy, extack);
5509
5510         ifm = nlmsg_data(nlh);
5511         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5512                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5513                 return -EINVAL;
5514         }
5515
5516         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5517                                             ifa_ipv6_policy, extack);
5518         if (err)
5519                 return err;
5520
5521         for (i = 0; i <= IFA_MAX; i++) {
5522                 if (!tb[i])
5523                         continue;
5524
5525                 switch (i) {
5526                 case IFA_TARGET_NETNSID:
5527                 case IFA_ADDRESS:
5528                 case IFA_LOCAL:
5529                         break;
5530                 default:
5531                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5532                         return -EINVAL;
5533                 }
5534         }
5535
5536         return 0;
5537 }
5538
5539 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5540                              struct netlink_ext_ack *extack)
5541 {
5542         struct net *tgt_net = sock_net(in_skb->sk);
5543         struct inet6_fill_args fillargs = {
5544                 .portid = NETLINK_CB(in_skb).portid,
5545                 .seq = nlh->nlmsg_seq,
5546                 .event = RTM_NEWADDR,
5547                 .flags = 0,
5548                 .netnsid = -1,
5549         };
5550         struct ifaddrmsg *ifm;
5551         struct nlattr *tb[IFA_MAX+1];
5552         struct in6_addr *addr = NULL, *peer;
5553         struct net_device *dev = NULL;
5554         struct inet6_ifaddr *ifa;
5555         struct sk_buff *skb;
5556         int err;
5557
5558         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5559         if (err < 0)
5560                 return err;
5561
5562         if (tb[IFA_TARGET_NETNSID]) {
5563                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5564
5565                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5566                                                   fillargs.netnsid);
5567                 if (IS_ERR(tgt_net))
5568                         return PTR_ERR(tgt_net);
5569         }
5570
5571         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5572         if (!addr) {
5573                 err = -EINVAL;
5574                 goto errout;
5575         }
5576         ifm = nlmsg_data(nlh);
5577         if (ifm->ifa_index)
5578                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5579
5580         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5581         if (!ifa) {
5582                 err = -EADDRNOTAVAIL;
5583                 goto errout;
5584         }
5585
5586         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5587         if (!skb) {
5588                 err = -ENOBUFS;
5589                 goto errout_ifa;
5590         }
5591
5592         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5593         if (err < 0) {
5594                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5595                 WARN_ON(err == -EMSGSIZE);
5596                 kfree_skb(skb);
5597                 goto errout_ifa;
5598         }
5599         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5600 errout_ifa:
5601         in6_ifa_put(ifa);
5602 errout:
5603         dev_put(dev);
5604         if (fillargs.netnsid >= 0)
5605                 put_net(tgt_net);
5606
5607         return err;
5608 }
5609
5610 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5611 {
5612         struct sk_buff *skb;
5613         struct net *net = dev_net(ifa->idev->dev);
5614         struct inet6_fill_args fillargs = {
5615                 .portid = 0,
5616                 .seq = 0,
5617                 .event = event,
5618                 .flags = 0,
5619                 .netnsid = -1,
5620         };
5621         int err = -ENOBUFS;
5622
5623         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5624         if (!skb)
5625                 goto errout;
5626
5627         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5628         if (err < 0) {
5629                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5630                 WARN_ON(err == -EMSGSIZE);
5631                 kfree_skb(skb);
5632                 goto errout;
5633         }
5634         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5635         return;
5636 errout:
5637         rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5638 }
5639
5640 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5641                                __s32 *array, int bytes)
5642 {
5643         BUG_ON(bytes < (DEVCONF_MAX * 4));
5644
5645         memset(array, 0, bytes);
5646         array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5647         array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5648         array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5649         array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5650         array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5651         array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5652         array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5653         array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5654         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5655                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5656         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5657                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5658         array[DEVCONF_RTR_SOLICIT_DELAY] =
5659                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5660         array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5661         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5662                 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5663         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5664                 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5665         array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5666         array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5667         array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5668         array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5669         array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5670         array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5671         array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5672         array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5673         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5674                 READ_ONCE(cnf->accept_ra_min_hop_limit);
5675         array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5676 #ifdef CONFIG_IPV6_ROUTER_PREF
5677         array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5678         array[DEVCONF_RTR_PROBE_INTERVAL] =
5679                 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5680 #ifdef CONFIG_IPV6_ROUTE_INFO
5681         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5682                 READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5683         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5684                 READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5685 #endif
5686 #endif
5687         array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5688         array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5689                 READ_ONCE(cnf->accept_source_route);
5690 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5691         array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5692         array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5693 #endif
5694 #ifdef CONFIG_IPV6_MROUTE
5695         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5696 #endif
5697         array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5698         array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5699         array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5700         array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5701         array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5702                 READ_ONCE(cnf->suppress_frag_ndisc);
5703         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5704                 READ_ONCE(cnf->accept_ra_from_local);
5705         array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5706         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5707                 READ_ONCE(cnf->ignore_routes_with_linkdown);
5708         /* we omit DEVCONF_STABLE_SECRET for now */
5709         array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5710         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5711                 READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5712         array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5713         array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5714         array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5715 #ifdef CONFIG_IPV6_SEG6_HMAC
5716         array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5717 #endif
5718         array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5719         array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5720         array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5721         array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5722         array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5723         array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5724         array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5725         array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5726         array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5727                 READ_ONCE(cnf->ndisc_evict_nocarrier);
5728         array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5729                 READ_ONCE(cnf->accept_untracked_na);
5730         array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5731 }
5732
5733 static inline size_t inet6_ifla6_size(void)
5734 {
5735         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5736              + nla_total_size(sizeof(struct ifla_cacheinfo))
5737              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5738              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5739              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5740              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5741              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5742              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5743              + 0;
5744 }
5745
5746 static inline size_t inet6_if_nlmsg_size(void)
5747 {
5748         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5749                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5750                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5751                + nla_total_size(4) /* IFLA_MTU */
5752                + nla_total_size(4) /* IFLA_LINK */
5753                + nla_total_size(1) /* IFLA_OPERSTATE */
5754                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5755 }
5756
5757 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5758                                         int bytes)
5759 {
5760         int i;
5761         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5762         BUG_ON(pad < 0);
5763
5764         /* Use put_unaligned() because stats may not be aligned for u64. */
5765         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5766         for (i = 1; i < ICMP6_MIB_MAX; i++)
5767                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5768
5769         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5770 }
5771
5772 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5773                                         int bytes, size_t syncpoff)
5774 {
5775         int i, c;
5776         u64 buff[IPSTATS_MIB_MAX];
5777         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5778
5779         BUG_ON(pad < 0);
5780
5781         memset(buff, 0, sizeof(buff));
5782         buff[0] = IPSTATS_MIB_MAX;
5783
5784         for_each_possible_cpu(c) {
5785                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5786                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5787         }
5788
5789         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5790         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5791 }
5792
5793 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5794                              int bytes)
5795 {
5796         switch (attrtype) {
5797         case IFLA_INET6_STATS:
5798                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5799                                      offsetof(struct ipstats_mib, syncp));
5800                 break;
5801         case IFLA_INET6_ICMP6STATS:
5802                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5803                 break;
5804         }
5805 }
5806
5807 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5808                                   u32 ext_filter_mask)
5809 {
5810         struct ifla_cacheinfo ci;
5811         struct nlattr *nla;
5812         u32 ra_mtu;
5813
5814         if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5815                 goto nla_put_failure;
5816         ci.max_reasm_len = IPV6_MAXPLEN;
5817         ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5818         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5819         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5820         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5821                 goto nla_put_failure;
5822         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5823         if (!nla)
5824                 goto nla_put_failure;
5825         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5826
5827         /* XXX - MC not implemented */
5828
5829         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5830                 return 0;
5831
5832         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5833         if (!nla)
5834                 goto nla_put_failure;
5835         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5836
5837         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5838         if (!nla)
5839                 goto nla_put_failure;
5840         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5841
5842         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5843         if (!nla)
5844                 goto nla_put_failure;
5845         read_lock_bh(&idev->lock);
5846         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5847         read_unlock_bh(&idev->lock);
5848
5849         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5850                        READ_ONCE(idev->cnf.addr_gen_mode)))
5851                 goto nla_put_failure;
5852
5853         ra_mtu = READ_ONCE(idev->ra_mtu);
5854         if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5855                 goto nla_put_failure;
5856
5857         return 0;
5858
5859 nla_put_failure:
5860         return -EMSGSIZE;
5861 }
5862
5863 static size_t inet6_get_link_af_size(const struct net_device *dev,
5864                                      u32 ext_filter_mask)
5865 {
5866         if (!__in6_dev_get(dev))
5867                 return 0;
5868
5869         return inet6_ifla6_size();
5870 }
5871
5872 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5873                               u32 ext_filter_mask)
5874 {
5875         struct inet6_dev *idev = __in6_dev_get(dev);
5876
5877         if (!idev)
5878                 return -ENODATA;
5879
5880         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5881                 return -EMSGSIZE;
5882
5883         return 0;
5884 }
5885
5886 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5887                              struct netlink_ext_ack *extack)
5888 {
5889         struct inet6_ifaddr *ifp;
5890         struct net_device *dev = idev->dev;
5891         bool clear_token, update_rs = false;
5892         struct in6_addr ll_addr;
5893
5894         ASSERT_RTNL();
5895
5896         if (!token)
5897                 return -EINVAL;
5898
5899         if (dev->flags & IFF_LOOPBACK) {
5900                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5901                 return -EINVAL;
5902         }
5903
5904         if (dev->flags & IFF_NOARP) {
5905                 NL_SET_ERR_MSG_MOD(extack,
5906                                    "Device does not do neighbour discovery");
5907                 return -EINVAL;
5908         }
5909
5910         if (!ipv6_accept_ra(idev)) {
5911                 NL_SET_ERR_MSG_MOD(extack,
5912                                    "Router advertisement is disabled on device");
5913                 return -EINVAL;
5914         }
5915
5916         if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5917                 NL_SET_ERR_MSG(extack,
5918                                "Router solicitation is disabled on device");
5919                 return -EINVAL;
5920         }
5921
5922         write_lock_bh(&idev->lock);
5923
5924         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5925         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5926
5927         write_unlock_bh(&idev->lock);
5928
5929         clear_token = ipv6_addr_any(token);
5930         if (clear_token)
5931                 goto update_lft;
5932
5933         if (!idev->dead && (idev->if_flags & IF_READY) &&
5934             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5935                              IFA_F_OPTIMISTIC)) {
5936                 /* If we're not ready, then normal ifup will take care
5937                  * of this. Otherwise, we need to request our rs here.
5938                  */
5939                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5940                 update_rs = true;
5941         }
5942
5943 update_lft:
5944         write_lock_bh(&idev->lock);
5945
5946         if (update_rs) {
5947                 idev->if_flags |= IF_RS_SENT;
5948                 idev->rs_interval = rfc3315_s14_backoff_init(
5949                         READ_ONCE(idev->cnf.rtr_solicit_interval));
5950                 idev->rs_probes = 1;
5951                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5952         }
5953
5954         /* Well, that's kinda nasty ... */
5955         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5956                 spin_lock(&ifp->lock);
5957                 if (ifp->tokenized) {
5958                         ifp->valid_lft = 0;
5959                         ifp->prefered_lft = 0;
5960                 }
5961                 spin_unlock(&ifp->lock);
5962         }
5963
5964         write_unlock_bh(&idev->lock);
5965         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5966         addrconf_verify_rtnl(dev_net(dev));
5967         return 0;
5968 }
5969
5970 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5971         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5972         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5973         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5974                                             .reject_message =
5975                                                 "IFLA_INET6_RA_MTU can not be set" },
5976 };
5977
5978 static int check_addr_gen_mode(int mode)
5979 {
5980         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5981             mode != IN6_ADDR_GEN_MODE_NONE &&
5982             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5983             mode != IN6_ADDR_GEN_MODE_RANDOM)
5984                 return -EINVAL;
5985         return 1;
5986 }
5987
5988 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5989                                 int mode)
5990 {
5991         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5992             !idev->cnf.stable_secret.initialized &&
5993             !net->ipv6.devconf_dflt->stable_secret.initialized)
5994                 return -EINVAL;
5995         return 1;
5996 }
5997
5998 static int inet6_validate_link_af(const struct net_device *dev,
5999                                   const struct nlattr *nla,
6000                                   struct netlink_ext_ack *extack)
6001 {
6002         struct nlattr *tb[IFLA_INET6_MAX + 1];
6003         struct inet6_dev *idev = NULL;
6004         int err;
6005
6006         if (dev) {
6007                 idev = __in6_dev_get(dev);
6008                 if (!idev)
6009                         return -EAFNOSUPPORT;
6010         }
6011
6012         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
6013                                           inet6_af_policy, extack);
6014         if (err)
6015                 return err;
6016
6017         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6018                 return -EINVAL;
6019
6020         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6021                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6022
6023                 if (check_addr_gen_mode(mode) < 0)
6024                         return -EINVAL;
6025                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6026                         return -EINVAL;
6027         }
6028
6029         return 0;
6030 }
6031
6032 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6033                              struct netlink_ext_ack *extack)
6034 {
6035         struct inet6_dev *idev = __in6_dev_get(dev);
6036         struct nlattr *tb[IFLA_INET6_MAX + 1];
6037         int err;
6038
6039         if (!idev)
6040                 return -EAFNOSUPPORT;
6041
6042         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6043                 return -EINVAL;
6044
6045         if (tb[IFLA_INET6_TOKEN]) {
6046                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6047                                         extack);
6048                 if (err)
6049                         return err;
6050         }
6051
6052         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6053                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6054
6055                 WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6056         }
6057
6058         return 0;
6059 }
6060
6061 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6062                              u32 portid, u32 seq, int event, unsigned int flags)
6063 {
6064         struct net_device *dev = idev->dev;
6065         struct ifinfomsg *hdr;
6066         struct nlmsghdr *nlh;
6067         int ifindex, iflink;
6068         void *protoinfo;
6069
6070         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6071         if (!nlh)
6072                 return -EMSGSIZE;
6073
6074         hdr = nlmsg_data(nlh);
6075         hdr->ifi_family = AF_INET6;
6076         hdr->__ifi_pad = 0;
6077         hdr->ifi_type = dev->type;
6078         ifindex = READ_ONCE(dev->ifindex);
6079         hdr->ifi_index = ifindex;
6080         hdr->ifi_flags = dev_get_flags(dev);
6081         hdr->ifi_change = 0;
6082
6083         iflink = dev_get_iflink(dev);
6084         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6085             (dev->addr_len &&
6086              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6087             nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6088             (ifindex != iflink &&
6089              nla_put_u32(skb, IFLA_LINK, iflink)) ||
6090             nla_put_u8(skb, IFLA_OPERSTATE,
6091                        netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6092                 goto nla_put_failure;
6093         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6094         if (!protoinfo)
6095                 goto nla_put_failure;
6096
6097         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6098                 goto nla_put_failure;
6099
6100         nla_nest_end(skb, protoinfo);
6101         nlmsg_end(skb, nlh);
6102         return 0;
6103
6104 nla_put_failure:
6105         nlmsg_cancel(skb, nlh);
6106         return -EMSGSIZE;
6107 }
6108
6109 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6110                                    struct netlink_ext_ack *extack)
6111 {
6112         struct ifinfomsg *ifm;
6113
6114         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6115                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6116                 return -EINVAL;
6117         }
6118
6119         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6120                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6121                 return -EINVAL;
6122         }
6123
6124         ifm = nlmsg_data(nlh);
6125         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6126             ifm->ifi_change || ifm->ifi_index) {
6127                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6128                 return -EINVAL;
6129         }
6130
6131         return 0;
6132 }
6133
6134 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6135 {
6136         struct net *net = sock_net(skb->sk);
6137         struct {
6138                 unsigned long ifindex;
6139         } *ctx = (void *)cb->ctx;
6140         struct net_device *dev;
6141         struct inet6_dev *idev;
6142         int err;
6143
6144         /* only requests using strict checking can pass data to
6145          * influence the dump
6146          */
6147         if (cb->strict_check) {
6148                 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6149
6150                 if (err < 0)
6151                         return err;
6152         }
6153
6154         err = 0;
6155         rcu_read_lock();
6156         for_each_netdev_dump(net, dev, ctx->ifindex) {
6157                 idev = __in6_dev_get(dev);
6158                 if (!idev)
6159                         continue;
6160                 err = inet6_fill_ifinfo(skb, idev,
6161                                         NETLINK_CB(cb->skb).portid,
6162                                         cb->nlh->nlmsg_seq,
6163                                         RTM_NEWLINK, NLM_F_MULTI);
6164                 if (err < 0)
6165                         break;
6166         }
6167         rcu_read_unlock();
6168
6169         return err;
6170 }
6171
6172 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6173 {
6174         struct sk_buff *skb;
6175         struct net *net = dev_net(idev->dev);
6176         int err = -ENOBUFS;
6177
6178         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6179         if (!skb)
6180                 goto errout;
6181
6182         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6183         if (err < 0) {
6184                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6185                 WARN_ON(err == -EMSGSIZE);
6186                 kfree_skb(skb);
6187                 goto errout;
6188         }
6189         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6190         return;
6191 errout:
6192         rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6193 }
6194
6195 static inline size_t inet6_prefix_nlmsg_size(void)
6196 {
6197         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6198                + nla_total_size(sizeof(struct in6_addr))
6199                + nla_total_size(sizeof(struct prefix_cacheinfo));
6200 }
6201
6202 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6203                              struct prefix_info *pinfo, u32 portid, u32 seq,
6204                              int event, unsigned int flags)
6205 {
6206         struct prefixmsg *pmsg;
6207         struct nlmsghdr *nlh;
6208         struct prefix_cacheinfo ci;
6209
6210         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6211         if (!nlh)
6212                 return -EMSGSIZE;
6213
6214         pmsg = nlmsg_data(nlh);
6215         pmsg->prefix_family = AF_INET6;
6216         pmsg->prefix_pad1 = 0;
6217         pmsg->prefix_pad2 = 0;
6218         pmsg->prefix_ifindex = idev->dev->ifindex;
6219         pmsg->prefix_len = pinfo->prefix_len;
6220         pmsg->prefix_type = pinfo->type;
6221         pmsg->prefix_pad3 = 0;
6222         pmsg->prefix_flags = pinfo->flags;
6223
6224         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6225                 goto nla_put_failure;
6226         ci.preferred_time = ntohl(pinfo->prefered);
6227         ci.valid_time = ntohl(pinfo->valid);
6228         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6229                 goto nla_put_failure;
6230         nlmsg_end(skb, nlh);
6231         return 0;
6232
6233 nla_put_failure:
6234         nlmsg_cancel(skb, nlh);
6235         return -EMSGSIZE;
6236 }
6237
6238 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6239                          struct prefix_info *pinfo)
6240 {
6241         struct sk_buff *skb;
6242         struct net *net = dev_net(idev->dev);
6243         int err = -ENOBUFS;
6244
6245         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6246         if (!skb)
6247                 goto errout;
6248
6249         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6250         if (err < 0) {
6251                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6252                 WARN_ON(err == -EMSGSIZE);
6253                 kfree_skb(skb);
6254                 goto errout;
6255         }
6256         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6257         return;
6258 errout:
6259         rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6260 }
6261
6262 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6263 {
6264         struct net *net = dev_net(ifp->idev->dev);
6265
6266         if (event)
6267                 ASSERT_RTNL();
6268
6269         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6270
6271         switch (event) {
6272         case RTM_NEWADDR:
6273                 /*
6274                  * If the address was optimistic we inserted the route at the
6275                  * start of our DAD process, so we don't need to do it again.
6276                  * If the device was taken down in the middle of the DAD
6277                  * cycle there is a race where we could get here without a
6278                  * host route, so nothing to insert. That will be fixed when
6279                  * the device is brought up.
6280                  */
6281                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6282                         ip6_ins_rt(net, ifp->rt);
6283                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6284                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6285                                 &ifp->addr, ifp->idev->dev->name);
6286                 }
6287
6288                 if (ifp->idev->cnf.forwarding)
6289                         addrconf_join_anycast(ifp);
6290                 if (!ipv6_addr_any(&ifp->peer_addr))
6291                         addrconf_prefix_route(&ifp->peer_addr, 128,
6292                                               ifp->rt_priority, ifp->idev->dev,
6293                                               0, 0, GFP_ATOMIC);
6294                 break;
6295         case RTM_DELADDR:
6296                 if (ifp->idev->cnf.forwarding)
6297                         addrconf_leave_anycast(ifp);
6298                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6299                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6300                         struct fib6_info *rt;
6301
6302                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6303                                                        ifp->idev->dev, 0, 0,
6304                                                        false);
6305                         if (rt)
6306                                 ip6_del_rt(net, rt, false);
6307                 }
6308                 if (ifp->rt) {
6309                         ip6_del_rt(net, ifp->rt, false);
6310                         ifp->rt = NULL;
6311                 }
6312                 rt_genid_bump_ipv6(net);
6313                 break;
6314         }
6315         atomic_inc(&net->ipv6.dev_addr_genid);
6316 }
6317
6318 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6319 {
6320         if (likely(ifp->idev->dead == 0))
6321                 __ipv6_ifa_notify(event, ifp);
6322 }
6323
6324 #ifdef CONFIG_SYSCTL
6325
6326 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6327                 void *buffer, size_t *lenp, loff_t *ppos)
6328 {
6329         int *valp = ctl->data;
6330         int val = *valp;
6331         loff_t pos = *ppos;
6332         struct ctl_table lctl;
6333         int ret;
6334
6335         /*
6336          * ctl->data points to idev->cnf.forwarding, we should
6337          * not modify it until we get the rtnl lock.
6338          */
6339         lctl = *ctl;
6340         lctl.data = &val;
6341
6342         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6343
6344         if (write)
6345                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6346         if (ret)
6347                 *ppos = pos;
6348         return ret;
6349 }
6350
6351 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6352                 void *buffer, size_t *lenp, loff_t *ppos)
6353 {
6354         struct inet6_dev *idev = ctl->extra1;
6355         int min_mtu = IPV6_MIN_MTU;
6356         struct ctl_table lctl;
6357
6358         lctl = *ctl;
6359         lctl.extra1 = &min_mtu;
6360         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6361
6362         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6363 }
6364
6365 static void dev_disable_change(struct inet6_dev *idev)
6366 {
6367         struct netdev_notifier_info info;
6368
6369         if (!idev || !idev->dev)
6370                 return;
6371
6372         netdev_notifier_info_init(&info, idev->dev);
6373         if (idev->cnf.disable_ipv6)
6374                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6375         else
6376                 addrconf_notify(NULL, NETDEV_UP, &info);
6377 }
6378
6379 static void addrconf_disable_change(struct net *net, __s32 newf)
6380 {
6381         struct net_device *dev;
6382         struct inet6_dev *idev;
6383
6384         for_each_netdev(net, dev) {
6385                 idev = __in6_dev_get(dev);
6386                 if (idev) {
6387                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6388
6389                         WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6390                         if (changed)
6391                                 dev_disable_change(idev);
6392                 }
6393         }
6394 }
6395
6396 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6397 {
6398         struct net *net = (struct net *)table->extra2;
6399         int old;
6400
6401         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6402                 WRITE_ONCE(*p, newf);
6403                 return 0;
6404         }
6405
6406         if (!rtnl_trylock())
6407                 return restart_syscall();
6408
6409         old = *p;
6410         WRITE_ONCE(*p, newf);
6411
6412         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6413                 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6414                 addrconf_disable_change(net, newf);
6415         } else if ((!newf) ^ (!old))
6416                 dev_disable_change((struct inet6_dev *)table->extra1);
6417
6418         rtnl_unlock();
6419         return 0;
6420 }
6421
6422 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6423                 void *buffer, size_t *lenp, loff_t *ppos)
6424 {
6425         int *valp = ctl->data;
6426         int val = *valp;
6427         loff_t pos = *ppos;
6428         struct ctl_table lctl;
6429         int ret;
6430
6431         /*
6432          * ctl->data points to idev->cnf.disable_ipv6, we should
6433          * not modify it until we get the rtnl lock.
6434          */
6435         lctl = *ctl;
6436         lctl.data = &val;
6437
6438         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6439
6440         if (write)
6441                 ret = addrconf_disable_ipv6(ctl, valp, val);
6442         if (ret)
6443                 *ppos = pos;
6444         return ret;
6445 }
6446
6447 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6448                 void *buffer, size_t *lenp, loff_t *ppos)
6449 {
6450         int *valp = ctl->data;
6451         int ret;
6452         int old, new;
6453
6454         old = *valp;
6455         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6456         new = *valp;
6457
6458         if (write && old != new) {
6459                 struct net *net = ctl->extra2;
6460
6461                 if (!rtnl_trylock())
6462                         return restart_syscall();
6463
6464                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6465                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6466                                                      NETCONFA_PROXY_NEIGH,
6467                                                      NETCONFA_IFINDEX_DEFAULT,
6468                                                      net->ipv6.devconf_dflt);
6469                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6470                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6471                                                      NETCONFA_PROXY_NEIGH,
6472                                                      NETCONFA_IFINDEX_ALL,
6473                                                      net->ipv6.devconf_all);
6474                 else {
6475                         struct inet6_dev *idev = ctl->extra1;
6476
6477                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6478                                                      NETCONFA_PROXY_NEIGH,
6479                                                      idev->dev->ifindex,
6480                                                      &idev->cnf);
6481                 }
6482                 rtnl_unlock();
6483         }
6484
6485         return ret;
6486 }
6487
6488 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6489                                          void *buffer, size_t *lenp,
6490                                          loff_t *ppos)
6491 {
6492         int ret = 0;
6493         u32 new_val;
6494         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6495         struct net *net = (struct net *)ctl->extra2;
6496         struct ctl_table tmp = {
6497                 .data = &new_val,
6498                 .maxlen = sizeof(new_val),
6499                 .mode = ctl->mode,
6500         };
6501
6502         if (!rtnl_trylock())
6503                 return restart_syscall();
6504
6505         new_val = *((u32 *)ctl->data);
6506
6507         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6508         if (ret != 0)
6509                 goto out;
6510
6511         if (write) {
6512                 if (check_addr_gen_mode(new_val) < 0) {
6513                         ret = -EINVAL;
6514                         goto out;
6515                 }
6516
6517                 if (idev) {
6518                         if (check_stable_privacy(idev, net, new_val) < 0) {
6519                                 ret = -EINVAL;
6520                                 goto out;
6521                         }
6522
6523                         if (idev->cnf.addr_gen_mode != new_val) {
6524                                 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6525                                 addrconf_init_auto_addrs(idev->dev);
6526                         }
6527                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6528                         struct net_device *dev;
6529
6530                         WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6531                         for_each_netdev(net, dev) {
6532                                 idev = __in6_dev_get(dev);
6533                                 if (idev &&
6534                                     idev->cnf.addr_gen_mode != new_val) {
6535                                         WRITE_ONCE(idev->cnf.addr_gen_mode,
6536                                                   new_val);
6537                                         addrconf_init_auto_addrs(idev->dev);
6538                                 }
6539                         }
6540                 }
6541
6542                 WRITE_ONCE(*((u32 *)ctl->data), new_val);
6543         }
6544
6545 out:
6546         rtnl_unlock();
6547
6548         return ret;
6549 }
6550
6551 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6552                                          void *buffer, size_t *lenp,
6553                                          loff_t *ppos)
6554 {
6555         int err;
6556         struct in6_addr addr;
6557         char str[IPV6_MAX_STRLEN];
6558         struct ctl_table lctl = *ctl;
6559         struct net *net = ctl->extra2;
6560         struct ipv6_stable_secret *secret = ctl->data;
6561
6562         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6563                 return -EIO;
6564
6565         lctl.maxlen = IPV6_MAX_STRLEN;
6566         lctl.data = str;
6567
6568         if (!rtnl_trylock())
6569                 return restart_syscall();
6570
6571         if (!write && !secret->initialized) {
6572                 err = -EIO;
6573                 goto out;
6574         }
6575
6576         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6577         if (err >= sizeof(str)) {
6578                 err = -EIO;
6579                 goto out;
6580         }
6581
6582         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6583         if (err || !write)
6584                 goto out;
6585
6586         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6587                 err = -EIO;
6588                 goto out;
6589         }
6590
6591         secret->initialized = true;
6592         secret->secret = addr;
6593
6594         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6595                 struct net_device *dev;
6596
6597                 for_each_netdev(net, dev) {
6598                         struct inet6_dev *idev = __in6_dev_get(dev);
6599
6600                         if (idev) {
6601                                 WRITE_ONCE(idev->cnf.addr_gen_mode,
6602                                            IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6603                         }
6604                 }
6605         } else {
6606                 struct inet6_dev *idev = ctl->extra1;
6607
6608                 WRITE_ONCE(idev->cnf.addr_gen_mode,
6609                            IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6610         }
6611
6612 out:
6613         rtnl_unlock();
6614
6615         return err;
6616 }
6617
6618 static
6619 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6620                                                 int write, void *buffer,
6621                                                 size_t *lenp,
6622                                                 loff_t *ppos)
6623 {
6624         int *valp = ctl->data;
6625         int val = *valp;
6626         loff_t pos = *ppos;
6627         struct ctl_table lctl;
6628         int ret;
6629
6630         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6631          * we should not modify it until we get the rtnl lock.
6632          */
6633         lctl = *ctl;
6634         lctl.data = &val;
6635
6636         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6637
6638         if (write)
6639                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6640         if (ret)
6641                 *ppos = pos;
6642         return ret;
6643 }
6644
6645 static
6646 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6647 {
6648         if (rt) {
6649                 if (action)
6650                         rt->dst.flags |= DST_NOPOLICY;
6651                 else
6652                         rt->dst.flags &= ~DST_NOPOLICY;
6653         }
6654 }
6655
6656 static
6657 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6658 {
6659         struct inet6_ifaddr *ifa;
6660
6661         read_lock_bh(&idev->lock);
6662         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6663                 spin_lock(&ifa->lock);
6664                 if (ifa->rt) {
6665                         /* host routes only use builtin fib6_nh */
6666                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6667                         int cpu;
6668
6669                         rcu_read_lock();
6670                         ifa->rt->dst_nopolicy = val ? true : false;
6671                         if (nh->rt6i_pcpu) {
6672                                 for_each_possible_cpu(cpu) {
6673                                         struct rt6_info **rtp;
6674
6675                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6676                                         addrconf_set_nopolicy(*rtp, val);
6677                                 }
6678                         }
6679                         rcu_read_unlock();
6680                 }
6681                 spin_unlock(&ifa->lock);
6682         }
6683         read_unlock_bh(&idev->lock);
6684 }
6685
6686 static
6687 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6688 {
6689         struct net *net = (struct net *)ctl->extra2;
6690         struct inet6_dev *idev;
6691
6692         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6693                 WRITE_ONCE(*valp, val);
6694                 return 0;
6695         }
6696
6697         if (!rtnl_trylock())
6698                 return restart_syscall();
6699
6700         WRITE_ONCE(*valp, val);
6701
6702         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6703                 struct net_device *dev;
6704
6705                 for_each_netdev(net, dev) {
6706                         idev = __in6_dev_get(dev);
6707                         if (idev)
6708                                 addrconf_disable_policy_idev(idev, val);
6709                 }
6710         } else {
6711                 idev = (struct inet6_dev *)ctl->extra1;
6712                 addrconf_disable_policy_idev(idev, val);
6713         }
6714
6715         rtnl_unlock();
6716         return 0;
6717 }
6718
6719 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6720                                    void *buffer, size_t *lenp, loff_t *ppos)
6721 {
6722         int *valp = ctl->data;
6723         int val = *valp;
6724         loff_t pos = *ppos;
6725         struct ctl_table lctl;
6726         int ret;
6727
6728         lctl = *ctl;
6729         lctl.data = &val;
6730         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6731
6732         if (write && (*valp != val))
6733                 ret = addrconf_disable_policy(ctl, valp, val);
6734
6735         if (ret)
6736                 *ppos = pos;
6737
6738         return ret;
6739 }
6740
6741 static int minus_one = -1;
6742 static const int two_five_five = 255;
6743 static u32 ioam6_if_id_max = U16_MAX;
6744
6745 static const struct ctl_table addrconf_sysctl[] = {
6746         {
6747                 .procname       = "forwarding",
6748                 .data           = &ipv6_devconf.forwarding,
6749                 .maxlen         = sizeof(int),
6750                 .mode           = 0644,
6751                 .proc_handler   = addrconf_sysctl_forward,
6752         },
6753         {
6754                 .procname       = "hop_limit",
6755                 .data           = &ipv6_devconf.hop_limit,
6756                 .maxlen         = sizeof(int),
6757                 .mode           = 0644,
6758                 .proc_handler   = proc_dointvec_minmax,
6759                 .extra1         = (void *)SYSCTL_ONE,
6760                 .extra2         = (void *)&two_five_five,
6761         },
6762         {
6763                 .procname       = "mtu",
6764                 .data           = &ipv6_devconf.mtu6,
6765                 .maxlen         = sizeof(int),
6766                 .mode           = 0644,
6767                 .proc_handler   = addrconf_sysctl_mtu,
6768         },
6769         {
6770                 .procname       = "accept_ra",
6771                 .data           = &ipv6_devconf.accept_ra,
6772                 .maxlen         = sizeof(int),
6773                 .mode           = 0644,
6774                 .proc_handler   = proc_dointvec,
6775         },
6776         {
6777                 .procname       = "accept_redirects",
6778                 .data           = &ipv6_devconf.accept_redirects,
6779                 .maxlen         = sizeof(int),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_dointvec,
6782         },
6783         {
6784                 .procname       = "autoconf",
6785                 .data           = &ipv6_devconf.autoconf,
6786                 .maxlen         = sizeof(int),
6787                 .mode           = 0644,
6788                 .proc_handler   = proc_dointvec,
6789         },
6790         {
6791                 .procname       = "dad_transmits",
6792                 .data           = &ipv6_devconf.dad_transmits,
6793                 .maxlen         = sizeof(int),
6794                 .mode           = 0644,
6795                 .proc_handler   = proc_dointvec,
6796         },
6797         {
6798                 .procname       = "router_solicitations",
6799                 .data           = &ipv6_devconf.rtr_solicits,
6800                 .maxlen         = sizeof(int),
6801                 .mode           = 0644,
6802                 .proc_handler   = proc_dointvec_minmax,
6803                 .extra1         = &minus_one,
6804         },
6805         {
6806                 .procname       = "router_solicitation_interval",
6807                 .data           = &ipv6_devconf.rtr_solicit_interval,
6808                 .maxlen         = sizeof(int),
6809                 .mode           = 0644,
6810                 .proc_handler   = proc_dointvec_jiffies,
6811         },
6812         {
6813                 .procname       = "router_solicitation_max_interval",
6814                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6815                 .maxlen         = sizeof(int),
6816                 .mode           = 0644,
6817                 .proc_handler   = proc_dointvec_jiffies,
6818         },
6819         {
6820                 .procname       = "router_solicitation_delay",
6821                 .data           = &ipv6_devconf.rtr_solicit_delay,
6822                 .maxlen         = sizeof(int),
6823                 .mode           = 0644,
6824                 .proc_handler   = proc_dointvec_jiffies,
6825         },
6826         {
6827                 .procname       = "force_mld_version",
6828                 .data           = &ipv6_devconf.force_mld_version,
6829                 .maxlen         = sizeof(int),
6830                 .mode           = 0644,
6831                 .proc_handler   = proc_dointvec,
6832         },
6833         {
6834                 .procname       = "mldv1_unsolicited_report_interval",
6835                 .data           =
6836                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6837                 .maxlen         = sizeof(int),
6838                 .mode           = 0644,
6839                 .proc_handler   = proc_dointvec_ms_jiffies,
6840         },
6841         {
6842                 .procname       = "mldv2_unsolicited_report_interval",
6843                 .data           =
6844                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6845                 .maxlen         = sizeof(int),
6846                 .mode           = 0644,
6847                 .proc_handler   = proc_dointvec_ms_jiffies,
6848         },
6849         {
6850                 .procname       = "use_tempaddr",
6851                 .data           = &ipv6_devconf.use_tempaddr,
6852                 .maxlen         = sizeof(int),
6853                 .mode           = 0644,
6854                 .proc_handler   = proc_dointvec,
6855         },
6856         {
6857                 .procname       = "temp_valid_lft",
6858                 .data           = &ipv6_devconf.temp_valid_lft,
6859                 .maxlen         = sizeof(int),
6860                 .mode           = 0644,
6861                 .proc_handler   = proc_dointvec,
6862         },
6863         {
6864                 .procname       = "temp_prefered_lft",
6865                 .data           = &ipv6_devconf.temp_prefered_lft,
6866                 .maxlen         = sizeof(int),
6867                 .mode           = 0644,
6868                 .proc_handler   = proc_dointvec,
6869         },
6870         {
6871                 .procname       = "regen_min_advance",
6872                 .data           = &ipv6_devconf.regen_min_advance,
6873                 .maxlen         = sizeof(int),
6874                 .mode           = 0644,
6875                 .proc_handler   = proc_dointvec,
6876         },
6877         {
6878                 .procname       = "regen_max_retry",
6879                 .data           = &ipv6_devconf.regen_max_retry,
6880                 .maxlen         = sizeof(int),
6881                 .mode           = 0644,
6882                 .proc_handler   = proc_dointvec,
6883         },
6884         {
6885                 .procname       = "max_desync_factor",
6886                 .data           = &ipv6_devconf.max_desync_factor,
6887                 .maxlen         = sizeof(int),
6888                 .mode           = 0644,
6889                 .proc_handler   = proc_dointvec,
6890         },
6891         {
6892                 .procname       = "max_addresses",
6893                 .data           = &ipv6_devconf.max_addresses,
6894                 .maxlen         = sizeof(int),
6895                 .mode           = 0644,
6896                 .proc_handler   = proc_dointvec,
6897         },
6898         {
6899                 .procname       = "accept_ra_defrtr",
6900                 .data           = &ipv6_devconf.accept_ra_defrtr,
6901                 .maxlen         = sizeof(int),
6902                 .mode           = 0644,
6903                 .proc_handler   = proc_dointvec,
6904         },
6905         {
6906                 .procname       = "ra_defrtr_metric",
6907                 .data           = &ipv6_devconf.ra_defrtr_metric,
6908                 .maxlen         = sizeof(u32),
6909                 .mode           = 0644,
6910                 .proc_handler   = proc_douintvec_minmax,
6911                 .extra1         = (void *)SYSCTL_ONE,
6912         },
6913         {
6914                 .procname       = "accept_ra_min_hop_limit",
6915                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6916                 .maxlen         = sizeof(int),
6917                 .mode           = 0644,
6918                 .proc_handler   = proc_dointvec,
6919         },
6920         {
6921                 .procname       = "accept_ra_min_lft",
6922                 .data           = &ipv6_devconf.accept_ra_min_lft,
6923                 .maxlen         = sizeof(int),
6924                 .mode           = 0644,
6925                 .proc_handler   = proc_dointvec,
6926         },
6927         {
6928                 .procname       = "accept_ra_pinfo",
6929                 .data           = &ipv6_devconf.accept_ra_pinfo,
6930                 .maxlen         = sizeof(int),
6931                 .mode           = 0644,
6932                 .proc_handler   = proc_dointvec,
6933         },
6934         {
6935                 .procname       = "ra_honor_pio_life",
6936                 .data           = &ipv6_devconf.ra_honor_pio_life,
6937                 .maxlen         = sizeof(u8),
6938                 .mode           = 0644,
6939                 .proc_handler   = proc_dou8vec_minmax,
6940                 .extra1         = SYSCTL_ZERO,
6941                 .extra2         = SYSCTL_ONE,
6942         },
6943         {
6944                 .procname       = "ra_honor_pio_pflag",
6945                 .data           = &ipv6_devconf.ra_honor_pio_pflag,
6946                 .maxlen         = sizeof(u8),
6947                 .mode           = 0644,
6948                 .proc_handler   = proc_dou8vec_minmax,
6949                 .extra1         = SYSCTL_ZERO,
6950                 .extra2         = SYSCTL_ONE,
6951         },
6952 #ifdef CONFIG_IPV6_ROUTER_PREF
6953         {
6954                 .procname       = "accept_ra_rtr_pref",
6955                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6956                 .maxlen         = sizeof(int),
6957                 .mode           = 0644,
6958                 .proc_handler   = proc_dointvec,
6959         },
6960         {
6961                 .procname       = "router_probe_interval",
6962                 .data           = &ipv6_devconf.rtr_probe_interval,
6963                 .maxlen         = sizeof(int),
6964                 .mode           = 0644,
6965                 .proc_handler   = proc_dointvec_jiffies,
6966         },
6967 #ifdef CONFIG_IPV6_ROUTE_INFO
6968         {
6969                 .procname       = "accept_ra_rt_info_min_plen",
6970                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6971                 .maxlen         = sizeof(int),
6972                 .mode           = 0644,
6973                 .proc_handler   = proc_dointvec,
6974         },
6975         {
6976                 .procname       = "accept_ra_rt_info_max_plen",
6977                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6978                 .maxlen         = sizeof(int),
6979                 .mode           = 0644,
6980                 .proc_handler   = proc_dointvec,
6981         },
6982 #endif
6983 #endif
6984         {
6985                 .procname       = "proxy_ndp",
6986                 .data           = &ipv6_devconf.proxy_ndp,
6987                 .maxlen         = sizeof(int),
6988                 .mode           = 0644,
6989                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6990         },
6991         {
6992                 .procname       = "accept_source_route",
6993                 .data           = &ipv6_devconf.accept_source_route,
6994                 .maxlen         = sizeof(int),
6995                 .mode           = 0644,
6996                 .proc_handler   = proc_dointvec,
6997         },
6998 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6999         {
7000                 .procname       = "optimistic_dad",
7001                 .data           = &ipv6_devconf.optimistic_dad,
7002                 .maxlen         = sizeof(int),
7003                 .mode           = 0644,
7004                 .proc_handler   = proc_dointvec,
7005         },
7006         {
7007                 .procname       = "use_optimistic",
7008                 .data           = &ipv6_devconf.use_optimistic,
7009                 .maxlen         = sizeof(int),
7010                 .mode           = 0644,
7011                 .proc_handler   = proc_dointvec,
7012         },
7013 #endif
7014 #ifdef CONFIG_IPV6_MROUTE
7015         {
7016                 .procname       = "mc_forwarding",
7017                 .data           = &ipv6_devconf.mc_forwarding,
7018                 .maxlen         = sizeof(int),
7019                 .mode           = 0444,
7020                 .proc_handler   = proc_dointvec,
7021         },
7022 #endif
7023         {
7024                 .procname       = "disable_ipv6",
7025                 .data           = &ipv6_devconf.disable_ipv6,
7026                 .maxlen         = sizeof(int),
7027                 .mode           = 0644,
7028                 .proc_handler   = addrconf_sysctl_disable,
7029         },
7030         {
7031                 .procname       = "accept_dad",
7032                 .data           = &ipv6_devconf.accept_dad,
7033                 .maxlen         = sizeof(int),
7034                 .mode           = 0644,
7035                 .proc_handler   = proc_dointvec,
7036         },
7037         {
7038                 .procname       = "force_tllao",
7039                 .data           = &ipv6_devconf.force_tllao,
7040                 .maxlen         = sizeof(int),
7041                 .mode           = 0644,
7042                 .proc_handler   = proc_dointvec
7043         },
7044         {
7045                 .procname       = "ndisc_notify",
7046                 .data           = &ipv6_devconf.ndisc_notify,
7047                 .maxlen         = sizeof(int),
7048                 .mode           = 0644,
7049                 .proc_handler   = proc_dointvec
7050         },
7051         {
7052                 .procname       = "suppress_frag_ndisc",
7053                 .data           = &ipv6_devconf.suppress_frag_ndisc,
7054                 .maxlen         = sizeof(int),
7055                 .mode           = 0644,
7056                 .proc_handler   = proc_dointvec
7057         },
7058         {
7059                 .procname       = "accept_ra_from_local",
7060                 .data           = &ipv6_devconf.accept_ra_from_local,
7061                 .maxlen         = sizeof(int),
7062                 .mode           = 0644,
7063                 .proc_handler   = proc_dointvec,
7064         },
7065         {
7066                 .procname       = "accept_ra_mtu",
7067                 .data           = &ipv6_devconf.accept_ra_mtu,
7068                 .maxlen         = sizeof(int),
7069                 .mode           = 0644,
7070                 .proc_handler   = proc_dointvec,
7071         },
7072         {
7073                 .procname       = "stable_secret",
7074                 .data           = &ipv6_devconf.stable_secret,
7075                 .maxlen         = IPV6_MAX_STRLEN,
7076                 .mode           = 0600,
7077                 .proc_handler   = addrconf_sysctl_stable_secret,
7078         },
7079         {
7080                 .procname       = "use_oif_addrs_only",
7081                 .data           = &ipv6_devconf.use_oif_addrs_only,
7082                 .maxlen         = sizeof(int),
7083                 .mode           = 0644,
7084                 .proc_handler   = proc_dointvec,
7085         },
7086         {
7087                 .procname       = "ignore_routes_with_linkdown",
7088                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
7089                 .maxlen         = sizeof(int),
7090                 .mode           = 0644,
7091                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
7092         },
7093         {
7094                 .procname       = "drop_unicast_in_l2_multicast",
7095                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
7096                 .maxlen         = sizeof(int),
7097                 .mode           = 0644,
7098                 .proc_handler   = proc_dointvec,
7099         },
7100         {
7101                 .procname       = "drop_unsolicited_na",
7102                 .data           = &ipv6_devconf.drop_unsolicited_na,
7103                 .maxlen         = sizeof(int),
7104                 .mode           = 0644,
7105                 .proc_handler   = proc_dointvec,
7106         },
7107         {
7108                 .procname       = "keep_addr_on_down",
7109                 .data           = &ipv6_devconf.keep_addr_on_down,
7110                 .maxlen         = sizeof(int),
7111                 .mode           = 0644,
7112                 .proc_handler   = proc_dointvec,
7113
7114         },
7115         {
7116                 .procname       = "seg6_enabled",
7117                 .data           = &ipv6_devconf.seg6_enabled,
7118                 .maxlen         = sizeof(int),
7119                 .mode           = 0644,
7120                 .proc_handler   = proc_dointvec,
7121         },
7122 #ifdef CONFIG_IPV6_SEG6_HMAC
7123         {
7124                 .procname       = "seg6_require_hmac",
7125                 .data           = &ipv6_devconf.seg6_require_hmac,
7126                 .maxlen         = sizeof(int),
7127                 .mode           = 0644,
7128                 .proc_handler   = proc_dointvec,
7129         },
7130 #endif
7131         {
7132                 .procname       = "enhanced_dad",
7133                 .data           = &ipv6_devconf.enhanced_dad,
7134                 .maxlen         = sizeof(int),
7135                 .mode           = 0644,
7136                 .proc_handler   = proc_dointvec,
7137         },
7138         {
7139                 .procname       = "addr_gen_mode",
7140                 .data           = &ipv6_devconf.addr_gen_mode,
7141                 .maxlen         = sizeof(int),
7142                 .mode           = 0644,
7143                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
7144         },
7145         {
7146                 .procname       = "disable_policy",
7147                 .data           = &ipv6_devconf.disable_policy,
7148                 .maxlen         = sizeof(int),
7149                 .mode           = 0644,
7150                 .proc_handler   = addrconf_sysctl_disable_policy,
7151         },
7152         {
7153                 .procname       = "ndisc_tclass",
7154                 .data           = &ipv6_devconf.ndisc_tclass,
7155                 .maxlen         = sizeof(int),
7156                 .mode           = 0644,
7157                 .proc_handler   = proc_dointvec_minmax,
7158                 .extra1         = (void *)SYSCTL_ZERO,
7159                 .extra2         = (void *)&two_five_five,
7160         },
7161         {
7162                 .procname       = "rpl_seg_enabled",
7163                 .data           = &ipv6_devconf.rpl_seg_enabled,
7164                 .maxlen         = sizeof(int),
7165                 .mode           = 0644,
7166                 .proc_handler   = proc_dointvec,
7167         },
7168         {
7169                 .procname       = "ioam6_enabled",
7170                 .data           = &ipv6_devconf.ioam6_enabled,
7171                 .maxlen         = sizeof(u8),
7172                 .mode           = 0644,
7173                 .proc_handler   = proc_dou8vec_minmax,
7174                 .extra1         = (void *)SYSCTL_ZERO,
7175                 .extra2         = (void *)SYSCTL_ONE,
7176         },
7177         {
7178                 .procname       = "ioam6_id",
7179                 .data           = &ipv6_devconf.ioam6_id,
7180                 .maxlen         = sizeof(u32),
7181                 .mode           = 0644,
7182                 .proc_handler   = proc_douintvec_minmax,
7183                 .extra1         = (void *)SYSCTL_ZERO,
7184                 .extra2         = (void *)&ioam6_if_id_max,
7185         },
7186         {
7187                 .procname       = "ioam6_id_wide",
7188                 .data           = &ipv6_devconf.ioam6_id_wide,
7189                 .maxlen         = sizeof(u32),
7190                 .mode           = 0644,
7191                 .proc_handler   = proc_douintvec,
7192         },
7193         {
7194                 .procname       = "ndisc_evict_nocarrier",
7195                 .data           = &ipv6_devconf.ndisc_evict_nocarrier,
7196                 .maxlen         = sizeof(u8),
7197                 .mode           = 0644,
7198                 .proc_handler   = proc_dou8vec_minmax,
7199                 .extra1         = (void *)SYSCTL_ZERO,
7200                 .extra2         = (void *)SYSCTL_ONE,
7201         },
7202         {
7203                 .procname       = "accept_untracked_na",
7204                 .data           = &ipv6_devconf.accept_untracked_na,
7205                 .maxlen         = sizeof(int),
7206                 .mode           = 0644,
7207                 .proc_handler   = proc_dointvec_minmax,
7208                 .extra1         = SYSCTL_ZERO,
7209                 .extra2         = SYSCTL_TWO,
7210         },
7211 };
7212
7213 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7214                 struct inet6_dev *idev, struct ipv6_devconf *p)
7215 {
7216         size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7217         int i, ifindex;
7218         struct ctl_table *table;
7219         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7220
7221         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7222         if (!table)
7223                 goto out;
7224
7225         for (i = 0; i < table_size; i++) {
7226                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7227                 /* If one of these is already set, then it is not safe to
7228                  * overwrite either of them: this makes proc_dointvec_minmax
7229                  * usable.
7230                  */
7231                 if (!table[i].extra1 && !table[i].extra2) {
7232                         table[i].extra1 = idev; /* embedded; no ref */
7233                         table[i].extra2 = net;
7234                 }
7235         }
7236
7237         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7238
7239         p->sysctl_header = register_net_sysctl_sz(net, path, table,
7240                                                   table_size);
7241         if (!p->sysctl_header)
7242                 goto free;
7243
7244         if (!strcmp(dev_name, "all"))
7245                 ifindex = NETCONFA_IFINDEX_ALL;
7246         else if (!strcmp(dev_name, "default"))
7247                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7248         else
7249                 ifindex = idev->dev->ifindex;
7250         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7251                                      ifindex, p);
7252         return 0;
7253
7254 free:
7255         kfree(table);
7256 out:
7257         return -ENOBUFS;
7258 }
7259
7260 static void __addrconf_sysctl_unregister(struct net *net,
7261                                          struct ipv6_devconf *p, int ifindex)
7262 {
7263         const struct ctl_table *table;
7264
7265         if (!p->sysctl_header)
7266                 return;
7267
7268         table = p->sysctl_header->ctl_table_arg;
7269         unregister_net_sysctl_table(p->sysctl_header);
7270         p->sysctl_header = NULL;
7271         kfree(table);
7272
7273         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7274 }
7275
7276 static int addrconf_sysctl_register(struct inet6_dev *idev)
7277 {
7278         int err;
7279
7280         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7281                 return -EINVAL;
7282
7283         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7284                                     &ndisc_ifinfo_sysctl_change);
7285         if (err)
7286                 return err;
7287         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7288                                          idev, &idev->cnf);
7289         if (err)
7290                 neigh_sysctl_unregister(idev->nd_parms);
7291
7292         return err;
7293 }
7294
7295 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7296 {
7297         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7298                                      idev->dev->ifindex);
7299         neigh_sysctl_unregister(idev->nd_parms);
7300 }
7301
7302
7303 #endif
7304
7305 static int __net_init addrconf_init_net(struct net *net)
7306 {
7307         int err = -ENOMEM;
7308         struct ipv6_devconf *all, *dflt;
7309
7310         spin_lock_init(&net->ipv6.addrconf_hash_lock);
7311         INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7312         net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7313                                            sizeof(struct hlist_head),
7314                                            GFP_KERNEL);
7315         if (!net->ipv6.inet6_addr_lst)
7316                 goto err_alloc_addr;
7317
7318         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7319         if (!all)
7320                 goto err_alloc_all;
7321
7322         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7323         if (!dflt)
7324                 goto err_alloc_dflt;
7325
7326         if (!net_eq(net, &init_net)) {
7327                 switch (net_inherit_devconf()) {
7328                 case 1:  /* copy from init_net */
7329                         memcpy(all, init_net.ipv6.devconf_all,
7330                                sizeof(ipv6_devconf));
7331                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7332                                sizeof(ipv6_devconf_dflt));
7333                         break;
7334                 case 3: /* copy from the current netns */
7335                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7336                                sizeof(ipv6_devconf));
7337                         memcpy(dflt,
7338                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7339                                sizeof(ipv6_devconf_dflt));
7340                         break;
7341                 case 0:
7342                 case 2:
7343                         /* use compiled values */
7344                         break;
7345                 }
7346         }
7347
7348         /* these will be inherited by all namespaces */
7349         dflt->autoconf = ipv6_defaults.autoconf;
7350         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7351
7352         dflt->stable_secret.initialized = false;
7353         all->stable_secret.initialized = false;
7354
7355         net->ipv6.devconf_all = all;
7356         net->ipv6.devconf_dflt = dflt;
7357
7358 #ifdef CONFIG_SYSCTL
7359         err = __addrconf_sysctl_register(net, "all", NULL, all);
7360         if (err < 0)
7361                 goto err_reg_all;
7362
7363         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7364         if (err < 0)
7365                 goto err_reg_dflt;
7366 #endif
7367         return 0;
7368
7369 #ifdef CONFIG_SYSCTL
7370 err_reg_dflt:
7371         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7372 err_reg_all:
7373         kfree(dflt);
7374         net->ipv6.devconf_dflt = NULL;
7375 #endif
7376 err_alloc_dflt:
7377         kfree(all);
7378         net->ipv6.devconf_all = NULL;
7379 err_alloc_all:
7380         kfree(net->ipv6.inet6_addr_lst);
7381 err_alloc_addr:
7382         return err;
7383 }
7384
7385 static void __net_exit addrconf_exit_net(struct net *net)
7386 {
7387         int i;
7388
7389 #ifdef CONFIG_SYSCTL
7390         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7391                                      NETCONFA_IFINDEX_DEFAULT);
7392         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7393                                      NETCONFA_IFINDEX_ALL);
7394 #endif
7395         kfree(net->ipv6.devconf_dflt);
7396         net->ipv6.devconf_dflt = NULL;
7397         kfree(net->ipv6.devconf_all);
7398         net->ipv6.devconf_all = NULL;
7399
7400         cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7401         /*
7402          *      Check hash table, then free it.
7403          */
7404         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7405                 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7406
7407         kfree(net->ipv6.inet6_addr_lst);
7408         net->ipv6.inet6_addr_lst = NULL;
7409 }
7410
7411 static struct pernet_operations addrconf_ops = {
7412         .init = addrconf_init_net,
7413         .exit = addrconf_exit_net,
7414 };
7415
7416 static struct rtnl_af_ops inet6_ops __read_mostly = {
7417         .family           = AF_INET6,
7418         .fill_link_af     = inet6_fill_link_af,
7419         .get_link_af_size = inet6_get_link_af_size,
7420         .validate_link_af = inet6_validate_link_af,
7421         .set_link_af      = inet6_set_link_af,
7422 };
7423
7424 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = {
7425         {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK,
7426          .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED},
7427         {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR,
7428          .doit = inet6_rtm_newaddr},
7429         {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR,
7430          .doit = inet6_rtm_deladdr},
7431         {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR,
7432          .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr,
7433          .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7434         {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST,
7435          .dumpit = inet6_dump_ifmcaddr,
7436          .flags = RTNL_FLAG_DUMP_UNLOCKED},
7437         {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST,
7438          .dumpit = inet6_dump_ifacaddr,
7439          .flags = RTNL_FLAG_DUMP_UNLOCKED},
7440         {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF,
7441          .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf,
7442          .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7443 };
7444
7445 /*
7446  *      Init / cleanup code
7447  */
7448
7449 int __init addrconf_init(void)
7450 {
7451         struct inet6_dev *idev;
7452         int err;
7453
7454         err = ipv6_addr_label_init();
7455         if (err < 0) {
7456                 pr_crit("%s: cannot initialize default policy table: %d\n",
7457                         __func__, err);
7458                 goto out;
7459         }
7460
7461         err = register_pernet_subsys(&addrconf_ops);
7462         if (err < 0)
7463                 goto out_addrlabel;
7464
7465         /* All works using addrconf_wq need to lock rtnl. */
7466         addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7467         if (!addrconf_wq) {
7468                 err = -ENOMEM;
7469                 goto out_nowq;
7470         }
7471
7472         rtnl_lock();
7473         idev = ipv6_add_dev(blackhole_netdev);
7474         rtnl_unlock();
7475         if (IS_ERR(idev)) {
7476                 err = PTR_ERR(idev);
7477                 goto errlo;
7478         }
7479
7480         ip6_route_init_special_entries();
7481
7482         register_netdevice_notifier(&ipv6_dev_notf);
7483
7484         addrconf_verify(&init_net);
7485
7486         err = rtnl_af_register(&inet6_ops);
7487         if (err)
7488                 goto erraf;
7489
7490         err = rtnl_register_many(addrconf_rtnl_msg_handlers);
7491         if (err)
7492                 goto errout;
7493
7494         err = ipv6_addr_label_rtnl_register();
7495         if (err < 0)
7496                 goto errout;
7497
7498         return 0;
7499 errout:
7500         rtnl_unregister_all(PF_INET6);
7501         rtnl_af_unregister(&inet6_ops);
7502 erraf:
7503         unregister_netdevice_notifier(&ipv6_dev_notf);
7504 errlo:
7505         destroy_workqueue(addrconf_wq);
7506 out_nowq:
7507         unregister_pernet_subsys(&addrconf_ops);
7508 out_addrlabel:
7509         ipv6_addr_label_cleanup();
7510 out:
7511         return err;
7512 }
7513
7514 void addrconf_cleanup(void)
7515 {
7516         struct net_device *dev;
7517
7518         unregister_netdevice_notifier(&ipv6_dev_notf);
7519         unregister_pernet_subsys(&addrconf_ops);
7520         ipv6_addr_label_cleanup();
7521
7522         rtnl_af_unregister(&inet6_ops);
7523
7524         rtnl_lock();
7525
7526         /* clean dev list */
7527         for_each_netdev(&init_net, dev) {
7528                 if (__in6_dev_get(dev) == NULL)
7529                         continue;
7530                 addrconf_ifdown(dev, true);
7531         }
7532         addrconf_ifdown(init_net.loopback_dev, true);
7533
7534         rtnl_unlock();
7535
7536         destroy_workqueue(addrconf_wq);
7537 }