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