net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92 #include <linux/ioam6.h>
  93
  94 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  95
  96 #define IPV6_MAX_STRLEN \
  97         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  98
  99 static inline u32 cstamp_delta(unsigned long cstamp)
 100 {
 101         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 102 }
 103
 104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 105 {
 106         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 107         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 108         do_div(tmp, 1000000);
 109         return (s32)tmp;
 110 }
 111
 112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 113 {
 114         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 115         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 116         do_div(tmp, 1000000);
 117         if ((s32)tmp > mrt) {
 118                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 119                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 120                 do_div(tmp, 1000000);
 121         }
 122         return (s32)tmp;
 123 }
 124
 125 #ifdef CONFIG_SYSCTL
 126 static int addrconf_sysctl_register(struct inet6_dev *idev);
 127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 128 #else
 129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 130 {
 131         return 0;
 132 }
 133
 134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 135 {
 136 }
 137 #endif
 138
 139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 140
 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 142 static int ipv6_count_addresses(const struct inet6_dev *idev);
 143 static int ipv6_generate_stable_address(struct in6_addr *addr,
 144                                         u8 dad_count,
 145                                         const struct inet6_dev *idev);
 146
 147 #define IN6_ADDR_HSIZE_SHIFT    8
 148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 149 /*
 150  *      Configured unicast address hash table
 151  */
 152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 153 static DEFINE_SPINLOCK(addrconf_hash_lock);
 154
 155 static void addrconf_verify(void);
 156 static void addrconf_verify_rtnl(void);
 157 static void addrconf_verify_work(struct work_struct *);
 158
 159 static struct workqueue_struct *addrconf_wq;
 160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 161
 162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 164
 165 static void addrconf_type_change(struct net_device *dev,
 166                                  unsigned long event);
 167 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 168
 169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 170                                                   int plen,
 171                                                   const struct net_device *dev,
 172                                                   u32 flags, u32 noflags,
 173                                                   bool no_gw);
 174
 175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 176 static void addrconf_dad_work(struct work_struct *w);
 177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 178                                    bool send_na);
 179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 180 static void addrconf_rs_timer(struct timer_list *t);
 181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 183
 184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 185                                 struct prefix_info *pinfo);
 186
 187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 188         .forwarding             = 0,
 189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 190         .mtu6                   = IPV6_MIN_MTU,
 191         .accept_ra              = 1,
 192         .accept_redirects       = 1,
 193         .autoconf               = 1,
 194         .force_mld_version      = 0,
 195         .mldv1_unsolicited_report_interval = 10 * HZ,
 196         .mldv2_unsolicited_report_interval = HZ,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 210         .accept_ra_from_local   = 0,
 211         .accept_ra_min_hop_limit= 1,
 212         .accept_ra_pinfo        = 1,
 213 #ifdef CONFIG_IPV6_ROUTER_PREF
 214         .accept_ra_rtr_pref     = 1,
 215         .rtr_probe_interval     = 60 * HZ,
 216 #ifdef CONFIG_IPV6_ROUTE_INFO
 217         .accept_ra_rt_info_min_plen = 0,
 218         .accept_ra_rt_info_max_plen = 0,
 219 #endif
 220 #endif
 221         .proxy_ndp              = 0,
 222         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 223         .disable_ipv6           = 0,
 224         .accept_dad             = 0,
 225         .suppress_frag_ndisc    = 1,
 226         .accept_ra_mtu          = 1,
 227         .stable_secret          = {
 228                 .initialized = false,
 229         },
 230         .use_oif_addrs_only     = 0,
 231         .ignore_routes_with_linkdown = 0,
 232         .keep_addr_on_down      = 0,
 233         .seg6_enabled           = 0,
 234 #ifdef CONFIG_IPV6_SEG6_HMAC
 235         .seg6_require_hmac      = 0,
 236 #endif
 237         .enhanced_dad           = 1,
 238         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 239         .disable_policy         = 0,
 240         .rpl_seg_enabled        = 0,
 241         .ioam6_enabled          = 0,
 242         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 243         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 244 };
 245
 246 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 247         .forwarding             = 0,
 248         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 249         .mtu6                   = IPV6_MIN_MTU,
 250         .accept_ra              = 1,
 251         .accept_redirects       = 1,
 252         .autoconf               = 1,
 253         .force_mld_version      = 0,
 254         .mldv1_unsolicited_report_interval = 10 * HZ,
 255         .mldv2_unsolicited_report_interval = HZ,
 256         .dad_transmits          = 1,
 257         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 258         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 259         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 260         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 261         .use_tempaddr           = 0,
 262         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 263         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 264         .regen_max_retry        = REGEN_MAX_RETRY,
 265         .max_desync_factor      = MAX_DESYNC_FACTOR,
 266         .max_addresses          = IPV6_MAX_ADDRESSES,
 267         .accept_ra_defrtr       = 1,
 268         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 269         .accept_ra_from_local   = 0,
 270         .accept_ra_min_hop_limit= 1,
 271         .accept_ra_pinfo        = 1,
 272 #ifdef CONFIG_IPV6_ROUTER_PREF
 273         .accept_ra_rtr_pref     = 1,
 274         .rtr_probe_interval     = 60 * HZ,
 275 #ifdef CONFIG_IPV6_ROUTE_INFO
 276         .accept_ra_rt_info_min_plen = 0,
 277         .accept_ra_rt_info_max_plen = 0,
 278 #endif
 279 #endif
 280         .proxy_ndp              = 0,
 281         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 282         .disable_ipv6           = 0,
 283         .accept_dad             = 1,
 284         .suppress_frag_ndisc    = 1,
 285         .accept_ra_mtu          = 1,
 286         .stable_secret          = {
 287                 .initialized = false,
 288         },
 289         .use_oif_addrs_only     = 0,
 290         .ignore_routes_with_linkdown = 0,
 291         .keep_addr_on_down      = 0,
 292         .seg6_enabled           = 0,
 293 #ifdef CONFIG_IPV6_SEG6_HMAC
 294         .seg6_require_hmac      = 0,
 295 #endif
 296         .enhanced_dad           = 1,
 297         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 298         .disable_policy         = 0,
 299         .rpl_seg_enabled        = 0,
 300         .ioam6_enabled          = 0,
 301         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 302         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 303 };
 304
 305 /* Check if link is ready: is it up and is a valid qdisc available */
 306 static inline bool addrconf_link_ready(const struct net_device *dev)
 307 {
 308         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 309 }
 310
 311 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 312 {
 313         if (del_timer(&idev->rs_timer))
 314                 __in6_dev_put(idev);
 315 }
 316
 317 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 318 {
 319         if (cancel_delayed_work(&ifp->dad_work))
 320                 __in6_ifa_put(ifp);
 321 }
 322
 323 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 324                                   unsigned long when)
 325 {
 326         if (!timer_pending(&idev->rs_timer))
 327                 in6_dev_hold(idev);
 328         mod_timer(&idev->rs_timer, jiffies + when);
 329 }
 330
 331 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 332                                    unsigned long delay)
 333 {
 334         in6_ifa_hold(ifp);
 335         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 336                 in6_ifa_put(ifp);
 337 }
 338
 339 static int snmp6_alloc_dev(struct inet6_dev *idev)
 340 {
 341         int i;
 342
 343         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 344         if (!idev->stats.ipv6)
 345                 goto err_ip;
 346
 347         for_each_possible_cpu(i) {
 348                 struct ipstats_mib *addrconf_stats;
 349                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 350                 u64_stats_init(&addrconf_stats->syncp);
 351         }
 352
 353
 354         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 355                                         GFP_KERNEL);
 356         if (!idev->stats.icmpv6dev)
 357                 goto err_icmp;
 358         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 359                                            GFP_KERNEL);
 360         if (!idev->stats.icmpv6msgdev)
 361                 goto err_icmpmsg;
 362
 363         return 0;
 364
 365 err_icmpmsg:
 366         kfree(idev->stats.icmpv6dev);
 367 err_icmp:
 368         free_percpu(idev->stats.ipv6);
 369 err_ip:
 370         return -ENOMEM;
 371 }
 372
 373 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 374 {
 375         struct inet6_dev *ndev;
 376         int err = -ENOMEM;
 377
 378         ASSERT_RTNL();
 379
 380         if (dev->mtu < IPV6_MIN_MTU)
 381                 return ERR_PTR(-EINVAL);
 382
 383         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 384         if (!ndev)
 385                 return ERR_PTR(err);
 386
 387         rwlock_init(&ndev->lock);
 388         ndev->dev = dev;
 389         INIT_LIST_HEAD(&ndev->addr_list);
 390         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 391         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 392
 393         if (ndev->cnf.stable_secret.initialized)
 394                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 395
 396         ndev->cnf.mtu6 = dev->mtu;
 397         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 398         if (!ndev->nd_parms) {
 399                 kfree(ndev);
 400                 return ERR_PTR(err);
 401         }
 402         if (ndev->cnf.forwarding)
 403                 dev_disable_lro(dev);
 404         /* We refer to the device */
 405         dev_hold(dev);
 406
 407         if (snmp6_alloc_dev(ndev) < 0) {
 408                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 409                            __func__);
 410                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 411                 dev_put(dev);
 412                 kfree(ndev);
 413                 return ERR_PTR(err);
 414         }
 415
 416         if (snmp6_register_dev(ndev) < 0) {
 417                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 418                            __func__, dev->name);
 419                 goto err_release;
 420         }
 421
 422         /* One reference from device. */
 423         refcount_set(&ndev->refcnt, 1);
 424
 425         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 426                 ndev->cnf.accept_dad = -1;
 427
 428 #if IS_ENABLED(CONFIG_IPV6_SIT)
 429         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 430                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 431                 ndev->cnf.rtr_solicits = 0;
 432         }
 433 #endif
 434
 435         INIT_LIST_HEAD(&ndev->tempaddr_list);
 436         ndev->desync_factor = U32_MAX;
 437         if ((dev->flags&IFF_LOOPBACK) ||
 438             dev->type == ARPHRD_TUNNEL ||
 439             dev->type == ARPHRD_TUNNEL6 ||
 440             dev->type == ARPHRD_SIT ||
 441             dev->type == ARPHRD_NONE) {
 442                 ndev->cnf.use_tempaddr = -1;
 443         }
 444
 445         ndev->token = in6addr_any;
 446
 447         if (netif_running(dev) && addrconf_link_ready(dev))
 448                 ndev->if_flags |= IF_READY;
 449
 450         ipv6_mc_init_dev(ndev);
 451         ndev->tstamp = jiffies;
 452         err = addrconf_sysctl_register(ndev);
 453         if (err) {
 454                 ipv6_mc_destroy_dev(ndev);
 455                 snmp6_unregister_dev(ndev);
 456                 goto err_release;
 457         }
 458         /* protected by rtnl_lock */
 459         rcu_assign_pointer(dev->ip6_ptr, ndev);
 460
 461         /* Join interface-local all-node multicast group */
 462         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 463
 464         /* Join all-node multicast group */
 465         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 466
 467         /* Join all-router multicast group if forwarding is set */
 468         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 469                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 470
 471         return ndev;
 472
 473 err_release:
 474         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 475         ndev->dead = 1;
 476         in6_dev_finish_destroy(ndev);
 477         return ERR_PTR(err);
 478 }
 479
 480 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 481 {
 482         struct inet6_dev *idev;
 483
 484         ASSERT_RTNL();
 485
 486         idev = __in6_dev_get(dev);
 487         if (!idev) {
 488                 idev = ipv6_add_dev(dev);
 489                 if (IS_ERR(idev))
 490                         return idev;
 491         }
 492
 493         if (dev->flags&IFF_UP)
 494                 ipv6_mc_up(idev);
 495         return idev;
 496 }
 497
 498 static int inet6_netconf_msgsize_devconf(int type)
 499 {
 500         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 501                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 502         bool all = false;
 503
 504         if (type == NETCONFA_ALL)
 505                 all = true;
 506
 507         if (all || type == NETCONFA_FORWARDING)
 508                 size += nla_total_size(4);
 509 #ifdef CONFIG_IPV6_MROUTE
 510         if (all || type == NETCONFA_MC_FORWARDING)
 511                 size += nla_total_size(4);
 512 #endif
 513         if (all || type == NETCONFA_PROXY_NEIGH)
 514                 size += nla_total_size(4);
 515
 516         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 517                 size += nla_total_size(4);
 518
 519         return size;
 520 }
 521
 522 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 523                                       struct ipv6_devconf *devconf, u32 portid,
 524                                       u32 seq, int event, unsigned int flags,
 525                                       int type)
 526 {
 527         struct nlmsghdr  *nlh;
 528         struct netconfmsg *ncm;
 529         bool all = false;
 530
 531         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 532                         flags);
 533         if (!nlh)
 534                 return -EMSGSIZE;
 535
 536         if (type == NETCONFA_ALL)
 537                 all = true;
 538
 539         ncm = nlmsg_data(nlh);
 540         ncm->ncm_family = AF_INET6;
 541
 542         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 543                 goto nla_put_failure;
 544
 545         if (!devconf)
 546                 goto out;
 547
 548         if ((all || type == NETCONFA_FORWARDING) &&
 549             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 550                 goto nla_put_failure;
 551 #ifdef CONFIG_IPV6_MROUTE
 552         if ((all || type == NETCONFA_MC_FORWARDING) &&
 553             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 554                         devconf->mc_forwarding) < 0)
 555                 goto nla_put_failure;
 556 #endif
 557         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 558             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 559                 goto nla_put_failure;
 560
 561         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 562             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 563                         devconf->ignore_routes_with_linkdown) < 0)
 564                 goto nla_put_failure;
 565
 566 out:
 567         nlmsg_end(skb, nlh);
 568         return 0;
 569
 570 nla_put_failure:
 571         nlmsg_cancel(skb, nlh);
 572         return -EMSGSIZE;
 573 }
 574
 575 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 576                                   int ifindex, struct ipv6_devconf *devconf)
 577 {
 578         struct sk_buff *skb;
 579         int err = -ENOBUFS;
 580
 581         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 582         if (!skb)
 583                 goto errout;
 584
 585         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 586                                          event, 0, type);
 587         if (err < 0) {
 588                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 589                 WARN_ON(err == -EMSGSIZE);
 590                 kfree_skb(skb);
 591                 goto errout;
 592         }
 593         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 594         return;
 595 errout:
 596         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 597 }
 598
 599 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 600         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 601         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 602         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 603         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 604 };
 605
 606 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 607                                        const struct nlmsghdr *nlh,
 608                                        struct nlattr **tb,
 609                                        struct netlink_ext_ack *extack)
 610 {
 611         int i, err;
 612
 613         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 614                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 615                 return -EINVAL;
 616         }
 617
 618         if (!netlink_strict_get_check(skb))
 619                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 620                                               tb, NETCONFA_MAX,
 621                                               devconf_ipv6_policy, extack);
 622
 623         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 624                                             tb, NETCONFA_MAX,
 625                                             devconf_ipv6_policy, extack);
 626         if (err)
 627                 return err;
 628
 629         for (i = 0; i <= NETCONFA_MAX; i++) {
 630                 if (!tb[i])
 631                         continue;
 632
 633                 switch (i) {
 634                 case NETCONFA_IFINDEX:
 635                         break;
 636                 default:
 637                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 638                         return -EINVAL;
 639                 }
 640         }
 641
 642         return 0;
 643 }
 644
 645 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 646                                      struct nlmsghdr *nlh,
 647                                      struct netlink_ext_ack *extack)
 648 {
 649         struct net *net = sock_net(in_skb->sk);
 650         struct nlattr *tb[NETCONFA_MAX+1];
 651         struct inet6_dev *in6_dev = NULL;
 652         struct net_device *dev = NULL;
 653         struct sk_buff *skb;
 654         struct ipv6_devconf *devconf;
 655         int ifindex;
 656         int err;
 657
 658         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 659         if (err < 0)
 660                 return err;
 661
 662         if (!tb[NETCONFA_IFINDEX])
 663                 return -EINVAL;
 664
 665         err = -EINVAL;
 666         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 667         switch (ifindex) {
 668         case NETCONFA_IFINDEX_ALL:
 669                 devconf = net->ipv6.devconf_all;
 670                 break;
 671         case NETCONFA_IFINDEX_DEFAULT:
 672                 devconf = net->ipv6.devconf_dflt;
 673                 break;
 674         default:
 675                 dev = dev_get_by_index(net, ifindex);
 676                 if (!dev)
 677                         return -EINVAL;
 678                 in6_dev = in6_dev_get(dev);
 679                 if (!in6_dev)
 680                         goto errout;
 681                 devconf = &in6_dev->cnf;
 682                 break;
 683         }
 684
 685         err = -ENOBUFS;
 686         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 687         if (!skb)
 688                 goto errout;
 689
 690         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 691                                          NETLINK_CB(in_skb).portid,
 692                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 693                                          NETCONFA_ALL);
 694         if (err < 0) {
 695                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 696                 WARN_ON(err == -EMSGSIZE);
 697                 kfree_skb(skb);
 698                 goto errout;
 699         }
 700         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 701 errout:
 702         if (in6_dev)
 703                 in6_dev_put(in6_dev);
 704         dev_put(dev);
 705         return err;
 706 }
 707
 708 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 709                                       struct netlink_callback *cb)
 710 {
 711         const struct nlmsghdr *nlh = cb->nlh;
 712         struct net *net = sock_net(skb->sk);
 713         int h, s_h;
 714         int idx, s_idx;
 715         struct net_device *dev;
 716         struct inet6_dev *idev;
 717         struct hlist_head *head;
 718
 719         if (cb->strict_check) {
 720                 struct netlink_ext_ack *extack = cb->extack;
 721                 struct netconfmsg *ncm;
 722
 723                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 724                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 725                         return -EINVAL;
 726                 }
 727
 728                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 729                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 730                         return -EINVAL;
 731                 }
 732         }
 733
 734         s_h = cb->args[0];
 735         s_idx = idx = cb->args[1];
 736
 737         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 738                 idx = 0;
 739                 head = &net->dev_index_head[h];
 740                 rcu_read_lock();
 741                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 742                           net->dev_base_seq;
 743                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 744                         if (idx < s_idx)
 745                                 goto cont;
 746                         idev = __in6_dev_get(dev);
 747                         if (!idev)
 748                                 goto cont;
 749
 750                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 751                                                        &idev->cnf,
 752                                                        NETLINK_CB(cb->skb).portid,
 753                                                        nlh->nlmsg_seq,
 754                                                        RTM_NEWNETCONF,
 755                                                        NLM_F_MULTI,
 756                                                        NETCONFA_ALL) < 0) {
 757                                 rcu_read_unlock();
 758                                 goto done;
 759                         }
 760                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 761 cont:
 762                         idx++;
 763                 }
 764                 rcu_read_unlock();
 765         }
 766         if (h == NETDEV_HASHENTRIES) {
 767                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 768                                                net->ipv6.devconf_all,
 769                                                NETLINK_CB(cb->skb).portid,
 770                                                nlh->nlmsg_seq,
 771                                                RTM_NEWNETCONF, NLM_F_MULTI,
 772                                                NETCONFA_ALL) < 0)
 773                         goto done;
 774                 else
 775                         h++;
 776         }
 777         if (h == NETDEV_HASHENTRIES + 1) {
 778                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 779                                                net->ipv6.devconf_dflt,
 780                                                NETLINK_CB(cb->skb).portid,
 781                                                nlh->nlmsg_seq,
 782                                                RTM_NEWNETCONF, NLM_F_MULTI,
 783                                                NETCONFA_ALL) < 0)
 784                         goto done;
 785                 else
 786                         h++;
 787         }
 788 done:
 789         cb->args[0] = h;
 790         cb->args[1] = idx;
 791
 792         return skb->len;
 793 }
 794
 795 #ifdef CONFIG_SYSCTL
 796 static void dev_forward_change(struct inet6_dev *idev)
 797 {
 798         struct net_device *dev;
 799         struct inet6_ifaddr *ifa;
 800
 801         if (!idev)
 802                 return;
 803         dev = idev->dev;
 804         if (idev->cnf.forwarding)
 805                 dev_disable_lro(dev);
 806         if (dev->flags & IFF_MULTICAST) {
 807                 if (idev->cnf.forwarding) {
 808                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 809                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 810                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 811                 } else {
 812                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 813                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 814                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 815                 }
 816         }
 817
 818         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 819                 if (ifa->flags&IFA_F_TENTATIVE)
 820                         continue;
 821                 if (idev->cnf.forwarding)
 822                         addrconf_join_anycast(ifa);
 823                 else
 824                         addrconf_leave_anycast(ifa);
 825         }
 826         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 827                                      NETCONFA_FORWARDING,
 828                                      dev->ifindex, &idev->cnf);
 829 }
 830
 831
 832 static void addrconf_forward_change(struct net *net, __s32 newf)
 833 {
 834         struct net_device *dev;
 835         struct inet6_dev *idev;
 836
 837         for_each_netdev(net, dev) {
 838                 idev = __in6_dev_get(dev);
 839                 if (idev) {
 840                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 841                         idev->cnf.forwarding = newf;
 842                         if (changed)
 843                                 dev_forward_change(idev);
 844                 }
 845         }
 846 }
 847
 848 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 849 {
 850         struct net *net;
 851         int old;
 852
 853         if (!rtnl_trylock())
 854                 return restart_syscall();
 855
 856         net = (struct net *)table->extra2;
 857         old = *p;
 858         *p = newf;
 859
 860         if (p == &net->ipv6.devconf_dflt->forwarding) {
 861                 if ((!newf) ^ (!old))
 862                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 863                                                      NETCONFA_FORWARDING,
 864                                                      NETCONFA_IFINDEX_DEFAULT,
 865                                                      net->ipv6.devconf_dflt);
 866                 rtnl_unlock();
 867                 return 0;
 868         }
 869
 870         if (p == &net->ipv6.devconf_all->forwarding) {
 871                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 872
 873                 net->ipv6.devconf_dflt->forwarding = newf;
 874                 if ((!newf) ^ (!old_dflt))
 875                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 876                                                      NETCONFA_FORWARDING,
 877                                                      NETCONFA_IFINDEX_DEFAULT,
 878                                                      net->ipv6.devconf_dflt);
 879
 880                 addrconf_forward_change(net, newf);
 881                 if ((!newf) ^ (!old))
 882                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 883                                                      NETCONFA_FORWARDING,
 884                                                      NETCONFA_IFINDEX_ALL,
 885                                                      net->ipv6.devconf_all);
 886         } else if ((!newf) ^ (!old))
 887                 dev_forward_change((struct inet6_dev *)table->extra1);
 888         rtnl_unlock();
 889
 890         if (newf)
 891                 rt6_purge_dflt_routers(net);
 892         return 1;
 893 }
 894
 895 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 896 {
 897         struct net_device *dev;
 898         struct inet6_dev *idev;
 899
 900         for_each_netdev(net, dev) {
 901                 idev = __in6_dev_get(dev);
 902                 if (idev) {
 903                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 904
 905                         idev->cnf.ignore_routes_with_linkdown = newf;
 906                         if (changed)
 907                                 inet6_netconf_notify_devconf(dev_net(dev),
 908                                                              RTM_NEWNETCONF,
 909                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 910                                                              dev->ifindex,
 911                                                              &idev->cnf);
 912                 }
 913         }
 914 }
 915
 916 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 917 {
 918         struct net *net;
 919         int old;
 920
 921         if (!rtnl_trylock())
 922                 return restart_syscall();
 923
 924         net = (struct net *)table->extra2;
 925         old = *p;
 926         *p = newf;
 927
 928         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 929                 if ((!newf) ^ (!old))
 930                         inet6_netconf_notify_devconf(net,
 931                                                      RTM_NEWNETCONF,
 932                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 933                                                      NETCONFA_IFINDEX_DEFAULT,
 934                                                      net->ipv6.devconf_dflt);
 935                 rtnl_unlock();
 936                 return 0;
 937         }
 938
 939         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 940                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 941                 addrconf_linkdown_change(net, newf);
 942                 if ((!newf) ^ (!old))
 943                         inet6_netconf_notify_devconf(net,
 944                                                      RTM_NEWNETCONF,
 945                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 946                                                      NETCONFA_IFINDEX_ALL,
 947                                                      net->ipv6.devconf_all);
 948         }
 949         rtnl_unlock();
 950
 951         return 1;
 952 }
 953
 954 #endif
 955
 956 /* Nobody refers to this ifaddr, destroy it */
 957 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 958 {
 959         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 960
 961 #ifdef NET_REFCNT_DEBUG
 962         pr_debug("%s\n", __func__);
 963 #endif
 964
 965         in6_dev_put(ifp->idev);
 966
 967         if (cancel_delayed_work(&ifp->dad_work))
 968                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 969                           ifp);
 970
 971         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 972                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 973                 return;
 974         }
 975
 976         kfree_rcu(ifp, rcu);
 977 }
 978
 979 static void
 980 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 981 {
 982         struct list_head *p;
 983         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 984
 985         /*
 986          * Each device address list is sorted in order of scope -
 987          * global before linklocal.
 988          */
 989         list_for_each(p, &idev->addr_list) {
 990                 struct inet6_ifaddr *ifa
 991                         = list_entry(p, struct inet6_ifaddr, if_list);
 992                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 993                         break;
 994         }
 995
 996         list_add_tail_rcu(&ifp->if_list, p);
 997 }
 998
 999 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1000 {
1001         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1002
1003         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1004 }
1005
1006 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1007                                struct net_device *dev, unsigned int hash)
1008 {
1009         struct inet6_ifaddr *ifp;
1010
1011         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1012                 if (!net_eq(dev_net(ifp->idev->dev), net))
1013                         continue;
1014                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1015                         if (!dev || ifp->idev->dev == dev)
1016                                 return true;
1017                 }
1018         }
1019         return false;
1020 }
1021
1022 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1023 {
1024         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1025         int err = 0;
1026
1027         spin_lock(&addrconf_hash_lock);
1028
1029         /* Ignore adding duplicate addresses on an interface */
1030         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1031                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1032                 err = -EEXIST;
1033         } else {
1034                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1035         }
1036
1037         spin_unlock(&addrconf_hash_lock);
1038
1039         return err;
1040 }
1041
1042 /* On success it returns ifp with increased reference count */
1043
1044 static struct inet6_ifaddr *
1045 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1046               bool can_block, struct netlink_ext_ack *extack)
1047 {
1048         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1049         int addr_type = ipv6_addr_type(cfg->pfx);
1050         struct net *net = dev_net(idev->dev);
1051         struct inet6_ifaddr *ifa = NULL;
1052         struct fib6_info *f6i = NULL;
1053         int err = 0;
1054
1055         if (addr_type == IPV6_ADDR_ANY ||
1056             (addr_type & IPV6_ADDR_MULTICAST &&
1057              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1058             (!(idev->dev->flags & IFF_LOOPBACK) &&
1059              !netif_is_l3_master(idev->dev) &&
1060              addr_type & IPV6_ADDR_LOOPBACK))
1061                 return ERR_PTR(-EADDRNOTAVAIL);
1062
1063         if (idev->dead) {
1064                 err = -ENODEV;                  /*XXX*/
1065                 goto out;
1066         }
1067
1068         if (idev->cnf.disable_ipv6) {
1069                 err = -EACCES;
1070                 goto out;
1071         }
1072
1073         /* validator notifier needs to be blocking;
1074          * do not call in atomic context
1075          */
1076         if (can_block) {
1077                 struct in6_validator_info i6vi = {
1078                         .i6vi_addr = *cfg->pfx,
1079                         .i6vi_dev = idev,
1080                         .extack = extack,
1081                 };
1082
1083                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1084                 err = notifier_to_errno(err);
1085                 if (err < 0)
1086                         goto out;
1087         }
1088
1089         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1090         if (!ifa) {
1091                 err = -ENOBUFS;
1092                 goto out;
1093         }
1094
1095         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1096         if (IS_ERR(f6i)) {
1097                 err = PTR_ERR(f6i);
1098                 f6i = NULL;
1099                 goto out;
1100         }
1101
1102         if (net->ipv6.devconf_all->disable_policy ||
1103             idev->cnf.disable_policy)
1104                 f6i->dst_nopolicy = true;
1105
1106         neigh_parms_data_state_setall(idev->nd_parms);
1107
1108         ifa->addr = *cfg->pfx;
1109         if (cfg->peer_pfx)
1110                 ifa->peer_addr = *cfg->peer_pfx;
1111
1112         spin_lock_init(&ifa->lock);
1113         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1114         INIT_HLIST_NODE(&ifa->addr_lst);
1115         ifa->scope = cfg->scope;
1116         ifa->prefix_len = cfg->plen;
1117         ifa->rt_priority = cfg->rt_priority;
1118         ifa->flags = cfg->ifa_flags;
1119         /* No need to add the TENTATIVE flag for addresses with NODAD */
1120         if (!(cfg->ifa_flags & IFA_F_NODAD))
1121                 ifa->flags |= IFA_F_TENTATIVE;
1122         ifa->valid_lft = cfg->valid_lft;
1123         ifa->prefered_lft = cfg->preferred_lft;
1124         ifa->cstamp = ifa->tstamp = jiffies;
1125         ifa->tokenized = false;
1126
1127         ifa->rt = f6i;
1128
1129         ifa->idev = idev;
1130         in6_dev_hold(idev);
1131
1132         /* For caller */
1133         refcount_set(&ifa->refcnt, 1);
1134
1135         rcu_read_lock_bh();
1136
1137         err = ipv6_add_addr_hash(idev->dev, ifa);
1138         if (err < 0) {
1139                 rcu_read_unlock_bh();
1140                 goto out;
1141         }
1142
1143         write_lock(&idev->lock);
1144
1145         /* Add to inet6_dev unicast addr list. */
1146         ipv6_link_dev_addr(idev, ifa);
1147
1148         if (ifa->flags&IFA_F_TEMPORARY) {
1149                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1150                 in6_ifa_hold(ifa);
1151         }
1152
1153         in6_ifa_hold(ifa);
1154         write_unlock(&idev->lock);
1155
1156         rcu_read_unlock_bh();
1157
1158         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1159 out:
1160         if (unlikely(err < 0)) {
1161                 fib6_info_release(f6i);
1162
1163                 if (ifa) {
1164                         if (ifa->idev)
1165                                 in6_dev_put(ifa->idev);
1166                         kfree(ifa);
1167                 }
1168                 ifa = ERR_PTR(err);
1169         }
1170
1171         return ifa;
1172 }
1173
1174 enum cleanup_prefix_rt_t {
1175         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1176         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1177         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1178 };
1179
1180 /*
1181  * Check, whether the prefix for ifp would still need a prefix route
1182  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1183  * constants.
1184  *
1185  * 1) we don't purge prefix if address was not permanent.
1186  *    prefix is managed by its own lifetime.
1187  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1188  * 3) if there are no addresses, delete prefix.
1189  * 4) if there are still other permanent address(es),
1190  *    corresponding prefix is still permanent.
1191  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1192  *    don't purge the prefix, assume user space is managing it.
1193  * 6) otherwise, update prefix lifetime to the
1194  *    longest valid lifetime among the corresponding
1195  *    addresses on the device.
1196  *    Note: subsequent RA will update lifetime.
1197  **/
1198 static enum cleanup_prefix_rt_t
1199 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1200 {
1201         struct inet6_ifaddr *ifa;
1202         struct inet6_dev *idev = ifp->idev;
1203         unsigned long lifetime;
1204         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1205
1206         *expires = jiffies;
1207
1208         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1209                 if (ifa == ifp)
1210                         continue;
1211                 if (ifa->prefix_len != ifp->prefix_len ||
1212                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1213                                        ifp->prefix_len))
1214                         continue;
1215                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1216                         return CLEANUP_PREFIX_RT_NOP;
1217
1218                 action = CLEANUP_PREFIX_RT_EXPIRE;
1219
1220                 spin_lock(&ifa->lock);
1221
1222                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1223                 /*
1224                  * Note: Because this address is
1225                  * not permanent, lifetime <
1226                  * LONG_MAX / HZ here.
1227                  */
1228                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1229                         *expires = ifa->tstamp + lifetime * HZ;
1230                 spin_unlock(&ifa->lock);
1231         }
1232
1233         return action;
1234 }
1235
1236 static void
1237 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1238                      bool del_rt, bool del_peer)
1239 {
1240         struct fib6_info *f6i;
1241
1242         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1243                                         ifp->prefix_len,
1244                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1245         if (f6i) {
1246                 if (del_rt)
1247                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1248                 else {
1249                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1250                                 fib6_set_expires(f6i, expires);
1251                         fib6_info_release(f6i);
1252                 }
1253         }
1254 }
1255
1256
1257 /* This function wants to get referenced ifp and releases it before return */
1258
1259 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1260 {
1261         int state;
1262         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1263         unsigned long expires;
1264
1265         ASSERT_RTNL();
1266
1267         spin_lock_bh(&ifp->lock);
1268         state = ifp->state;
1269         ifp->state = INET6_IFADDR_STATE_DEAD;
1270         spin_unlock_bh(&ifp->lock);
1271
1272         if (state == INET6_IFADDR_STATE_DEAD)
1273                 goto out;
1274
1275         spin_lock_bh(&addrconf_hash_lock);
1276         hlist_del_init_rcu(&ifp->addr_lst);
1277         spin_unlock_bh(&addrconf_hash_lock);
1278
1279         write_lock_bh(&ifp->idev->lock);
1280
1281         if (ifp->flags&IFA_F_TEMPORARY) {
1282                 list_del(&ifp->tmp_list);
1283                 if (ifp->ifpub) {
1284                         in6_ifa_put(ifp->ifpub);
1285                         ifp->ifpub = NULL;
1286                 }
1287                 __in6_ifa_put(ifp);
1288         }
1289
1290         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1291                 action = check_cleanup_prefix_route(ifp, &expires);
1292
1293         list_del_rcu(&ifp->if_list);
1294         __in6_ifa_put(ifp);
1295
1296         write_unlock_bh(&ifp->idev->lock);
1297
1298         addrconf_del_dad_work(ifp);
1299
1300         ipv6_ifa_notify(RTM_DELADDR, ifp);
1301
1302         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1303
1304         if (action != CLEANUP_PREFIX_RT_NOP) {
1305                 cleanup_prefix_route(ifp, expires,
1306                         action == CLEANUP_PREFIX_RT_DEL, false);
1307         }
1308
1309         /* clean up prefsrc entries */
1310         rt6_remove_prefsrc(ifp);
1311 out:
1312         in6_ifa_put(ifp);
1313 }
1314
1315 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1316 {
1317         struct inet6_dev *idev = ifp->idev;
1318         unsigned long tmp_tstamp, age;
1319         unsigned long regen_advance;
1320         unsigned long now = jiffies;
1321         s32 cnf_temp_preferred_lft;
1322         struct inet6_ifaddr *ift;
1323         struct ifa6_config cfg;
1324         long max_desync_factor;
1325         struct in6_addr addr;
1326         int ret = 0;
1327
1328         write_lock_bh(&idev->lock);
1329
1330 retry:
1331         in6_dev_hold(idev);
1332         if (idev->cnf.use_tempaddr <= 0) {
1333                 write_unlock_bh(&idev->lock);
1334                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1335                 in6_dev_put(idev);
1336                 ret = -1;
1337                 goto out;
1338         }
1339         spin_lock_bh(&ifp->lock);
1340         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1341                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1342                 spin_unlock_bh(&ifp->lock);
1343                 write_unlock_bh(&idev->lock);
1344                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1345                         __func__);
1346                 in6_dev_put(idev);
1347                 ret = -1;
1348                 goto out;
1349         }
1350         in6_ifa_hold(ifp);
1351         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1352         ipv6_gen_rnd_iid(&addr);
1353
1354         age = (now - ifp->tstamp) / HZ;
1355
1356         regen_advance = idev->cnf.regen_max_retry *
1357                         idev->cnf.dad_transmits *
1358                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1359
1360         /* recalculate max_desync_factor each time and update
1361          * idev->desync_factor if it's larger
1362          */
1363         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1364         max_desync_factor = min_t(__u32,
1365                                   idev->cnf.max_desync_factor,
1366                                   cnf_temp_preferred_lft - regen_advance);
1367
1368         if (unlikely(idev->desync_factor > max_desync_factor)) {
1369                 if (max_desync_factor > 0) {
1370                         get_random_bytes(&idev->desync_factor,
1371                                          sizeof(idev->desync_factor));
1372                         idev->desync_factor %= max_desync_factor;
1373                 } else {
1374                         idev->desync_factor = 0;
1375                 }
1376         }
1377
1378         memset(&cfg, 0, sizeof(cfg));
1379         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1380                               idev->cnf.temp_valid_lft + age);
1381         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1382         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1383
1384         cfg.plen = ifp->prefix_len;
1385         tmp_tstamp = ifp->tstamp;
1386         spin_unlock_bh(&ifp->lock);
1387
1388         write_unlock_bh(&idev->lock);
1389
1390         /* A temporary address is created only if this calculated Preferred
1391          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1392          * an implementation must not create a temporary address with a zero
1393          * Preferred Lifetime.
1394          * Use age calculation as in addrconf_verify to avoid unnecessary
1395          * temporary addresses being generated.
1396          */
1397         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1398         if (cfg.preferred_lft <= regen_advance + age) {
1399                 in6_ifa_put(ifp);
1400                 in6_dev_put(idev);
1401                 ret = -1;
1402                 goto out;
1403         }
1404
1405         cfg.ifa_flags = IFA_F_TEMPORARY;
1406         /* set in addrconf_prefix_rcv() */
1407         if (ifp->flags & IFA_F_OPTIMISTIC)
1408                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1409
1410         cfg.pfx = &addr;
1411         cfg.scope = ipv6_addr_scope(cfg.pfx);
1412
1413         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1414         if (IS_ERR(ift)) {
1415                 in6_ifa_put(ifp);
1416                 in6_dev_put(idev);
1417                 pr_info("%s: retry temporary address regeneration\n", __func__);
1418                 write_lock_bh(&idev->lock);
1419                 goto retry;
1420         }
1421
1422         spin_lock_bh(&ift->lock);
1423         ift->ifpub = ifp;
1424         ift->cstamp = now;
1425         ift->tstamp = tmp_tstamp;
1426         spin_unlock_bh(&ift->lock);
1427
1428         addrconf_dad_start(ift);
1429         in6_ifa_put(ift);
1430         in6_dev_put(idev);
1431 out:
1432         return ret;
1433 }
1434
1435 /*
1436  *      Choose an appropriate source address (RFC3484)
1437  */
1438 enum {
1439         IPV6_SADDR_RULE_INIT = 0,
1440         IPV6_SADDR_RULE_LOCAL,
1441         IPV6_SADDR_RULE_SCOPE,
1442         IPV6_SADDR_RULE_PREFERRED,
1443 #ifdef CONFIG_IPV6_MIP6
1444         IPV6_SADDR_RULE_HOA,
1445 #endif
1446         IPV6_SADDR_RULE_OIF,
1447         IPV6_SADDR_RULE_LABEL,
1448         IPV6_SADDR_RULE_PRIVACY,
1449         IPV6_SADDR_RULE_ORCHID,
1450         IPV6_SADDR_RULE_PREFIX,
1451 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1452         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1453 #endif
1454         IPV6_SADDR_RULE_MAX
1455 };
1456
1457 struct ipv6_saddr_score {
1458         int                     rule;
1459         int                     addr_type;
1460         struct inet6_ifaddr     *ifa;
1461         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1462         int                     scopedist;
1463         int                     matchlen;
1464 };
1465
1466 struct ipv6_saddr_dst {
1467         const struct in6_addr *addr;
1468         int ifindex;
1469         int scope;
1470         int label;
1471         unsigned int prefs;
1472 };
1473
1474 static inline int ipv6_saddr_preferred(int type)
1475 {
1476         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1477                 return 1;
1478         return 0;
1479 }
1480
1481 static bool ipv6_use_optimistic_addr(struct net *net,
1482                                      struct inet6_dev *idev)
1483 {
1484 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1485         if (!idev)
1486                 return false;
1487         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1488                 return false;
1489         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1490                 return false;
1491
1492         return true;
1493 #else
1494         return false;
1495 #endif
1496 }
1497
1498 static bool ipv6_allow_optimistic_dad(struct net *net,
1499                                       struct inet6_dev *idev)
1500 {
1501 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1502         if (!idev)
1503                 return false;
1504         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1505                 return false;
1506
1507         return true;
1508 #else
1509         return false;
1510 #endif
1511 }
1512
1513 static int ipv6_get_saddr_eval(struct net *net,
1514                                struct ipv6_saddr_score *score,
1515                                struct ipv6_saddr_dst *dst,
1516                                int i)
1517 {
1518         int ret;
1519
1520         if (i <= score->rule) {
1521                 switch (i) {
1522                 case IPV6_SADDR_RULE_SCOPE:
1523                         ret = score->scopedist;
1524                         break;
1525                 case IPV6_SADDR_RULE_PREFIX:
1526                         ret = score->matchlen;
1527                         break;
1528                 default:
1529                         ret = !!test_bit(i, score->scorebits);
1530                 }
1531                 goto out;
1532         }
1533
1534         switch (i) {
1535         case IPV6_SADDR_RULE_INIT:
1536                 /* Rule 0: remember if hiscore is not ready yet */
1537                 ret = !!score->ifa;
1538                 break;
1539         case IPV6_SADDR_RULE_LOCAL:
1540                 /* Rule 1: Prefer same address */
1541                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1542                 break;
1543         case IPV6_SADDR_RULE_SCOPE:
1544                 /* Rule 2: Prefer appropriate scope
1545                  *
1546                  *      ret
1547                  *       ^
1548                  *    -1 |  d 15
1549                  *    ---+--+-+---> scope
1550                  *       |
1551                  *       |             d is scope of the destination.
1552                  *  B-d  |  \
1553                  *       |   \      <- smaller scope is better if
1554                  *  B-15 |    \        if scope is enough for destination.
1555                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1556                  * d-C-1 | /
1557                  *       |/         <- greater is better
1558                  *   -C  /             if scope is not enough for destination.
1559                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1560                  *
1561                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1562                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1563                  * Assume B = 0 and we get C > 29.
1564                  */
1565                 ret = __ipv6_addr_src_scope(score->addr_type);
1566                 if (ret >= dst->scope)
1567                         ret = -ret;
1568                 else
1569                         ret -= 128;     /* 30 is enough */
1570                 score->scopedist = ret;
1571                 break;
1572         case IPV6_SADDR_RULE_PREFERRED:
1573             {
1574                 /* Rule 3: Avoid deprecated and optimistic addresses */
1575                 u8 avoid = IFA_F_DEPRECATED;
1576
1577                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1578                         avoid |= IFA_F_OPTIMISTIC;
1579                 ret = ipv6_saddr_preferred(score->addr_type) ||
1580                       !(score->ifa->flags & avoid);
1581                 break;
1582             }
1583 #ifdef CONFIG_IPV6_MIP6
1584         case IPV6_SADDR_RULE_HOA:
1585             {
1586                 /* Rule 4: Prefer home address */
1587                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1588                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1589                 break;
1590             }
1591 #endif
1592         case IPV6_SADDR_RULE_OIF:
1593                 /* Rule 5: Prefer outgoing interface */
1594                 ret = (!dst->ifindex ||
1595                        dst->ifindex == score->ifa->idev->dev->ifindex);
1596                 break;
1597         case IPV6_SADDR_RULE_LABEL:
1598                 /* Rule 6: Prefer matching label */
1599                 ret = ipv6_addr_label(net,
1600                                       &score->ifa->addr, score->addr_type,
1601                                       score->ifa->idev->dev->ifindex) == dst->label;
1602                 break;
1603         case IPV6_SADDR_RULE_PRIVACY:
1604             {
1605                 /* Rule 7: Prefer public address
1606                  * Note: prefer temporary address if use_tempaddr >= 2
1607                  */
1608                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1609                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1610                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1611                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1612                 break;
1613             }
1614         case IPV6_SADDR_RULE_ORCHID:
1615                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1616                  *          non-ORCHID vs non-ORCHID
1617                  */
1618                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1619                         ipv6_addr_orchid(dst->addr));
1620                 break;
1621         case IPV6_SADDR_RULE_PREFIX:
1622                 /* Rule 8: Use longest matching prefix */
1623                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1624                 if (ret > score->ifa->prefix_len)
1625                         ret = score->ifa->prefix_len;
1626                 score->matchlen = ret;
1627                 break;
1628 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1629         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1630                 /* Optimistic addresses still have lower precedence than other
1631                  * preferred addresses.
1632                  */
1633                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1634                 break;
1635 #endif
1636         default:
1637                 ret = 0;
1638         }
1639
1640         if (ret)
1641                 __set_bit(i, score->scorebits);
1642         score->rule = i;
1643 out:
1644         return ret;
1645 }
1646
1647 static int __ipv6_dev_get_saddr(struct net *net,
1648                                 struct ipv6_saddr_dst *dst,
1649                                 struct inet6_dev *idev,
1650                                 struct ipv6_saddr_score *scores,
1651                                 int hiscore_idx)
1652 {
1653         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1654
1655         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1656                 int i;
1657
1658                 /*
1659                  * - Tentative Address (RFC2462 section 5.4)
1660                  *  - A tentative address is not considered
1661                  *    "assigned to an interface" in the traditional
1662                  *    sense, unless it is also flagged as optimistic.
1663                  * - Candidate Source Address (section 4)
1664                  *  - In any case, anycast addresses, multicast
1665                  *    addresses, and the unspecified address MUST
1666                  *    NOT be included in a candidate set.
1667                  */
1668                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1669                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1670                         continue;
1671
1672                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1673
1674                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1675                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1676                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1677                                             idev->dev->name);
1678                         continue;
1679                 }
1680
1681                 score->rule = -1;
1682                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1683
1684                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1685                         int minihiscore, miniscore;
1686
1687                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1688                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1689
1690                         if (minihiscore > miniscore) {
1691                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1692                                     score->scopedist > 0) {
1693                                         /*
1694                                          * special case:
1695                                          * each remaining entry
1696                                          * has too small (not enough)
1697                                          * scope, because ifa entries
1698                                          * are sorted by their scope
1699                                          * values.
1700                                          */
1701                                         goto out;
1702                                 }
1703                                 break;
1704                         } else if (minihiscore < miniscore) {
1705                                 swap(hiscore, score);
1706                                 hiscore_idx = 1 - hiscore_idx;
1707
1708                                 /* restore our iterator */
1709                                 score->ifa = hiscore->ifa;
1710
1711                                 break;
1712                         }
1713                 }
1714         }
1715 out:
1716         return hiscore_idx;
1717 }
1718
1719 static int ipv6_get_saddr_master(struct net *net,
1720                                  const struct net_device *dst_dev,
1721                                  const struct net_device *master,
1722                                  struct ipv6_saddr_dst *dst,
1723                                  struct ipv6_saddr_score *scores,
1724                                  int hiscore_idx)
1725 {
1726         struct inet6_dev *idev;
1727
1728         idev = __in6_dev_get(dst_dev);
1729         if (idev)
1730                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1731                                                    scores, hiscore_idx);
1732
1733         idev = __in6_dev_get(master);
1734         if (idev)
1735                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1736                                                    scores, hiscore_idx);
1737
1738         return hiscore_idx;
1739 }
1740
1741 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1742                        const struct in6_addr *daddr, unsigned int prefs,
1743                        struct in6_addr *saddr)
1744 {
1745         struct ipv6_saddr_score scores[2], *hiscore;
1746         struct ipv6_saddr_dst dst;
1747         struct inet6_dev *idev;
1748         struct net_device *dev;
1749         int dst_type;
1750         bool use_oif_addr = false;
1751         int hiscore_idx = 0;
1752         int ret = 0;
1753
1754         dst_type = __ipv6_addr_type(daddr);
1755         dst.addr = daddr;
1756         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1757         dst.scope = __ipv6_addr_src_scope(dst_type);
1758         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1759         dst.prefs = prefs;
1760
1761         scores[hiscore_idx].rule = -1;
1762         scores[hiscore_idx].ifa = NULL;
1763
1764         rcu_read_lock();
1765
1766         /* Candidate Source Address (section 4)
1767          *  - multicast and link-local destination address,
1768          *    the set of candidate source address MUST only
1769          *    include addresses assigned to interfaces
1770          *    belonging to the same link as the outgoing
1771          *    interface.
1772          * (- For site-local destination addresses, the
1773          *    set of candidate source addresses MUST only
1774          *    include addresses assigned to interfaces
1775          *    belonging to the same site as the outgoing
1776          *    interface.)
1777          *  - "It is RECOMMENDED that the candidate source addresses
1778          *    be the set of unicast addresses assigned to the
1779          *    interface that will be used to send to the destination
1780          *    (the 'outgoing' interface)." (RFC 6724)
1781          */
1782         if (dst_dev) {
1783                 idev = __in6_dev_get(dst_dev);
1784                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1785                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1786                     (idev && idev->cnf.use_oif_addrs_only)) {
1787                         use_oif_addr = true;
1788                 }
1789         }
1790
1791         if (use_oif_addr) {
1792                 if (idev)
1793                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1794         } else {
1795                 const struct net_device *master;
1796                 int master_idx = 0;
1797
1798                 /* if dst_dev exists and is enslaved to an L3 device, then
1799                  * prefer addresses from dst_dev and then the master over
1800                  * any other enslaved devices in the L3 domain.
1801                  */
1802                 master = l3mdev_master_dev_rcu(dst_dev);
1803                 if (master) {
1804                         master_idx = master->ifindex;
1805
1806                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1807                                                             master, &dst,
1808                                                             scores, hiscore_idx);
1809
1810                         if (scores[hiscore_idx].ifa)
1811                                 goto out;
1812                 }
1813
1814                 for_each_netdev_rcu(net, dev) {
1815                         /* only consider addresses on devices in the
1816                          * same L3 domain
1817                          */
1818                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1819                                 continue;
1820                         idev = __in6_dev_get(dev);
1821                         if (!idev)
1822                                 continue;
1823                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1824                 }
1825         }
1826
1827 out:
1828         hiscore = &scores[hiscore_idx];
1829         if (!hiscore->ifa)
1830                 ret = -EADDRNOTAVAIL;
1831         else
1832                 *saddr = hiscore->ifa->addr;
1833
1834         rcu_read_unlock();
1835         return ret;
1836 }
1837 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1838
1839 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1840                       u32 banned_flags)
1841 {
1842         struct inet6_ifaddr *ifp;
1843         int err = -EADDRNOTAVAIL;
1844
1845         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1846                 if (ifp->scope > IFA_LINK)
1847                         break;
1848                 if (ifp->scope == IFA_LINK &&
1849                     !(ifp->flags & banned_flags)) {
1850                         *addr = ifp->addr;
1851                         err = 0;
1852                         break;
1853                 }
1854         }
1855         return err;
1856 }
1857
1858 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1859                     u32 banned_flags)
1860 {
1861         struct inet6_dev *idev;
1862         int err = -EADDRNOTAVAIL;
1863
1864         rcu_read_lock();
1865         idev = __in6_dev_get(dev);
1866         if (idev) {
1867                 read_lock_bh(&idev->lock);
1868                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1869                 read_unlock_bh(&idev->lock);
1870         }
1871         rcu_read_unlock();
1872         return err;
1873 }
1874
1875 static int ipv6_count_addresses(const struct inet6_dev *idev)
1876 {
1877         const struct inet6_ifaddr *ifp;
1878         int cnt = 0;
1879
1880         rcu_read_lock();
1881         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1882                 cnt++;
1883         rcu_read_unlock();
1884         return cnt;
1885 }
1886
1887 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1888                   const struct net_device *dev, int strict)
1889 {
1890         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1891                                        strict, IFA_F_TENTATIVE);
1892 }
1893 EXPORT_SYMBOL(ipv6_chk_addr);
1894
1895 /* device argument is used to find the L3 domain of interest. If
1896  * skip_dev_check is set, then the ifp device is not checked against
1897  * the passed in dev argument. So the 2 cases for addresses checks are:
1898  *   1. does the address exist in the L3 domain that dev is part of
1899  *      (skip_dev_check = true), or
1900  *
1901  *   2. does the address exist on the specific device
1902  *      (skip_dev_check = false)
1903  */
1904 static struct net_device *
1905 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1906                           const struct net_device *dev, bool skip_dev_check,
1907                           int strict, u32 banned_flags)
1908 {
1909         unsigned int hash = inet6_addr_hash(net, addr);
1910         struct net_device *l3mdev, *ndev;
1911         struct inet6_ifaddr *ifp;
1912         u32 ifp_flags;
1913
1914         rcu_read_lock();
1915
1916         l3mdev = l3mdev_master_dev_rcu(dev);
1917         if (skip_dev_check)
1918                 dev = NULL;
1919
1920         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1921                 ndev = ifp->idev->dev;
1922                 if (!net_eq(dev_net(ndev), net))
1923                         continue;
1924
1925                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1926                         continue;
1927
1928                 /* Decouple optimistic from tentative for evaluation here.
1929                  * Ban optimistic addresses explicitly, when required.
1930                  */
1931                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1932                             ? (ifp->flags&~IFA_F_TENTATIVE)
1933                             : ifp->flags;
1934                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1935                     !(ifp_flags&banned_flags) &&
1936                     (!dev || ndev == dev ||
1937                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1938                         rcu_read_unlock();
1939                         return ndev;
1940                 }
1941         }
1942
1943         rcu_read_unlock();
1944         return NULL;
1945 }
1946
1947 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1948                             const struct net_device *dev, bool skip_dev_check,
1949                             int strict, u32 banned_flags)
1950 {
1951         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1952                                          strict, banned_flags) ? 1 : 0;
1953 }
1954 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1955
1956
1957 /* Compares an address/prefix_len with addresses on device @dev.
1958  * If one is found it returns true.
1959  */
1960 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1961         const unsigned int prefix_len, struct net_device *dev)
1962 {
1963         const struct inet6_ifaddr *ifa;
1964         const struct inet6_dev *idev;
1965         bool ret = false;
1966
1967         rcu_read_lock();
1968         idev = __in6_dev_get(dev);
1969         if (idev) {
1970                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1971                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1972                         if (ret)
1973                                 break;
1974                 }
1975         }
1976         rcu_read_unlock();
1977
1978         return ret;
1979 }
1980 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1981
1982 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1983 {
1984         const struct inet6_ifaddr *ifa;
1985         const struct inet6_dev *idev;
1986         int     onlink;
1987
1988         onlink = 0;
1989         rcu_read_lock();
1990         idev = __in6_dev_get(dev);
1991         if (idev) {
1992                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1993                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1994                                                    ifa->prefix_len);
1995                         if (onlink)
1996                                 break;
1997                 }
1998         }
1999         rcu_read_unlock();
2000         return onlink;
2001 }
2002 EXPORT_SYMBOL(ipv6_chk_prefix);
2003
2004 /**
2005  * ipv6_dev_find - find the first device with a given source address.
2006  * @net: the net namespace
2007  * @addr: the source address
2008  * @dev: used to find the L3 domain of interest
2009  *
2010  * The caller should be protected by RCU, or RTNL.
2011  */
2012 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2013                                  struct net_device *dev)
2014 {
2015         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2016                                          IFA_F_TENTATIVE);
2017 }
2018 EXPORT_SYMBOL(ipv6_dev_find);
2019
2020 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2021                                      struct net_device *dev, int strict)
2022 {
2023         unsigned int hash = inet6_addr_hash(net, addr);
2024         struct inet6_ifaddr *ifp, *result = NULL;
2025
2026         rcu_read_lock();
2027         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2028                 if (!net_eq(dev_net(ifp->idev->dev), net))
2029                         continue;
2030                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2031                         if (!dev || ifp->idev->dev == dev ||
2032                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2033                                 result = ifp;
2034                                 in6_ifa_hold(ifp);
2035                                 break;
2036                         }
2037                 }
2038         }
2039         rcu_read_unlock();
2040
2041         return result;
2042 }
2043
2044 /* Gets referenced address, destroys ifaddr */
2045
2046 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2047 {
2048         if (dad_failed)
2049                 ifp->flags |= IFA_F_DADFAILED;
2050
2051         if (ifp->flags&IFA_F_TEMPORARY) {
2052                 struct inet6_ifaddr *ifpub;
2053                 spin_lock_bh(&ifp->lock);
2054                 ifpub = ifp->ifpub;
2055                 if (ifpub) {
2056                         in6_ifa_hold(ifpub);
2057                         spin_unlock_bh(&ifp->lock);
2058                         ipv6_create_tempaddr(ifpub, true);
2059                         in6_ifa_put(ifpub);
2060                 } else {
2061                         spin_unlock_bh(&ifp->lock);
2062                 }
2063                 ipv6_del_addr(ifp);
2064         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2065                 spin_lock_bh(&ifp->lock);
2066                 addrconf_del_dad_work(ifp);
2067                 ifp->flags |= IFA_F_TENTATIVE;
2068                 if (dad_failed)
2069                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2070                 spin_unlock_bh(&ifp->lock);
2071                 if (dad_failed)
2072                         ipv6_ifa_notify(0, ifp);
2073                 in6_ifa_put(ifp);
2074         } else {
2075                 ipv6_del_addr(ifp);
2076         }
2077 }
2078
2079 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2080 {
2081         int err = -ENOENT;
2082
2083         spin_lock_bh(&ifp->lock);
2084         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2085                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2086                 err = 0;
2087         }
2088         spin_unlock_bh(&ifp->lock);
2089
2090         return err;
2091 }
2092
2093 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2094 {
2095         struct inet6_dev *idev = ifp->idev;
2096         struct net *net = dev_net(ifp->idev->dev);
2097
2098         if (addrconf_dad_end(ifp)) {
2099                 in6_ifa_put(ifp);
2100                 return;
2101         }
2102
2103         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2104                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2105
2106         spin_lock_bh(&ifp->lock);
2107
2108         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2109                 struct in6_addr new_addr;
2110                 struct inet6_ifaddr *ifp2;
2111                 int retries = ifp->stable_privacy_retry + 1;
2112                 struct ifa6_config cfg = {
2113                         .pfx = &new_addr,
2114                         .plen = ifp->prefix_len,
2115                         .ifa_flags = ifp->flags,
2116                         .valid_lft = ifp->valid_lft,
2117                         .preferred_lft = ifp->prefered_lft,
2118                         .scope = ifp->scope,
2119                 };
2120
2121                 if (retries > net->ipv6.sysctl.idgen_retries) {
2122                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2123                                              ifp->idev->dev->name);
2124                         goto errdad;
2125                 }
2126
2127                 new_addr = ifp->addr;
2128                 if (ipv6_generate_stable_address(&new_addr, retries,
2129                                                  idev))
2130                         goto errdad;
2131
2132                 spin_unlock_bh(&ifp->lock);
2133
2134                 if (idev->cnf.max_addresses &&
2135                     ipv6_count_addresses(idev) >=
2136                     idev->cnf.max_addresses)
2137                         goto lock_errdad;
2138
2139                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2140                                      ifp->idev->dev->name);
2141
2142                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2143                 if (IS_ERR(ifp2))
2144                         goto lock_errdad;
2145
2146                 spin_lock_bh(&ifp2->lock);
2147                 ifp2->stable_privacy_retry = retries;
2148                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2149                 spin_unlock_bh(&ifp2->lock);
2150
2151                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2152                 in6_ifa_put(ifp2);
2153 lock_errdad:
2154                 spin_lock_bh(&ifp->lock);
2155         }
2156
2157 errdad:
2158         /* transition from _POSTDAD to _ERRDAD */
2159         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2160         spin_unlock_bh(&ifp->lock);
2161
2162         addrconf_mod_dad_work(ifp, 0);
2163         in6_ifa_put(ifp);
2164 }
2165
2166 /* Join to solicited addr multicast group.
2167  * caller must hold RTNL */
2168 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2169 {
2170         struct in6_addr maddr;
2171
2172         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2173                 return;
2174
2175         addrconf_addr_solict_mult(addr, &maddr);
2176         ipv6_dev_mc_inc(dev, &maddr);
2177 }
2178
2179 /* caller must hold RTNL */
2180 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2181 {
2182         struct in6_addr maddr;
2183
2184         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2185                 return;
2186
2187         addrconf_addr_solict_mult(addr, &maddr);
2188         __ipv6_dev_mc_dec(idev, &maddr);
2189 }
2190
2191 /* caller must hold RTNL */
2192 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2193 {
2194         struct in6_addr addr;
2195
2196         if (ifp->prefix_len >= 127) /* RFC 6164 */
2197                 return;
2198         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2199         if (ipv6_addr_any(&addr))
2200                 return;
2201         __ipv6_dev_ac_inc(ifp->idev, &addr);
2202 }
2203
2204 /* caller must hold RTNL */
2205 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2206 {
2207         struct in6_addr addr;
2208
2209         if (ifp->prefix_len >= 127) /* RFC 6164 */
2210                 return;
2211         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2212         if (ipv6_addr_any(&addr))
2213                 return;
2214         __ipv6_dev_ac_dec(ifp->idev, &addr);
2215 }
2216
2217 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2218 {
2219         switch (dev->addr_len) {
2220         case ETH_ALEN:
2221                 memcpy(eui, dev->dev_addr, 3);
2222                 eui[3] = 0xFF;
2223                 eui[4] = 0xFE;
2224                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2225                 break;
2226         case EUI64_ADDR_LEN:
2227                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2228                 eui[0] ^= 2;
2229                 break;
2230         default:
2231                 return -1;
2232         }
2233
2234         return 0;
2235 }
2236
2237 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2238 {
2239         union fwnet_hwaddr *ha;
2240
2241         if (dev->addr_len != FWNET_ALEN)
2242                 return -1;
2243
2244         ha = (union fwnet_hwaddr *)dev->dev_addr;
2245
2246         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2247         eui[0] ^= 2;
2248         return 0;
2249 }
2250
2251 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2252 {
2253         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2254         if (dev->addr_len != ARCNET_ALEN)
2255                 return -1;
2256         memset(eui, 0, 7);
2257         eui[7] = *(u8 *)dev->dev_addr;
2258         return 0;
2259 }
2260
2261 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2262 {
2263         if (dev->addr_len != INFINIBAND_ALEN)
2264                 return -1;
2265         memcpy(eui, dev->dev_addr + 12, 8);
2266         eui[0] |= 2;
2267         return 0;
2268 }
2269
2270 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2271 {
2272         if (addr == 0)
2273                 return -1;
2274         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2275                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2276                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2277                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2278                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2279                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2280         eui[1] = 0;
2281         eui[2] = 0x5E;
2282         eui[3] = 0xFE;
2283         memcpy(eui + 4, &addr, 4);
2284         return 0;
2285 }
2286
2287 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2288 {
2289         if (dev->priv_flags & IFF_ISATAP)
2290                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2291         return -1;
2292 }
2293
2294 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2295 {
2296         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2297 }
2298
2299 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2300 {
2301         memcpy(eui, dev->perm_addr, 3);
2302         memcpy(eui + 5, dev->perm_addr + 3, 3);
2303         eui[3] = 0xFF;
2304         eui[4] = 0xFE;
2305         eui[0] ^= 2;
2306         return 0;
2307 }
2308
2309 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2310 {
2311         switch (dev->type) {
2312         case ARPHRD_ETHER:
2313         case ARPHRD_FDDI:
2314                 return addrconf_ifid_eui48(eui, dev);
2315         case ARPHRD_ARCNET:
2316                 return addrconf_ifid_arcnet(eui, dev);
2317         case ARPHRD_INFINIBAND:
2318                 return addrconf_ifid_infiniband(eui, dev);
2319         case ARPHRD_SIT:
2320                 return addrconf_ifid_sit(eui, dev);
2321         case ARPHRD_IPGRE:
2322         case ARPHRD_TUNNEL:
2323                 return addrconf_ifid_gre(eui, dev);
2324         case ARPHRD_6LOWPAN:
2325                 return addrconf_ifid_6lowpan(eui, dev);
2326         case ARPHRD_IEEE1394:
2327                 return addrconf_ifid_ieee1394(eui, dev);
2328         case ARPHRD_TUNNEL6:
2329         case ARPHRD_IP6GRE:
2330         case ARPHRD_RAWIP:
2331                 return addrconf_ifid_ip6tnl(eui, dev);
2332         }
2333         return -1;
2334 }
2335
2336 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2337 {
2338         int err = -1;
2339         struct inet6_ifaddr *ifp;
2340
2341         read_lock_bh(&idev->lock);
2342         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2343                 if (ifp->scope > IFA_LINK)
2344                         break;
2345                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2346                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2347                         err = 0;
2348                         break;
2349                 }
2350         }
2351         read_unlock_bh(&idev->lock);
2352         return err;
2353 }
2354
2355 /* Generation of a randomized Interface Identifier
2356  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2357  */
2358
2359 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2360 {
2361 regen:
2362         get_random_bytes(&addr->s6_addr[8], 8);
2363
2364         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2365          * check if generated address is not inappropriate:
2366          *
2367          * - Reserved IPv6 Interface Identifiers
2368          * - XXX: already assigned to an address on the device
2369          */
2370
2371         /* Subnet-router anycast: 0000:0000:0000:0000 */
2372         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2373                 goto regen;
2374
2375         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2376          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2377          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2378          */
2379         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2380             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2381                 goto regen;
2382
2383         /* Reserved subnet anycast addresses */
2384         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2385             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2386                 goto regen;
2387 }
2388
2389 /*
2390  *      Add prefix route.
2391  */
2392
2393 static void
2394 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2395                       struct net_device *dev, unsigned long expires,
2396                       u32 flags, gfp_t gfp_flags)
2397 {
2398         struct fib6_config cfg = {
2399                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2400                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2401                 .fc_ifindex = dev->ifindex,
2402                 .fc_expires = expires,
2403                 .fc_dst_len = plen,
2404                 .fc_flags = RTF_UP | flags,
2405                 .fc_nlinfo.nl_net = dev_net(dev),
2406                 .fc_protocol = RTPROT_KERNEL,
2407                 .fc_type = RTN_UNICAST,
2408         };
2409
2410         cfg.fc_dst = *pfx;
2411
2412         /* Prevent useless cloning on PtP SIT.
2413            This thing is done here expecting that the whole
2414            class of non-broadcast devices need not cloning.
2415          */
2416 #if IS_ENABLED(CONFIG_IPV6_SIT)
2417         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2418                 cfg.fc_flags |= RTF_NONEXTHOP;
2419 #endif
2420
2421         ip6_route_add(&cfg, gfp_flags, NULL);
2422 }
2423
2424
2425 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2426                                                   int plen,
2427                                                   const struct net_device *dev,
2428                                                   u32 flags, u32 noflags,
2429                                                   bool no_gw)
2430 {
2431         struct fib6_node *fn;
2432         struct fib6_info *rt = NULL;
2433         struct fib6_table *table;
2434         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2435
2436         table = fib6_get_table(dev_net(dev), tb_id);
2437         if (!table)
2438                 return NULL;
2439
2440         rcu_read_lock();
2441         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2442         if (!fn)
2443                 goto out;
2444
2445         for_each_fib6_node_rt_rcu(fn) {
2446                 /* prefix routes only use builtin fib6_nh */
2447                 if (rt->nh)
2448                         continue;
2449
2450                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2451                         continue;
2452                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2453                         continue;
2454                 if ((rt->fib6_flags & flags) != flags)
2455                         continue;
2456                 if ((rt->fib6_flags & noflags) != 0)
2457                         continue;
2458                 if (!fib6_info_hold_safe(rt))
2459                         continue;
2460                 break;
2461         }
2462 out:
2463         rcu_read_unlock();
2464         return rt;
2465 }
2466
2467
2468 /* Create "default" multicast route to the interface */
2469
2470 static void addrconf_add_mroute(struct net_device *dev)
2471 {
2472         struct fib6_config cfg = {
2473                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2474                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2475                 .fc_ifindex = dev->ifindex,
2476                 .fc_dst_len = 8,
2477                 .fc_flags = RTF_UP,
2478                 .fc_type = RTN_MULTICAST,
2479                 .fc_nlinfo.nl_net = dev_net(dev),
2480                 .fc_protocol = RTPROT_KERNEL,
2481         };
2482
2483         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2484
2485         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2486 }
2487
2488 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2489 {
2490         struct inet6_dev *idev;
2491
2492         ASSERT_RTNL();
2493
2494         idev = ipv6_find_idev(dev);
2495         if (IS_ERR(idev))
2496                 return idev;
2497
2498         if (idev->cnf.disable_ipv6)
2499                 return ERR_PTR(-EACCES);
2500
2501         /* Add default multicast route */
2502         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2503                 addrconf_add_mroute(dev);
2504
2505         return idev;
2506 }
2507
2508 static void manage_tempaddrs(struct inet6_dev *idev,
2509                              struct inet6_ifaddr *ifp,
2510                              __u32 valid_lft, __u32 prefered_lft,
2511                              bool create, unsigned long now)
2512 {
2513         u32 flags;
2514         struct inet6_ifaddr *ift;
2515
2516         read_lock_bh(&idev->lock);
2517         /* update all temporary addresses in the list */
2518         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2519                 int age, max_valid, max_prefered;
2520
2521                 if (ifp != ift->ifpub)
2522                         continue;
2523
2524                 /* RFC 4941 section 3.3:
2525                  * If a received option will extend the lifetime of a public
2526                  * address, the lifetimes of temporary addresses should
2527                  * be extended, subject to the overall constraint that no
2528                  * temporary addresses should ever remain "valid" or "preferred"
2529                  * for a time longer than (TEMP_VALID_LIFETIME) or
2530                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2531                  */
2532                 age = (now - ift->cstamp) / HZ;
2533                 max_valid = idev->cnf.temp_valid_lft - age;
2534                 if (max_valid < 0)
2535                         max_valid = 0;
2536
2537                 max_prefered = idev->cnf.temp_prefered_lft -
2538                                idev->desync_factor - age;
2539                 if (max_prefered < 0)
2540                         max_prefered = 0;
2541
2542                 if (valid_lft > max_valid)
2543                         valid_lft = max_valid;
2544
2545                 if (prefered_lft > max_prefered)
2546                         prefered_lft = max_prefered;
2547
2548                 spin_lock(&ift->lock);
2549                 flags = ift->flags;
2550                 ift->valid_lft = valid_lft;
2551                 ift->prefered_lft = prefered_lft;
2552                 ift->tstamp = now;
2553                 if (prefered_lft > 0)
2554                         ift->flags &= ~IFA_F_DEPRECATED;
2555
2556                 spin_unlock(&ift->lock);
2557                 if (!(flags&IFA_F_TENTATIVE))
2558                         ipv6_ifa_notify(0, ift);
2559         }
2560
2561         if ((create || list_empty(&idev->tempaddr_list)) &&
2562             idev->cnf.use_tempaddr > 0) {
2563                 /* When a new public address is created as described
2564                  * in [ADDRCONF], also create a new temporary address.
2565                  * Also create a temporary address if it's enabled but
2566                  * no temporary address currently exists.
2567                  */
2568                 read_unlock_bh(&idev->lock);
2569                 ipv6_create_tempaddr(ifp, false);
2570         } else {
2571                 read_unlock_bh(&idev->lock);
2572         }
2573 }
2574
2575 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2576 {
2577         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2578                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2579 }
2580
2581 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2582                                  const struct prefix_info *pinfo,
2583                                  struct inet6_dev *in6_dev,
2584                                  const struct in6_addr *addr, int addr_type,
2585                                  u32 addr_flags, bool sllao, bool tokenized,
2586                                  __u32 valid_lft, u32 prefered_lft)
2587 {
2588         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2589         int create = 0;
2590
2591         if (!ifp && valid_lft) {
2592                 int max_addresses = in6_dev->cnf.max_addresses;
2593                 struct ifa6_config cfg = {
2594                         .pfx = addr,
2595                         .plen = pinfo->prefix_len,
2596                         .ifa_flags = addr_flags,
2597                         .valid_lft = valid_lft,
2598                         .preferred_lft = prefered_lft,
2599                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2600                 };
2601
2602 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2603                 if ((net->ipv6.devconf_all->optimistic_dad ||
2604                      in6_dev->cnf.optimistic_dad) &&
2605                     !net->ipv6.devconf_all->forwarding && sllao)
2606                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2607 #endif
2608
2609                 /* Do not allow to create too much of autoconfigured
2610                  * addresses; this would be too easy way to crash kernel.
2611                  */
2612                 if (!max_addresses ||
2613                     ipv6_count_addresses(in6_dev) < max_addresses)
2614                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2615
2616                 if (IS_ERR_OR_NULL(ifp))
2617                         return -1;
2618
2619                 create = 1;
2620                 spin_lock_bh(&ifp->lock);
2621                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2622                 ifp->cstamp = jiffies;
2623                 ifp->tokenized = tokenized;
2624                 spin_unlock_bh(&ifp->lock);
2625                 addrconf_dad_start(ifp);
2626         }
2627
2628         if (ifp) {
2629                 u32 flags;
2630                 unsigned long now;
2631                 u32 stored_lft;
2632
2633                 /* Update lifetime (RFC4862 5.5.3 e)
2634                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2635                  * improve the reaction of SLAAC to renumbering events
2636                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2637                  */
2638                 spin_lock_bh(&ifp->lock);
2639                 now = jiffies;
2640                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2641                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2642                 else
2643                         stored_lft = 0;
2644
2645                 if (!create && stored_lft) {
2646                         ifp->valid_lft = valid_lft;
2647                         ifp->prefered_lft = prefered_lft;
2648                         ifp->tstamp = now;
2649                         flags = ifp->flags;
2650                         ifp->flags &= ~IFA_F_DEPRECATED;
2651                         spin_unlock_bh(&ifp->lock);
2652
2653                         if (!(flags&IFA_F_TENTATIVE))
2654                                 ipv6_ifa_notify(0, ifp);
2655                 } else
2656                         spin_unlock_bh(&ifp->lock);
2657
2658                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2659                                  create, now);
2660
2661                 in6_ifa_put(ifp);
2662                 addrconf_verify();
2663         }
2664
2665         return 0;
2666 }
2667 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2668
2669 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2670 {
2671         struct prefix_info *pinfo;
2672         __u32 valid_lft;
2673         __u32 prefered_lft;
2674         int addr_type, err;
2675         u32 addr_flags = 0;
2676         struct inet6_dev *in6_dev;
2677         struct net *net = dev_net(dev);
2678
2679         pinfo = (struct prefix_info *) opt;
2680
2681         if (len < sizeof(struct prefix_info)) {
2682                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2683                 return;
2684         }
2685
2686         /*
2687          *      Validation checks ([ADDRCONF], page 19)
2688          */
2689
2690         addr_type = ipv6_addr_type(&pinfo->prefix);
2691
2692         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2693                 return;
2694
2695         valid_lft = ntohl(pinfo->valid);
2696         prefered_lft = ntohl(pinfo->prefered);
2697
2698         if (prefered_lft > valid_lft) {
2699                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2700                 return;
2701         }
2702
2703         in6_dev = in6_dev_get(dev);
2704
2705         if (!in6_dev) {
2706                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2707                                     dev->name);
2708                 return;
2709         }
2710
2711         /*
2712          *      Two things going on here:
2713          *      1) Add routes for on-link prefixes
2714          *      2) Configure prefixes with the auto flag set
2715          */
2716
2717         if (pinfo->onlink) {
2718                 struct fib6_info *rt;
2719                 unsigned long rt_expires;
2720
2721                 /* Avoid arithmetic overflow. Really, we could
2722                  * save rt_expires in seconds, likely valid_lft,
2723                  * but it would require division in fib gc, that it
2724                  * not good.
2725                  */
2726                 if (HZ > USER_HZ)
2727                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2728                 else
2729                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2730
2731                 if (addrconf_finite_timeout(rt_expires))
2732                         rt_expires *= HZ;
2733
2734                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2735                                                pinfo->prefix_len,
2736                                                dev,
2737                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2738                                                RTF_DEFAULT, true);
2739
2740                 if (rt) {
2741                         /* Autoconf prefix route */
2742                         if (valid_lft == 0) {
2743                                 ip6_del_rt(net, rt, false);
2744                                 rt = NULL;
2745                         } else if (addrconf_finite_timeout(rt_expires)) {
2746                                 /* not infinity */
2747                                 fib6_set_expires(rt, jiffies + rt_expires);
2748                         } else {
2749                                 fib6_clean_expires(rt);
2750                         }
2751                 } else if (valid_lft) {
2752                         clock_t expires = 0;
2753                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2754                         if (addrconf_finite_timeout(rt_expires)) {
2755                                 /* not infinity */
2756                                 flags |= RTF_EXPIRES;
2757                                 expires = jiffies_to_clock_t(rt_expires);
2758                         }
2759                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2760                                               0, dev, expires, flags,
2761                                               GFP_ATOMIC);
2762                 }
2763                 fib6_info_release(rt);
2764         }
2765
2766         /* Try to figure out our local address for this prefix */
2767
2768         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2769                 struct in6_addr addr;
2770                 bool tokenized = false, dev_addr_generated = false;
2771
2772                 if (pinfo->prefix_len == 64) {
2773                         memcpy(&addr, &pinfo->prefix, 8);
2774
2775                         if (!ipv6_addr_any(&in6_dev->token)) {
2776                                 read_lock_bh(&in6_dev->lock);
2777                                 memcpy(addr.s6_addr + 8,
2778                                        in6_dev->token.s6_addr + 8, 8);
2779                                 read_unlock_bh(&in6_dev->lock);
2780                                 tokenized = true;
2781                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2782                                    !ipv6_generate_stable_address(&addr, 0,
2783                                                                  in6_dev)) {
2784                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2785                                 goto ok;
2786                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2787                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2788                                 goto put;
2789                         } else {
2790                                 dev_addr_generated = true;
2791                         }
2792                         goto ok;
2793                 }
2794                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2795                                     pinfo->prefix_len);
2796                 goto put;
2797
2798 ok:
2799                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2800                                                    &addr, addr_type,
2801                                                    addr_flags, sllao,
2802                                                    tokenized, valid_lft,
2803                                                    prefered_lft);
2804                 if (err)
2805                         goto put;
2806
2807                 /* Ignore error case here because previous prefix add addr was
2808                  * successful which will be notified.
2809                  */
2810                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2811                                               addr_type, addr_flags, sllao,
2812                                               tokenized, valid_lft,
2813                                               prefered_lft,
2814                                               dev_addr_generated);
2815         }
2816         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2817 put:
2818         in6_dev_put(in6_dev);
2819 }
2820
2821 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2822                 struct in6_ifreq *ireq)
2823 {
2824         struct ip_tunnel_parm p = { };
2825         int err;
2826
2827         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2828                 return -EADDRNOTAVAIL;
2829
2830         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2831         p.iph.version = 4;
2832         p.iph.ihl = 5;
2833         p.iph.protocol = IPPROTO_IPV6;
2834         p.iph.ttl = 64;
2835
2836         if (!dev->netdev_ops->ndo_tunnel_ctl)
2837                 return -EOPNOTSUPP;
2838         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2839         if (err)
2840                 return err;
2841
2842         dev = __dev_get_by_name(net, p.name);
2843         if (!dev)
2844                 return -ENOBUFS;
2845         return dev_open(dev, NULL);
2846 }
2847
2848 /*
2849  *      Set destination address.
2850  *      Special case for SIT interfaces where we create a new "virtual"
2851  *      device.
2852  */
2853 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2854 {
2855         struct net_device *dev;
2856         struct in6_ifreq ireq;
2857         int err = -ENODEV;
2858
2859         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2860                 return -ENODEV;
2861         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2862                 return -EFAULT;
2863
2864         rtnl_lock();
2865         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2866         if (dev && dev->type == ARPHRD_SIT)
2867                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2868         rtnl_unlock();
2869         return err;
2870 }
2871
2872 static int ipv6_mc_config(struct sock *sk, bool join,
2873                           const struct in6_addr *addr, int ifindex)
2874 {
2875         int ret;
2876
2877         ASSERT_RTNL();
2878
2879         lock_sock(sk);
2880         if (join)
2881                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2882         else
2883                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2884         release_sock(sk);
2885
2886         return ret;
2887 }
2888
2889 /*
2890  *      Manual configuration of address on an interface
2891  */
2892 static int inet6_addr_add(struct net *net, int ifindex,
2893                           struct ifa6_config *cfg,
2894                           struct netlink_ext_ack *extack)
2895 {
2896         struct inet6_ifaddr *ifp;
2897         struct inet6_dev *idev;
2898         struct net_device *dev;
2899         unsigned long timeout;
2900         clock_t expires;
2901         u32 flags;
2902
2903         ASSERT_RTNL();
2904
2905         if (cfg->plen > 128)
2906                 return -EINVAL;
2907
2908         /* check the lifetime */
2909         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2910                 return -EINVAL;
2911
2912         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2913                 return -EINVAL;
2914
2915         dev = __dev_get_by_index(net, ifindex);
2916         if (!dev)
2917                 return -ENODEV;
2918
2919         idev = addrconf_add_dev(dev);
2920         if (IS_ERR(idev))
2921                 return PTR_ERR(idev);
2922
2923         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2924                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2925                                          true, cfg->pfx, ifindex);
2926
2927                 if (ret < 0)
2928                         return ret;
2929         }
2930
2931         cfg->scope = ipv6_addr_scope(cfg->pfx);
2932
2933         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2934         if (addrconf_finite_timeout(timeout)) {
2935                 expires = jiffies_to_clock_t(timeout * HZ);
2936                 cfg->valid_lft = timeout;
2937                 flags = RTF_EXPIRES;
2938         } else {
2939                 expires = 0;
2940                 flags = 0;
2941                 cfg->ifa_flags |= IFA_F_PERMANENT;
2942         }
2943
2944         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2945         if (addrconf_finite_timeout(timeout)) {
2946                 if (timeout == 0)
2947                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2948                 cfg->preferred_lft = timeout;
2949         }
2950
2951         ifp = ipv6_add_addr(idev, cfg, true, extack);
2952         if (!IS_ERR(ifp)) {
2953                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2954                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2955                                               ifp->rt_priority, dev, expires,
2956                                               flags, GFP_KERNEL);
2957                 }
2958
2959                 /* Send a netlink notification if DAD is enabled and
2960                  * optimistic flag is not set
2961                  */
2962                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2963                         ipv6_ifa_notify(0, ifp);
2964                 /*
2965                  * Note that section 3.1 of RFC 4429 indicates
2966                  * that the Optimistic flag should not be set for
2967                  * manually configured addresses
2968                  */
2969                 addrconf_dad_start(ifp);
2970                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2971                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2972                                          cfg->preferred_lft, true, jiffies);
2973                 in6_ifa_put(ifp);
2974                 addrconf_verify_rtnl();
2975                 return 0;
2976         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2977                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2978                                cfg->pfx, ifindex);
2979         }
2980
2981         return PTR_ERR(ifp);
2982 }
2983
2984 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2985                           const struct in6_addr *pfx, unsigned int plen)
2986 {
2987         struct inet6_ifaddr *ifp;
2988         struct inet6_dev *idev;
2989         struct net_device *dev;
2990
2991         if (plen > 128)
2992                 return -EINVAL;
2993
2994         dev = __dev_get_by_index(net, ifindex);
2995         if (!dev)
2996                 return -ENODEV;
2997
2998         idev = __in6_dev_get(dev);
2999         if (!idev)
3000                 return -ENXIO;
3001
3002         read_lock_bh(&idev->lock);
3003         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3004                 if (ifp->prefix_len == plen &&
3005                     ipv6_addr_equal(pfx, &ifp->addr)) {
3006                         in6_ifa_hold(ifp);
3007                         read_unlock_bh(&idev->lock);
3008
3009                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3010                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3011                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3012                                                  jiffies);
3013                         ipv6_del_addr(ifp);
3014                         addrconf_verify_rtnl();
3015                         if (ipv6_addr_is_multicast(pfx)) {
3016                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3017                                                false, pfx, dev->ifindex);
3018                         }
3019                         return 0;
3020                 }
3021         }
3022         read_unlock_bh(&idev->lock);
3023         return -EADDRNOTAVAIL;
3024 }
3025
3026
3027 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3028 {
3029         struct ifa6_config cfg = {
3030                 .ifa_flags = IFA_F_PERMANENT,
3031                 .preferred_lft = INFINITY_LIFE_TIME,
3032                 .valid_lft = INFINITY_LIFE_TIME,
3033         };
3034         struct in6_ifreq ireq;
3035         int err;
3036
3037         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3038                 return -EPERM;
3039
3040         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3041                 return -EFAULT;
3042
3043         cfg.pfx = &ireq.ifr6_addr;
3044         cfg.plen = ireq.ifr6_prefixlen;
3045
3046         rtnl_lock();
3047         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3048         rtnl_unlock();
3049         return err;
3050 }
3051
3052 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3053 {
3054         struct in6_ifreq ireq;
3055         int err;
3056
3057         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3058                 return -EPERM;
3059
3060         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3061                 return -EFAULT;
3062
3063         rtnl_lock();
3064         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3065                              ireq.ifr6_prefixlen);
3066         rtnl_unlock();
3067         return err;
3068 }
3069
3070 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3071                      int plen, int scope)
3072 {
3073         struct inet6_ifaddr *ifp;
3074         struct ifa6_config cfg = {
3075                 .pfx = addr,
3076                 .plen = plen,
3077                 .ifa_flags = IFA_F_PERMANENT,
3078                 .valid_lft = INFINITY_LIFE_TIME,
3079                 .preferred_lft = INFINITY_LIFE_TIME,
3080                 .scope = scope
3081         };
3082
3083         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3084         if (!IS_ERR(ifp)) {
3085                 spin_lock_bh(&ifp->lock);
3086                 ifp->flags &= ~IFA_F_TENTATIVE;
3087                 spin_unlock_bh(&ifp->lock);
3088                 rt_genid_bump_ipv6(dev_net(idev->dev));
3089                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3090                 in6_ifa_put(ifp);
3091         }
3092 }
3093
3094 #if IS_ENABLED(CONFIG_IPV6_SIT)
3095 static void sit_add_v4_addrs(struct inet6_dev *idev)
3096 {
3097         struct in6_addr addr;
3098         struct net_device *dev;
3099         struct net *net = dev_net(idev->dev);
3100         int scope, plen;
3101         u32 pflags = 0;
3102
3103         ASSERT_RTNL();
3104
3105         memset(&addr, 0, sizeof(struct in6_addr));
3106         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3107
3108         if (idev->dev->flags&IFF_POINTOPOINT) {
3109                 addr.s6_addr32[0] = htonl(0xfe800000);
3110                 scope = IFA_LINK;
3111                 plen = 64;
3112         } else {
3113                 scope = IPV6_ADDR_COMPATv4;
3114                 plen = 96;
3115                 pflags |= RTF_NONEXTHOP;
3116         }
3117
3118         if (addr.s6_addr32[3]) {
3119                 add_addr(idev, &addr, plen, scope);
3120                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3121                                       GFP_KERNEL);
3122                 return;
3123         }
3124
3125         for_each_netdev(net, dev) {
3126                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3127                 if (in_dev && (dev->flags & IFF_UP)) {
3128                         struct in_ifaddr *ifa;
3129                         int flag = scope;
3130
3131                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3132                                 addr.s6_addr32[3] = ifa->ifa_local;
3133
3134                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3135                                         continue;
3136                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3137                                         if (idev->dev->flags&IFF_POINTOPOINT)
3138                                                 continue;
3139                                         flag |= IFA_HOST;
3140                                 }
3141
3142                                 add_addr(idev, &addr, plen, flag);
3143                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3144                                                       0, pflags, GFP_KERNEL);
3145                         }
3146                 }
3147         }
3148 }
3149 #endif
3150
3151 static void init_loopback(struct net_device *dev)
3152 {
3153         struct inet6_dev  *idev;
3154
3155         /* ::1 */
3156
3157         ASSERT_RTNL();
3158
3159         idev = ipv6_find_idev(dev);
3160         if (IS_ERR(idev)) {
3161                 pr_debug("%s: add_dev failed\n", __func__);
3162                 return;
3163         }
3164
3165         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3166 }
3167
3168 void addrconf_add_linklocal(struct inet6_dev *idev,
3169                             const struct in6_addr *addr, u32 flags)
3170 {
3171         struct ifa6_config cfg = {
3172                 .pfx = addr,
3173                 .plen = 64,
3174                 .ifa_flags = flags | IFA_F_PERMANENT,
3175                 .valid_lft = INFINITY_LIFE_TIME,
3176                 .preferred_lft = INFINITY_LIFE_TIME,
3177                 .scope = IFA_LINK
3178         };
3179         struct inet6_ifaddr *ifp;
3180
3181 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3182         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3183              idev->cnf.optimistic_dad) &&
3184             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3185                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3186 #endif
3187
3188         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3189         if (!IS_ERR(ifp)) {
3190                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3191                                       0, 0, GFP_ATOMIC);
3192                 addrconf_dad_start(ifp);
3193                 in6_ifa_put(ifp);
3194         }
3195 }
3196 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3197
3198 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3199 {
3200         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3201                 return true;
3202
3203         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3204             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3205                 return true;
3206
3207         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3208             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3209                 return true;
3210
3211         return false;
3212 }
3213
3214 static int ipv6_generate_stable_address(struct in6_addr *address,
3215                                         u8 dad_count,
3216                                         const struct inet6_dev *idev)
3217 {
3218         static DEFINE_SPINLOCK(lock);
3219         static __u32 digest[SHA1_DIGEST_WORDS];
3220         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3221
3222         static union {
3223                 char __data[SHA1_BLOCK_SIZE];
3224                 struct {
3225                         struct in6_addr secret;
3226                         __be32 prefix[2];
3227                         unsigned char hwaddr[MAX_ADDR_LEN];
3228                         u8 dad_count;
3229                 } __packed;
3230         } data;
3231
3232         struct in6_addr secret;
3233         struct in6_addr temp;
3234         struct net *net = dev_net(idev->dev);
3235
3236         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3237
3238         if (idev->cnf.stable_secret.initialized)
3239                 secret = idev->cnf.stable_secret.secret;
3240         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3241                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3242         else
3243                 return -1;
3244
3245 retry:
3246         spin_lock_bh(&lock);
3247
3248         sha1_init(digest);
3249         memset(&data, 0, sizeof(data));
3250         memset(workspace, 0, sizeof(workspace));
3251         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3252         data.prefix[0] = address->s6_addr32[0];
3253         data.prefix[1] = address->s6_addr32[1];
3254         data.secret = secret;
3255         data.dad_count = dad_count;
3256
3257         sha1_transform(digest, data.__data, workspace);
3258
3259         temp = *address;
3260         temp.s6_addr32[2] = (__force __be32)digest[0];
3261         temp.s6_addr32[3] = (__force __be32)digest[1];
3262
3263         spin_unlock_bh(&lock);
3264
3265         if (ipv6_reserved_interfaceid(temp)) {
3266                 dad_count++;
3267                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3268                         return -1;
3269                 goto retry;
3270         }
3271
3272         *address = temp;
3273         return 0;
3274 }
3275
3276 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3277 {
3278         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3279
3280         if (s->initialized)
3281                 return;
3282         s = &idev->cnf.stable_secret;
3283         get_random_bytes(&s->secret, sizeof(s->secret));
3284         s->initialized = true;
3285 }
3286
3287 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3288 {
3289         struct in6_addr addr;
3290
3291         /* no link local addresses on L3 master devices */
3292         if (netif_is_l3_master(idev->dev))
3293                 return;
3294
3295         /* no link local addresses on devices flagged as slaves */
3296         if (idev->dev->flags & IFF_SLAVE)
3297                 return;
3298
3299         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3300
3301         switch (idev->cnf.addr_gen_mode) {
3302         case IN6_ADDR_GEN_MODE_RANDOM:
3303                 ipv6_gen_mode_random_init(idev);
3304                 fallthrough;
3305         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3306                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3307                         addrconf_add_linklocal(idev, &addr,
3308                                                IFA_F_STABLE_PRIVACY);
3309                 else if (prefix_route)
3310                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3311                                               0, 0, GFP_KERNEL);
3312                 break;
3313         case IN6_ADDR_GEN_MODE_EUI64:
3314                 /* addrconf_add_linklocal also adds a prefix_route and we
3315                  * only need to care about prefix routes if ipv6_generate_eui64
3316                  * couldn't generate one.
3317                  */
3318                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3319                         addrconf_add_linklocal(idev, &addr, 0);
3320                 else if (prefix_route)
3321                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3322                                               0, 0, GFP_KERNEL);
3323                 break;
3324         case IN6_ADDR_GEN_MODE_NONE:
3325         default:
3326                 /* will not add any link local address */
3327                 break;
3328         }
3329 }
3330
3331 static void addrconf_dev_config(struct net_device *dev)
3332 {
3333         struct inet6_dev *idev;
3334
3335         ASSERT_RTNL();
3336
3337         if ((dev->type != ARPHRD_ETHER) &&
3338             (dev->type != ARPHRD_FDDI) &&
3339             (dev->type != ARPHRD_ARCNET) &&
3340             (dev->type != ARPHRD_INFINIBAND) &&
3341             (dev->type != ARPHRD_IEEE1394) &&
3342             (dev->type != ARPHRD_TUNNEL6) &&
3343             (dev->type != ARPHRD_6LOWPAN) &&
3344             (dev->type != ARPHRD_IP6GRE) &&
3345             (dev->type != ARPHRD_IPGRE) &&
3346             (dev->type != ARPHRD_TUNNEL) &&
3347             (dev->type != ARPHRD_NONE) &&
3348             (dev->type != ARPHRD_RAWIP)) {
3349                 /* Alas, we support only Ethernet autoconfiguration. */
3350                 idev = __in6_dev_get(dev);
3351                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3352                     dev->flags & IFF_MULTICAST)
3353                         ipv6_mc_up(idev);
3354                 return;
3355         }
3356
3357         idev = addrconf_add_dev(dev);
3358         if (IS_ERR(idev))
3359                 return;
3360
3361         /* this device type has no EUI support */
3362         if (dev->type == ARPHRD_NONE &&
3363             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3364                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3365
3366         addrconf_addr_gen(idev, false);
3367 }
3368
3369 #if IS_ENABLED(CONFIG_IPV6_SIT)
3370 static void addrconf_sit_config(struct net_device *dev)
3371 {
3372         struct inet6_dev *idev;
3373
3374         ASSERT_RTNL();
3375
3376         /*
3377          * Configure the tunnel with one of our IPv4
3378          * addresses... we should configure all of
3379          * our v4 addrs in the tunnel
3380          */
3381
3382         idev = ipv6_find_idev(dev);
3383         if (IS_ERR(idev)) {
3384                 pr_debug("%s: add_dev failed\n", __func__);
3385                 return;
3386         }
3387
3388         if (dev->priv_flags & IFF_ISATAP) {
3389                 addrconf_addr_gen(idev, false);
3390                 return;
3391         }
3392
3393         sit_add_v4_addrs(idev);
3394
3395         if (dev->flags&IFF_POINTOPOINT)
3396                 addrconf_add_mroute(dev);
3397 }
3398 #endif
3399
3400 #if IS_ENABLED(CONFIG_NET_IPGRE)
3401 static void addrconf_gre_config(struct net_device *dev)
3402 {
3403         struct inet6_dev *idev;
3404
3405         ASSERT_RTNL();
3406
3407         idev = ipv6_find_idev(dev);
3408         if (IS_ERR(idev)) {
3409                 pr_debug("%s: add_dev failed\n", __func__);
3410                 return;
3411         }
3412
3413         addrconf_addr_gen(idev, true);
3414         if (dev->flags & IFF_POINTOPOINT)
3415                 addrconf_add_mroute(dev);
3416 }
3417 #endif
3418
3419 static int fixup_permanent_addr(struct net *net,
3420                                 struct inet6_dev *idev,
3421                                 struct inet6_ifaddr *ifp)
3422 {
3423         /* !fib6_node means the host route was removed from the
3424          * FIB, for example, if 'lo' device is taken down. In that
3425          * case regenerate the host route.
3426          */
3427         if (!ifp->rt || !ifp->rt->fib6_node) {
3428                 struct fib6_info *f6i, *prev;
3429
3430                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3431                                          GFP_ATOMIC);
3432                 if (IS_ERR(f6i))
3433                         return PTR_ERR(f6i);
3434
3435                 /* ifp->rt can be accessed outside of rtnl */
3436                 spin_lock(&ifp->lock);
3437                 prev = ifp->rt;
3438                 ifp->rt = f6i;
3439                 spin_unlock(&ifp->lock);
3440
3441                 fib6_info_release(prev);
3442         }
3443
3444         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3445                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3446                                       ifp->rt_priority, idev->dev, 0, 0,
3447                                       GFP_ATOMIC);
3448         }
3449
3450         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3451                 addrconf_dad_start(ifp);
3452
3453         return 0;
3454 }
3455
3456 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3457 {
3458         struct inet6_ifaddr *ifp, *tmp;
3459         struct inet6_dev *idev;
3460
3461         idev = __in6_dev_get(dev);
3462         if (!idev)
3463                 return;
3464
3465         write_lock_bh(&idev->lock);
3466
3467         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3468                 if ((ifp->flags & IFA_F_PERMANENT) &&
3469                     fixup_permanent_addr(net, idev, ifp) < 0) {
3470                         write_unlock_bh(&idev->lock);
3471                         in6_ifa_hold(ifp);
3472                         ipv6_del_addr(ifp);
3473                         write_lock_bh(&idev->lock);
3474
3475                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3476                                              idev->dev->name, &ifp->addr);
3477                 }
3478         }
3479
3480         write_unlock_bh(&idev->lock);
3481 }
3482
3483 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3484                            void *ptr)
3485 {
3486         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3487         struct netdev_notifier_change_info *change_info;
3488         struct netdev_notifier_changeupper_info *info;
3489         struct inet6_dev *idev = __in6_dev_get(dev);
3490         struct net *net = dev_net(dev);
3491         int run_pending = 0;
3492         int err;
3493
3494         switch (event) {
3495         case NETDEV_REGISTER:
3496                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3497                         idev = ipv6_add_dev(dev);
3498                         if (IS_ERR(idev))
3499                                 return notifier_from_errno(PTR_ERR(idev));
3500                 }
3501                 break;
3502
3503         case NETDEV_CHANGEMTU:
3504                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3505                 if (dev->mtu < IPV6_MIN_MTU) {
3506                         addrconf_ifdown(dev, dev != net->loopback_dev);
3507                         break;
3508                 }
3509
3510                 if (idev) {
3511                         rt6_mtu_change(dev, dev->mtu);
3512                         idev->cnf.mtu6 = dev->mtu;
3513                         break;
3514                 }
3515
3516                 /* allocate new idev */
3517                 idev = ipv6_add_dev(dev);
3518                 if (IS_ERR(idev))
3519                         break;
3520
3521                 /* device is still not ready */
3522                 if (!(idev->if_flags & IF_READY))
3523                         break;
3524
3525                 run_pending = 1;
3526                 fallthrough;
3527         case NETDEV_UP:
3528         case NETDEV_CHANGE:
3529                 if (dev->flags & IFF_SLAVE)
3530                         break;
3531
3532                 if (idev && idev->cnf.disable_ipv6)
3533                         break;
3534
3535                 if (event == NETDEV_UP) {
3536                         /* restore routes for permanent addresses */
3537                         addrconf_permanent_addr(net, dev);
3538
3539                         if (!addrconf_link_ready(dev)) {
3540                                 /* device is not ready yet. */
3541                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3542                                          dev->name);
3543                                 break;
3544                         }
3545
3546                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3547                                 idev = ipv6_add_dev(dev);
3548
3549                         if (!IS_ERR_OR_NULL(idev)) {
3550                                 idev->if_flags |= IF_READY;
3551                                 run_pending = 1;
3552                         }
3553                 } else if (event == NETDEV_CHANGE) {
3554                         if (!addrconf_link_ready(dev)) {
3555                                 /* device is still not ready. */
3556                                 rt6_sync_down_dev(dev, event);
3557                                 break;
3558                         }
3559
3560                         if (!IS_ERR_OR_NULL(idev)) {
3561                                 if (idev->if_flags & IF_READY) {
3562                                         /* device is already configured -
3563                                          * but resend MLD reports, we might
3564                                          * have roamed and need to update
3565                                          * multicast snooping switches
3566                                          */
3567                                         ipv6_mc_up(idev);
3568                                         change_info = ptr;
3569                                         if (change_info->flags_changed & IFF_NOARP)
3570                                                 addrconf_dad_run(idev, true);
3571                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3572                                         break;
3573                                 }
3574                                 idev->if_flags |= IF_READY;
3575                         }
3576
3577                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3578                                 dev->name);
3579
3580                         run_pending = 1;
3581                 }
3582
3583                 switch (dev->type) {
3584 #if IS_ENABLED(CONFIG_IPV6_SIT)
3585                 case ARPHRD_SIT:
3586                         addrconf_sit_config(dev);
3587                         break;
3588 #endif
3589 #if IS_ENABLED(CONFIG_NET_IPGRE)
3590                 case ARPHRD_IPGRE:
3591                         addrconf_gre_config(dev);
3592                         break;
3593 #endif
3594                 case ARPHRD_LOOPBACK:
3595                         init_loopback(dev);
3596                         break;
3597
3598                 default:
3599                         addrconf_dev_config(dev);
3600                         break;
3601                 }
3602
3603                 if (!IS_ERR_OR_NULL(idev)) {
3604                         if (run_pending)
3605                                 addrconf_dad_run(idev, false);
3606
3607                         /* Device has an address by now */
3608                         rt6_sync_up(dev, RTNH_F_DEAD);
3609
3610                         /*
3611                          * If the MTU changed during the interface down,
3612                          * when the interface up, the changed MTU must be
3613                          * reflected in the idev as well as routers.
3614                          */
3615                         if (idev->cnf.mtu6 != dev->mtu &&
3616                             dev->mtu >= IPV6_MIN_MTU) {
3617                                 rt6_mtu_change(dev, dev->mtu);
3618                                 idev->cnf.mtu6 = dev->mtu;
3619                         }
3620                         idev->tstamp = jiffies;
3621                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3622
3623                         /*
3624                          * If the changed mtu during down is lower than
3625                          * IPV6_MIN_MTU stop IPv6 on this interface.
3626                          */
3627                         if (dev->mtu < IPV6_MIN_MTU)
3628                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3629                 }
3630                 break;
3631
3632         case NETDEV_DOWN:
3633         case NETDEV_UNREGISTER:
3634                 /*
3635                  *      Remove all addresses from this interface.
3636                  */
3637                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3638                 break;
3639
3640         case NETDEV_CHANGENAME:
3641                 if (idev) {
3642                         snmp6_unregister_dev(idev);
3643                         addrconf_sysctl_unregister(idev);
3644                         err = addrconf_sysctl_register(idev);
3645                         if (err)
3646                                 return notifier_from_errno(err);
3647                         err = snmp6_register_dev(idev);
3648                         if (err) {
3649                                 addrconf_sysctl_unregister(idev);
3650                                 return notifier_from_errno(err);
3651                         }
3652                 }
3653                 break;
3654
3655         case NETDEV_PRE_TYPE_CHANGE:
3656         case NETDEV_POST_TYPE_CHANGE:
3657                 if (idev)
3658                         addrconf_type_change(dev, event);
3659                 break;
3660
3661         case NETDEV_CHANGEUPPER:
3662                 info = ptr;
3663
3664                 /* flush all routes if dev is linked to or unlinked from
3665                  * an L3 master device (e.g., VRF)
3666                  */
3667                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3668                         addrconf_ifdown(dev, false);
3669         }
3670
3671         return NOTIFY_OK;
3672 }
3673
3674 /*
3675  *      addrconf module should be notified of a device going up
3676  */
3677 static struct notifier_block ipv6_dev_notf = {
3678         .notifier_call = addrconf_notify,
3679         .priority = ADDRCONF_NOTIFY_PRIORITY,
3680 };
3681
3682 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3683 {
3684         struct inet6_dev *idev;
3685         ASSERT_RTNL();
3686
3687         idev = __in6_dev_get(dev);
3688
3689         if (event == NETDEV_POST_TYPE_CHANGE)
3690                 ipv6_mc_remap(idev);
3691         else if (event == NETDEV_PRE_TYPE_CHANGE)
3692                 ipv6_mc_unmap(idev);
3693 }
3694
3695 static bool addr_is_local(const struct in6_addr *addr)
3696 {
3697         return ipv6_addr_type(addr) &
3698                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3699 }
3700
3701 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3702 {
3703         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3704         struct net *net = dev_net(dev);
3705         struct inet6_dev *idev;
3706         struct inet6_ifaddr *ifa, *tmp;
3707         bool keep_addr = false;
3708         int state, i;
3709
3710         ASSERT_RTNL();
3711
3712         rt6_disable_ip(dev, event);
3713
3714         idev = __in6_dev_get(dev);
3715         if (!idev)
3716                 return -ENODEV;
3717
3718         /*
3719          * Step 1: remove reference to ipv6 device from parent device.
3720          *         Do not dev_put!
3721          */
3722         if (unregister) {
3723                 idev->dead = 1;
3724
3725                 /* protected by rtnl_lock */
3726                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3727
3728                 /* Step 1.5: remove snmp6 entry */
3729                 snmp6_unregister_dev(idev);
3730
3731         }
3732
3733         /* combine the user config with event to determine if permanent
3734          * addresses are to be removed from address hash table
3735          */
3736         if (!unregister && !idev->cnf.disable_ipv6) {
3737                 /* aggregate the system setting and interface setting */
3738                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3739
3740                 if (!_keep_addr)
3741                         _keep_addr = idev->cnf.keep_addr_on_down;
3742
3743                 keep_addr = (_keep_addr > 0);
3744         }
3745
3746         /* Step 2: clear hash table */
3747         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3748                 struct hlist_head *h = &inet6_addr_lst[i];
3749
3750                 spin_lock_bh(&addrconf_hash_lock);
3751 restart:
3752                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3753                         if (ifa->idev == idev) {
3754                                 addrconf_del_dad_work(ifa);
3755                                 /* combined flag + permanent flag decide if
3756                                  * address is retained on a down event
3757                                  */
3758                                 if (!keep_addr ||
3759                                     !(ifa->flags & IFA_F_PERMANENT) ||
3760                                     addr_is_local(&ifa->addr)) {
3761                                         hlist_del_init_rcu(&ifa->addr_lst);
3762                                         goto restart;
3763                                 }
3764                         }
3765                 }
3766                 spin_unlock_bh(&addrconf_hash_lock);
3767         }
3768
3769         write_lock_bh(&idev->lock);
3770
3771         addrconf_del_rs_timer(idev);
3772
3773         /* Step 2: clear flags for stateless addrconf */
3774         if (!unregister)
3775                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3776
3777         /* Step 3: clear tempaddr list */
3778         while (!list_empty(&idev->tempaddr_list)) {
3779                 ifa = list_first_entry(&idev->tempaddr_list,
3780                                        struct inet6_ifaddr, tmp_list);
3781                 list_del(&ifa->tmp_list);
3782                 write_unlock_bh(&idev->lock);
3783                 spin_lock_bh(&ifa->lock);
3784
3785                 if (ifa->ifpub) {
3786                         in6_ifa_put(ifa->ifpub);
3787                         ifa->ifpub = NULL;
3788                 }
3789                 spin_unlock_bh(&ifa->lock);
3790                 in6_ifa_put(ifa);
3791                 write_lock_bh(&idev->lock);
3792         }
3793
3794         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3795                 struct fib6_info *rt = NULL;
3796                 bool keep;
3797
3798                 addrconf_del_dad_work(ifa);
3799
3800                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3801                         !addr_is_local(&ifa->addr);
3802
3803                 write_unlock_bh(&idev->lock);
3804                 spin_lock_bh(&ifa->lock);
3805
3806                 if (keep) {
3807                         /* set state to skip the notifier below */
3808                         state = INET6_IFADDR_STATE_DEAD;
3809                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3810                         if (!(ifa->flags & IFA_F_NODAD))
3811                                 ifa->flags |= IFA_F_TENTATIVE;
3812
3813                         rt = ifa->rt;
3814                         ifa->rt = NULL;
3815                 } else {
3816                         state = ifa->state;
3817                         ifa->state = INET6_IFADDR_STATE_DEAD;
3818                 }
3819
3820                 spin_unlock_bh(&ifa->lock);
3821
3822                 if (rt)
3823                         ip6_del_rt(net, rt, false);
3824
3825                 if (state != INET6_IFADDR_STATE_DEAD) {
3826                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3827                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3828                 } else {
3829                         if (idev->cnf.forwarding)
3830                                 addrconf_leave_anycast(ifa);
3831                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3832                 }
3833
3834                 write_lock_bh(&idev->lock);
3835                 if (!keep) {
3836                         list_del_rcu(&ifa->if_list);
3837                         in6_ifa_put(ifa);
3838                 }
3839         }
3840
3841         write_unlock_bh(&idev->lock);
3842
3843         /* Step 5: Discard anycast and multicast list */
3844         if (unregister) {
3845                 ipv6_ac_destroy_dev(idev);
3846                 ipv6_mc_destroy_dev(idev);
3847         } else {
3848                 ipv6_mc_down(idev);
3849         }
3850
3851         idev->tstamp = jiffies;
3852
3853         /* Last: Shot the device (if unregistered) */
3854         if (unregister) {
3855                 addrconf_sysctl_unregister(idev);
3856                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3857                 neigh_ifdown(&nd_tbl, dev);
3858                 in6_dev_put(idev);
3859         }
3860         return 0;
3861 }
3862
3863 static void addrconf_rs_timer(struct timer_list *t)
3864 {
3865         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3866         struct net_device *dev = idev->dev;
3867         struct in6_addr lladdr;
3868
3869         write_lock(&idev->lock);
3870         if (idev->dead || !(idev->if_flags & IF_READY))
3871                 goto out;
3872
3873         if (!ipv6_accept_ra(idev))
3874                 goto out;
3875
3876         /* Announcement received after solicitation was sent */
3877         if (idev->if_flags & IF_RA_RCVD)
3878                 goto out;
3879
3880         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3881                 write_unlock(&idev->lock);
3882                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3883                         ndisc_send_rs(dev, &lladdr,
3884                                       &in6addr_linklocal_allrouters);
3885                 else
3886                         goto put;
3887
3888                 write_lock(&idev->lock);
3889                 idev->rs_interval = rfc3315_s14_backoff_update(
3890                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3891                 /* The wait after the last probe can be shorter */
3892                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3893                                              idev->cnf.rtr_solicits) ?
3894                                       idev->cnf.rtr_solicit_delay :
3895                                       idev->rs_interval);
3896         } else {
3897                 /*
3898                  * Note: we do not support deprecated "all on-link"
3899                  * assumption any longer.
3900                  */
3901                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3902         }
3903
3904 out:
3905         write_unlock(&idev->lock);
3906 put:
3907         in6_dev_put(idev);
3908 }
3909
3910 /*
3911  *      Duplicate Address Detection
3912  */
3913 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3914 {
3915         unsigned long rand_num;
3916         struct inet6_dev *idev = ifp->idev;
3917         u64 nonce;
3918
3919         if (ifp->flags & IFA_F_OPTIMISTIC)
3920                 rand_num = 0;
3921         else
3922                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3923
3924         nonce = 0;
3925         if (idev->cnf.enhanced_dad ||
3926             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3927                 do
3928                         get_random_bytes(&nonce, 6);
3929                 while (nonce == 0);
3930         }
3931         ifp->dad_nonce = nonce;
3932         ifp->dad_probes = idev->cnf.dad_transmits;
3933         addrconf_mod_dad_work(ifp, rand_num);
3934 }
3935
3936 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3937 {
3938         struct inet6_dev *idev = ifp->idev;
3939         struct net_device *dev = idev->dev;
3940         bool bump_id, notify = false;
3941         struct net *net;
3942
3943         addrconf_join_solict(dev, &ifp->addr);
3944
3945         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3946
3947         read_lock_bh(&idev->lock);
3948         spin_lock(&ifp->lock);
3949         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3950                 goto out;
3951
3952         net = dev_net(dev);
3953         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3954             (net->ipv6.devconf_all->accept_dad < 1 &&
3955              idev->cnf.accept_dad < 1) ||
3956             !(ifp->flags&IFA_F_TENTATIVE) ||
3957             ifp->flags & IFA_F_NODAD) {
3958                 bool send_na = false;
3959
3960                 if (ifp->flags & IFA_F_TENTATIVE &&
3961                     !(ifp->flags & IFA_F_OPTIMISTIC))
3962                         send_na = true;
3963                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3964                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3965                 spin_unlock(&ifp->lock);
3966                 read_unlock_bh(&idev->lock);
3967
3968                 addrconf_dad_completed(ifp, bump_id, send_na);
3969                 return;
3970         }
3971
3972         if (!(idev->if_flags & IF_READY)) {
3973                 spin_unlock(&ifp->lock);
3974                 read_unlock_bh(&idev->lock);
3975                 /*
3976                  * If the device is not ready:
3977                  * - keep it tentative if it is a permanent address.
3978                  * - otherwise, kill it.
3979                  */
3980                 in6_ifa_hold(ifp);
3981                 addrconf_dad_stop(ifp, 0);
3982                 return;
3983         }
3984
3985         /*
3986          * Optimistic nodes can start receiving
3987          * Frames right away
3988          */
3989         if (ifp->flags & IFA_F_OPTIMISTIC) {
3990                 ip6_ins_rt(net, ifp->rt);
3991                 if (ipv6_use_optimistic_addr(net, idev)) {
3992                         /* Because optimistic nodes can use this address,
3993                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3994                          */
3995                         notify = true;
3996                 }
3997         }
3998
3999         addrconf_dad_kick(ifp);
4000 out:
4001         spin_unlock(&ifp->lock);
4002         read_unlock_bh(&idev->lock);
4003         if (notify)
4004                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4005 }
4006
4007 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4008 {
4009         bool begin_dad = false;
4010
4011         spin_lock_bh(&ifp->lock);
4012         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4013                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4014                 begin_dad = true;
4015         }
4016         spin_unlock_bh(&ifp->lock);
4017
4018         if (begin_dad)
4019                 addrconf_mod_dad_work(ifp, 0);
4020 }
4021
4022 static void addrconf_dad_work(struct work_struct *w)
4023 {
4024         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4025                                                 struct inet6_ifaddr,
4026                                                 dad_work);
4027         struct inet6_dev *idev = ifp->idev;
4028         bool bump_id, disable_ipv6 = false;
4029         struct in6_addr mcaddr;
4030
4031         enum {
4032                 DAD_PROCESS,
4033                 DAD_BEGIN,
4034                 DAD_ABORT,
4035         } action = DAD_PROCESS;
4036
4037         rtnl_lock();
4038
4039         spin_lock_bh(&ifp->lock);
4040         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4041                 action = DAD_BEGIN;
4042                 ifp->state = INET6_IFADDR_STATE_DAD;
4043         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4044                 action = DAD_ABORT;
4045                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4046
4047                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4048                      idev->cnf.accept_dad > 1) &&
4049                     !idev->cnf.disable_ipv6 &&
4050                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4051                         struct in6_addr addr;
4052
4053                         addr.s6_addr32[0] = htonl(0xfe800000);
4054                         addr.s6_addr32[1] = 0;
4055
4056                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4057                             ipv6_addr_equal(&ifp->addr, &addr)) {
4058                                 /* DAD failed for link-local based on MAC */
4059                                 idev->cnf.disable_ipv6 = 1;
4060
4061                                 pr_info("%s: IPv6 being disabled!\n",
4062                                         ifp->idev->dev->name);
4063                                 disable_ipv6 = true;
4064                         }
4065                 }
4066         }
4067         spin_unlock_bh(&ifp->lock);
4068
4069         if (action == DAD_BEGIN) {
4070                 addrconf_dad_begin(ifp);
4071                 goto out;
4072         } else if (action == DAD_ABORT) {
4073                 in6_ifa_hold(ifp);
4074                 addrconf_dad_stop(ifp, 1);
4075                 if (disable_ipv6)
4076                         addrconf_ifdown(idev->dev, false);
4077                 goto out;
4078         }
4079
4080         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4081                 goto out;
4082
4083         write_lock_bh(&idev->lock);
4084         if (idev->dead || !(idev->if_flags & IF_READY)) {
4085                 write_unlock_bh(&idev->lock);
4086                 goto out;
4087         }
4088
4089         spin_lock(&ifp->lock);
4090         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4091                 spin_unlock(&ifp->lock);
4092                 write_unlock_bh(&idev->lock);
4093                 goto out;
4094         }
4095
4096         if (ifp->dad_probes == 0) {
4097                 bool send_na = false;
4098
4099                 /*
4100                  * DAD was successful
4101                  */
4102
4103                 if (ifp->flags & IFA_F_TENTATIVE &&
4104                     !(ifp->flags & IFA_F_OPTIMISTIC))
4105                         send_na = true;
4106                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4107                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4108                 spin_unlock(&ifp->lock);
4109                 write_unlock_bh(&idev->lock);
4110
4111                 addrconf_dad_completed(ifp, bump_id, send_na);
4112
4113                 goto out;
4114         }
4115
4116         ifp->dad_probes--;
4117         addrconf_mod_dad_work(ifp,
4118                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4119                                   HZ/100));
4120         spin_unlock(&ifp->lock);
4121         write_unlock_bh(&idev->lock);
4122
4123         /* send a neighbour solicitation for our addr */
4124         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4125         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4126                       ifp->dad_nonce);
4127 out:
4128         in6_ifa_put(ifp);
4129         rtnl_unlock();
4130 }
4131
4132 /* ifp->idev must be at least read locked */
4133 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4134 {
4135         struct inet6_ifaddr *ifpiter;
4136         struct inet6_dev *idev = ifp->idev;
4137
4138         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4139                 if (ifpiter->scope > IFA_LINK)
4140                         break;
4141                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4142                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4143                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4144                     IFA_F_PERMANENT)
4145                         return false;
4146         }
4147         return true;
4148 }
4149
4150 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4151                                    bool send_na)
4152 {
4153         struct net_device *dev = ifp->idev->dev;
4154         struct in6_addr lladdr;
4155         bool send_rs, send_mld;
4156
4157         addrconf_del_dad_work(ifp);
4158
4159         /*
4160          *      Configure the address for reception. Now it is valid.
4161          */
4162
4163         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4164
4165         /* If added prefix is link local and we are prepared to process
4166            router advertisements, start sending router solicitations.
4167          */
4168
4169         read_lock_bh(&ifp->idev->lock);
4170         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4171         send_rs = send_mld &&
4172                   ipv6_accept_ra(ifp->idev) &&
4173                   ifp->idev->cnf.rtr_solicits != 0 &&
4174                   (dev->flags&IFF_LOOPBACK) == 0;
4175         read_unlock_bh(&ifp->idev->lock);
4176
4177         /* While dad is in progress mld report's source address is in6_addrany.
4178          * Resend with proper ll now.
4179          */
4180         if (send_mld)
4181                 ipv6_mc_dad_complete(ifp->idev);
4182
4183         /* send unsolicited NA if enabled */
4184         if (send_na &&
4185             (ifp->idev->cnf.ndisc_notify ||
4186              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4187                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4188                               /*router=*/ !!ifp->idev->cnf.forwarding,
4189                               /*solicited=*/ false, /*override=*/ true,
4190                               /*inc_opt=*/ true);
4191         }
4192
4193         if (send_rs) {
4194                 /*
4195                  *      If a host as already performed a random delay
4196                  *      [...] as part of DAD [...] there is no need
4197                  *      to delay again before sending the first RS
4198                  */
4199                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4200                         return;
4201                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4202
4203                 write_lock_bh(&ifp->idev->lock);
4204                 spin_lock(&ifp->lock);
4205                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4206                         ifp->idev->cnf.rtr_solicit_interval);
4207                 ifp->idev->rs_probes = 1;
4208                 ifp->idev->if_flags |= IF_RS_SENT;
4209                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4210                 spin_unlock(&ifp->lock);
4211                 write_unlock_bh(&ifp->idev->lock);
4212         }
4213
4214         if (bump_id)
4215                 rt_genid_bump_ipv6(dev_net(dev));
4216
4217         /* Make sure that a new temporary address will be created
4218          * before this temporary address becomes deprecated.
4219          */
4220         if (ifp->flags & IFA_F_TEMPORARY)
4221                 addrconf_verify_rtnl();
4222 }
4223
4224 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4225 {
4226         struct inet6_ifaddr *ifp;
4227
4228         read_lock_bh(&idev->lock);
4229         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4230                 spin_lock(&ifp->lock);
4231                 if ((ifp->flags & IFA_F_TENTATIVE &&
4232                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4233                         if (restart)
4234                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4235                         addrconf_dad_kick(ifp);
4236                 }
4237                 spin_unlock(&ifp->lock);
4238         }
4239         read_unlock_bh(&idev->lock);
4240 }
4241
4242 #ifdef CONFIG_PROC_FS
4243 struct if6_iter_state {
4244         struct seq_net_private p;
4245         int bucket;
4246         int offset;
4247 };
4248
4249 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4250 {
4251         struct if6_iter_state *state = seq->private;
4252         struct net *net = seq_file_net(seq);
4253         struct inet6_ifaddr *ifa = NULL;
4254         int p = 0;
4255
4256         /* initial bucket if pos is 0 */
4257         if (pos == 0) {
4258                 state->bucket = 0;
4259                 state->offset = 0;
4260         }
4261
4262         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4263                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4264                                          addr_lst) {
4265                         if (!net_eq(dev_net(ifa->idev->dev), net))
4266                                 continue;
4267                         /* sync with offset */
4268                         if (p < state->offset) {
4269                                 p++;
4270                                 continue;
4271                         }
4272                         return ifa;
4273                 }
4274
4275                 /* prepare for next bucket */
4276                 state->offset = 0;
4277                 p = 0;
4278         }
4279         return NULL;
4280 }
4281
4282 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4283                                          struct inet6_ifaddr *ifa)
4284 {
4285         struct if6_iter_state *state = seq->private;
4286         struct net *net = seq_file_net(seq);
4287
4288         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4289                 if (!net_eq(dev_net(ifa->idev->dev), net))
4290                         continue;
4291                 state->offset++;
4292                 return ifa;
4293         }
4294
4295         state->offset = 0;
4296         while (++state->bucket < IN6_ADDR_HSIZE) {
4297                 hlist_for_each_entry_rcu(ifa,
4298                                      &inet6_addr_lst[state->bucket], addr_lst) {
4299                         if (!net_eq(dev_net(ifa->idev->dev), net))
4300                                 continue;
4301                         return ifa;
4302                 }
4303         }
4304
4305         return NULL;
4306 }
4307
4308 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4309         __acquires(rcu)
4310 {
4311         rcu_read_lock();
4312         return if6_get_first(seq, *pos);
4313 }
4314
4315 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4316 {
4317         struct inet6_ifaddr *ifa;
4318
4319         ifa = if6_get_next(seq, v);
4320         ++*pos;
4321         return ifa;
4322 }
4323
4324 static void if6_seq_stop(struct seq_file *seq, void *v)
4325         __releases(rcu)
4326 {
4327         rcu_read_unlock();
4328 }
4329
4330 static int if6_seq_show(struct seq_file *seq, void *v)
4331 {
4332         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4333         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4334                    &ifp->addr,
4335                    ifp->idev->dev->ifindex,
4336                    ifp->prefix_len,
4337                    ifp->scope,
4338                    (u8) ifp->flags,
4339                    ifp->idev->dev->name);
4340         return 0;
4341 }
4342
4343 static const struct seq_operations if6_seq_ops = {
4344         .start  = if6_seq_start,
4345         .next   = if6_seq_next,
4346         .show   = if6_seq_show,
4347         .stop   = if6_seq_stop,
4348 };
4349
4350 static int __net_init if6_proc_net_init(struct net *net)
4351 {
4352         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4353                         sizeof(struct if6_iter_state)))
4354                 return -ENOMEM;
4355         return 0;
4356 }
4357
4358 static void __net_exit if6_proc_net_exit(struct net *net)
4359 {
4360         remove_proc_entry("if_inet6", net->proc_net);
4361 }
4362
4363 static struct pernet_operations if6_proc_net_ops = {
4364         .init = if6_proc_net_init,
4365         .exit = if6_proc_net_exit,
4366 };
4367
4368 int __init if6_proc_init(void)
4369 {
4370         return register_pernet_subsys(&if6_proc_net_ops);
4371 }
4372
4373 void if6_proc_exit(void)
4374 {
4375         unregister_pernet_subsys(&if6_proc_net_ops);
4376 }
4377 #endif  /* CONFIG_PROC_FS */
4378
4379 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4380 /* Check if address is a home address configured on any interface. */
4381 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4382 {
4383         unsigned int hash = inet6_addr_hash(net, addr);
4384         struct inet6_ifaddr *ifp = NULL;
4385         int ret = 0;
4386
4387         rcu_read_lock();
4388         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4389                 if (!net_eq(dev_net(ifp->idev->dev), net))
4390                         continue;
4391                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4392                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4393                         ret = 1;
4394                         break;
4395                 }
4396         }
4397         rcu_read_unlock();
4398         return ret;
4399 }
4400 #endif
4401
4402 /* RFC6554 has some algorithm to avoid loops in segment routing by
4403  * checking if the segments contains any of a local interface address.
4404  *
4405  * Quote:
4406  *
4407  * To detect loops in the SRH, a router MUST determine if the SRH
4408  * includes multiple addresses assigned to any interface on that router.
4409  * If such addresses appear more than once and are separated by at least
4410  * one address not assigned to that router.
4411  */
4412 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4413                           unsigned char nsegs)
4414 {
4415         const struct in6_addr *addr;
4416         int i, ret = 0, found = 0;
4417         struct inet6_ifaddr *ifp;
4418         bool separated = false;
4419         unsigned int hash;
4420         bool hash_found;
4421
4422         rcu_read_lock();
4423         for (i = 0; i < nsegs; i++) {
4424                 addr = &segs[i];
4425                 hash = inet6_addr_hash(net, addr);
4426
4427                 hash_found = false;
4428                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4429                         if (!net_eq(dev_net(ifp->idev->dev), net))
4430                                 continue;
4431
4432                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4433                                 hash_found = true;
4434                                 break;
4435                         }
4436                 }
4437
4438                 if (hash_found) {
4439                         if (found > 1 && separated) {
4440                                 ret = 1;
4441                                 break;
4442                         }
4443
4444                         separated = false;
4445                         found++;
4446                 } else {
4447                         separated = true;
4448                 }
4449         }
4450         rcu_read_unlock();
4451
4452         return ret;
4453 }
4454
4455 /*
4456  *      Periodic address status verification
4457  */
4458
4459 static void addrconf_verify_rtnl(void)
4460 {
4461         unsigned long now, next, next_sec, next_sched;
4462         struct inet6_ifaddr *ifp;
4463         int i;
4464
4465         ASSERT_RTNL();
4466
4467         rcu_read_lock_bh();
4468         now = jiffies;
4469         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4470
4471         cancel_delayed_work(&addr_chk_work);
4472
4473         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4474 restart:
4475                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4476                         unsigned long age;
4477
4478                         /* When setting preferred_lft to a value not zero or
4479                          * infinity, while valid_lft is infinity
4480                          * IFA_F_PERMANENT has a non-infinity life time.
4481                          */
4482                         if ((ifp->flags & IFA_F_PERMANENT) &&
4483                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4484                                 continue;
4485
4486                         spin_lock(&ifp->lock);
4487                         /* We try to batch several events at once. */
4488                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4489
4490                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4491                             age >= ifp->valid_lft) {
4492                                 spin_unlock(&ifp->lock);
4493                                 in6_ifa_hold(ifp);
4494                                 rcu_read_unlock_bh();
4495                                 ipv6_del_addr(ifp);
4496                                 rcu_read_lock_bh();
4497                                 goto restart;
4498                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4499                                 spin_unlock(&ifp->lock);
4500                                 continue;
4501                         } else if (age >= ifp->prefered_lft) {
4502                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4503                                 int deprecate = 0;
4504
4505                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4506                                         deprecate = 1;
4507                                         ifp->flags |= IFA_F_DEPRECATED;
4508                                 }
4509
4510                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4511                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4512                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4513
4514                                 spin_unlock(&ifp->lock);
4515
4516                                 if (deprecate) {
4517                                         in6_ifa_hold(ifp);
4518
4519                                         ipv6_ifa_notify(0, ifp);
4520                                         in6_ifa_put(ifp);
4521                                         goto restart;
4522                                 }
4523                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4524                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4525                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4526                                         ifp->idev->cnf.dad_transmits *
4527                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4528
4529                                 if (age >= ifp->prefered_lft - regen_advance) {
4530                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4531                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4532                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4533                                         if (!ifp->regen_count && ifpub) {
4534                                                 ifp->regen_count++;
4535                                                 in6_ifa_hold(ifp);
4536                                                 in6_ifa_hold(ifpub);
4537                                                 spin_unlock(&ifp->lock);
4538
4539                                                 spin_lock(&ifpub->lock);
4540                                                 ifpub->regen_count = 0;
4541                                                 spin_unlock(&ifpub->lock);
4542                                                 rcu_read_unlock_bh();
4543                                                 ipv6_create_tempaddr(ifpub, true);
4544                                                 in6_ifa_put(ifpub);
4545                                                 in6_ifa_put(ifp);
4546                                                 rcu_read_lock_bh();
4547                                                 goto restart;
4548                                         }
4549                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4550                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4551                                 spin_unlock(&ifp->lock);
4552                         } else {
4553                                 /* ifp->prefered_lft <= ifp->valid_lft */
4554                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4555                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4556                                 spin_unlock(&ifp->lock);
4557                         }
4558                 }
4559         }
4560
4561         next_sec = round_jiffies_up(next);
4562         next_sched = next;
4563
4564         /* If rounded timeout is accurate enough, accept it. */
4565         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4566                 next_sched = next_sec;
4567
4568         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4569         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4570                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4571
4572         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4573                  now, next, next_sec, next_sched);
4574         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4575         rcu_read_unlock_bh();
4576 }
4577
4578 static void addrconf_verify_work(struct work_struct *w)
4579 {
4580         rtnl_lock();
4581         addrconf_verify_rtnl();
4582         rtnl_unlock();
4583 }
4584
4585 static void addrconf_verify(void)
4586 {
4587         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4588 }
4589
4590 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4591                                      struct in6_addr **peer_pfx)
4592 {
4593         struct in6_addr *pfx = NULL;
4594
4595         *peer_pfx = NULL;
4596
4597         if (addr)
4598                 pfx = nla_data(addr);
4599
4600         if (local) {
4601                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4602                         *peer_pfx = pfx;
4603                 pfx = nla_data(local);
4604         }
4605
4606         return pfx;
4607 }
4608
4609 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4610         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4611         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4612         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4613         [IFA_FLAGS]             = { .len = sizeof(u32) },
4614         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4615         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4616 };
4617
4618 static int
4619 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4620                   struct netlink_ext_ack *extack)
4621 {
4622         struct net *net = sock_net(skb->sk);
4623         struct ifaddrmsg *ifm;
4624         struct nlattr *tb[IFA_MAX+1];
4625         struct in6_addr *pfx, *peer_pfx;
4626         u32 ifa_flags;
4627         int err;
4628
4629         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4630                                      ifa_ipv6_policy, extack);
4631         if (err < 0)
4632                 return err;
4633
4634         ifm = nlmsg_data(nlh);
4635         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4636         if (!pfx)
4637                 return -EINVAL;
4638
4639         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4640
4641         /* We ignore other flags so far. */
4642         ifa_flags &= IFA_F_MANAGETEMPADDR;
4643
4644         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4645                               ifm->ifa_prefixlen);
4646 }
4647
4648 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4649                                unsigned long expires, u32 flags,
4650                                bool modify_peer)
4651 {
4652         struct fib6_info *f6i;
4653         u32 prio;
4654
4655         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4656                                         ifp->prefix_len,
4657                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4658         if (!f6i)
4659                 return -ENOENT;
4660
4661         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4662         if (f6i->fib6_metric != prio) {
4663                 /* delete old one */
4664                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4665
4666                 /* add new one */
4667                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4668                                       ifp->prefix_len,
4669                                       ifp->rt_priority, ifp->idev->dev,
4670                                       expires, flags, GFP_KERNEL);
4671         } else {
4672                 if (!expires)
4673                         fib6_clean_expires(f6i);
4674                 else
4675                         fib6_set_expires(f6i, expires);
4676
4677                 fib6_info_release(f6i);
4678         }
4679
4680         return 0;
4681 }
4682
4683 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4684 {
4685         u32 flags;
4686         clock_t expires;
4687         unsigned long timeout;
4688         bool was_managetempaddr;
4689         bool had_prefixroute;
4690         bool new_peer = false;
4691
4692         ASSERT_RTNL();
4693
4694         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4695                 return -EINVAL;
4696
4697         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4698             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4699                 return -EINVAL;
4700
4701         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4702                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4703
4704         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4705         if (addrconf_finite_timeout(timeout)) {
4706                 expires = jiffies_to_clock_t(timeout * HZ);
4707                 cfg->valid_lft = timeout;
4708                 flags = RTF_EXPIRES;
4709         } else {
4710                 expires = 0;
4711                 flags = 0;
4712                 cfg->ifa_flags |= IFA_F_PERMANENT;
4713         }
4714
4715         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4716         if (addrconf_finite_timeout(timeout)) {
4717                 if (timeout == 0)
4718                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4719                 cfg->preferred_lft = timeout;
4720         }
4721
4722         if (cfg->peer_pfx &&
4723             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4724                 if (!ipv6_addr_any(&ifp->peer_addr))
4725                         cleanup_prefix_route(ifp, expires, true, true);
4726                 new_peer = true;
4727         }
4728
4729         spin_lock_bh(&ifp->lock);
4730         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4731         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4732                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4733         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4734                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4735                         IFA_F_NOPREFIXROUTE);
4736         ifp->flags |= cfg->ifa_flags;
4737         ifp->tstamp = jiffies;
4738         ifp->valid_lft = cfg->valid_lft;
4739         ifp->prefered_lft = cfg->preferred_lft;
4740
4741         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4742                 ifp->rt_priority = cfg->rt_priority;
4743
4744         if (new_peer)
4745                 ifp->peer_addr = *cfg->peer_pfx;
4746
4747         spin_unlock_bh(&ifp->lock);
4748         if (!(ifp->flags&IFA_F_TENTATIVE))
4749                 ipv6_ifa_notify(0, ifp);
4750
4751         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4752                 int rc = -ENOENT;
4753
4754                 if (had_prefixroute)
4755                         rc = modify_prefix_route(ifp, expires, flags, false);
4756
4757                 /* prefix route could have been deleted; if so restore it */
4758                 if (rc == -ENOENT) {
4759                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4760                                               ifp->rt_priority, ifp->idev->dev,
4761                                               expires, flags, GFP_KERNEL);
4762                 }
4763
4764                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4765                         rc = modify_prefix_route(ifp, expires, flags, true);
4766
4767                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4768                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4769                                               ifp->rt_priority, ifp->idev->dev,
4770                                               expires, flags, GFP_KERNEL);
4771                 }
4772         } else if (had_prefixroute) {
4773                 enum cleanup_prefix_rt_t action;
4774                 unsigned long rt_expires;
4775
4776                 write_lock_bh(&ifp->idev->lock);
4777                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4778                 write_unlock_bh(&ifp->idev->lock);
4779
4780                 if (action != CLEANUP_PREFIX_RT_NOP) {
4781                         cleanup_prefix_route(ifp, rt_expires,
4782                                 action == CLEANUP_PREFIX_RT_DEL, false);
4783                 }
4784         }
4785
4786         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4787                 if (was_managetempaddr &&
4788                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4789                         cfg->valid_lft = 0;
4790                         cfg->preferred_lft = 0;
4791                 }
4792                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4793                                  cfg->preferred_lft, !was_managetempaddr,
4794                                  jiffies);
4795         }
4796
4797         addrconf_verify_rtnl();
4798
4799         return 0;
4800 }
4801
4802 static int
4803 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4804                   struct netlink_ext_ack *extack)
4805 {
4806         struct net *net = sock_net(skb->sk);
4807         struct ifaddrmsg *ifm;
4808         struct nlattr *tb[IFA_MAX+1];
4809         struct in6_addr *peer_pfx;
4810         struct inet6_ifaddr *ifa;
4811         struct net_device *dev;
4812         struct inet6_dev *idev;
4813         struct ifa6_config cfg;
4814         int err;
4815
4816         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4817                                      ifa_ipv6_policy, extack);
4818         if (err < 0)
4819                 return err;
4820
4821         memset(&cfg, 0, sizeof(cfg));
4822
4823         ifm = nlmsg_data(nlh);
4824         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4825         if (!cfg.pfx)
4826                 return -EINVAL;
4827
4828         cfg.peer_pfx = peer_pfx;
4829         cfg.plen = ifm->ifa_prefixlen;
4830         if (tb[IFA_RT_PRIORITY])
4831                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4832
4833         cfg.valid_lft = INFINITY_LIFE_TIME;
4834         cfg.preferred_lft = INFINITY_LIFE_TIME;
4835
4836         if (tb[IFA_CACHEINFO]) {
4837                 struct ifa_cacheinfo *ci;
4838
4839                 ci = nla_data(tb[IFA_CACHEINFO]);
4840                 cfg.valid_lft = ci->ifa_valid;
4841                 cfg.preferred_lft = ci->ifa_prefered;
4842         }
4843
4844         dev =  __dev_get_by_index(net, ifm->ifa_index);
4845         if (!dev)
4846                 return -ENODEV;
4847
4848         if (tb[IFA_FLAGS])
4849                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4850         else
4851                 cfg.ifa_flags = ifm->ifa_flags;
4852
4853         /* We ignore other flags so far. */
4854         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4855                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4856                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4857
4858         idev = ipv6_find_idev(dev);
4859         if (IS_ERR(idev))
4860                 return PTR_ERR(idev);
4861
4862         if (!ipv6_allow_optimistic_dad(net, idev))
4863                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4864
4865         if (cfg.ifa_flags & IFA_F_NODAD &&
4866             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4867                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4868                 return -EINVAL;
4869         }
4870
4871         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4872         if (!ifa) {
4873                 /*
4874                  * It would be best to check for !NLM_F_CREATE here but
4875                  * userspace already relies on not having to provide this.
4876                  */
4877                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4878         }
4879
4880         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4881             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4882                 err = -EEXIST;
4883         else
4884                 err = inet6_addr_modify(ifa, &cfg);
4885
4886         in6_ifa_put(ifa);
4887
4888         return err;
4889 }
4890
4891 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4892                           u8 scope, int ifindex)
4893 {
4894         struct ifaddrmsg *ifm;
4895
4896         ifm = nlmsg_data(nlh);
4897         ifm->ifa_family = AF_INET6;
4898         ifm->ifa_prefixlen = prefixlen;
4899         ifm->ifa_flags = flags;
4900         ifm->ifa_scope = scope;
4901         ifm->ifa_index = ifindex;
4902 }
4903
4904 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4905                          unsigned long tstamp, u32 preferred, u32 valid)
4906 {
4907         struct ifa_cacheinfo ci;
4908
4909         ci.cstamp = cstamp_delta(cstamp);
4910         ci.tstamp = cstamp_delta(tstamp);
4911         ci.ifa_prefered = preferred;
4912         ci.ifa_valid = valid;
4913
4914         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4915 }
4916
4917 static inline int rt_scope(int ifa_scope)
4918 {
4919         if (ifa_scope & IFA_HOST)
4920                 return RT_SCOPE_HOST;
4921         else if (ifa_scope & IFA_LINK)
4922                 return RT_SCOPE_LINK;
4923         else if (ifa_scope & IFA_SITE)
4924                 return RT_SCOPE_SITE;
4925         else
4926                 return RT_SCOPE_UNIVERSE;
4927 }
4928
4929 static inline int inet6_ifaddr_msgsize(void)
4930 {
4931         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4932                + nla_total_size(16) /* IFA_LOCAL */
4933                + nla_total_size(16) /* IFA_ADDRESS */
4934                + nla_total_size(sizeof(struct ifa_cacheinfo))
4935                + nla_total_size(4)  /* IFA_FLAGS */
4936                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4937 }
4938
4939 enum addr_type_t {
4940         UNICAST_ADDR,
4941         MULTICAST_ADDR,
4942         ANYCAST_ADDR,
4943 };
4944
4945 struct inet6_fill_args {
4946         u32 portid;
4947         u32 seq;
4948         int event;
4949         unsigned int flags;
4950         int netnsid;
4951         int ifindex;
4952         enum addr_type_t type;
4953 };
4954
4955 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4956                              struct inet6_fill_args *args)
4957 {
4958         struct nlmsghdr  *nlh;
4959         u32 preferred, valid;
4960
4961         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4962                         sizeof(struct ifaddrmsg), args->flags);
4963         if (!nlh)
4964                 return -EMSGSIZE;
4965
4966         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4967                       ifa->idev->dev->ifindex);
4968
4969         if (args->netnsid >= 0 &&
4970             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4971                 goto error;
4972
4973         if (!((ifa->flags&IFA_F_PERMANENT) &&
4974               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4975                 preferred = ifa->prefered_lft;
4976                 valid = ifa->valid_lft;
4977                 if (preferred != INFINITY_LIFE_TIME) {
4978                         long tval = (jiffies - ifa->tstamp)/HZ;
4979                         if (preferred > tval)
4980                                 preferred -= tval;
4981                         else
4982                                 preferred = 0;
4983                         if (valid != INFINITY_LIFE_TIME) {
4984                                 if (valid > tval)
4985                                         valid -= tval;
4986                                 else
4987                                         valid = 0;
4988                         }
4989                 }
4990         } else {
4991                 preferred = INFINITY_LIFE_TIME;
4992                 valid = INFINITY_LIFE_TIME;
4993         }
4994
4995         if (!ipv6_addr_any(&ifa->peer_addr)) {
4996                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4997                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4998                         goto error;
4999         } else
5000                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5001                         goto error;
5002
5003         if (ifa->rt_priority &&
5004             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5005                 goto error;
5006
5007         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5008                 goto error;
5009
5010         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5011                 goto error;
5012
5013         nlmsg_end(skb, nlh);
5014         return 0;
5015
5016 error:
5017         nlmsg_cancel(skb, nlh);
5018         return -EMSGSIZE;
5019 }
5020
5021 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5022                                struct inet6_fill_args *args)
5023 {
5024         struct nlmsghdr  *nlh;
5025         u8 scope = RT_SCOPE_UNIVERSE;
5026         int ifindex = ifmca->idev->dev->ifindex;
5027
5028         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5029                 scope = RT_SCOPE_SITE;
5030
5031         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5032                         sizeof(struct ifaddrmsg), args->flags);
5033         if (!nlh)
5034                 return -EMSGSIZE;
5035
5036         if (args->netnsid >= 0 &&
5037             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5038                 nlmsg_cancel(skb, nlh);
5039                 return -EMSGSIZE;
5040         }
5041
5042         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5043         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5044             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5045                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5046                 nlmsg_cancel(skb, nlh);
5047                 return -EMSGSIZE;
5048         }
5049
5050         nlmsg_end(skb, nlh);
5051         return 0;
5052 }
5053
5054 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5055                                struct inet6_fill_args *args)
5056 {
5057         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5058         int ifindex = dev ? dev->ifindex : 1;
5059         struct nlmsghdr  *nlh;
5060         u8 scope = RT_SCOPE_UNIVERSE;
5061
5062         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5063                 scope = RT_SCOPE_SITE;
5064
5065         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5066                         sizeof(struct ifaddrmsg), args->flags);
5067         if (!nlh)
5068                 return -EMSGSIZE;
5069
5070         if (args->netnsid >= 0 &&
5071             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5072                 nlmsg_cancel(skb, nlh);
5073                 return -EMSGSIZE;
5074         }
5075
5076         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5077         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5078             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5079                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5080                 nlmsg_cancel(skb, nlh);
5081                 return -EMSGSIZE;
5082         }
5083
5084         nlmsg_end(skb, nlh);
5085         return 0;
5086 }
5087
5088 /* called with rcu_read_lock() */
5089 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5090                           struct netlink_callback *cb, int s_ip_idx,
5091                           struct inet6_fill_args *fillargs)
5092 {
5093         struct ifmcaddr6 *ifmca;
5094         struct ifacaddr6 *ifaca;
5095         int ip_idx = 0;
5096         int err = 1;
5097
5098         read_lock_bh(&idev->lock);
5099         switch (fillargs->type) {
5100         case UNICAST_ADDR: {
5101                 struct inet6_ifaddr *ifa;
5102                 fillargs->event = RTM_NEWADDR;
5103
5104                 /* unicast address incl. temp addr */
5105                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5106                         if (ip_idx < s_ip_idx)
5107                                 goto next;
5108                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5109                         if (err < 0)
5110                                 break;
5111                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5112 next:
5113                         ip_idx++;
5114                 }
5115                 break;
5116         }
5117         case MULTICAST_ADDR:
5118                 read_unlock_bh(&idev->lock);
5119                 fillargs->event = RTM_GETMULTICAST;
5120
5121                 /* multicast address */
5122                 for (ifmca = rcu_dereference(idev->mc_list);
5123                      ifmca;
5124                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5125                         if (ip_idx < s_ip_idx)
5126                                 continue;
5127                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5128                         if (err < 0)
5129                                 break;
5130                 }
5131                 read_lock_bh(&idev->lock);
5132                 break;
5133         case ANYCAST_ADDR:
5134                 fillargs->event = RTM_GETANYCAST;
5135                 /* anycast address */
5136                 for (ifaca = idev->ac_list; ifaca;
5137                      ifaca = ifaca->aca_next, ip_idx++) {
5138                         if (ip_idx < s_ip_idx)
5139                                 continue;
5140                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5141                         if (err < 0)
5142                                 break;
5143                 }
5144                 break;
5145         default:
5146                 break;
5147         }
5148         read_unlock_bh(&idev->lock);
5149         cb->args[2] = ip_idx;
5150         return err;
5151 }
5152
5153 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5154                                        struct inet6_fill_args *fillargs,
5155                                        struct net **tgt_net, struct sock *sk,
5156                                        struct netlink_callback *cb)
5157 {
5158         struct netlink_ext_ack *extack = cb->extack;
5159         struct nlattr *tb[IFA_MAX+1];
5160         struct ifaddrmsg *ifm;
5161         int err, i;
5162
5163         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5164                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5165                 return -EINVAL;
5166         }
5167
5168         ifm = nlmsg_data(nlh);
5169         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5170                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5171                 return -EINVAL;
5172         }
5173
5174         fillargs->ifindex = ifm->ifa_index;
5175         if (fillargs->ifindex) {
5176                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5177                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5178         }
5179
5180         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5181                                             ifa_ipv6_policy, extack);
5182         if (err < 0)
5183                 return err;
5184
5185         for (i = 0; i <= IFA_MAX; ++i) {
5186                 if (!tb[i])
5187                         continue;
5188
5189                 if (i == IFA_TARGET_NETNSID) {
5190                         struct net *net;
5191
5192                         fillargs->netnsid = nla_get_s32(tb[i]);
5193                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5194                         if (IS_ERR(net)) {
5195                                 fillargs->netnsid = -1;
5196                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5197                                 return PTR_ERR(net);
5198                         }
5199                         *tgt_net = net;
5200                 } else {
5201                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5202                         return -EINVAL;
5203                 }
5204         }
5205
5206         return 0;
5207 }
5208
5209 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5210                            enum addr_type_t type)
5211 {
5212         const struct nlmsghdr *nlh = cb->nlh;
5213         struct inet6_fill_args fillargs = {
5214                 .portid = NETLINK_CB(cb->skb).portid,
5215                 .seq = cb->nlh->nlmsg_seq,
5216                 .flags = NLM_F_MULTI,
5217                 .netnsid = -1,
5218                 .type = type,
5219         };
5220         struct net *tgt_net = sock_net(skb->sk);
5221         int idx, s_idx, s_ip_idx;
5222         int h, s_h;
5223         struct net_device *dev;
5224         struct inet6_dev *idev;
5225         struct hlist_head *head;
5226         int err = 0;
5227
5228         s_h = cb->args[0];
5229         s_idx = idx = cb->args[1];
5230         s_ip_idx = cb->args[2];
5231
5232         if (cb->strict_check) {
5233                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5234                                                   skb->sk, cb);
5235                 if (err < 0)
5236                         goto put_tgt_net;
5237
5238                 err = 0;
5239                 if (fillargs.ifindex) {
5240                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5241                         if (!dev) {
5242                                 err = -ENODEV;
5243                                 goto put_tgt_net;
5244                         }
5245                         idev = __in6_dev_get(dev);
5246                         if (idev) {
5247                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5248                                                      &fillargs);
5249                                 if (err > 0)
5250                                         err = 0;
5251                         }
5252                         goto put_tgt_net;
5253                 }
5254         }
5255
5256         rcu_read_lock();
5257         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5258         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5259                 idx = 0;
5260                 head = &tgt_net->dev_index_head[h];
5261                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5262                         if (idx < s_idx)
5263                                 goto cont;
5264                         if (h > s_h || idx > s_idx)
5265                                 s_ip_idx = 0;
5266                         idev = __in6_dev_get(dev);
5267                         if (!idev)
5268                                 goto cont;
5269
5270                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5271                                            &fillargs) < 0)
5272                                 goto done;
5273 cont:
5274                         idx++;
5275                 }
5276         }
5277 done:
5278         rcu_read_unlock();
5279         cb->args[0] = h;
5280         cb->args[1] = idx;
5281 put_tgt_net:
5282         if (fillargs.netnsid >= 0)
5283                 put_net(tgt_net);
5284
5285         return skb->len ? : err;
5286 }
5287
5288 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5289 {
5290         enum addr_type_t type = UNICAST_ADDR;
5291
5292         return inet6_dump_addr(skb, cb, type);
5293 }
5294
5295 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5296 {
5297         enum addr_type_t type = MULTICAST_ADDR;
5298
5299         return inet6_dump_addr(skb, cb, type);
5300 }
5301
5302
5303 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5304 {
5305         enum addr_type_t type = ANYCAST_ADDR;
5306
5307         return inet6_dump_addr(skb, cb, type);
5308 }
5309
5310 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5311                                        const struct nlmsghdr *nlh,
5312                                        struct nlattr **tb,
5313                                        struct netlink_ext_ack *extack)
5314 {
5315         struct ifaddrmsg *ifm;
5316         int i, err;
5317
5318         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5319                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5320                 return -EINVAL;
5321         }
5322
5323         if (!netlink_strict_get_check(skb))
5324                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5325                                               ifa_ipv6_policy, extack);
5326
5327         ifm = nlmsg_data(nlh);
5328         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5329                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5330                 return -EINVAL;
5331         }
5332
5333         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5334                                             ifa_ipv6_policy, extack);
5335         if (err)
5336                 return err;
5337
5338         for (i = 0; i <= IFA_MAX; i++) {
5339                 if (!tb[i])
5340                         continue;
5341
5342                 switch (i) {
5343                 case IFA_TARGET_NETNSID:
5344                 case IFA_ADDRESS:
5345                 case IFA_LOCAL:
5346                         break;
5347                 default:
5348                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5349                         return -EINVAL;
5350                 }
5351         }
5352
5353         return 0;
5354 }
5355
5356 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5357                              struct netlink_ext_ack *extack)
5358 {
5359         struct net *tgt_net = sock_net(in_skb->sk);
5360         struct inet6_fill_args fillargs = {
5361                 .portid = NETLINK_CB(in_skb).portid,
5362                 .seq = nlh->nlmsg_seq,
5363                 .event = RTM_NEWADDR,
5364                 .flags = 0,
5365                 .netnsid = -1,
5366         };
5367         struct ifaddrmsg *ifm;
5368         struct nlattr *tb[IFA_MAX+1];
5369         struct in6_addr *addr = NULL, *peer;
5370         struct net_device *dev = NULL;
5371         struct inet6_ifaddr *ifa;
5372         struct sk_buff *skb;
5373         int err;
5374
5375         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5376         if (err < 0)
5377                 return err;
5378
5379         if (tb[IFA_TARGET_NETNSID]) {
5380                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5381
5382                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5383                                                   fillargs.netnsid);
5384                 if (IS_ERR(tgt_net))
5385                         return PTR_ERR(tgt_net);
5386         }
5387
5388         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5389         if (!addr)
5390                 return -EINVAL;
5391
5392         ifm = nlmsg_data(nlh);
5393         if (ifm->ifa_index)
5394                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5395
5396         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5397         if (!ifa) {
5398                 err = -EADDRNOTAVAIL;
5399                 goto errout;
5400         }
5401
5402         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5403         if (!skb) {
5404                 err = -ENOBUFS;
5405                 goto errout_ifa;
5406         }
5407
5408         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5409         if (err < 0) {
5410                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5411                 WARN_ON(err == -EMSGSIZE);
5412                 kfree_skb(skb);
5413                 goto errout_ifa;
5414         }
5415         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5416 errout_ifa:
5417         in6_ifa_put(ifa);
5418 errout:
5419         dev_put(dev);
5420         if (fillargs.netnsid >= 0)
5421                 put_net(tgt_net);
5422
5423         return err;
5424 }
5425
5426 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5427 {
5428         struct sk_buff *skb;
5429         struct net *net = dev_net(ifa->idev->dev);
5430         struct inet6_fill_args fillargs = {
5431                 .portid = 0,
5432                 .seq = 0,
5433                 .event = event,
5434                 .flags = 0,
5435                 .netnsid = -1,
5436         };
5437         int err = -ENOBUFS;
5438
5439         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5440         if (!skb)
5441                 goto errout;
5442
5443         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5444         if (err < 0) {
5445                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5446                 WARN_ON(err == -EMSGSIZE);
5447                 kfree_skb(skb);
5448                 goto errout;
5449         }
5450         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5451         return;
5452 errout:
5453         if (err < 0)
5454                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5455 }
5456
5457 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5458                                 __s32 *array, int bytes)
5459 {
5460         BUG_ON(bytes < (DEVCONF_MAX * 4));
5461
5462         memset(array, 0, bytes);
5463         array[DEVCONF_FORWARDING] = cnf->forwarding;
5464         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5465         array[DEVCONF_MTU6] = cnf->mtu6;
5466         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5467         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5468         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5469         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5470         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5471         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5472                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5473         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5474                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5475         array[DEVCONF_RTR_SOLICIT_DELAY] =
5476                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5477         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5478         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5479                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5480         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5481                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5482         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5483         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5484         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5485         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5486         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5487         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5488         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5489         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5490         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5491         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5492 #ifdef CONFIG_IPV6_ROUTER_PREF
5493         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5494         array[DEVCONF_RTR_PROBE_INTERVAL] =
5495                 jiffies_to_msecs(cnf->rtr_probe_interval);
5496 #ifdef CONFIG_IPV6_ROUTE_INFO
5497         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5498         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5499 #endif
5500 #endif
5501         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5502         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5503 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5504         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5505         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5506 #endif
5507 #ifdef CONFIG_IPV6_MROUTE
5508         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5509 #endif
5510         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5511         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5512         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5513         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5514         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5515         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5516         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5517         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5518         /* we omit DEVCONF_STABLE_SECRET for now */
5519         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5520         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5521         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5522         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5523         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5524 #ifdef CONFIG_IPV6_SEG6_HMAC
5525         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5526 #endif
5527         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5528         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5529         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5530         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5531         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5532         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5533         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5534         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5535 }
5536
5537 static inline size_t inet6_ifla6_size(void)
5538 {
5539         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5540              + nla_total_size(sizeof(struct ifla_cacheinfo))
5541              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5542              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5543              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5544              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5545              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5546              + 0;
5547 }
5548
5549 static inline size_t inet6_if_nlmsg_size(void)
5550 {
5551         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5552                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5553                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5554                + nla_total_size(4) /* IFLA_MTU */
5555                + nla_total_size(4) /* IFLA_LINK */
5556                + nla_total_size(1) /* IFLA_OPERSTATE */
5557                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5558 }
5559
5560 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5561                                         int bytes)
5562 {
5563         int i;
5564         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5565         BUG_ON(pad < 0);
5566
5567         /* Use put_unaligned() because stats may not be aligned for u64. */
5568         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5569         for (i = 1; i < ICMP6_MIB_MAX; i++)
5570                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5571
5572         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5573 }
5574
5575 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5576                                         int bytes, size_t syncpoff)
5577 {
5578         int i, c;
5579         u64 buff[IPSTATS_MIB_MAX];
5580         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5581
5582         BUG_ON(pad < 0);
5583
5584         memset(buff, 0, sizeof(buff));
5585         buff[0] = IPSTATS_MIB_MAX;
5586
5587         for_each_possible_cpu(c) {
5588                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5589                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5590         }
5591
5592         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5593         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5594 }
5595
5596 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5597                              int bytes)
5598 {
5599         switch (attrtype) {
5600         case IFLA_INET6_STATS:
5601                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5602                                      offsetof(struct ipstats_mib, syncp));
5603                 break;
5604         case IFLA_INET6_ICMP6STATS:
5605                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5606                 break;
5607         }
5608 }
5609
5610 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5611                                   u32 ext_filter_mask)
5612 {
5613         struct nlattr *nla;
5614         struct ifla_cacheinfo ci;
5615
5616         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5617                 goto nla_put_failure;
5618         ci.max_reasm_len = IPV6_MAXPLEN;
5619         ci.tstamp = cstamp_delta(idev->tstamp);
5620         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5621         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5622         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5623                 goto nla_put_failure;
5624         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5625         if (!nla)
5626                 goto nla_put_failure;
5627         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5628
5629         /* XXX - MC not implemented */
5630
5631         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5632                 return 0;
5633
5634         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5635         if (!nla)
5636                 goto nla_put_failure;
5637         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5638
5639         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5640         if (!nla)
5641                 goto nla_put_failure;
5642         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5643
5644         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5645         if (!nla)
5646                 goto nla_put_failure;
5647         read_lock_bh(&idev->lock);
5648         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5649         read_unlock_bh(&idev->lock);
5650
5651         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5652                 goto nla_put_failure;
5653
5654         return 0;
5655
5656 nla_put_failure:
5657         return -EMSGSIZE;
5658 }
5659
5660 static size_t inet6_get_link_af_size(const struct net_device *dev,
5661                                      u32 ext_filter_mask)
5662 {
5663         if (!__in6_dev_get(dev))
5664                 return 0;
5665
5666         return inet6_ifla6_size();
5667 }
5668
5669 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5670                               u32 ext_filter_mask)
5671 {
5672         struct inet6_dev *idev = __in6_dev_get(dev);
5673
5674         if (!idev)
5675                 return -ENODATA;
5676
5677         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5678                 return -EMSGSIZE;
5679
5680         return 0;
5681 }
5682
5683 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5684                              struct netlink_ext_ack *extack)
5685 {
5686         struct inet6_ifaddr *ifp;
5687         struct net_device *dev = idev->dev;
5688         bool clear_token, update_rs = false;
5689         struct in6_addr ll_addr;
5690
5691         ASSERT_RTNL();
5692
5693         if (!token)
5694                 return -EINVAL;
5695
5696         if (dev->flags & IFF_LOOPBACK) {
5697                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5698                 return -EINVAL;
5699         }
5700
5701         if (dev->flags & IFF_NOARP) {
5702                 NL_SET_ERR_MSG_MOD(extack,
5703                                    "Device does not do neighbour discovery");
5704                 return -EINVAL;
5705         }
5706
5707         if (!ipv6_accept_ra(idev)) {
5708                 NL_SET_ERR_MSG_MOD(extack,
5709                                    "Router advertisement is disabled on device");
5710                 return -EINVAL;
5711         }
5712
5713         if (idev->cnf.rtr_solicits == 0) {
5714                 NL_SET_ERR_MSG(extack,
5715                                "Router solicitation is disabled on device");
5716                 return -EINVAL;
5717         }
5718
5719         write_lock_bh(&idev->lock);
5720
5721         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5722         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5723
5724         write_unlock_bh(&idev->lock);
5725
5726         clear_token = ipv6_addr_any(token);
5727         if (clear_token)
5728                 goto update_lft;
5729
5730         if (!idev->dead && (idev->if_flags & IF_READY) &&
5731             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5732                              IFA_F_OPTIMISTIC)) {
5733                 /* If we're not ready, then normal ifup will take care
5734                  * of this. Otherwise, we need to request our rs here.
5735                  */
5736                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5737                 update_rs = true;
5738         }
5739
5740 update_lft:
5741         write_lock_bh(&idev->lock);
5742
5743         if (update_rs) {
5744                 idev->if_flags |= IF_RS_SENT;
5745                 idev->rs_interval = rfc3315_s14_backoff_init(
5746                         idev->cnf.rtr_solicit_interval);
5747                 idev->rs_probes = 1;
5748                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5749         }
5750
5751         /* Well, that's kinda nasty ... */
5752         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5753                 spin_lock(&ifp->lock);
5754                 if (ifp->tokenized) {
5755                         ifp->valid_lft = 0;
5756                         ifp->prefered_lft = 0;
5757                 }
5758                 spin_unlock(&ifp->lock);
5759         }
5760
5761         write_unlock_bh(&idev->lock);
5762         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5763         addrconf_verify_rtnl();
5764         return 0;
5765 }
5766
5767 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5768         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5769         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5770 };
5771
5772 static int check_addr_gen_mode(int mode)
5773 {
5774         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5775             mode != IN6_ADDR_GEN_MODE_NONE &&
5776             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5777             mode != IN6_ADDR_GEN_MODE_RANDOM)
5778                 return -EINVAL;
5779         return 1;
5780 }
5781
5782 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5783                                 int mode)
5784 {
5785         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5786             !idev->cnf.stable_secret.initialized &&
5787             !net->ipv6.devconf_dflt->stable_secret.initialized)
5788                 return -EINVAL;
5789         return 1;
5790 }
5791
5792 static int inet6_validate_link_af(const struct net_device *dev,
5793                                   const struct nlattr *nla,
5794                                   struct netlink_ext_ack *extack)
5795 {
5796         struct nlattr *tb[IFLA_INET6_MAX + 1];
5797         struct inet6_dev *idev = NULL;
5798         int err;
5799
5800         if (dev) {
5801                 idev = __in6_dev_get(dev);
5802                 if (!idev)
5803                         return -EAFNOSUPPORT;
5804         }
5805
5806         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5807                                           inet6_af_policy, extack);
5808         if (err)
5809                 return err;
5810
5811         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5812                 return -EINVAL;
5813
5814         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5815                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5816
5817                 if (check_addr_gen_mode(mode) < 0)
5818                         return -EINVAL;
5819                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5820                         return -EINVAL;
5821         }
5822
5823         return 0;
5824 }
5825
5826 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5827                              struct netlink_ext_ack *extack)
5828 {
5829         struct inet6_dev *idev = __in6_dev_get(dev);
5830         struct nlattr *tb[IFLA_INET6_MAX + 1];
5831         int err;
5832
5833         if (!idev)
5834                 return -EAFNOSUPPORT;
5835
5836         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5837                 return -EINVAL;
5838
5839         if (tb[IFLA_INET6_TOKEN]) {
5840                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5841                                         extack);
5842                 if (err)
5843                         return err;
5844         }
5845
5846         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5847                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5848
5849                 idev->cnf.addr_gen_mode = mode;
5850         }
5851
5852         return 0;
5853 }
5854
5855 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5856                              u32 portid, u32 seq, int event, unsigned int flags)
5857 {
5858         struct net_device *dev = idev->dev;
5859         struct ifinfomsg *hdr;
5860         struct nlmsghdr *nlh;
5861         void *protoinfo;
5862
5863         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5864         if (!nlh)
5865                 return -EMSGSIZE;
5866
5867         hdr = nlmsg_data(nlh);
5868         hdr->ifi_family = AF_INET6;
5869         hdr->__ifi_pad = 0;
5870         hdr->ifi_type = dev->type;
5871         hdr->ifi_index = dev->ifindex;
5872         hdr->ifi_flags = dev_get_flags(dev);
5873         hdr->ifi_change = 0;
5874
5875         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5876             (dev->addr_len &&
5877              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5878             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5879             (dev->ifindex != dev_get_iflink(dev) &&
5880              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5881             nla_put_u8(skb, IFLA_OPERSTATE,
5882                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5883                 goto nla_put_failure;
5884         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5885         if (!protoinfo)
5886                 goto nla_put_failure;
5887
5888         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5889                 goto nla_put_failure;
5890
5891         nla_nest_end(skb, protoinfo);
5892         nlmsg_end(skb, nlh);
5893         return 0;
5894
5895 nla_put_failure:
5896         nlmsg_cancel(skb, nlh);
5897         return -EMSGSIZE;
5898 }
5899
5900 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5901                                    struct netlink_ext_ack *extack)
5902 {
5903         struct ifinfomsg *ifm;
5904
5905         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5906                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5907                 return -EINVAL;
5908         }
5909
5910         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5911                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5912                 return -EINVAL;
5913         }
5914
5915         ifm = nlmsg_data(nlh);
5916         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5917             ifm->ifi_change || ifm->ifi_index) {
5918                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5919                 return -EINVAL;
5920         }
5921
5922         return 0;
5923 }
5924
5925 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5926 {
5927         struct net *net = sock_net(skb->sk);
5928         int h, s_h;
5929         int idx = 0, s_idx;
5930         struct net_device *dev;
5931         struct inet6_dev *idev;
5932         struct hlist_head *head;
5933
5934         /* only requests using strict checking can pass data to
5935          * influence the dump
5936          */
5937         if (cb->strict_check) {
5938                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5939
5940                 if (err < 0)
5941                         return err;
5942         }
5943
5944         s_h = cb->args[0];
5945         s_idx = cb->args[1];
5946
5947         rcu_read_lock();
5948         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5949                 idx = 0;
5950                 head = &net->dev_index_head[h];
5951                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5952                         if (idx < s_idx)
5953                                 goto cont;
5954                         idev = __in6_dev_get(dev);
5955                         if (!idev)
5956                                 goto cont;
5957                         if (inet6_fill_ifinfo(skb, idev,
5958                                               NETLINK_CB(cb->skb).portid,
5959                                               cb->nlh->nlmsg_seq,
5960                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5961                                 goto out;
5962 cont:
5963                         idx++;
5964                 }
5965         }
5966 out:
5967         rcu_read_unlock();
5968         cb->args[1] = idx;
5969         cb->args[0] = h;
5970
5971         return skb->len;
5972 }
5973
5974 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5975 {
5976         struct sk_buff *skb;
5977         struct net *net = dev_net(idev->dev);
5978         int err = -ENOBUFS;
5979
5980         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5981         if (!skb)
5982                 goto errout;
5983
5984         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5985         if (err < 0) {
5986                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5987                 WARN_ON(err == -EMSGSIZE);
5988                 kfree_skb(skb);
5989                 goto errout;
5990         }
5991         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5992         return;
5993 errout:
5994         if (err < 0)
5995                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5996 }
5997
5998 static inline size_t inet6_prefix_nlmsg_size(void)
5999 {
6000         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6001                + nla_total_size(sizeof(struct in6_addr))
6002                + nla_total_size(sizeof(struct prefix_cacheinfo));
6003 }
6004
6005 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6006                              struct prefix_info *pinfo, u32 portid, u32 seq,
6007                              int event, unsigned int flags)
6008 {
6009         struct prefixmsg *pmsg;
6010         struct nlmsghdr *nlh;
6011         struct prefix_cacheinfo ci;
6012
6013         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6014         if (!nlh)
6015                 return -EMSGSIZE;
6016
6017         pmsg = nlmsg_data(nlh);
6018         pmsg->prefix_family = AF_INET6;
6019         pmsg->prefix_pad1 = 0;
6020         pmsg->prefix_pad2 = 0;
6021         pmsg->prefix_ifindex = idev->dev->ifindex;
6022         pmsg->prefix_len = pinfo->prefix_len;
6023         pmsg->prefix_type = pinfo->type;
6024         pmsg->prefix_pad3 = 0;
6025         pmsg->prefix_flags = 0;
6026         if (pinfo->onlink)
6027                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6028         if (pinfo->autoconf)
6029                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6030
6031         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6032                 goto nla_put_failure;
6033         ci.preferred_time = ntohl(pinfo->prefered);
6034         ci.valid_time = ntohl(pinfo->valid);
6035         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6036                 goto nla_put_failure;
6037         nlmsg_end(skb, nlh);
6038         return 0;
6039
6040 nla_put_failure:
6041         nlmsg_cancel(skb, nlh);
6042         return -EMSGSIZE;
6043 }
6044
6045 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6046                          struct prefix_info *pinfo)
6047 {
6048         struct sk_buff *skb;
6049         struct net *net = dev_net(idev->dev);
6050         int err = -ENOBUFS;
6051
6052         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6053         if (!skb)
6054                 goto errout;
6055
6056         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6057         if (err < 0) {
6058                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6059                 WARN_ON(err == -EMSGSIZE);
6060                 kfree_skb(skb);
6061                 goto errout;
6062         }
6063         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6064         return;
6065 errout:
6066         if (err < 0)
6067                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6068 }
6069
6070 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6071 {
6072         struct net *net = dev_net(ifp->idev->dev);
6073
6074         if (event)
6075                 ASSERT_RTNL();
6076
6077         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6078
6079         switch (event) {
6080         case RTM_NEWADDR:
6081                 /*
6082                  * If the address was optimistic we inserted the route at the
6083                  * start of our DAD process, so we don't need to do it again.
6084                  * If the device was taken down in the middle of the DAD
6085                  * cycle there is a race where we could get here without a
6086                  * host route, so nothing to insert. That will be fixed when
6087                  * the device is brought up.
6088                  */
6089                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6090                         ip6_ins_rt(net, ifp->rt);
6091                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6092                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6093                                 &ifp->addr, ifp->idev->dev->name);
6094                 }
6095
6096                 if (ifp->idev->cnf.forwarding)
6097                         addrconf_join_anycast(ifp);
6098                 if (!ipv6_addr_any(&ifp->peer_addr))
6099                         addrconf_prefix_route(&ifp->peer_addr, 128,
6100                                               ifp->rt_priority, ifp->idev->dev,
6101                                               0, 0, GFP_ATOMIC);
6102                 break;
6103         case RTM_DELADDR:
6104                 if (ifp->idev->cnf.forwarding)
6105                         addrconf_leave_anycast(ifp);
6106                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6107                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6108                         struct fib6_info *rt;
6109
6110                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6111                                                        ifp->idev->dev, 0, 0,
6112                                                        false);
6113                         if (rt)
6114                                 ip6_del_rt(net, rt, false);
6115                 }
6116                 if (ifp->rt) {
6117                         ip6_del_rt(net, ifp->rt, false);
6118                         ifp->rt = NULL;
6119                 }
6120                 rt_genid_bump_ipv6(net);
6121                 break;
6122         }
6123         atomic_inc(&net->ipv6.dev_addr_genid);
6124 }
6125
6126 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6127 {
6128         if (likely(ifp->idev->dead == 0))
6129                 __ipv6_ifa_notify(event, ifp);
6130 }
6131
6132 #ifdef CONFIG_SYSCTL
6133
6134 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6135                 void *buffer, size_t *lenp, loff_t *ppos)
6136 {
6137         int *valp = ctl->data;
6138         int val = *valp;
6139         loff_t pos = *ppos;
6140         struct ctl_table lctl;
6141         int ret;
6142
6143         /*
6144          * ctl->data points to idev->cnf.forwarding, we should
6145          * not modify it until we get the rtnl lock.
6146          */
6147         lctl = *ctl;
6148         lctl.data = &val;
6149
6150         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6151
6152         if (write)
6153                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6154         if (ret)
6155                 *ppos = pos;
6156         return ret;
6157 }
6158
6159 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6160                 void *buffer, size_t *lenp, loff_t *ppos)
6161 {
6162         struct inet6_dev *idev = ctl->extra1;
6163         int min_mtu = IPV6_MIN_MTU;
6164         struct ctl_table lctl;
6165
6166         lctl = *ctl;
6167         lctl.extra1 = &min_mtu;
6168         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6169
6170         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6171 }
6172
6173 static void dev_disable_change(struct inet6_dev *idev)
6174 {
6175         struct netdev_notifier_info info;
6176
6177         if (!idev || !idev->dev)
6178                 return;
6179
6180         netdev_notifier_info_init(&info, idev->dev);
6181         if (idev->cnf.disable_ipv6)
6182                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6183         else
6184                 addrconf_notify(NULL, NETDEV_UP, &info);
6185 }
6186
6187 static void addrconf_disable_change(struct net *net, __s32 newf)
6188 {
6189         struct net_device *dev;
6190         struct inet6_dev *idev;
6191
6192         for_each_netdev(net, dev) {
6193                 idev = __in6_dev_get(dev);
6194                 if (idev) {
6195                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6196                         idev->cnf.disable_ipv6 = newf;
6197                         if (changed)
6198                                 dev_disable_change(idev);
6199                 }
6200         }
6201 }
6202
6203 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6204 {
6205         struct net *net;
6206         int old;
6207
6208         if (!rtnl_trylock())
6209                 return restart_syscall();
6210
6211         net = (struct net *)table->extra2;
6212         old = *p;
6213         *p = newf;
6214
6215         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6216                 rtnl_unlock();
6217                 return 0;
6218         }
6219
6220         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6221                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6222                 addrconf_disable_change(net, newf);
6223         } else if ((!newf) ^ (!old))
6224                 dev_disable_change((struct inet6_dev *)table->extra1);
6225
6226         rtnl_unlock();
6227         return 0;
6228 }
6229
6230 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6231                 void *buffer, size_t *lenp, loff_t *ppos)
6232 {
6233         int *valp = ctl->data;
6234         int val = *valp;
6235         loff_t pos = *ppos;
6236         struct ctl_table lctl;
6237         int ret;
6238
6239         /*
6240          * ctl->data points to idev->cnf.disable_ipv6, we should
6241          * not modify it until we get the rtnl lock.
6242          */
6243         lctl = *ctl;
6244         lctl.data = &val;
6245
6246         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6247
6248         if (write)
6249                 ret = addrconf_disable_ipv6(ctl, valp, val);
6250         if (ret)
6251                 *ppos = pos;
6252         return ret;
6253 }
6254
6255 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6256                 void *buffer, size_t *lenp, loff_t *ppos)
6257 {
6258         int *valp = ctl->data;
6259         int ret;
6260         int old, new;
6261
6262         old = *valp;
6263         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6264         new = *valp;
6265
6266         if (write && old != new) {
6267                 struct net *net = ctl->extra2;
6268
6269                 if (!rtnl_trylock())
6270                         return restart_syscall();
6271
6272                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6273                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6274                                                      NETCONFA_PROXY_NEIGH,
6275                                                      NETCONFA_IFINDEX_DEFAULT,
6276                                                      net->ipv6.devconf_dflt);
6277                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6278                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6279                                                      NETCONFA_PROXY_NEIGH,
6280                                                      NETCONFA_IFINDEX_ALL,
6281                                                      net->ipv6.devconf_all);
6282                 else {
6283                         struct inet6_dev *idev = ctl->extra1;
6284
6285                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6286                                                      NETCONFA_PROXY_NEIGH,
6287                                                      idev->dev->ifindex,
6288                                                      &idev->cnf);
6289                 }
6290                 rtnl_unlock();
6291         }
6292
6293         return ret;
6294 }
6295
6296 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6297                                          void *buffer, size_t *lenp,
6298                                          loff_t *ppos)
6299 {
6300         int ret = 0;
6301         u32 new_val;
6302         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6303         struct net *net = (struct net *)ctl->extra2;
6304         struct ctl_table tmp = {
6305                 .data = &new_val,
6306                 .maxlen = sizeof(new_val),
6307                 .mode = ctl->mode,
6308         };
6309
6310         if (!rtnl_trylock())
6311                 return restart_syscall();
6312
6313         new_val = *((u32 *)ctl->data);
6314
6315         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6316         if (ret != 0)
6317                 goto out;
6318
6319         if (write) {
6320                 if (check_addr_gen_mode(new_val) < 0) {
6321                         ret = -EINVAL;
6322                         goto out;
6323                 }
6324
6325                 if (idev) {
6326                         if (check_stable_privacy(idev, net, new_val) < 0) {
6327                                 ret = -EINVAL;
6328                                 goto out;
6329                         }
6330
6331                         if (idev->cnf.addr_gen_mode != new_val) {
6332                                 idev->cnf.addr_gen_mode = new_val;
6333                                 addrconf_dev_config(idev->dev);
6334                         }
6335                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6336                         struct net_device *dev;
6337
6338                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6339                         for_each_netdev(net, dev) {
6340                                 idev = __in6_dev_get(dev);
6341                                 if (idev &&
6342                                     idev->cnf.addr_gen_mode != new_val) {
6343                                         idev->cnf.addr_gen_mode = new_val;
6344                                         addrconf_dev_config(idev->dev);
6345                                 }
6346                         }
6347                 }
6348
6349                 *((u32 *)ctl->data) = new_val;
6350         }
6351
6352 out:
6353         rtnl_unlock();
6354
6355         return ret;
6356 }
6357
6358 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6359                                          void *buffer, size_t *lenp,
6360                                          loff_t *ppos)
6361 {
6362         int err;
6363         struct in6_addr addr;
6364         char str[IPV6_MAX_STRLEN];
6365         struct ctl_table lctl = *ctl;
6366         struct net *net = ctl->extra2;
6367         struct ipv6_stable_secret *secret = ctl->data;
6368
6369         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6370                 return -EIO;
6371
6372         lctl.maxlen = IPV6_MAX_STRLEN;
6373         lctl.data = str;
6374
6375         if (!rtnl_trylock())
6376                 return restart_syscall();
6377
6378         if (!write && !secret->initialized) {
6379                 err = -EIO;
6380                 goto out;
6381         }
6382
6383         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6384         if (err >= sizeof(str)) {
6385                 err = -EIO;
6386                 goto out;
6387         }
6388
6389         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6390         if (err || !write)
6391                 goto out;
6392
6393         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6394                 err = -EIO;
6395                 goto out;
6396         }
6397
6398         secret->initialized = true;
6399         secret->secret = addr;
6400
6401         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6402                 struct net_device *dev;
6403
6404                 for_each_netdev(net, dev) {
6405                         struct inet6_dev *idev = __in6_dev_get(dev);
6406
6407                         if (idev) {
6408                                 idev->cnf.addr_gen_mode =
6409                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6410                         }
6411                 }
6412         } else {
6413                 struct inet6_dev *idev = ctl->extra1;
6414
6415                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6416         }
6417
6418 out:
6419         rtnl_unlock();
6420
6421         return err;
6422 }
6423
6424 static
6425 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6426                                                 int write, void *buffer,
6427                                                 size_t *lenp,
6428                                                 loff_t *ppos)
6429 {
6430         int *valp = ctl->data;
6431         int val = *valp;
6432         loff_t pos = *ppos;
6433         struct ctl_table lctl;
6434         int ret;
6435
6436         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6437          * we should not modify it until we get the rtnl lock.
6438          */
6439         lctl = *ctl;
6440         lctl.data = &val;
6441
6442         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6443
6444         if (write)
6445                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6446         if (ret)
6447                 *ppos = pos;
6448         return ret;
6449 }
6450
6451 static
6452 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6453 {
6454         if (rt) {
6455                 if (action)
6456                         rt->dst.flags |= DST_NOPOLICY;
6457                 else
6458                         rt->dst.flags &= ~DST_NOPOLICY;
6459         }
6460 }
6461
6462 static
6463 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6464 {
6465         struct inet6_ifaddr *ifa;
6466
6467         read_lock_bh(&idev->lock);
6468         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6469                 spin_lock(&ifa->lock);
6470                 if (ifa->rt) {
6471                         /* host routes only use builtin fib6_nh */
6472                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6473                         int cpu;
6474
6475                         rcu_read_lock();
6476                         ifa->rt->dst_nopolicy = val ? true : false;
6477                         if (nh->rt6i_pcpu) {
6478                                 for_each_possible_cpu(cpu) {
6479                                         struct rt6_info **rtp;
6480
6481                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6482                                         addrconf_set_nopolicy(*rtp, val);
6483                                 }
6484                         }
6485                         rcu_read_unlock();
6486                 }
6487                 spin_unlock(&ifa->lock);
6488         }
6489         read_unlock_bh(&idev->lock);
6490 }
6491
6492 static
6493 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6494 {
6495         struct inet6_dev *idev;
6496         struct net *net;
6497
6498         if (!rtnl_trylock())
6499                 return restart_syscall();
6500
6501         *valp = val;
6502
6503         net = (struct net *)ctl->extra2;
6504         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6505                 rtnl_unlock();
6506                 return 0;
6507         }
6508
6509         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6510                 struct net_device *dev;
6511
6512                 for_each_netdev(net, dev) {
6513                         idev = __in6_dev_get(dev);
6514                         if (idev)
6515                                 addrconf_disable_policy_idev(idev, val);
6516                 }
6517         } else {
6518                 idev = (struct inet6_dev *)ctl->extra1;
6519                 addrconf_disable_policy_idev(idev, val);
6520         }
6521
6522         rtnl_unlock();
6523         return 0;
6524 }
6525
6526 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6527                                    void *buffer, size_t *lenp, loff_t *ppos)
6528 {
6529         int *valp = ctl->data;
6530         int val = *valp;
6531         loff_t pos = *ppos;
6532         struct ctl_table lctl;
6533         int ret;
6534
6535         lctl = *ctl;
6536         lctl.data = &val;
6537         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6538
6539         if (write && (*valp != val))
6540                 ret = addrconf_disable_policy(ctl, valp, val);
6541
6542         if (ret)
6543                 *ppos = pos;
6544
6545         return ret;
6546 }
6547
6548 static int minus_one = -1;
6549 static const int two_five_five = 255;
6550 static u32 ioam6_if_id_max = U16_MAX;
6551
6552 static const struct ctl_table addrconf_sysctl[] = {
6553         {
6554                 .procname       = "forwarding",
6555                 .data           = &ipv6_devconf.forwarding,
6556                 .maxlen         = sizeof(int),
6557                 .mode           = 0644,
6558                 .proc_handler   = addrconf_sysctl_forward,
6559         },
6560         {
6561                 .procname       = "hop_limit",
6562                 .data           = &ipv6_devconf.hop_limit,
6563                 .maxlen         = sizeof(int),
6564                 .mode           = 0644,
6565                 .proc_handler   = proc_dointvec_minmax,
6566                 .extra1         = (void *)SYSCTL_ONE,
6567                 .extra2         = (void *)&two_five_five,
6568         },
6569         {
6570                 .procname       = "mtu",
6571                 .data           = &ipv6_devconf.mtu6,
6572                 .maxlen         = sizeof(int),
6573                 .mode           = 0644,
6574                 .proc_handler   = addrconf_sysctl_mtu,
6575         },
6576         {
6577                 .procname       = "accept_ra",
6578                 .data           = &ipv6_devconf.accept_ra,
6579                 .maxlen         = sizeof(int),
6580                 .mode           = 0644,
6581                 .proc_handler   = proc_dointvec,
6582         },
6583         {
6584                 .procname       = "accept_redirects",
6585                 .data           = &ipv6_devconf.accept_redirects,
6586                 .maxlen         = sizeof(int),
6587                 .mode           = 0644,
6588                 .proc_handler   = proc_dointvec,
6589         },
6590         {
6591                 .procname       = "autoconf",
6592                 .data           = &ipv6_devconf.autoconf,
6593                 .maxlen         = sizeof(int),
6594                 .mode           = 0644,
6595                 .proc_handler   = proc_dointvec,
6596         },
6597         {
6598                 .procname       = "dad_transmits",
6599                 .data           = &ipv6_devconf.dad_transmits,
6600                 .maxlen         = sizeof(int),
6601                 .mode           = 0644,
6602                 .proc_handler   = proc_dointvec,
6603         },
6604         {
6605                 .procname       = "router_solicitations",
6606                 .data           = &ipv6_devconf.rtr_solicits,
6607                 .maxlen         = sizeof(int),
6608                 .mode           = 0644,
6609                 .proc_handler   = proc_dointvec_minmax,
6610                 .extra1         = &minus_one,
6611         },
6612         {
6613                 .procname       = "router_solicitation_interval",
6614                 .data           = &ipv6_devconf.rtr_solicit_interval,
6615                 .maxlen         = sizeof(int),
6616                 .mode           = 0644,
6617                 .proc_handler   = proc_dointvec_jiffies,
6618         },
6619         {
6620                 .procname       = "router_solicitation_max_interval",
6621                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6622                 .maxlen         = sizeof(int),
6623                 .mode           = 0644,
6624                 .proc_handler   = proc_dointvec_jiffies,
6625         },
6626         {
6627                 .procname       = "router_solicitation_delay",
6628                 .data           = &ipv6_devconf.rtr_solicit_delay,
6629                 .maxlen         = sizeof(int),
6630                 .mode           = 0644,
6631                 .proc_handler   = proc_dointvec_jiffies,
6632         },
6633         {
6634                 .procname       = "force_mld_version",
6635                 .data           = &ipv6_devconf.force_mld_version,
6636                 .maxlen         = sizeof(int),
6637                 .mode           = 0644,
6638                 .proc_handler   = proc_dointvec,
6639         },
6640         {
6641                 .procname       = "mldv1_unsolicited_report_interval",
6642                 .data           =
6643                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6644                 .maxlen         = sizeof(int),
6645                 .mode           = 0644,
6646                 .proc_handler   = proc_dointvec_ms_jiffies,
6647         },
6648         {
6649                 .procname       = "mldv2_unsolicited_report_interval",
6650                 .data           =
6651                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6652                 .maxlen         = sizeof(int),
6653                 .mode           = 0644,
6654                 .proc_handler   = proc_dointvec_ms_jiffies,
6655         },
6656         {
6657                 .procname       = "use_tempaddr",
6658                 .data           = &ipv6_devconf.use_tempaddr,
6659                 .maxlen         = sizeof(int),
6660                 .mode           = 0644,
6661                 .proc_handler   = proc_dointvec,
6662         },
6663         {
6664                 .procname       = "temp_valid_lft",
6665                 .data           = &ipv6_devconf.temp_valid_lft,
6666                 .maxlen         = sizeof(int),
6667                 .mode           = 0644,
6668                 .proc_handler   = proc_dointvec,
6669         },
6670         {
6671                 .procname       = "temp_prefered_lft",
6672                 .data           = &ipv6_devconf.temp_prefered_lft,
6673                 .maxlen         = sizeof(int),
6674                 .mode           = 0644,
6675                 .proc_handler   = proc_dointvec,
6676         },
6677         {
6678                 .procname       = "regen_max_retry",
6679                 .data           = &ipv6_devconf.regen_max_retry,
6680                 .maxlen         = sizeof(int),
6681                 .mode           = 0644,
6682                 .proc_handler   = proc_dointvec,
6683         },
6684         {
6685                 .procname       = "max_desync_factor",
6686                 .data           = &ipv6_devconf.max_desync_factor,
6687                 .maxlen         = sizeof(int),
6688                 .mode           = 0644,
6689                 .proc_handler   = proc_dointvec,
6690         },
6691         {
6692                 .procname       = "max_addresses",
6693                 .data           = &ipv6_devconf.max_addresses,
6694                 .maxlen         = sizeof(int),
6695                 .mode           = 0644,
6696                 .proc_handler   = proc_dointvec,
6697         },
6698         {
6699                 .procname       = "accept_ra_defrtr",
6700                 .data           = &ipv6_devconf.accept_ra_defrtr,
6701                 .maxlen         = sizeof(int),
6702                 .mode           = 0644,
6703                 .proc_handler   = proc_dointvec,
6704         },
6705         {
6706                 .procname       = "ra_defrtr_metric",
6707                 .data           = &ipv6_devconf.ra_defrtr_metric,
6708                 .maxlen         = sizeof(u32),
6709                 .mode           = 0644,
6710                 .proc_handler   = proc_douintvec_minmax,
6711                 .extra1         = (void *)SYSCTL_ONE,
6712         },
6713         {
6714                 .procname       = "accept_ra_min_hop_limit",
6715                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6716                 .maxlen         = sizeof(int),
6717                 .mode           = 0644,
6718                 .proc_handler   = proc_dointvec,
6719         },
6720         {
6721                 .procname       = "accept_ra_pinfo",
6722                 .data           = &ipv6_devconf.accept_ra_pinfo,
6723                 .maxlen         = sizeof(int),
6724                 .mode           = 0644,
6725                 .proc_handler   = proc_dointvec,
6726         },
6727 #ifdef CONFIG_IPV6_ROUTER_PREF
6728         {
6729                 .procname       = "accept_ra_rtr_pref",
6730                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6731                 .maxlen         = sizeof(int),
6732                 .mode           = 0644,
6733                 .proc_handler   = proc_dointvec,
6734         },
6735         {
6736                 .procname       = "router_probe_interval",
6737                 .data           = &ipv6_devconf.rtr_probe_interval,
6738                 .maxlen         = sizeof(int),
6739                 .mode           = 0644,
6740                 .proc_handler   = proc_dointvec_jiffies,
6741         },
6742 #ifdef CONFIG_IPV6_ROUTE_INFO
6743         {
6744                 .procname       = "accept_ra_rt_info_min_plen",
6745                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6746                 .maxlen         = sizeof(int),
6747                 .mode           = 0644,
6748                 .proc_handler   = proc_dointvec,
6749         },
6750         {
6751                 .procname       = "accept_ra_rt_info_max_plen",
6752                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6753                 .maxlen         = sizeof(int),
6754                 .mode           = 0644,
6755                 .proc_handler   = proc_dointvec,
6756         },
6757 #endif
6758 #endif
6759         {
6760                 .procname       = "proxy_ndp",
6761                 .data           = &ipv6_devconf.proxy_ndp,
6762                 .maxlen         = sizeof(int),
6763                 .mode           = 0644,
6764                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6765         },
6766         {
6767                 .procname       = "accept_source_route",
6768                 .data           = &ipv6_devconf.accept_source_route,
6769                 .maxlen         = sizeof(int),
6770                 .mode           = 0644,
6771                 .proc_handler   = proc_dointvec,
6772         },
6773 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6774         {
6775                 .procname       = "optimistic_dad",
6776                 .data           = &ipv6_devconf.optimistic_dad,
6777                 .maxlen         = sizeof(int),
6778                 .mode           = 0644,
6779                 .proc_handler   = proc_dointvec,
6780         },
6781         {
6782                 .procname       = "use_optimistic",
6783                 .data           = &ipv6_devconf.use_optimistic,
6784                 .maxlen         = sizeof(int),
6785                 .mode           = 0644,
6786                 .proc_handler   = proc_dointvec,
6787         },
6788 #endif
6789 #ifdef CONFIG_IPV6_MROUTE
6790         {
6791                 .procname       = "mc_forwarding",
6792                 .data           = &ipv6_devconf.mc_forwarding,
6793                 .maxlen         = sizeof(int),
6794                 .mode           = 0444,
6795                 .proc_handler   = proc_dointvec,
6796         },
6797 #endif
6798         {
6799                 .procname       = "disable_ipv6",
6800                 .data           = &ipv6_devconf.disable_ipv6,
6801                 .maxlen         = sizeof(int),
6802                 .mode           = 0644,
6803                 .proc_handler   = addrconf_sysctl_disable,
6804         },
6805         {
6806                 .procname       = "accept_dad",
6807                 .data           = &ipv6_devconf.accept_dad,
6808                 .maxlen         = sizeof(int),
6809                 .mode           = 0644,
6810                 .proc_handler   = proc_dointvec,
6811         },
6812         {
6813                 .procname       = "force_tllao",
6814                 .data           = &ipv6_devconf.force_tllao,
6815                 .maxlen         = sizeof(int),
6816                 .mode           = 0644,
6817                 .proc_handler   = proc_dointvec
6818         },
6819         {
6820                 .procname       = "ndisc_notify",
6821                 .data           = &ipv6_devconf.ndisc_notify,
6822                 .maxlen         = sizeof(int),
6823                 .mode           = 0644,
6824                 .proc_handler   = proc_dointvec
6825         },
6826         {
6827                 .procname       = "suppress_frag_ndisc",
6828                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6829                 .maxlen         = sizeof(int),
6830                 .mode           = 0644,
6831                 .proc_handler   = proc_dointvec
6832         },
6833         {
6834                 .procname       = "accept_ra_from_local",
6835                 .data           = &ipv6_devconf.accept_ra_from_local,
6836                 .maxlen         = sizeof(int),
6837                 .mode           = 0644,
6838                 .proc_handler   = proc_dointvec,
6839         },
6840         {
6841                 .procname       = "accept_ra_mtu",
6842                 .data           = &ipv6_devconf.accept_ra_mtu,
6843                 .maxlen         = sizeof(int),
6844                 .mode           = 0644,
6845                 .proc_handler   = proc_dointvec,
6846         },
6847         {
6848                 .procname       = "stable_secret",
6849                 .data           = &ipv6_devconf.stable_secret,
6850                 .maxlen         = IPV6_MAX_STRLEN,
6851                 .mode           = 0600,
6852                 .proc_handler   = addrconf_sysctl_stable_secret,
6853         },
6854         {
6855                 .procname       = "use_oif_addrs_only",
6856                 .data           = &ipv6_devconf.use_oif_addrs_only,
6857                 .maxlen         = sizeof(int),
6858                 .mode           = 0644,
6859                 .proc_handler   = proc_dointvec,
6860         },
6861         {
6862                 .procname       = "ignore_routes_with_linkdown",
6863                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6864                 .maxlen         = sizeof(int),
6865                 .mode           = 0644,
6866                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6867         },
6868         {
6869                 .procname       = "drop_unicast_in_l2_multicast",
6870                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6871                 .maxlen         = sizeof(int),
6872                 .mode           = 0644,
6873                 .proc_handler   = proc_dointvec,
6874         },
6875         {
6876                 .procname       = "drop_unsolicited_na",
6877                 .data           = &ipv6_devconf.drop_unsolicited_na,
6878                 .maxlen         = sizeof(int),
6879                 .mode           = 0644,
6880                 .proc_handler   = proc_dointvec,
6881         },
6882         {
6883                 .procname       = "keep_addr_on_down",
6884                 .data           = &ipv6_devconf.keep_addr_on_down,
6885                 .maxlen         = sizeof(int),
6886                 .mode           = 0644,
6887                 .proc_handler   = proc_dointvec,
6888
6889         },
6890         {
6891                 .procname       = "seg6_enabled",
6892                 .data           = &ipv6_devconf.seg6_enabled,
6893                 .maxlen         = sizeof(int),
6894                 .mode           = 0644,
6895                 .proc_handler   = proc_dointvec,
6896         },
6897 #ifdef CONFIG_IPV6_SEG6_HMAC
6898         {
6899                 .procname       = "seg6_require_hmac",
6900                 .data           = &ipv6_devconf.seg6_require_hmac,
6901                 .maxlen         = sizeof(int),
6902                 .mode           = 0644,
6903                 .proc_handler   = proc_dointvec,
6904         },
6905 #endif
6906         {
6907                 .procname       = "enhanced_dad",
6908                 .data           = &ipv6_devconf.enhanced_dad,
6909                 .maxlen         = sizeof(int),
6910                 .mode           = 0644,
6911                 .proc_handler   = proc_dointvec,
6912         },
6913         {
6914                 .procname       = "addr_gen_mode",
6915                 .data           = &ipv6_devconf.addr_gen_mode,
6916                 .maxlen         = sizeof(int),
6917                 .mode           = 0644,
6918                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6919         },
6920         {
6921                 .procname       = "disable_policy",
6922                 .data           = &ipv6_devconf.disable_policy,
6923                 .maxlen         = sizeof(int),
6924                 .mode           = 0644,
6925                 .proc_handler   = addrconf_sysctl_disable_policy,
6926         },
6927         {
6928                 .procname       = "ndisc_tclass",
6929                 .data           = &ipv6_devconf.ndisc_tclass,
6930                 .maxlen         = sizeof(int),
6931                 .mode           = 0644,
6932                 .proc_handler   = proc_dointvec_minmax,
6933                 .extra1         = (void *)SYSCTL_ZERO,
6934                 .extra2         = (void *)&two_five_five,
6935         },
6936         {
6937                 .procname       = "rpl_seg_enabled",
6938                 .data           = &ipv6_devconf.rpl_seg_enabled,
6939                 .maxlen         = sizeof(int),
6940                 .mode           = 0644,
6941                 .proc_handler   = proc_dointvec,
6942         },
6943         {
6944                 .procname       = "ioam6_enabled",
6945                 .data           = &ipv6_devconf.ioam6_enabled,
6946                 .maxlen         = sizeof(u8),
6947                 .mode           = 0644,
6948                 .proc_handler   = proc_dou8vec_minmax,
6949                 .extra1         = (void *)SYSCTL_ZERO,
6950                 .extra2         = (void *)SYSCTL_ONE,
6951         },
6952         {
6953                 .procname       = "ioam6_id",
6954                 .data           = &ipv6_devconf.ioam6_id,
6955                 .maxlen         = sizeof(u32),
6956                 .mode           = 0644,
6957                 .proc_handler   = proc_douintvec_minmax,
6958                 .extra1         = (void *)SYSCTL_ZERO,
6959                 .extra2         = (void *)&ioam6_if_id_max,
6960         },
6961         {
6962                 .procname       = "ioam6_id_wide",
6963                 .data           = &ipv6_devconf.ioam6_id_wide,
6964                 .maxlen         = sizeof(u32),
6965                 .mode           = 0644,
6966                 .proc_handler   = proc_douintvec,
6967         },
6968         {
6969                 /* sentinel */
6970         }
6971 };
6972
6973 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6974                 struct inet6_dev *idev, struct ipv6_devconf *p)
6975 {
6976         int i, ifindex;
6977         struct ctl_table *table;
6978         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6979
6980         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6981         if (!table)
6982                 goto out;
6983
6984         for (i = 0; table[i].data; i++) {
6985                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6986                 /* If one of these is already set, then it is not safe to
6987                  * overwrite either of them: this makes proc_dointvec_minmax
6988                  * usable.
6989                  */
6990                 if (!table[i].extra1 && !table[i].extra2) {
6991                         table[i].extra1 = idev; /* embedded; no ref */
6992                         table[i].extra2 = net;
6993                 }
6994         }
6995
6996         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6997
6998         p->sysctl_header = register_net_sysctl(net, path, table);
6999         if (!p->sysctl_header)
7000                 goto free;
7001
7002         if (!strcmp(dev_name, "all"))
7003                 ifindex = NETCONFA_IFINDEX_ALL;
7004         else if (!strcmp(dev_name, "default"))
7005                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7006         else
7007                 ifindex = idev->dev->ifindex;
7008         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7009                                      ifindex, p);
7010         return 0;
7011
7012 free:
7013         kfree(table);
7014 out:
7015         return -ENOBUFS;
7016 }
7017
7018 static void __addrconf_sysctl_unregister(struct net *net,
7019                                          struct ipv6_devconf *p, int ifindex)
7020 {
7021         struct ctl_table *table;
7022
7023         if (!p->sysctl_header)
7024                 return;
7025
7026         table = p->sysctl_header->ctl_table_arg;
7027         unregister_net_sysctl_table(p->sysctl_header);
7028         p->sysctl_header = NULL;
7029         kfree(table);
7030
7031         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7032 }
7033
7034 static int addrconf_sysctl_register(struct inet6_dev *idev)
7035 {
7036         int err;
7037
7038         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7039                 return -EINVAL;
7040
7041         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7042                                     &ndisc_ifinfo_sysctl_change);
7043         if (err)
7044                 return err;
7045         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7046                                          idev, &idev->cnf);
7047         if (err)
7048                 neigh_sysctl_unregister(idev->nd_parms);
7049
7050         return err;
7051 }
7052
7053 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7054 {
7055         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7056                                      idev->dev->ifindex);
7057         neigh_sysctl_unregister(idev->nd_parms);
7058 }
7059
7060
7061 #endif
7062
7063 static int __net_init addrconf_init_net(struct net *net)
7064 {
7065         int err = -ENOMEM;
7066         struct ipv6_devconf *all, *dflt;
7067
7068         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7069         if (!all)
7070                 goto err_alloc_all;
7071
7072         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7073         if (!dflt)
7074                 goto err_alloc_dflt;
7075
7076         if (IS_ENABLED(CONFIG_SYSCTL) &&
7077             !net_eq(net, &init_net)) {
7078                 switch (sysctl_devconf_inherit_init_net) {
7079                 case 1:  /* copy from init_net */
7080                         memcpy(all, init_net.ipv6.devconf_all,
7081                                sizeof(ipv6_devconf));
7082                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7083                                sizeof(ipv6_devconf_dflt));
7084                         break;
7085                 case 3: /* copy from the current netns */
7086                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7087                                sizeof(ipv6_devconf));
7088                         memcpy(dflt,
7089                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7090                                sizeof(ipv6_devconf_dflt));
7091                         break;
7092                 case 0:
7093                 case 2:
7094                         /* use compiled values */
7095                         break;
7096                 }
7097         }
7098
7099         /* these will be inherited by all namespaces */
7100         dflt->autoconf = ipv6_defaults.autoconf;
7101         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7102
7103         dflt->stable_secret.initialized = false;
7104         all->stable_secret.initialized = false;
7105
7106         net->ipv6.devconf_all = all;
7107         net->ipv6.devconf_dflt = dflt;
7108
7109 #ifdef CONFIG_SYSCTL
7110         err = __addrconf_sysctl_register(net, "all", NULL, all);
7111         if (err < 0)
7112                 goto err_reg_all;
7113
7114         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7115         if (err < 0)
7116                 goto err_reg_dflt;
7117 #endif
7118         return 0;
7119
7120 #ifdef CONFIG_SYSCTL
7121 err_reg_dflt:
7122         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7123 err_reg_all:
7124         kfree(dflt);
7125 #endif
7126 err_alloc_dflt:
7127         kfree(all);
7128 err_alloc_all:
7129         return err;
7130 }
7131
7132 static void __net_exit addrconf_exit_net(struct net *net)
7133 {
7134 #ifdef CONFIG_SYSCTL
7135         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7136                                      NETCONFA_IFINDEX_DEFAULT);
7137         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7138                                      NETCONFA_IFINDEX_ALL);
7139 #endif
7140         kfree(net->ipv6.devconf_dflt);
7141         kfree(net->ipv6.devconf_all);
7142 }
7143
7144 static struct pernet_operations addrconf_ops = {
7145         .init = addrconf_init_net,
7146         .exit = addrconf_exit_net,
7147 };
7148
7149 static struct rtnl_af_ops inet6_ops __read_mostly = {
7150         .family           = AF_INET6,
7151         .fill_link_af     = inet6_fill_link_af,
7152         .get_link_af_size = inet6_get_link_af_size,
7153         .validate_link_af = inet6_validate_link_af,
7154         .set_link_af      = inet6_set_link_af,
7155 };
7156
7157 /*
7158  *      Init / cleanup code
7159  */
7160
7161 int __init addrconf_init(void)
7162 {
7163         struct inet6_dev *idev;
7164         int i, err;
7165
7166         err = ipv6_addr_label_init();
7167         if (err < 0) {
7168                 pr_crit("%s: cannot initialize default policy table: %d\n",
7169                         __func__, err);
7170                 goto out;
7171         }
7172
7173         err = register_pernet_subsys(&addrconf_ops);
7174         if (err < 0)
7175                 goto out_addrlabel;
7176
7177         addrconf_wq = create_workqueue("ipv6_addrconf");
7178         if (!addrconf_wq) {
7179                 err = -ENOMEM;
7180                 goto out_nowq;
7181         }
7182
7183         /* The addrconf netdev notifier requires that loopback_dev
7184          * has it's ipv6 private information allocated and setup
7185          * before it can bring up and give link-local addresses
7186          * to other devices which are up.
7187          *
7188          * Unfortunately, loopback_dev is not necessarily the first
7189          * entry in the global dev_base list of net devices.  In fact,
7190          * it is likely to be the very last entry on that list.
7191          * So this causes the notifier registry below to try and
7192          * give link-local addresses to all devices besides loopback_dev
7193          * first, then loopback_dev, which cases all the non-loopback_dev
7194          * devices to fail to get a link-local address.
7195          *
7196          * So, as a temporary fix, allocate the ipv6 structure for
7197          * loopback_dev first by hand.
7198          * Longer term, all of the dependencies ipv6 has upon the loopback
7199          * device and it being up should be removed.
7200          */
7201         rtnl_lock();
7202         idev = ipv6_add_dev(init_net.loopback_dev);
7203         rtnl_unlock();
7204         if (IS_ERR(idev)) {
7205                 err = PTR_ERR(idev);
7206                 goto errlo;
7207         }
7208
7209         ip6_route_init_special_entries();
7210
7211         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7212                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7213
7214         register_netdevice_notifier(&ipv6_dev_notf);
7215
7216         addrconf_verify();
7217
7218         rtnl_af_register(&inet6_ops);
7219
7220         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7221                                    NULL, inet6_dump_ifinfo, 0);
7222         if (err < 0)
7223                 goto errout;
7224
7225         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7226                                    inet6_rtm_newaddr, NULL, 0);
7227         if (err < 0)
7228                 goto errout;
7229         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7230                                    inet6_rtm_deladdr, NULL, 0);
7231         if (err < 0)
7232                 goto errout;
7233         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7234                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7235                                    RTNL_FLAG_DOIT_UNLOCKED);
7236         if (err < 0)
7237                 goto errout;
7238         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7239                                    NULL, inet6_dump_ifmcaddr, 0);
7240         if (err < 0)
7241                 goto errout;
7242         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7243                                    NULL, inet6_dump_ifacaddr, 0);
7244         if (err < 0)
7245                 goto errout;
7246         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7247                                    inet6_netconf_get_devconf,
7248                                    inet6_netconf_dump_devconf,
7249                                    RTNL_FLAG_DOIT_UNLOCKED);
7250         if (err < 0)
7251                 goto errout;
7252         err = ipv6_addr_label_rtnl_register();
7253         if (err < 0)
7254                 goto errout;
7255
7256         return 0;
7257 errout:
7258         rtnl_unregister_all(PF_INET6);
7259         rtnl_af_unregister(&inet6_ops);
7260         unregister_netdevice_notifier(&ipv6_dev_notf);
7261 errlo:
7262         destroy_workqueue(addrconf_wq);
7263 out_nowq:
7264         unregister_pernet_subsys(&addrconf_ops);
7265 out_addrlabel:
7266         ipv6_addr_label_cleanup();
7267 out:
7268         return err;
7269 }
7270
7271 void addrconf_cleanup(void)
7272 {
7273         struct net_device *dev;
7274         int i;
7275
7276         unregister_netdevice_notifier(&ipv6_dev_notf);
7277         unregister_pernet_subsys(&addrconf_ops);
7278         ipv6_addr_label_cleanup();
7279
7280         rtnl_af_unregister(&inet6_ops);
7281
7282         rtnl_lock();
7283
7284         /* clean dev list */
7285         for_each_netdev(&init_net, dev) {
7286                 if (__in6_dev_get(dev) == NULL)
7287                         continue;
7288                 addrconf_ifdown(dev, true);
7289         }
7290         addrconf_ifdown(init_net.loopback_dev, true);
7291
7292         /*
7293          *      Check hash table.
7294          */
7295         spin_lock_bh(&addrconf_hash_lock);
7296         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7297                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7298         spin_unlock_bh(&addrconf_hash_lock);
7299         cancel_delayed_work(&addr_chk_work);
7300         rtnl_unlock();
7301
7302         destroy_workqueue(addrconf_wq);
7303 }