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 u32 ioam6_if_id_max = U16_MAX;
 100
 101 static inline u32 cstamp_delta(unsigned long cstamp)
 102 {
 103         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 104 }
 105
 106 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 107 {
 108         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 109         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 110         do_div(tmp, 1000000);
 111         return (s32)tmp;
 112 }
 113
 114 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 115 {
 116         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 117         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 118         do_div(tmp, 1000000);
 119         if ((s32)tmp > mrt) {
 120                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 121                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 122                 do_div(tmp, 1000000);
 123         }
 124         return (s32)tmp;
 125 }
 126
 127 #ifdef CONFIG_SYSCTL
 128 static int addrconf_sysctl_register(struct inet6_dev *idev);
 129 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 130 #else
 131 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 132 {
 133         return 0;
 134 }
 135
 136 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 137 {
 138 }
 139 #endif
 140
 141 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
 142
 143 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 144 static int ipv6_count_addresses(const struct inet6_dev *idev);
 145 static int ipv6_generate_stable_address(struct in6_addr *addr,
 146                                         u8 dad_count,
 147                                         const struct inet6_dev *idev);
 148
 149 #define IN6_ADDR_HSIZE_SHIFT    8
 150 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 151 /*
 152  *      Configured unicast address hash table
 153  */
 154 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 155 static DEFINE_SPINLOCK(addrconf_hash_lock);
 156
 157 static void addrconf_verify(void);
 158 static void addrconf_verify_rtnl(void);
 159 static void addrconf_verify_work(struct work_struct *);
 160
 161 static struct workqueue_struct *addrconf_wq;
 162 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 163
 164 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 165 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 166
 167 static void addrconf_type_change(struct net_device *dev,
 168                                  unsigned long event);
 169 static int addrconf_ifdown(struct net_device *dev, bool unregister);
 170
 171 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 172                                                   int plen,
 173                                                   const struct net_device *dev,
 174                                                   u32 flags, u32 noflags,
 175                                                   bool no_gw);
 176
 177 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 178 static void addrconf_dad_work(struct work_struct *w);
 179 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 180                                    bool send_na);
 181 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 182 static void addrconf_rs_timer(struct timer_list *t);
 183 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 184 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 185
 186 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 187                                 struct prefix_info *pinfo);
 188
 189 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 190         .forwarding             = 0,
 191         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 192         .mtu6                   = IPV6_MIN_MTU,
 193         .accept_ra              = 1,
 194         .accept_redirects       = 1,
 195         .autoconf               = 1,
 196         .force_mld_version      = 0,
 197         .mldv1_unsolicited_report_interval = 10 * HZ,
 198         .mldv2_unsolicited_report_interval = HZ,
 199         .dad_transmits          = 1,
 200         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 201         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 202         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 203         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 204         .use_tempaddr           = 0,
 205         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 206         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 207         .regen_max_retry        = REGEN_MAX_RETRY,
 208         .max_desync_factor      = MAX_DESYNC_FACTOR,
 209         .max_addresses          = IPV6_MAX_ADDRESSES,
 210         .accept_ra_defrtr       = 1,
 211         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 212         .accept_ra_from_local   = 0,
 213         .accept_ra_min_hop_limit= 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_min_plen = 0,
 220         .accept_ra_rt_info_max_plen = 0,
 221 #endif
 222 #endif
 223         .proxy_ndp              = 0,
 224         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 225         .disable_ipv6           = 0,
 226         .accept_dad             = 0,
 227         .suppress_frag_ndisc    = 1,
 228         .accept_ra_mtu          = 1,
 229         .stable_secret          = {
 230                 .initialized = false,
 231         },
 232         .use_oif_addrs_only     = 0,
 233         .ignore_routes_with_linkdown = 0,
 234         .keep_addr_on_down      = 0,
 235         .seg6_enabled           = 0,
 236 #ifdef CONFIG_IPV6_SEG6_HMAC
 237         .seg6_require_hmac      = 0,
 238 #endif
 239         .enhanced_dad           = 1,
 240         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 241         .disable_policy         = 0,
 242         .rpl_seg_enabled        = 0,
 243         .ioam6_enabled          = 0,
 244         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 245         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 246 };
 247
 248 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 249         .forwarding             = 0,
 250         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 251         .mtu6                   = IPV6_MIN_MTU,
 252         .accept_ra              = 1,
 253         .accept_redirects       = 1,
 254         .autoconf               = 1,
 255         .force_mld_version      = 0,
 256         .mldv1_unsolicited_report_interval = 10 * HZ,
 257         .mldv2_unsolicited_report_interval = HZ,
 258         .dad_transmits          = 1,
 259         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 260         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 261         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 262         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 263         .use_tempaddr           = 0,
 264         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 265         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 266         .regen_max_retry        = REGEN_MAX_RETRY,
 267         .max_desync_factor      = MAX_DESYNC_FACTOR,
 268         .max_addresses          = IPV6_MAX_ADDRESSES,
 269         .accept_ra_defrtr       = 1,
 270         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
 271         .accept_ra_from_local   = 0,
 272         .accept_ra_min_hop_limit= 1,
 273         .accept_ra_pinfo        = 1,
 274 #ifdef CONFIG_IPV6_ROUTER_PREF
 275         .accept_ra_rtr_pref     = 1,
 276         .rtr_probe_interval     = 60 * HZ,
 277 #ifdef CONFIG_IPV6_ROUTE_INFO
 278         .accept_ra_rt_info_min_plen = 0,
 279         .accept_ra_rt_info_max_plen = 0,
 280 #endif
 281 #endif
 282         .proxy_ndp              = 0,
 283         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 284         .disable_ipv6           = 0,
 285         .accept_dad             = 1,
 286         .suppress_frag_ndisc    = 1,
 287         .accept_ra_mtu          = 1,
 288         .stable_secret          = {
 289                 .initialized = false,
 290         },
 291         .use_oif_addrs_only     = 0,
 292         .ignore_routes_with_linkdown = 0,
 293         .keep_addr_on_down      = 0,
 294         .seg6_enabled           = 0,
 295 #ifdef CONFIG_IPV6_SEG6_HMAC
 296         .seg6_require_hmac      = 0,
 297 #endif
 298         .enhanced_dad           = 1,
 299         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 300         .disable_policy         = 0,
 301         .rpl_seg_enabled        = 0,
 302         .ioam6_enabled          = 0,
 303         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
 304         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
 305 };
 306
 307 /* Check if link is ready: is it up and is a valid qdisc available */
 308 static inline bool addrconf_link_ready(const struct net_device *dev)
 309 {
 310         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 311 }
 312
 313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 314 {
 315         if (del_timer(&idev->rs_timer))
 316                 __in6_dev_put(idev);
 317 }
 318
 319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 320 {
 321         if (cancel_delayed_work(&ifp->dad_work))
 322                 __in6_ifa_put(ifp);
 323 }
 324
 325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 326                                   unsigned long when)
 327 {
 328         if (!timer_pending(&idev->rs_timer))
 329                 in6_dev_hold(idev);
 330         mod_timer(&idev->rs_timer, jiffies + when);
 331 }
 332
 333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 334                                    unsigned long delay)
 335 {
 336         in6_ifa_hold(ifp);
 337         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 338                 in6_ifa_put(ifp);
 339 }
 340
 341 static int snmp6_alloc_dev(struct inet6_dev *idev)
 342 {
 343         int i;
 344
 345         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 346         if (!idev->stats.ipv6)
 347                 goto err_ip;
 348
 349         for_each_possible_cpu(i) {
 350                 struct ipstats_mib *addrconf_stats;
 351                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 352                 u64_stats_init(&addrconf_stats->syncp);
 353         }
 354
 355
 356         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 357                                         GFP_KERNEL);
 358         if (!idev->stats.icmpv6dev)
 359                 goto err_icmp;
 360         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 361                                            GFP_KERNEL);
 362         if (!idev->stats.icmpv6msgdev)
 363                 goto err_icmpmsg;
 364
 365         return 0;
 366
 367 err_icmpmsg:
 368         kfree(idev->stats.icmpv6dev);
 369 err_icmp:
 370         free_percpu(idev->stats.ipv6);
 371 err_ip:
 372         return -ENOMEM;
 373 }
 374
 375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 376 {
 377         struct inet6_dev *ndev;
 378         int err = -ENOMEM;
 379
 380         ASSERT_RTNL();
 381
 382         if (dev->mtu < IPV6_MIN_MTU)
 383                 return ERR_PTR(-EINVAL);
 384
 385         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 386         if (!ndev)
 387                 return ERR_PTR(err);
 388
 389         rwlock_init(&ndev->lock);
 390         ndev->dev = dev;
 391         INIT_LIST_HEAD(&ndev->addr_list);
 392         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 393         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 394
 395         if (ndev->cnf.stable_secret.initialized)
 396                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 397
 398         ndev->cnf.mtu6 = dev->mtu;
 399         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 400         if (!ndev->nd_parms) {
 401                 kfree(ndev);
 402                 return ERR_PTR(err);
 403         }
 404         if (ndev->cnf.forwarding)
 405                 dev_disable_lro(dev);
 406         /* We refer to the device */
 407         dev_hold(dev);
 408
 409         if (snmp6_alloc_dev(ndev) < 0) {
 410                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 411                            __func__);
 412                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 413                 dev_put(dev);
 414                 kfree(ndev);
 415                 return ERR_PTR(err);
 416         }
 417
 418         if (snmp6_register_dev(ndev) < 0) {
 419                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 420                            __func__, dev->name);
 421                 goto err_release;
 422         }
 423
 424         /* One reference from device. */
 425         refcount_set(&ndev->refcnt, 1);
 426
 427         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 428                 ndev->cnf.accept_dad = -1;
 429
 430 #if IS_ENABLED(CONFIG_IPV6_SIT)
 431         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 432                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 433                 ndev->cnf.rtr_solicits = 0;
 434         }
 435 #endif
 436
 437         INIT_LIST_HEAD(&ndev->tempaddr_list);
 438         ndev->desync_factor = U32_MAX;
 439         if ((dev->flags&IFF_LOOPBACK) ||
 440             dev->type == ARPHRD_TUNNEL ||
 441             dev->type == ARPHRD_TUNNEL6 ||
 442             dev->type == ARPHRD_SIT ||
 443             dev->type == ARPHRD_NONE) {
 444                 ndev->cnf.use_tempaddr = -1;
 445         }
 446
 447         ndev->token = in6addr_any;
 448
 449         if (netif_running(dev) && addrconf_link_ready(dev))
 450                 ndev->if_flags |= IF_READY;
 451
 452         ipv6_mc_init_dev(ndev);
 453         ndev->tstamp = jiffies;
 454         err = addrconf_sysctl_register(ndev);
 455         if (err) {
 456                 ipv6_mc_destroy_dev(ndev);
 457                 snmp6_unregister_dev(ndev);
 458                 goto err_release;
 459         }
 460         /* protected by rtnl_lock */
 461         rcu_assign_pointer(dev->ip6_ptr, ndev);
 462
 463         /* Join interface-local all-node multicast group */
 464         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 465
 466         /* Join all-node multicast group */
 467         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 468
 469         /* Join all-router multicast group if forwarding is set */
 470         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 471                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 472
 473         return ndev;
 474
 475 err_release:
 476         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 477         ndev->dead = 1;
 478         in6_dev_finish_destroy(ndev);
 479         return ERR_PTR(err);
 480 }
 481
 482 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 483 {
 484         struct inet6_dev *idev;
 485
 486         ASSERT_RTNL();
 487
 488         idev = __in6_dev_get(dev);
 489         if (!idev) {
 490                 idev = ipv6_add_dev(dev);
 491                 if (IS_ERR(idev))
 492                         return idev;
 493         }
 494
 495         if (dev->flags&IFF_UP)
 496                 ipv6_mc_up(idev);
 497         return idev;
 498 }
 499
 500 static int inet6_netconf_msgsize_devconf(int type)
 501 {
 502         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 503                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 504         bool all = false;
 505
 506         if (type == NETCONFA_ALL)
 507                 all = true;
 508
 509         if (all || type == NETCONFA_FORWARDING)
 510                 size += nla_total_size(4);
 511 #ifdef CONFIG_IPV6_MROUTE
 512         if (all || type == NETCONFA_MC_FORWARDING)
 513                 size += nla_total_size(4);
 514 #endif
 515         if (all || type == NETCONFA_PROXY_NEIGH)
 516                 size += nla_total_size(4);
 517
 518         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 519                 size += nla_total_size(4);
 520
 521         return size;
 522 }
 523
 524 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 525                                       struct ipv6_devconf *devconf, u32 portid,
 526                                       u32 seq, int event, unsigned int flags,
 527                                       int type)
 528 {
 529         struct nlmsghdr  *nlh;
 530         struct netconfmsg *ncm;
 531         bool all = false;
 532
 533         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 534                         flags);
 535         if (!nlh)
 536                 return -EMSGSIZE;
 537
 538         if (type == NETCONFA_ALL)
 539                 all = true;
 540
 541         ncm = nlmsg_data(nlh);
 542         ncm->ncm_family = AF_INET6;
 543
 544         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 545                 goto nla_put_failure;
 546
 547         if (!devconf)
 548                 goto out;
 549
 550         if ((all || type == NETCONFA_FORWARDING) &&
 551             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 552                 goto nla_put_failure;
 553 #ifdef CONFIG_IPV6_MROUTE
 554         if ((all || type == NETCONFA_MC_FORWARDING) &&
 555             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 556                         devconf->mc_forwarding) < 0)
 557                 goto nla_put_failure;
 558 #endif
 559         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 560             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 561                 goto nla_put_failure;
 562
 563         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 564             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 565                         devconf->ignore_routes_with_linkdown) < 0)
 566                 goto nla_put_failure;
 567
 568 out:
 569         nlmsg_end(skb, nlh);
 570         return 0;
 571
 572 nla_put_failure:
 573         nlmsg_cancel(skb, nlh);
 574         return -EMSGSIZE;
 575 }
 576
 577 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 578                                   int ifindex, struct ipv6_devconf *devconf)
 579 {
 580         struct sk_buff *skb;
 581         int err = -ENOBUFS;
 582
 583         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 584         if (!skb)
 585                 goto errout;
 586
 587         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 588                                          event, 0, type);
 589         if (err < 0) {
 590                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 591                 WARN_ON(err == -EMSGSIZE);
 592                 kfree_skb(skb);
 593                 goto errout;
 594         }
 595         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 596         return;
 597 errout:
 598         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 599 }
 600
 601 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 602         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 603         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 604         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 605         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 606 };
 607
 608 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 609                                        const struct nlmsghdr *nlh,
 610                                        struct nlattr **tb,
 611                                        struct netlink_ext_ack *extack)
 612 {
 613         int i, err;
 614
 615         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 616                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 617                 return -EINVAL;
 618         }
 619
 620         if (!netlink_strict_get_check(skb))
 621                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 622                                               tb, NETCONFA_MAX,
 623                                               devconf_ipv6_policy, extack);
 624
 625         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 626                                             tb, NETCONFA_MAX,
 627                                             devconf_ipv6_policy, extack);
 628         if (err)
 629                 return err;
 630
 631         for (i = 0; i <= NETCONFA_MAX; i++) {
 632                 if (!tb[i])
 633                         continue;
 634
 635                 switch (i) {
 636                 case NETCONFA_IFINDEX:
 637                         break;
 638                 default:
 639                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 640                         return -EINVAL;
 641                 }
 642         }
 643
 644         return 0;
 645 }
 646
 647 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 648                                      struct nlmsghdr *nlh,
 649                                      struct netlink_ext_ack *extack)
 650 {
 651         struct net *net = sock_net(in_skb->sk);
 652         struct nlattr *tb[NETCONFA_MAX+1];
 653         struct inet6_dev *in6_dev = NULL;
 654         struct net_device *dev = NULL;
 655         struct sk_buff *skb;
 656         struct ipv6_devconf *devconf;
 657         int ifindex;
 658         int err;
 659
 660         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 661         if (err < 0)
 662                 return err;
 663
 664         if (!tb[NETCONFA_IFINDEX])
 665                 return -EINVAL;
 666
 667         err = -EINVAL;
 668         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 669         switch (ifindex) {
 670         case NETCONFA_IFINDEX_ALL:
 671                 devconf = net->ipv6.devconf_all;
 672                 break;
 673         case NETCONFA_IFINDEX_DEFAULT:
 674                 devconf = net->ipv6.devconf_dflt;
 675                 break;
 676         default:
 677                 dev = dev_get_by_index(net, ifindex);
 678                 if (!dev)
 679                         return -EINVAL;
 680                 in6_dev = in6_dev_get(dev);
 681                 if (!in6_dev)
 682                         goto errout;
 683                 devconf = &in6_dev->cnf;
 684                 break;
 685         }
 686
 687         err = -ENOBUFS;
 688         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 689         if (!skb)
 690                 goto errout;
 691
 692         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 693                                          NETLINK_CB(in_skb).portid,
 694                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 695                                          NETCONFA_ALL);
 696         if (err < 0) {
 697                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 698                 WARN_ON(err == -EMSGSIZE);
 699                 kfree_skb(skb);
 700                 goto errout;
 701         }
 702         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 703 errout:
 704         if (in6_dev)
 705                 in6_dev_put(in6_dev);
 706         if (dev)
 707                 dev_put(dev);
 708         return err;
 709 }
 710
 711 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 712                                       struct netlink_callback *cb)
 713 {
 714         const struct nlmsghdr *nlh = cb->nlh;
 715         struct net *net = sock_net(skb->sk);
 716         int h, s_h;
 717         int idx, s_idx;
 718         struct net_device *dev;
 719         struct inet6_dev *idev;
 720         struct hlist_head *head;
 721
 722         if (cb->strict_check) {
 723                 struct netlink_ext_ack *extack = cb->extack;
 724                 struct netconfmsg *ncm;
 725
 726                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 727                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 728                         return -EINVAL;
 729                 }
 730
 731                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 732                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 733                         return -EINVAL;
 734                 }
 735         }
 736
 737         s_h = cb->args[0];
 738         s_idx = idx = cb->args[1];
 739
 740         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 741                 idx = 0;
 742                 head = &net->dev_index_head[h];
 743                 rcu_read_lock();
 744                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 745                           net->dev_base_seq;
 746                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 747                         if (idx < s_idx)
 748                                 goto cont;
 749                         idev = __in6_dev_get(dev);
 750                         if (!idev)
 751                                 goto cont;
 752
 753                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 754                                                        &idev->cnf,
 755                                                        NETLINK_CB(cb->skb).portid,
 756                                                        nlh->nlmsg_seq,
 757                                                        RTM_NEWNETCONF,
 758                                                        NLM_F_MULTI,
 759                                                        NETCONFA_ALL) < 0) {
 760                                 rcu_read_unlock();
 761                                 goto done;
 762                         }
 763                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 764 cont:
 765                         idx++;
 766                 }
 767                 rcu_read_unlock();
 768         }
 769         if (h == NETDEV_HASHENTRIES) {
 770                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 771                                                net->ipv6.devconf_all,
 772                                                NETLINK_CB(cb->skb).portid,
 773                                                nlh->nlmsg_seq,
 774                                                RTM_NEWNETCONF, NLM_F_MULTI,
 775                                                NETCONFA_ALL) < 0)
 776                         goto done;
 777                 else
 778                         h++;
 779         }
 780         if (h == NETDEV_HASHENTRIES + 1) {
 781                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 782                                                net->ipv6.devconf_dflt,
 783                                                NETLINK_CB(cb->skb).portid,
 784                                                nlh->nlmsg_seq,
 785                                                RTM_NEWNETCONF, NLM_F_MULTI,
 786                                                NETCONFA_ALL) < 0)
 787                         goto done;
 788                 else
 789                         h++;
 790         }
 791 done:
 792         cb->args[0] = h;
 793         cb->args[1] = idx;
 794
 795         return skb->len;
 796 }
 797
 798 #ifdef CONFIG_SYSCTL
 799 static void dev_forward_change(struct inet6_dev *idev)
 800 {
 801         struct net_device *dev;
 802         struct inet6_ifaddr *ifa;
 803
 804         if (!idev)
 805                 return;
 806         dev = idev->dev;
 807         if (idev->cnf.forwarding)
 808                 dev_disable_lro(dev);
 809         if (dev->flags & IFF_MULTICAST) {
 810                 if (idev->cnf.forwarding) {
 811                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 812                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 813                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 814                 } else {
 815                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 816                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 817                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 818                 }
 819         }
 820
 821         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 822                 if (ifa->flags&IFA_F_TENTATIVE)
 823                         continue;
 824                 if (idev->cnf.forwarding)
 825                         addrconf_join_anycast(ifa);
 826                 else
 827                         addrconf_leave_anycast(ifa);
 828         }
 829         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 830                                      NETCONFA_FORWARDING,
 831                                      dev->ifindex, &idev->cnf);
 832 }
 833
 834
 835 static void addrconf_forward_change(struct net *net, __s32 newf)
 836 {
 837         struct net_device *dev;
 838         struct inet6_dev *idev;
 839
 840         for_each_netdev(net, dev) {
 841                 idev = __in6_dev_get(dev);
 842                 if (idev) {
 843                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 844                         idev->cnf.forwarding = newf;
 845                         if (changed)
 846                                 dev_forward_change(idev);
 847                 }
 848         }
 849 }
 850
 851 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 852 {
 853         struct net *net;
 854         int old;
 855
 856         if (!rtnl_trylock())
 857                 return restart_syscall();
 858
 859         net = (struct net *)table->extra2;
 860         old = *p;
 861         *p = newf;
 862
 863         if (p == &net->ipv6.devconf_dflt->forwarding) {
 864                 if ((!newf) ^ (!old))
 865                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 866                                                      NETCONFA_FORWARDING,
 867                                                      NETCONFA_IFINDEX_DEFAULT,
 868                                                      net->ipv6.devconf_dflt);
 869                 rtnl_unlock();
 870                 return 0;
 871         }
 872
 873         if (p == &net->ipv6.devconf_all->forwarding) {
 874                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 875
 876                 net->ipv6.devconf_dflt->forwarding = newf;
 877                 if ((!newf) ^ (!old_dflt))
 878                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 879                                                      NETCONFA_FORWARDING,
 880                                                      NETCONFA_IFINDEX_DEFAULT,
 881                                                      net->ipv6.devconf_dflt);
 882
 883                 addrconf_forward_change(net, newf);
 884                 if ((!newf) ^ (!old))
 885                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 886                                                      NETCONFA_FORWARDING,
 887                                                      NETCONFA_IFINDEX_ALL,
 888                                                      net->ipv6.devconf_all);
 889         } else if ((!newf) ^ (!old))
 890                 dev_forward_change((struct inet6_dev *)table->extra1);
 891         rtnl_unlock();
 892
 893         if (newf)
 894                 rt6_purge_dflt_routers(net);
 895         return 1;
 896 }
 897
 898 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 899 {
 900         struct net_device *dev;
 901         struct inet6_dev *idev;
 902
 903         for_each_netdev(net, dev) {
 904                 idev = __in6_dev_get(dev);
 905                 if (idev) {
 906                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 907
 908                         idev->cnf.ignore_routes_with_linkdown = newf;
 909                         if (changed)
 910                                 inet6_netconf_notify_devconf(dev_net(dev),
 911                                                              RTM_NEWNETCONF,
 912                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 913                                                              dev->ifindex,
 914                                                              &idev->cnf);
 915                 }
 916         }
 917 }
 918
 919 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 920 {
 921         struct net *net;
 922         int old;
 923
 924         if (!rtnl_trylock())
 925                 return restart_syscall();
 926
 927         net = (struct net *)table->extra2;
 928         old = *p;
 929         *p = newf;
 930
 931         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 932                 if ((!newf) ^ (!old))
 933                         inet6_netconf_notify_devconf(net,
 934                                                      RTM_NEWNETCONF,
 935                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 936                                                      NETCONFA_IFINDEX_DEFAULT,
 937                                                      net->ipv6.devconf_dflt);
 938                 rtnl_unlock();
 939                 return 0;
 940         }
 941
 942         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 943                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 944                 addrconf_linkdown_change(net, newf);
 945                 if ((!newf) ^ (!old))
 946                         inet6_netconf_notify_devconf(net,
 947                                                      RTM_NEWNETCONF,
 948                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 949                                                      NETCONFA_IFINDEX_ALL,
 950                                                      net->ipv6.devconf_all);
 951         }
 952         rtnl_unlock();
 953
 954         return 1;
 955 }
 956
 957 #endif
 958
 959 /* Nobody refers to this ifaddr, destroy it */
 960 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 961 {
 962         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 963
 964 #ifdef NET_REFCNT_DEBUG
 965         pr_debug("%s\n", __func__);
 966 #endif
 967
 968         in6_dev_put(ifp->idev);
 969
 970         if (cancel_delayed_work(&ifp->dad_work))
 971                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 972                           ifp);
 973
 974         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 975                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 976                 return;
 977         }
 978
 979         kfree_rcu(ifp, rcu);
 980 }
 981
 982 static void
 983 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 984 {
 985         struct list_head *p;
 986         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 987
 988         /*
 989          * Each device address list is sorted in order of scope -
 990          * global before linklocal.
 991          */
 992         list_for_each(p, &idev->addr_list) {
 993                 struct inet6_ifaddr *ifa
 994                         = list_entry(p, struct inet6_ifaddr, if_list);
 995                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 996                         break;
 997         }
 998
 999         list_add_tail_rcu(&ifp->if_list, p);
1000 }
1001
1002 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1003 {
1004         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1005
1006         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1007 }
1008
1009 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1010                                struct net_device *dev, unsigned int hash)
1011 {
1012         struct inet6_ifaddr *ifp;
1013
1014         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1015                 if (!net_eq(dev_net(ifp->idev->dev), net))
1016                         continue;
1017                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1018                         if (!dev || ifp->idev->dev == dev)
1019                                 return true;
1020                 }
1021         }
1022         return false;
1023 }
1024
1025 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1026 {
1027         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1028         int err = 0;
1029
1030         spin_lock(&addrconf_hash_lock);
1031
1032         /* Ignore adding duplicate addresses on an interface */
1033         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1034                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1035                 err = -EEXIST;
1036         } else {
1037                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1038         }
1039
1040         spin_unlock(&addrconf_hash_lock);
1041
1042         return err;
1043 }
1044
1045 /* On success it returns ifp with increased reference count */
1046
1047 static struct inet6_ifaddr *
1048 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1049               bool can_block, struct netlink_ext_ack *extack)
1050 {
1051         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1052         int addr_type = ipv6_addr_type(cfg->pfx);
1053         struct net *net = dev_net(idev->dev);
1054         struct inet6_ifaddr *ifa = NULL;
1055         struct fib6_info *f6i = NULL;
1056         int err = 0;
1057
1058         if (addr_type == IPV6_ADDR_ANY ||
1059             (addr_type & IPV6_ADDR_MULTICAST &&
1060              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1061             (!(idev->dev->flags & IFF_LOOPBACK) &&
1062              !netif_is_l3_master(idev->dev) &&
1063              addr_type & IPV6_ADDR_LOOPBACK))
1064                 return ERR_PTR(-EADDRNOTAVAIL);
1065
1066         if (idev->dead) {
1067                 err = -ENODEV;                  /*XXX*/
1068                 goto out;
1069         }
1070
1071         if (idev->cnf.disable_ipv6) {
1072                 err = -EACCES;
1073                 goto out;
1074         }
1075
1076         /* validator notifier needs to be blocking;
1077          * do not call in atomic context
1078          */
1079         if (can_block) {
1080                 struct in6_validator_info i6vi = {
1081                         .i6vi_addr = *cfg->pfx,
1082                         .i6vi_dev = idev,
1083                         .extack = extack,
1084                 };
1085
1086                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1087                 err = notifier_to_errno(err);
1088                 if (err < 0)
1089                         goto out;
1090         }
1091
1092         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1093         if (!ifa) {
1094                 err = -ENOBUFS;
1095                 goto out;
1096         }
1097
1098         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1099         if (IS_ERR(f6i)) {
1100                 err = PTR_ERR(f6i);
1101                 f6i = NULL;
1102                 goto out;
1103         }
1104
1105         if (net->ipv6.devconf_all->disable_policy ||
1106             idev->cnf.disable_policy)
1107                 f6i->dst_nopolicy = true;
1108
1109         neigh_parms_data_state_setall(idev->nd_parms);
1110
1111         ifa->addr = *cfg->pfx;
1112         if (cfg->peer_pfx)
1113                 ifa->peer_addr = *cfg->peer_pfx;
1114
1115         spin_lock_init(&ifa->lock);
1116         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1117         INIT_HLIST_NODE(&ifa->addr_lst);
1118         ifa->scope = cfg->scope;
1119         ifa->prefix_len = cfg->plen;
1120         ifa->rt_priority = cfg->rt_priority;
1121         ifa->flags = cfg->ifa_flags;
1122         /* No need to add the TENTATIVE flag for addresses with NODAD */
1123         if (!(cfg->ifa_flags & IFA_F_NODAD))
1124                 ifa->flags |= IFA_F_TENTATIVE;
1125         ifa->valid_lft = cfg->valid_lft;
1126         ifa->prefered_lft = cfg->preferred_lft;
1127         ifa->cstamp = ifa->tstamp = jiffies;
1128         ifa->tokenized = false;
1129
1130         ifa->rt = f6i;
1131
1132         ifa->idev = idev;
1133         in6_dev_hold(idev);
1134
1135         /* For caller */
1136         refcount_set(&ifa->refcnt, 1);
1137
1138         rcu_read_lock_bh();
1139
1140         err = ipv6_add_addr_hash(idev->dev, ifa);
1141         if (err < 0) {
1142                 rcu_read_unlock_bh();
1143                 goto out;
1144         }
1145
1146         write_lock(&idev->lock);
1147
1148         /* Add to inet6_dev unicast addr list. */
1149         ipv6_link_dev_addr(idev, ifa);
1150
1151         if (ifa->flags&IFA_F_TEMPORARY) {
1152                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1153                 in6_ifa_hold(ifa);
1154         }
1155
1156         in6_ifa_hold(ifa);
1157         write_unlock(&idev->lock);
1158
1159         rcu_read_unlock_bh();
1160
1161         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1162 out:
1163         if (unlikely(err < 0)) {
1164                 fib6_info_release(f6i);
1165
1166                 if (ifa) {
1167                         if (ifa->idev)
1168                                 in6_dev_put(ifa->idev);
1169                         kfree(ifa);
1170                 }
1171                 ifa = ERR_PTR(err);
1172         }
1173
1174         return ifa;
1175 }
1176
1177 enum cleanup_prefix_rt_t {
1178         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1179         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1180         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1181 };
1182
1183 /*
1184  * Check, whether the prefix for ifp would still need a prefix route
1185  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1186  * constants.
1187  *
1188  * 1) we don't purge prefix if address was not permanent.
1189  *    prefix is managed by its own lifetime.
1190  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1191  * 3) if there are no addresses, delete prefix.
1192  * 4) if there are still other permanent address(es),
1193  *    corresponding prefix is still permanent.
1194  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1195  *    don't purge the prefix, assume user space is managing it.
1196  * 6) otherwise, update prefix lifetime to the
1197  *    longest valid lifetime among the corresponding
1198  *    addresses on the device.
1199  *    Note: subsequent RA will update lifetime.
1200  **/
1201 static enum cleanup_prefix_rt_t
1202 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1203 {
1204         struct inet6_ifaddr *ifa;
1205         struct inet6_dev *idev = ifp->idev;
1206         unsigned long lifetime;
1207         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1208
1209         *expires = jiffies;
1210
1211         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1212                 if (ifa == ifp)
1213                         continue;
1214                 if (ifa->prefix_len != ifp->prefix_len ||
1215                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1216                                        ifp->prefix_len))
1217                         continue;
1218                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1219                         return CLEANUP_PREFIX_RT_NOP;
1220
1221                 action = CLEANUP_PREFIX_RT_EXPIRE;
1222
1223                 spin_lock(&ifa->lock);
1224
1225                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1226                 /*
1227                  * Note: Because this address is
1228                  * not permanent, lifetime <
1229                  * LONG_MAX / HZ here.
1230                  */
1231                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1232                         *expires = ifa->tstamp + lifetime * HZ;
1233                 spin_unlock(&ifa->lock);
1234         }
1235
1236         return action;
1237 }
1238
1239 static void
1240 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1241                      bool del_rt, bool del_peer)
1242 {
1243         struct fib6_info *f6i;
1244
1245         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1246                                         ifp->prefix_len,
1247                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1248         if (f6i) {
1249                 if (del_rt)
1250                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1251                 else {
1252                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1253                                 fib6_set_expires(f6i, expires);
1254                         fib6_info_release(f6i);
1255                 }
1256         }
1257 }
1258
1259
1260 /* This function wants to get referenced ifp and releases it before return */
1261
1262 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1263 {
1264         int state;
1265         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1266         unsigned long expires;
1267
1268         ASSERT_RTNL();
1269
1270         spin_lock_bh(&ifp->lock);
1271         state = ifp->state;
1272         ifp->state = INET6_IFADDR_STATE_DEAD;
1273         spin_unlock_bh(&ifp->lock);
1274
1275         if (state == INET6_IFADDR_STATE_DEAD)
1276                 goto out;
1277
1278         spin_lock_bh(&addrconf_hash_lock);
1279         hlist_del_init_rcu(&ifp->addr_lst);
1280         spin_unlock_bh(&addrconf_hash_lock);
1281
1282         write_lock_bh(&ifp->idev->lock);
1283
1284         if (ifp->flags&IFA_F_TEMPORARY) {
1285                 list_del(&ifp->tmp_list);
1286                 if (ifp->ifpub) {
1287                         in6_ifa_put(ifp->ifpub);
1288                         ifp->ifpub = NULL;
1289                 }
1290                 __in6_ifa_put(ifp);
1291         }
1292
1293         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1294                 action = check_cleanup_prefix_route(ifp, &expires);
1295
1296         list_del_rcu(&ifp->if_list);
1297         __in6_ifa_put(ifp);
1298
1299         write_unlock_bh(&ifp->idev->lock);
1300
1301         addrconf_del_dad_work(ifp);
1302
1303         ipv6_ifa_notify(RTM_DELADDR, ifp);
1304
1305         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1306
1307         if (action != CLEANUP_PREFIX_RT_NOP) {
1308                 cleanup_prefix_route(ifp, expires,
1309                         action == CLEANUP_PREFIX_RT_DEL, false);
1310         }
1311
1312         /* clean up prefsrc entries */
1313         rt6_remove_prefsrc(ifp);
1314 out:
1315         in6_ifa_put(ifp);
1316 }
1317
1318 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1319 {
1320         struct inet6_dev *idev = ifp->idev;
1321         unsigned long tmp_tstamp, age;
1322         unsigned long regen_advance;
1323         unsigned long now = jiffies;
1324         s32 cnf_temp_preferred_lft;
1325         struct inet6_ifaddr *ift;
1326         struct ifa6_config cfg;
1327         long max_desync_factor;
1328         struct in6_addr addr;
1329         int ret = 0;
1330
1331         write_lock_bh(&idev->lock);
1332
1333 retry:
1334         in6_dev_hold(idev);
1335         if (idev->cnf.use_tempaddr <= 0) {
1336                 write_unlock_bh(&idev->lock);
1337                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1338                 in6_dev_put(idev);
1339                 ret = -1;
1340                 goto out;
1341         }
1342         spin_lock_bh(&ifp->lock);
1343         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1344                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1345                 spin_unlock_bh(&ifp->lock);
1346                 write_unlock_bh(&idev->lock);
1347                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1348                         __func__);
1349                 in6_dev_put(idev);
1350                 ret = -1;
1351                 goto out;
1352         }
1353         in6_ifa_hold(ifp);
1354         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1355         ipv6_gen_rnd_iid(&addr);
1356
1357         age = (now - ifp->tstamp) / HZ;
1358
1359         regen_advance = idev->cnf.regen_max_retry *
1360                         idev->cnf.dad_transmits *
1361                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1362
1363         /* recalculate max_desync_factor each time and update
1364          * idev->desync_factor if it's larger
1365          */
1366         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1367         max_desync_factor = min_t(__u32,
1368                                   idev->cnf.max_desync_factor,
1369                                   cnf_temp_preferred_lft - regen_advance);
1370
1371         if (unlikely(idev->desync_factor > max_desync_factor)) {
1372                 if (max_desync_factor > 0) {
1373                         get_random_bytes(&idev->desync_factor,
1374                                          sizeof(idev->desync_factor));
1375                         idev->desync_factor %= max_desync_factor;
1376                 } else {
1377                         idev->desync_factor = 0;
1378                 }
1379         }
1380
1381         memset(&cfg, 0, sizeof(cfg));
1382         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1383                               idev->cnf.temp_valid_lft + age);
1384         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1385         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1386
1387         cfg.plen = ifp->prefix_len;
1388         tmp_tstamp = ifp->tstamp;
1389         spin_unlock_bh(&ifp->lock);
1390
1391         write_unlock_bh(&idev->lock);
1392
1393         /* A temporary address is created only if this calculated Preferred
1394          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1395          * an implementation must not create a temporary address with a zero
1396          * Preferred Lifetime.
1397          * Use age calculation as in addrconf_verify to avoid unnecessary
1398          * temporary addresses being generated.
1399          */
1400         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1401         if (cfg.preferred_lft <= regen_advance + age) {
1402                 in6_ifa_put(ifp);
1403                 in6_dev_put(idev);
1404                 ret = -1;
1405                 goto out;
1406         }
1407
1408         cfg.ifa_flags = IFA_F_TEMPORARY;
1409         /* set in addrconf_prefix_rcv() */
1410         if (ifp->flags & IFA_F_OPTIMISTIC)
1411                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1412
1413         cfg.pfx = &addr;
1414         cfg.scope = ipv6_addr_scope(cfg.pfx);
1415
1416         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1417         if (IS_ERR(ift)) {
1418                 in6_ifa_put(ifp);
1419                 in6_dev_put(idev);
1420                 pr_info("%s: retry temporary address regeneration\n", __func__);
1421                 write_lock_bh(&idev->lock);
1422                 goto retry;
1423         }
1424
1425         spin_lock_bh(&ift->lock);
1426         ift->ifpub = ifp;
1427         ift->cstamp = now;
1428         ift->tstamp = tmp_tstamp;
1429         spin_unlock_bh(&ift->lock);
1430
1431         addrconf_dad_start(ift);
1432         in6_ifa_put(ift);
1433         in6_dev_put(idev);
1434 out:
1435         return ret;
1436 }
1437
1438 /*
1439  *      Choose an appropriate source address (RFC3484)
1440  */
1441 enum {
1442         IPV6_SADDR_RULE_INIT = 0,
1443         IPV6_SADDR_RULE_LOCAL,
1444         IPV6_SADDR_RULE_SCOPE,
1445         IPV6_SADDR_RULE_PREFERRED,
1446 #ifdef CONFIG_IPV6_MIP6
1447         IPV6_SADDR_RULE_HOA,
1448 #endif
1449         IPV6_SADDR_RULE_OIF,
1450         IPV6_SADDR_RULE_LABEL,
1451         IPV6_SADDR_RULE_PRIVACY,
1452         IPV6_SADDR_RULE_ORCHID,
1453         IPV6_SADDR_RULE_PREFIX,
1454 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1455         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1456 #endif
1457         IPV6_SADDR_RULE_MAX
1458 };
1459
1460 struct ipv6_saddr_score {
1461         int                     rule;
1462         int                     addr_type;
1463         struct inet6_ifaddr     *ifa;
1464         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1465         int                     scopedist;
1466         int                     matchlen;
1467 };
1468
1469 struct ipv6_saddr_dst {
1470         const struct in6_addr *addr;
1471         int ifindex;
1472         int scope;
1473         int label;
1474         unsigned int prefs;
1475 };
1476
1477 static inline int ipv6_saddr_preferred(int type)
1478 {
1479         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1480                 return 1;
1481         return 0;
1482 }
1483
1484 static bool ipv6_use_optimistic_addr(struct net *net,
1485                                      struct inet6_dev *idev)
1486 {
1487 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1488         if (!idev)
1489                 return false;
1490         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1491                 return false;
1492         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1493                 return false;
1494
1495         return true;
1496 #else
1497         return false;
1498 #endif
1499 }
1500
1501 static bool ipv6_allow_optimistic_dad(struct net *net,
1502                                       struct inet6_dev *idev)
1503 {
1504 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1505         if (!idev)
1506                 return false;
1507         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1508                 return false;
1509
1510         return true;
1511 #else
1512         return false;
1513 #endif
1514 }
1515
1516 static int ipv6_get_saddr_eval(struct net *net,
1517                                struct ipv6_saddr_score *score,
1518                                struct ipv6_saddr_dst *dst,
1519                                int i)
1520 {
1521         int ret;
1522
1523         if (i <= score->rule) {
1524                 switch (i) {
1525                 case IPV6_SADDR_RULE_SCOPE:
1526                         ret = score->scopedist;
1527                         break;
1528                 case IPV6_SADDR_RULE_PREFIX:
1529                         ret = score->matchlen;
1530                         break;
1531                 default:
1532                         ret = !!test_bit(i, score->scorebits);
1533                 }
1534                 goto out;
1535         }
1536
1537         switch (i) {
1538         case IPV6_SADDR_RULE_INIT:
1539                 /* Rule 0: remember if hiscore is not ready yet */
1540                 ret = !!score->ifa;
1541                 break;
1542         case IPV6_SADDR_RULE_LOCAL:
1543                 /* Rule 1: Prefer same address */
1544                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1545                 break;
1546         case IPV6_SADDR_RULE_SCOPE:
1547                 /* Rule 2: Prefer appropriate scope
1548                  *
1549                  *      ret
1550                  *       ^
1551                  *    -1 |  d 15
1552                  *    ---+--+-+---> scope
1553                  *       |
1554                  *       |             d is scope of the destination.
1555                  *  B-d  |  \
1556                  *       |   \      <- smaller scope is better if
1557                  *  B-15 |    \        if scope is enough for destination.
1558                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1559                  * d-C-1 | /
1560                  *       |/         <- greater is better
1561                  *   -C  /             if scope is not enough for destination.
1562                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1563                  *
1564                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1565                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1566                  * Assume B = 0 and we get C > 29.
1567                  */
1568                 ret = __ipv6_addr_src_scope(score->addr_type);
1569                 if (ret >= dst->scope)
1570                         ret = -ret;
1571                 else
1572                         ret -= 128;     /* 30 is enough */
1573                 score->scopedist = ret;
1574                 break;
1575         case IPV6_SADDR_RULE_PREFERRED:
1576             {
1577                 /* Rule 3: Avoid deprecated and optimistic addresses */
1578                 u8 avoid = IFA_F_DEPRECATED;
1579
1580                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1581                         avoid |= IFA_F_OPTIMISTIC;
1582                 ret = ipv6_saddr_preferred(score->addr_type) ||
1583                       !(score->ifa->flags & avoid);
1584                 break;
1585             }
1586 #ifdef CONFIG_IPV6_MIP6
1587         case IPV6_SADDR_RULE_HOA:
1588             {
1589                 /* Rule 4: Prefer home address */
1590                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1591                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1592                 break;
1593             }
1594 #endif
1595         case IPV6_SADDR_RULE_OIF:
1596                 /* Rule 5: Prefer outgoing interface */
1597                 ret = (!dst->ifindex ||
1598                        dst->ifindex == score->ifa->idev->dev->ifindex);
1599                 break;
1600         case IPV6_SADDR_RULE_LABEL:
1601                 /* Rule 6: Prefer matching label */
1602                 ret = ipv6_addr_label(net,
1603                                       &score->ifa->addr, score->addr_type,
1604                                       score->ifa->idev->dev->ifindex) == dst->label;
1605                 break;
1606         case IPV6_SADDR_RULE_PRIVACY:
1607             {
1608                 /* Rule 7: Prefer public address
1609                  * Note: prefer temporary address if use_tempaddr >= 2
1610                  */
1611                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1612                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1613                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1614                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1615                 break;
1616             }
1617         case IPV6_SADDR_RULE_ORCHID:
1618                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1619                  *          non-ORCHID vs non-ORCHID
1620                  */
1621                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1622                         ipv6_addr_orchid(dst->addr));
1623                 break;
1624         case IPV6_SADDR_RULE_PREFIX:
1625                 /* Rule 8: Use longest matching prefix */
1626                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1627                 if (ret > score->ifa->prefix_len)
1628                         ret = score->ifa->prefix_len;
1629                 score->matchlen = ret;
1630                 break;
1631 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1632         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1633                 /* Optimistic addresses still have lower precedence than other
1634                  * preferred addresses.
1635                  */
1636                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1637                 break;
1638 #endif
1639         default:
1640                 ret = 0;
1641         }
1642
1643         if (ret)
1644                 __set_bit(i, score->scorebits);
1645         score->rule = i;
1646 out:
1647         return ret;
1648 }
1649
1650 static int __ipv6_dev_get_saddr(struct net *net,
1651                                 struct ipv6_saddr_dst *dst,
1652                                 struct inet6_dev *idev,
1653                                 struct ipv6_saddr_score *scores,
1654                                 int hiscore_idx)
1655 {
1656         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1657
1658         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1659                 int i;
1660
1661                 /*
1662                  * - Tentative Address (RFC2462 section 5.4)
1663                  *  - A tentative address is not considered
1664                  *    "assigned to an interface" in the traditional
1665                  *    sense, unless it is also flagged as optimistic.
1666                  * - Candidate Source Address (section 4)
1667                  *  - In any case, anycast addresses, multicast
1668                  *    addresses, and the unspecified address MUST
1669                  *    NOT be included in a candidate set.
1670                  */
1671                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1672                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1673                         continue;
1674
1675                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1676
1677                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1678                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1679                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1680                                             idev->dev->name);
1681                         continue;
1682                 }
1683
1684                 score->rule = -1;
1685                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1686
1687                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1688                         int minihiscore, miniscore;
1689
1690                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1691                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1692
1693                         if (minihiscore > miniscore) {
1694                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1695                                     score->scopedist > 0) {
1696                                         /*
1697                                          * special case:
1698                                          * each remaining entry
1699                                          * has too small (not enough)
1700                                          * scope, because ifa entries
1701                                          * are sorted by their scope
1702                                          * values.
1703                                          */
1704                                         goto out;
1705                                 }
1706                                 break;
1707                         } else if (minihiscore < miniscore) {
1708                                 swap(hiscore, score);
1709                                 hiscore_idx = 1 - hiscore_idx;
1710
1711                                 /* restore our iterator */
1712                                 score->ifa = hiscore->ifa;
1713
1714                                 break;
1715                         }
1716                 }
1717         }
1718 out:
1719         return hiscore_idx;
1720 }
1721
1722 static int ipv6_get_saddr_master(struct net *net,
1723                                  const struct net_device *dst_dev,
1724                                  const struct net_device *master,
1725                                  struct ipv6_saddr_dst *dst,
1726                                  struct ipv6_saddr_score *scores,
1727                                  int hiscore_idx)
1728 {
1729         struct inet6_dev *idev;
1730
1731         idev = __in6_dev_get(dst_dev);
1732         if (idev)
1733                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1734                                                    scores, hiscore_idx);
1735
1736         idev = __in6_dev_get(master);
1737         if (idev)
1738                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1739                                                    scores, hiscore_idx);
1740
1741         return hiscore_idx;
1742 }
1743
1744 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1745                        const struct in6_addr *daddr, unsigned int prefs,
1746                        struct in6_addr *saddr)
1747 {
1748         struct ipv6_saddr_score scores[2], *hiscore;
1749         struct ipv6_saddr_dst dst;
1750         struct inet6_dev *idev;
1751         struct net_device *dev;
1752         int dst_type;
1753         bool use_oif_addr = false;
1754         int hiscore_idx = 0;
1755         int ret = 0;
1756
1757         dst_type = __ipv6_addr_type(daddr);
1758         dst.addr = daddr;
1759         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1760         dst.scope = __ipv6_addr_src_scope(dst_type);
1761         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1762         dst.prefs = prefs;
1763
1764         scores[hiscore_idx].rule = -1;
1765         scores[hiscore_idx].ifa = NULL;
1766
1767         rcu_read_lock();
1768
1769         /* Candidate Source Address (section 4)
1770          *  - multicast and link-local destination address,
1771          *    the set of candidate source address MUST only
1772          *    include addresses assigned to interfaces
1773          *    belonging to the same link as the outgoing
1774          *    interface.
1775          * (- For site-local destination addresses, the
1776          *    set of candidate source addresses MUST only
1777          *    include addresses assigned to interfaces
1778          *    belonging to the same site as the outgoing
1779          *    interface.)
1780          *  - "It is RECOMMENDED that the candidate source addresses
1781          *    be the set of unicast addresses assigned to the
1782          *    interface that will be used to send to the destination
1783          *    (the 'outgoing' interface)." (RFC 6724)
1784          */
1785         if (dst_dev) {
1786                 idev = __in6_dev_get(dst_dev);
1787                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1788                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1789                     (idev && idev->cnf.use_oif_addrs_only)) {
1790                         use_oif_addr = true;
1791                 }
1792         }
1793
1794         if (use_oif_addr) {
1795                 if (idev)
1796                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1797         } else {
1798                 const struct net_device *master;
1799                 int master_idx = 0;
1800
1801                 /* if dst_dev exists and is enslaved to an L3 device, then
1802                  * prefer addresses from dst_dev and then the master over
1803                  * any other enslaved devices in the L3 domain.
1804                  */
1805                 master = l3mdev_master_dev_rcu(dst_dev);
1806                 if (master) {
1807                         master_idx = master->ifindex;
1808
1809                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1810                                                             master, &dst,
1811                                                             scores, hiscore_idx);
1812
1813                         if (scores[hiscore_idx].ifa)
1814                                 goto out;
1815                 }
1816
1817                 for_each_netdev_rcu(net, dev) {
1818                         /* only consider addresses on devices in the
1819                          * same L3 domain
1820                          */
1821                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1822                                 continue;
1823                         idev = __in6_dev_get(dev);
1824                         if (!idev)
1825                                 continue;
1826                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1827                 }
1828         }
1829
1830 out:
1831         hiscore = &scores[hiscore_idx];
1832         if (!hiscore->ifa)
1833                 ret = -EADDRNOTAVAIL;
1834         else
1835                 *saddr = hiscore->ifa->addr;
1836
1837         rcu_read_unlock();
1838         return ret;
1839 }
1840 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1841
1842 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1843                       u32 banned_flags)
1844 {
1845         struct inet6_ifaddr *ifp;
1846         int err = -EADDRNOTAVAIL;
1847
1848         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1849                 if (ifp->scope > IFA_LINK)
1850                         break;
1851                 if (ifp->scope == IFA_LINK &&
1852                     !(ifp->flags & banned_flags)) {
1853                         *addr = ifp->addr;
1854                         err = 0;
1855                         break;
1856                 }
1857         }
1858         return err;
1859 }
1860
1861 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1862                     u32 banned_flags)
1863 {
1864         struct inet6_dev *idev;
1865         int err = -EADDRNOTAVAIL;
1866
1867         rcu_read_lock();
1868         idev = __in6_dev_get(dev);
1869         if (idev) {
1870                 read_lock_bh(&idev->lock);
1871                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1872                 read_unlock_bh(&idev->lock);
1873         }
1874         rcu_read_unlock();
1875         return err;
1876 }
1877
1878 static int ipv6_count_addresses(const struct inet6_dev *idev)
1879 {
1880         const struct inet6_ifaddr *ifp;
1881         int cnt = 0;
1882
1883         rcu_read_lock();
1884         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1885                 cnt++;
1886         rcu_read_unlock();
1887         return cnt;
1888 }
1889
1890 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1891                   const struct net_device *dev, int strict)
1892 {
1893         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1894                                        strict, IFA_F_TENTATIVE);
1895 }
1896 EXPORT_SYMBOL(ipv6_chk_addr);
1897
1898 /* device argument is used to find the L3 domain of interest. If
1899  * skip_dev_check is set, then the ifp device is not checked against
1900  * the passed in dev argument. So the 2 cases for addresses checks are:
1901  *   1. does the address exist in the L3 domain that dev is part of
1902  *      (skip_dev_check = true), or
1903  *
1904  *   2. does the address exist on the specific device
1905  *      (skip_dev_check = false)
1906  */
1907 static struct net_device *
1908 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1909                           const struct net_device *dev, bool skip_dev_check,
1910                           int strict, u32 banned_flags)
1911 {
1912         unsigned int hash = inet6_addr_hash(net, addr);
1913         struct net_device *l3mdev, *ndev;
1914         struct inet6_ifaddr *ifp;
1915         u32 ifp_flags;
1916
1917         rcu_read_lock();
1918
1919         l3mdev = l3mdev_master_dev_rcu(dev);
1920         if (skip_dev_check)
1921                 dev = NULL;
1922
1923         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1924                 ndev = ifp->idev->dev;
1925                 if (!net_eq(dev_net(ndev), net))
1926                         continue;
1927
1928                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1929                         continue;
1930
1931                 /* Decouple optimistic from tentative for evaluation here.
1932                  * Ban optimistic addresses explicitly, when required.
1933                  */
1934                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1935                             ? (ifp->flags&~IFA_F_TENTATIVE)
1936                             : ifp->flags;
1937                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1938                     !(ifp_flags&banned_flags) &&
1939                     (!dev || ndev == dev ||
1940                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1941                         rcu_read_unlock();
1942                         return ndev;
1943                 }
1944         }
1945
1946         rcu_read_unlock();
1947         return NULL;
1948 }
1949
1950 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1951                             const struct net_device *dev, bool skip_dev_check,
1952                             int strict, u32 banned_flags)
1953 {
1954         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1955                                          strict, banned_flags) ? 1 : 0;
1956 }
1957 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1958
1959
1960 /* Compares an address/prefix_len with addresses on device @dev.
1961  * If one is found it returns true.
1962  */
1963 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1964         const unsigned int prefix_len, struct net_device *dev)
1965 {
1966         const struct inet6_ifaddr *ifa;
1967         const struct inet6_dev *idev;
1968         bool ret = false;
1969
1970         rcu_read_lock();
1971         idev = __in6_dev_get(dev);
1972         if (idev) {
1973                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1974                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1975                         if (ret)
1976                                 break;
1977                 }
1978         }
1979         rcu_read_unlock();
1980
1981         return ret;
1982 }
1983 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1984
1985 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1986 {
1987         const struct inet6_ifaddr *ifa;
1988         const struct inet6_dev *idev;
1989         int     onlink;
1990
1991         onlink = 0;
1992         rcu_read_lock();
1993         idev = __in6_dev_get(dev);
1994         if (idev) {
1995                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1996                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1997                                                    ifa->prefix_len);
1998                         if (onlink)
1999                                 break;
2000                 }
2001         }
2002         rcu_read_unlock();
2003         return onlink;
2004 }
2005 EXPORT_SYMBOL(ipv6_chk_prefix);
2006
2007 /**
2008  * ipv6_dev_find - find the first device with a given source address.
2009  * @net: the net namespace
2010  * @addr: the source address
2011  * @dev: used to find the L3 domain of interest
2012  *
2013  * The caller should be protected by RCU, or RTNL.
2014  */
2015 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2016                                  struct net_device *dev)
2017 {
2018         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2019                                          IFA_F_TENTATIVE);
2020 }
2021 EXPORT_SYMBOL(ipv6_dev_find);
2022
2023 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2024                                      struct net_device *dev, int strict)
2025 {
2026         unsigned int hash = inet6_addr_hash(net, addr);
2027         struct inet6_ifaddr *ifp, *result = NULL;
2028
2029         rcu_read_lock();
2030         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2031                 if (!net_eq(dev_net(ifp->idev->dev), net))
2032                         continue;
2033                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2034                         if (!dev || ifp->idev->dev == dev ||
2035                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2036                                 result = ifp;
2037                                 in6_ifa_hold(ifp);
2038                                 break;
2039                         }
2040                 }
2041         }
2042         rcu_read_unlock();
2043
2044         return result;
2045 }
2046
2047 /* Gets referenced address, destroys ifaddr */
2048
2049 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2050 {
2051         if (dad_failed)
2052                 ifp->flags |= IFA_F_DADFAILED;
2053
2054         if (ifp->flags&IFA_F_TEMPORARY) {
2055                 struct inet6_ifaddr *ifpub;
2056                 spin_lock_bh(&ifp->lock);
2057                 ifpub = ifp->ifpub;
2058                 if (ifpub) {
2059                         in6_ifa_hold(ifpub);
2060                         spin_unlock_bh(&ifp->lock);
2061                         ipv6_create_tempaddr(ifpub, true);
2062                         in6_ifa_put(ifpub);
2063                 } else {
2064                         spin_unlock_bh(&ifp->lock);
2065                 }
2066                 ipv6_del_addr(ifp);
2067         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2068                 spin_lock_bh(&ifp->lock);
2069                 addrconf_del_dad_work(ifp);
2070                 ifp->flags |= IFA_F_TENTATIVE;
2071                 if (dad_failed)
2072                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2073                 spin_unlock_bh(&ifp->lock);
2074                 if (dad_failed)
2075                         ipv6_ifa_notify(0, ifp);
2076                 in6_ifa_put(ifp);
2077         } else {
2078                 ipv6_del_addr(ifp);
2079         }
2080 }
2081
2082 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2083 {
2084         int err = -ENOENT;
2085
2086         spin_lock_bh(&ifp->lock);
2087         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2088                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2089                 err = 0;
2090         }
2091         spin_unlock_bh(&ifp->lock);
2092
2093         return err;
2094 }
2095
2096 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2097 {
2098         struct inet6_dev *idev = ifp->idev;
2099         struct net *net = dev_net(ifp->idev->dev);
2100
2101         if (addrconf_dad_end(ifp)) {
2102                 in6_ifa_put(ifp);
2103                 return;
2104         }
2105
2106         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2107                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2108
2109         spin_lock_bh(&ifp->lock);
2110
2111         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2112                 struct in6_addr new_addr;
2113                 struct inet6_ifaddr *ifp2;
2114                 int retries = ifp->stable_privacy_retry + 1;
2115                 struct ifa6_config cfg = {
2116                         .pfx = &new_addr,
2117                         .plen = ifp->prefix_len,
2118                         .ifa_flags = ifp->flags,
2119                         .valid_lft = ifp->valid_lft,
2120                         .preferred_lft = ifp->prefered_lft,
2121                         .scope = ifp->scope,
2122                 };
2123
2124                 if (retries > net->ipv6.sysctl.idgen_retries) {
2125                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2126                                              ifp->idev->dev->name);
2127                         goto errdad;
2128                 }
2129
2130                 new_addr = ifp->addr;
2131                 if (ipv6_generate_stable_address(&new_addr, retries,
2132                                                  idev))
2133                         goto errdad;
2134
2135                 spin_unlock_bh(&ifp->lock);
2136
2137                 if (idev->cnf.max_addresses &&
2138                     ipv6_count_addresses(idev) >=
2139                     idev->cnf.max_addresses)
2140                         goto lock_errdad;
2141
2142                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2143                                      ifp->idev->dev->name);
2144
2145                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2146                 if (IS_ERR(ifp2))
2147                         goto lock_errdad;
2148
2149                 spin_lock_bh(&ifp2->lock);
2150                 ifp2->stable_privacy_retry = retries;
2151                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2152                 spin_unlock_bh(&ifp2->lock);
2153
2154                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2155                 in6_ifa_put(ifp2);
2156 lock_errdad:
2157                 spin_lock_bh(&ifp->lock);
2158         }
2159
2160 errdad:
2161         /* transition from _POSTDAD to _ERRDAD */
2162         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2163         spin_unlock_bh(&ifp->lock);
2164
2165         addrconf_mod_dad_work(ifp, 0);
2166         in6_ifa_put(ifp);
2167 }
2168
2169 /* Join to solicited addr multicast group.
2170  * caller must hold RTNL */
2171 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2172 {
2173         struct in6_addr maddr;
2174
2175         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2176                 return;
2177
2178         addrconf_addr_solict_mult(addr, &maddr);
2179         ipv6_dev_mc_inc(dev, &maddr);
2180 }
2181
2182 /* caller must hold RTNL */
2183 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2184 {
2185         struct in6_addr maddr;
2186
2187         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2188                 return;
2189
2190         addrconf_addr_solict_mult(addr, &maddr);
2191         __ipv6_dev_mc_dec(idev, &maddr);
2192 }
2193
2194 /* caller must hold RTNL */
2195 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2196 {
2197         struct in6_addr addr;
2198
2199         if (ifp->prefix_len >= 127) /* RFC 6164 */
2200                 return;
2201         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2202         if (ipv6_addr_any(&addr))
2203                 return;
2204         __ipv6_dev_ac_inc(ifp->idev, &addr);
2205 }
2206
2207 /* caller must hold RTNL */
2208 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2209 {
2210         struct in6_addr addr;
2211
2212         if (ifp->prefix_len >= 127) /* RFC 6164 */
2213                 return;
2214         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2215         if (ipv6_addr_any(&addr))
2216                 return;
2217         __ipv6_dev_ac_dec(ifp->idev, &addr);
2218 }
2219
2220 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2221 {
2222         switch (dev->addr_len) {
2223         case ETH_ALEN:
2224                 memcpy(eui, dev->dev_addr, 3);
2225                 eui[3] = 0xFF;
2226                 eui[4] = 0xFE;
2227                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2228                 break;
2229         case EUI64_ADDR_LEN:
2230                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2231                 eui[0] ^= 2;
2232                 break;
2233         default:
2234                 return -1;
2235         }
2236
2237         return 0;
2238 }
2239
2240 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2241 {
2242         union fwnet_hwaddr *ha;
2243
2244         if (dev->addr_len != FWNET_ALEN)
2245                 return -1;
2246
2247         ha = (union fwnet_hwaddr *)dev->dev_addr;
2248
2249         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2250         eui[0] ^= 2;
2251         return 0;
2252 }
2253
2254 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2255 {
2256         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2257         if (dev->addr_len != ARCNET_ALEN)
2258                 return -1;
2259         memset(eui, 0, 7);
2260         eui[7] = *(u8 *)dev->dev_addr;
2261         return 0;
2262 }
2263
2264 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2265 {
2266         if (dev->addr_len != INFINIBAND_ALEN)
2267                 return -1;
2268         memcpy(eui, dev->dev_addr + 12, 8);
2269         eui[0] |= 2;
2270         return 0;
2271 }
2272
2273 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2274 {
2275         if (addr == 0)
2276                 return -1;
2277         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2278                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2279                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2280                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2281                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2282                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2283         eui[1] = 0;
2284         eui[2] = 0x5E;
2285         eui[3] = 0xFE;
2286         memcpy(eui + 4, &addr, 4);
2287         return 0;
2288 }
2289
2290 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2291 {
2292         if (dev->priv_flags & IFF_ISATAP)
2293                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2294         return -1;
2295 }
2296
2297 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2298 {
2299         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2300 }
2301
2302 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2303 {
2304         memcpy(eui, dev->perm_addr, 3);
2305         memcpy(eui + 5, dev->perm_addr + 3, 3);
2306         eui[3] = 0xFF;
2307         eui[4] = 0xFE;
2308         eui[0] ^= 2;
2309         return 0;
2310 }
2311
2312 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2313 {
2314         switch (dev->type) {
2315         case ARPHRD_ETHER:
2316         case ARPHRD_FDDI:
2317                 return addrconf_ifid_eui48(eui, dev);
2318         case ARPHRD_ARCNET:
2319                 return addrconf_ifid_arcnet(eui, dev);
2320         case ARPHRD_INFINIBAND:
2321                 return addrconf_ifid_infiniband(eui, dev);
2322         case ARPHRD_SIT:
2323                 return addrconf_ifid_sit(eui, dev);
2324         case ARPHRD_IPGRE:
2325         case ARPHRD_TUNNEL:
2326                 return addrconf_ifid_gre(eui, dev);
2327         case ARPHRD_6LOWPAN:
2328                 return addrconf_ifid_6lowpan(eui, dev);
2329         case ARPHRD_IEEE1394:
2330                 return addrconf_ifid_ieee1394(eui, dev);
2331         case ARPHRD_TUNNEL6:
2332         case ARPHRD_IP6GRE:
2333         case ARPHRD_RAWIP:
2334                 return addrconf_ifid_ip6tnl(eui, dev);
2335         }
2336         return -1;
2337 }
2338
2339 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2340 {
2341         int err = -1;
2342         struct inet6_ifaddr *ifp;
2343
2344         read_lock_bh(&idev->lock);
2345         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2346                 if (ifp->scope > IFA_LINK)
2347                         break;
2348                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2349                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2350                         err = 0;
2351                         break;
2352                 }
2353         }
2354         read_unlock_bh(&idev->lock);
2355         return err;
2356 }
2357
2358 /* Generation of a randomized Interface Identifier
2359  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2360  */
2361
2362 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2363 {
2364 regen:
2365         get_random_bytes(&addr->s6_addr[8], 8);
2366
2367         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2368          * check if generated address is not inappropriate:
2369          *
2370          * - Reserved IPv6 Interface Identifiers
2371          * - XXX: already assigned to an address on the device
2372          */
2373
2374         /* Subnet-router anycast: 0000:0000:0000:0000 */
2375         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2376                 goto regen;
2377
2378         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2379          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2380          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2381          */
2382         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2383             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2384                 goto regen;
2385
2386         /* Reserved subnet anycast addresses */
2387         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2388             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2389                 goto regen;
2390 }
2391
2392 /*
2393  *      Add prefix route.
2394  */
2395
2396 static void
2397 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2398                       struct net_device *dev, unsigned long expires,
2399                       u32 flags, gfp_t gfp_flags)
2400 {
2401         struct fib6_config cfg = {
2402                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2403                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2404                 .fc_ifindex = dev->ifindex,
2405                 .fc_expires = expires,
2406                 .fc_dst_len = plen,
2407                 .fc_flags = RTF_UP | flags,
2408                 .fc_nlinfo.nl_net = dev_net(dev),
2409                 .fc_protocol = RTPROT_KERNEL,
2410                 .fc_type = RTN_UNICAST,
2411         };
2412
2413         cfg.fc_dst = *pfx;
2414
2415         /* Prevent useless cloning on PtP SIT.
2416            This thing is done here expecting that the whole
2417            class of non-broadcast devices need not cloning.
2418          */
2419 #if IS_ENABLED(CONFIG_IPV6_SIT)
2420         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2421                 cfg.fc_flags |= RTF_NONEXTHOP;
2422 #endif
2423
2424         ip6_route_add(&cfg, gfp_flags, NULL);
2425 }
2426
2427
2428 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2429                                                   int plen,
2430                                                   const struct net_device *dev,
2431                                                   u32 flags, u32 noflags,
2432                                                   bool no_gw)
2433 {
2434         struct fib6_node *fn;
2435         struct fib6_info *rt = NULL;
2436         struct fib6_table *table;
2437         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2438
2439         table = fib6_get_table(dev_net(dev), tb_id);
2440         if (!table)
2441                 return NULL;
2442
2443         rcu_read_lock();
2444         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2445         if (!fn)
2446                 goto out;
2447
2448         for_each_fib6_node_rt_rcu(fn) {
2449                 /* prefix routes only use builtin fib6_nh */
2450                 if (rt->nh)
2451                         continue;
2452
2453                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2454                         continue;
2455                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2456                         continue;
2457                 if ((rt->fib6_flags & flags) != flags)
2458                         continue;
2459                 if ((rt->fib6_flags & noflags) != 0)
2460                         continue;
2461                 if (!fib6_info_hold_safe(rt))
2462                         continue;
2463                 break;
2464         }
2465 out:
2466         rcu_read_unlock();
2467         return rt;
2468 }
2469
2470
2471 /* Create "default" multicast route to the interface */
2472
2473 static void addrconf_add_mroute(struct net_device *dev)
2474 {
2475         struct fib6_config cfg = {
2476                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2477                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2478                 .fc_ifindex = dev->ifindex,
2479                 .fc_dst_len = 8,
2480                 .fc_flags = RTF_UP,
2481                 .fc_type = RTN_MULTICAST,
2482                 .fc_nlinfo.nl_net = dev_net(dev),
2483                 .fc_protocol = RTPROT_KERNEL,
2484         };
2485
2486         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2487
2488         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2489 }
2490
2491 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2492 {
2493         struct inet6_dev *idev;
2494
2495         ASSERT_RTNL();
2496
2497         idev = ipv6_find_idev(dev);
2498         if (IS_ERR(idev))
2499                 return idev;
2500
2501         if (idev->cnf.disable_ipv6)
2502                 return ERR_PTR(-EACCES);
2503
2504         /* Add default multicast route */
2505         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2506                 addrconf_add_mroute(dev);
2507
2508         return idev;
2509 }
2510
2511 static void manage_tempaddrs(struct inet6_dev *idev,
2512                              struct inet6_ifaddr *ifp,
2513                              __u32 valid_lft, __u32 prefered_lft,
2514                              bool create, unsigned long now)
2515 {
2516         u32 flags;
2517         struct inet6_ifaddr *ift;
2518
2519         read_lock_bh(&idev->lock);
2520         /* update all temporary addresses in the list */
2521         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2522                 int age, max_valid, max_prefered;
2523
2524                 if (ifp != ift->ifpub)
2525                         continue;
2526
2527                 /* RFC 4941 section 3.3:
2528                  * If a received option will extend the lifetime of a public
2529                  * address, the lifetimes of temporary addresses should
2530                  * be extended, subject to the overall constraint that no
2531                  * temporary addresses should ever remain "valid" or "preferred"
2532                  * for a time longer than (TEMP_VALID_LIFETIME) or
2533                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2534                  */
2535                 age = (now - ift->cstamp) / HZ;
2536                 max_valid = idev->cnf.temp_valid_lft - age;
2537                 if (max_valid < 0)
2538                         max_valid = 0;
2539
2540                 max_prefered = idev->cnf.temp_prefered_lft -
2541                                idev->desync_factor - age;
2542                 if (max_prefered < 0)
2543                         max_prefered = 0;
2544
2545                 if (valid_lft > max_valid)
2546                         valid_lft = max_valid;
2547
2548                 if (prefered_lft > max_prefered)
2549                         prefered_lft = max_prefered;
2550
2551                 spin_lock(&ift->lock);
2552                 flags = ift->flags;
2553                 ift->valid_lft = valid_lft;
2554                 ift->prefered_lft = prefered_lft;
2555                 ift->tstamp = now;
2556                 if (prefered_lft > 0)
2557                         ift->flags &= ~IFA_F_DEPRECATED;
2558
2559                 spin_unlock(&ift->lock);
2560                 if (!(flags&IFA_F_TENTATIVE))
2561                         ipv6_ifa_notify(0, ift);
2562         }
2563
2564         if ((create || list_empty(&idev->tempaddr_list)) &&
2565             idev->cnf.use_tempaddr > 0) {
2566                 /* When a new public address is created as described
2567                  * in [ADDRCONF], also create a new temporary address.
2568                  * Also create a temporary address if it's enabled but
2569                  * no temporary address currently exists.
2570                  */
2571                 read_unlock_bh(&idev->lock);
2572                 ipv6_create_tempaddr(ifp, false);
2573         } else {
2574                 read_unlock_bh(&idev->lock);
2575         }
2576 }
2577
2578 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2579 {
2580         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2581                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2582 }
2583
2584 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2585                                  const struct prefix_info *pinfo,
2586                                  struct inet6_dev *in6_dev,
2587                                  const struct in6_addr *addr, int addr_type,
2588                                  u32 addr_flags, bool sllao, bool tokenized,
2589                                  __u32 valid_lft, u32 prefered_lft)
2590 {
2591         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2592         int create = 0;
2593
2594         if (!ifp && valid_lft) {
2595                 int max_addresses = in6_dev->cnf.max_addresses;
2596                 struct ifa6_config cfg = {
2597                         .pfx = addr,
2598                         .plen = pinfo->prefix_len,
2599                         .ifa_flags = addr_flags,
2600                         .valid_lft = valid_lft,
2601                         .preferred_lft = prefered_lft,
2602                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2603                 };
2604
2605 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2606                 if ((net->ipv6.devconf_all->optimistic_dad ||
2607                      in6_dev->cnf.optimistic_dad) &&
2608                     !net->ipv6.devconf_all->forwarding && sllao)
2609                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2610 #endif
2611
2612                 /* Do not allow to create too much of autoconfigured
2613                  * addresses; this would be too easy way to crash kernel.
2614                  */
2615                 if (!max_addresses ||
2616                     ipv6_count_addresses(in6_dev) < max_addresses)
2617                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2618
2619                 if (IS_ERR_OR_NULL(ifp))
2620                         return -1;
2621
2622                 create = 1;
2623                 spin_lock_bh(&ifp->lock);
2624                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2625                 ifp->cstamp = jiffies;
2626                 ifp->tokenized = tokenized;
2627                 spin_unlock_bh(&ifp->lock);
2628                 addrconf_dad_start(ifp);
2629         }
2630
2631         if (ifp) {
2632                 u32 flags;
2633                 unsigned long now;
2634                 u32 stored_lft;
2635
2636                 /* Update lifetime (RFC4862 5.5.3 e)
2637                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2638                  * improve the reaction of SLAAC to renumbering events
2639                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2640                  */
2641                 spin_lock_bh(&ifp->lock);
2642                 now = jiffies;
2643                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2644                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2645                 else
2646                         stored_lft = 0;
2647
2648                 if (!create && stored_lft) {
2649                         ifp->valid_lft = valid_lft;
2650                         ifp->prefered_lft = prefered_lft;
2651                         ifp->tstamp = now;
2652                         flags = ifp->flags;
2653                         ifp->flags &= ~IFA_F_DEPRECATED;
2654                         spin_unlock_bh(&ifp->lock);
2655
2656                         if (!(flags&IFA_F_TENTATIVE))
2657                                 ipv6_ifa_notify(0, ifp);
2658                 } else
2659                         spin_unlock_bh(&ifp->lock);
2660
2661                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2662                                  create, now);
2663
2664                 in6_ifa_put(ifp);
2665                 addrconf_verify();
2666         }
2667
2668         return 0;
2669 }
2670 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2671
2672 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2673 {
2674         struct prefix_info *pinfo;
2675         __u32 valid_lft;
2676         __u32 prefered_lft;
2677         int addr_type, err;
2678         u32 addr_flags = 0;
2679         struct inet6_dev *in6_dev;
2680         struct net *net = dev_net(dev);
2681
2682         pinfo = (struct prefix_info *) opt;
2683
2684         if (len < sizeof(struct prefix_info)) {
2685                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2686                 return;
2687         }
2688
2689         /*
2690          *      Validation checks ([ADDRCONF], page 19)
2691          */
2692
2693         addr_type = ipv6_addr_type(&pinfo->prefix);
2694
2695         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2696                 return;
2697
2698         valid_lft = ntohl(pinfo->valid);
2699         prefered_lft = ntohl(pinfo->prefered);
2700
2701         if (prefered_lft > valid_lft) {
2702                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2703                 return;
2704         }
2705
2706         in6_dev = in6_dev_get(dev);
2707
2708         if (!in6_dev) {
2709                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2710                                     dev->name);
2711                 return;
2712         }
2713
2714         /*
2715          *      Two things going on here:
2716          *      1) Add routes for on-link prefixes
2717          *      2) Configure prefixes with the auto flag set
2718          */
2719
2720         if (pinfo->onlink) {
2721                 struct fib6_info *rt;
2722                 unsigned long rt_expires;
2723
2724                 /* Avoid arithmetic overflow. Really, we could
2725                  * save rt_expires in seconds, likely valid_lft,
2726                  * but it would require division in fib gc, that it
2727                  * not good.
2728                  */
2729                 if (HZ > USER_HZ)
2730                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2731                 else
2732                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2733
2734                 if (addrconf_finite_timeout(rt_expires))
2735                         rt_expires *= HZ;
2736
2737                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2738                                                pinfo->prefix_len,
2739                                                dev,
2740                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2741                                                RTF_DEFAULT, true);
2742
2743                 if (rt) {
2744                         /* Autoconf prefix route */
2745                         if (valid_lft == 0) {
2746                                 ip6_del_rt(net, rt, false);
2747                                 rt = NULL;
2748                         } else if (addrconf_finite_timeout(rt_expires)) {
2749                                 /* not infinity */
2750                                 fib6_set_expires(rt, jiffies + rt_expires);
2751                         } else {
2752                                 fib6_clean_expires(rt);
2753                         }
2754                 } else if (valid_lft) {
2755                         clock_t expires = 0;
2756                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2757                         if (addrconf_finite_timeout(rt_expires)) {
2758                                 /* not infinity */
2759                                 flags |= RTF_EXPIRES;
2760                                 expires = jiffies_to_clock_t(rt_expires);
2761                         }
2762                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2763                                               0, dev, expires, flags,
2764                                               GFP_ATOMIC);
2765                 }
2766                 fib6_info_release(rt);
2767         }
2768
2769         /* Try to figure out our local address for this prefix */
2770
2771         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2772                 struct in6_addr addr;
2773                 bool tokenized = false, dev_addr_generated = false;
2774
2775                 if (pinfo->prefix_len == 64) {
2776                         memcpy(&addr, &pinfo->prefix, 8);
2777
2778                         if (!ipv6_addr_any(&in6_dev->token)) {
2779                                 read_lock_bh(&in6_dev->lock);
2780                                 memcpy(addr.s6_addr + 8,
2781                                        in6_dev->token.s6_addr + 8, 8);
2782                                 read_unlock_bh(&in6_dev->lock);
2783                                 tokenized = true;
2784                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2785                                    !ipv6_generate_stable_address(&addr, 0,
2786                                                                  in6_dev)) {
2787                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2788                                 goto ok;
2789                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2790                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2791                                 goto put;
2792                         } else {
2793                                 dev_addr_generated = true;
2794                         }
2795                         goto ok;
2796                 }
2797                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2798                                     pinfo->prefix_len);
2799                 goto put;
2800
2801 ok:
2802                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2803                                                    &addr, addr_type,
2804                                                    addr_flags, sllao,
2805                                                    tokenized, valid_lft,
2806                                                    prefered_lft);
2807                 if (err)
2808                         goto put;
2809
2810                 /* Ignore error case here because previous prefix add addr was
2811                  * successful which will be notified.
2812                  */
2813                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2814                                               addr_type, addr_flags, sllao,
2815                                               tokenized, valid_lft,
2816                                               prefered_lft,
2817                                               dev_addr_generated);
2818         }
2819         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2820 put:
2821         in6_dev_put(in6_dev);
2822 }
2823
2824 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2825                 struct in6_ifreq *ireq)
2826 {
2827         struct ip_tunnel_parm p = { };
2828         int err;
2829
2830         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2831                 return -EADDRNOTAVAIL;
2832
2833         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2834         p.iph.version = 4;
2835         p.iph.ihl = 5;
2836         p.iph.protocol = IPPROTO_IPV6;
2837         p.iph.ttl = 64;
2838
2839         if (!dev->netdev_ops->ndo_tunnel_ctl)
2840                 return -EOPNOTSUPP;
2841         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2842         if (err)
2843                 return err;
2844
2845         dev = __dev_get_by_name(net, p.name);
2846         if (!dev)
2847                 return -ENOBUFS;
2848         return dev_open(dev, NULL);
2849 }
2850
2851 /*
2852  *      Set destination address.
2853  *      Special case for SIT interfaces where we create a new "virtual"
2854  *      device.
2855  */
2856 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2857 {
2858         struct net_device *dev;
2859         struct in6_ifreq ireq;
2860         int err = -ENODEV;
2861
2862         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2863                 return -ENODEV;
2864         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2865                 return -EFAULT;
2866
2867         rtnl_lock();
2868         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2869         if (dev && dev->type == ARPHRD_SIT)
2870                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2871         rtnl_unlock();
2872         return err;
2873 }
2874
2875 static int ipv6_mc_config(struct sock *sk, bool join,
2876                           const struct in6_addr *addr, int ifindex)
2877 {
2878         int ret;
2879
2880         ASSERT_RTNL();
2881
2882         lock_sock(sk);
2883         if (join)
2884                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2885         else
2886                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2887         release_sock(sk);
2888
2889         return ret;
2890 }
2891
2892 /*
2893  *      Manual configuration of address on an interface
2894  */
2895 static int inet6_addr_add(struct net *net, int ifindex,
2896                           struct ifa6_config *cfg,
2897                           struct netlink_ext_ack *extack)
2898 {
2899         struct inet6_ifaddr *ifp;
2900         struct inet6_dev *idev;
2901         struct net_device *dev;
2902         unsigned long timeout;
2903         clock_t expires;
2904         u32 flags;
2905
2906         ASSERT_RTNL();
2907
2908         if (cfg->plen > 128)
2909                 return -EINVAL;
2910
2911         /* check the lifetime */
2912         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2913                 return -EINVAL;
2914
2915         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2916                 return -EINVAL;
2917
2918         dev = __dev_get_by_index(net, ifindex);
2919         if (!dev)
2920                 return -ENODEV;
2921
2922         idev = addrconf_add_dev(dev);
2923         if (IS_ERR(idev))
2924                 return PTR_ERR(idev);
2925
2926         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2927                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2928                                          true, cfg->pfx, ifindex);
2929
2930                 if (ret < 0)
2931                         return ret;
2932         }
2933
2934         cfg->scope = ipv6_addr_scope(cfg->pfx);
2935
2936         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2937         if (addrconf_finite_timeout(timeout)) {
2938                 expires = jiffies_to_clock_t(timeout * HZ);
2939                 cfg->valid_lft = timeout;
2940                 flags = RTF_EXPIRES;
2941         } else {
2942                 expires = 0;
2943                 flags = 0;
2944                 cfg->ifa_flags |= IFA_F_PERMANENT;
2945         }
2946
2947         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2948         if (addrconf_finite_timeout(timeout)) {
2949                 if (timeout == 0)
2950                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2951                 cfg->preferred_lft = timeout;
2952         }
2953
2954         ifp = ipv6_add_addr(idev, cfg, true, extack);
2955         if (!IS_ERR(ifp)) {
2956                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2957                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2958                                               ifp->rt_priority, dev, expires,
2959                                               flags, GFP_KERNEL);
2960                 }
2961
2962                 /* Send a netlink notification if DAD is enabled and
2963                  * optimistic flag is not set
2964                  */
2965                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2966                         ipv6_ifa_notify(0, ifp);
2967                 /*
2968                  * Note that section 3.1 of RFC 4429 indicates
2969                  * that the Optimistic flag should not be set for
2970                  * manually configured addresses
2971                  */
2972                 addrconf_dad_start(ifp);
2973                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2974                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2975                                          cfg->preferred_lft, true, jiffies);
2976                 in6_ifa_put(ifp);
2977                 addrconf_verify_rtnl();
2978                 return 0;
2979         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2980                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2981                                cfg->pfx, ifindex);
2982         }
2983
2984         return PTR_ERR(ifp);
2985 }
2986
2987 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2988                           const struct in6_addr *pfx, unsigned int plen)
2989 {
2990         struct inet6_ifaddr *ifp;
2991         struct inet6_dev *idev;
2992         struct net_device *dev;
2993
2994         if (plen > 128)
2995                 return -EINVAL;
2996
2997         dev = __dev_get_by_index(net, ifindex);
2998         if (!dev)
2999                 return -ENODEV;
3000
3001         idev = __in6_dev_get(dev);
3002         if (!idev)
3003                 return -ENXIO;
3004
3005         read_lock_bh(&idev->lock);
3006         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3007                 if (ifp->prefix_len == plen &&
3008                     ipv6_addr_equal(pfx, &ifp->addr)) {
3009                         in6_ifa_hold(ifp);
3010                         read_unlock_bh(&idev->lock);
3011
3012                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3013                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3014                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3015                                                  jiffies);
3016                         ipv6_del_addr(ifp);
3017                         addrconf_verify_rtnl();
3018                         if (ipv6_addr_is_multicast(pfx)) {
3019                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3020                                                false, pfx, dev->ifindex);
3021                         }
3022                         return 0;
3023                 }
3024         }
3025         read_unlock_bh(&idev->lock);
3026         return -EADDRNOTAVAIL;
3027 }
3028
3029
3030 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3031 {
3032         struct ifa6_config cfg = {
3033                 .ifa_flags = IFA_F_PERMANENT,
3034                 .preferred_lft = INFINITY_LIFE_TIME,
3035                 .valid_lft = INFINITY_LIFE_TIME,
3036         };
3037         struct in6_ifreq ireq;
3038         int err;
3039
3040         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3041                 return -EPERM;
3042
3043         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3044                 return -EFAULT;
3045
3046         cfg.pfx = &ireq.ifr6_addr;
3047         cfg.plen = ireq.ifr6_prefixlen;
3048
3049         rtnl_lock();
3050         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3051         rtnl_unlock();
3052         return err;
3053 }
3054
3055 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3056 {
3057         struct in6_ifreq ireq;
3058         int err;
3059
3060         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3061                 return -EPERM;
3062
3063         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3064                 return -EFAULT;
3065
3066         rtnl_lock();
3067         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3068                              ireq.ifr6_prefixlen);
3069         rtnl_unlock();
3070         return err;
3071 }
3072
3073 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3074                      int plen, int scope)
3075 {
3076         struct inet6_ifaddr *ifp;
3077         struct ifa6_config cfg = {
3078                 .pfx = addr,
3079                 .plen = plen,
3080                 .ifa_flags = IFA_F_PERMANENT,
3081                 .valid_lft = INFINITY_LIFE_TIME,
3082                 .preferred_lft = INFINITY_LIFE_TIME,
3083                 .scope = scope
3084         };
3085
3086         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3087         if (!IS_ERR(ifp)) {
3088                 spin_lock_bh(&ifp->lock);
3089                 ifp->flags &= ~IFA_F_TENTATIVE;
3090                 spin_unlock_bh(&ifp->lock);
3091                 rt_genid_bump_ipv6(dev_net(idev->dev));
3092                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3093                 in6_ifa_put(ifp);
3094         }
3095 }
3096
3097 #if IS_ENABLED(CONFIG_IPV6_SIT)
3098 static void sit_add_v4_addrs(struct inet6_dev *idev)
3099 {
3100         struct in6_addr addr;
3101         struct net_device *dev;
3102         struct net *net = dev_net(idev->dev);
3103         int scope, plen;
3104         u32 pflags = 0;
3105
3106         ASSERT_RTNL();
3107
3108         memset(&addr, 0, sizeof(struct in6_addr));
3109         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3110
3111         if (idev->dev->flags&IFF_POINTOPOINT) {
3112                 addr.s6_addr32[0] = htonl(0xfe800000);
3113                 scope = IFA_LINK;
3114                 plen = 64;
3115         } else {
3116                 scope = IPV6_ADDR_COMPATv4;
3117                 plen = 96;
3118                 pflags |= RTF_NONEXTHOP;
3119         }
3120
3121         if (addr.s6_addr32[3]) {
3122                 add_addr(idev, &addr, plen, scope);
3123                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3124                                       GFP_KERNEL);
3125                 return;
3126         }
3127
3128         for_each_netdev(net, dev) {
3129                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3130                 if (in_dev && (dev->flags & IFF_UP)) {
3131                         struct in_ifaddr *ifa;
3132                         int flag = scope;
3133
3134                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3135                                 addr.s6_addr32[3] = ifa->ifa_local;
3136
3137                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3138                                         continue;
3139                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3140                                         if (idev->dev->flags&IFF_POINTOPOINT)
3141                                                 continue;
3142                                         flag |= IFA_HOST;
3143                                 }
3144
3145                                 add_addr(idev, &addr, plen, flag);
3146                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3147                                                       0, pflags, GFP_KERNEL);
3148                         }
3149                 }
3150         }
3151 }
3152 #endif
3153
3154 static void init_loopback(struct net_device *dev)
3155 {
3156         struct inet6_dev  *idev;
3157
3158         /* ::1 */
3159
3160         ASSERT_RTNL();
3161
3162         idev = ipv6_find_idev(dev);
3163         if (IS_ERR(idev)) {
3164                 pr_debug("%s: add_dev failed\n", __func__);
3165                 return;
3166         }
3167
3168         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3169 }
3170
3171 void addrconf_add_linklocal(struct inet6_dev *idev,
3172                             const struct in6_addr *addr, u32 flags)
3173 {
3174         struct ifa6_config cfg = {
3175                 .pfx = addr,
3176                 .plen = 64,
3177                 .ifa_flags = flags | IFA_F_PERMANENT,
3178                 .valid_lft = INFINITY_LIFE_TIME,
3179                 .preferred_lft = INFINITY_LIFE_TIME,
3180                 .scope = IFA_LINK
3181         };
3182         struct inet6_ifaddr *ifp;
3183
3184 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3185         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3186              idev->cnf.optimistic_dad) &&
3187             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3188                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3189 #endif
3190
3191         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3192         if (!IS_ERR(ifp)) {
3193                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3194                                       0, 0, GFP_ATOMIC);
3195                 addrconf_dad_start(ifp);
3196                 in6_ifa_put(ifp);
3197         }
3198 }
3199 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3200
3201 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3202 {
3203         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3204                 return true;
3205
3206         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3207             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3208                 return true;
3209
3210         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3211             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3212                 return true;
3213
3214         return false;
3215 }
3216
3217 static int ipv6_generate_stable_address(struct in6_addr *address,
3218                                         u8 dad_count,
3219                                         const struct inet6_dev *idev)
3220 {
3221         static DEFINE_SPINLOCK(lock);
3222         static __u32 digest[SHA1_DIGEST_WORDS];
3223         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3224
3225         static union {
3226                 char __data[SHA1_BLOCK_SIZE];
3227                 struct {
3228                         struct in6_addr secret;
3229                         __be32 prefix[2];
3230                         unsigned char hwaddr[MAX_ADDR_LEN];
3231                         u8 dad_count;
3232                 } __packed;
3233         } data;
3234
3235         struct in6_addr secret;
3236         struct in6_addr temp;
3237         struct net *net = dev_net(idev->dev);
3238
3239         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3240
3241         if (idev->cnf.stable_secret.initialized)
3242                 secret = idev->cnf.stable_secret.secret;
3243         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3244                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3245         else
3246                 return -1;
3247
3248 retry:
3249         spin_lock_bh(&lock);
3250
3251         sha1_init(digest);
3252         memset(&data, 0, sizeof(data));
3253         memset(workspace, 0, sizeof(workspace));
3254         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3255         data.prefix[0] = address->s6_addr32[0];
3256         data.prefix[1] = address->s6_addr32[1];
3257         data.secret = secret;
3258         data.dad_count = dad_count;
3259
3260         sha1_transform(digest, data.__data, workspace);
3261
3262         temp = *address;
3263         temp.s6_addr32[2] = (__force __be32)digest[0];
3264         temp.s6_addr32[3] = (__force __be32)digest[1];
3265
3266         spin_unlock_bh(&lock);
3267
3268         if (ipv6_reserved_interfaceid(temp)) {
3269                 dad_count++;
3270                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3271                         return -1;
3272                 goto retry;
3273         }
3274
3275         *address = temp;
3276         return 0;
3277 }
3278
3279 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3280 {
3281         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3282
3283         if (s->initialized)
3284                 return;
3285         s = &idev->cnf.stable_secret;
3286         get_random_bytes(&s->secret, sizeof(s->secret));
3287         s->initialized = true;
3288 }
3289
3290 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3291 {
3292         struct in6_addr addr;
3293
3294         /* no link local addresses on L3 master devices */
3295         if (netif_is_l3_master(idev->dev))
3296                 return;
3297
3298         /* no link local addresses on devices flagged as slaves */
3299         if (idev->dev->flags & IFF_SLAVE)
3300                 return;
3301
3302         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3303
3304         switch (idev->cnf.addr_gen_mode) {
3305         case IN6_ADDR_GEN_MODE_RANDOM:
3306                 ipv6_gen_mode_random_init(idev);
3307                 fallthrough;
3308         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3309                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3310                         addrconf_add_linklocal(idev, &addr,
3311                                                IFA_F_STABLE_PRIVACY);
3312                 else if (prefix_route)
3313                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3314                                               0, 0, GFP_KERNEL);
3315                 break;
3316         case IN6_ADDR_GEN_MODE_EUI64:
3317                 /* addrconf_add_linklocal also adds a prefix_route and we
3318                  * only need to care about prefix routes if ipv6_generate_eui64
3319                  * couldn't generate one.
3320                  */
3321                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3322                         addrconf_add_linklocal(idev, &addr, 0);
3323                 else if (prefix_route)
3324                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3325                                               0, 0, GFP_KERNEL);
3326                 break;
3327         case IN6_ADDR_GEN_MODE_NONE:
3328         default:
3329                 /* will not add any link local address */
3330                 break;
3331         }
3332 }
3333
3334 static void addrconf_dev_config(struct net_device *dev)
3335 {
3336         struct inet6_dev *idev;
3337
3338         ASSERT_RTNL();
3339
3340         if ((dev->type != ARPHRD_ETHER) &&
3341             (dev->type != ARPHRD_FDDI) &&
3342             (dev->type != ARPHRD_ARCNET) &&
3343             (dev->type != ARPHRD_INFINIBAND) &&
3344             (dev->type != ARPHRD_IEEE1394) &&
3345             (dev->type != ARPHRD_TUNNEL6) &&
3346             (dev->type != ARPHRD_6LOWPAN) &&
3347             (dev->type != ARPHRD_IP6GRE) &&
3348             (dev->type != ARPHRD_IPGRE) &&
3349             (dev->type != ARPHRD_TUNNEL) &&
3350             (dev->type != ARPHRD_NONE) &&
3351             (dev->type != ARPHRD_RAWIP)) {
3352                 /* Alas, we support only Ethernet autoconfiguration. */
3353                 idev = __in6_dev_get(dev);
3354                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3355                     dev->flags & IFF_MULTICAST)
3356                         ipv6_mc_up(idev);
3357                 return;
3358         }
3359
3360         idev = addrconf_add_dev(dev);
3361         if (IS_ERR(idev))
3362                 return;
3363
3364         /* this device type has no EUI support */
3365         if (dev->type == ARPHRD_NONE &&
3366             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3367                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3368
3369         addrconf_addr_gen(idev, false);
3370 }
3371
3372 #if IS_ENABLED(CONFIG_IPV6_SIT)
3373 static void addrconf_sit_config(struct net_device *dev)
3374 {
3375         struct inet6_dev *idev;
3376
3377         ASSERT_RTNL();
3378
3379         /*
3380          * Configure the tunnel with one of our IPv4
3381          * addresses... we should configure all of
3382          * our v4 addrs in the tunnel
3383          */
3384
3385         idev = ipv6_find_idev(dev);
3386         if (IS_ERR(idev)) {
3387                 pr_debug("%s: add_dev failed\n", __func__);
3388                 return;
3389         }
3390
3391         if (dev->priv_flags & IFF_ISATAP) {
3392                 addrconf_addr_gen(idev, false);
3393                 return;
3394         }
3395
3396         sit_add_v4_addrs(idev);
3397
3398         if (dev->flags&IFF_POINTOPOINT)
3399                 addrconf_add_mroute(dev);
3400 }
3401 #endif
3402
3403 #if IS_ENABLED(CONFIG_NET_IPGRE)
3404 static void addrconf_gre_config(struct net_device *dev)
3405 {
3406         struct inet6_dev *idev;
3407
3408         ASSERT_RTNL();
3409
3410         idev = ipv6_find_idev(dev);
3411         if (IS_ERR(idev)) {
3412                 pr_debug("%s: add_dev failed\n", __func__);
3413                 return;
3414         }
3415
3416         addrconf_addr_gen(idev, true);
3417         if (dev->flags & IFF_POINTOPOINT)
3418                 addrconf_add_mroute(dev);
3419 }
3420 #endif
3421
3422 static int fixup_permanent_addr(struct net *net,
3423                                 struct inet6_dev *idev,
3424                                 struct inet6_ifaddr *ifp)
3425 {
3426         /* !fib6_node means the host route was removed from the
3427          * FIB, for example, if 'lo' device is taken down. In that
3428          * case regenerate the host route.
3429          */
3430         if (!ifp->rt || !ifp->rt->fib6_node) {
3431                 struct fib6_info *f6i, *prev;
3432
3433                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3434                                          GFP_ATOMIC);
3435                 if (IS_ERR(f6i))
3436                         return PTR_ERR(f6i);
3437
3438                 /* ifp->rt can be accessed outside of rtnl */
3439                 spin_lock(&ifp->lock);
3440                 prev = ifp->rt;
3441                 ifp->rt = f6i;
3442                 spin_unlock(&ifp->lock);
3443
3444                 fib6_info_release(prev);
3445         }
3446
3447         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3448                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3449                                       ifp->rt_priority, idev->dev, 0, 0,
3450                                       GFP_ATOMIC);
3451         }
3452
3453         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3454                 addrconf_dad_start(ifp);
3455
3456         return 0;
3457 }
3458
3459 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3460 {
3461         struct inet6_ifaddr *ifp, *tmp;
3462         struct inet6_dev *idev;
3463
3464         idev = __in6_dev_get(dev);
3465         if (!idev)
3466                 return;
3467
3468         write_lock_bh(&idev->lock);
3469
3470         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3471                 if ((ifp->flags & IFA_F_PERMANENT) &&
3472                     fixup_permanent_addr(net, idev, ifp) < 0) {
3473                         write_unlock_bh(&idev->lock);
3474                         in6_ifa_hold(ifp);
3475                         ipv6_del_addr(ifp);
3476                         write_lock_bh(&idev->lock);
3477
3478                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3479                                              idev->dev->name, &ifp->addr);
3480                 }
3481         }
3482
3483         write_unlock_bh(&idev->lock);
3484 }
3485
3486 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3487                            void *ptr)
3488 {
3489         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3490         struct netdev_notifier_change_info *change_info;
3491         struct netdev_notifier_changeupper_info *info;
3492         struct inet6_dev *idev = __in6_dev_get(dev);
3493         struct net *net = dev_net(dev);
3494         int run_pending = 0;
3495         int err;
3496
3497         switch (event) {
3498         case NETDEV_REGISTER:
3499                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3500                         idev = ipv6_add_dev(dev);
3501                         if (IS_ERR(idev))
3502                                 return notifier_from_errno(PTR_ERR(idev));
3503                 }
3504                 break;
3505
3506         case NETDEV_CHANGEMTU:
3507                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3508                 if (dev->mtu < IPV6_MIN_MTU) {
3509                         addrconf_ifdown(dev, dev != net->loopback_dev);
3510                         break;
3511                 }
3512
3513                 if (idev) {
3514                         rt6_mtu_change(dev, dev->mtu);
3515                         idev->cnf.mtu6 = dev->mtu;
3516                         break;
3517                 }
3518
3519                 /* allocate new idev */
3520                 idev = ipv6_add_dev(dev);
3521                 if (IS_ERR(idev))
3522                         break;
3523
3524                 /* device is still not ready */
3525                 if (!(idev->if_flags & IF_READY))
3526                         break;
3527
3528                 run_pending = 1;
3529                 fallthrough;
3530         case NETDEV_UP:
3531         case NETDEV_CHANGE:
3532                 if (dev->flags & IFF_SLAVE)
3533                         break;
3534
3535                 if (idev && idev->cnf.disable_ipv6)
3536                         break;
3537
3538                 if (event == NETDEV_UP) {
3539                         /* restore routes for permanent addresses */
3540                         addrconf_permanent_addr(net, dev);
3541
3542                         if (!addrconf_link_ready(dev)) {
3543                                 /* device is not ready yet. */
3544                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3545                                          dev->name);
3546                                 break;
3547                         }
3548
3549                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3550                                 idev = ipv6_add_dev(dev);
3551
3552                         if (!IS_ERR_OR_NULL(idev)) {
3553                                 idev->if_flags |= IF_READY;
3554                                 run_pending = 1;
3555                         }
3556                 } else if (event == NETDEV_CHANGE) {
3557                         if (!addrconf_link_ready(dev)) {
3558                                 /* device is still not ready. */
3559                                 rt6_sync_down_dev(dev, event);
3560                                 break;
3561                         }
3562
3563                         if (!IS_ERR_OR_NULL(idev)) {
3564                                 if (idev->if_flags & IF_READY) {
3565                                         /* device is already configured -
3566                                          * but resend MLD reports, we might
3567                                          * have roamed and need to update
3568                                          * multicast snooping switches
3569                                          */
3570                                         ipv6_mc_up(idev);
3571                                         change_info = ptr;
3572                                         if (change_info->flags_changed & IFF_NOARP)
3573                                                 addrconf_dad_run(idev, true);
3574                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3575                                         break;
3576                                 }
3577                                 idev->if_flags |= IF_READY;
3578                         }
3579
3580                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3581                                 dev->name);
3582
3583                         run_pending = 1;
3584                 }
3585
3586                 switch (dev->type) {
3587 #if IS_ENABLED(CONFIG_IPV6_SIT)
3588                 case ARPHRD_SIT:
3589                         addrconf_sit_config(dev);
3590                         break;
3591 #endif
3592 #if IS_ENABLED(CONFIG_NET_IPGRE)
3593                 case ARPHRD_IPGRE:
3594                         addrconf_gre_config(dev);
3595                         break;
3596 #endif
3597                 case ARPHRD_LOOPBACK:
3598                         init_loopback(dev);
3599                         break;
3600
3601                 default:
3602                         addrconf_dev_config(dev);
3603                         break;
3604                 }
3605
3606                 if (!IS_ERR_OR_NULL(idev)) {
3607                         if (run_pending)
3608                                 addrconf_dad_run(idev, false);
3609
3610                         /* Device has an address by now */
3611                         rt6_sync_up(dev, RTNH_F_DEAD);
3612
3613                         /*
3614                          * If the MTU changed during the interface down,
3615                          * when the interface up, the changed MTU must be
3616                          * reflected in the idev as well as routers.
3617                          */
3618                         if (idev->cnf.mtu6 != dev->mtu &&
3619                             dev->mtu >= IPV6_MIN_MTU) {
3620                                 rt6_mtu_change(dev, dev->mtu);
3621                                 idev->cnf.mtu6 = dev->mtu;
3622                         }
3623                         idev->tstamp = jiffies;
3624                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3625
3626                         /*
3627                          * If the changed mtu during down is lower than
3628                          * IPV6_MIN_MTU stop IPv6 on this interface.
3629                          */
3630                         if (dev->mtu < IPV6_MIN_MTU)
3631                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3632                 }
3633                 break;
3634
3635         case NETDEV_DOWN:
3636         case NETDEV_UNREGISTER:
3637                 /*
3638                  *      Remove all addresses from this interface.
3639                  */
3640                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3641                 break;
3642
3643         case NETDEV_CHANGENAME:
3644                 if (idev) {
3645                         snmp6_unregister_dev(idev);
3646                         addrconf_sysctl_unregister(idev);
3647                         err = addrconf_sysctl_register(idev);
3648                         if (err)
3649                                 return notifier_from_errno(err);
3650                         err = snmp6_register_dev(idev);
3651                         if (err) {
3652                                 addrconf_sysctl_unregister(idev);
3653                                 return notifier_from_errno(err);
3654                         }
3655                 }
3656                 break;
3657
3658         case NETDEV_PRE_TYPE_CHANGE:
3659         case NETDEV_POST_TYPE_CHANGE:
3660                 if (idev)
3661                         addrconf_type_change(dev, event);
3662                 break;
3663
3664         case NETDEV_CHANGEUPPER:
3665                 info = ptr;
3666
3667                 /* flush all routes if dev is linked to or unlinked from
3668                  * an L3 master device (e.g., VRF)
3669                  */
3670                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3671                         addrconf_ifdown(dev, false);
3672         }
3673
3674         return NOTIFY_OK;
3675 }
3676
3677 /*
3678  *      addrconf module should be notified of a device going up
3679  */
3680 static struct notifier_block ipv6_dev_notf = {
3681         .notifier_call = addrconf_notify,
3682         .priority = ADDRCONF_NOTIFY_PRIORITY,
3683 };
3684
3685 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3686 {
3687         struct inet6_dev *idev;
3688         ASSERT_RTNL();
3689
3690         idev = __in6_dev_get(dev);
3691
3692         if (event == NETDEV_POST_TYPE_CHANGE)
3693                 ipv6_mc_remap(idev);
3694         else if (event == NETDEV_PRE_TYPE_CHANGE)
3695                 ipv6_mc_unmap(idev);
3696 }
3697
3698 static bool addr_is_local(const struct in6_addr *addr)
3699 {
3700         return ipv6_addr_type(addr) &
3701                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3702 }
3703
3704 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3705 {
3706         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3707         struct net *net = dev_net(dev);
3708         struct inet6_dev *idev;
3709         struct inet6_ifaddr *ifa, *tmp;
3710         bool keep_addr = false;
3711         int state, i;
3712
3713         ASSERT_RTNL();
3714
3715         rt6_disable_ip(dev, event);
3716
3717         idev = __in6_dev_get(dev);
3718         if (!idev)
3719                 return -ENODEV;
3720
3721         /*
3722          * Step 1: remove reference to ipv6 device from parent device.
3723          *         Do not dev_put!
3724          */
3725         if (unregister) {
3726                 idev->dead = 1;
3727
3728                 /* protected by rtnl_lock */
3729                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3730
3731                 /* Step 1.5: remove snmp6 entry */
3732                 snmp6_unregister_dev(idev);
3733
3734         }
3735
3736         /* combine the user config with event to determine if permanent
3737          * addresses are to be removed from address hash table
3738          */
3739         if (!unregister && !idev->cnf.disable_ipv6) {
3740                 /* aggregate the system setting and interface setting */
3741                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3742
3743                 if (!_keep_addr)
3744                         _keep_addr = idev->cnf.keep_addr_on_down;
3745
3746                 keep_addr = (_keep_addr > 0);
3747         }
3748
3749         /* Step 2: clear hash table */
3750         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3751                 struct hlist_head *h = &inet6_addr_lst[i];
3752
3753                 spin_lock_bh(&addrconf_hash_lock);
3754 restart:
3755                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3756                         if (ifa->idev == idev) {
3757                                 addrconf_del_dad_work(ifa);
3758                                 /* combined flag + permanent flag decide if
3759                                  * address is retained on a down event
3760                                  */
3761                                 if (!keep_addr ||
3762                                     !(ifa->flags & IFA_F_PERMANENT) ||
3763                                     addr_is_local(&ifa->addr)) {
3764                                         hlist_del_init_rcu(&ifa->addr_lst);
3765                                         goto restart;
3766                                 }
3767                         }
3768                 }
3769                 spin_unlock_bh(&addrconf_hash_lock);
3770         }
3771
3772         write_lock_bh(&idev->lock);
3773
3774         addrconf_del_rs_timer(idev);
3775
3776         /* Step 2: clear flags for stateless addrconf */
3777         if (!unregister)
3778                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3779
3780         /* Step 3: clear tempaddr list */
3781         while (!list_empty(&idev->tempaddr_list)) {
3782                 ifa = list_first_entry(&idev->tempaddr_list,
3783                                        struct inet6_ifaddr, tmp_list);
3784                 list_del(&ifa->tmp_list);
3785                 write_unlock_bh(&idev->lock);
3786                 spin_lock_bh(&ifa->lock);
3787
3788                 if (ifa->ifpub) {
3789                         in6_ifa_put(ifa->ifpub);
3790                         ifa->ifpub = NULL;
3791                 }
3792                 spin_unlock_bh(&ifa->lock);
3793                 in6_ifa_put(ifa);
3794                 write_lock_bh(&idev->lock);
3795         }
3796
3797         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3798                 struct fib6_info *rt = NULL;
3799                 bool keep;
3800
3801                 addrconf_del_dad_work(ifa);
3802
3803                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3804                         !addr_is_local(&ifa->addr);
3805
3806                 write_unlock_bh(&idev->lock);
3807                 spin_lock_bh(&ifa->lock);
3808
3809                 if (keep) {
3810                         /* set state to skip the notifier below */
3811                         state = INET6_IFADDR_STATE_DEAD;
3812                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3813                         if (!(ifa->flags & IFA_F_NODAD))
3814                                 ifa->flags |= IFA_F_TENTATIVE;
3815
3816                         rt = ifa->rt;
3817                         ifa->rt = NULL;
3818                 } else {
3819                         state = ifa->state;
3820                         ifa->state = INET6_IFADDR_STATE_DEAD;
3821                 }
3822
3823                 spin_unlock_bh(&ifa->lock);
3824
3825                 if (rt)
3826                         ip6_del_rt(net, rt, false);
3827
3828                 if (state != INET6_IFADDR_STATE_DEAD) {
3829                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3830                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3831                 } else {
3832                         if (idev->cnf.forwarding)
3833                                 addrconf_leave_anycast(ifa);
3834                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3835                 }
3836
3837                 write_lock_bh(&idev->lock);
3838                 if (!keep) {
3839                         list_del_rcu(&ifa->if_list);
3840                         in6_ifa_put(ifa);
3841                 }
3842         }
3843
3844         write_unlock_bh(&idev->lock);
3845
3846         /* Step 5: Discard anycast and multicast list */
3847         if (unregister) {
3848                 ipv6_ac_destroy_dev(idev);
3849                 ipv6_mc_destroy_dev(idev);
3850         } else {
3851                 ipv6_mc_down(idev);
3852         }
3853
3854         idev->tstamp = jiffies;
3855
3856         /* Last: Shot the device (if unregistered) */
3857         if (unregister) {
3858                 addrconf_sysctl_unregister(idev);
3859                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3860                 neigh_ifdown(&nd_tbl, dev);
3861                 in6_dev_put(idev);
3862         }
3863         return 0;
3864 }
3865
3866 static void addrconf_rs_timer(struct timer_list *t)
3867 {
3868         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3869         struct net_device *dev = idev->dev;
3870         struct in6_addr lladdr;
3871
3872         write_lock(&idev->lock);
3873         if (idev->dead || !(idev->if_flags & IF_READY))
3874                 goto out;
3875
3876         if (!ipv6_accept_ra(idev))
3877                 goto out;
3878
3879         /* Announcement received after solicitation was sent */
3880         if (idev->if_flags & IF_RA_RCVD)
3881                 goto out;
3882
3883         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3884                 write_unlock(&idev->lock);
3885                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3886                         ndisc_send_rs(dev, &lladdr,
3887                                       &in6addr_linklocal_allrouters);
3888                 else
3889                         goto put;
3890
3891                 write_lock(&idev->lock);
3892                 idev->rs_interval = rfc3315_s14_backoff_update(
3893                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3894                 /* The wait after the last probe can be shorter */
3895                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3896                                              idev->cnf.rtr_solicits) ?
3897                                       idev->cnf.rtr_solicit_delay :
3898                                       idev->rs_interval);
3899         } else {
3900                 /*
3901                  * Note: we do not support deprecated "all on-link"
3902                  * assumption any longer.
3903                  */
3904                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3905         }
3906
3907 out:
3908         write_unlock(&idev->lock);
3909 put:
3910         in6_dev_put(idev);
3911 }
3912
3913 /*
3914  *      Duplicate Address Detection
3915  */
3916 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3917 {
3918         unsigned long rand_num;
3919         struct inet6_dev *idev = ifp->idev;
3920         u64 nonce;
3921
3922         if (ifp->flags & IFA_F_OPTIMISTIC)
3923                 rand_num = 0;
3924         else
3925                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3926
3927         nonce = 0;
3928         if (idev->cnf.enhanced_dad ||
3929             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3930                 do
3931                         get_random_bytes(&nonce, 6);
3932                 while (nonce == 0);
3933         }
3934         ifp->dad_nonce = nonce;
3935         ifp->dad_probes = idev->cnf.dad_transmits;
3936         addrconf_mod_dad_work(ifp, rand_num);
3937 }
3938
3939 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3940 {
3941         struct inet6_dev *idev = ifp->idev;
3942         struct net_device *dev = idev->dev;
3943         bool bump_id, notify = false;
3944         struct net *net;
3945
3946         addrconf_join_solict(dev, &ifp->addr);
3947
3948         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3949
3950         read_lock_bh(&idev->lock);
3951         spin_lock(&ifp->lock);
3952         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3953                 goto out;
3954
3955         net = dev_net(dev);
3956         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3957             (net->ipv6.devconf_all->accept_dad < 1 &&
3958              idev->cnf.accept_dad < 1) ||
3959             !(ifp->flags&IFA_F_TENTATIVE) ||
3960             ifp->flags & IFA_F_NODAD) {
3961                 bool send_na = false;
3962
3963                 if (ifp->flags & IFA_F_TENTATIVE &&
3964                     !(ifp->flags & IFA_F_OPTIMISTIC))
3965                         send_na = true;
3966                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3967                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3968                 spin_unlock(&ifp->lock);
3969                 read_unlock_bh(&idev->lock);
3970
3971                 addrconf_dad_completed(ifp, bump_id, send_na);
3972                 return;
3973         }
3974
3975         if (!(idev->if_flags & IF_READY)) {
3976                 spin_unlock(&ifp->lock);
3977                 read_unlock_bh(&idev->lock);
3978                 /*
3979                  * If the device is not ready:
3980                  * - keep it tentative if it is a permanent address.
3981                  * - otherwise, kill it.
3982                  */
3983                 in6_ifa_hold(ifp);
3984                 addrconf_dad_stop(ifp, 0);
3985                 return;
3986         }
3987
3988         /*
3989          * Optimistic nodes can start receiving
3990          * Frames right away
3991          */
3992         if (ifp->flags & IFA_F_OPTIMISTIC) {
3993                 ip6_ins_rt(net, ifp->rt);
3994                 if (ipv6_use_optimistic_addr(net, idev)) {
3995                         /* Because optimistic nodes can use this address,
3996                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3997                          */
3998                         notify = true;
3999                 }
4000         }
4001
4002         addrconf_dad_kick(ifp);
4003 out:
4004         spin_unlock(&ifp->lock);
4005         read_unlock_bh(&idev->lock);
4006         if (notify)
4007                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4008 }
4009
4010 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4011 {
4012         bool begin_dad = false;
4013
4014         spin_lock_bh(&ifp->lock);
4015         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4016                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4017                 begin_dad = true;
4018         }
4019         spin_unlock_bh(&ifp->lock);
4020
4021         if (begin_dad)
4022                 addrconf_mod_dad_work(ifp, 0);
4023 }
4024
4025 static void addrconf_dad_work(struct work_struct *w)
4026 {
4027         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4028                                                 struct inet6_ifaddr,
4029                                                 dad_work);
4030         struct inet6_dev *idev = ifp->idev;
4031         bool bump_id, disable_ipv6 = false;
4032         struct in6_addr mcaddr;
4033
4034         enum {
4035                 DAD_PROCESS,
4036                 DAD_BEGIN,
4037                 DAD_ABORT,
4038         } action = DAD_PROCESS;
4039
4040         rtnl_lock();
4041
4042         spin_lock_bh(&ifp->lock);
4043         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4044                 action = DAD_BEGIN;
4045                 ifp->state = INET6_IFADDR_STATE_DAD;
4046         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4047                 action = DAD_ABORT;
4048                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4049
4050                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4051                      idev->cnf.accept_dad > 1) &&
4052                     !idev->cnf.disable_ipv6 &&
4053                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4054                         struct in6_addr addr;
4055
4056                         addr.s6_addr32[0] = htonl(0xfe800000);
4057                         addr.s6_addr32[1] = 0;
4058
4059                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4060                             ipv6_addr_equal(&ifp->addr, &addr)) {
4061                                 /* DAD failed for link-local based on MAC */
4062                                 idev->cnf.disable_ipv6 = 1;
4063
4064                                 pr_info("%s: IPv6 being disabled!\n",
4065                                         ifp->idev->dev->name);
4066                                 disable_ipv6 = true;
4067                         }
4068                 }
4069         }
4070         spin_unlock_bh(&ifp->lock);
4071
4072         if (action == DAD_BEGIN) {
4073                 addrconf_dad_begin(ifp);
4074                 goto out;
4075         } else if (action == DAD_ABORT) {
4076                 in6_ifa_hold(ifp);
4077                 addrconf_dad_stop(ifp, 1);
4078                 if (disable_ipv6)
4079                         addrconf_ifdown(idev->dev, false);
4080                 goto out;
4081         }
4082
4083         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4084                 goto out;
4085
4086         write_lock_bh(&idev->lock);
4087         if (idev->dead || !(idev->if_flags & IF_READY)) {
4088                 write_unlock_bh(&idev->lock);
4089                 goto out;
4090         }
4091
4092         spin_lock(&ifp->lock);
4093         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4094                 spin_unlock(&ifp->lock);
4095                 write_unlock_bh(&idev->lock);
4096                 goto out;
4097         }
4098
4099         if (ifp->dad_probes == 0) {
4100                 bool send_na = false;
4101
4102                 /*
4103                  * DAD was successful
4104                  */
4105
4106                 if (ifp->flags & IFA_F_TENTATIVE &&
4107                     !(ifp->flags & IFA_F_OPTIMISTIC))
4108                         send_na = true;
4109                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4110                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4111                 spin_unlock(&ifp->lock);
4112                 write_unlock_bh(&idev->lock);
4113
4114                 addrconf_dad_completed(ifp, bump_id, send_na);
4115
4116                 goto out;
4117         }
4118
4119         ifp->dad_probes--;
4120         addrconf_mod_dad_work(ifp,
4121                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4122                                   HZ/100));
4123         spin_unlock(&ifp->lock);
4124         write_unlock_bh(&idev->lock);
4125
4126         /* send a neighbour solicitation for our addr */
4127         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4128         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4129                       ifp->dad_nonce);
4130 out:
4131         in6_ifa_put(ifp);
4132         rtnl_unlock();
4133 }
4134
4135 /* ifp->idev must be at least read locked */
4136 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4137 {
4138         struct inet6_ifaddr *ifpiter;
4139         struct inet6_dev *idev = ifp->idev;
4140
4141         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4142                 if (ifpiter->scope > IFA_LINK)
4143                         break;
4144                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4145                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4146                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4147                     IFA_F_PERMANENT)
4148                         return false;
4149         }
4150         return true;
4151 }
4152
4153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4154                                    bool send_na)
4155 {
4156         struct net_device *dev = ifp->idev->dev;
4157         struct in6_addr lladdr;
4158         bool send_rs, send_mld;
4159
4160         addrconf_del_dad_work(ifp);
4161
4162         /*
4163          *      Configure the address for reception. Now it is valid.
4164          */
4165
4166         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4167
4168         /* If added prefix is link local and we are prepared to process
4169            router advertisements, start sending router solicitations.
4170          */
4171
4172         read_lock_bh(&ifp->idev->lock);
4173         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4174         send_rs = send_mld &&
4175                   ipv6_accept_ra(ifp->idev) &&
4176                   ifp->idev->cnf.rtr_solicits != 0 &&
4177                   (dev->flags&IFF_LOOPBACK) == 0;
4178         read_unlock_bh(&ifp->idev->lock);
4179
4180         /* While dad is in progress mld report's source address is in6_addrany.
4181          * Resend with proper ll now.
4182          */
4183         if (send_mld)
4184                 ipv6_mc_dad_complete(ifp->idev);
4185
4186         /* send unsolicited NA if enabled */
4187         if (send_na &&
4188             (ifp->idev->cnf.ndisc_notify ||
4189              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4190                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4191                               /*router=*/ !!ifp->idev->cnf.forwarding,
4192                               /*solicited=*/ false, /*override=*/ true,
4193                               /*inc_opt=*/ true);
4194         }
4195
4196         if (send_rs) {
4197                 /*
4198                  *      If a host as already performed a random delay
4199                  *      [...] as part of DAD [...] there is no need
4200                  *      to delay again before sending the first RS
4201                  */
4202                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4203                         return;
4204                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4205
4206                 write_lock_bh(&ifp->idev->lock);
4207                 spin_lock(&ifp->lock);
4208                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4209                         ifp->idev->cnf.rtr_solicit_interval);
4210                 ifp->idev->rs_probes = 1;
4211                 ifp->idev->if_flags |= IF_RS_SENT;
4212                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4213                 spin_unlock(&ifp->lock);
4214                 write_unlock_bh(&ifp->idev->lock);
4215         }
4216
4217         if (bump_id)
4218                 rt_genid_bump_ipv6(dev_net(dev));
4219
4220         /* Make sure that a new temporary address will be created
4221          * before this temporary address becomes deprecated.
4222          */
4223         if (ifp->flags & IFA_F_TEMPORARY)
4224                 addrconf_verify_rtnl();
4225 }
4226
4227 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4228 {
4229         struct inet6_ifaddr *ifp;
4230
4231         read_lock_bh(&idev->lock);
4232         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4233                 spin_lock(&ifp->lock);
4234                 if ((ifp->flags & IFA_F_TENTATIVE &&
4235                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4236                         if (restart)
4237                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4238                         addrconf_dad_kick(ifp);
4239                 }
4240                 spin_unlock(&ifp->lock);
4241         }
4242         read_unlock_bh(&idev->lock);
4243 }
4244
4245 #ifdef CONFIG_PROC_FS
4246 struct if6_iter_state {
4247         struct seq_net_private p;
4248         int bucket;
4249         int offset;
4250 };
4251
4252 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4253 {
4254         struct if6_iter_state *state = seq->private;
4255         struct net *net = seq_file_net(seq);
4256         struct inet6_ifaddr *ifa = NULL;
4257         int p = 0;
4258
4259         /* initial bucket if pos is 0 */
4260         if (pos == 0) {
4261                 state->bucket = 0;
4262                 state->offset = 0;
4263         }
4264
4265         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4266                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4267                                          addr_lst) {
4268                         if (!net_eq(dev_net(ifa->idev->dev), net))
4269                                 continue;
4270                         /* sync with offset */
4271                         if (p < state->offset) {
4272                                 p++;
4273                                 continue;
4274                         }
4275                         return ifa;
4276                 }
4277
4278                 /* prepare for next bucket */
4279                 state->offset = 0;
4280                 p = 0;
4281         }
4282         return NULL;
4283 }
4284
4285 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4286                                          struct inet6_ifaddr *ifa)
4287 {
4288         struct if6_iter_state *state = seq->private;
4289         struct net *net = seq_file_net(seq);
4290
4291         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4292                 if (!net_eq(dev_net(ifa->idev->dev), net))
4293                         continue;
4294                 state->offset++;
4295                 return ifa;
4296         }
4297
4298         state->offset = 0;
4299         while (++state->bucket < IN6_ADDR_HSIZE) {
4300                 hlist_for_each_entry_rcu(ifa,
4301                                      &inet6_addr_lst[state->bucket], addr_lst) {
4302                         if (!net_eq(dev_net(ifa->idev->dev), net))
4303                                 continue;
4304                         return ifa;
4305                 }
4306         }
4307
4308         return NULL;
4309 }
4310
4311 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4312         __acquires(rcu)
4313 {
4314         rcu_read_lock();
4315         return if6_get_first(seq, *pos);
4316 }
4317
4318 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4319 {
4320         struct inet6_ifaddr *ifa;
4321
4322         ifa = if6_get_next(seq, v);
4323         ++*pos;
4324         return ifa;
4325 }
4326
4327 static void if6_seq_stop(struct seq_file *seq, void *v)
4328         __releases(rcu)
4329 {
4330         rcu_read_unlock();
4331 }
4332
4333 static int if6_seq_show(struct seq_file *seq, void *v)
4334 {
4335         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4336         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4337                    &ifp->addr,
4338                    ifp->idev->dev->ifindex,
4339                    ifp->prefix_len,
4340                    ifp->scope,
4341                    (u8) ifp->flags,
4342                    ifp->idev->dev->name);
4343         return 0;
4344 }
4345
4346 static const struct seq_operations if6_seq_ops = {
4347         .start  = if6_seq_start,
4348         .next   = if6_seq_next,
4349         .show   = if6_seq_show,
4350         .stop   = if6_seq_stop,
4351 };
4352
4353 static int __net_init if6_proc_net_init(struct net *net)
4354 {
4355         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4356                         sizeof(struct if6_iter_state)))
4357                 return -ENOMEM;
4358         return 0;
4359 }
4360
4361 static void __net_exit if6_proc_net_exit(struct net *net)
4362 {
4363         remove_proc_entry("if_inet6", net->proc_net);
4364 }
4365
4366 static struct pernet_operations if6_proc_net_ops = {
4367         .init = if6_proc_net_init,
4368         .exit = if6_proc_net_exit,
4369 };
4370
4371 int __init if6_proc_init(void)
4372 {
4373         return register_pernet_subsys(&if6_proc_net_ops);
4374 }
4375
4376 void if6_proc_exit(void)
4377 {
4378         unregister_pernet_subsys(&if6_proc_net_ops);
4379 }
4380 #endif  /* CONFIG_PROC_FS */
4381
4382 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4383 /* Check if address is a home address configured on any interface. */
4384 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4385 {
4386         unsigned int hash = inet6_addr_hash(net, addr);
4387         struct inet6_ifaddr *ifp = NULL;
4388         int ret = 0;
4389
4390         rcu_read_lock();
4391         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4392                 if (!net_eq(dev_net(ifp->idev->dev), net))
4393                         continue;
4394                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4395                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4396                         ret = 1;
4397                         break;
4398                 }
4399         }
4400         rcu_read_unlock();
4401         return ret;
4402 }
4403 #endif
4404
4405 /* RFC6554 has some algorithm to avoid loops in segment routing by
4406  * checking if the segments contains any of a local interface address.
4407  *
4408  * Quote:
4409  *
4410  * To detect loops in the SRH, a router MUST determine if the SRH
4411  * includes multiple addresses assigned to any interface on that router.
4412  * If such addresses appear more than once and are separated by at least
4413  * one address not assigned to that router.
4414  */
4415 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4416                           unsigned char nsegs)
4417 {
4418         const struct in6_addr *addr;
4419         int i, ret = 0, found = 0;
4420         struct inet6_ifaddr *ifp;
4421         bool separated = false;
4422         unsigned int hash;
4423         bool hash_found;
4424
4425         rcu_read_lock();
4426         for (i = 0; i < nsegs; i++) {
4427                 addr = &segs[i];
4428                 hash = inet6_addr_hash(net, addr);
4429
4430                 hash_found = false;
4431                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4432                         if (!net_eq(dev_net(ifp->idev->dev), net))
4433                                 continue;
4434
4435                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4436                                 hash_found = true;
4437                                 break;
4438                         }
4439                 }
4440
4441                 if (hash_found) {
4442                         if (found > 1 && separated) {
4443                                 ret = 1;
4444                                 break;
4445                         }
4446
4447                         separated = false;
4448                         found++;
4449                 } else {
4450                         separated = true;
4451                 }
4452         }
4453         rcu_read_unlock();
4454
4455         return ret;
4456 }
4457
4458 /*
4459  *      Periodic address status verification
4460  */
4461
4462 static void addrconf_verify_rtnl(void)
4463 {
4464         unsigned long now, next, next_sec, next_sched;
4465         struct inet6_ifaddr *ifp;
4466         int i;
4467
4468         ASSERT_RTNL();
4469
4470         rcu_read_lock_bh();
4471         now = jiffies;
4472         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4473
4474         cancel_delayed_work(&addr_chk_work);
4475
4476         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4477 restart:
4478                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4479                         unsigned long age;
4480
4481                         /* When setting preferred_lft to a value not zero or
4482                          * infinity, while valid_lft is infinity
4483                          * IFA_F_PERMANENT has a non-infinity life time.
4484                          */
4485                         if ((ifp->flags & IFA_F_PERMANENT) &&
4486                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4487                                 continue;
4488
4489                         spin_lock(&ifp->lock);
4490                         /* We try to batch several events at once. */
4491                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4492
4493                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4494                             age >= ifp->valid_lft) {
4495                                 spin_unlock(&ifp->lock);
4496                                 in6_ifa_hold(ifp);
4497                                 rcu_read_unlock_bh();
4498                                 ipv6_del_addr(ifp);
4499                                 rcu_read_lock_bh();
4500                                 goto restart;
4501                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4502                                 spin_unlock(&ifp->lock);
4503                                 continue;
4504                         } else if (age >= ifp->prefered_lft) {
4505                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4506                                 int deprecate = 0;
4507
4508                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4509                                         deprecate = 1;
4510                                         ifp->flags |= IFA_F_DEPRECATED;
4511                                 }
4512
4513                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4514                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4515                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4516
4517                                 spin_unlock(&ifp->lock);
4518
4519                                 if (deprecate) {
4520                                         in6_ifa_hold(ifp);
4521
4522                                         ipv6_ifa_notify(0, ifp);
4523                                         in6_ifa_put(ifp);
4524                                         goto restart;
4525                                 }
4526                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4527                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4528                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4529                                         ifp->idev->cnf.dad_transmits *
4530                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4531
4532                                 if (age >= ifp->prefered_lft - regen_advance) {
4533                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4534                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4535                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4536                                         if (!ifp->regen_count && ifpub) {
4537                                                 ifp->regen_count++;
4538                                                 in6_ifa_hold(ifp);
4539                                                 in6_ifa_hold(ifpub);
4540                                                 spin_unlock(&ifp->lock);
4541
4542                                                 spin_lock(&ifpub->lock);
4543                                                 ifpub->regen_count = 0;
4544                                                 spin_unlock(&ifpub->lock);
4545                                                 rcu_read_unlock_bh();
4546                                                 ipv6_create_tempaddr(ifpub, true);
4547                                                 in6_ifa_put(ifpub);
4548                                                 in6_ifa_put(ifp);
4549                                                 rcu_read_lock_bh();
4550                                                 goto restart;
4551                                         }
4552                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4553                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4554                                 spin_unlock(&ifp->lock);
4555                         } else {
4556                                 /* ifp->prefered_lft <= ifp->valid_lft */
4557                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4558                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4559                                 spin_unlock(&ifp->lock);
4560                         }
4561                 }
4562         }
4563
4564         next_sec = round_jiffies_up(next);
4565         next_sched = next;
4566
4567         /* If rounded timeout is accurate enough, accept it. */
4568         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4569                 next_sched = next_sec;
4570
4571         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4572         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4573                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4574
4575         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4576                  now, next, next_sec, next_sched);
4577         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4578         rcu_read_unlock_bh();
4579 }
4580
4581 static void addrconf_verify_work(struct work_struct *w)
4582 {
4583         rtnl_lock();
4584         addrconf_verify_rtnl();
4585         rtnl_unlock();
4586 }
4587
4588 static void addrconf_verify(void)
4589 {
4590         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4591 }
4592
4593 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4594                                      struct in6_addr **peer_pfx)
4595 {
4596         struct in6_addr *pfx = NULL;
4597
4598         *peer_pfx = NULL;
4599
4600         if (addr)
4601                 pfx = nla_data(addr);
4602
4603         if (local) {
4604                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4605                         *peer_pfx = pfx;
4606                 pfx = nla_data(local);
4607         }
4608
4609         return pfx;
4610 }
4611
4612 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4613         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4614         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4615         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4616         [IFA_FLAGS]             = { .len = sizeof(u32) },
4617         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4618         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4619 };
4620
4621 static int
4622 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4623                   struct netlink_ext_ack *extack)
4624 {
4625         struct net *net = sock_net(skb->sk);
4626         struct ifaddrmsg *ifm;
4627         struct nlattr *tb[IFA_MAX+1];
4628         struct in6_addr *pfx, *peer_pfx;
4629         u32 ifa_flags;
4630         int err;
4631
4632         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4633                                      ifa_ipv6_policy, extack);
4634         if (err < 0)
4635                 return err;
4636
4637         ifm = nlmsg_data(nlh);
4638         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4639         if (!pfx)
4640                 return -EINVAL;
4641
4642         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4643
4644         /* We ignore other flags so far. */
4645         ifa_flags &= IFA_F_MANAGETEMPADDR;
4646
4647         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4648                               ifm->ifa_prefixlen);
4649 }
4650
4651 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4652                                unsigned long expires, u32 flags,
4653                                bool modify_peer)
4654 {
4655         struct fib6_info *f6i;
4656         u32 prio;
4657
4658         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4659                                         ifp->prefix_len,
4660                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4661         if (!f6i)
4662                 return -ENOENT;
4663
4664         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4665         if (f6i->fib6_metric != prio) {
4666                 /* delete old one */
4667                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4668
4669                 /* add new one */
4670                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4671                                       ifp->prefix_len,
4672                                       ifp->rt_priority, ifp->idev->dev,
4673                                       expires, flags, GFP_KERNEL);
4674         } else {
4675                 if (!expires)
4676                         fib6_clean_expires(f6i);
4677                 else
4678                         fib6_set_expires(f6i, expires);
4679
4680                 fib6_info_release(f6i);
4681         }
4682
4683         return 0;
4684 }
4685
4686 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4687 {
4688         u32 flags;
4689         clock_t expires;
4690         unsigned long timeout;
4691         bool was_managetempaddr;
4692         bool had_prefixroute;
4693         bool new_peer = false;
4694
4695         ASSERT_RTNL();
4696
4697         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4698                 return -EINVAL;
4699
4700         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4701             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4702                 return -EINVAL;
4703
4704         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4705                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4706
4707         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4708         if (addrconf_finite_timeout(timeout)) {
4709                 expires = jiffies_to_clock_t(timeout * HZ);
4710                 cfg->valid_lft = timeout;
4711                 flags = RTF_EXPIRES;
4712         } else {
4713                 expires = 0;
4714                 flags = 0;
4715                 cfg->ifa_flags |= IFA_F_PERMANENT;
4716         }
4717
4718         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4719         if (addrconf_finite_timeout(timeout)) {
4720                 if (timeout == 0)
4721                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4722                 cfg->preferred_lft = timeout;
4723         }
4724
4725         if (cfg->peer_pfx &&
4726             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4727                 if (!ipv6_addr_any(&ifp->peer_addr))
4728                         cleanup_prefix_route(ifp, expires, true, true);
4729                 new_peer = true;
4730         }
4731
4732         spin_lock_bh(&ifp->lock);
4733         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4734         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4735                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4736         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4737                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4738                         IFA_F_NOPREFIXROUTE);
4739         ifp->flags |= cfg->ifa_flags;
4740         ifp->tstamp = jiffies;
4741         ifp->valid_lft = cfg->valid_lft;
4742         ifp->prefered_lft = cfg->preferred_lft;
4743
4744         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4745                 ifp->rt_priority = cfg->rt_priority;
4746
4747         if (new_peer)
4748                 ifp->peer_addr = *cfg->peer_pfx;
4749
4750         spin_unlock_bh(&ifp->lock);
4751         if (!(ifp->flags&IFA_F_TENTATIVE))
4752                 ipv6_ifa_notify(0, ifp);
4753
4754         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4755                 int rc = -ENOENT;
4756
4757                 if (had_prefixroute)
4758                         rc = modify_prefix_route(ifp, expires, flags, false);
4759
4760                 /* prefix route could have been deleted; if so restore it */
4761                 if (rc == -ENOENT) {
4762                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4763                                               ifp->rt_priority, ifp->idev->dev,
4764                                               expires, flags, GFP_KERNEL);
4765                 }
4766
4767                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4768                         rc = modify_prefix_route(ifp, expires, flags, true);
4769
4770                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4771                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4772                                               ifp->rt_priority, ifp->idev->dev,
4773                                               expires, flags, GFP_KERNEL);
4774                 }
4775         } else if (had_prefixroute) {
4776                 enum cleanup_prefix_rt_t action;
4777                 unsigned long rt_expires;
4778
4779                 write_lock_bh(&ifp->idev->lock);
4780                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4781                 write_unlock_bh(&ifp->idev->lock);
4782
4783                 if (action != CLEANUP_PREFIX_RT_NOP) {
4784                         cleanup_prefix_route(ifp, rt_expires,
4785                                 action == CLEANUP_PREFIX_RT_DEL, false);
4786                 }
4787         }
4788
4789         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4790                 if (was_managetempaddr &&
4791                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4792                         cfg->valid_lft = 0;
4793                         cfg->preferred_lft = 0;
4794                 }
4795                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4796                                  cfg->preferred_lft, !was_managetempaddr,
4797                                  jiffies);
4798         }
4799
4800         addrconf_verify_rtnl();
4801
4802         return 0;
4803 }
4804
4805 static int
4806 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4807                   struct netlink_ext_ack *extack)
4808 {
4809         struct net *net = sock_net(skb->sk);
4810         struct ifaddrmsg *ifm;
4811         struct nlattr *tb[IFA_MAX+1];
4812         struct in6_addr *peer_pfx;
4813         struct inet6_ifaddr *ifa;
4814         struct net_device *dev;
4815         struct inet6_dev *idev;
4816         struct ifa6_config cfg;
4817         int err;
4818
4819         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4820                                      ifa_ipv6_policy, extack);
4821         if (err < 0)
4822                 return err;
4823
4824         memset(&cfg, 0, sizeof(cfg));
4825
4826         ifm = nlmsg_data(nlh);
4827         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4828         if (!cfg.pfx)
4829                 return -EINVAL;
4830
4831         cfg.peer_pfx = peer_pfx;
4832         cfg.plen = ifm->ifa_prefixlen;
4833         if (tb[IFA_RT_PRIORITY])
4834                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4835
4836         cfg.valid_lft = INFINITY_LIFE_TIME;
4837         cfg.preferred_lft = INFINITY_LIFE_TIME;
4838
4839         if (tb[IFA_CACHEINFO]) {
4840                 struct ifa_cacheinfo *ci;
4841
4842                 ci = nla_data(tb[IFA_CACHEINFO]);
4843                 cfg.valid_lft = ci->ifa_valid;
4844                 cfg.preferred_lft = ci->ifa_prefered;
4845         }
4846
4847         dev =  __dev_get_by_index(net, ifm->ifa_index);
4848         if (!dev)
4849                 return -ENODEV;
4850
4851         if (tb[IFA_FLAGS])
4852                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4853         else
4854                 cfg.ifa_flags = ifm->ifa_flags;
4855
4856         /* We ignore other flags so far. */
4857         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4858                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4859                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4860
4861         idev = ipv6_find_idev(dev);
4862         if (IS_ERR(idev))
4863                 return PTR_ERR(idev);
4864
4865         if (!ipv6_allow_optimistic_dad(net, idev))
4866                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4867
4868         if (cfg.ifa_flags & IFA_F_NODAD &&
4869             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4870                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4871                 return -EINVAL;
4872         }
4873
4874         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4875         if (!ifa) {
4876                 /*
4877                  * It would be best to check for !NLM_F_CREATE here but
4878                  * userspace already relies on not having to provide this.
4879                  */
4880                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4881         }
4882
4883         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4884             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4885                 err = -EEXIST;
4886         else
4887                 err = inet6_addr_modify(ifa, &cfg);
4888
4889         in6_ifa_put(ifa);
4890
4891         return err;
4892 }
4893
4894 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4895                           u8 scope, int ifindex)
4896 {
4897         struct ifaddrmsg *ifm;
4898
4899         ifm = nlmsg_data(nlh);
4900         ifm->ifa_family = AF_INET6;
4901         ifm->ifa_prefixlen = prefixlen;
4902         ifm->ifa_flags = flags;
4903         ifm->ifa_scope = scope;
4904         ifm->ifa_index = ifindex;
4905 }
4906
4907 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4908                          unsigned long tstamp, u32 preferred, u32 valid)
4909 {
4910         struct ifa_cacheinfo ci;
4911
4912         ci.cstamp = cstamp_delta(cstamp);
4913         ci.tstamp = cstamp_delta(tstamp);
4914         ci.ifa_prefered = preferred;
4915         ci.ifa_valid = valid;
4916
4917         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4918 }
4919
4920 static inline int rt_scope(int ifa_scope)
4921 {
4922         if (ifa_scope & IFA_HOST)
4923                 return RT_SCOPE_HOST;
4924         else if (ifa_scope & IFA_LINK)
4925                 return RT_SCOPE_LINK;
4926         else if (ifa_scope & IFA_SITE)
4927                 return RT_SCOPE_SITE;
4928         else
4929                 return RT_SCOPE_UNIVERSE;
4930 }
4931
4932 static inline int inet6_ifaddr_msgsize(void)
4933 {
4934         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4935                + nla_total_size(16) /* IFA_LOCAL */
4936                + nla_total_size(16) /* IFA_ADDRESS */
4937                + nla_total_size(sizeof(struct ifa_cacheinfo))
4938                + nla_total_size(4)  /* IFA_FLAGS */
4939                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4940 }
4941
4942 enum addr_type_t {
4943         UNICAST_ADDR,
4944         MULTICAST_ADDR,
4945         ANYCAST_ADDR,
4946 };
4947
4948 struct inet6_fill_args {
4949         u32 portid;
4950         u32 seq;
4951         int event;
4952         unsigned int flags;
4953         int netnsid;
4954         int ifindex;
4955         enum addr_type_t type;
4956 };
4957
4958 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4959                              struct inet6_fill_args *args)
4960 {
4961         struct nlmsghdr  *nlh;
4962         u32 preferred, valid;
4963
4964         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4965                         sizeof(struct ifaddrmsg), args->flags);
4966         if (!nlh)
4967                 return -EMSGSIZE;
4968
4969         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4970                       ifa->idev->dev->ifindex);
4971
4972         if (args->netnsid >= 0 &&
4973             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4974                 goto error;
4975
4976         if (!((ifa->flags&IFA_F_PERMANENT) &&
4977               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4978                 preferred = ifa->prefered_lft;
4979                 valid = ifa->valid_lft;
4980                 if (preferred != INFINITY_LIFE_TIME) {
4981                         long tval = (jiffies - ifa->tstamp)/HZ;
4982                         if (preferred > tval)
4983                                 preferred -= tval;
4984                         else
4985                                 preferred = 0;
4986                         if (valid != INFINITY_LIFE_TIME) {
4987                                 if (valid > tval)
4988                                         valid -= tval;
4989                                 else
4990                                         valid = 0;
4991                         }
4992                 }
4993         } else {
4994                 preferred = INFINITY_LIFE_TIME;
4995                 valid = INFINITY_LIFE_TIME;
4996         }
4997
4998         if (!ipv6_addr_any(&ifa->peer_addr)) {
4999                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5000                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5001                         goto error;
5002         } else
5003                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5004                         goto error;
5005
5006         if (ifa->rt_priority &&
5007             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5008                 goto error;
5009
5010         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5011                 goto error;
5012
5013         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5014                 goto error;
5015
5016         nlmsg_end(skb, nlh);
5017         return 0;
5018
5019 error:
5020         nlmsg_cancel(skb, nlh);
5021         return -EMSGSIZE;
5022 }
5023
5024 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5025                                struct inet6_fill_args *args)
5026 {
5027         struct nlmsghdr  *nlh;
5028         u8 scope = RT_SCOPE_UNIVERSE;
5029         int ifindex = ifmca->idev->dev->ifindex;
5030
5031         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5032                 scope = RT_SCOPE_SITE;
5033
5034         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5035                         sizeof(struct ifaddrmsg), args->flags);
5036         if (!nlh)
5037                 return -EMSGSIZE;
5038
5039         if (args->netnsid >= 0 &&
5040             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5041                 nlmsg_cancel(skb, nlh);
5042                 return -EMSGSIZE;
5043         }
5044
5045         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5046         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5047             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5048                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5049                 nlmsg_cancel(skb, nlh);
5050                 return -EMSGSIZE;
5051         }
5052
5053         nlmsg_end(skb, nlh);
5054         return 0;
5055 }
5056
5057 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5058                                struct inet6_fill_args *args)
5059 {
5060         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5061         int ifindex = dev ? dev->ifindex : 1;
5062         struct nlmsghdr  *nlh;
5063         u8 scope = RT_SCOPE_UNIVERSE;
5064
5065         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5066                 scope = RT_SCOPE_SITE;
5067
5068         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5069                         sizeof(struct ifaddrmsg), args->flags);
5070         if (!nlh)
5071                 return -EMSGSIZE;
5072
5073         if (args->netnsid >= 0 &&
5074             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5075                 nlmsg_cancel(skb, nlh);
5076                 return -EMSGSIZE;
5077         }
5078
5079         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5080         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5081             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5082                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5083                 nlmsg_cancel(skb, nlh);
5084                 return -EMSGSIZE;
5085         }
5086
5087         nlmsg_end(skb, nlh);
5088         return 0;
5089 }
5090
5091 /* called with rcu_read_lock() */
5092 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5093                           struct netlink_callback *cb, int s_ip_idx,
5094                           struct inet6_fill_args *fillargs)
5095 {
5096         struct ifmcaddr6 *ifmca;
5097         struct ifacaddr6 *ifaca;
5098         int ip_idx = 0;
5099         int err = 1;
5100
5101         read_lock_bh(&idev->lock);
5102         switch (fillargs->type) {
5103         case UNICAST_ADDR: {
5104                 struct inet6_ifaddr *ifa;
5105                 fillargs->event = RTM_NEWADDR;
5106
5107                 /* unicast address incl. temp addr */
5108                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5109                         if (ip_idx < s_ip_idx)
5110                                 goto next;
5111                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5112                         if (err < 0)
5113                                 break;
5114                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5115 next:
5116                         ip_idx++;
5117                 }
5118                 break;
5119         }
5120         case MULTICAST_ADDR:
5121                 read_unlock_bh(&idev->lock);
5122                 fillargs->event = RTM_GETMULTICAST;
5123
5124                 /* multicast address */
5125                 for (ifmca = rcu_dereference(idev->mc_list);
5126                      ifmca;
5127                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5128                         if (ip_idx < s_ip_idx)
5129                                 continue;
5130                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5131                         if (err < 0)
5132                                 break;
5133                 }
5134                 read_lock_bh(&idev->lock);
5135                 break;
5136         case ANYCAST_ADDR:
5137                 fillargs->event = RTM_GETANYCAST;
5138                 /* anycast address */
5139                 for (ifaca = idev->ac_list; ifaca;
5140                      ifaca = ifaca->aca_next, ip_idx++) {
5141                         if (ip_idx < s_ip_idx)
5142                                 continue;
5143                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5144                         if (err < 0)
5145                                 break;
5146                 }
5147                 break;
5148         default:
5149                 break;
5150         }
5151         read_unlock_bh(&idev->lock);
5152         cb->args[2] = ip_idx;
5153         return err;
5154 }
5155
5156 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5157                                        struct inet6_fill_args *fillargs,
5158                                        struct net **tgt_net, struct sock *sk,
5159                                        struct netlink_callback *cb)
5160 {
5161         struct netlink_ext_ack *extack = cb->extack;
5162         struct nlattr *tb[IFA_MAX+1];
5163         struct ifaddrmsg *ifm;
5164         int err, i;
5165
5166         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5167                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5168                 return -EINVAL;
5169         }
5170
5171         ifm = nlmsg_data(nlh);
5172         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5173                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5174                 return -EINVAL;
5175         }
5176
5177         fillargs->ifindex = ifm->ifa_index;
5178         if (fillargs->ifindex) {
5179                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5180                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5181         }
5182
5183         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5184                                             ifa_ipv6_policy, extack);
5185         if (err < 0)
5186                 return err;
5187
5188         for (i = 0; i <= IFA_MAX; ++i) {
5189                 if (!tb[i])
5190                         continue;
5191
5192                 if (i == IFA_TARGET_NETNSID) {
5193                         struct net *net;
5194
5195                         fillargs->netnsid = nla_get_s32(tb[i]);
5196                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5197                         if (IS_ERR(net)) {
5198                                 fillargs->netnsid = -1;
5199                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5200                                 return PTR_ERR(net);
5201                         }
5202                         *tgt_net = net;
5203                 } else {
5204                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5205                         return -EINVAL;
5206                 }
5207         }
5208
5209         return 0;
5210 }
5211
5212 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5213                            enum addr_type_t type)
5214 {
5215         const struct nlmsghdr *nlh = cb->nlh;
5216         struct inet6_fill_args fillargs = {
5217                 .portid = NETLINK_CB(cb->skb).portid,
5218                 .seq = cb->nlh->nlmsg_seq,
5219                 .flags = NLM_F_MULTI,
5220                 .netnsid = -1,
5221                 .type = type,
5222         };
5223         struct net *tgt_net = sock_net(skb->sk);
5224         int idx, s_idx, s_ip_idx;
5225         int h, s_h;
5226         struct net_device *dev;
5227         struct inet6_dev *idev;
5228         struct hlist_head *head;
5229         int err = 0;
5230
5231         s_h = cb->args[0];
5232         s_idx = idx = cb->args[1];
5233         s_ip_idx = cb->args[2];
5234
5235         if (cb->strict_check) {
5236                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5237                                                   skb->sk, cb);
5238                 if (err < 0)
5239                         goto put_tgt_net;
5240
5241                 err = 0;
5242                 if (fillargs.ifindex) {
5243                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5244                         if (!dev) {
5245                                 err = -ENODEV;
5246                                 goto put_tgt_net;
5247                         }
5248                         idev = __in6_dev_get(dev);
5249                         if (idev) {
5250                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5251                                                      &fillargs);
5252                                 if (err > 0)
5253                                         err = 0;
5254                         }
5255                         goto put_tgt_net;
5256                 }
5257         }
5258
5259         rcu_read_lock();
5260         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5261         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5262                 idx = 0;
5263                 head = &tgt_net->dev_index_head[h];
5264                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5265                         if (idx < s_idx)
5266                                 goto cont;
5267                         if (h > s_h || idx > s_idx)
5268                                 s_ip_idx = 0;
5269                         idev = __in6_dev_get(dev);
5270                         if (!idev)
5271                                 goto cont;
5272
5273                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5274                                            &fillargs) < 0)
5275                                 goto done;
5276 cont:
5277                         idx++;
5278                 }
5279         }
5280 done:
5281         rcu_read_unlock();
5282         cb->args[0] = h;
5283         cb->args[1] = idx;
5284 put_tgt_net:
5285         if (fillargs.netnsid >= 0)
5286                 put_net(tgt_net);
5287
5288         return skb->len ? : err;
5289 }
5290
5291 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5292 {
5293         enum addr_type_t type = UNICAST_ADDR;
5294
5295         return inet6_dump_addr(skb, cb, type);
5296 }
5297
5298 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5299 {
5300         enum addr_type_t type = MULTICAST_ADDR;
5301
5302         return inet6_dump_addr(skb, cb, type);
5303 }
5304
5305
5306 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5307 {
5308         enum addr_type_t type = ANYCAST_ADDR;
5309
5310         return inet6_dump_addr(skb, cb, type);
5311 }
5312
5313 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5314                                        const struct nlmsghdr *nlh,
5315                                        struct nlattr **tb,
5316                                        struct netlink_ext_ack *extack)
5317 {
5318         struct ifaddrmsg *ifm;
5319         int i, err;
5320
5321         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5322                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5323                 return -EINVAL;
5324         }
5325
5326         if (!netlink_strict_get_check(skb))
5327                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5328                                               ifa_ipv6_policy, extack);
5329
5330         ifm = nlmsg_data(nlh);
5331         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5332                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5333                 return -EINVAL;
5334         }
5335
5336         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5337                                             ifa_ipv6_policy, extack);
5338         if (err)
5339                 return err;
5340
5341         for (i = 0; i <= IFA_MAX; i++) {
5342                 if (!tb[i])
5343                         continue;
5344
5345                 switch (i) {
5346                 case IFA_TARGET_NETNSID:
5347                 case IFA_ADDRESS:
5348                 case IFA_LOCAL:
5349                         break;
5350                 default:
5351                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5352                         return -EINVAL;
5353                 }
5354         }
5355
5356         return 0;
5357 }
5358
5359 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5360                              struct netlink_ext_ack *extack)
5361 {
5362         struct net *tgt_net = sock_net(in_skb->sk);
5363         struct inet6_fill_args fillargs = {
5364                 .portid = NETLINK_CB(in_skb).portid,
5365                 .seq = nlh->nlmsg_seq,
5366                 .event = RTM_NEWADDR,
5367                 .flags = 0,
5368                 .netnsid = -1,
5369         };
5370         struct ifaddrmsg *ifm;
5371         struct nlattr *tb[IFA_MAX+1];
5372         struct in6_addr *addr = NULL, *peer;
5373         struct net_device *dev = NULL;
5374         struct inet6_ifaddr *ifa;
5375         struct sk_buff *skb;
5376         int err;
5377
5378         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5379         if (err < 0)
5380                 return err;
5381
5382         if (tb[IFA_TARGET_NETNSID]) {
5383                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5384
5385                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5386                                                   fillargs.netnsid);
5387                 if (IS_ERR(tgt_net))
5388                         return PTR_ERR(tgt_net);
5389         }
5390
5391         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5392         if (!addr)
5393                 return -EINVAL;
5394
5395         ifm = nlmsg_data(nlh);
5396         if (ifm->ifa_index)
5397                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5398
5399         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5400         if (!ifa) {
5401                 err = -EADDRNOTAVAIL;
5402                 goto errout;
5403         }
5404
5405         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5406         if (!skb) {
5407                 err = -ENOBUFS;
5408                 goto errout_ifa;
5409         }
5410
5411         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5412         if (err < 0) {
5413                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5414                 WARN_ON(err == -EMSGSIZE);
5415                 kfree_skb(skb);
5416                 goto errout_ifa;
5417         }
5418         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5419 errout_ifa:
5420         in6_ifa_put(ifa);
5421 errout:
5422         if (dev)
5423                 dev_put(dev);
5424         if (fillargs.netnsid >= 0)
5425                 put_net(tgt_net);
5426
5427         return err;
5428 }
5429
5430 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5431 {
5432         struct sk_buff *skb;
5433         struct net *net = dev_net(ifa->idev->dev);
5434         struct inet6_fill_args fillargs = {
5435                 .portid = 0,
5436                 .seq = 0,
5437                 .event = event,
5438                 .flags = 0,
5439                 .netnsid = -1,
5440         };
5441         int err = -ENOBUFS;
5442
5443         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5444         if (!skb)
5445                 goto errout;
5446
5447         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5448         if (err < 0) {
5449                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5450                 WARN_ON(err == -EMSGSIZE);
5451                 kfree_skb(skb);
5452                 goto errout;
5453         }
5454         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5455         return;
5456 errout:
5457         if (err < 0)
5458                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5459 }
5460
5461 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5462                                 __s32 *array, int bytes)
5463 {
5464         BUG_ON(bytes < (DEVCONF_MAX * 4));
5465
5466         memset(array, 0, bytes);
5467         array[DEVCONF_FORWARDING] = cnf->forwarding;
5468         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5469         array[DEVCONF_MTU6] = cnf->mtu6;
5470         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5471         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5472         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5473         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5474         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5475         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5476                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5477         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5478                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5479         array[DEVCONF_RTR_SOLICIT_DELAY] =
5480                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5481         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5482         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5483                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5484         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5485                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5486         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5487         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5488         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5489         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5490         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5491         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5492         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5493         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5494         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5495         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5496 #ifdef CONFIG_IPV6_ROUTER_PREF
5497         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5498         array[DEVCONF_RTR_PROBE_INTERVAL] =
5499                 jiffies_to_msecs(cnf->rtr_probe_interval);
5500 #ifdef CONFIG_IPV6_ROUTE_INFO
5501         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5502         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5503 #endif
5504 #endif
5505         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5506         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5507 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5508         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5509         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5510 #endif
5511 #ifdef CONFIG_IPV6_MROUTE
5512         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5513 #endif
5514         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5515         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5516         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5517         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5518         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5519         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5520         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5521         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5522         /* we omit DEVCONF_STABLE_SECRET for now */
5523         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5524         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5525         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5526         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5527         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5528 #ifdef CONFIG_IPV6_SEG6_HMAC
5529         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5530 #endif
5531         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5532         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5533         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5534         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5535         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5536         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5537         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5538         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5539 }
5540
5541 static inline size_t inet6_ifla6_size(void)
5542 {
5543         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5544              + nla_total_size(sizeof(struct ifla_cacheinfo))
5545              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5546              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5547              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5548              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5549              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5550              + 0;
5551 }
5552
5553 static inline size_t inet6_if_nlmsg_size(void)
5554 {
5555         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5556                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5557                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5558                + nla_total_size(4) /* IFLA_MTU */
5559                + nla_total_size(4) /* IFLA_LINK */
5560                + nla_total_size(1) /* IFLA_OPERSTATE */
5561                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5562 }
5563
5564 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5565                                         int bytes)
5566 {
5567         int i;
5568         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5569         BUG_ON(pad < 0);
5570
5571         /* Use put_unaligned() because stats may not be aligned for u64. */
5572         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5573         for (i = 1; i < ICMP6_MIB_MAX; i++)
5574                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5575
5576         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5577 }
5578
5579 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5580                                         int bytes, size_t syncpoff)
5581 {
5582         int i, c;
5583         u64 buff[IPSTATS_MIB_MAX];
5584         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5585
5586         BUG_ON(pad < 0);
5587
5588         memset(buff, 0, sizeof(buff));
5589         buff[0] = IPSTATS_MIB_MAX;
5590
5591         for_each_possible_cpu(c) {
5592                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5593                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5594         }
5595
5596         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5597         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5598 }
5599
5600 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5601                              int bytes)
5602 {
5603         switch (attrtype) {
5604         case IFLA_INET6_STATS:
5605                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5606                                      offsetof(struct ipstats_mib, syncp));
5607                 break;
5608         case IFLA_INET6_ICMP6STATS:
5609                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5610                 break;
5611         }
5612 }
5613
5614 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5615                                   u32 ext_filter_mask)
5616 {
5617         struct nlattr *nla;
5618         struct ifla_cacheinfo ci;
5619
5620         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5621                 goto nla_put_failure;
5622         ci.max_reasm_len = IPV6_MAXPLEN;
5623         ci.tstamp = cstamp_delta(idev->tstamp);
5624         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5625         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5626         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5627                 goto nla_put_failure;
5628         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5629         if (!nla)
5630                 goto nla_put_failure;
5631         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5632
5633         /* XXX - MC not implemented */
5634
5635         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5636                 return 0;
5637
5638         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5639         if (!nla)
5640                 goto nla_put_failure;
5641         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5642
5643         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5644         if (!nla)
5645                 goto nla_put_failure;
5646         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5647
5648         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5649         if (!nla)
5650                 goto nla_put_failure;
5651         read_lock_bh(&idev->lock);
5652         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5653         read_unlock_bh(&idev->lock);
5654
5655         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5656                 goto nla_put_failure;
5657
5658         return 0;
5659
5660 nla_put_failure:
5661         return -EMSGSIZE;
5662 }
5663
5664 static size_t inet6_get_link_af_size(const struct net_device *dev,
5665                                      u32 ext_filter_mask)
5666 {
5667         if (!__in6_dev_get(dev))
5668                 return 0;
5669
5670         return inet6_ifla6_size();
5671 }
5672
5673 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5674                               u32 ext_filter_mask)
5675 {
5676         struct inet6_dev *idev = __in6_dev_get(dev);
5677
5678         if (!idev)
5679                 return -ENODATA;
5680
5681         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5682                 return -EMSGSIZE;
5683
5684         return 0;
5685 }
5686
5687 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5688                              struct netlink_ext_ack *extack)
5689 {
5690         struct inet6_ifaddr *ifp;
5691         struct net_device *dev = idev->dev;
5692         bool clear_token, update_rs = false;
5693         struct in6_addr ll_addr;
5694
5695         ASSERT_RTNL();
5696
5697         if (!token)
5698                 return -EINVAL;
5699
5700         if (dev->flags & IFF_LOOPBACK) {
5701                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5702                 return -EINVAL;
5703         }
5704
5705         if (dev->flags & IFF_NOARP) {
5706                 NL_SET_ERR_MSG_MOD(extack,
5707                                    "Device does not do neighbour discovery");
5708                 return -EINVAL;
5709         }
5710
5711         if (!ipv6_accept_ra(idev)) {
5712                 NL_SET_ERR_MSG_MOD(extack,
5713                                    "Router advertisement is disabled on device");
5714                 return -EINVAL;
5715         }
5716
5717         if (idev->cnf.rtr_solicits == 0) {
5718                 NL_SET_ERR_MSG(extack,
5719                                "Router solicitation is disabled on device");
5720                 return -EINVAL;
5721         }
5722
5723         write_lock_bh(&idev->lock);
5724
5725         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5726         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5727
5728         write_unlock_bh(&idev->lock);
5729
5730         clear_token = ipv6_addr_any(token);
5731         if (clear_token)
5732                 goto update_lft;
5733
5734         if (!idev->dead && (idev->if_flags & IF_READY) &&
5735             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5736                              IFA_F_OPTIMISTIC)) {
5737                 /* If we're not ready, then normal ifup will take care
5738                  * of this. Otherwise, we need to request our rs here.
5739                  */
5740                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5741                 update_rs = true;
5742         }
5743
5744 update_lft:
5745         write_lock_bh(&idev->lock);
5746
5747         if (update_rs) {
5748                 idev->if_flags |= IF_RS_SENT;
5749                 idev->rs_interval = rfc3315_s14_backoff_init(
5750                         idev->cnf.rtr_solicit_interval);
5751                 idev->rs_probes = 1;
5752                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5753         }
5754
5755         /* Well, that's kinda nasty ... */
5756         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5757                 spin_lock(&ifp->lock);
5758                 if (ifp->tokenized) {
5759                         ifp->valid_lft = 0;
5760                         ifp->prefered_lft = 0;
5761                 }
5762                 spin_unlock(&ifp->lock);
5763         }
5764
5765         write_unlock_bh(&idev->lock);
5766         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5767         addrconf_verify_rtnl();
5768         return 0;
5769 }
5770
5771 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5772         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5773         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5774 };
5775
5776 static int check_addr_gen_mode(int mode)
5777 {
5778         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5779             mode != IN6_ADDR_GEN_MODE_NONE &&
5780             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5781             mode != IN6_ADDR_GEN_MODE_RANDOM)
5782                 return -EINVAL;
5783         return 1;
5784 }
5785
5786 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5787                                 int mode)
5788 {
5789         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5790             !idev->cnf.stable_secret.initialized &&
5791             !net->ipv6.devconf_dflt->stable_secret.initialized)
5792                 return -EINVAL;
5793         return 1;
5794 }
5795
5796 static int inet6_validate_link_af(const struct net_device *dev,
5797                                   const struct nlattr *nla)
5798 {
5799         struct nlattr *tb[IFLA_INET6_MAX + 1];
5800         struct inet6_dev *idev = NULL;
5801         int err;
5802
5803         if (dev) {
5804                 idev = __in6_dev_get(dev);
5805                 if (!idev)
5806                         return -EAFNOSUPPORT;
5807         }
5808
5809         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5810                                           inet6_af_policy, NULL);
5811         if (err)
5812                 return err;
5813
5814         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5815                 return -EINVAL;
5816
5817         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5818                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5819
5820                 if (check_addr_gen_mode(mode) < 0)
5821                         return -EINVAL;
5822                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5823                         return -EINVAL;
5824         }
5825
5826         return 0;
5827 }
5828
5829 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5830                              struct netlink_ext_ack *extack)
5831 {
5832         struct inet6_dev *idev = __in6_dev_get(dev);
5833         struct nlattr *tb[IFLA_INET6_MAX + 1];
5834         int err;
5835
5836         if (!idev)
5837                 return -EAFNOSUPPORT;
5838
5839         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5840                 return -EINVAL;
5841
5842         if (tb[IFLA_INET6_TOKEN]) {
5843                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5844                                         extack);
5845                 if (err)
5846                         return err;
5847         }
5848
5849         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5850                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5851
5852                 idev->cnf.addr_gen_mode = mode;
5853         }
5854
5855         return 0;
5856 }
5857
5858 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5859                              u32 portid, u32 seq, int event, unsigned int flags)
5860 {
5861         struct net_device *dev = idev->dev;
5862         struct ifinfomsg *hdr;
5863         struct nlmsghdr *nlh;
5864         void *protoinfo;
5865
5866         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5867         if (!nlh)
5868                 return -EMSGSIZE;
5869
5870         hdr = nlmsg_data(nlh);
5871         hdr->ifi_family = AF_INET6;
5872         hdr->__ifi_pad = 0;
5873         hdr->ifi_type = dev->type;
5874         hdr->ifi_index = dev->ifindex;
5875         hdr->ifi_flags = dev_get_flags(dev);
5876         hdr->ifi_change = 0;
5877
5878         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5879             (dev->addr_len &&
5880              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5881             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5882             (dev->ifindex != dev_get_iflink(dev) &&
5883              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5884             nla_put_u8(skb, IFLA_OPERSTATE,
5885                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5886                 goto nla_put_failure;
5887         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5888         if (!protoinfo)
5889                 goto nla_put_failure;
5890
5891         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5892                 goto nla_put_failure;
5893
5894         nla_nest_end(skb, protoinfo);
5895         nlmsg_end(skb, nlh);
5896         return 0;
5897
5898 nla_put_failure:
5899         nlmsg_cancel(skb, nlh);
5900         return -EMSGSIZE;
5901 }
5902
5903 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5904                                    struct netlink_ext_ack *extack)
5905 {
5906         struct ifinfomsg *ifm;
5907
5908         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5909                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5910                 return -EINVAL;
5911         }
5912
5913         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5914                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5915                 return -EINVAL;
5916         }
5917
5918         ifm = nlmsg_data(nlh);
5919         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5920             ifm->ifi_change || ifm->ifi_index) {
5921                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5922                 return -EINVAL;
5923         }
5924
5925         return 0;
5926 }
5927
5928 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5929 {
5930         struct net *net = sock_net(skb->sk);
5931         int h, s_h;
5932         int idx = 0, s_idx;
5933         struct net_device *dev;
5934         struct inet6_dev *idev;
5935         struct hlist_head *head;
5936
5937         /* only requests using strict checking can pass data to
5938          * influence the dump
5939          */
5940         if (cb->strict_check) {
5941                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5942
5943                 if (err < 0)
5944                         return err;
5945         }
5946
5947         s_h = cb->args[0];
5948         s_idx = cb->args[1];
5949
5950         rcu_read_lock();
5951         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5952                 idx = 0;
5953                 head = &net->dev_index_head[h];
5954                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5955                         if (idx < s_idx)
5956                                 goto cont;
5957                         idev = __in6_dev_get(dev);
5958                         if (!idev)
5959                                 goto cont;
5960                         if (inet6_fill_ifinfo(skb, idev,
5961                                               NETLINK_CB(cb->skb).portid,
5962                                               cb->nlh->nlmsg_seq,
5963                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5964                                 goto out;
5965 cont:
5966                         idx++;
5967                 }
5968         }
5969 out:
5970         rcu_read_unlock();
5971         cb->args[1] = idx;
5972         cb->args[0] = h;
5973
5974         return skb->len;
5975 }
5976
5977 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5978 {
5979         struct sk_buff *skb;
5980         struct net *net = dev_net(idev->dev);
5981         int err = -ENOBUFS;
5982
5983         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5984         if (!skb)
5985                 goto errout;
5986
5987         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5988         if (err < 0) {
5989                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5990                 WARN_ON(err == -EMSGSIZE);
5991                 kfree_skb(skb);
5992                 goto errout;
5993         }
5994         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5995         return;
5996 errout:
5997         if (err < 0)
5998                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5999 }
6000
6001 static inline size_t inet6_prefix_nlmsg_size(void)
6002 {
6003         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6004                + nla_total_size(sizeof(struct in6_addr))
6005                + nla_total_size(sizeof(struct prefix_cacheinfo));
6006 }
6007
6008 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6009                              struct prefix_info *pinfo, u32 portid, u32 seq,
6010                              int event, unsigned int flags)
6011 {
6012         struct prefixmsg *pmsg;
6013         struct nlmsghdr *nlh;
6014         struct prefix_cacheinfo ci;
6015
6016         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6017         if (!nlh)
6018                 return -EMSGSIZE;
6019
6020         pmsg = nlmsg_data(nlh);
6021         pmsg->prefix_family = AF_INET6;
6022         pmsg->prefix_pad1 = 0;
6023         pmsg->prefix_pad2 = 0;
6024         pmsg->prefix_ifindex = idev->dev->ifindex;
6025         pmsg->prefix_len = pinfo->prefix_len;
6026         pmsg->prefix_type = pinfo->type;
6027         pmsg->prefix_pad3 = 0;
6028         pmsg->prefix_flags = 0;
6029         if (pinfo->onlink)
6030                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6031         if (pinfo->autoconf)
6032                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6033
6034         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6035                 goto nla_put_failure;
6036         ci.preferred_time = ntohl(pinfo->prefered);
6037         ci.valid_time = ntohl(pinfo->valid);
6038         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6039                 goto nla_put_failure;
6040         nlmsg_end(skb, nlh);
6041         return 0;
6042
6043 nla_put_failure:
6044         nlmsg_cancel(skb, nlh);
6045         return -EMSGSIZE;
6046 }
6047
6048 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6049                          struct prefix_info *pinfo)
6050 {
6051         struct sk_buff *skb;
6052         struct net *net = dev_net(idev->dev);
6053         int err = -ENOBUFS;
6054
6055         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6056         if (!skb)
6057                 goto errout;
6058
6059         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6060         if (err < 0) {
6061                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6062                 WARN_ON(err == -EMSGSIZE);
6063                 kfree_skb(skb);
6064                 goto errout;
6065         }
6066         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6067         return;
6068 errout:
6069         if (err < 0)
6070                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6071 }
6072
6073 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6074 {
6075         struct net *net = dev_net(ifp->idev->dev);
6076
6077         if (event)
6078                 ASSERT_RTNL();
6079
6080         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6081
6082         switch (event) {
6083         case RTM_NEWADDR:
6084                 /*
6085                  * If the address was optimistic we inserted the route at the
6086                  * start of our DAD process, so we don't need to do it again.
6087                  * If the device was taken down in the middle of the DAD
6088                  * cycle there is a race where we could get here without a
6089                  * host route, so nothing to insert. That will be fixed when
6090                  * the device is brought up.
6091                  */
6092                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6093                         ip6_ins_rt(net, ifp->rt);
6094                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6095                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6096                                 &ifp->addr, ifp->idev->dev->name);
6097                 }
6098
6099                 if (ifp->idev->cnf.forwarding)
6100                         addrconf_join_anycast(ifp);
6101                 if (!ipv6_addr_any(&ifp->peer_addr))
6102                         addrconf_prefix_route(&ifp->peer_addr, 128,
6103                                               ifp->rt_priority, ifp->idev->dev,
6104                                               0, 0, GFP_ATOMIC);
6105                 break;
6106         case RTM_DELADDR:
6107                 if (ifp->idev->cnf.forwarding)
6108                         addrconf_leave_anycast(ifp);
6109                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6110                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6111                         struct fib6_info *rt;
6112
6113                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6114                                                        ifp->idev->dev, 0, 0,
6115                                                        false);
6116                         if (rt)
6117                                 ip6_del_rt(net, rt, false);
6118                 }
6119                 if (ifp->rt) {
6120                         ip6_del_rt(net, ifp->rt, false);
6121                         ifp->rt = NULL;
6122                 }
6123                 rt_genid_bump_ipv6(net);
6124                 break;
6125         }
6126         atomic_inc(&net->ipv6.dev_addr_genid);
6127 }
6128
6129 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6130 {
6131         if (likely(ifp->idev->dead == 0))
6132                 __ipv6_ifa_notify(event, ifp);
6133 }
6134
6135 #ifdef CONFIG_SYSCTL
6136
6137 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6138                 void *buffer, size_t *lenp, loff_t *ppos)
6139 {
6140         int *valp = ctl->data;
6141         int val = *valp;
6142         loff_t pos = *ppos;
6143         struct ctl_table lctl;
6144         int ret;
6145
6146         /*
6147          * ctl->data points to idev->cnf.forwarding, we should
6148          * not modify it until we get the rtnl lock.
6149          */
6150         lctl = *ctl;
6151         lctl.data = &val;
6152
6153         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6154
6155         if (write)
6156                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6157         if (ret)
6158                 *ppos = pos;
6159         return ret;
6160 }
6161
6162 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6163                 void *buffer, size_t *lenp, loff_t *ppos)
6164 {
6165         struct inet6_dev *idev = ctl->extra1;
6166         int min_mtu = IPV6_MIN_MTU;
6167         struct ctl_table lctl;
6168
6169         lctl = *ctl;
6170         lctl.extra1 = &min_mtu;
6171         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6172
6173         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6174 }
6175
6176 static void dev_disable_change(struct inet6_dev *idev)
6177 {
6178         struct netdev_notifier_info info;
6179
6180         if (!idev || !idev->dev)
6181                 return;
6182
6183         netdev_notifier_info_init(&info, idev->dev);
6184         if (idev->cnf.disable_ipv6)
6185                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6186         else
6187                 addrconf_notify(NULL, NETDEV_UP, &info);
6188 }
6189
6190 static void addrconf_disable_change(struct net *net, __s32 newf)
6191 {
6192         struct net_device *dev;
6193         struct inet6_dev *idev;
6194
6195         for_each_netdev(net, dev) {
6196                 idev = __in6_dev_get(dev);
6197                 if (idev) {
6198                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6199                         idev->cnf.disable_ipv6 = newf;
6200                         if (changed)
6201                                 dev_disable_change(idev);
6202                 }
6203         }
6204 }
6205
6206 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6207 {
6208         struct net *net;
6209         int old;
6210
6211         if (!rtnl_trylock())
6212                 return restart_syscall();
6213
6214         net = (struct net *)table->extra2;
6215         old = *p;
6216         *p = newf;
6217
6218         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6219                 rtnl_unlock();
6220                 return 0;
6221         }
6222
6223         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6224                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6225                 addrconf_disable_change(net, newf);
6226         } else if ((!newf) ^ (!old))
6227                 dev_disable_change((struct inet6_dev *)table->extra1);
6228
6229         rtnl_unlock();
6230         return 0;
6231 }
6232
6233 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6234                 void *buffer, size_t *lenp, loff_t *ppos)
6235 {
6236         int *valp = ctl->data;
6237         int val = *valp;
6238         loff_t pos = *ppos;
6239         struct ctl_table lctl;
6240         int ret;
6241
6242         /*
6243          * ctl->data points to idev->cnf.disable_ipv6, we should
6244          * not modify it until we get the rtnl lock.
6245          */
6246         lctl = *ctl;
6247         lctl.data = &val;
6248
6249         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6250
6251         if (write)
6252                 ret = addrconf_disable_ipv6(ctl, valp, val);
6253         if (ret)
6254                 *ppos = pos;
6255         return ret;
6256 }
6257
6258 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6259                 void *buffer, size_t *lenp, loff_t *ppos)
6260 {
6261         int *valp = ctl->data;
6262         int ret;
6263         int old, new;
6264
6265         old = *valp;
6266         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6267         new = *valp;
6268
6269         if (write && old != new) {
6270                 struct net *net = ctl->extra2;
6271
6272                 if (!rtnl_trylock())
6273                         return restart_syscall();
6274
6275                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6276                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6277                                                      NETCONFA_PROXY_NEIGH,
6278                                                      NETCONFA_IFINDEX_DEFAULT,
6279                                                      net->ipv6.devconf_dflt);
6280                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6281                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6282                                                      NETCONFA_PROXY_NEIGH,
6283                                                      NETCONFA_IFINDEX_ALL,
6284                                                      net->ipv6.devconf_all);
6285                 else {
6286                         struct inet6_dev *idev = ctl->extra1;
6287
6288                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6289                                                      NETCONFA_PROXY_NEIGH,
6290                                                      idev->dev->ifindex,
6291                                                      &idev->cnf);
6292                 }
6293                 rtnl_unlock();
6294         }
6295
6296         return ret;
6297 }
6298
6299 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6300                                          void *buffer, size_t *lenp,
6301                                          loff_t *ppos)
6302 {
6303         int ret = 0;
6304         u32 new_val;
6305         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6306         struct net *net = (struct net *)ctl->extra2;
6307         struct ctl_table tmp = {
6308                 .data = &new_val,
6309                 .maxlen = sizeof(new_val),
6310                 .mode = ctl->mode,
6311         };
6312
6313         if (!rtnl_trylock())
6314                 return restart_syscall();
6315
6316         new_val = *((u32 *)ctl->data);
6317
6318         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6319         if (ret != 0)
6320                 goto out;
6321
6322         if (write) {
6323                 if (check_addr_gen_mode(new_val) < 0) {
6324                         ret = -EINVAL;
6325                         goto out;
6326                 }
6327
6328                 if (idev) {
6329                         if (check_stable_privacy(idev, net, new_val) < 0) {
6330                                 ret = -EINVAL;
6331                                 goto out;
6332                         }
6333
6334                         if (idev->cnf.addr_gen_mode != new_val) {
6335                                 idev->cnf.addr_gen_mode = new_val;
6336                                 addrconf_dev_config(idev->dev);
6337                         }
6338                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6339                         struct net_device *dev;
6340
6341                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6342                         for_each_netdev(net, dev) {
6343                                 idev = __in6_dev_get(dev);
6344                                 if (idev &&
6345                                     idev->cnf.addr_gen_mode != new_val) {
6346                                         idev->cnf.addr_gen_mode = new_val;
6347                                         addrconf_dev_config(idev->dev);
6348                                 }
6349                         }
6350                 }
6351
6352                 *((u32 *)ctl->data) = new_val;
6353         }
6354
6355 out:
6356         rtnl_unlock();
6357
6358         return ret;
6359 }
6360
6361 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6362                                          void *buffer, size_t *lenp,
6363                                          loff_t *ppos)
6364 {
6365         int err;
6366         struct in6_addr addr;
6367         char str[IPV6_MAX_STRLEN];
6368         struct ctl_table lctl = *ctl;
6369         struct net *net = ctl->extra2;
6370         struct ipv6_stable_secret *secret = ctl->data;
6371
6372         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6373                 return -EIO;
6374
6375         lctl.maxlen = IPV6_MAX_STRLEN;
6376         lctl.data = str;
6377
6378         if (!rtnl_trylock())
6379                 return restart_syscall();
6380
6381         if (!write && !secret->initialized) {
6382                 err = -EIO;
6383                 goto out;
6384         }
6385
6386         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6387         if (err >= sizeof(str)) {
6388                 err = -EIO;
6389                 goto out;
6390         }
6391
6392         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6393         if (err || !write)
6394                 goto out;
6395
6396         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6397                 err = -EIO;
6398                 goto out;
6399         }
6400
6401         secret->initialized = true;
6402         secret->secret = addr;
6403
6404         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6405                 struct net_device *dev;
6406
6407                 for_each_netdev(net, dev) {
6408                         struct inet6_dev *idev = __in6_dev_get(dev);
6409
6410                         if (idev) {
6411                                 idev->cnf.addr_gen_mode =
6412                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6413                         }
6414                 }
6415         } else {
6416                 struct inet6_dev *idev = ctl->extra1;
6417
6418                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6419         }
6420
6421 out:
6422         rtnl_unlock();
6423
6424         return err;
6425 }
6426
6427 static
6428 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6429                                                 int write, void *buffer,
6430                                                 size_t *lenp,
6431                                                 loff_t *ppos)
6432 {
6433         int *valp = ctl->data;
6434         int val = *valp;
6435         loff_t pos = *ppos;
6436         struct ctl_table lctl;
6437         int ret;
6438
6439         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6440          * we should not modify it until we get the rtnl lock.
6441          */
6442         lctl = *ctl;
6443         lctl.data = &val;
6444
6445         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6446
6447         if (write)
6448                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6449         if (ret)
6450                 *ppos = pos;
6451         return ret;
6452 }
6453
6454 static
6455 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6456 {
6457         if (rt) {
6458                 if (action)
6459                         rt->dst.flags |= DST_NOPOLICY;
6460                 else
6461                         rt->dst.flags &= ~DST_NOPOLICY;
6462         }
6463 }
6464
6465 static
6466 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6467 {
6468         struct inet6_ifaddr *ifa;
6469
6470         read_lock_bh(&idev->lock);
6471         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6472                 spin_lock(&ifa->lock);
6473                 if (ifa->rt) {
6474                         /* host routes only use builtin fib6_nh */
6475                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6476                         int cpu;
6477
6478                         rcu_read_lock();
6479                         ifa->rt->dst_nopolicy = val ? true : false;
6480                         if (nh->rt6i_pcpu) {
6481                                 for_each_possible_cpu(cpu) {
6482                                         struct rt6_info **rtp;
6483
6484                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6485                                         addrconf_set_nopolicy(*rtp, val);
6486                                 }
6487                         }
6488                         rcu_read_unlock();
6489                 }
6490                 spin_unlock(&ifa->lock);
6491         }
6492         read_unlock_bh(&idev->lock);
6493 }
6494
6495 static
6496 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6497 {
6498         struct inet6_dev *idev;
6499         struct net *net;
6500
6501         if (!rtnl_trylock())
6502                 return restart_syscall();
6503
6504         *valp = val;
6505
6506         net = (struct net *)ctl->extra2;
6507         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6508                 rtnl_unlock();
6509                 return 0;
6510         }
6511
6512         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6513                 struct net_device *dev;
6514
6515                 for_each_netdev(net, dev) {
6516                         idev = __in6_dev_get(dev);
6517                         if (idev)
6518                                 addrconf_disable_policy_idev(idev, val);
6519                 }
6520         } else {
6521                 idev = (struct inet6_dev *)ctl->extra1;
6522                 addrconf_disable_policy_idev(idev, val);
6523         }
6524
6525         rtnl_unlock();
6526         return 0;
6527 }
6528
6529 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6530                                    void *buffer, size_t *lenp, loff_t *ppos)
6531 {
6532         int *valp = ctl->data;
6533         int val = *valp;
6534         loff_t pos = *ppos;
6535         struct ctl_table lctl;
6536         int ret;
6537
6538         lctl = *ctl;
6539         lctl.data = &val;
6540         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6541
6542         if (write && (*valp != val))
6543                 ret = addrconf_disable_policy(ctl, valp, val);
6544
6545         if (ret)
6546                 *ppos = pos;
6547
6548         return ret;
6549 }
6550
6551 static int minus_one = -1;
6552 static const int two_five_five = 255;
6553
6554 static const struct ctl_table addrconf_sysctl[] = {
6555         {
6556                 .procname       = "forwarding",
6557                 .data           = &ipv6_devconf.forwarding,
6558                 .maxlen         = sizeof(int),
6559                 .mode           = 0644,
6560                 .proc_handler   = addrconf_sysctl_forward,
6561         },
6562         {
6563                 .procname       = "hop_limit",
6564                 .data           = &ipv6_devconf.hop_limit,
6565                 .maxlen         = sizeof(int),
6566                 .mode           = 0644,
6567                 .proc_handler   = proc_dointvec_minmax,
6568                 .extra1         = (void *)SYSCTL_ONE,
6569                 .extra2         = (void *)&two_five_five,
6570         },
6571         {
6572                 .procname       = "mtu",
6573                 .data           = &ipv6_devconf.mtu6,
6574                 .maxlen         = sizeof(int),
6575                 .mode           = 0644,
6576                 .proc_handler   = addrconf_sysctl_mtu,
6577         },
6578         {
6579                 .procname       = "accept_ra",
6580                 .data           = &ipv6_devconf.accept_ra,
6581                 .maxlen         = sizeof(int),
6582                 .mode           = 0644,
6583                 .proc_handler   = proc_dointvec,
6584         },
6585         {
6586                 .procname       = "accept_redirects",
6587                 .data           = &ipv6_devconf.accept_redirects,
6588                 .maxlen         = sizeof(int),
6589                 .mode           = 0644,
6590                 .proc_handler   = proc_dointvec,
6591         },
6592         {
6593                 .procname       = "autoconf",
6594                 .data           = &ipv6_devconf.autoconf,
6595                 .maxlen         = sizeof(int),
6596                 .mode           = 0644,
6597                 .proc_handler   = proc_dointvec,
6598         },
6599         {
6600                 .procname       = "dad_transmits",
6601                 .data           = &ipv6_devconf.dad_transmits,
6602                 .maxlen         = sizeof(int),
6603                 .mode           = 0644,
6604                 .proc_handler   = proc_dointvec,
6605         },
6606         {
6607                 .procname       = "router_solicitations",
6608                 .data           = &ipv6_devconf.rtr_solicits,
6609                 .maxlen         = sizeof(int),
6610                 .mode           = 0644,
6611                 .proc_handler   = proc_dointvec_minmax,
6612                 .extra1         = &minus_one,
6613         },
6614         {
6615                 .procname       = "router_solicitation_interval",
6616                 .data           = &ipv6_devconf.rtr_solicit_interval,
6617                 .maxlen         = sizeof(int),
6618                 .mode           = 0644,
6619                 .proc_handler   = proc_dointvec_jiffies,
6620         },
6621         {
6622                 .procname       = "router_solicitation_max_interval",
6623                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6624                 .maxlen         = sizeof(int),
6625                 .mode           = 0644,
6626                 .proc_handler   = proc_dointvec_jiffies,
6627         },
6628         {
6629                 .procname       = "router_solicitation_delay",
6630                 .data           = &ipv6_devconf.rtr_solicit_delay,
6631                 .maxlen         = sizeof(int),
6632                 .mode           = 0644,
6633                 .proc_handler   = proc_dointvec_jiffies,
6634         },
6635         {
6636                 .procname       = "force_mld_version",
6637                 .data           = &ipv6_devconf.force_mld_version,
6638                 .maxlen         = sizeof(int),
6639                 .mode           = 0644,
6640                 .proc_handler   = proc_dointvec,
6641         },
6642         {
6643                 .procname       = "mldv1_unsolicited_report_interval",
6644                 .data           =
6645                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6646                 .maxlen         = sizeof(int),
6647                 .mode           = 0644,
6648                 .proc_handler   = proc_dointvec_ms_jiffies,
6649         },
6650         {
6651                 .procname       = "mldv2_unsolicited_report_interval",
6652                 .data           =
6653                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6654                 .maxlen         = sizeof(int),
6655                 .mode           = 0644,
6656                 .proc_handler   = proc_dointvec_ms_jiffies,
6657         },
6658         {
6659                 .procname       = "use_tempaddr",
6660                 .data           = &ipv6_devconf.use_tempaddr,
6661                 .maxlen         = sizeof(int),
6662                 .mode           = 0644,
6663                 .proc_handler   = proc_dointvec,
6664         },
6665         {
6666                 .procname       = "temp_valid_lft",
6667                 .data           = &ipv6_devconf.temp_valid_lft,
6668                 .maxlen         = sizeof(int),
6669                 .mode           = 0644,
6670                 .proc_handler   = proc_dointvec,
6671         },
6672         {
6673                 .procname       = "temp_prefered_lft",
6674                 .data           = &ipv6_devconf.temp_prefered_lft,
6675                 .maxlen         = sizeof(int),
6676                 .mode           = 0644,
6677                 .proc_handler   = proc_dointvec,
6678         },
6679         {
6680                 .procname       = "regen_max_retry",
6681                 .data           = &ipv6_devconf.regen_max_retry,
6682                 .maxlen         = sizeof(int),
6683                 .mode           = 0644,
6684                 .proc_handler   = proc_dointvec,
6685         },
6686         {
6687                 .procname       = "max_desync_factor",
6688                 .data           = &ipv6_devconf.max_desync_factor,
6689                 .maxlen         = sizeof(int),
6690                 .mode           = 0644,
6691                 .proc_handler   = proc_dointvec,
6692         },
6693         {
6694                 .procname       = "max_addresses",
6695                 .data           = &ipv6_devconf.max_addresses,
6696                 .maxlen         = sizeof(int),
6697                 .mode           = 0644,
6698                 .proc_handler   = proc_dointvec,
6699         },
6700         {
6701                 .procname       = "accept_ra_defrtr",
6702                 .data           = &ipv6_devconf.accept_ra_defrtr,
6703                 .maxlen         = sizeof(int),
6704                 .mode           = 0644,
6705                 .proc_handler   = proc_dointvec,
6706         },
6707         {
6708                 .procname       = "ra_defrtr_metric",
6709                 .data           = &ipv6_devconf.ra_defrtr_metric,
6710                 .maxlen         = sizeof(u32),
6711                 .mode           = 0644,
6712                 .proc_handler   = proc_douintvec_minmax,
6713                 .extra1         = (void *)SYSCTL_ONE,
6714         },
6715         {
6716                 .procname       = "accept_ra_min_hop_limit",
6717                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6718                 .maxlen         = sizeof(int),
6719                 .mode           = 0644,
6720                 .proc_handler   = proc_dointvec,
6721         },
6722         {
6723                 .procname       = "accept_ra_pinfo",
6724                 .data           = &ipv6_devconf.accept_ra_pinfo,
6725                 .maxlen         = sizeof(int),
6726                 .mode           = 0644,
6727                 .proc_handler   = proc_dointvec,
6728         },
6729 #ifdef CONFIG_IPV6_ROUTER_PREF
6730         {
6731                 .procname       = "accept_ra_rtr_pref",
6732                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6733                 .maxlen         = sizeof(int),
6734                 .mode           = 0644,
6735                 .proc_handler   = proc_dointvec,
6736         },
6737         {
6738                 .procname       = "router_probe_interval",
6739                 .data           = &ipv6_devconf.rtr_probe_interval,
6740                 .maxlen         = sizeof(int),
6741                 .mode           = 0644,
6742                 .proc_handler   = proc_dointvec_jiffies,
6743         },
6744 #ifdef CONFIG_IPV6_ROUTE_INFO
6745         {
6746                 .procname       = "accept_ra_rt_info_min_plen",
6747                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6748                 .maxlen         = sizeof(int),
6749                 .mode           = 0644,
6750                 .proc_handler   = proc_dointvec,
6751         },
6752         {
6753                 .procname       = "accept_ra_rt_info_max_plen",
6754                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6755                 .maxlen         = sizeof(int),
6756                 .mode           = 0644,
6757                 .proc_handler   = proc_dointvec,
6758         },
6759 #endif
6760 #endif
6761         {
6762                 .procname       = "proxy_ndp",
6763                 .data           = &ipv6_devconf.proxy_ndp,
6764                 .maxlen         = sizeof(int),
6765                 .mode           = 0644,
6766                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6767         },
6768         {
6769                 .procname       = "accept_source_route",
6770                 .data           = &ipv6_devconf.accept_source_route,
6771                 .maxlen         = sizeof(int),
6772                 .mode           = 0644,
6773                 .proc_handler   = proc_dointvec,
6774         },
6775 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6776         {
6777                 .procname       = "optimistic_dad",
6778                 .data           = &ipv6_devconf.optimistic_dad,
6779                 .maxlen         = sizeof(int),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_dointvec,
6782         },
6783         {
6784                 .procname       = "use_optimistic",
6785                 .data           = &ipv6_devconf.use_optimistic,
6786                 .maxlen         = sizeof(int),
6787                 .mode           = 0644,
6788                 .proc_handler   = proc_dointvec,
6789         },
6790 #endif
6791 #ifdef CONFIG_IPV6_MROUTE
6792         {
6793                 .procname       = "mc_forwarding",
6794                 .data           = &ipv6_devconf.mc_forwarding,
6795                 .maxlen         = sizeof(int),
6796                 .mode           = 0444,
6797                 .proc_handler   = proc_dointvec,
6798         },
6799 #endif
6800         {
6801                 .procname       = "disable_ipv6",
6802                 .data           = &ipv6_devconf.disable_ipv6,
6803                 .maxlen         = sizeof(int),
6804                 .mode           = 0644,
6805                 .proc_handler   = addrconf_sysctl_disable,
6806         },
6807         {
6808                 .procname       = "accept_dad",
6809                 .data           = &ipv6_devconf.accept_dad,
6810                 .maxlen         = sizeof(int),
6811                 .mode           = 0644,
6812                 .proc_handler   = proc_dointvec,
6813         },
6814         {
6815                 .procname       = "force_tllao",
6816                 .data           = &ipv6_devconf.force_tllao,
6817                 .maxlen         = sizeof(int),
6818                 .mode           = 0644,
6819                 .proc_handler   = proc_dointvec
6820         },
6821         {
6822                 .procname       = "ndisc_notify",
6823                 .data           = &ipv6_devconf.ndisc_notify,
6824                 .maxlen         = sizeof(int),
6825                 .mode           = 0644,
6826                 .proc_handler   = proc_dointvec
6827         },
6828         {
6829                 .procname       = "suppress_frag_ndisc",
6830                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6831                 .maxlen         = sizeof(int),
6832                 .mode           = 0644,
6833                 .proc_handler   = proc_dointvec
6834         },
6835         {
6836                 .procname       = "accept_ra_from_local",
6837                 .data           = &ipv6_devconf.accept_ra_from_local,
6838                 .maxlen         = sizeof(int),
6839                 .mode           = 0644,
6840                 .proc_handler   = proc_dointvec,
6841         },
6842         {
6843                 .procname       = "accept_ra_mtu",
6844                 .data           = &ipv6_devconf.accept_ra_mtu,
6845                 .maxlen         = sizeof(int),
6846                 .mode           = 0644,
6847                 .proc_handler   = proc_dointvec,
6848         },
6849         {
6850                 .procname       = "stable_secret",
6851                 .data           = &ipv6_devconf.stable_secret,
6852                 .maxlen         = IPV6_MAX_STRLEN,
6853                 .mode           = 0600,
6854                 .proc_handler   = addrconf_sysctl_stable_secret,
6855         },
6856         {
6857                 .procname       = "use_oif_addrs_only",
6858                 .data           = &ipv6_devconf.use_oif_addrs_only,
6859                 .maxlen         = sizeof(int),
6860                 .mode           = 0644,
6861                 .proc_handler   = proc_dointvec,
6862         },
6863         {
6864                 .procname       = "ignore_routes_with_linkdown",
6865                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6866                 .maxlen         = sizeof(int),
6867                 .mode           = 0644,
6868                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6869         },
6870         {
6871                 .procname       = "drop_unicast_in_l2_multicast",
6872                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6873                 .maxlen         = sizeof(int),
6874                 .mode           = 0644,
6875                 .proc_handler   = proc_dointvec,
6876         },
6877         {
6878                 .procname       = "drop_unsolicited_na",
6879                 .data           = &ipv6_devconf.drop_unsolicited_na,
6880                 .maxlen         = sizeof(int),
6881                 .mode           = 0644,
6882                 .proc_handler   = proc_dointvec,
6883         },
6884         {
6885                 .procname       = "keep_addr_on_down",
6886                 .data           = &ipv6_devconf.keep_addr_on_down,
6887                 .maxlen         = sizeof(int),
6888                 .mode           = 0644,
6889                 .proc_handler   = proc_dointvec,
6890
6891         },
6892         {
6893                 .procname       = "seg6_enabled",
6894                 .data           = &ipv6_devconf.seg6_enabled,
6895                 .maxlen         = sizeof(int),
6896                 .mode           = 0644,
6897                 .proc_handler   = proc_dointvec,
6898         },
6899 #ifdef CONFIG_IPV6_SEG6_HMAC
6900         {
6901                 .procname       = "seg6_require_hmac",
6902                 .data           = &ipv6_devconf.seg6_require_hmac,
6903                 .maxlen         = sizeof(int),
6904                 .mode           = 0644,
6905                 .proc_handler   = proc_dointvec,
6906         },
6907 #endif
6908         {
6909                 .procname       = "enhanced_dad",
6910                 .data           = &ipv6_devconf.enhanced_dad,
6911                 .maxlen         = sizeof(int),
6912                 .mode           = 0644,
6913                 .proc_handler   = proc_dointvec,
6914         },
6915         {
6916                 .procname       = "addr_gen_mode",
6917                 .data           = &ipv6_devconf.addr_gen_mode,
6918                 .maxlen         = sizeof(int),
6919                 .mode           = 0644,
6920                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6921         },
6922         {
6923                 .procname       = "disable_policy",
6924                 .data           = &ipv6_devconf.disable_policy,
6925                 .maxlen         = sizeof(int),
6926                 .mode           = 0644,
6927                 .proc_handler   = addrconf_sysctl_disable_policy,
6928         },
6929         {
6930                 .procname       = "ndisc_tclass",
6931                 .data           = &ipv6_devconf.ndisc_tclass,
6932                 .maxlen         = sizeof(int),
6933                 .mode           = 0644,
6934                 .proc_handler   = proc_dointvec_minmax,
6935                 .extra1         = (void *)SYSCTL_ZERO,
6936                 .extra2         = (void *)&two_five_five,
6937         },
6938         {
6939                 .procname       = "rpl_seg_enabled",
6940                 .data           = &ipv6_devconf.rpl_seg_enabled,
6941                 .maxlen         = sizeof(int),
6942                 .mode           = 0644,
6943                 .proc_handler   = proc_dointvec,
6944         },
6945         {
6946                 .procname       = "ioam6_enabled",
6947                 .data           = &ipv6_devconf.ioam6_enabled,
6948                 .maxlen         = sizeof(u8),
6949                 .mode           = 0644,
6950                 .proc_handler   = proc_dou8vec_minmax,
6951                 .extra1         = (void *)SYSCTL_ZERO,
6952                 .extra2         = (void *)SYSCTL_ONE,
6953         },
6954         {
6955                 .procname       = "ioam6_id",
6956                 .data           = &ipv6_devconf.ioam6_id,
6957                 .maxlen         = sizeof(u32),
6958                 .mode           = 0644,
6959                 .proc_handler   = proc_douintvec_minmax,
6960                 .extra1         = (void *)SYSCTL_ZERO,
6961                 .extra2         = (void *)&ioam6_if_id_max,
6962         },
6963         {
6964                 .procname       = "ioam6_id_wide",
6965                 .data           = &ipv6_devconf.ioam6_id_wide,
6966                 .maxlen         = sizeof(u32),
6967                 .mode           = 0644,
6968                 .proc_handler   = proc_douintvec,
6969         },
6970         {
6971                 /* sentinel */
6972         }
6973 };
6974
6975 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6976                 struct inet6_dev *idev, struct ipv6_devconf *p)
6977 {
6978         int i, ifindex;
6979         struct ctl_table *table;
6980         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6981
6982         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6983         if (!table)
6984                 goto out;
6985
6986         for (i = 0; table[i].data; i++) {
6987                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6988                 /* If one of these is already set, then it is not safe to
6989                  * overwrite either of them: this makes proc_dointvec_minmax
6990                  * usable.
6991                  */
6992                 if (!table[i].extra1 && !table[i].extra2) {
6993                         table[i].extra1 = idev; /* embedded; no ref */
6994                         table[i].extra2 = net;
6995                 }
6996         }
6997
6998         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6999
7000         p->sysctl_header = register_net_sysctl(net, path, table);
7001         if (!p->sysctl_header)
7002                 goto free;
7003
7004         if (!strcmp(dev_name, "all"))
7005                 ifindex = NETCONFA_IFINDEX_ALL;
7006         else if (!strcmp(dev_name, "default"))
7007                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7008         else
7009                 ifindex = idev->dev->ifindex;
7010         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7011                                      ifindex, p);
7012         return 0;
7013
7014 free:
7015         kfree(table);
7016 out:
7017         return -ENOBUFS;
7018 }
7019
7020 static void __addrconf_sysctl_unregister(struct net *net,
7021                                          struct ipv6_devconf *p, int ifindex)
7022 {
7023         struct ctl_table *table;
7024
7025         if (!p->sysctl_header)
7026                 return;
7027
7028         table = p->sysctl_header->ctl_table_arg;
7029         unregister_net_sysctl_table(p->sysctl_header);
7030         p->sysctl_header = NULL;
7031         kfree(table);
7032
7033         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7034 }
7035
7036 static int addrconf_sysctl_register(struct inet6_dev *idev)
7037 {
7038         int err;
7039
7040         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7041                 return -EINVAL;
7042
7043         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7044                                     &ndisc_ifinfo_sysctl_change);
7045         if (err)
7046                 return err;
7047         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7048                                          idev, &idev->cnf);
7049         if (err)
7050                 neigh_sysctl_unregister(idev->nd_parms);
7051
7052         return err;
7053 }
7054
7055 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7056 {
7057         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7058                                      idev->dev->ifindex);
7059         neigh_sysctl_unregister(idev->nd_parms);
7060 }
7061
7062
7063 #endif
7064
7065 static int __net_init addrconf_init_net(struct net *net)
7066 {
7067         int err = -ENOMEM;
7068         struct ipv6_devconf *all, *dflt;
7069
7070         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7071         if (!all)
7072                 goto err_alloc_all;
7073
7074         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7075         if (!dflt)
7076                 goto err_alloc_dflt;
7077
7078         if (IS_ENABLED(CONFIG_SYSCTL) &&
7079             !net_eq(net, &init_net)) {
7080                 switch (sysctl_devconf_inherit_init_net) {
7081                 case 1:  /* copy from init_net */
7082                         memcpy(all, init_net.ipv6.devconf_all,
7083                                sizeof(ipv6_devconf));
7084                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7085                                sizeof(ipv6_devconf_dflt));
7086                         break;
7087                 case 3: /* copy from the current netns */
7088                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7089                                sizeof(ipv6_devconf));
7090                         memcpy(dflt,
7091                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7092                                sizeof(ipv6_devconf_dflt));
7093                         break;
7094                 case 0:
7095                 case 2:
7096                         /* use compiled values */
7097                         break;
7098                 }
7099         }
7100
7101         /* these will be inherited by all namespaces */
7102         dflt->autoconf = ipv6_defaults.autoconf;
7103         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7104
7105         dflt->stable_secret.initialized = false;
7106         all->stable_secret.initialized = false;
7107
7108         net->ipv6.devconf_all = all;
7109         net->ipv6.devconf_dflt = dflt;
7110
7111 #ifdef CONFIG_SYSCTL
7112         err = __addrconf_sysctl_register(net, "all", NULL, all);
7113         if (err < 0)
7114                 goto err_reg_all;
7115
7116         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7117         if (err < 0)
7118                 goto err_reg_dflt;
7119 #endif
7120         return 0;
7121
7122 #ifdef CONFIG_SYSCTL
7123 err_reg_dflt:
7124         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7125 err_reg_all:
7126         kfree(dflt);
7127 #endif
7128 err_alloc_dflt:
7129         kfree(all);
7130 err_alloc_all:
7131         return err;
7132 }
7133
7134 static void __net_exit addrconf_exit_net(struct net *net)
7135 {
7136 #ifdef CONFIG_SYSCTL
7137         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7138                                      NETCONFA_IFINDEX_DEFAULT);
7139         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7140                                      NETCONFA_IFINDEX_ALL);
7141 #endif
7142         kfree(net->ipv6.devconf_dflt);
7143         kfree(net->ipv6.devconf_all);
7144 }
7145
7146 static struct pernet_operations addrconf_ops = {
7147         .init = addrconf_init_net,
7148         .exit = addrconf_exit_net,
7149 };
7150
7151 static struct rtnl_af_ops inet6_ops __read_mostly = {
7152         .family           = AF_INET6,
7153         .fill_link_af     = inet6_fill_link_af,
7154         .get_link_af_size = inet6_get_link_af_size,
7155         .validate_link_af = inet6_validate_link_af,
7156         .set_link_af      = inet6_set_link_af,
7157 };
7158
7159 /*
7160  *      Init / cleanup code
7161  */
7162
7163 int __init addrconf_init(void)
7164 {
7165         struct inet6_dev *idev;
7166         int i, err;
7167
7168         err = ipv6_addr_label_init();
7169         if (err < 0) {
7170                 pr_crit("%s: cannot initialize default policy table: %d\n",
7171                         __func__, err);
7172                 goto out;
7173         }
7174
7175         err = register_pernet_subsys(&addrconf_ops);
7176         if (err < 0)
7177                 goto out_addrlabel;
7178
7179         addrconf_wq = create_workqueue("ipv6_addrconf");
7180         if (!addrconf_wq) {
7181                 err = -ENOMEM;
7182                 goto out_nowq;
7183         }
7184
7185         /* The addrconf netdev notifier requires that loopback_dev
7186          * has it's ipv6 private information allocated and setup
7187          * before it can bring up and give link-local addresses
7188          * to other devices which are up.
7189          *
7190          * Unfortunately, loopback_dev is not necessarily the first
7191          * entry in the global dev_base list of net devices.  In fact,
7192          * it is likely to be the very last entry on that list.
7193          * So this causes the notifier registry below to try and
7194          * give link-local addresses to all devices besides loopback_dev
7195          * first, then loopback_dev, which cases all the non-loopback_dev
7196          * devices to fail to get a link-local address.
7197          *
7198          * So, as a temporary fix, allocate the ipv6 structure for
7199          * loopback_dev first by hand.
7200          * Longer term, all of the dependencies ipv6 has upon the loopback
7201          * device and it being up should be removed.
7202          */
7203         rtnl_lock();
7204         idev = ipv6_add_dev(init_net.loopback_dev);
7205         rtnl_unlock();
7206         if (IS_ERR(idev)) {
7207                 err = PTR_ERR(idev);
7208                 goto errlo;
7209         }
7210
7211         ip6_route_init_special_entries();
7212
7213         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7214                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7215
7216         register_netdevice_notifier(&ipv6_dev_notf);
7217
7218         addrconf_verify();
7219
7220         rtnl_af_register(&inet6_ops);
7221
7222         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7223                                    NULL, inet6_dump_ifinfo, 0);
7224         if (err < 0)
7225                 goto errout;
7226
7227         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7228                                    inet6_rtm_newaddr, NULL, 0);
7229         if (err < 0)
7230                 goto errout;
7231         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7232                                    inet6_rtm_deladdr, NULL, 0);
7233         if (err < 0)
7234                 goto errout;
7235         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7236                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7237                                    RTNL_FLAG_DOIT_UNLOCKED);
7238         if (err < 0)
7239                 goto errout;
7240         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7241                                    NULL, inet6_dump_ifmcaddr, 0);
7242         if (err < 0)
7243                 goto errout;
7244         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7245                                    NULL, inet6_dump_ifacaddr, 0);
7246         if (err < 0)
7247                 goto errout;
7248         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7249                                    inet6_netconf_get_devconf,
7250                                    inet6_netconf_dump_devconf,
7251                                    RTNL_FLAG_DOIT_UNLOCKED);
7252         if (err < 0)
7253                 goto errout;
7254         err = ipv6_addr_label_rtnl_register();
7255         if (err < 0)
7256                 goto errout;
7257
7258         return 0;
7259 errout:
7260         rtnl_unregister_all(PF_INET6);
7261         rtnl_af_unregister(&inet6_ops);
7262         unregister_netdevice_notifier(&ipv6_dev_notf);
7263 errlo:
7264         destroy_workqueue(addrconf_wq);
7265 out_nowq:
7266         unregister_pernet_subsys(&addrconf_ops);
7267 out_addrlabel:
7268         ipv6_addr_label_cleanup();
7269 out:
7270         return err;
7271 }
7272
7273 void addrconf_cleanup(void)
7274 {
7275         struct net_device *dev;
7276         int i;
7277
7278         unregister_netdevice_notifier(&ipv6_dev_notf);
7279         unregister_pernet_subsys(&addrconf_ops);
7280         ipv6_addr_label_cleanup();
7281
7282         rtnl_af_unregister(&inet6_ops);
7283
7284         rtnl_lock();
7285
7286         /* clean dev list */
7287         for_each_netdev(&init_net, dev) {
7288                 if (__in6_dev_get(dev) == NULL)
7289                         continue;
7290                 addrconf_ifdown(dev, true);
7291         }
7292         addrconf_ifdown(init_net.loopback_dev, true);
7293
7294         /*
7295          *      Check hash table.
7296          */
7297         spin_lock_bh(&addrconf_hash_lock);
7298         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7299                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7300         spin_unlock_bh(&addrconf_hash_lock);
7301         cancel_delayed_work(&addr_chk_work);
7302         rtnl_unlock();
7303
7304         destroy_workqueue(addrconf_wq);
7305 }