net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92
  93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  94
  95 #define IPV6_MAX_STRLEN \
  96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  97
  98 static inline u32 cstamp_delta(unsigned long cstamp)
  99 {
 100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 101 }
 102
 103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 104 {
 105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 107         do_div(tmp, 1000000);
 108         return (s32)tmp;
 109 }
 110
 111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 112 {
 113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 119                 do_div(tmp, 1000000);
 120         }
 121         return (s32)tmp;
 122 }
 123
 124 #ifdef CONFIG_SYSCTL
 125 static int addrconf_sysctl_register(struct inet6_dev *idev);
 126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 127 #else
 128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 129 {
 130         return 0;
 131 }
 132
 133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 134 {
 135 }
 136 #endif
 137
 138 static void ipv6_regen_rndid(struct inet6_dev *idev);
 139 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 140
 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 142 static int ipv6_count_addresses(const struct inet6_dev *idev);
 143 static int ipv6_generate_stable_address(struct in6_addr *addr,
 144                                         u8 dad_count,
 145                                         const struct inet6_dev *idev);
 146
 147 #define IN6_ADDR_HSIZE_SHIFT    8
 148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 149 /*
 150  *      Configured unicast address hash table
 151  */
 152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 153 static DEFINE_SPINLOCK(addrconf_hash_lock);
 154
 155 static void addrconf_verify(void);
 156 static void addrconf_verify_rtnl(void);
 157 static void addrconf_verify_work(struct work_struct *);
 158
 159 static struct workqueue_struct *addrconf_wq;
 160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 161
 162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 164
 165 static void addrconf_type_change(struct net_device *dev,
 166                                  unsigned long event);
 167 static int addrconf_ifdown(struct net_device *dev, int how);
 168
 169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 170                                                   int plen,
 171                                                   const struct net_device *dev,
 172                                                   u32 flags, u32 noflags,
 173                                                   bool no_gw);
 174
 175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 176 static void addrconf_dad_work(struct work_struct *w);
 177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 178                                    bool send_na);
 179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 180 static void addrconf_rs_timer(struct timer_list *t);
 181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 183
 184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 185                                 struct prefix_info *pinfo);
 186
 187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 188         .forwarding             = 0,
 189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 190         .mtu6                   = IPV6_MIN_MTU,
 191         .accept_ra              = 1,
 192         .accept_redirects       = 1,
 193         .autoconf               = 1,
 194         .force_mld_version      = 0,
 195         .mldv1_unsolicited_report_interval = 10 * HZ,
 196         .mldv2_unsolicited_report_interval = HZ,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .accept_ra_from_local   = 0,
 210         .accept_ra_min_hop_limit= 1,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_min_plen = 0,
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 0,
 224         .suppress_frag_ndisc    = 1,
 225         .accept_ra_mtu          = 1,
 226         .stable_secret          = {
 227                 .initialized = false,
 228         },
 229         .use_oif_addrs_only     = 0,
 230         .ignore_routes_with_linkdown = 0,
 231         .keep_addr_on_down      = 0,
 232         .seg6_enabled           = 0,
 233 #ifdef CONFIG_IPV6_SEG6_HMAC
 234         .seg6_require_hmac      = 0,
 235 #endif
 236         .enhanced_dad           = 1,
 237         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 238         .disable_policy         = 0,
 239         .rpl_seg_enabled        = 0,
 240 };
 241
 242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 243         .forwarding             = 0,
 244         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 245         .mtu6                   = IPV6_MIN_MTU,
 246         .accept_ra              = 1,
 247         .accept_redirects       = 1,
 248         .autoconf               = 1,
 249         .force_mld_version      = 0,
 250         .mldv1_unsolicited_report_interval = 10 * HZ,
 251         .mldv2_unsolicited_report_interval = HZ,
 252         .dad_transmits          = 1,
 253         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 254         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 255         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 256         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 257         .use_tempaddr           = 0,
 258         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 259         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 260         .regen_max_retry        = REGEN_MAX_RETRY,
 261         .max_desync_factor      = MAX_DESYNC_FACTOR,
 262         .max_addresses          = IPV6_MAX_ADDRESSES,
 263         .accept_ra_defrtr       = 1,
 264         .accept_ra_from_local   = 0,
 265         .accept_ra_min_hop_limit= 1,
 266         .accept_ra_pinfo        = 1,
 267 #ifdef CONFIG_IPV6_ROUTER_PREF
 268         .accept_ra_rtr_pref     = 1,
 269         .rtr_probe_interval     = 60 * HZ,
 270 #ifdef CONFIG_IPV6_ROUTE_INFO
 271         .accept_ra_rt_info_min_plen = 0,
 272         .accept_ra_rt_info_max_plen = 0,
 273 #endif
 274 #endif
 275         .proxy_ndp              = 0,
 276         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 277         .disable_ipv6           = 0,
 278         .accept_dad             = 1,
 279         .suppress_frag_ndisc    = 1,
 280         .accept_ra_mtu          = 1,
 281         .stable_secret          = {
 282                 .initialized = false,
 283         },
 284         .use_oif_addrs_only     = 0,
 285         .ignore_routes_with_linkdown = 0,
 286         .keep_addr_on_down      = 0,
 287         .seg6_enabled           = 0,
 288 #ifdef CONFIG_IPV6_SEG6_HMAC
 289         .seg6_require_hmac      = 0,
 290 #endif
 291         .enhanced_dad           = 1,
 292         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 293         .disable_policy         = 0,
 294         .rpl_seg_enabled        = 0,
 295 };
 296
 297 /* Check if link is ready: is it up and is a valid qdisc available */
 298 static inline bool addrconf_link_ready(const struct net_device *dev)
 299 {
 300         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 301 }
 302
 303 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 304 {
 305         if (del_timer(&idev->rs_timer))
 306                 __in6_dev_put(idev);
 307 }
 308
 309 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 310 {
 311         if (cancel_delayed_work(&ifp->dad_work))
 312                 __in6_ifa_put(ifp);
 313 }
 314
 315 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 316                                   unsigned long when)
 317 {
 318         if (!timer_pending(&idev->rs_timer))
 319                 in6_dev_hold(idev);
 320         mod_timer(&idev->rs_timer, jiffies + when);
 321 }
 322
 323 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 324                                    unsigned long delay)
 325 {
 326         in6_ifa_hold(ifp);
 327         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 328                 in6_ifa_put(ifp);
 329 }
 330
 331 static int snmp6_alloc_dev(struct inet6_dev *idev)
 332 {
 333         int i;
 334
 335         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 336         if (!idev->stats.ipv6)
 337                 goto err_ip;
 338
 339         for_each_possible_cpu(i) {
 340                 struct ipstats_mib *addrconf_stats;
 341                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 342                 u64_stats_init(&addrconf_stats->syncp);
 343         }
 344
 345
 346         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 347                                         GFP_KERNEL);
 348         if (!idev->stats.icmpv6dev)
 349                 goto err_icmp;
 350         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 351                                            GFP_KERNEL);
 352         if (!idev->stats.icmpv6msgdev)
 353                 goto err_icmpmsg;
 354
 355         return 0;
 356
 357 err_icmpmsg:
 358         kfree(idev->stats.icmpv6dev);
 359 err_icmp:
 360         free_percpu(idev->stats.ipv6);
 361 err_ip:
 362         return -ENOMEM;
 363 }
 364
 365 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 366 {
 367         struct inet6_dev *ndev;
 368         int err = -ENOMEM;
 369
 370         ASSERT_RTNL();
 371
 372         if (dev->mtu < IPV6_MIN_MTU)
 373                 return ERR_PTR(-EINVAL);
 374
 375         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 376         if (!ndev)
 377                 return ERR_PTR(err);
 378
 379         rwlock_init(&ndev->lock);
 380         ndev->dev = dev;
 381         INIT_LIST_HEAD(&ndev->addr_list);
 382         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 383         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 384
 385         if (ndev->cnf.stable_secret.initialized)
 386                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 387
 388         ndev->cnf.mtu6 = dev->mtu;
 389         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 390         if (!ndev->nd_parms) {
 391                 kfree(ndev);
 392                 return ERR_PTR(err);
 393         }
 394         if (ndev->cnf.forwarding)
 395                 dev_disable_lro(dev);
 396         /* We refer to the device */
 397         dev_hold(dev);
 398
 399         if (snmp6_alloc_dev(ndev) < 0) {
 400                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 401                            __func__);
 402                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 403                 dev_put(dev);
 404                 kfree(ndev);
 405                 return ERR_PTR(err);
 406         }
 407
 408         if (snmp6_register_dev(ndev) < 0) {
 409                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 410                            __func__, dev->name);
 411                 goto err_release;
 412         }
 413
 414         /* One reference from device. */
 415         refcount_set(&ndev->refcnt, 1);
 416
 417         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 418                 ndev->cnf.accept_dad = -1;
 419
 420 #if IS_ENABLED(CONFIG_IPV6_SIT)
 421         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 422                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 423                 ndev->cnf.rtr_solicits = 0;
 424         }
 425 #endif
 426
 427         INIT_LIST_HEAD(&ndev->tempaddr_list);
 428         ndev->desync_factor = U32_MAX;
 429         if ((dev->flags&IFF_LOOPBACK) ||
 430             dev->type == ARPHRD_TUNNEL ||
 431             dev->type == ARPHRD_TUNNEL6 ||
 432             dev->type == ARPHRD_SIT ||
 433             dev->type == ARPHRD_NONE) {
 434                 ndev->cnf.use_tempaddr = -1;
 435         } else
 436                 ipv6_regen_rndid(ndev);
 437
 438         ndev->token = in6addr_any;
 439
 440         if (netif_running(dev) && addrconf_link_ready(dev))
 441                 ndev->if_flags |= IF_READY;
 442
 443         ipv6_mc_init_dev(ndev);
 444         ndev->tstamp = jiffies;
 445         err = addrconf_sysctl_register(ndev);
 446         if (err) {
 447                 ipv6_mc_destroy_dev(ndev);
 448                 snmp6_unregister_dev(ndev);
 449                 goto err_release;
 450         }
 451         /* protected by rtnl_lock */
 452         rcu_assign_pointer(dev->ip6_ptr, ndev);
 453
 454         /* Join interface-local all-node multicast group */
 455         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 456
 457         /* Join all-node multicast group */
 458         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 459
 460         /* Join all-router multicast group if forwarding is set */
 461         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 462                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 463
 464         return ndev;
 465
 466 err_release:
 467         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 468         ndev->dead = 1;
 469         in6_dev_finish_destroy(ndev);
 470         return ERR_PTR(err);
 471 }
 472
 473 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 474 {
 475         struct inet6_dev *idev;
 476
 477         ASSERT_RTNL();
 478
 479         idev = __in6_dev_get(dev);
 480         if (!idev) {
 481                 idev = ipv6_add_dev(dev);
 482                 if (IS_ERR(idev))
 483                         return idev;
 484         }
 485
 486         if (dev->flags&IFF_UP)
 487                 ipv6_mc_up(idev);
 488         return idev;
 489 }
 490
 491 static int inet6_netconf_msgsize_devconf(int type)
 492 {
 493         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 494                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 495         bool all = false;
 496
 497         if (type == NETCONFA_ALL)
 498                 all = true;
 499
 500         if (all || type == NETCONFA_FORWARDING)
 501                 size += nla_total_size(4);
 502 #ifdef CONFIG_IPV6_MROUTE
 503         if (all || type == NETCONFA_MC_FORWARDING)
 504                 size += nla_total_size(4);
 505 #endif
 506         if (all || type == NETCONFA_PROXY_NEIGH)
 507                 size += nla_total_size(4);
 508
 509         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 510                 size += nla_total_size(4);
 511
 512         return size;
 513 }
 514
 515 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 516                                       struct ipv6_devconf *devconf, u32 portid,
 517                                       u32 seq, int event, unsigned int flags,
 518                                       int type)
 519 {
 520         struct nlmsghdr  *nlh;
 521         struct netconfmsg *ncm;
 522         bool all = false;
 523
 524         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 525                         flags);
 526         if (!nlh)
 527                 return -EMSGSIZE;
 528
 529         if (type == NETCONFA_ALL)
 530                 all = true;
 531
 532         ncm = nlmsg_data(nlh);
 533         ncm->ncm_family = AF_INET6;
 534
 535         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 536                 goto nla_put_failure;
 537
 538         if (!devconf)
 539                 goto out;
 540
 541         if ((all || type == NETCONFA_FORWARDING) &&
 542             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 543                 goto nla_put_failure;
 544 #ifdef CONFIG_IPV6_MROUTE
 545         if ((all || type == NETCONFA_MC_FORWARDING) &&
 546             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 547                         devconf->mc_forwarding) < 0)
 548                 goto nla_put_failure;
 549 #endif
 550         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 551             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 552                 goto nla_put_failure;
 553
 554         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 555             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 556                         devconf->ignore_routes_with_linkdown) < 0)
 557                 goto nla_put_failure;
 558
 559 out:
 560         nlmsg_end(skb, nlh);
 561         return 0;
 562
 563 nla_put_failure:
 564         nlmsg_cancel(skb, nlh);
 565         return -EMSGSIZE;
 566 }
 567
 568 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 569                                   int ifindex, struct ipv6_devconf *devconf)
 570 {
 571         struct sk_buff *skb;
 572         int err = -ENOBUFS;
 573
 574         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 575         if (!skb)
 576                 goto errout;
 577
 578         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 579                                          event, 0, type);
 580         if (err < 0) {
 581                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 582                 WARN_ON(err == -EMSGSIZE);
 583                 kfree_skb(skb);
 584                 goto errout;
 585         }
 586         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 587         return;
 588 errout:
 589         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 590 }
 591
 592 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 593         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 594         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 595         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 596         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 597 };
 598
 599 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 600                                        const struct nlmsghdr *nlh,
 601                                        struct nlattr **tb,
 602                                        struct netlink_ext_ack *extack)
 603 {
 604         int i, err;
 605
 606         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 607                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 608                 return -EINVAL;
 609         }
 610
 611         if (!netlink_strict_get_check(skb))
 612                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
 613                                               tb, NETCONFA_MAX,
 614                                               devconf_ipv6_policy, extack);
 615
 616         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
 617                                             tb, NETCONFA_MAX,
 618                                             devconf_ipv6_policy, extack);
 619         if (err)
 620                 return err;
 621
 622         for (i = 0; i <= NETCONFA_MAX; i++) {
 623                 if (!tb[i])
 624                         continue;
 625
 626                 switch (i) {
 627                 case NETCONFA_IFINDEX:
 628                         break;
 629                 default:
 630                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 631                         return -EINVAL;
 632                 }
 633         }
 634
 635         return 0;
 636 }
 637
 638 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 639                                      struct nlmsghdr *nlh,
 640                                      struct netlink_ext_ack *extack)
 641 {
 642         struct net *net = sock_net(in_skb->sk);
 643         struct nlattr *tb[NETCONFA_MAX+1];
 644         struct inet6_dev *in6_dev = NULL;
 645         struct net_device *dev = NULL;
 646         struct sk_buff *skb;
 647         struct ipv6_devconf *devconf;
 648         int ifindex;
 649         int err;
 650
 651         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 652         if (err < 0)
 653                 return err;
 654
 655         if (!tb[NETCONFA_IFINDEX])
 656                 return -EINVAL;
 657
 658         err = -EINVAL;
 659         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 660         switch (ifindex) {
 661         case NETCONFA_IFINDEX_ALL:
 662                 devconf = net->ipv6.devconf_all;
 663                 break;
 664         case NETCONFA_IFINDEX_DEFAULT:
 665                 devconf = net->ipv6.devconf_dflt;
 666                 break;
 667         default:
 668                 dev = dev_get_by_index(net, ifindex);
 669                 if (!dev)
 670                         return -EINVAL;
 671                 in6_dev = in6_dev_get(dev);
 672                 if (!in6_dev)
 673                         goto errout;
 674                 devconf = &in6_dev->cnf;
 675                 break;
 676         }
 677
 678         err = -ENOBUFS;
 679         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 680         if (!skb)
 681                 goto errout;
 682
 683         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 684                                          NETLINK_CB(in_skb).portid,
 685                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 686                                          NETCONFA_ALL);
 687         if (err < 0) {
 688                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 689                 WARN_ON(err == -EMSGSIZE);
 690                 kfree_skb(skb);
 691                 goto errout;
 692         }
 693         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 694 errout:
 695         if (in6_dev)
 696                 in6_dev_put(in6_dev);
 697         if (dev)
 698                 dev_put(dev);
 699         return err;
 700 }
 701
 702 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 703                                       struct netlink_callback *cb)
 704 {
 705         const struct nlmsghdr *nlh = cb->nlh;
 706         struct net *net = sock_net(skb->sk);
 707         int h, s_h;
 708         int idx, s_idx;
 709         struct net_device *dev;
 710         struct inet6_dev *idev;
 711         struct hlist_head *head;
 712
 713         if (cb->strict_check) {
 714                 struct netlink_ext_ack *extack = cb->extack;
 715                 struct netconfmsg *ncm;
 716
 717                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 718                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 719                         return -EINVAL;
 720                 }
 721
 722                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 723                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 724                         return -EINVAL;
 725                 }
 726         }
 727
 728         s_h = cb->args[0];
 729         s_idx = idx = cb->args[1];
 730
 731         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 732                 idx = 0;
 733                 head = &net->dev_index_head[h];
 734                 rcu_read_lock();
 735                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 736                           net->dev_base_seq;
 737                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 738                         if (idx < s_idx)
 739                                 goto cont;
 740                         idev = __in6_dev_get(dev);
 741                         if (!idev)
 742                                 goto cont;
 743
 744                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 745                                                        &idev->cnf,
 746                                                        NETLINK_CB(cb->skb).portid,
 747                                                        nlh->nlmsg_seq,
 748                                                        RTM_NEWNETCONF,
 749                                                        NLM_F_MULTI,
 750                                                        NETCONFA_ALL) < 0) {
 751                                 rcu_read_unlock();
 752                                 goto done;
 753                         }
 754                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 755 cont:
 756                         idx++;
 757                 }
 758                 rcu_read_unlock();
 759         }
 760         if (h == NETDEV_HASHENTRIES) {
 761                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 762                                                net->ipv6.devconf_all,
 763                                                NETLINK_CB(cb->skb).portid,
 764                                                nlh->nlmsg_seq,
 765                                                RTM_NEWNETCONF, NLM_F_MULTI,
 766                                                NETCONFA_ALL) < 0)
 767                         goto done;
 768                 else
 769                         h++;
 770         }
 771         if (h == NETDEV_HASHENTRIES + 1) {
 772                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 773                                                net->ipv6.devconf_dflt,
 774                                                NETLINK_CB(cb->skb).portid,
 775                                                nlh->nlmsg_seq,
 776                                                RTM_NEWNETCONF, NLM_F_MULTI,
 777                                                NETCONFA_ALL) < 0)
 778                         goto done;
 779                 else
 780                         h++;
 781         }
 782 done:
 783         cb->args[0] = h;
 784         cb->args[1] = idx;
 785
 786         return skb->len;
 787 }
 788
 789 #ifdef CONFIG_SYSCTL
 790 static void dev_forward_change(struct inet6_dev *idev)
 791 {
 792         struct net_device *dev;
 793         struct inet6_ifaddr *ifa;
 794
 795         if (!idev)
 796                 return;
 797         dev = idev->dev;
 798         if (idev->cnf.forwarding)
 799                 dev_disable_lro(dev);
 800         if (dev->flags & IFF_MULTICAST) {
 801                 if (idev->cnf.forwarding) {
 802                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 803                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 804                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 805                 } else {
 806                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 807                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 808                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 809                 }
 810         }
 811
 812         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 813                 if (ifa->flags&IFA_F_TENTATIVE)
 814                         continue;
 815                 if (idev->cnf.forwarding)
 816                         addrconf_join_anycast(ifa);
 817                 else
 818                         addrconf_leave_anycast(ifa);
 819         }
 820         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 821                                      NETCONFA_FORWARDING,
 822                                      dev->ifindex, &idev->cnf);
 823 }
 824
 825
 826 static void addrconf_forward_change(struct net *net, __s32 newf)
 827 {
 828         struct net_device *dev;
 829         struct inet6_dev *idev;
 830
 831         for_each_netdev(net, dev) {
 832                 idev = __in6_dev_get(dev);
 833                 if (idev) {
 834                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 835                         idev->cnf.forwarding = newf;
 836                         if (changed)
 837                                 dev_forward_change(idev);
 838                 }
 839         }
 840 }
 841
 842 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 843 {
 844         struct net *net;
 845         int old;
 846
 847         if (!rtnl_trylock())
 848                 return restart_syscall();
 849
 850         net = (struct net *)table->extra2;
 851         old = *p;
 852         *p = newf;
 853
 854         if (p == &net->ipv6.devconf_dflt->forwarding) {
 855                 if ((!newf) ^ (!old))
 856                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 857                                                      NETCONFA_FORWARDING,
 858                                                      NETCONFA_IFINDEX_DEFAULT,
 859                                                      net->ipv6.devconf_dflt);
 860                 rtnl_unlock();
 861                 return 0;
 862         }
 863
 864         if (p == &net->ipv6.devconf_all->forwarding) {
 865                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 866
 867                 net->ipv6.devconf_dflt->forwarding = newf;
 868                 if ((!newf) ^ (!old_dflt))
 869                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 870                                                      NETCONFA_FORWARDING,
 871                                                      NETCONFA_IFINDEX_DEFAULT,
 872                                                      net->ipv6.devconf_dflt);
 873
 874                 addrconf_forward_change(net, newf);
 875                 if ((!newf) ^ (!old))
 876                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 877                                                      NETCONFA_FORWARDING,
 878                                                      NETCONFA_IFINDEX_ALL,
 879                                                      net->ipv6.devconf_all);
 880         } else if ((!newf) ^ (!old))
 881                 dev_forward_change((struct inet6_dev *)table->extra1);
 882         rtnl_unlock();
 883
 884         if (newf)
 885                 rt6_purge_dflt_routers(net);
 886         return 1;
 887 }
 888
 889 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 890 {
 891         struct net_device *dev;
 892         struct inet6_dev *idev;
 893
 894         for_each_netdev(net, dev) {
 895                 idev = __in6_dev_get(dev);
 896                 if (idev) {
 897                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 898
 899                         idev->cnf.ignore_routes_with_linkdown = newf;
 900                         if (changed)
 901                                 inet6_netconf_notify_devconf(dev_net(dev),
 902                                                              RTM_NEWNETCONF,
 903                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 904                                                              dev->ifindex,
 905                                                              &idev->cnf);
 906                 }
 907         }
 908 }
 909
 910 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 911 {
 912         struct net *net;
 913         int old;
 914
 915         if (!rtnl_trylock())
 916                 return restart_syscall();
 917
 918         net = (struct net *)table->extra2;
 919         old = *p;
 920         *p = newf;
 921
 922         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 923                 if ((!newf) ^ (!old))
 924                         inet6_netconf_notify_devconf(net,
 925                                                      RTM_NEWNETCONF,
 926                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 927                                                      NETCONFA_IFINDEX_DEFAULT,
 928                                                      net->ipv6.devconf_dflt);
 929                 rtnl_unlock();
 930                 return 0;
 931         }
 932
 933         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 934                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 935                 addrconf_linkdown_change(net, newf);
 936                 if ((!newf) ^ (!old))
 937                         inet6_netconf_notify_devconf(net,
 938                                                      RTM_NEWNETCONF,
 939                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 940                                                      NETCONFA_IFINDEX_ALL,
 941                                                      net->ipv6.devconf_all);
 942         }
 943         rtnl_unlock();
 944
 945         return 1;
 946 }
 947
 948 #endif
 949
 950 /* Nobody refers to this ifaddr, destroy it */
 951 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 952 {
 953         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 954
 955 #ifdef NET_REFCNT_DEBUG
 956         pr_debug("%s\n", __func__);
 957 #endif
 958
 959         in6_dev_put(ifp->idev);
 960
 961         if (cancel_delayed_work(&ifp->dad_work))
 962                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 963                           ifp);
 964
 965         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 966                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 967                 return;
 968         }
 969
 970         kfree_rcu(ifp, rcu);
 971 }
 972
 973 static void
 974 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 975 {
 976         struct list_head *p;
 977         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 978
 979         /*
 980          * Each device address list is sorted in order of scope -
 981          * global before linklocal.
 982          */
 983         list_for_each(p, &idev->addr_list) {
 984                 struct inet6_ifaddr *ifa
 985                         = list_entry(p, struct inet6_ifaddr, if_list);
 986                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 987                         break;
 988         }
 989
 990         list_add_tail_rcu(&ifp->if_list, p);
 991 }
 992
 993 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 994 {
 995         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 996
 997         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 998 }
 999
1000 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1001                                struct net_device *dev, unsigned int hash)
1002 {
1003         struct inet6_ifaddr *ifp;
1004
1005         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1006                 if (!net_eq(dev_net(ifp->idev->dev), net))
1007                         continue;
1008                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1009                         if (!dev || ifp->idev->dev == dev)
1010                                 return true;
1011                 }
1012         }
1013         return false;
1014 }
1015
1016 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1017 {
1018         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1019         int err = 0;
1020
1021         spin_lock(&addrconf_hash_lock);
1022
1023         /* Ignore adding duplicate addresses on an interface */
1024         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1025                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1026                 err = -EEXIST;
1027         } else {
1028                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1029         }
1030
1031         spin_unlock(&addrconf_hash_lock);
1032
1033         return err;
1034 }
1035
1036 /* On success it returns ifp with increased reference count */
1037
1038 static struct inet6_ifaddr *
1039 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1040               bool can_block, struct netlink_ext_ack *extack)
1041 {
1042         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1043         int addr_type = ipv6_addr_type(cfg->pfx);
1044         struct net *net = dev_net(idev->dev);
1045         struct inet6_ifaddr *ifa = NULL;
1046         struct fib6_info *f6i = NULL;
1047         int err = 0;
1048
1049         if (addr_type == IPV6_ADDR_ANY ||
1050             (addr_type & IPV6_ADDR_MULTICAST &&
1051              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1052             (!(idev->dev->flags & IFF_LOOPBACK) &&
1053              !netif_is_l3_master(idev->dev) &&
1054              addr_type & IPV6_ADDR_LOOPBACK))
1055                 return ERR_PTR(-EADDRNOTAVAIL);
1056
1057         if (idev->dead) {
1058                 err = -ENODEV;                  /*XXX*/
1059                 goto out;
1060         }
1061
1062         if (idev->cnf.disable_ipv6) {
1063                 err = -EACCES;
1064                 goto out;
1065         }
1066
1067         /* validator notifier needs to be blocking;
1068          * do not call in atomic context
1069          */
1070         if (can_block) {
1071                 struct in6_validator_info i6vi = {
1072                         .i6vi_addr = *cfg->pfx,
1073                         .i6vi_dev = idev,
1074                         .extack = extack,
1075                 };
1076
1077                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1078                 err = notifier_to_errno(err);
1079                 if (err < 0)
1080                         goto out;
1081         }
1082
1083         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1084         if (!ifa) {
1085                 err = -ENOBUFS;
1086                 goto out;
1087         }
1088
1089         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1090         if (IS_ERR(f6i)) {
1091                 err = PTR_ERR(f6i);
1092                 f6i = NULL;
1093                 goto out;
1094         }
1095
1096         if (net->ipv6.devconf_all->disable_policy ||
1097             idev->cnf.disable_policy)
1098                 f6i->dst_nopolicy = true;
1099
1100         neigh_parms_data_state_setall(idev->nd_parms);
1101
1102         ifa->addr = *cfg->pfx;
1103         if (cfg->peer_pfx)
1104                 ifa->peer_addr = *cfg->peer_pfx;
1105
1106         spin_lock_init(&ifa->lock);
1107         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1108         INIT_HLIST_NODE(&ifa->addr_lst);
1109         ifa->scope = cfg->scope;
1110         ifa->prefix_len = cfg->plen;
1111         ifa->rt_priority = cfg->rt_priority;
1112         ifa->flags = cfg->ifa_flags;
1113         /* No need to add the TENTATIVE flag for addresses with NODAD */
1114         if (!(cfg->ifa_flags & IFA_F_NODAD))
1115                 ifa->flags |= IFA_F_TENTATIVE;
1116         ifa->valid_lft = cfg->valid_lft;
1117         ifa->prefered_lft = cfg->preferred_lft;
1118         ifa->cstamp = ifa->tstamp = jiffies;
1119         ifa->tokenized = false;
1120
1121         ifa->rt = f6i;
1122
1123         ifa->idev = idev;
1124         in6_dev_hold(idev);
1125
1126         /* For caller */
1127         refcount_set(&ifa->refcnt, 1);
1128
1129         rcu_read_lock_bh();
1130
1131         err = ipv6_add_addr_hash(idev->dev, ifa);
1132         if (err < 0) {
1133                 rcu_read_unlock_bh();
1134                 goto out;
1135         }
1136
1137         write_lock(&idev->lock);
1138
1139         /* Add to inet6_dev unicast addr list. */
1140         ipv6_link_dev_addr(idev, ifa);
1141
1142         if (ifa->flags&IFA_F_TEMPORARY) {
1143                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1144                 in6_ifa_hold(ifa);
1145         }
1146
1147         in6_ifa_hold(ifa);
1148         write_unlock(&idev->lock);
1149
1150         rcu_read_unlock_bh();
1151
1152         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1153 out:
1154         if (unlikely(err < 0)) {
1155                 fib6_info_release(f6i);
1156
1157                 if (ifa) {
1158                         if (ifa->idev)
1159                                 in6_dev_put(ifa->idev);
1160                         kfree(ifa);
1161                 }
1162                 ifa = ERR_PTR(err);
1163         }
1164
1165         return ifa;
1166 }
1167
1168 enum cleanup_prefix_rt_t {
1169         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1170         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1171         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1172 };
1173
1174 /*
1175  * Check, whether the prefix for ifp would still need a prefix route
1176  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1177  * constants.
1178  *
1179  * 1) we don't purge prefix if address was not permanent.
1180  *    prefix is managed by its own lifetime.
1181  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1182  * 3) if there are no addresses, delete prefix.
1183  * 4) if there are still other permanent address(es),
1184  *    corresponding prefix is still permanent.
1185  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1186  *    don't purge the prefix, assume user space is managing it.
1187  * 6) otherwise, update prefix lifetime to the
1188  *    longest valid lifetime among the corresponding
1189  *    addresses on the device.
1190  *    Note: subsequent RA will update lifetime.
1191  **/
1192 static enum cleanup_prefix_rt_t
1193 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1194 {
1195         struct inet6_ifaddr *ifa;
1196         struct inet6_dev *idev = ifp->idev;
1197         unsigned long lifetime;
1198         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1199
1200         *expires = jiffies;
1201
1202         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1203                 if (ifa == ifp)
1204                         continue;
1205                 if (ifa->prefix_len != ifp->prefix_len ||
1206                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1207                                        ifp->prefix_len))
1208                         continue;
1209                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1210                         return CLEANUP_PREFIX_RT_NOP;
1211
1212                 action = CLEANUP_PREFIX_RT_EXPIRE;
1213
1214                 spin_lock(&ifa->lock);
1215
1216                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1217                 /*
1218                  * Note: Because this address is
1219                  * not permanent, lifetime <
1220                  * LONG_MAX / HZ here.
1221                  */
1222                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1223                         *expires = ifa->tstamp + lifetime * HZ;
1224                 spin_unlock(&ifa->lock);
1225         }
1226
1227         return action;
1228 }
1229
1230 static void
1231 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1232                      bool del_rt, bool del_peer)
1233 {
1234         struct fib6_info *f6i;
1235
1236         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1237                                         ifp->prefix_len,
1238                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1239         if (f6i) {
1240                 if (del_rt)
1241                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1242                 else {
1243                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1244                                 fib6_set_expires(f6i, expires);
1245                         fib6_info_release(f6i);
1246                 }
1247         }
1248 }
1249
1250
1251 /* This function wants to get referenced ifp and releases it before return */
1252
1253 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1254 {
1255         int state;
1256         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1257         unsigned long expires;
1258
1259         ASSERT_RTNL();
1260
1261         spin_lock_bh(&ifp->lock);
1262         state = ifp->state;
1263         ifp->state = INET6_IFADDR_STATE_DEAD;
1264         spin_unlock_bh(&ifp->lock);
1265
1266         if (state == INET6_IFADDR_STATE_DEAD)
1267                 goto out;
1268
1269         spin_lock_bh(&addrconf_hash_lock);
1270         hlist_del_init_rcu(&ifp->addr_lst);
1271         spin_unlock_bh(&addrconf_hash_lock);
1272
1273         write_lock_bh(&ifp->idev->lock);
1274
1275         if (ifp->flags&IFA_F_TEMPORARY) {
1276                 list_del(&ifp->tmp_list);
1277                 if (ifp->ifpub) {
1278                         in6_ifa_put(ifp->ifpub);
1279                         ifp->ifpub = NULL;
1280                 }
1281                 __in6_ifa_put(ifp);
1282         }
1283
1284         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1285                 action = check_cleanup_prefix_route(ifp, &expires);
1286
1287         list_del_rcu(&ifp->if_list);
1288         __in6_ifa_put(ifp);
1289
1290         write_unlock_bh(&ifp->idev->lock);
1291
1292         addrconf_del_dad_work(ifp);
1293
1294         ipv6_ifa_notify(RTM_DELADDR, ifp);
1295
1296         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1297
1298         if (action != CLEANUP_PREFIX_RT_NOP) {
1299                 cleanup_prefix_route(ifp, expires,
1300                         action == CLEANUP_PREFIX_RT_DEL, false);
1301         }
1302
1303         /* clean up prefsrc entries */
1304         rt6_remove_prefsrc(ifp);
1305 out:
1306         in6_ifa_put(ifp);
1307 }
1308
1309 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1310                                 struct inet6_ifaddr *ift,
1311                                 bool block)
1312 {
1313         struct inet6_dev *idev = ifp->idev;
1314         struct in6_addr addr, *tmpaddr;
1315         unsigned long tmp_tstamp, age;
1316         unsigned long regen_advance;
1317         struct ifa6_config cfg;
1318         int ret = 0;
1319         unsigned long now = jiffies;
1320         long max_desync_factor;
1321         s32 cnf_temp_preferred_lft;
1322
1323         write_lock_bh(&idev->lock);
1324         if (ift) {
1325                 spin_lock_bh(&ift->lock);
1326                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1327                 spin_unlock_bh(&ift->lock);
1328                 tmpaddr = &addr;
1329         } else {
1330                 tmpaddr = NULL;
1331         }
1332 retry:
1333         in6_dev_hold(idev);
1334         if (idev->cnf.use_tempaddr <= 0) {
1335                 write_unlock_bh(&idev->lock);
1336                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1337                 in6_dev_put(idev);
1338                 ret = -1;
1339                 goto out;
1340         }
1341         spin_lock_bh(&ifp->lock);
1342         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1343                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1344                 spin_unlock_bh(&ifp->lock);
1345                 write_unlock_bh(&idev->lock);
1346                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1347                         __func__);
1348                 in6_dev_put(idev);
1349                 ret = -1;
1350                 goto out;
1351         }
1352         in6_ifa_hold(ifp);
1353         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1354         ipv6_try_regen_rndid(idev, tmpaddr);
1355         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1356         age = (now - ifp->tstamp) / HZ;
1357
1358         regen_advance = idev->cnf.regen_max_retry *
1359                         idev->cnf.dad_transmits *
1360                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1361
1362         /* recalculate max_desync_factor each time and update
1363          * idev->desync_factor if it's larger
1364          */
1365         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1366         max_desync_factor = min_t(__u32,
1367                                   idev->cnf.max_desync_factor,
1368                                   cnf_temp_preferred_lft - regen_advance);
1369
1370         if (unlikely(idev->desync_factor > max_desync_factor)) {
1371                 if (max_desync_factor > 0) {
1372                         get_random_bytes(&idev->desync_factor,
1373                                          sizeof(idev->desync_factor));
1374                         idev->desync_factor %= max_desync_factor;
1375                 } else {
1376                         idev->desync_factor = 0;
1377                 }
1378         }
1379
1380         memset(&cfg, 0, sizeof(cfg));
1381         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1382                               idev->cnf.temp_valid_lft + age);
1383         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1384         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1385
1386         cfg.plen = ifp->prefix_len;
1387         tmp_tstamp = ifp->tstamp;
1388         spin_unlock_bh(&ifp->lock);
1389
1390         write_unlock_bh(&idev->lock);
1391
1392         /* A temporary address is created only if this calculated Preferred
1393          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1394          * an implementation must not create a temporary address with a zero
1395          * Preferred Lifetime.
1396          * Use age calculation as in addrconf_verify to avoid unnecessary
1397          * temporary addresses being generated.
1398          */
1399         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1400         if (cfg.preferred_lft <= regen_advance + age) {
1401                 in6_ifa_put(ifp);
1402                 in6_dev_put(idev);
1403                 ret = -1;
1404                 goto out;
1405         }
1406
1407         cfg.ifa_flags = IFA_F_TEMPORARY;
1408         /* set in addrconf_prefix_rcv() */
1409         if (ifp->flags & IFA_F_OPTIMISTIC)
1410                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1411
1412         cfg.pfx = &addr;
1413         cfg.scope = ipv6_addr_scope(cfg.pfx);
1414
1415         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1416         if (IS_ERR(ift)) {
1417                 in6_ifa_put(ifp);
1418                 in6_dev_put(idev);
1419                 pr_info("%s: retry temporary address regeneration\n", __func__);
1420                 tmpaddr = &addr;
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 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1908                             const struct net_device *dev, bool skip_dev_check,
1909                             int strict, u32 banned_flags)
1910 {
1911         unsigned int hash = inet6_addr_hash(net, addr);
1912         const struct net_device *l3mdev;
1913         struct inet6_ifaddr *ifp;
1914         u32 ifp_flags;
1915
1916         rcu_read_lock();
1917
1918         l3mdev = l3mdev_master_dev_rcu(dev);
1919         if (skip_dev_check)
1920                 dev = NULL;
1921
1922         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1923                 if (!net_eq(dev_net(ifp->idev->dev), net))
1924                         continue;
1925
1926                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1927                         continue;
1928
1929                 /* Decouple optimistic from tentative for evaluation here.
1930                  * Ban optimistic addresses explicitly, when required.
1931                  */
1932                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1933                             ? (ifp->flags&~IFA_F_TENTATIVE)
1934                             : ifp->flags;
1935                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1936                     !(ifp_flags&banned_flags) &&
1937                     (!dev || ifp->idev->dev == dev ||
1938                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1939                         rcu_read_unlock();
1940                         return 1;
1941                 }
1942         }
1943
1944         rcu_read_unlock();
1945         return 0;
1946 }
1947 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1948
1949
1950 /* Compares an address/prefix_len with addresses on device @dev.
1951  * If one is found it returns true.
1952  */
1953 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1954         const unsigned int prefix_len, struct net_device *dev)
1955 {
1956         const struct inet6_ifaddr *ifa;
1957         const struct inet6_dev *idev;
1958         bool ret = false;
1959
1960         rcu_read_lock();
1961         idev = __in6_dev_get(dev);
1962         if (idev) {
1963                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1964                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1965                         if (ret)
1966                                 break;
1967                 }
1968         }
1969         rcu_read_unlock();
1970
1971         return ret;
1972 }
1973 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1974
1975 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1976 {
1977         const struct inet6_ifaddr *ifa;
1978         const struct inet6_dev *idev;
1979         int     onlink;
1980
1981         onlink = 0;
1982         rcu_read_lock();
1983         idev = __in6_dev_get(dev);
1984         if (idev) {
1985                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1986                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1987                                                    ifa->prefix_len);
1988                         if (onlink)
1989                                 break;
1990                 }
1991         }
1992         rcu_read_unlock();
1993         return onlink;
1994 }
1995 EXPORT_SYMBOL(ipv6_chk_prefix);
1996
1997 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1998                                      struct net_device *dev, int strict)
1999 {
2000         unsigned int hash = inet6_addr_hash(net, addr);
2001         struct inet6_ifaddr *ifp, *result = NULL;
2002
2003         rcu_read_lock();
2004         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2005                 if (!net_eq(dev_net(ifp->idev->dev), net))
2006                         continue;
2007                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2008                         if (!dev || ifp->idev->dev == dev ||
2009                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2010                                 result = ifp;
2011                                 in6_ifa_hold(ifp);
2012                                 break;
2013                         }
2014                 }
2015         }
2016         rcu_read_unlock();
2017
2018         return result;
2019 }
2020
2021 /* Gets referenced address, destroys ifaddr */
2022
2023 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2024 {
2025         if (dad_failed)
2026                 ifp->flags |= IFA_F_DADFAILED;
2027
2028         if (ifp->flags&IFA_F_TEMPORARY) {
2029                 struct inet6_ifaddr *ifpub;
2030                 spin_lock_bh(&ifp->lock);
2031                 ifpub = ifp->ifpub;
2032                 if (ifpub) {
2033                         in6_ifa_hold(ifpub);
2034                         spin_unlock_bh(&ifp->lock);
2035                         ipv6_create_tempaddr(ifpub, ifp, true);
2036                         in6_ifa_put(ifpub);
2037                 } else {
2038                         spin_unlock_bh(&ifp->lock);
2039                 }
2040                 ipv6_del_addr(ifp);
2041         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2042                 spin_lock_bh(&ifp->lock);
2043                 addrconf_del_dad_work(ifp);
2044                 ifp->flags |= IFA_F_TENTATIVE;
2045                 if (dad_failed)
2046                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2047                 spin_unlock_bh(&ifp->lock);
2048                 if (dad_failed)
2049                         ipv6_ifa_notify(0, ifp);
2050                 in6_ifa_put(ifp);
2051         } else {
2052                 ipv6_del_addr(ifp);
2053         }
2054 }
2055
2056 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2057 {
2058         int err = -ENOENT;
2059
2060         spin_lock_bh(&ifp->lock);
2061         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2062                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2063                 err = 0;
2064         }
2065         spin_unlock_bh(&ifp->lock);
2066
2067         return err;
2068 }
2069
2070 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2071 {
2072         struct inet6_dev *idev = ifp->idev;
2073         struct net *net = dev_net(ifp->idev->dev);
2074
2075         if (addrconf_dad_end(ifp)) {
2076                 in6_ifa_put(ifp);
2077                 return;
2078         }
2079
2080         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2081                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2082
2083         spin_lock_bh(&ifp->lock);
2084
2085         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2086                 struct in6_addr new_addr;
2087                 struct inet6_ifaddr *ifp2;
2088                 int retries = ifp->stable_privacy_retry + 1;
2089                 struct ifa6_config cfg = {
2090                         .pfx = &new_addr,
2091                         .plen = ifp->prefix_len,
2092                         .ifa_flags = ifp->flags,
2093                         .valid_lft = ifp->valid_lft,
2094                         .preferred_lft = ifp->prefered_lft,
2095                         .scope = ifp->scope,
2096                 };
2097
2098                 if (retries > net->ipv6.sysctl.idgen_retries) {
2099                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2100                                              ifp->idev->dev->name);
2101                         goto errdad;
2102                 }
2103
2104                 new_addr = ifp->addr;
2105                 if (ipv6_generate_stable_address(&new_addr, retries,
2106                                                  idev))
2107                         goto errdad;
2108
2109                 spin_unlock_bh(&ifp->lock);
2110
2111                 if (idev->cnf.max_addresses &&
2112                     ipv6_count_addresses(idev) >=
2113                     idev->cnf.max_addresses)
2114                         goto lock_errdad;
2115
2116                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2117                                      ifp->idev->dev->name);
2118
2119                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2120                 if (IS_ERR(ifp2))
2121                         goto lock_errdad;
2122
2123                 spin_lock_bh(&ifp2->lock);
2124                 ifp2->stable_privacy_retry = retries;
2125                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2126                 spin_unlock_bh(&ifp2->lock);
2127
2128                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2129                 in6_ifa_put(ifp2);
2130 lock_errdad:
2131                 spin_lock_bh(&ifp->lock);
2132         }
2133
2134 errdad:
2135         /* transition from _POSTDAD to _ERRDAD */
2136         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2137         spin_unlock_bh(&ifp->lock);
2138
2139         addrconf_mod_dad_work(ifp, 0);
2140         in6_ifa_put(ifp);
2141 }
2142
2143 /* Join to solicited addr multicast group.
2144  * caller must hold RTNL */
2145 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2146 {
2147         struct in6_addr maddr;
2148
2149         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2150                 return;
2151
2152         addrconf_addr_solict_mult(addr, &maddr);
2153         ipv6_dev_mc_inc(dev, &maddr);
2154 }
2155
2156 /* caller must hold RTNL */
2157 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2158 {
2159         struct in6_addr maddr;
2160
2161         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2162                 return;
2163
2164         addrconf_addr_solict_mult(addr, &maddr);
2165         __ipv6_dev_mc_dec(idev, &maddr);
2166 }
2167
2168 /* caller must hold RTNL */
2169 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2170 {
2171         struct in6_addr addr;
2172
2173         if (ifp->prefix_len >= 127) /* RFC 6164 */
2174                 return;
2175         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2176         if (ipv6_addr_any(&addr))
2177                 return;
2178         __ipv6_dev_ac_inc(ifp->idev, &addr);
2179 }
2180
2181 /* caller must hold RTNL */
2182 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2183 {
2184         struct in6_addr addr;
2185
2186         if (ifp->prefix_len >= 127) /* RFC 6164 */
2187                 return;
2188         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2189         if (ipv6_addr_any(&addr))
2190                 return;
2191         __ipv6_dev_ac_dec(ifp->idev, &addr);
2192 }
2193
2194 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2195 {
2196         switch (dev->addr_len) {
2197         case ETH_ALEN:
2198                 memcpy(eui, dev->dev_addr, 3);
2199                 eui[3] = 0xFF;
2200                 eui[4] = 0xFE;
2201                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2202                 break;
2203         case EUI64_ADDR_LEN:
2204                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2205                 eui[0] ^= 2;
2206                 break;
2207         default:
2208                 return -1;
2209         }
2210
2211         return 0;
2212 }
2213
2214 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2215 {
2216         union fwnet_hwaddr *ha;
2217
2218         if (dev->addr_len != FWNET_ALEN)
2219                 return -1;
2220
2221         ha = (union fwnet_hwaddr *)dev->dev_addr;
2222
2223         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2224         eui[0] ^= 2;
2225         return 0;
2226 }
2227
2228 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2229 {
2230         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2231         if (dev->addr_len != ARCNET_ALEN)
2232                 return -1;
2233         memset(eui, 0, 7);
2234         eui[7] = *(u8 *)dev->dev_addr;
2235         return 0;
2236 }
2237
2238 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2239 {
2240         if (dev->addr_len != INFINIBAND_ALEN)
2241                 return -1;
2242         memcpy(eui, dev->dev_addr + 12, 8);
2243         eui[0] |= 2;
2244         return 0;
2245 }
2246
2247 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2248 {
2249         if (addr == 0)
2250                 return -1;
2251         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2252                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2253                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2254                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2255                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2256                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2257         eui[1] = 0;
2258         eui[2] = 0x5E;
2259         eui[3] = 0xFE;
2260         memcpy(eui + 4, &addr, 4);
2261         return 0;
2262 }
2263
2264 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2265 {
2266         if (dev->priv_flags & IFF_ISATAP)
2267                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2268         return -1;
2269 }
2270
2271 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2272 {
2273         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2274 }
2275
2276 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2277 {
2278         memcpy(eui, dev->perm_addr, 3);
2279         memcpy(eui + 5, dev->perm_addr + 3, 3);
2280         eui[3] = 0xFF;
2281         eui[4] = 0xFE;
2282         eui[0] ^= 2;
2283         return 0;
2284 }
2285
2286 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2287 {
2288         switch (dev->type) {
2289         case ARPHRD_ETHER:
2290         case ARPHRD_FDDI:
2291                 return addrconf_ifid_eui48(eui, dev);
2292         case ARPHRD_ARCNET:
2293                 return addrconf_ifid_arcnet(eui, dev);
2294         case ARPHRD_INFINIBAND:
2295                 return addrconf_ifid_infiniband(eui, dev);
2296         case ARPHRD_SIT:
2297                 return addrconf_ifid_sit(eui, dev);
2298         case ARPHRD_IPGRE:
2299         case ARPHRD_TUNNEL:
2300                 return addrconf_ifid_gre(eui, dev);
2301         case ARPHRD_6LOWPAN:
2302                 return addrconf_ifid_6lowpan(eui, dev);
2303         case ARPHRD_IEEE1394:
2304                 return addrconf_ifid_ieee1394(eui, dev);
2305         case ARPHRD_TUNNEL6:
2306         case ARPHRD_IP6GRE:
2307         case ARPHRD_RAWIP:
2308                 return addrconf_ifid_ip6tnl(eui, dev);
2309         }
2310         return -1;
2311 }
2312
2313 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2314 {
2315         int err = -1;
2316         struct inet6_ifaddr *ifp;
2317
2318         read_lock_bh(&idev->lock);
2319         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2320                 if (ifp->scope > IFA_LINK)
2321                         break;
2322                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2323                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2324                         err = 0;
2325                         break;
2326                 }
2327         }
2328         read_unlock_bh(&idev->lock);
2329         return err;
2330 }
2331
2332 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2333 static void ipv6_regen_rndid(struct inet6_dev *idev)
2334 {
2335 regen:
2336         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2337         idev->rndid[0] &= ~0x02;
2338
2339         /*
2340          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2341          * check if generated address is not inappropriate
2342          *
2343          *  - Reserved subnet anycast (RFC 2526)
2344          *      11111101 11....11 1xxxxxxx
2345          *  - ISATAP (RFC4214) 6.1
2346          *      00-00-5E-FE-xx-xx-xx-xx
2347          *  - value 0
2348          *  - XXX: already assigned to an address on the device
2349          */
2350         if (idev->rndid[0] == 0xfd &&
2351             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2352             (idev->rndid[7]&0x80))
2353                 goto regen;
2354         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2355                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2356                         goto regen;
2357                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2358                         goto regen;
2359         }
2360 }
2361
2362 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2363 {
2364         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2365                 ipv6_regen_rndid(idev);
2366 }
2367
2368 /*
2369  *      Add prefix route.
2370  */
2371
2372 static void
2373 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2374                       struct net_device *dev, unsigned long expires,
2375                       u32 flags, gfp_t gfp_flags)
2376 {
2377         struct fib6_config cfg = {
2378                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2379                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2380                 .fc_ifindex = dev->ifindex,
2381                 .fc_expires = expires,
2382                 .fc_dst_len = plen,
2383                 .fc_flags = RTF_UP | flags,
2384                 .fc_nlinfo.nl_net = dev_net(dev),
2385                 .fc_protocol = RTPROT_KERNEL,
2386                 .fc_type = RTN_UNICAST,
2387         };
2388
2389         cfg.fc_dst = *pfx;
2390
2391         /* Prevent useless cloning on PtP SIT.
2392            This thing is done here expecting that the whole
2393            class of non-broadcast devices need not cloning.
2394          */
2395 #if IS_ENABLED(CONFIG_IPV6_SIT)
2396         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2397                 cfg.fc_flags |= RTF_NONEXTHOP;
2398 #endif
2399
2400         ip6_route_add(&cfg, gfp_flags, NULL);
2401 }
2402
2403
2404 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2405                                                   int plen,
2406                                                   const struct net_device *dev,
2407                                                   u32 flags, u32 noflags,
2408                                                   bool no_gw)
2409 {
2410         struct fib6_node *fn;
2411         struct fib6_info *rt = NULL;
2412         struct fib6_table *table;
2413         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2414
2415         table = fib6_get_table(dev_net(dev), tb_id);
2416         if (!table)
2417                 return NULL;
2418
2419         rcu_read_lock();
2420         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2421         if (!fn)
2422                 goto out;
2423
2424         for_each_fib6_node_rt_rcu(fn) {
2425                 /* prefix routes only use builtin fib6_nh */
2426                 if (rt->nh)
2427                         continue;
2428
2429                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2430                         continue;
2431                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2432                         continue;
2433                 if ((rt->fib6_flags & flags) != flags)
2434                         continue;
2435                 if ((rt->fib6_flags & noflags) != 0)
2436                         continue;
2437                 if (!fib6_info_hold_safe(rt))
2438                         continue;
2439                 break;
2440         }
2441 out:
2442         rcu_read_unlock();
2443         return rt;
2444 }
2445
2446
2447 /* Create "default" multicast route to the interface */
2448
2449 static void addrconf_add_mroute(struct net_device *dev)
2450 {
2451         struct fib6_config cfg = {
2452                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2453                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2454                 .fc_ifindex = dev->ifindex,
2455                 .fc_dst_len = 8,
2456                 .fc_flags = RTF_UP,
2457                 .fc_type = RTN_UNICAST,
2458                 .fc_nlinfo.nl_net = dev_net(dev),
2459         };
2460
2461         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2462
2463         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2464 }
2465
2466 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2467 {
2468         struct inet6_dev *idev;
2469
2470         ASSERT_RTNL();
2471
2472         idev = ipv6_find_idev(dev);
2473         if (IS_ERR(idev))
2474                 return idev;
2475
2476         if (idev->cnf.disable_ipv6)
2477                 return ERR_PTR(-EACCES);
2478
2479         /* Add default multicast route */
2480         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2481                 addrconf_add_mroute(dev);
2482
2483         return idev;
2484 }
2485
2486 static void manage_tempaddrs(struct inet6_dev *idev,
2487                              struct inet6_ifaddr *ifp,
2488                              __u32 valid_lft, __u32 prefered_lft,
2489                              bool create, unsigned long now)
2490 {
2491         u32 flags;
2492         struct inet6_ifaddr *ift;
2493
2494         read_lock_bh(&idev->lock);
2495         /* update all temporary addresses in the list */
2496         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2497                 int age, max_valid, max_prefered;
2498
2499                 if (ifp != ift->ifpub)
2500                         continue;
2501
2502                 /* RFC 4941 section 3.3:
2503                  * If a received option will extend the lifetime of a public
2504                  * address, the lifetimes of temporary addresses should
2505                  * be extended, subject to the overall constraint that no
2506                  * temporary addresses should ever remain "valid" or "preferred"
2507                  * for a time longer than (TEMP_VALID_LIFETIME) or
2508                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2509                  */
2510                 age = (now - ift->cstamp) / HZ;
2511                 max_valid = idev->cnf.temp_valid_lft - age;
2512                 if (max_valid < 0)
2513                         max_valid = 0;
2514
2515                 max_prefered = idev->cnf.temp_prefered_lft -
2516                                idev->desync_factor - age;
2517                 if (max_prefered < 0)
2518                         max_prefered = 0;
2519
2520                 if (valid_lft > max_valid)
2521                         valid_lft = max_valid;
2522
2523                 if (prefered_lft > max_prefered)
2524                         prefered_lft = max_prefered;
2525
2526                 spin_lock(&ift->lock);
2527                 flags = ift->flags;
2528                 ift->valid_lft = valid_lft;
2529                 ift->prefered_lft = prefered_lft;
2530                 ift->tstamp = now;
2531                 if (prefered_lft > 0)
2532                         ift->flags &= ~IFA_F_DEPRECATED;
2533
2534                 spin_unlock(&ift->lock);
2535                 if (!(flags&IFA_F_TENTATIVE))
2536                         ipv6_ifa_notify(0, ift);
2537         }
2538
2539         if ((create || list_empty(&idev->tempaddr_list)) &&
2540             idev->cnf.use_tempaddr > 0) {
2541                 /* When a new public address is created as described
2542                  * in [ADDRCONF], also create a new temporary address.
2543                  * Also create a temporary address if it's enabled but
2544                  * no temporary address currently exists.
2545                  */
2546                 read_unlock_bh(&idev->lock);
2547                 ipv6_create_tempaddr(ifp, NULL, false);
2548         } else {
2549                 read_unlock_bh(&idev->lock);
2550         }
2551 }
2552
2553 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2554 {
2555         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2556                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2557 }
2558
2559 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2560                                  const struct prefix_info *pinfo,
2561                                  struct inet6_dev *in6_dev,
2562                                  const struct in6_addr *addr, int addr_type,
2563                                  u32 addr_flags, bool sllao, bool tokenized,
2564                                  __u32 valid_lft, u32 prefered_lft)
2565 {
2566         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2567         int create = 0, update_lft = 0;
2568
2569         if (!ifp && valid_lft) {
2570                 int max_addresses = in6_dev->cnf.max_addresses;
2571                 struct ifa6_config cfg = {
2572                         .pfx = addr,
2573                         .plen = pinfo->prefix_len,
2574                         .ifa_flags = addr_flags,
2575                         .valid_lft = valid_lft,
2576                         .preferred_lft = prefered_lft,
2577                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2578                 };
2579
2580 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2581                 if ((net->ipv6.devconf_all->optimistic_dad ||
2582                      in6_dev->cnf.optimistic_dad) &&
2583                     !net->ipv6.devconf_all->forwarding && sllao)
2584                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2585 #endif
2586
2587                 /* Do not allow to create too much of autoconfigured
2588                  * addresses; this would be too easy way to crash kernel.
2589                  */
2590                 if (!max_addresses ||
2591                     ipv6_count_addresses(in6_dev) < max_addresses)
2592                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2593
2594                 if (IS_ERR_OR_NULL(ifp))
2595                         return -1;
2596
2597                 create = 1;
2598                 spin_lock_bh(&ifp->lock);
2599                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2600                 ifp->cstamp = jiffies;
2601                 ifp->tokenized = tokenized;
2602                 spin_unlock_bh(&ifp->lock);
2603                 addrconf_dad_start(ifp);
2604         }
2605
2606         if (ifp) {
2607                 u32 flags;
2608                 unsigned long now;
2609                 u32 stored_lft;
2610
2611                 /* update lifetime (RFC2462 5.5.3 e) */
2612                 spin_lock_bh(&ifp->lock);
2613                 now = jiffies;
2614                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2615                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2616                 else
2617                         stored_lft = 0;
2618                 if (!create && stored_lft) {
2619                         const u32 minimum_lft = min_t(u32,
2620                                 stored_lft, MIN_VALID_LIFETIME);
2621                         valid_lft = max(valid_lft, minimum_lft);
2622
2623                         /* RFC4862 Section 5.5.3e:
2624                          * "Note that the preferred lifetime of the
2625                          *  corresponding address is always reset to
2626                          *  the Preferred Lifetime in the received
2627                          *  Prefix Information option, regardless of
2628                          *  whether the valid lifetime is also reset or
2629                          *  ignored."
2630                          *
2631                          * So we should always update prefered_lft here.
2632                          */
2633                         update_lft = 1;
2634                 }
2635
2636                 if (update_lft) {
2637                         ifp->valid_lft = valid_lft;
2638                         ifp->prefered_lft = prefered_lft;
2639                         ifp->tstamp = now;
2640                         flags = ifp->flags;
2641                         ifp->flags &= ~IFA_F_DEPRECATED;
2642                         spin_unlock_bh(&ifp->lock);
2643
2644                         if (!(flags&IFA_F_TENTATIVE))
2645                                 ipv6_ifa_notify(0, ifp);
2646                 } else
2647                         spin_unlock_bh(&ifp->lock);
2648
2649                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2650                                  create, now);
2651
2652                 in6_ifa_put(ifp);
2653                 addrconf_verify();
2654         }
2655
2656         return 0;
2657 }
2658 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2659
2660 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2661 {
2662         struct prefix_info *pinfo;
2663         __u32 valid_lft;
2664         __u32 prefered_lft;
2665         int addr_type, err;
2666         u32 addr_flags = 0;
2667         struct inet6_dev *in6_dev;
2668         struct net *net = dev_net(dev);
2669
2670         pinfo = (struct prefix_info *) opt;
2671
2672         if (len < sizeof(struct prefix_info)) {
2673                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2674                 return;
2675         }
2676
2677         /*
2678          *      Validation checks ([ADDRCONF], page 19)
2679          */
2680
2681         addr_type = ipv6_addr_type(&pinfo->prefix);
2682
2683         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2684                 return;
2685
2686         valid_lft = ntohl(pinfo->valid);
2687         prefered_lft = ntohl(pinfo->prefered);
2688
2689         if (prefered_lft > valid_lft) {
2690                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2691                 return;
2692         }
2693
2694         in6_dev = in6_dev_get(dev);
2695
2696         if (!in6_dev) {
2697                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2698                                     dev->name);
2699                 return;
2700         }
2701
2702         /*
2703          *      Two things going on here:
2704          *      1) Add routes for on-link prefixes
2705          *      2) Configure prefixes with the auto flag set
2706          */
2707
2708         if (pinfo->onlink) {
2709                 struct fib6_info *rt;
2710                 unsigned long rt_expires;
2711
2712                 /* Avoid arithmetic overflow. Really, we could
2713                  * save rt_expires in seconds, likely valid_lft,
2714                  * but it would require division in fib gc, that it
2715                  * not good.
2716                  */
2717                 if (HZ > USER_HZ)
2718                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2719                 else
2720                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2721
2722                 if (addrconf_finite_timeout(rt_expires))
2723                         rt_expires *= HZ;
2724
2725                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2726                                                pinfo->prefix_len,
2727                                                dev,
2728                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2729                                                RTF_DEFAULT, true);
2730
2731                 if (rt) {
2732                         /* Autoconf prefix route */
2733                         if (valid_lft == 0) {
2734                                 ip6_del_rt(net, rt);
2735                                 rt = NULL;
2736                         } else if (addrconf_finite_timeout(rt_expires)) {
2737                                 /* not infinity */
2738                                 fib6_set_expires(rt, jiffies + rt_expires);
2739                         } else {
2740                                 fib6_clean_expires(rt);
2741                         }
2742                 } else if (valid_lft) {
2743                         clock_t expires = 0;
2744                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2745                         if (addrconf_finite_timeout(rt_expires)) {
2746                                 /* not infinity */
2747                                 flags |= RTF_EXPIRES;
2748                                 expires = jiffies_to_clock_t(rt_expires);
2749                         }
2750                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2751                                               0, dev, expires, flags,
2752                                               GFP_ATOMIC);
2753                 }
2754                 fib6_info_release(rt);
2755         }
2756
2757         /* Try to figure out our local address for this prefix */
2758
2759         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2760                 struct in6_addr addr;
2761                 bool tokenized = false, dev_addr_generated = false;
2762
2763                 if (pinfo->prefix_len == 64) {
2764                         memcpy(&addr, &pinfo->prefix, 8);
2765
2766                         if (!ipv6_addr_any(&in6_dev->token)) {
2767                                 read_lock_bh(&in6_dev->lock);
2768                                 memcpy(addr.s6_addr + 8,
2769                                        in6_dev->token.s6_addr + 8, 8);
2770                                 read_unlock_bh(&in6_dev->lock);
2771                                 tokenized = true;
2772                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2773                                    !ipv6_generate_stable_address(&addr, 0,
2774                                                                  in6_dev)) {
2775                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2776                                 goto ok;
2777                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2778                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2779                                 goto put;
2780                         } else {
2781                                 dev_addr_generated = true;
2782                         }
2783                         goto ok;
2784                 }
2785                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2786                                     pinfo->prefix_len);
2787                 goto put;
2788
2789 ok:
2790                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2791                                                    &addr, addr_type,
2792                                                    addr_flags, sllao,
2793                                                    tokenized, valid_lft,
2794                                                    prefered_lft);
2795                 if (err)
2796                         goto put;
2797
2798                 /* Ignore error case here because previous prefix add addr was
2799                  * successful which will be notified.
2800                  */
2801                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2802                                               addr_type, addr_flags, sllao,
2803                                               tokenized, valid_lft,
2804                                               prefered_lft,
2805                                               dev_addr_generated);
2806         }
2807         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2808 put:
2809         in6_dev_put(in6_dev);
2810 }
2811
2812 /*
2813  *      Set destination address.
2814  *      Special case for SIT interfaces where we create a new "virtual"
2815  *      device.
2816  */
2817 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2818 {
2819         struct in6_ifreq ireq;
2820         struct net_device *dev;
2821         int err = -EINVAL;
2822
2823         rtnl_lock();
2824
2825         err = -EFAULT;
2826         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2827                 goto err_exit;
2828
2829         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2830
2831         err = -ENODEV;
2832         if (!dev)
2833                 goto err_exit;
2834
2835 #if IS_ENABLED(CONFIG_IPV6_SIT)
2836         if (dev->type == ARPHRD_SIT) {
2837                 const struct net_device_ops *ops = dev->netdev_ops;
2838                 struct ifreq ifr;
2839                 struct ip_tunnel_parm p;
2840
2841                 err = -EADDRNOTAVAIL;
2842                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2843                         goto err_exit;
2844
2845                 memset(&p, 0, sizeof(p));
2846                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2847                 p.iph.saddr = 0;
2848                 p.iph.version = 4;
2849                 p.iph.ihl = 5;
2850                 p.iph.protocol = IPPROTO_IPV6;
2851                 p.iph.ttl = 64;
2852                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2853
2854                 if (ops->ndo_do_ioctl) {
2855                         mm_segment_t oldfs = get_fs();
2856
2857                         set_fs(KERNEL_DS);
2858                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2859                         set_fs(oldfs);
2860                 } else
2861                         err = -EOPNOTSUPP;
2862
2863                 if (err == 0) {
2864                         err = -ENOBUFS;
2865                         dev = __dev_get_by_name(net, p.name);
2866                         if (!dev)
2867                                 goto err_exit;
2868                         err = dev_open(dev, NULL);
2869                 }
2870         }
2871 #endif
2872
2873 err_exit:
2874         rtnl_unlock();
2875         return err;
2876 }
2877
2878 static int ipv6_mc_config(struct sock *sk, bool join,
2879                           const struct in6_addr *addr, int ifindex)
2880 {
2881         int ret;
2882
2883         ASSERT_RTNL();
2884
2885         lock_sock(sk);
2886         if (join)
2887                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2888         else
2889                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2890         release_sock(sk);
2891
2892         return ret;
2893 }
2894
2895 /*
2896  *      Manual configuration of address on an interface
2897  */
2898 static int inet6_addr_add(struct net *net, int ifindex,
2899                           struct ifa6_config *cfg,
2900                           struct netlink_ext_ack *extack)
2901 {
2902         struct inet6_ifaddr *ifp;
2903         struct inet6_dev *idev;
2904         struct net_device *dev;
2905         unsigned long timeout;
2906         clock_t expires;
2907         u32 flags;
2908
2909         ASSERT_RTNL();
2910
2911         if (cfg->plen > 128)
2912                 return -EINVAL;
2913
2914         /* check the lifetime */
2915         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2916                 return -EINVAL;
2917
2918         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2919                 return -EINVAL;
2920
2921         dev = __dev_get_by_index(net, ifindex);
2922         if (!dev)
2923                 return -ENODEV;
2924
2925         idev = addrconf_add_dev(dev);
2926         if (IS_ERR(idev))
2927                 return PTR_ERR(idev);
2928
2929         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2930                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2931                                          true, cfg->pfx, ifindex);
2932
2933                 if (ret < 0)
2934                         return ret;
2935         }
2936
2937         cfg->scope = ipv6_addr_scope(cfg->pfx);
2938
2939         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2940         if (addrconf_finite_timeout(timeout)) {
2941                 expires = jiffies_to_clock_t(timeout * HZ);
2942                 cfg->valid_lft = timeout;
2943                 flags = RTF_EXPIRES;
2944         } else {
2945                 expires = 0;
2946                 flags = 0;
2947                 cfg->ifa_flags |= IFA_F_PERMANENT;
2948         }
2949
2950         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2951         if (addrconf_finite_timeout(timeout)) {
2952                 if (timeout == 0)
2953                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2954                 cfg->preferred_lft = timeout;
2955         }
2956
2957         ifp = ipv6_add_addr(idev, cfg, true, extack);
2958         if (!IS_ERR(ifp)) {
2959                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2960                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2961                                               ifp->rt_priority, dev, expires,
2962                                               flags, GFP_KERNEL);
2963                 }
2964
2965                 /* Send a netlink notification if DAD is enabled and
2966                  * optimistic flag is not set
2967                  */
2968                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2969                         ipv6_ifa_notify(0, ifp);
2970                 /*
2971                  * Note that section 3.1 of RFC 4429 indicates
2972                  * that the Optimistic flag should not be set for
2973                  * manually configured addresses
2974                  */
2975                 addrconf_dad_start(ifp);
2976                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2977                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2978                                          cfg->preferred_lft, true, jiffies);
2979                 in6_ifa_put(ifp);
2980                 addrconf_verify_rtnl();
2981                 return 0;
2982         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2983                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2984                                cfg->pfx, ifindex);
2985         }
2986
2987         return PTR_ERR(ifp);
2988 }
2989
2990 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2991                           const struct in6_addr *pfx, unsigned int plen)
2992 {
2993         struct inet6_ifaddr *ifp;
2994         struct inet6_dev *idev;
2995         struct net_device *dev;
2996
2997         if (plen > 128)
2998                 return -EINVAL;
2999
3000         dev = __dev_get_by_index(net, ifindex);
3001         if (!dev)
3002                 return -ENODEV;
3003
3004         idev = __in6_dev_get(dev);
3005         if (!idev)
3006                 return -ENXIO;
3007
3008         read_lock_bh(&idev->lock);
3009         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3010                 if (ifp->prefix_len == plen &&
3011                     ipv6_addr_equal(pfx, &ifp->addr)) {
3012                         in6_ifa_hold(ifp);
3013                         read_unlock_bh(&idev->lock);
3014
3015                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3016                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3017                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3018                                                  jiffies);
3019                         ipv6_del_addr(ifp);
3020                         addrconf_verify_rtnl();
3021                         if (ipv6_addr_is_multicast(pfx)) {
3022                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3023                                                false, pfx, dev->ifindex);
3024                         }
3025                         return 0;
3026                 }
3027         }
3028         read_unlock_bh(&idev->lock);
3029         return -EADDRNOTAVAIL;
3030 }
3031
3032
3033 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3034 {
3035         struct ifa6_config cfg = {
3036                 .ifa_flags = IFA_F_PERMANENT,
3037                 .preferred_lft = INFINITY_LIFE_TIME,
3038                 .valid_lft = INFINITY_LIFE_TIME,
3039         };
3040         struct in6_ifreq ireq;
3041         int err;
3042
3043         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3044                 return -EPERM;
3045
3046         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3047                 return -EFAULT;
3048
3049         cfg.pfx = &ireq.ifr6_addr;
3050         cfg.plen = ireq.ifr6_prefixlen;
3051
3052         rtnl_lock();
3053         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3054         rtnl_unlock();
3055         return err;
3056 }
3057
3058 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3059 {
3060         struct in6_ifreq ireq;
3061         int err;
3062
3063         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3064                 return -EPERM;
3065
3066         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3067                 return -EFAULT;
3068
3069         rtnl_lock();
3070         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3071                              ireq.ifr6_prefixlen);
3072         rtnl_unlock();
3073         return err;
3074 }
3075
3076 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3077                      int plen, int scope)
3078 {
3079         struct inet6_ifaddr *ifp;
3080         struct ifa6_config cfg = {
3081                 .pfx = addr,
3082                 .plen = plen,
3083                 .ifa_flags = IFA_F_PERMANENT,
3084                 .valid_lft = INFINITY_LIFE_TIME,
3085                 .preferred_lft = INFINITY_LIFE_TIME,
3086                 .scope = scope
3087         };
3088
3089         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3090         if (!IS_ERR(ifp)) {
3091                 spin_lock_bh(&ifp->lock);
3092                 ifp->flags &= ~IFA_F_TENTATIVE;
3093                 spin_unlock_bh(&ifp->lock);
3094                 rt_genid_bump_ipv6(dev_net(idev->dev));
3095                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3096                 in6_ifa_put(ifp);
3097         }
3098 }
3099
3100 #if IS_ENABLED(CONFIG_IPV6_SIT)
3101 static void sit_add_v4_addrs(struct inet6_dev *idev)
3102 {
3103         struct in6_addr addr;
3104         struct net_device *dev;
3105         struct net *net = dev_net(idev->dev);
3106         int scope, plen;
3107         u32 pflags = 0;
3108
3109         ASSERT_RTNL();
3110
3111         memset(&addr, 0, sizeof(struct in6_addr));
3112         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3113
3114         if (idev->dev->flags&IFF_POINTOPOINT) {
3115                 addr.s6_addr32[0] = htonl(0xfe800000);
3116                 scope = IFA_LINK;
3117                 plen = 64;
3118         } else {
3119                 scope = IPV6_ADDR_COMPATv4;
3120                 plen = 96;
3121                 pflags |= RTF_NONEXTHOP;
3122         }
3123
3124         if (addr.s6_addr32[3]) {
3125                 add_addr(idev, &addr, plen, scope);
3126                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3127                                       GFP_KERNEL);
3128                 return;
3129         }
3130
3131         for_each_netdev(net, dev) {
3132                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3133                 if (in_dev && (dev->flags & IFF_UP)) {
3134                         struct in_ifaddr *ifa;
3135                         int flag = scope;
3136
3137                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3138                                 addr.s6_addr32[3] = ifa->ifa_local;
3139
3140                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3141                                         continue;
3142                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3143                                         if (idev->dev->flags&IFF_POINTOPOINT)
3144                                                 continue;
3145                                         flag |= IFA_HOST;
3146                                 }
3147
3148                                 add_addr(idev, &addr, plen, flag);
3149                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3150                                                       0, pflags, GFP_KERNEL);
3151                         }
3152                 }
3153         }
3154 }
3155 #endif
3156
3157 static void init_loopback(struct net_device *dev)
3158 {
3159         struct inet6_dev  *idev;
3160
3161         /* ::1 */
3162
3163         ASSERT_RTNL();
3164
3165         idev = ipv6_find_idev(dev);
3166         if (IS_ERR(idev)) {
3167                 pr_debug("%s: add_dev failed\n", __func__);
3168                 return;
3169         }
3170
3171         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3172 }
3173
3174 void addrconf_add_linklocal(struct inet6_dev *idev,
3175                             const struct in6_addr *addr, u32 flags)
3176 {
3177         struct ifa6_config cfg = {
3178                 .pfx = addr,
3179                 .plen = 64,
3180                 .ifa_flags = flags | IFA_F_PERMANENT,
3181                 .valid_lft = INFINITY_LIFE_TIME,
3182                 .preferred_lft = INFINITY_LIFE_TIME,
3183                 .scope = IFA_LINK
3184         };
3185         struct inet6_ifaddr *ifp;
3186
3187 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3188         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3189              idev->cnf.optimistic_dad) &&
3190             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3191                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3192 #endif
3193
3194         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3195         if (!IS_ERR(ifp)) {
3196                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3197                                       0, 0, GFP_ATOMIC);
3198                 addrconf_dad_start(ifp);
3199                 in6_ifa_put(ifp);
3200         }
3201 }
3202 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3203
3204 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3205 {
3206         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3207                 return true;
3208
3209         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3210             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3211                 return true;
3212
3213         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3214             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3215                 return true;
3216
3217         return false;
3218 }
3219
3220 static int ipv6_generate_stable_address(struct in6_addr *address,
3221                                         u8 dad_count,
3222                                         const struct inet6_dev *idev)
3223 {
3224         static DEFINE_SPINLOCK(lock);
3225         static __u32 digest[SHA_DIGEST_WORDS];
3226         static __u32 workspace[SHA_WORKSPACE_WORDS];
3227
3228         static union {
3229                 char __data[SHA_MESSAGE_BYTES];
3230                 struct {
3231                         struct in6_addr secret;
3232                         __be32 prefix[2];
3233                         unsigned char hwaddr[MAX_ADDR_LEN];
3234                         u8 dad_count;
3235                 } __packed;
3236         } data;
3237
3238         struct in6_addr secret;
3239         struct in6_addr temp;
3240         struct net *net = dev_net(idev->dev);
3241
3242         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3243
3244         if (idev->cnf.stable_secret.initialized)
3245                 secret = idev->cnf.stable_secret.secret;
3246         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3247                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3248         else
3249                 return -1;
3250
3251 retry:
3252         spin_lock_bh(&lock);
3253
3254         sha_init(digest);
3255         memset(&data, 0, sizeof(data));
3256         memset(workspace, 0, sizeof(workspace));
3257         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3258         data.prefix[0] = address->s6_addr32[0];
3259         data.prefix[1] = address->s6_addr32[1];
3260         data.secret = secret;
3261         data.dad_count = dad_count;
3262
3263         sha_transform(digest, data.__data, workspace);
3264
3265         temp = *address;
3266         temp.s6_addr32[2] = (__force __be32)digest[0];
3267         temp.s6_addr32[3] = (__force __be32)digest[1];
3268
3269         spin_unlock_bh(&lock);
3270
3271         if (ipv6_reserved_interfaceid(temp)) {
3272                 dad_count++;
3273                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3274                         return -1;
3275                 goto retry;
3276         }
3277
3278         *address = temp;
3279         return 0;
3280 }
3281
3282 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3283 {
3284         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3285
3286         if (s->initialized)
3287                 return;
3288         s = &idev->cnf.stable_secret;
3289         get_random_bytes(&s->secret, sizeof(s->secret));
3290         s->initialized = true;
3291 }
3292
3293 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3294 {
3295         struct in6_addr addr;
3296
3297         /* no link local addresses on L3 master devices */
3298         if (netif_is_l3_master(idev->dev))
3299                 return;
3300
3301         /* no link local addresses on devices flagged as slaves */
3302         if (idev->dev->flags & IFF_SLAVE)
3303                 return;
3304
3305         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3306
3307         switch (idev->cnf.addr_gen_mode) {
3308         case IN6_ADDR_GEN_MODE_RANDOM:
3309                 ipv6_gen_mode_random_init(idev);
3310                 fallthrough;
3311         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3312                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3313                         addrconf_add_linklocal(idev, &addr,
3314                                                IFA_F_STABLE_PRIVACY);
3315                 else if (prefix_route)
3316                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3317                                               0, 0, GFP_KERNEL);
3318                 break;
3319         case IN6_ADDR_GEN_MODE_EUI64:
3320                 /* addrconf_add_linklocal also adds a prefix_route and we
3321                  * only need to care about prefix routes if ipv6_generate_eui64
3322                  * couldn't generate one.
3323                  */
3324                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3325                         addrconf_add_linklocal(idev, &addr, 0);
3326                 else if (prefix_route)
3327                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3328                                               0, 0, GFP_KERNEL);
3329                 break;
3330         case IN6_ADDR_GEN_MODE_NONE:
3331         default:
3332                 /* will not add any link local address */
3333                 break;
3334         }
3335 }
3336
3337 static void addrconf_dev_config(struct net_device *dev)
3338 {
3339         struct inet6_dev *idev;
3340
3341         ASSERT_RTNL();
3342
3343         if ((dev->type != ARPHRD_ETHER) &&
3344             (dev->type != ARPHRD_FDDI) &&
3345             (dev->type != ARPHRD_ARCNET) &&
3346             (dev->type != ARPHRD_INFINIBAND) &&
3347             (dev->type != ARPHRD_IEEE1394) &&
3348             (dev->type != ARPHRD_TUNNEL6) &&
3349             (dev->type != ARPHRD_6LOWPAN) &&
3350             (dev->type != ARPHRD_IP6GRE) &&
3351             (dev->type != ARPHRD_IPGRE) &&
3352             (dev->type != ARPHRD_TUNNEL) &&
3353             (dev->type != ARPHRD_NONE) &&
3354             (dev->type != ARPHRD_RAWIP)) {
3355                 /* Alas, we support only Ethernet autoconfiguration. */
3356                 idev = __in6_dev_get(dev);
3357                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3358                     dev->flags & IFF_MULTICAST)
3359                         ipv6_mc_up(idev);
3360                 return;
3361         }
3362
3363         idev = addrconf_add_dev(dev);
3364         if (IS_ERR(idev))
3365                 return;
3366
3367         /* this device type has no EUI support */
3368         if (dev->type == ARPHRD_NONE &&
3369             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3370                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3371
3372         addrconf_addr_gen(idev, false);
3373 }
3374
3375 #if IS_ENABLED(CONFIG_IPV6_SIT)
3376 static void addrconf_sit_config(struct net_device *dev)
3377 {
3378         struct inet6_dev *idev;
3379
3380         ASSERT_RTNL();
3381
3382         /*
3383          * Configure the tunnel with one of our IPv4
3384          * addresses... we should configure all of
3385          * our v4 addrs in the tunnel
3386          */
3387
3388         idev = ipv6_find_idev(dev);
3389         if (IS_ERR(idev)) {
3390                 pr_debug("%s: add_dev failed\n", __func__);
3391                 return;
3392         }
3393
3394         if (dev->priv_flags & IFF_ISATAP) {
3395                 addrconf_addr_gen(idev, false);
3396                 return;
3397         }
3398
3399         sit_add_v4_addrs(idev);
3400
3401         if (dev->flags&IFF_POINTOPOINT)
3402                 addrconf_add_mroute(dev);
3403 }
3404 #endif
3405
3406 #if IS_ENABLED(CONFIG_NET_IPGRE)
3407 static void addrconf_gre_config(struct net_device *dev)
3408 {
3409         struct inet6_dev *idev;
3410
3411         ASSERT_RTNL();
3412
3413         idev = ipv6_find_idev(dev);
3414         if (IS_ERR(idev)) {
3415                 pr_debug("%s: add_dev failed\n", __func__);
3416                 return;
3417         }
3418
3419         addrconf_addr_gen(idev, true);
3420         if (dev->flags & IFF_POINTOPOINT)
3421                 addrconf_add_mroute(dev);
3422 }
3423 #endif
3424
3425 static int fixup_permanent_addr(struct net *net,
3426                                 struct inet6_dev *idev,
3427                                 struct inet6_ifaddr *ifp)
3428 {
3429         /* !fib6_node means the host route was removed from the
3430          * FIB, for example, if 'lo' device is taken down. In that
3431          * case regenerate the host route.
3432          */
3433         if (!ifp->rt || !ifp->rt->fib6_node) {
3434                 struct fib6_info *f6i, *prev;
3435
3436                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3437                                          GFP_ATOMIC);
3438                 if (IS_ERR(f6i))
3439                         return PTR_ERR(f6i);
3440
3441                 /* ifp->rt can be accessed outside of rtnl */
3442                 spin_lock(&ifp->lock);
3443                 prev = ifp->rt;
3444                 ifp->rt = f6i;
3445                 spin_unlock(&ifp->lock);
3446
3447                 fib6_info_release(prev);
3448         }
3449
3450         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3451                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3452                                       ifp->rt_priority, idev->dev, 0, 0,
3453                                       GFP_ATOMIC);
3454         }
3455
3456         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3457                 addrconf_dad_start(ifp);
3458
3459         return 0;
3460 }
3461
3462 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3463 {
3464         struct inet6_ifaddr *ifp, *tmp;
3465         struct inet6_dev *idev;
3466
3467         idev = __in6_dev_get(dev);
3468         if (!idev)
3469                 return;
3470
3471         write_lock_bh(&idev->lock);
3472
3473         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3474                 if ((ifp->flags & IFA_F_PERMANENT) &&
3475                     fixup_permanent_addr(net, idev, ifp) < 0) {
3476                         write_unlock_bh(&idev->lock);
3477                         in6_ifa_hold(ifp);
3478                         ipv6_del_addr(ifp);
3479                         write_lock_bh(&idev->lock);
3480
3481                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3482                                              idev->dev->name, &ifp->addr);
3483                 }
3484         }
3485
3486         write_unlock_bh(&idev->lock);
3487 }
3488
3489 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3490                            void *ptr)
3491 {
3492         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3493         struct netdev_notifier_change_info *change_info;
3494         struct netdev_notifier_changeupper_info *info;
3495         struct inet6_dev *idev = __in6_dev_get(dev);
3496         struct net *net = dev_net(dev);
3497         int run_pending = 0;
3498         int err;
3499
3500         switch (event) {
3501         case NETDEV_REGISTER:
3502                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3503                         idev = ipv6_add_dev(dev);
3504                         if (IS_ERR(idev))
3505                                 return notifier_from_errno(PTR_ERR(idev));
3506                 }
3507                 break;
3508
3509         case NETDEV_CHANGEMTU:
3510                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3511                 if (dev->mtu < IPV6_MIN_MTU) {
3512                         addrconf_ifdown(dev, dev != net->loopback_dev);
3513                         break;
3514                 }
3515
3516                 if (idev) {
3517                         rt6_mtu_change(dev, dev->mtu);
3518                         idev->cnf.mtu6 = dev->mtu;
3519                         break;
3520                 }
3521
3522                 /* allocate new idev */
3523                 idev = ipv6_add_dev(dev);
3524                 if (IS_ERR(idev))
3525                         break;
3526
3527                 /* device is still not ready */
3528                 if (!(idev->if_flags & IF_READY))
3529                         break;
3530
3531                 run_pending = 1;
3532                 fallthrough;
3533         case NETDEV_UP:
3534         case NETDEV_CHANGE:
3535                 if (dev->flags & IFF_SLAVE)
3536                         break;
3537
3538                 if (idev && idev->cnf.disable_ipv6)
3539                         break;
3540
3541                 if (event == NETDEV_UP) {
3542                         /* restore routes for permanent addresses */
3543                         addrconf_permanent_addr(net, dev);
3544
3545                         if (!addrconf_link_ready(dev)) {
3546                                 /* device is not ready yet. */
3547                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3548                                          dev->name);
3549                                 break;
3550                         }
3551
3552                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3553                                 idev = ipv6_add_dev(dev);
3554
3555                         if (!IS_ERR_OR_NULL(idev)) {
3556                                 idev->if_flags |= IF_READY;
3557                                 run_pending = 1;
3558                         }
3559                 } else if (event == NETDEV_CHANGE) {
3560                         if (!addrconf_link_ready(dev)) {
3561                                 /* device is still not ready. */
3562                                 rt6_sync_down_dev(dev, event);
3563                                 break;
3564                         }
3565
3566                         if (!IS_ERR_OR_NULL(idev)) {
3567                                 if (idev->if_flags & IF_READY) {
3568                                         /* device is already configured -
3569                                          * but resend MLD reports, we might
3570                                          * have roamed and need to update
3571                                          * multicast snooping switches
3572                                          */
3573                                         ipv6_mc_up(idev);
3574                                         change_info = ptr;
3575                                         if (change_info->flags_changed & IFF_NOARP)
3576                                                 addrconf_dad_run(idev, true);
3577                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3578                                         break;
3579                                 }
3580                                 idev->if_flags |= IF_READY;
3581                         }
3582
3583                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3584                                 dev->name);
3585
3586                         run_pending = 1;
3587                 }
3588
3589                 switch (dev->type) {
3590 #if IS_ENABLED(CONFIG_IPV6_SIT)
3591                 case ARPHRD_SIT:
3592                         addrconf_sit_config(dev);
3593                         break;
3594 #endif
3595 #if IS_ENABLED(CONFIG_NET_IPGRE)
3596                 case ARPHRD_IPGRE:
3597                         addrconf_gre_config(dev);
3598                         break;
3599 #endif
3600                 case ARPHRD_LOOPBACK:
3601                         init_loopback(dev);
3602                         break;
3603
3604                 default:
3605                         addrconf_dev_config(dev);
3606                         break;
3607                 }
3608
3609                 if (!IS_ERR_OR_NULL(idev)) {
3610                         if (run_pending)
3611                                 addrconf_dad_run(idev, false);
3612
3613                         /* Device has an address by now */
3614                         rt6_sync_up(dev, RTNH_F_DEAD);
3615
3616                         /*
3617                          * If the MTU changed during the interface down,
3618                          * when the interface up, the changed MTU must be
3619                          * reflected in the idev as well as routers.
3620                          */
3621                         if (idev->cnf.mtu6 != dev->mtu &&
3622                             dev->mtu >= IPV6_MIN_MTU) {
3623                                 rt6_mtu_change(dev, dev->mtu);
3624                                 idev->cnf.mtu6 = dev->mtu;
3625                         }
3626                         idev->tstamp = jiffies;
3627                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3628
3629                         /*
3630                          * If the changed mtu during down is lower than
3631                          * IPV6_MIN_MTU stop IPv6 on this interface.
3632                          */
3633                         if (dev->mtu < IPV6_MIN_MTU)
3634                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3635                 }
3636                 break;
3637
3638         case NETDEV_DOWN:
3639         case NETDEV_UNREGISTER:
3640                 /*
3641                  *      Remove all addresses from this interface.
3642                  */
3643                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3644                 break;
3645
3646         case NETDEV_CHANGENAME:
3647                 if (idev) {
3648                         snmp6_unregister_dev(idev);
3649                         addrconf_sysctl_unregister(idev);
3650                         err = addrconf_sysctl_register(idev);
3651                         if (err)
3652                                 return notifier_from_errno(err);
3653                         err = snmp6_register_dev(idev);
3654                         if (err) {
3655                                 addrconf_sysctl_unregister(idev);
3656                                 return notifier_from_errno(err);
3657                         }
3658                 }
3659                 break;
3660
3661         case NETDEV_PRE_TYPE_CHANGE:
3662         case NETDEV_POST_TYPE_CHANGE:
3663                 if (idev)
3664                         addrconf_type_change(dev, event);
3665                 break;
3666
3667         case NETDEV_CHANGEUPPER:
3668                 info = ptr;
3669
3670                 /* flush all routes if dev is linked to or unlinked from
3671                  * an L3 master device (e.g., VRF)
3672                  */
3673                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3674                         addrconf_ifdown(dev, 0);
3675         }
3676
3677         return NOTIFY_OK;
3678 }
3679
3680 /*
3681  *      addrconf module should be notified of a device going up
3682  */
3683 static struct notifier_block ipv6_dev_notf = {
3684         .notifier_call = addrconf_notify,
3685         .priority = ADDRCONF_NOTIFY_PRIORITY,
3686 };
3687
3688 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3689 {
3690         struct inet6_dev *idev;
3691         ASSERT_RTNL();
3692
3693         idev = __in6_dev_get(dev);
3694
3695         if (event == NETDEV_POST_TYPE_CHANGE)
3696                 ipv6_mc_remap(idev);
3697         else if (event == NETDEV_PRE_TYPE_CHANGE)
3698                 ipv6_mc_unmap(idev);
3699 }
3700
3701 static bool addr_is_local(const struct in6_addr *addr)
3702 {
3703         return ipv6_addr_type(addr) &
3704                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3705 }
3706
3707 static int addrconf_ifdown(struct net_device *dev, int how)
3708 {
3709         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3710         struct net *net = dev_net(dev);
3711         struct inet6_dev *idev;
3712         struct inet6_ifaddr *ifa, *tmp;
3713         bool keep_addr = false;
3714         int state, i;
3715
3716         ASSERT_RTNL();
3717
3718         rt6_disable_ip(dev, event);
3719
3720         idev = __in6_dev_get(dev);
3721         if (!idev)
3722                 return -ENODEV;
3723
3724         /*
3725          * Step 1: remove reference to ipv6 device from parent device.
3726          *         Do not dev_put!
3727          */
3728         if (how) {
3729                 idev->dead = 1;
3730
3731                 /* protected by rtnl_lock */
3732                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3733
3734                 /* Step 1.5: remove snmp6 entry */
3735                 snmp6_unregister_dev(idev);
3736
3737         }
3738
3739         /* combine the user config with event to determine if permanent
3740          * addresses are to be removed from address hash table
3741          */
3742         if (!how && !idev->cnf.disable_ipv6) {
3743                 /* aggregate the system setting and interface setting */
3744                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3745
3746                 if (!_keep_addr)
3747                         _keep_addr = idev->cnf.keep_addr_on_down;
3748
3749                 keep_addr = (_keep_addr > 0);
3750         }
3751
3752         /* Step 2: clear hash table */
3753         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3754                 struct hlist_head *h = &inet6_addr_lst[i];
3755
3756                 spin_lock_bh(&addrconf_hash_lock);
3757 restart:
3758                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3759                         if (ifa->idev == idev) {
3760                                 addrconf_del_dad_work(ifa);
3761                                 /* combined flag + permanent flag decide if
3762                                  * address is retained on a down event
3763                                  */
3764                                 if (!keep_addr ||
3765                                     !(ifa->flags & IFA_F_PERMANENT) ||
3766                                     addr_is_local(&ifa->addr)) {
3767                                         hlist_del_init_rcu(&ifa->addr_lst);
3768                                         goto restart;
3769                                 }
3770                         }
3771                 }
3772                 spin_unlock_bh(&addrconf_hash_lock);
3773         }
3774
3775         write_lock_bh(&idev->lock);
3776
3777         addrconf_del_rs_timer(idev);
3778
3779         /* Step 2: clear flags for stateless addrconf */
3780         if (!how)
3781                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3782
3783         /* Step 3: clear tempaddr list */
3784         while (!list_empty(&idev->tempaddr_list)) {
3785                 ifa = list_first_entry(&idev->tempaddr_list,
3786                                        struct inet6_ifaddr, tmp_list);
3787                 list_del(&ifa->tmp_list);
3788                 write_unlock_bh(&idev->lock);
3789                 spin_lock_bh(&ifa->lock);
3790
3791                 if (ifa->ifpub) {
3792                         in6_ifa_put(ifa->ifpub);
3793                         ifa->ifpub = NULL;
3794                 }
3795                 spin_unlock_bh(&ifa->lock);
3796                 in6_ifa_put(ifa);
3797                 write_lock_bh(&idev->lock);
3798         }
3799
3800         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3801                 struct fib6_info *rt = NULL;
3802                 bool keep;
3803
3804                 addrconf_del_dad_work(ifa);
3805
3806                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3807                         !addr_is_local(&ifa->addr);
3808
3809                 write_unlock_bh(&idev->lock);
3810                 spin_lock_bh(&ifa->lock);
3811
3812                 if (keep) {
3813                         /* set state to skip the notifier below */
3814                         state = INET6_IFADDR_STATE_DEAD;
3815                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3816                         if (!(ifa->flags & IFA_F_NODAD))
3817                                 ifa->flags |= IFA_F_TENTATIVE;
3818
3819                         rt = ifa->rt;
3820                         ifa->rt = NULL;
3821                 } else {
3822                         state = ifa->state;
3823                         ifa->state = INET6_IFADDR_STATE_DEAD;
3824                 }
3825
3826                 spin_unlock_bh(&ifa->lock);
3827
3828                 if (rt)
3829                         ip6_del_rt(net, rt);
3830
3831                 if (state != INET6_IFADDR_STATE_DEAD) {
3832                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3833                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3834                 } else {
3835                         if (idev->cnf.forwarding)
3836                                 addrconf_leave_anycast(ifa);
3837                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3838                 }
3839
3840                 write_lock_bh(&idev->lock);
3841                 if (!keep) {
3842                         list_del_rcu(&ifa->if_list);
3843                         in6_ifa_put(ifa);
3844                 }
3845         }
3846
3847         write_unlock_bh(&idev->lock);
3848
3849         /* Step 5: Discard anycast and multicast list */
3850         if (how) {
3851                 ipv6_ac_destroy_dev(idev);
3852                 ipv6_mc_destroy_dev(idev);
3853         } else {
3854                 ipv6_mc_down(idev);
3855         }
3856
3857         idev->tstamp = jiffies;
3858
3859         /* Last: Shot the device (if unregistered) */
3860         if (how) {
3861                 addrconf_sysctl_unregister(idev);
3862                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3863                 neigh_ifdown(&nd_tbl, dev);
3864                 in6_dev_put(idev);
3865         }
3866         return 0;
3867 }
3868
3869 static void addrconf_rs_timer(struct timer_list *t)
3870 {
3871         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3872         struct net_device *dev = idev->dev;
3873         struct in6_addr lladdr;
3874
3875         write_lock(&idev->lock);
3876         if (idev->dead || !(idev->if_flags & IF_READY))
3877                 goto out;
3878
3879         if (!ipv6_accept_ra(idev))
3880                 goto out;
3881
3882         /* Announcement received after solicitation was sent */
3883         if (idev->if_flags & IF_RA_RCVD)
3884                 goto out;
3885
3886         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3887                 write_unlock(&idev->lock);
3888                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3889                         ndisc_send_rs(dev, &lladdr,
3890                                       &in6addr_linklocal_allrouters);
3891                 else
3892                         goto put;
3893
3894                 write_lock(&idev->lock);
3895                 idev->rs_interval = rfc3315_s14_backoff_update(
3896                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3897                 /* The wait after the last probe can be shorter */
3898                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3899                                              idev->cnf.rtr_solicits) ?
3900                                       idev->cnf.rtr_solicit_delay :
3901                                       idev->rs_interval);
3902         } else {
3903                 /*
3904                  * Note: we do not support deprecated "all on-link"
3905                  * assumption any longer.
3906                  */
3907                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3908         }
3909
3910 out:
3911         write_unlock(&idev->lock);
3912 put:
3913         in6_dev_put(idev);
3914 }
3915
3916 /*
3917  *      Duplicate Address Detection
3918  */
3919 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3920 {
3921         unsigned long rand_num;
3922         struct inet6_dev *idev = ifp->idev;
3923         u64 nonce;
3924
3925         if (ifp->flags & IFA_F_OPTIMISTIC)
3926                 rand_num = 0;
3927         else
3928                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3929
3930         nonce = 0;
3931         if (idev->cnf.enhanced_dad ||
3932             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3933                 do
3934                         get_random_bytes(&nonce, 6);
3935                 while (nonce == 0);
3936         }
3937         ifp->dad_nonce = nonce;
3938         ifp->dad_probes = idev->cnf.dad_transmits;
3939         addrconf_mod_dad_work(ifp, rand_num);
3940 }
3941
3942 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3943 {
3944         struct inet6_dev *idev = ifp->idev;
3945         struct net_device *dev = idev->dev;
3946         bool bump_id, notify = false;
3947         struct net *net;
3948
3949         addrconf_join_solict(dev, &ifp->addr);
3950
3951         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3952
3953         read_lock_bh(&idev->lock);
3954         spin_lock(&ifp->lock);
3955         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3956                 goto out;
3957
3958         net = dev_net(dev);
3959         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3960             (net->ipv6.devconf_all->accept_dad < 1 &&
3961              idev->cnf.accept_dad < 1) ||
3962             !(ifp->flags&IFA_F_TENTATIVE) ||
3963             ifp->flags & IFA_F_NODAD) {
3964                 bool send_na = false;
3965
3966                 if (ifp->flags & IFA_F_TENTATIVE &&
3967                     !(ifp->flags & IFA_F_OPTIMISTIC))
3968                         send_na = true;
3969                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3970                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3971                 spin_unlock(&ifp->lock);
3972                 read_unlock_bh(&idev->lock);
3973
3974                 addrconf_dad_completed(ifp, bump_id, send_na);
3975                 return;
3976         }
3977
3978         if (!(idev->if_flags & IF_READY)) {
3979                 spin_unlock(&ifp->lock);
3980                 read_unlock_bh(&idev->lock);
3981                 /*
3982                  * If the device is not ready:
3983                  * - keep it tentative if it is a permanent address.
3984                  * - otherwise, kill it.
3985                  */
3986                 in6_ifa_hold(ifp);
3987                 addrconf_dad_stop(ifp, 0);
3988                 return;
3989         }
3990
3991         /*
3992          * Optimistic nodes can start receiving
3993          * Frames right away
3994          */
3995         if (ifp->flags & IFA_F_OPTIMISTIC) {
3996                 ip6_ins_rt(net, ifp->rt);
3997                 if (ipv6_use_optimistic_addr(net, idev)) {
3998                         /* Because optimistic nodes can use this address,
3999                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4000                          */
4001                         notify = true;
4002                 }
4003         }
4004
4005         addrconf_dad_kick(ifp);
4006 out:
4007         spin_unlock(&ifp->lock);
4008         read_unlock_bh(&idev->lock);
4009         if (notify)
4010                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4011 }
4012
4013 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4014 {
4015         bool begin_dad = false;
4016
4017         spin_lock_bh(&ifp->lock);
4018         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4019                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4020                 begin_dad = true;
4021         }
4022         spin_unlock_bh(&ifp->lock);
4023
4024         if (begin_dad)
4025                 addrconf_mod_dad_work(ifp, 0);
4026 }
4027
4028 static void addrconf_dad_work(struct work_struct *w)
4029 {
4030         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4031                                                 struct inet6_ifaddr,
4032                                                 dad_work);
4033         struct inet6_dev *idev = ifp->idev;
4034         bool bump_id, disable_ipv6 = false;
4035         struct in6_addr mcaddr;
4036
4037         enum {
4038                 DAD_PROCESS,
4039                 DAD_BEGIN,
4040                 DAD_ABORT,
4041         } action = DAD_PROCESS;
4042
4043         rtnl_lock();
4044
4045         spin_lock_bh(&ifp->lock);
4046         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4047                 action = DAD_BEGIN;
4048                 ifp->state = INET6_IFADDR_STATE_DAD;
4049         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4050                 action = DAD_ABORT;
4051                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4052
4053                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4054                      idev->cnf.accept_dad > 1) &&
4055                     !idev->cnf.disable_ipv6 &&
4056                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4057                         struct in6_addr addr;
4058
4059                         addr.s6_addr32[0] = htonl(0xfe800000);
4060                         addr.s6_addr32[1] = 0;
4061
4062                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4063                             ipv6_addr_equal(&ifp->addr, &addr)) {
4064                                 /* DAD failed for link-local based on MAC */
4065                                 idev->cnf.disable_ipv6 = 1;
4066
4067                                 pr_info("%s: IPv6 being disabled!\n",
4068                                         ifp->idev->dev->name);
4069                                 disable_ipv6 = true;
4070                         }
4071                 }
4072         }
4073         spin_unlock_bh(&ifp->lock);
4074
4075         if (action == DAD_BEGIN) {
4076                 addrconf_dad_begin(ifp);
4077                 goto out;
4078         } else if (action == DAD_ABORT) {
4079                 in6_ifa_hold(ifp);
4080                 addrconf_dad_stop(ifp, 1);
4081                 if (disable_ipv6)
4082                         addrconf_ifdown(idev->dev, 0);
4083                 goto out;
4084         }
4085
4086         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4087                 goto out;
4088
4089         write_lock_bh(&idev->lock);
4090         if (idev->dead || !(idev->if_flags & IF_READY)) {
4091                 write_unlock_bh(&idev->lock);
4092                 goto out;
4093         }
4094
4095         spin_lock(&ifp->lock);
4096         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4097                 spin_unlock(&ifp->lock);
4098                 write_unlock_bh(&idev->lock);
4099                 goto out;
4100         }
4101
4102         if (ifp->dad_probes == 0) {
4103                 bool send_na = false;
4104
4105                 /*
4106                  * DAD was successful
4107                  */
4108
4109                 if (ifp->flags & IFA_F_TENTATIVE &&
4110                     !(ifp->flags & IFA_F_OPTIMISTIC))
4111                         send_na = true;
4112                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4113                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4114                 spin_unlock(&ifp->lock);
4115                 write_unlock_bh(&idev->lock);
4116
4117                 addrconf_dad_completed(ifp, bump_id, send_na);
4118
4119                 goto out;
4120         }
4121
4122         ifp->dad_probes--;
4123         addrconf_mod_dad_work(ifp,
4124                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4125                                   HZ/100));
4126         spin_unlock(&ifp->lock);
4127         write_unlock_bh(&idev->lock);
4128
4129         /* send a neighbour solicitation for our addr */
4130         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4131         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4132                       ifp->dad_nonce);
4133 out:
4134         in6_ifa_put(ifp);
4135         rtnl_unlock();
4136 }
4137
4138 /* ifp->idev must be at least read locked */
4139 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4140 {
4141         struct inet6_ifaddr *ifpiter;
4142         struct inet6_dev *idev = ifp->idev;
4143
4144         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4145                 if (ifpiter->scope > IFA_LINK)
4146                         break;
4147                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4148                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4149                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4150                     IFA_F_PERMANENT)
4151                         return false;
4152         }
4153         return true;
4154 }
4155
4156 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4157                                    bool send_na)
4158 {
4159         struct net_device *dev = ifp->idev->dev;
4160         struct in6_addr lladdr;
4161         bool send_rs, send_mld;
4162
4163         addrconf_del_dad_work(ifp);
4164
4165         /*
4166          *      Configure the address for reception. Now it is valid.
4167          */
4168
4169         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4170
4171         /* If added prefix is link local and we are prepared to process
4172            router advertisements, start sending router solicitations.
4173          */
4174
4175         read_lock_bh(&ifp->idev->lock);
4176         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4177         send_rs = send_mld &&
4178                   ipv6_accept_ra(ifp->idev) &&
4179                   ifp->idev->cnf.rtr_solicits != 0 &&
4180                   (dev->flags&IFF_LOOPBACK) == 0;
4181         read_unlock_bh(&ifp->idev->lock);
4182
4183         /* While dad is in progress mld report's source address is in6_addrany.
4184          * Resend with proper ll now.
4185          */
4186         if (send_mld)
4187                 ipv6_mc_dad_complete(ifp->idev);
4188
4189         /* send unsolicited NA if enabled */
4190         if (send_na &&
4191             (ifp->idev->cnf.ndisc_notify ||
4192              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4193                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4194                               /*router=*/ !!ifp->idev->cnf.forwarding,
4195                               /*solicited=*/ false, /*override=*/ true,
4196                               /*inc_opt=*/ true);
4197         }
4198
4199         if (send_rs) {
4200                 /*
4201                  *      If a host as already performed a random delay
4202                  *      [...] as part of DAD [...] there is no need
4203                  *      to delay again before sending the first RS
4204                  */
4205                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4206                         return;
4207                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4208
4209                 write_lock_bh(&ifp->idev->lock);
4210                 spin_lock(&ifp->lock);
4211                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4212                         ifp->idev->cnf.rtr_solicit_interval);
4213                 ifp->idev->rs_probes = 1;
4214                 ifp->idev->if_flags |= IF_RS_SENT;
4215                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4216                 spin_unlock(&ifp->lock);
4217                 write_unlock_bh(&ifp->idev->lock);
4218         }
4219
4220         if (bump_id)
4221                 rt_genid_bump_ipv6(dev_net(dev));
4222
4223         /* Make sure that a new temporary address will be created
4224          * before this temporary address becomes deprecated.
4225          */
4226         if (ifp->flags & IFA_F_TEMPORARY)
4227                 addrconf_verify_rtnl();
4228 }
4229
4230 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4231 {
4232         struct inet6_ifaddr *ifp;
4233
4234         read_lock_bh(&idev->lock);
4235         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4236                 spin_lock(&ifp->lock);
4237                 if ((ifp->flags & IFA_F_TENTATIVE &&
4238                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4239                         if (restart)
4240                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4241                         addrconf_dad_kick(ifp);
4242                 }
4243                 spin_unlock(&ifp->lock);
4244         }
4245         read_unlock_bh(&idev->lock);
4246 }
4247
4248 #ifdef CONFIG_PROC_FS
4249 struct if6_iter_state {
4250         struct seq_net_private p;
4251         int bucket;
4252         int offset;
4253 };
4254
4255 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4256 {
4257         struct if6_iter_state *state = seq->private;
4258         struct net *net = seq_file_net(seq);
4259         struct inet6_ifaddr *ifa = NULL;
4260         int p = 0;
4261
4262         /* initial bucket if pos is 0 */
4263         if (pos == 0) {
4264                 state->bucket = 0;
4265                 state->offset = 0;
4266         }
4267
4268         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4269                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4270                                          addr_lst) {
4271                         if (!net_eq(dev_net(ifa->idev->dev), net))
4272                                 continue;
4273                         /* sync with offset */
4274                         if (p < state->offset) {
4275                                 p++;
4276                                 continue;
4277                         }
4278                         return ifa;
4279                 }
4280
4281                 /* prepare for next bucket */
4282                 state->offset = 0;
4283                 p = 0;
4284         }
4285         return NULL;
4286 }
4287
4288 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4289                                          struct inet6_ifaddr *ifa)
4290 {
4291         struct if6_iter_state *state = seq->private;
4292         struct net *net = seq_file_net(seq);
4293
4294         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4295                 if (!net_eq(dev_net(ifa->idev->dev), net))
4296                         continue;
4297                 state->offset++;
4298                 return ifa;
4299         }
4300
4301         state->offset = 0;
4302         while (++state->bucket < IN6_ADDR_HSIZE) {
4303                 hlist_for_each_entry_rcu(ifa,
4304                                      &inet6_addr_lst[state->bucket], addr_lst) {
4305                         if (!net_eq(dev_net(ifa->idev->dev), net))
4306                                 continue;
4307                         return ifa;
4308                 }
4309         }
4310
4311         return NULL;
4312 }
4313
4314 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4315         __acquires(rcu)
4316 {
4317         rcu_read_lock();
4318         return if6_get_first(seq, *pos);
4319 }
4320
4321 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4322 {
4323         struct inet6_ifaddr *ifa;
4324
4325         ifa = if6_get_next(seq, v);
4326         ++*pos;
4327         return ifa;
4328 }
4329
4330 static void if6_seq_stop(struct seq_file *seq, void *v)
4331         __releases(rcu)
4332 {
4333         rcu_read_unlock();
4334 }
4335
4336 static int if6_seq_show(struct seq_file *seq, void *v)
4337 {
4338         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4339         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4340                    &ifp->addr,
4341                    ifp->idev->dev->ifindex,
4342                    ifp->prefix_len,
4343                    ifp->scope,
4344                    (u8) ifp->flags,
4345                    ifp->idev->dev->name);
4346         return 0;
4347 }
4348
4349 static const struct seq_operations if6_seq_ops = {
4350         .start  = if6_seq_start,
4351         .next   = if6_seq_next,
4352         .show   = if6_seq_show,
4353         .stop   = if6_seq_stop,
4354 };
4355
4356 static int __net_init if6_proc_net_init(struct net *net)
4357 {
4358         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4359                         sizeof(struct if6_iter_state)))
4360                 return -ENOMEM;
4361         return 0;
4362 }
4363
4364 static void __net_exit if6_proc_net_exit(struct net *net)
4365 {
4366         remove_proc_entry("if_inet6", net->proc_net);
4367 }
4368
4369 static struct pernet_operations if6_proc_net_ops = {
4370         .init = if6_proc_net_init,
4371         .exit = if6_proc_net_exit,
4372 };
4373
4374 int __init if6_proc_init(void)
4375 {
4376         return register_pernet_subsys(&if6_proc_net_ops);
4377 }
4378
4379 void if6_proc_exit(void)
4380 {
4381         unregister_pernet_subsys(&if6_proc_net_ops);
4382 }
4383 #endif  /* CONFIG_PROC_FS */
4384
4385 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4386 /* Check if address is a home address configured on any interface. */
4387 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4388 {
4389         unsigned int hash = inet6_addr_hash(net, addr);
4390         struct inet6_ifaddr *ifp = NULL;
4391         int ret = 0;
4392
4393         rcu_read_lock();
4394         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4395                 if (!net_eq(dev_net(ifp->idev->dev), net))
4396                         continue;
4397                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4398                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4399                         ret = 1;
4400                         break;
4401                 }
4402         }
4403         rcu_read_unlock();
4404         return ret;
4405 }
4406 #endif
4407
4408 /* RFC6554 has some algorithm to avoid loops in segment routing by
4409  * checking if the segments contains any of a local interface address.
4410  *
4411  * Quote:
4412  *
4413  * To detect loops in the SRH, a router MUST determine if the SRH
4414  * includes multiple addresses assigned to any interface on that router.
4415  * If such addresses appear more than once and are separated by at least
4416  * one address not assigned to that router.
4417  */
4418 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4419                           unsigned char nsegs)
4420 {
4421         const struct in6_addr *addr;
4422         int i, ret = 0, found = 0;
4423         struct inet6_ifaddr *ifp;
4424         bool separated = false;
4425         unsigned int hash;
4426         bool hash_found;
4427
4428         rcu_read_lock();
4429         for (i = 0; i < nsegs; i++) {
4430                 addr = &segs[i];
4431                 hash = inet6_addr_hash(net, addr);
4432
4433                 hash_found = false;
4434                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4435                         if (!net_eq(dev_net(ifp->idev->dev), net))
4436                                 continue;
4437
4438                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4439                                 hash_found = true;
4440                                 break;
4441                         }
4442                 }
4443
4444                 if (hash_found) {
4445                         if (found > 1 && separated) {
4446                                 ret = 1;
4447                                 break;
4448                         }
4449
4450                         separated = false;
4451                         found++;
4452                 } else {
4453                         separated = true;
4454                 }
4455         }
4456         rcu_read_unlock();
4457
4458         return ret;
4459 }
4460
4461 /*
4462  *      Periodic address status verification
4463  */
4464
4465 static void addrconf_verify_rtnl(void)
4466 {
4467         unsigned long now, next, next_sec, next_sched;
4468         struct inet6_ifaddr *ifp;
4469         int i;
4470
4471         ASSERT_RTNL();
4472
4473         rcu_read_lock_bh();
4474         now = jiffies;
4475         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4476
4477         cancel_delayed_work(&addr_chk_work);
4478
4479         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4480 restart:
4481                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4482                         unsigned long age;
4483
4484                         /* When setting preferred_lft to a value not zero or
4485                          * infinity, while valid_lft is infinity
4486                          * IFA_F_PERMANENT has a non-infinity life time.
4487                          */
4488                         if ((ifp->flags & IFA_F_PERMANENT) &&
4489                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4490                                 continue;
4491
4492                         spin_lock(&ifp->lock);
4493                         /* We try to batch several events at once. */
4494                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4495
4496                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4497                             age >= ifp->valid_lft) {
4498                                 spin_unlock(&ifp->lock);
4499                                 in6_ifa_hold(ifp);
4500                                 ipv6_del_addr(ifp);
4501                                 goto restart;
4502                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4503                                 spin_unlock(&ifp->lock);
4504                                 continue;
4505                         } else if (age >= ifp->prefered_lft) {
4506                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4507                                 int deprecate = 0;
4508
4509                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4510                                         deprecate = 1;
4511                                         ifp->flags |= IFA_F_DEPRECATED;
4512                                 }
4513
4514                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4515                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4516                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4517
4518                                 spin_unlock(&ifp->lock);
4519
4520                                 if (deprecate) {
4521                                         in6_ifa_hold(ifp);
4522
4523                                         ipv6_ifa_notify(0, ifp);
4524                                         in6_ifa_put(ifp);
4525                                         goto restart;
4526                                 }
4527                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4528                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4529                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4530                                         ifp->idev->cnf.dad_transmits *
4531                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4532
4533                                 if (age >= ifp->prefered_lft - regen_advance) {
4534                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4535                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4536                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4537                                         if (!ifp->regen_count && ifpub) {
4538                                                 ifp->regen_count++;
4539                                                 in6_ifa_hold(ifp);
4540                                                 in6_ifa_hold(ifpub);
4541                                                 spin_unlock(&ifp->lock);
4542
4543                                                 spin_lock(&ifpub->lock);
4544                                                 ifpub->regen_count = 0;
4545                                                 spin_unlock(&ifpub->lock);
4546                                                 rcu_read_unlock_bh();
4547                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4548                                                 in6_ifa_put(ifpub);
4549                                                 in6_ifa_put(ifp);
4550                                                 rcu_read_lock_bh();
4551                                                 goto restart;
4552                                         }
4553                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4554                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4555                                 spin_unlock(&ifp->lock);
4556                         } else {
4557                                 /* ifp->prefered_lft <= ifp->valid_lft */
4558                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4559                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4560                                 spin_unlock(&ifp->lock);
4561                         }
4562                 }
4563         }
4564
4565         next_sec = round_jiffies_up(next);
4566         next_sched = next;
4567
4568         /* If rounded timeout is accurate enough, accept it. */
4569         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4570                 next_sched = next_sec;
4571
4572         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4573         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4574                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4575
4576         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4577                  now, next, next_sec, next_sched);
4578         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4579         rcu_read_unlock_bh();
4580 }
4581
4582 static void addrconf_verify_work(struct work_struct *w)
4583 {
4584         rtnl_lock();
4585         addrconf_verify_rtnl();
4586         rtnl_unlock();
4587 }
4588
4589 static void addrconf_verify(void)
4590 {
4591         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4592 }
4593
4594 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4595                                      struct in6_addr **peer_pfx)
4596 {
4597         struct in6_addr *pfx = NULL;
4598
4599         *peer_pfx = NULL;
4600
4601         if (addr)
4602                 pfx = nla_data(addr);
4603
4604         if (local) {
4605                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4606                         *peer_pfx = pfx;
4607                 pfx = nla_data(local);
4608         }
4609
4610         return pfx;
4611 }
4612
4613 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4614         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4615         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4616         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4617         [IFA_FLAGS]             = { .len = sizeof(u32) },
4618         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4619         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4620 };
4621
4622 static int
4623 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4624                   struct netlink_ext_ack *extack)
4625 {
4626         struct net *net = sock_net(skb->sk);
4627         struct ifaddrmsg *ifm;
4628         struct nlattr *tb[IFA_MAX+1];
4629         struct in6_addr *pfx, *peer_pfx;
4630         u32 ifa_flags;
4631         int err;
4632
4633         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4634                                      ifa_ipv6_policy, extack);
4635         if (err < 0)
4636                 return err;
4637
4638         ifm = nlmsg_data(nlh);
4639         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4640         if (!pfx)
4641                 return -EINVAL;
4642
4643         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4644
4645         /* We ignore other flags so far. */
4646         ifa_flags &= IFA_F_MANAGETEMPADDR;
4647
4648         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4649                               ifm->ifa_prefixlen);
4650 }
4651
4652 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4653                                unsigned long expires, u32 flags,
4654                                bool modify_peer)
4655 {
4656         struct fib6_info *f6i;
4657         u32 prio;
4658
4659         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4660                                         ifp->prefix_len,
4661                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4662         if (!f6i)
4663                 return -ENOENT;
4664
4665         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4666         if (f6i->fib6_metric != prio) {
4667                 /* delete old one */
4668                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4669
4670                 /* add new one */
4671                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4672                                       ifp->prefix_len,
4673                                       ifp->rt_priority, ifp->idev->dev,
4674                                       expires, flags, GFP_KERNEL);
4675         } else {
4676                 if (!expires)
4677                         fib6_clean_expires(f6i);
4678                 else
4679                         fib6_set_expires(f6i, expires);
4680
4681                 fib6_info_release(f6i);
4682         }
4683
4684         return 0;
4685 }
4686
4687 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4688 {
4689         u32 flags;
4690         clock_t expires;
4691         unsigned long timeout;
4692         bool was_managetempaddr;
4693         bool had_prefixroute;
4694         bool new_peer = false;
4695
4696         ASSERT_RTNL();
4697
4698         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4699                 return -EINVAL;
4700
4701         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4702             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4703                 return -EINVAL;
4704
4705         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4706                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4707
4708         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4709         if (addrconf_finite_timeout(timeout)) {
4710                 expires = jiffies_to_clock_t(timeout * HZ);
4711                 cfg->valid_lft = timeout;
4712                 flags = RTF_EXPIRES;
4713         } else {
4714                 expires = 0;
4715                 flags = 0;
4716                 cfg->ifa_flags |= IFA_F_PERMANENT;
4717         }
4718
4719         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4720         if (addrconf_finite_timeout(timeout)) {
4721                 if (timeout == 0)
4722                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4723                 cfg->preferred_lft = timeout;
4724         }
4725
4726         if (cfg->peer_pfx &&
4727             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4728                 if (!ipv6_addr_any(&ifp->peer_addr))
4729                         cleanup_prefix_route(ifp, expires, true, true);
4730                 new_peer = true;
4731         }
4732
4733         spin_lock_bh(&ifp->lock);
4734         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4735         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4736                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4737         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4738                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4739                         IFA_F_NOPREFIXROUTE);
4740         ifp->flags |= cfg->ifa_flags;
4741         ifp->tstamp = jiffies;
4742         ifp->valid_lft = cfg->valid_lft;
4743         ifp->prefered_lft = cfg->preferred_lft;
4744
4745         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4746                 ifp->rt_priority = cfg->rt_priority;
4747
4748         if (new_peer)
4749                 ifp->peer_addr = *cfg->peer_pfx;
4750
4751         spin_unlock_bh(&ifp->lock);
4752         if (!(ifp->flags&IFA_F_TENTATIVE))
4753                 ipv6_ifa_notify(0, ifp);
4754
4755         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4756                 int rc = -ENOENT;
4757
4758                 if (had_prefixroute)
4759                         rc = modify_prefix_route(ifp, expires, flags, false);
4760
4761                 /* prefix route could have been deleted; if so restore it */
4762                 if (rc == -ENOENT) {
4763                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4764                                               ifp->rt_priority, ifp->idev->dev,
4765                                               expires, flags, GFP_KERNEL);
4766                 }
4767
4768                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4769                         rc = modify_prefix_route(ifp, expires, flags, true);
4770
4771                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4772                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4773                                               ifp->rt_priority, ifp->idev->dev,
4774                                               expires, flags, GFP_KERNEL);
4775                 }
4776         } else if (had_prefixroute) {
4777                 enum cleanup_prefix_rt_t action;
4778                 unsigned long rt_expires;
4779
4780                 write_lock_bh(&ifp->idev->lock);
4781                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4782                 write_unlock_bh(&ifp->idev->lock);
4783
4784                 if (action != CLEANUP_PREFIX_RT_NOP) {
4785                         cleanup_prefix_route(ifp, rt_expires,
4786                                 action == CLEANUP_PREFIX_RT_DEL, false);
4787                 }
4788         }
4789
4790         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4791                 if (was_managetempaddr &&
4792                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4793                         cfg->valid_lft = 0;
4794                         cfg->preferred_lft = 0;
4795                 }
4796                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4797                                  cfg->preferred_lft, !was_managetempaddr,
4798                                  jiffies);
4799         }
4800
4801         addrconf_verify_rtnl();
4802
4803         return 0;
4804 }
4805
4806 static int
4807 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4808                   struct netlink_ext_ack *extack)
4809 {
4810         struct net *net = sock_net(skb->sk);
4811         struct ifaddrmsg *ifm;
4812         struct nlattr *tb[IFA_MAX+1];
4813         struct in6_addr *peer_pfx;
4814         struct inet6_ifaddr *ifa;
4815         struct net_device *dev;
4816         struct inet6_dev *idev;
4817         struct ifa6_config cfg;
4818         int err;
4819
4820         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4821                                      ifa_ipv6_policy, extack);
4822         if (err < 0)
4823                 return err;
4824
4825         memset(&cfg, 0, sizeof(cfg));
4826
4827         ifm = nlmsg_data(nlh);
4828         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4829         if (!cfg.pfx)
4830                 return -EINVAL;
4831
4832         cfg.peer_pfx = peer_pfx;
4833         cfg.plen = ifm->ifa_prefixlen;
4834         if (tb[IFA_RT_PRIORITY])
4835                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4836
4837         cfg.valid_lft = INFINITY_LIFE_TIME;
4838         cfg.preferred_lft = INFINITY_LIFE_TIME;
4839
4840         if (tb[IFA_CACHEINFO]) {
4841                 struct ifa_cacheinfo *ci;
4842
4843                 ci = nla_data(tb[IFA_CACHEINFO]);
4844                 cfg.valid_lft = ci->ifa_valid;
4845                 cfg.preferred_lft = ci->ifa_prefered;
4846         }
4847
4848         dev =  __dev_get_by_index(net, ifm->ifa_index);
4849         if (!dev)
4850                 return -ENODEV;
4851
4852         if (tb[IFA_FLAGS])
4853                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4854         else
4855                 cfg.ifa_flags = ifm->ifa_flags;
4856
4857         /* We ignore other flags so far. */
4858         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4859                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4860                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4861
4862         idev = ipv6_find_idev(dev);
4863         if (IS_ERR(idev))
4864                 return PTR_ERR(idev);
4865
4866         if (!ipv6_allow_optimistic_dad(net, idev))
4867                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4868
4869         if (cfg.ifa_flags & IFA_F_NODAD &&
4870             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4871                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4872                 return -EINVAL;
4873         }
4874
4875         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4876         if (!ifa) {
4877                 /*
4878                  * It would be best to check for !NLM_F_CREATE here but
4879                  * userspace already relies on not having to provide this.
4880                  */
4881                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4882         }
4883
4884         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4885             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4886                 err = -EEXIST;
4887         else
4888                 err = inet6_addr_modify(ifa, &cfg);
4889
4890         in6_ifa_put(ifa);
4891
4892         return err;
4893 }
4894
4895 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4896                           u8 scope, int ifindex)
4897 {
4898         struct ifaddrmsg *ifm;
4899
4900         ifm = nlmsg_data(nlh);
4901         ifm->ifa_family = AF_INET6;
4902         ifm->ifa_prefixlen = prefixlen;
4903         ifm->ifa_flags = flags;
4904         ifm->ifa_scope = scope;
4905         ifm->ifa_index = ifindex;
4906 }
4907
4908 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4909                          unsigned long tstamp, u32 preferred, u32 valid)
4910 {
4911         struct ifa_cacheinfo ci;
4912
4913         ci.cstamp = cstamp_delta(cstamp);
4914         ci.tstamp = cstamp_delta(tstamp);
4915         ci.ifa_prefered = preferred;
4916         ci.ifa_valid = valid;
4917
4918         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4919 }
4920
4921 static inline int rt_scope(int ifa_scope)
4922 {
4923         if (ifa_scope & IFA_HOST)
4924                 return RT_SCOPE_HOST;
4925         else if (ifa_scope & IFA_LINK)
4926                 return RT_SCOPE_LINK;
4927         else if (ifa_scope & IFA_SITE)
4928                 return RT_SCOPE_SITE;
4929         else
4930                 return RT_SCOPE_UNIVERSE;
4931 }
4932
4933 static inline int inet6_ifaddr_msgsize(void)
4934 {
4935         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4936                + nla_total_size(16) /* IFA_LOCAL */
4937                + nla_total_size(16) /* IFA_ADDRESS */
4938                + nla_total_size(sizeof(struct ifa_cacheinfo))
4939                + nla_total_size(4)  /* IFA_FLAGS */
4940                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4941 }
4942
4943 enum addr_type_t {
4944         UNICAST_ADDR,
4945         MULTICAST_ADDR,
4946         ANYCAST_ADDR,
4947 };
4948
4949 struct inet6_fill_args {
4950         u32 portid;
4951         u32 seq;
4952         int event;
4953         unsigned int flags;
4954         int netnsid;
4955         int ifindex;
4956         enum addr_type_t type;
4957 };
4958
4959 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4960                              struct inet6_fill_args *args)
4961 {
4962         struct nlmsghdr  *nlh;
4963         u32 preferred, valid;
4964
4965         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4966                         sizeof(struct ifaddrmsg), args->flags);
4967         if (!nlh)
4968                 return -EMSGSIZE;
4969
4970         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4971                       ifa->idev->dev->ifindex);
4972
4973         if (args->netnsid >= 0 &&
4974             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4975                 goto error;
4976
4977         if (!((ifa->flags&IFA_F_PERMANENT) &&
4978               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4979                 preferred = ifa->prefered_lft;
4980                 valid = ifa->valid_lft;
4981                 if (preferred != INFINITY_LIFE_TIME) {
4982                         long tval = (jiffies - ifa->tstamp)/HZ;
4983                         if (preferred > tval)
4984                                 preferred -= tval;
4985                         else
4986                                 preferred = 0;
4987                         if (valid != INFINITY_LIFE_TIME) {
4988                                 if (valid > tval)
4989                                         valid -= tval;
4990                                 else
4991                                         valid = 0;
4992                         }
4993                 }
4994         } else {
4995                 preferred = INFINITY_LIFE_TIME;
4996                 valid = INFINITY_LIFE_TIME;
4997         }
4998
4999         if (!ipv6_addr_any(&ifa->peer_addr)) {
5000                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5001                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5002                         goto error;
5003         } else
5004                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5005                         goto error;
5006
5007         if (ifa->rt_priority &&
5008             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5009                 goto error;
5010
5011         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5012                 goto error;
5013
5014         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5015                 goto error;
5016
5017         nlmsg_end(skb, nlh);
5018         return 0;
5019
5020 error:
5021         nlmsg_cancel(skb, nlh);
5022         return -EMSGSIZE;
5023 }
5024
5025 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5026                                struct inet6_fill_args *args)
5027 {
5028         struct nlmsghdr  *nlh;
5029         u8 scope = RT_SCOPE_UNIVERSE;
5030         int ifindex = ifmca->idev->dev->ifindex;
5031
5032         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5033                 scope = RT_SCOPE_SITE;
5034
5035         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5036                         sizeof(struct ifaddrmsg), args->flags);
5037         if (!nlh)
5038                 return -EMSGSIZE;
5039
5040         if (args->netnsid >= 0 &&
5041             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5042                 return -EMSGSIZE;
5043
5044         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5045         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5046             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5047                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5048                 nlmsg_cancel(skb, nlh);
5049                 return -EMSGSIZE;
5050         }
5051
5052         nlmsg_end(skb, nlh);
5053         return 0;
5054 }
5055
5056 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5057                                struct inet6_fill_args *args)
5058 {
5059         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5060         int ifindex = dev ? dev->ifindex : 1;
5061         struct nlmsghdr  *nlh;
5062         u8 scope = RT_SCOPE_UNIVERSE;
5063
5064         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5065                 scope = RT_SCOPE_SITE;
5066
5067         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5068                         sizeof(struct ifaddrmsg), args->flags);
5069         if (!nlh)
5070                 return -EMSGSIZE;
5071
5072         if (args->netnsid >= 0 &&
5073             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5074                 return -EMSGSIZE;
5075
5076         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5077         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5078             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5079                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5080                 nlmsg_cancel(skb, nlh);
5081                 return -EMSGSIZE;
5082         }
5083
5084         nlmsg_end(skb, nlh);
5085         return 0;
5086 }
5087
5088 /* called with rcu_read_lock() */
5089 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5090                           struct netlink_callback *cb, int s_ip_idx,
5091                           struct inet6_fill_args *fillargs)
5092 {
5093         struct ifmcaddr6 *ifmca;
5094         struct ifacaddr6 *ifaca;
5095         int ip_idx = 0;
5096         int err = 1;
5097
5098         read_lock_bh(&idev->lock);
5099         switch (fillargs->type) {
5100         case UNICAST_ADDR: {
5101                 struct inet6_ifaddr *ifa;
5102                 fillargs->event = RTM_NEWADDR;
5103
5104                 /* unicast address incl. temp addr */
5105                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5106                         if (ip_idx < s_ip_idx)
5107                                 goto next;
5108                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5109                         if (err < 0)
5110                                 break;
5111                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5112 next:
5113                         ip_idx++;
5114                 }
5115                 break;
5116         }
5117         case MULTICAST_ADDR:
5118                 fillargs->event = RTM_GETMULTICAST;
5119
5120                 /* multicast address */
5121                 for (ifmca = idev->mc_list; ifmca;
5122                      ifmca = ifmca->next, ip_idx++) {
5123                         if (ip_idx < s_ip_idx)
5124                                 continue;
5125                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5126                         if (err < 0)
5127                                 break;
5128                 }
5129                 break;
5130         case ANYCAST_ADDR:
5131                 fillargs->event = RTM_GETANYCAST;
5132                 /* anycast address */
5133                 for (ifaca = idev->ac_list; ifaca;
5134                      ifaca = ifaca->aca_next, ip_idx++) {
5135                         if (ip_idx < s_ip_idx)
5136                                 continue;
5137                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5138                         if (err < 0)
5139                                 break;
5140                 }
5141                 break;
5142         default:
5143                 break;
5144         }
5145         read_unlock_bh(&idev->lock);
5146         cb->args[2] = ip_idx;
5147         return err;
5148 }
5149
5150 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5151                                        struct inet6_fill_args *fillargs,
5152                                        struct net **tgt_net, struct sock *sk,
5153                                        struct netlink_callback *cb)
5154 {
5155         struct netlink_ext_ack *extack = cb->extack;
5156         struct nlattr *tb[IFA_MAX+1];
5157         struct ifaddrmsg *ifm;
5158         int err, i;
5159
5160         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5161                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5162                 return -EINVAL;
5163         }
5164
5165         ifm = nlmsg_data(nlh);
5166         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5167                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5168                 return -EINVAL;
5169         }
5170
5171         fillargs->ifindex = ifm->ifa_index;
5172         if (fillargs->ifindex) {
5173                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5174                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5175         }
5176
5177         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5178                                             ifa_ipv6_policy, extack);
5179         if (err < 0)
5180                 return err;
5181
5182         for (i = 0; i <= IFA_MAX; ++i) {
5183                 if (!tb[i])
5184                         continue;
5185
5186                 if (i == IFA_TARGET_NETNSID) {
5187                         struct net *net;
5188
5189                         fillargs->netnsid = nla_get_s32(tb[i]);
5190                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5191                         if (IS_ERR(net)) {
5192                                 fillargs->netnsid = -1;
5193                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5194                                 return PTR_ERR(net);
5195                         }
5196                         *tgt_net = net;
5197                 } else {
5198                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5199                         return -EINVAL;
5200                 }
5201         }
5202
5203         return 0;
5204 }
5205
5206 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5207                            enum addr_type_t type)
5208 {
5209         const struct nlmsghdr *nlh = cb->nlh;
5210         struct inet6_fill_args fillargs = {
5211                 .portid = NETLINK_CB(cb->skb).portid,
5212                 .seq = cb->nlh->nlmsg_seq,
5213                 .flags = NLM_F_MULTI,
5214                 .netnsid = -1,
5215                 .type = type,
5216         };
5217         struct net *net = sock_net(skb->sk);
5218         struct net *tgt_net = net;
5219         int idx, s_idx, s_ip_idx;
5220         int h, s_h;
5221         struct net_device *dev;
5222         struct inet6_dev *idev;
5223         struct hlist_head *head;
5224         int err = 0;
5225
5226         s_h = cb->args[0];
5227         s_idx = idx = cb->args[1];
5228         s_ip_idx = cb->args[2];
5229
5230         if (cb->strict_check) {
5231                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5232                                                   skb->sk, cb);
5233                 if (err < 0)
5234                         goto put_tgt_net;
5235
5236                 err = 0;
5237                 if (fillargs.ifindex) {
5238                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5239                         if (!dev) {
5240                                 err = -ENODEV;
5241                                 goto put_tgt_net;
5242                         }
5243                         idev = __in6_dev_get(dev);
5244                         if (idev) {
5245                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5246                                                      &fillargs);
5247                                 if (err > 0)
5248                                         err = 0;
5249                         }
5250                         goto put_tgt_net;
5251                 }
5252         }
5253
5254         rcu_read_lock();
5255         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5256         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5257                 idx = 0;
5258                 head = &tgt_net->dev_index_head[h];
5259                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5260                         if (idx < s_idx)
5261                                 goto cont;
5262                         if (h > s_h || idx > s_idx)
5263                                 s_ip_idx = 0;
5264                         idev = __in6_dev_get(dev);
5265                         if (!idev)
5266                                 goto cont;
5267
5268                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5269                                            &fillargs) < 0)
5270                                 goto done;
5271 cont:
5272                         idx++;
5273                 }
5274         }
5275 done:
5276         rcu_read_unlock();
5277         cb->args[0] = h;
5278         cb->args[1] = idx;
5279 put_tgt_net:
5280         if (fillargs.netnsid >= 0)
5281                 put_net(tgt_net);
5282
5283         return skb->len ? : err;
5284 }
5285
5286 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5287 {
5288         enum addr_type_t type = UNICAST_ADDR;
5289
5290         return inet6_dump_addr(skb, cb, type);
5291 }
5292
5293 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5294 {
5295         enum addr_type_t type = MULTICAST_ADDR;
5296
5297         return inet6_dump_addr(skb, cb, type);
5298 }
5299
5300
5301 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5302 {
5303         enum addr_type_t type = ANYCAST_ADDR;
5304
5305         return inet6_dump_addr(skb, cb, type);
5306 }
5307
5308 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5309                                        const struct nlmsghdr *nlh,
5310                                        struct nlattr **tb,
5311                                        struct netlink_ext_ack *extack)
5312 {
5313         struct ifaddrmsg *ifm;
5314         int i, err;
5315
5316         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5317                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5318                 return -EINVAL;
5319         }
5320
5321         if (!netlink_strict_get_check(skb))
5322                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5323                                               ifa_ipv6_policy, extack);
5324
5325         ifm = nlmsg_data(nlh);
5326         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5327                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5328                 return -EINVAL;
5329         }
5330
5331         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5332                                             ifa_ipv6_policy, extack);
5333         if (err)
5334                 return err;
5335
5336         for (i = 0; i <= IFA_MAX; i++) {
5337                 if (!tb[i])
5338                         continue;
5339
5340                 switch (i) {
5341                 case IFA_TARGET_NETNSID:
5342                 case IFA_ADDRESS:
5343                 case IFA_LOCAL:
5344                         break;
5345                 default:
5346                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5347                         return -EINVAL;
5348                 }
5349         }
5350
5351         return 0;
5352 }
5353
5354 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5355                              struct netlink_ext_ack *extack)
5356 {
5357         struct net *net = sock_net(in_skb->sk);
5358         struct inet6_fill_args fillargs = {
5359                 .portid = NETLINK_CB(in_skb).portid,
5360                 .seq = nlh->nlmsg_seq,
5361                 .event = RTM_NEWADDR,
5362                 .flags = 0,
5363                 .netnsid = -1,
5364         };
5365         struct net *tgt_net = net;
5366         struct ifaddrmsg *ifm;
5367         struct nlattr *tb[IFA_MAX+1];
5368         struct in6_addr *addr = NULL, *peer;
5369         struct net_device *dev = NULL;
5370         struct inet6_ifaddr *ifa;
5371         struct sk_buff *skb;
5372         int err;
5373
5374         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5375         if (err < 0)
5376                 return err;
5377
5378         if (tb[IFA_TARGET_NETNSID]) {
5379                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5380
5381                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5382                                                   fillargs.netnsid);
5383                 if (IS_ERR(tgt_net))
5384                         return PTR_ERR(tgt_net);
5385         }
5386
5387         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5388         if (!addr)
5389                 return -EINVAL;
5390
5391         ifm = nlmsg_data(nlh);
5392         if (ifm->ifa_index)
5393                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5394
5395         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5396         if (!ifa) {
5397                 err = -EADDRNOTAVAIL;
5398                 goto errout;
5399         }
5400
5401         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5402         if (!skb) {
5403                 err = -ENOBUFS;
5404                 goto errout_ifa;
5405         }
5406
5407         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5408         if (err < 0) {
5409                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5410                 WARN_ON(err == -EMSGSIZE);
5411                 kfree_skb(skb);
5412                 goto errout_ifa;
5413         }
5414         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5415 errout_ifa:
5416         in6_ifa_put(ifa);
5417 errout:
5418         if (dev)
5419                 dev_put(dev);
5420         if (fillargs.netnsid >= 0)
5421                 put_net(tgt_net);
5422
5423         return err;
5424 }
5425
5426 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5427 {
5428         struct sk_buff *skb;
5429         struct net *net = dev_net(ifa->idev->dev);
5430         struct inet6_fill_args fillargs = {
5431                 .portid = 0,
5432                 .seq = 0,
5433                 .event = event,
5434                 .flags = 0,
5435                 .netnsid = -1,
5436         };
5437         int err = -ENOBUFS;
5438
5439         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5440         if (!skb)
5441                 goto errout;
5442
5443         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5444         if (err < 0) {
5445                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5446                 WARN_ON(err == -EMSGSIZE);
5447                 kfree_skb(skb);
5448                 goto errout;
5449         }
5450         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5451         return;
5452 errout:
5453         if (err < 0)
5454                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5455 }
5456
5457 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5458                                 __s32 *array, int bytes)
5459 {
5460         BUG_ON(bytes < (DEVCONF_MAX * 4));
5461
5462         memset(array, 0, bytes);
5463         array[DEVCONF_FORWARDING] = cnf->forwarding;
5464         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5465         array[DEVCONF_MTU6] = cnf->mtu6;
5466         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5467         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5468         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5469         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5470         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5471         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5472                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5473         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5474                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5475         array[DEVCONF_RTR_SOLICIT_DELAY] =
5476                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5477         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5478         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5479                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5480         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5481                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5482         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5483         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5484         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5485         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5486         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5487         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5488         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5489         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5490         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5491 #ifdef CONFIG_IPV6_ROUTER_PREF
5492         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5493         array[DEVCONF_RTR_PROBE_INTERVAL] =
5494                 jiffies_to_msecs(cnf->rtr_probe_interval);
5495 #ifdef CONFIG_IPV6_ROUTE_INFO
5496         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5497         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5498 #endif
5499 #endif
5500         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5501         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5503         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5504         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5505 #endif
5506 #ifdef CONFIG_IPV6_MROUTE
5507         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5508 #endif
5509         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5510         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5511         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5512         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5513         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5514         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5515         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5516         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5517         /* we omit DEVCONF_STABLE_SECRET for now */
5518         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5519         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5520         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5521         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5522         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5523 #ifdef CONFIG_IPV6_SEG6_HMAC
5524         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5525 #endif
5526         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5527         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5528         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5529         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5530         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5531 }
5532
5533 static inline size_t inet6_ifla6_size(void)
5534 {
5535         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5536              + nla_total_size(sizeof(struct ifla_cacheinfo))
5537              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5538              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5539              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5540              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5541              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5542              + 0;
5543 }
5544
5545 static inline size_t inet6_if_nlmsg_size(void)
5546 {
5547         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5548                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5549                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5550                + nla_total_size(4) /* IFLA_MTU */
5551                + nla_total_size(4) /* IFLA_LINK */
5552                + nla_total_size(1) /* IFLA_OPERSTATE */
5553                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5554 }
5555
5556 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5557                                         int bytes)
5558 {
5559         int i;
5560         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5561         BUG_ON(pad < 0);
5562
5563         /* Use put_unaligned() because stats may not be aligned for u64. */
5564         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5565         for (i = 1; i < ICMP6_MIB_MAX; i++)
5566                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5567
5568         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5569 }
5570
5571 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5572                                         int bytes, size_t syncpoff)
5573 {
5574         int i, c;
5575         u64 buff[IPSTATS_MIB_MAX];
5576         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5577
5578         BUG_ON(pad < 0);
5579
5580         memset(buff, 0, sizeof(buff));
5581         buff[0] = IPSTATS_MIB_MAX;
5582
5583         for_each_possible_cpu(c) {
5584                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5585                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5586         }
5587
5588         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5589         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5590 }
5591
5592 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5593                              int bytes)
5594 {
5595         switch (attrtype) {
5596         case IFLA_INET6_STATS:
5597                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5598                                      offsetof(struct ipstats_mib, syncp));
5599                 break;
5600         case IFLA_INET6_ICMP6STATS:
5601                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5602                 break;
5603         }
5604 }
5605
5606 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5607                                   u32 ext_filter_mask)
5608 {
5609         struct nlattr *nla;
5610         struct ifla_cacheinfo ci;
5611
5612         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5613                 goto nla_put_failure;
5614         ci.max_reasm_len = IPV6_MAXPLEN;
5615         ci.tstamp = cstamp_delta(idev->tstamp);
5616         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5617         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5618         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5619                 goto nla_put_failure;
5620         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5621         if (!nla)
5622                 goto nla_put_failure;
5623         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5624
5625         /* XXX - MC not implemented */
5626
5627         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5628                 return 0;
5629
5630         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5631         if (!nla)
5632                 goto nla_put_failure;
5633         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5634
5635         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5636         if (!nla)
5637                 goto nla_put_failure;
5638         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5639
5640         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5641         if (!nla)
5642                 goto nla_put_failure;
5643         read_lock_bh(&idev->lock);
5644         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5645         read_unlock_bh(&idev->lock);
5646
5647         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5648                 goto nla_put_failure;
5649
5650         return 0;
5651
5652 nla_put_failure:
5653         return -EMSGSIZE;
5654 }
5655
5656 static size_t inet6_get_link_af_size(const struct net_device *dev,
5657                                      u32 ext_filter_mask)
5658 {
5659         if (!__in6_dev_get(dev))
5660                 return 0;
5661
5662         return inet6_ifla6_size();
5663 }
5664
5665 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5666                               u32 ext_filter_mask)
5667 {
5668         struct inet6_dev *idev = __in6_dev_get(dev);
5669
5670         if (!idev)
5671                 return -ENODATA;
5672
5673         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5674                 return -EMSGSIZE;
5675
5676         return 0;
5677 }
5678
5679 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5680 {
5681         struct inet6_ifaddr *ifp;
5682         struct net_device *dev = idev->dev;
5683         bool clear_token, update_rs = false;
5684         struct in6_addr ll_addr;
5685
5686         ASSERT_RTNL();
5687
5688         if (!token)
5689                 return -EINVAL;
5690         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5691                 return -EINVAL;
5692         if (!ipv6_accept_ra(idev))
5693                 return -EINVAL;
5694         if (idev->cnf.rtr_solicits == 0)
5695                 return -EINVAL;
5696
5697         write_lock_bh(&idev->lock);
5698
5699         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5700         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5701
5702         write_unlock_bh(&idev->lock);
5703
5704         clear_token = ipv6_addr_any(token);
5705         if (clear_token)
5706                 goto update_lft;
5707
5708         if (!idev->dead && (idev->if_flags & IF_READY) &&
5709             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5710                              IFA_F_OPTIMISTIC)) {
5711                 /* If we're not ready, then normal ifup will take care
5712                  * of this. Otherwise, we need to request our rs here.
5713                  */
5714                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5715                 update_rs = true;
5716         }
5717
5718 update_lft:
5719         write_lock_bh(&idev->lock);
5720
5721         if (update_rs) {
5722                 idev->if_flags |= IF_RS_SENT;
5723                 idev->rs_interval = rfc3315_s14_backoff_init(
5724                         idev->cnf.rtr_solicit_interval);
5725                 idev->rs_probes = 1;
5726                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5727         }
5728
5729         /* Well, that's kinda nasty ... */
5730         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5731                 spin_lock(&ifp->lock);
5732                 if (ifp->tokenized) {
5733                         ifp->valid_lft = 0;
5734                         ifp->prefered_lft = 0;
5735                 }
5736                 spin_unlock(&ifp->lock);
5737         }
5738
5739         write_unlock_bh(&idev->lock);
5740         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5741         addrconf_verify_rtnl();
5742         return 0;
5743 }
5744
5745 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5746         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5747         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5748 };
5749
5750 static int check_addr_gen_mode(int mode)
5751 {
5752         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5753             mode != IN6_ADDR_GEN_MODE_NONE &&
5754             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5755             mode != IN6_ADDR_GEN_MODE_RANDOM)
5756                 return -EINVAL;
5757         return 1;
5758 }
5759
5760 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5761                                 int mode)
5762 {
5763         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5764             !idev->cnf.stable_secret.initialized &&
5765             !net->ipv6.devconf_dflt->stable_secret.initialized)
5766                 return -EINVAL;
5767         return 1;
5768 }
5769
5770 static int inet6_validate_link_af(const struct net_device *dev,
5771                                   const struct nlattr *nla)
5772 {
5773         struct nlattr *tb[IFLA_INET6_MAX + 1];
5774         struct inet6_dev *idev = NULL;
5775         int err;
5776
5777         if (dev) {
5778                 idev = __in6_dev_get(dev);
5779                 if (!idev)
5780                         return -EAFNOSUPPORT;
5781         }
5782
5783         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5784                                           inet6_af_policy, NULL);
5785         if (err)
5786                 return err;
5787
5788         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5789                 return -EINVAL;
5790
5791         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5792                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5793
5794                 if (check_addr_gen_mode(mode) < 0)
5795                         return -EINVAL;
5796                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5797                         return -EINVAL;
5798         }
5799
5800         return 0;
5801 }
5802
5803 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5804 {
5805         struct inet6_dev *idev = __in6_dev_get(dev);
5806         struct nlattr *tb[IFLA_INET6_MAX + 1];
5807         int err;
5808
5809         if (!idev)
5810                 return -EAFNOSUPPORT;
5811
5812         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5813                 BUG();
5814
5815         if (tb[IFLA_INET6_TOKEN]) {
5816                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5817                 if (err)
5818                         return err;
5819         }
5820
5821         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5822                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5823
5824                 idev->cnf.addr_gen_mode = mode;
5825         }
5826
5827         return 0;
5828 }
5829
5830 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5831                              u32 portid, u32 seq, int event, unsigned int flags)
5832 {
5833         struct net_device *dev = idev->dev;
5834         struct ifinfomsg *hdr;
5835         struct nlmsghdr *nlh;
5836         void *protoinfo;
5837
5838         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5839         if (!nlh)
5840                 return -EMSGSIZE;
5841
5842         hdr = nlmsg_data(nlh);
5843         hdr->ifi_family = AF_INET6;
5844         hdr->__ifi_pad = 0;
5845         hdr->ifi_type = dev->type;
5846         hdr->ifi_index = dev->ifindex;
5847         hdr->ifi_flags = dev_get_flags(dev);
5848         hdr->ifi_change = 0;
5849
5850         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5851             (dev->addr_len &&
5852              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5853             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5854             (dev->ifindex != dev_get_iflink(dev) &&
5855              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5856             nla_put_u8(skb, IFLA_OPERSTATE,
5857                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5858                 goto nla_put_failure;
5859         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5860         if (!protoinfo)
5861                 goto nla_put_failure;
5862
5863         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5864                 goto nla_put_failure;
5865
5866         nla_nest_end(skb, protoinfo);
5867         nlmsg_end(skb, nlh);
5868         return 0;
5869
5870 nla_put_failure:
5871         nlmsg_cancel(skb, nlh);
5872         return -EMSGSIZE;
5873 }
5874
5875 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5876                                    struct netlink_ext_ack *extack)
5877 {
5878         struct ifinfomsg *ifm;
5879
5880         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5881                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5882                 return -EINVAL;
5883         }
5884
5885         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5886                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5887                 return -EINVAL;
5888         }
5889
5890         ifm = nlmsg_data(nlh);
5891         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5892             ifm->ifi_change || ifm->ifi_index) {
5893                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5894                 return -EINVAL;
5895         }
5896
5897         return 0;
5898 }
5899
5900 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5901 {
5902         struct net *net = sock_net(skb->sk);
5903         int h, s_h;
5904         int idx = 0, s_idx;
5905         struct net_device *dev;
5906         struct inet6_dev *idev;
5907         struct hlist_head *head;
5908
5909         /* only requests using strict checking can pass data to
5910          * influence the dump
5911          */
5912         if (cb->strict_check) {
5913                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5914
5915                 if (err < 0)
5916                         return err;
5917         }
5918
5919         s_h = cb->args[0];
5920         s_idx = cb->args[1];
5921
5922         rcu_read_lock();
5923         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5924                 idx = 0;
5925                 head = &net->dev_index_head[h];
5926                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5927                         if (idx < s_idx)
5928                                 goto cont;
5929                         idev = __in6_dev_get(dev);
5930                         if (!idev)
5931                                 goto cont;
5932                         if (inet6_fill_ifinfo(skb, idev,
5933                                               NETLINK_CB(cb->skb).portid,
5934                                               cb->nlh->nlmsg_seq,
5935                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5936                                 goto out;
5937 cont:
5938                         idx++;
5939                 }
5940         }
5941 out:
5942         rcu_read_unlock();
5943         cb->args[1] = idx;
5944         cb->args[0] = h;
5945
5946         return skb->len;
5947 }
5948
5949 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5950 {
5951         struct sk_buff *skb;
5952         struct net *net = dev_net(idev->dev);
5953         int err = -ENOBUFS;
5954
5955         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5956         if (!skb)
5957                 goto errout;
5958
5959         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5960         if (err < 0) {
5961                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5962                 WARN_ON(err == -EMSGSIZE);
5963                 kfree_skb(skb);
5964                 goto errout;
5965         }
5966         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5967         return;
5968 errout:
5969         if (err < 0)
5970                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5971 }
5972
5973 static inline size_t inet6_prefix_nlmsg_size(void)
5974 {
5975         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5976                + nla_total_size(sizeof(struct in6_addr))
5977                + nla_total_size(sizeof(struct prefix_cacheinfo));
5978 }
5979
5980 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5981                              struct prefix_info *pinfo, u32 portid, u32 seq,
5982                              int event, unsigned int flags)
5983 {
5984         struct prefixmsg *pmsg;
5985         struct nlmsghdr *nlh;
5986         struct prefix_cacheinfo ci;
5987
5988         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5989         if (!nlh)
5990                 return -EMSGSIZE;
5991
5992         pmsg = nlmsg_data(nlh);
5993         pmsg->prefix_family = AF_INET6;
5994         pmsg->prefix_pad1 = 0;
5995         pmsg->prefix_pad2 = 0;
5996         pmsg->prefix_ifindex = idev->dev->ifindex;
5997         pmsg->prefix_len = pinfo->prefix_len;
5998         pmsg->prefix_type = pinfo->type;
5999         pmsg->prefix_pad3 = 0;
6000         pmsg->prefix_flags = 0;
6001         if (pinfo->onlink)
6002                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6003         if (pinfo->autoconf)
6004                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6005
6006         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6007                 goto nla_put_failure;
6008         ci.preferred_time = ntohl(pinfo->prefered);
6009         ci.valid_time = ntohl(pinfo->valid);
6010         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6011                 goto nla_put_failure;
6012         nlmsg_end(skb, nlh);
6013         return 0;
6014
6015 nla_put_failure:
6016         nlmsg_cancel(skb, nlh);
6017         return -EMSGSIZE;
6018 }
6019
6020 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6021                          struct prefix_info *pinfo)
6022 {
6023         struct sk_buff *skb;
6024         struct net *net = dev_net(idev->dev);
6025         int err = -ENOBUFS;
6026
6027         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6028         if (!skb)
6029                 goto errout;
6030
6031         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6032         if (err < 0) {
6033                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6034                 WARN_ON(err == -EMSGSIZE);
6035                 kfree_skb(skb);
6036                 goto errout;
6037         }
6038         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6039         return;
6040 errout:
6041         if (err < 0)
6042                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6043 }
6044
6045 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6046 {
6047         struct net *net = dev_net(ifp->idev->dev);
6048
6049         if (event)
6050                 ASSERT_RTNL();
6051
6052         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6053
6054         switch (event) {
6055         case RTM_NEWADDR:
6056                 /*
6057                  * If the address was optimistic we inserted the route at the
6058                  * start of our DAD process, so we don't need to do it again.
6059                  * If the device was taken down in the middle of the DAD
6060                  * cycle there is a race where we could get here without a
6061                  * host route, so nothing to insert. That will be fixed when
6062                  * the device is brought up.
6063                  */
6064                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6065                         ip6_ins_rt(net, ifp->rt);
6066                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6067                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6068                                 &ifp->addr, ifp->idev->dev->name);
6069                 }
6070
6071                 if (ifp->idev->cnf.forwarding)
6072                         addrconf_join_anycast(ifp);
6073                 if (!ipv6_addr_any(&ifp->peer_addr))
6074                         addrconf_prefix_route(&ifp->peer_addr, 128,
6075                                               ifp->rt_priority, ifp->idev->dev,
6076                                               0, 0, GFP_ATOMIC);
6077                 break;
6078         case RTM_DELADDR:
6079                 if (ifp->idev->cnf.forwarding)
6080                         addrconf_leave_anycast(ifp);
6081                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6082                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6083                         struct fib6_info *rt;
6084
6085                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6086                                                        ifp->idev->dev, 0, 0,
6087                                                        false);
6088                         if (rt)
6089                                 ip6_del_rt(net, rt);
6090                 }
6091                 if (ifp->rt) {
6092                         ip6_del_rt(net, ifp->rt);
6093                         ifp->rt = NULL;
6094                 }
6095                 rt_genid_bump_ipv6(net);
6096                 break;
6097         }
6098         atomic_inc(&net->ipv6.dev_addr_genid);
6099 }
6100
6101 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6102 {
6103         rcu_read_lock_bh();
6104         if (likely(ifp->idev->dead == 0))
6105                 __ipv6_ifa_notify(event, ifp);
6106         rcu_read_unlock_bh();
6107 }
6108
6109 #ifdef CONFIG_SYSCTL
6110
6111 static
6112 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6113                            void __user *buffer, size_t *lenp, loff_t *ppos)
6114 {
6115         int *valp = ctl->data;
6116         int val = *valp;
6117         loff_t pos = *ppos;
6118         struct ctl_table lctl;
6119         int ret;
6120
6121         /*
6122          * ctl->data points to idev->cnf.forwarding, we should
6123          * not modify it until we get the rtnl lock.
6124          */
6125         lctl = *ctl;
6126         lctl.data = &val;
6127
6128         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6129
6130         if (write)
6131                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6132         if (ret)
6133                 *ppos = pos;
6134         return ret;
6135 }
6136
6137 static
6138 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6139                         void __user *buffer, size_t *lenp, loff_t *ppos)
6140 {
6141         struct inet6_dev *idev = ctl->extra1;
6142         int min_mtu = IPV6_MIN_MTU;
6143         struct ctl_table lctl;
6144
6145         lctl = *ctl;
6146         lctl.extra1 = &min_mtu;
6147         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6148
6149         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6150 }
6151
6152 static void dev_disable_change(struct inet6_dev *idev)
6153 {
6154         struct netdev_notifier_info info;
6155
6156         if (!idev || !idev->dev)
6157                 return;
6158
6159         netdev_notifier_info_init(&info, idev->dev);
6160         if (idev->cnf.disable_ipv6)
6161                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6162         else
6163                 addrconf_notify(NULL, NETDEV_UP, &info);
6164 }
6165
6166 static void addrconf_disable_change(struct net *net, __s32 newf)
6167 {
6168         struct net_device *dev;
6169         struct inet6_dev *idev;
6170
6171         for_each_netdev(net, dev) {
6172                 idev = __in6_dev_get(dev);
6173                 if (idev) {
6174                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6175                         idev->cnf.disable_ipv6 = newf;
6176                         if (changed)
6177                                 dev_disable_change(idev);
6178                 }
6179         }
6180 }
6181
6182 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6183 {
6184         struct net *net;
6185         int old;
6186
6187         if (!rtnl_trylock())
6188                 return restart_syscall();
6189
6190         net = (struct net *)table->extra2;
6191         old = *p;
6192         *p = newf;
6193
6194         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6195                 rtnl_unlock();
6196                 return 0;
6197         }
6198
6199         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6200                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6201                 addrconf_disable_change(net, newf);
6202         } else if ((!newf) ^ (!old))
6203                 dev_disable_change((struct inet6_dev *)table->extra1);
6204
6205         rtnl_unlock();
6206         return 0;
6207 }
6208
6209 static
6210 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6211                             void __user *buffer, size_t *lenp, loff_t *ppos)
6212 {
6213         int *valp = ctl->data;
6214         int val = *valp;
6215         loff_t pos = *ppos;
6216         struct ctl_table lctl;
6217         int ret;
6218
6219         /*
6220          * ctl->data points to idev->cnf.disable_ipv6, we should
6221          * not modify it until we get the rtnl lock.
6222          */
6223         lctl = *ctl;
6224         lctl.data = &val;
6225
6226         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6227
6228         if (write)
6229                 ret = addrconf_disable_ipv6(ctl, valp, val);
6230         if (ret)
6231                 *ppos = pos;
6232         return ret;
6233 }
6234
6235 static
6236 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6237                               void __user *buffer, size_t *lenp, loff_t *ppos)
6238 {
6239         int *valp = ctl->data;
6240         int ret;
6241         int old, new;
6242
6243         old = *valp;
6244         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6245         new = *valp;
6246
6247         if (write && old != new) {
6248                 struct net *net = ctl->extra2;
6249
6250                 if (!rtnl_trylock())
6251                         return restart_syscall();
6252
6253                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6254                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6255                                                      NETCONFA_PROXY_NEIGH,
6256                                                      NETCONFA_IFINDEX_DEFAULT,
6257                                                      net->ipv6.devconf_dflt);
6258                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6259                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6260                                                      NETCONFA_PROXY_NEIGH,
6261                                                      NETCONFA_IFINDEX_ALL,
6262                                                      net->ipv6.devconf_all);
6263                 else {
6264                         struct inet6_dev *idev = ctl->extra1;
6265
6266                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6267                                                      NETCONFA_PROXY_NEIGH,
6268                                                      idev->dev->ifindex,
6269                                                      &idev->cnf);
6270                 }
6271                 rtnl_unlock();
6272         }
6273
6274         return ret;
6275 }
6276
6277 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6278                                          void __user *buffer, size_t *lenp,
6279                                          loff_t *ppos)
6280 {
6281         int ret = 0;
6282         u32 new_val;
6283         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6284         struct net *net = (struct net *)ctl->extra2;
6285         struct ctl_table tmp = {
6286                 .data = &new_val,
6287                 .maxlen = sizeof(new_val),
6288                 .mode = ctl->mode,
6289         };
6290
6291         if (!rtnl_trylock())
6292                 return restart_syscall();
6293
6294         new_val = *((u32 *)ctl->data);
6295
6296         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6297         if (ret != 0)
6298                 goto out;
6299
6300         if (write) {
6301                 if (check_addr_gen_mode(new_val) < 0) {
6302                         ret = -EINVAL;
6303                         goto out;
6304                 }
6305
6306                 if (idev) {
6307                         if (check_stable_privacy(idev, net, new_val) < 0) {
6308                                 ret = -EINVAL;
6309                                 goto out;
6310                         }
6311
6312                         if (idev->cnf.addr_gen_mode != new_val) {
6313                                 idev->cnf.addr_gen_mode = new_val;
6314                                 addrconf_dev_config(idev->dev);
6315                         }
6316                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6317                         struct net_device *dev;
6318
6319                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6320                         for_each_netdev(net, dev) {
6321                                 idev = __in6_dev_get(dev);
6322                                 if (idev &&
6323                                     idev->cnf.addr_gen_mode != new_val) {
6324                                         idev->cnf.addr_gen_mode = new_val;
6325                                         addrconf_dev_config(idev->dev);
6326                                 }
6327                         }
6328                 }
6329
6330                 *((u32 *)ctl->data) = new_val;
6331         }
6332
6333 out:
6334         rtnl_unlock();
6335
6336         return ret;
6337 }
6338
6339 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6340                                          void __user *buffer, size_t *lenp,
6341                                          loff_t *ppos)
6342 {
6343         int err;
6344         struct in6_addr addr;
6345         char str[IPV6_MAX_STRLEN];
6346         struct ctl_table lctl = *ctl;
6347         struct net *net = ctl->extra2;
6348         struct ipv6_stable_secret *secret = ctl->data;
6349
6350         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6351                 return -EIO;
6352
6353         lctl.maxlen = IPV6_MAX_STRLEN;
6354         lctl.data = str;
6355
6356         if (!rtnl_trylock())
6357                 return restart_syscall();
6358
6359         if (!write && !secret->initialized) {
6360                 err = -EIO;
6361                 goto out;
6362         }
6363
6364         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6365         if (err >= sizeof(str)) {
6366                 err = -EIO;
6367                 goto out;
6368         }
6369
6370         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6371         if (err || !write)
6372                 goto out;
6373
6374         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6375                 err = -EIO;
6376                 goto out;
6377         }
6378
6379         secret->initialized = true;
6380         secret->secret = addr;
6381
6382         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6383                 struct net_device *dev;
6384
6385                 for_each_netdev(net, dev) {
6386                         struct inet6_dev *idev = __in6_dev_get(dev);
6387
6388                         if (idev) {
6389                                 idev->cnf.addr_gen_mode =
6390                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6391                         }
6392                 }
6393         } else {
6394                 struct inet6_dev *idev = ctl->extra1;
6395
6396                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6397         }
6398
6399 out:
6400         rtnl_unlock();
6401
6402         return err;
6403 }
6404
6405 static
6406 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6407                                                 int write,
6408                                                 void __user *buffer,
6409                                                 size_t *lenp,
6410                                                 loff_t *ppos)
6411 {
6412         int *valp = ctl->data;
6413         int val = *valp;
6414         loff_t pos = *ppos;
6415         struct ctl_table lctl;
6416         int ret;
6417
6418         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6419          * we should not modify it until we get the rtnl lock.
6420          */
6421         lctl = *ctl;
6422         lctl.data = &val;
6423
6424         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6425
6426         if (write)
6427                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6428         if (ret)
6429                 *ppos = pos;
6430         return ret;
6431 }
6432
6433 static
6434 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6435 {
6436         if (rt) {
6437                 if (action)
6438                         rt->dst.flags |= DST_NOPOLICY;
6439                 else
6440                         rt->dst.flags &= ~DST_NOPOLICY;
6441         }
6442 }
6443
6444 static
6445 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6446 {
6447         struct inet6_ifaddr *ifa;
6448
6449         read_lock_bh(&idev->lock);
6450         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6451                 spin_lock(&ifa->lock);
6452                 if (ifa->rt) {
6453                         /* host routes only use builtin fib6_nh */
6454                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6455                         int cpu;
6456
6457                         rcu_read_lock();
6458                         ifa->rt->dst_nopolicy = val ? true : false;
6459                         if (nh->rt6i_pcpu) {
6460                                 for_each_possible_cpu(cpu) {
6461                                         struct rt6_info **rtp;
6462
6463                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6464                                         addrconf_set_nopolicy(*rtp, val);
6465                                 }
6466                         }
6467                         rcu_read_unlock();
6468                 }
6469                 spin_unlock(&ifa->lock);
6470         }
6471         read_unlock_bh(&idev->lock);
6472 }
6473
6474 static
6475 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6476 {
6477         struct inet6_dev *idev;
6478         struct net *net;
6479
6480         if (!rtnl_trylock())
6481                 return restart_syscall();
6482
6483         *valp = val;
6484
6485         net = (struct net *)ctl->extra2;
6486         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6487                 rtnl_unlock();
6488                 return 0;
6489         }
6490
6491         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6492                 struct net_device *dev;
6493
6494                 for_each_netdev(net, dev) {
6495                         idev = __in6_dev_get(dev);
6496                         if (idev)
6497                                 addrconf_disable_policy_idev(idev, val);
6498                 }
6499         } else {
6500                 idev = (struct inet6_dev *)ctl->extra1;
6501                 addrconf_disable_policy_idev(idev, val);
6502         }
6503
6504         rtnl_unlock();
6505         return 0;
6506 }
6507
6508 static
6509 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6510                                    void __user *buffer, size_t *lenp,
6511                                    loff_t *ppos)
6512 {
6513         int *valp = ctl->data;
6514         int val = *valp;
6515         loff_t pos = *ppos;
6516         struct ctl_table lctl;
6517         int ret;
6518
6519         lctl = *ctl;
6520         lctl.data = &val;
6521         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6522
6523         if (write && (*valp != val))
6524                 ret = addrconf_disable_policy(ctl, valp, val);
6525
6526         if (ret)
6527                 *ppos = pos;
6528
6529         return ret;
6530 }
6531
6532 static int minus_one = -1;
6533 static const int two_five_five = 255;
6534
6535 static const struct ctl_table addrconf_sysctl[] = {
6536         {
6537                 .procname       = "forwarding",
6538                 .data           = &ipv6_devconf.forwarding,
6539                 .maxlen         = sizeof(int),
6540                 .mode           = 0644,
6541                 .proc_handler   = addrconf_sysctl_forward,
6542         },
6543         {
6544                 .procname       = "hop_limit",
6545                 .data           = &ipv6_devconf.hop_limit,
6546                 .maxlen         = sizeof(int),
6547                 .mode           = 0644,
6548                 .proc_handler   = proc_dointvec_minmax,
6549                 .extra1         = (void *)SYSCTL_ONE,
6550                 .extra2         = (void *)&two_five_five,
6551         },
6552         {
6553                 .procname       = "mtu",
6554                 .data           = &ipv6_devconf.mtu6,
6555                 .maxlen         = sizeof(int),
6556                 .mode           = 0644,
6557                 .proc_handler   = addrconf_sysctl_mtu,
6558         },
6559         {
6560                 .procname       = "accept_ra",
6561                 .data           = &ipv6_devconf.accept_ra,
6562                 .maxlen         = sizeof(int),
6563                 .mode           = 0644,
6564                 .proc_handler   = proc_dointvec,
6565         },
6566         {
6567                 .procname       = "accept_redirects",
6568                 .data           = &ipv6_devconf.accept_redirects,
6569                 .maxlen         = sizeof(int),
6570                 .mode           = 0644,
6571                 .proc_handler   = proc_dointvec,
6572         },
6573         {
6574                 .procname       = "autoconf",
6575                 .data           = &ipv6_devconf.autoconf,
6576                 .maxlen         = sizeof(int),
6577                 .mode           = 0644,
6578                 .proc_handler   = proc_dointvec,
6579         },
6580         {
6581                 .procname       = "dad_transmits",
6582                 .data           = &ipv6_devconf.dad_transmits,
6583                 .maxlen         = sizeof(int),
6584                 .mode           = 0644,
6585                 .proc_handler   = proc_dointvec,
6586         },
6587         {
6588                 .procname       = "router_solicitations",
6589                 .data           = &ipv6_devconf.rtr_solicits,
6590                 .maxlen         = sizeof(int),
6591                 .mode           = 0644,
6592                 .proc_handler   = proc_dointvec_minmax,
6593                 .extra1         = &minus_one,
6594         },
6595         {
6596                 .procname       = "router_solicitation_interval",
6597                 .data           = &ipv6_devconf.rtr_solicit_interval,
6598                 .maxlen         = sizeof(int),
6599                 .mode           = 0644,
6600                 .proc_handler   = proc_dointvec_jiffies,
6601         },
6602         {
6603                 .procname       = "router_solicitation_max_interval",
6604                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6605                 .maxlen         = sizeof(int),
6606                 .mode           = 0644,
6607                 .proc_handler   = proc_dointvec_jiffies,
6608         },
6609         {
6610                 .procname       = "router_solicitation_delay",
6611                 .data           = &ipv6_devconf.rtr_solicit_delay,
6612                 .maxlen         = sizeof(int),
6613                 .mode           = 0644,
6614                 .proc_handler   = proc_dointvec_jiffies,
6615         },
6616         {
6617                 .procname       = "force_mld_version",
6618                 .data           = &ipv6_devconf.force_mld_version,
6619                 .maxlen         = sizeof(int),
6620                 .mode           = 0644,
6621                 .proc_handler   = proc_dointvec,
6622         },
6623         {
6624                 .procname       = "mldv1_unsolicited_report_interval",
6625                 .data           =
6626                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6627                 .maxlen         = sizeof(int),
6628                 .mode           = 0644,
6629                 .proc_handler   = proc_dointvec_ms_jiffies,
6630         },
6631         {
6632                 .procname       = "mldv2_unsolicited_report_interval",
6633                 .data           =
6634                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6635                 .maxlen         = sizeof(int),
6636                 .mode           = 0644,
6637                 .proc_handler   = proc_dointvec_ms_jiffies,
6638         },
6639         {
6640                 .procname       = "use_tempaddr",
6641                 .data           = &ipv6_devconf.use_tempaddr,
6642                 .maxlen         = sizeof(int),
6643                 .mode           = 0644,
6644                 .proc_handler   = proc_dointvec,
6645         },
6646         {
6647                 .procname       = "temp_valid_lft",
6648                 .data           = &ipv6_devconf.temp_valid_lft,
6649                 .maxlen         = sizeof(int),
6650                 .mode           = 0644,
6651                 .proc_handler   = proc_dointvec,
6652         },
6653         {
6654                 .procname       = "temp_prefered_lft",
6655                 .data           = &ipv6_devconf.temp_prefered_lft,
6656                 .maxlen         = sizeof(int),
6657                 .mode           = 0644,
6658                 .proc_handler   = proc_dointvec,
6659         },
6660         {
6661                 .procname       = "regen_max_retry",
6662                 .data           = &ipv6_devconf.regen_max_retry,
6663                 .maxlen         = sizeof(int),
6664                 .mode           = 0644,
6665                 .proc_handler   = proc_dointvec,
6666         },
6667         {
6668                 .procname       = "max_desync_factor",
6669                 .data           = &ipv6_devconf.max_desync_factor,
6670                 .maxlen         = sizeof(int),
6671                 .mode           = 0644,
6672                 .proc_handler   = proc_dointvec,
6673         },
6674         {
6675                 .procname       = "max_addresses",
6676                 .data           = &ipv6_devconf.max_addresses,
6677                 .maxlen         = sizeof(int),
6678                 .mode           = 0644,
6679                 .proc_handler   = proc_dointvec,
6680         },
6681         {
6682                 .procname       = "accept_ra_defrtr",
6683                 .data           = &ipv6_devconf.accept_ra_defrtr,
6684                 .maxlen         = sizeof(int),
6685                 .mode           = 0644,
6686                 .proc_handler   = proc_dointvec,
6687         },
6688         {
6689                 .procname       = "accept_ra_min_hop_limit",
6690                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6691                 .maxlen         = sizeof(int),
6692                 .mode           = 0644,
6693                 .proc_handler   = proc_dointvec,
6694         },
6695         {
6696                 .procname       = "accept_ra_pinfo",
6697                 .data           = &ipv6_devconf.accept_ra_pinfo,
6698                 .maxlen         = sizeof(int),
6699                 .mode           = 0644,
6700                 .proc_handler   = proc_dointvec,
6701         },
6702 #ifdef CONFIG_IPV6_ROUTER_PREF
6703         {
6704                 .procname       = "accept_ra_rtr_pref",
6705                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6706                 .maxlen         = sizeof(int),
6707                 .mode           = 0644,
6708                 .proc_handler   = proc_dointvec,
6709         },
6710         {
6711                 .procname       = "router_probe_interval",
6712                 .data           = &ipv6_devconf.rtr_probe_interval,
6713                 .maxlen         = sizeof(int),
6714                 .mode           = 0644,
6715                 .proc_handler   = proc_dointvec_jiffies,
6716         },
6717 #ifdef CONFIG_IPV6_ROUTE_INFO
6718         {
6719                 .procname       = "accept_ra_rt_info_min_plen",
6720                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6721                 .maxlen         = sizeof(int),
6722                 .mode           = 0644,
6723                 .proc_handler   = proc_dointvec,
6724         },
6725         {
6726                 .procname       = "accept_ra_rt_info_max_plen",
6727                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6728                 .maxlen         = sizeof(int),
6729                 .mode           = 0644,
6730                 .proc_handler   = proc_dointvec,
6731         },
6732 #endif
6733 #endif
6734         {
6735                 .procname       = "proxy_ndp",
6736                 .data           = &ipv6_devconf.proxy_ndp,
6737                 .maxlen         = sizeof(int),
6738                 .mode           = 0644,
6739                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6740         },
6741         {
6742                 .procname       = "accept_source_route",
6743                 .data           = &ipv6_devconf.accept_source_route,
6744                 .maxlen         = sizeof(int),
6745                 .mode           = 0644,
6746                 .proc_handler   = proc_dointvec,
6747         },
6748 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6749         {
6750                 .procname       = "optimistic_dad",
6751                 .data           = &ipv6_devconf.optimistic_dad,
6752                 .maxlen         = sizeof(int),
6753                 .mode           = 0644,
6754                 .proc_handler   = proc_dointvec,
6755         },
6756         {
6757                 .procname       = "use_optimistic",
6758                 .data           = &ipv6_devconf.use_optimistic,
6759                 .maxlen         = sizeof(int),
6760                 .mode           = 0644,
6761                 .proc_handler   = proc_dointvec,
6762         },
6763 #endif
6764 #ifdef CONFIG_IPV6_MROUTE
6765         {
6766                 .procname       = "mc_forwarding",
6767                 .data           = &ipv6_devconf.mc_forwarding,
6768                 .maxlen         = sizeof(int),
6769                 .mode           = 0444,
6770                 .proc_handler   = proc_dointvec,
6771         },
6772 #endif
6773         {
6774                 .procname       = "disable_ipv6",
6775                 .data           = &ipv6_devconf.disable_ipv6,
6776                 .maxlen         = sizeof(int),
6777                 .mode           = 0644,
6778                 .proc_handler   = addrconf_sysctl_disable,
6779         },
6780         {
6781                 .procname       = "accept_dad",
6782                 .data           = &ipv6_devconf.accept_dad,
6783                 .maxlen         = sizeof(int),
6784                 .mode           = 0644,
6785                 .proc_handler   = proc_dointvec,
6786         },
6787         {
6788                 .procname       = "force_tllao",
6789                 .data           = &ipv6_devconf.force_tllao,
6790                 .maxlen         = sizeof(int),
6791                 .mode           = 0644,
6792                 .proc_handler   = proc_dointvec
6793         },
6794         {
6795                 .procname       = "ndisc_notify",
6796                 .data           = &ipv6_devconf.ndisc_notify,
6797                 .maxlen         = sizeof(int),
6798                 .mode           = 0644,
6799                 .proc_handler   = proc_dointvec
6800         },
6801         {
6802                 .procname       = "suppress_frag_ndisc",
6803                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6804                 .maxlen         = sizeof(int),
6805                 .mode           = 0644,
6806                 .proc_handler   = proc_dointvec
6807         },
6808         {
6809                 .procname       = "accept_ra_from_local",
6810                 .data           = &ipv6_devconf.accept_ra_from_local,
6811                 .maxlen         = sizeof(int),
6812                 .mode           = 0644,
6813                 .proc_handler   = proc_dointvec,
6814         },
6815         {
6816                 .procname       = "accept_ra_mtu",
6817                 .data           = &ipv6_devconf.accept_ra_mtu,
6818                 .maxlen         = sizeof(int),
6819                 .mode           = 0644,
6820                 .proc_handler   = proc_dointvec,
6821         },
6822         {
6823                 .procname       = "stable_secret",
6824                 .data           = &ipv6_devconf.stable_secret,
6825                 .maxlen         = IPV6_MAX_STRLEN,
6826                 .mode           = 0600,
6827                 .proc_handler   = addrconf_sysctl_stable_secret,
6828         },
6829         {
6830                 .procname       = "use_oif_addrs_only",
6831                 .data           = &ipv6_devconf.use_oif_addrs_only,
6832                 .maxlen         = sizeof(int),
6833                 .mode           = 0644,
6834                 .proc_handler   = proc_dointvec,
6835         },
6836         {
6837                 .procname       = "ignore_routes_with_linkdown",
6838                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6839                 .maxlen         = sizeof(int),
6840                 .mode           = 0644,
6841                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6842         },
6843         {
6844                 .procname       = "drop_unicast_in_l2_multicast",
6845                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6846                 .maxlen         = sizeof(int),
6847                 .mode           = 0644,
6848                 .proc_handler   = proc_dointvec,
6849         },
6850         {
6851                 .procname       = "drop_unsolicited_na",
6852                 .data           = &ipv6_devconf.drop_unsolicited_na,
6853                 .maxlen         = sizeof(int),
6854                 .mode           = 0644,
6855                 .proc_handler   = proc_dointvec,
6856         },
6857         {
6858                 .procname       = "keep_addr_on_down",
6859                 .data           = &ipv6_devconf.keep_addr_on_down,
6860                 .maxlen         = sizeof(int),
6861                 .mode           = 0644,
6862                 .proc_handler   = proc_dointvec,
6863
6864         },
6865         {
6866                 .procname       = "seg6_enabled",
6867                 .data           = &ipv6_devconf.seg6_enabled,
6868                 .maxlen         = sizeof(int),
6869                 .mode           = 0644,
6870                 .proc_handler   = proc_dointvec,
6871         },
6872 #ifdef CONFIG_IPV6_SEG6_HMAC
6873         {
6874                 .procname       = "seg6_require_hmac",
6875                 .data           = &ipv6_devconf.seg6_require_hmac,
6876                 .maxlen         = sizeof(int),
6877                 .mode           = 0644,
6878                 .proc_handler   = proc_dointvec,
6879         },
6880 #endif
6881         {
6882                 .procname       = "enhanced_dad",
6883                 .data           = &ipv6_devconf.enhanced_dad,
6884                 .maxlen         = sizeof(int),
6885                 .mode           = 0644,
6886                 .proc_handler   = proc_dointvec,
6887         },
6888         {
6889                 .procname               = "addr_gen_mode",
6890                 .data                   = &ipv6_devconf.addr_gen_mode,
6891                 .maxlen                 = sizeof(int),
6892                 .mode                   = 0644,
6893                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6894         },
6895         {
6896                 .procname       = "disable_policy",
6897                 .data           = &ipv6_devconf.disable_policy,
6898                 .maxlen         = sizeof(int),
6899                 .mode           = 0644,
6900                 .proc_handler   = addrconf_sysctl_disable_policy,
6901         },
6902         {
6903                 .procname       = "ndisc_tclass",
6904                 .data           = &ipv6_devconf.ndisc_tclass,
6905                 .maxlen         = sizeof(int),
6906                 .mode           = 0644,
6907                 .proc_handler   = proc_dointvec_minmax,
6908                 .extra1         = (void *)SYSCTL_ZERO,
6909                 .extra2         = (void *)&two_five_five,
6910         },
6911         {
6912                 .procname       = "rpl_seg_enabled",
6913                 .data           = &ipv6_devconf.rpl_seg_enabled,
6914                 .maxlen         = sizeof(int),
6915                 .mode           = 0644,
6916                 .proc_handler   = proc_dointvec,
6917         },
6918         {
6919                 /* sentinel */
6920         }
6921 };
6922
6923 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6924                 struct inet6_dev *idev, struct ipv6_devconf *p)
6925 {
6926         int i, ifindex;
6927         struct ctl_table *table;
6928         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6929
6930         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6931         if (!table)
6932                 goto out;
6933
6934         for (i = 0; table[i].data; i++) {
6935                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6936                 /* If one of these is already set, then it is not safe to
6937                  * overwrite either of them: this makes proc_dointvec_minmax
6938                  * usable.
6939                  */
6940                 if (!table[i].extra1 && !table[i].extra2) {
6941                         table[i].extra1 = idev; /* embedded; no ref */
6942                         table[i].extra2 = net;
6943                 }
6944         }
6945
6946         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6947
6948         p->sysctl_header = register_net_sysctl(net, path, table);
6949         if (!p->sysctl_header)
6950                 goto free;
6951
6952         if (!strcmp(dev_name, "all"))
6953                 ifindex = NETCONFA_IFINDEX_ALL;
6954         else if (!strcmp(dev_name, "default"))
6955                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6956         else
6957                 ifindex = idev->dev->ifindex;
6958         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6959                                      ifindex, p);
6960         return 0;
6961
6962 free:
6963         kfree(table);
6964 out:
6965         return -ENOBUFS;
6966 }
6967
6968 static void __addrconf_sysctl_unregister(struct net *net,
6969                                          struct ipv6_devconf *p, int ifindex)
6970 {
6971         struct ctl_table *table;
6972
6973         if (!p->sysctl_header)
6974                 return;
6975
6976         table = p->sysctl_header->ctl_table_arg;
6977         unregister_net_sysctl_table(p->sysctl_header);
6978         p->sysctl_header = NULL;
6979         kfree(table);
6980
6981         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6982 }
6983
6984 static int addrconf_sysctl_register(struct inet6_dev *idev)
6985 {
6986         int err;
6987
6988         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6989                 return -EINVAL;
6990
6991         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6992                                     &ndisc_ifinfo_sysctl_change);
6993         if (err)
6994                 return err;
6995         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6996                                          idev, &idev->cnf);
6997         if (err)
6998                 neigh_sysctl_unregister(idev->nd_parms);
6999
7000         return err;
7001 }
7002
7003 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7004 {
7005         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7006                                      idev->dev->ifindex);
7007         neigh_sysctl_unregister(idev->nd_parms);
7008 }
7009
7010
7011 #endif
7012
7013 static int __net_init addrconf_init_net(struct net *net)
7014 {
7015         int err = -ENOMEM;
7016         struct ipv6_devconf *all, *dflt;
7017
7018         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7019         if (!all)
7020                 goto err_alloc_all;
7021
7022         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7023         if (!dflt)
7024                 goto err_alloc_dflt;
7025
7026         if (IS_ENABLED(CONFIG_SYSCTL) &&
7027             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
7028                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
7029                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
7030         }
7031
7032         /* these will be inherited by all namespaces */
7033         dflt->autoconf = ipv6_defaults.autoconf;
7034         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7035
7036         dflt->stable_secret.initialized = false;
7037         all->stable_secret.initialized = false;
7038
7039         net->ipv6.devconf_all = all;
7040         net->ipv6.devconf_dflt = dflt;
7041
7042 #ifdef CONFIG_SYSCTL
7043         err = __addrconf_sysctl_register(net, "all", NULL, all);
7044         if (err < 0)
7045                 goto err_reg_all;
7046
7047         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7048         if (err < 0)
7049                 goto err_reg_dflt;
7050 #endif
7051         return 0;
7052
7053 #ifdef CONFIG_SYSCTL
7054 err_reg_dflt:
7055         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7056 err_reg_all:
7057         kfree(dflt);
7058 #endif
7059 err_alloc_dflt:
7060         kfree(all);
7061 err_alloc_all:
7062         return err;
7063 }
7064
7065 static void __net_exit addrconf_exit_net(struct net *net)
7066 {
7067 #ifdef CONFIG_SYSCTL
7068         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7069                                      NETCONFA_IFINDEX_DEFAULT);
7070         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7071                                      NETCONFA_IFINDEX_ALL);
7072 #endif
7073         kfree(net->ipv6.devconf_dflt);
7074         kfree(net->ipv6.devconf_all);
7075 }
7076
7077 static struct pernet_operations addrconf_ops = {
7078         .init = addrconf_init_net,
7079         .exit = addrconf_exit_net,
7080 };
7081
7082 static struct rtnl_af_ops inet6_ops __read_mostly = {
7083         .family           = AF_INET6,
7084         .fill_link_af     = inet6_fill_link_af,
7085         .get_link_af_size = inet6_get_link_af_size,
7086         .validate_link_af = inet6_validate_link_af,
7087         .set_link_af      = inet6_set_link_af,
7088 };
7089
7090 /*
7091  *      Init / cleanup code
7092  */
7093
7094 int __init addrconf_init(void)
7095 {
7096         struct inet6_dev *idev;
7097         int i, err;
7098
7099         err = ipv6_addr_label_init();
7100         if (err < 0) {
7101                 pr_crit("%s: cannot initialize default policy table: %d\n",
7102                         __func__, err);
7103                 goto out;
7104         }
7105
7106         err = register_pernet_subsys(&addrconf_ops);
7107         if (err < 0)
7108                 goto out_addrlabel;
7109
7110         addrconf_wq = create_workqueue("ipv6_addrconf");
7111         if (!addrconf_wq) {
7112                 err = -ENOMEM;
7113                 goto out_nowq;
7114         }
7115
7116         /* The addrconf netdev notifier requires that loopback_dev
7117          * has it's ipv6 private information allocated and setup
7118          * before it can bring up and give link-local addresses
7119          * to other devices which are up.
7120          *
7121          * Unfortunately, loopback_dev is not necessarily the first
7122          * entry in the global dev_base list of net devices.  In fact,
7123          * it is likely to be the very last entry on that list.
7124          * So this causes the notifier registry below to try and
7125          * give link-local addresses to all devices besides loopback_dev
7126          * first, then loopback_dev, which cases all the non-loopback_dev
7127          * devices to fail to get a link-local address.
7128          *
7129          * So, as a temporary fix, allocate the ipv6 structure for
7130          * loopback_dev first by hand.
7131          * Longer term, all of the dependencies ipv6 has upon the loopback
7132          * device and it being up should be removed.
7133          */
7134         rtnl_lock();
7135         idev = ipv6_add_dev(init_net.loopback_dev);
7136         rtnl_unlock();
7137         if (IS_ERR(idev)) {
7138                 err = PTR_ERR(idev);
7139                 goto errlo;
7140         }
7141
7142         ip6_route_init_special_entries();
7143
7144         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7145                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7146
7147         register_netdevice_notifier(&ipv6_dev_notf);
7148
7149         addrconf_verify();
7150
7151         rtnl_af_register(&inet6_ops);
7152
7153         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7154                                    NULL, inet6_dump_ifinfo, 0);
7155         if (err < 0)
7156                 goto errout;
7157
7158         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7159                                    inet6_rtm_newaddr, NULL, 0);
7160         if (err < 0)
7161                 goto errout;
7162         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7163                                    inet6_rtm_deladdr, NULL, 0);
7164         if (err < 0)
7165                 goto errout;
7166         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7167                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7168                                    RTNL_FLAG_DOIT_UNLOCKED);
7169         if (err < 0)
7170                 goto errout;
7171         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7172                                    NULL, inet6_dump_ifmcaddr, 0);
7173         if (err < 0)
7174                 goto errout;
7175         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7176                                    NULL, inet6_dump_ifacaddr, 0);
7177         if (err < 0)
7178                 goto errout;
7179         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7180                                    inet6_netconf_get_devconf,
7181                                    inet6_netconf_dump_devconf,
7182                                    RTNL_FLAG_DOIT_UNLOCKED);
7183         if (err < 0)
7184                 goto errout;
7185         err = ipv6_addr_label_rtnl_register();
7186         if (err < 0)
7187                 goto errout;
7188
7189         return 0;
7190 errout:
7191         rtnl_unregister_all(PF_INET6);
7192         rtnl_af_unregister(&inet6_ops);
7193         unregister_netdevice_notifier(&ipv6_dev_notf);
7194 errlo:
7195         destroy_workqueue(addrconf_wq);
7196 out_nowq:
7197         unregister_pernet_subsys(&addrconf_ops);
7198 out_addrlabel:
7199         ipv6_addr_label_cleanup();
7200 out:
7201         return err;
7202 }
7203
7204 void addrconf_cleanup(void)
7205 {
7206         struct net_device *dev;
7207         int i;
7208
7209         unregister_netdevice_notifier(&ipv6_dev_notf);
7210         unregister_pernet_subsys(&addrconf_ops);
7211         ipv6_addr_label_cleanup();
7212
7213         rtnl_af_unregister(&inet6_ops);
7214
7215         rtnl_lock();
7216
7217         /* clean dev list */
7218         for_each_netdev(&init_net, dev) {
7219                 if (__in6_dev_get(dev) == NULL)
7220                         continue;
7221                 addrconf_ifdown(dev, 1);
7222         }
7223         addrconf_ifdown(init_net.loopback_dev, 2);
7224
7225         /*
7226          *      Check hash table.
7227          */
7228         spin_lock_bh(&addrconf_hash_lock);
7229         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7230                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7231         spin_unlock_bh(&addrconf_hash_lock);
7232         cancel_delayed_work(&addr_chk_work);
7233         rtnl_unlock();
7234
7235         destroy_workqueue(addrconf_wq);
7236 }