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