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  * @dev: used to find the L3 domain of interest
2001  *
2002  * The caller should be protected by RCU, or RTNL.
2003  */
2004 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2005                                  struct net_device *dev)
2006 {
2007         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2008                                          IFA_F_TENTATIVE);
2009 }
2010 EXPORT_SYMBOL(ipv6_dev_find);
2011
2012 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2013                                      struct net_device *dev, int strict)
2014 {
2015         unsigned int hash = inet6_addr_hash(net, addr);
2016         struct inet6_ifaddr *ifp, *result = NULL;
2017
2018         rcu_read_lock();
2019         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2020                 if (!net_eq(dev_net(ifp->idev->dev), net))
2021                         continue;
2022                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2023                         if (!dev || ifp->idev->dev == dev ||
2024                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2025                                 result = ifp;
2026                                 in6_ifa_hold(ifp);
2027                                 break;
2028                         }
2029                 }
2030         }
2031         rcu_read_unlock();
2032
2033         return result;
2034 }
2035
2036 /* Gets referenced address, destroys ifaddr */
2037
2038 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2039 {
2040         if (dad_failed)
2041                 ifp->flags |= IFA_F_DADFAILED;
2042
2043         if (ifp->flags&IFA_F_TEMPORARY) {
2044                 struct inet6_ifaddr *ifpub;
2045                 spin_lock_bh(&ifp->lock);
2046                 ifpub = ifp->ifpub;
2047                 if (ifpub) {
2048                         in6_ifa_hold(ifpub);
2049                         spin_unlock_bh(&ifp->lock);
2050                         ipv6_create_tempaddr(ifpub, true);
2051                         in6_ifa_put(ifpub);
2052                 } else {
2053                         spin_unlock_bh(&ifp->lock);
2054                 }
2055                 ipv6_del_addr(ifp);
2056         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2057                 spin_lock_bh(&ifp->lock);
2058                 addrconf_del_dad_work(ifp);
2059                 ifp->flags |= IFA_F_TENTATIVE;
2060                 if (dad_failed)
2061                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2062                 spin_unlock_bh(&ifp->lock);
2063                 if (dad_failed)
2064                         ipv6_ifa_notify(0, ifp);
2065                 in6_ifa_put(ifp);
2066         } else {
2067                 ipv6_del_addr(ifp);
2068         }
2069 }
2070
2071 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2072 {
2073         int err = -ENOENT;
2074
2075         spin_lock_bh(&ifp->lock);
2076         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2077                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2078                 err = 0;
2079         }
2080         spin_unlock_bh(&ifp->lock);
2081
2082         return err;
2083 }
2084
2085 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2086 {
2087         struct inet6_dev *idev = ifp->idev;
2088         struct net *net = dev_net(ifp->idev->dev);
2089
2090         if (addrconf_dad_end(ifp)) {
2091                 in6_ifa_put(ifp);
2092                 return;
2093         }
2094
2095         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2096                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2097
2098         spin_lock_bh(&ifp->lock);
2099
2100         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2101                 struct in6_addr new_addr;
2102                 struct inet6_ifaddr *ifp2;
2103                 int retries = ifp->stable_privacy_retry + 1;
2104                 struct ifa6_config cfg = {
2105                         .pfx = &new_addr,
2106                         .plen = ifp->prefix_len,
2107                         .ifa_flags = ifp->flags,
2108                         .valid_lft = ifp->valid_lft,
2109                         .preferred_lft = ifp->prefered_lft,
2110                         .scope = ifp->scope,
2111                 };
2112
2113                 if (retries > net->ipv6.sysctl.idgen_retries) {
2114                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2115                                              ifp->idev->dev->name);
2116                         goto errdad;
2117                 }
2118
2119                 new_addr = ifp->addr;
2120                 if (ipv6_generate_stable_address(&new_addr, retries,
2121                                                  idev))
2122                         goto errdad;
2123
2124                 spin_unlock_bh(&ifp->lock);
2125
2126                 if (idev->cnf.max_addresses &&
2127                     ipv6_count_addresses(idev) >=
2128                     idev->cnf.max_addresses)
2129                         goto lock_errdad;
2130
2131                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2132                                      ifp->idev->dev->name);
2133
2134                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2135                 if (IS_ERR(ifp2))
2136                         goto lock_errdad;
2137
2138                 spin_lock_bh(&ifp2->lock);
2139                 ifp2->stable_privacy_retry = retries;
2140                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2141                 spin_unlock_bh(&ifp2->lock);
2142
2143                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2144                 in6_ifa_put(ifp2);
2145 lock_errdad:
2146                 spin_lock_bh(&ifp->lock);
2147         }
2148
2149 errdad:
2150         /* transition from _POSTDAD to _ERRDAD */
2151         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2152         spin_unlock_bh(&ifp->lock);
2153
2154         addrconf_mod_dad_work(ifp, 0);
2155         in6_ifa_put(ifp);
2156 }
2157
2158 /* Join to solicited addr multicast group.
2159  * caller must hold RTNL */
2160 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2161 {
2162         struct in6_addr maddr;
2163
2164         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2165                 return;
2166
2167         addrconf_addr_solict_mult(addr, &maddr);
2168         ipv6_dev_mc_inc(dev, &maddr);
2169 }
2170
2171 /* caller must hold RTNL */
2172 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2173 {
2174         struct in6_addr maddr;
2175
2176         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2177                 return;
2178
2179         addrconf_addr_solict_mult(addr, &maddr);
2180         __ipv6_dev_mc_dec(idev, &maddr);
2181 }
2182
2183 /* caller must hold RTNL */
2184 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2185 {
2186         struct in6_addr addr;
2187
2188         if (ifp->prefix_len >= 127) /* RFC 6164 */
2189                 return;
2190         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2191         if (ipv6_addr_any(&addr))
2192                 return;
2193         __ipv6_dev_ac_inc(ifp->idev, &addr);
2194 }
2195
2196 /* caller must hold RTNL */
2197 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2198 {
2199         struct in6_addr addr;
2200
2201         if (ifp->prefix_len >= 127) /* RFC 6164 */
2202                 return;
2203         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2204         if (ipv6_addr_any(&addr))
2205                 return;
2206         __ipv6_dev_ac_dec(ifp->idev, &addr);
2207 }
2208
2209 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2210 {
2211         switch (dev->addr_len) {
2212         case ETH_ALEN:
2213                 memcpy(eui, dev->dev_addr, 3);
2214                 eui[3] = 0xFF;
2215                 eui[4] = 0xFE;
2216                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2217                 break;
2218         case EUI64_ADDR_LEN:
2219                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2220                 eui[0] ^= 2;
2221                 break;
2222         default:
2223                 return -1;
2224         }
2225
2226         return 0;
2227 }
2228
2229 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2230 {
2231         union fwnet_hwaddr *ha;
2232
2233         if (dev->addr_len != FWNET_ALEN)
2234                 return -1;
2235
2236         ha = (union fwnet_hwaddr *)dev->dev_addr;
2237
2238         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2239         eui[0] ^= 2;
2240         return 0;
2241 }
2242
2243 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2244 {
2245         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2246         if (dev->addr_len != ARCNET_ALEN)
2247                 return -1;
2248         memset(eui, 0, 7);
2249         eui[7] = *(u8 *)dev->dev_addr;
2250         return 0;
2251 }
2252
2253 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2254 {
2255         if (dev->addr_len != INFINIBAND_ALEN)
2256                 return -1;
2257         memcpy(eui, dev->dev_addr + 12, 8);
2258         eui[0] |= 2;
2259         return 0;
2260 }
2261
2262 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2263 {
2264         if (addr == 0)
2265                 return -1;
2266         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2267                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2268                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2269                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2270                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2271                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2272         eui[1] = 0;
2273         eui[2] = 0x5E;
2274         eui[3] = 0xFE;
2275         memcpy(eui + 4, &addr, 4);
2276         return 0;
2277 }
2278
2279 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2280 {
2281         if (dev->priv_flags & IFF_ISATAP)
2282                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2283         return -1;
2284 }
2285
2286 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2287 {
2288         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2289 }
2290
2291 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2292 {
2293         memcpy(eui, dev->perm_addr, 3);
2294         memcpy(eui + 5, dev->perm_addr + 3, 3);
2295         eui[3] = 0xFF;
2296         eui[4] = 0xFE;
2297         eui[0] ^= 2;
2298         return 0;
2299 }
2300
2301 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2302 {
2303         switch (dev->type) {
2304         case ARPHRD_ETHER:
2305         case ARPHRD_FDDI:
2306                 return addrconf_ifid_eui48(eui, dev);
2307         case ARPHRD_ARCNET:
2308                 return addrconf_ifid_arcnet(eui, dev);
2309         case ARPHRD_INFINIBAND:
2310                 return addrconf_ifid_infiniband(eui, dev);
2311         case ARPHRD_SIT:
2312                 return addrconf_ifid_sit(eui, dev);
2313         case ARPHRD_IPGRE:
2314         case ARPHRD_TUNNEL:
2315                 return addrconf_ifid_gre(eui, dev);
2316         case ARPHRD_6LOWPAN:
2317                 return addrconf_ifid_6lowpan(eui, dev);
2318         case ARPHRD_IEEE1394:
2319                 return addrconf_ifid_ieee1394(eui, dev);
2320         case ARPHRD_TUNNEL6:
2321         case ARPHRD_IP6GRE:
2322         case ARPHRD_RAWIP:
2323                 return addrconf_ifid_ip6tnl(eui, dev);
2324         }
2325         return -1;
2326 }
2327
2328 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2329 {
2330         int err = -1;
2331         struct inet6_ifaddr *ifp;
2332
2333         read_lock_bh(&idev->lock);
2334         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2335                 if (ifp->scope > IFA_LINK)
2336                         break;
2337                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2338                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2339                         err = 0;
2340                         break;
2341                 }
2342         }
2343         read_unlock_bh(&idev->lock);
2344         return err;
2345 }
2346
2347 /* Generation of a randomized Interface Identifier
2348  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2349  */
2350
2351 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2352 {
2353 regen:
2354         get_random_bytes(&addr->s6_addr[8], 8);
2355
2356         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2357          * check if generated address is not inappropriate:
2358          *
2359          * - Reserved IPv6 Interface Identifers
2360          * - XXX: already assigned to an address on the device
2361          */
2362
2363         /* Subnet-router anycast: 0000:0000:0000:0000 */
2364         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2365                 goto regen;
2366
2367         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2368          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2369          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2370          */
2371         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2372             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2373                 goto regen;
2374
2375         /* Reserved subnet anycast addresses */
2376         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2377             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2378                 goto regen;
2379 }
2380
2381 /*
2382  *      Add prefix route.
2383  */
2384
2385 static void
2386 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2387                       struct net_device *dev, unsigned long expires,
2388                       u32 flags, gfp_t gfp_flags)
2389 {
2390         struct fib6_config cfg = {
2391                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2392                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2393                 .fc_ifindex = dev->ifindex,
2394                 .fc_expires = expires,
2395                 .fc_dst_len = plen,
2396                 .fc_flags = RTF_UP | flags,
2397                 .fc_nlinfo.nl_net = dev_net(dev),
2398                 .fc_protocol = RTPROT_KERNEL,
2399                 .fc_type = RTN_UNICAST,
2400         };
2401
2402         cfg.fc_dst = *pfx;
2403
2404         /* Prevent useless cloning on PtP SIT.
2405            This thing is done here expecting that the whole
2406            class of non-broadcast devices need not cloning.
2407          */
2408 #if IS_ENABLED(CONFIG_IPV6_SIT)
2409         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2410                 cfg.fc_flags |= RTF_NONEXTHOP;
2411 #endif
2412
2413         ip6_route_add(&cfg, gfp_flags, NULL);
2414 }
2415
2416
2417 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2418                                                   int plen,
2419                                                   const struct net_device *dev,
2420                                                   u32 flags, u32 noflags,
2421                                                   bool no_gw)
2422 {
2423         struct fib6_node *fn;
2424         struct fib6_info *rt = NULL;
2425         struct fib6_table *table;
2426         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2427
2428         table = fib6_get_table(dev_net(dev), tb_id);
2429         if (!table)
2430                 return NULL;
2431
2432         rcu_read_lock();
2433         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2434         if (!fn)
2435                 goto out;
2436
2437         for_each_fib6_node_rt_rcu(fn) {
2438                 /* prefix routes only use builtin fib6_nh */
2439                 if (rt->nh)
2440                         continue;
2441
2442                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2443                         continue;
2444                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2445                         continue;
2446                 if ((rt->fib6_flags & flags) != flags)
2447                         continue;
2448                 if ((rt->fib6_flags & noflags) != 0)
2449                         continue;
2450                 if (!fib6_info_hold_safe(rt))
2451                         continue;
2452                 break;
2453         }
2454 out:
2455         rcu_read_unlock();
2456         return rt;
2457 }
2458
2459
2460 /* Create "default" multicast route to the interface */
2461
2462 static void addrconf_add_mroute(struct net_device *dev)
2463 {
2464         struct fib6_config cfg = {
2465                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2466                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2467                 .fc_ifindex = dev->ifindex,
2468                 .fc_dst_len = 8,
2469                 .fc_flags = RTF_UP,
2470                 .fc_type = RTN_UNICAST,
2471                 .fc_nlinfo.nl_net = dev_net(dev),
2472         };
2473
2474         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2475
2476         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2477 }
2478
2479 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2480 {
2481         struct inet6_dev *idev;
2482
2483         ASSERT_RTNL();
2484
2485         idev = ipv6_find_idev(dev);
2486         if (IS_ERR(idev))
2487                 return idev;
2488
2489         if (idev->cnf.disable_ipv6)
2490                 return ERR_PTR(-EACCES);
2491
2492         /* Add default multicast route */
2493         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2494                 addrconf_add_mroute(dev);
2495
2496         return idev;
2497 }
2498
2499 static void manage_tempaddrs(struct inet6_dev *idev,
2500                              struct inet6_ifaddr *ifp,
2501                              __u32 valid_lft, __u32 prefered_lft,
2502                              bool create, unsigned long now)
2503 {
2504         u32 flags;
2505         struct inet6_ifaddr *ift;
2506
2507         read_lock_bh(&idev->lock);
2508         /* update all temporary addresses in the list */
2509         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2510                 int age, max_valid, max_prefered;
2511
2512                 if (ifp != ift->ifpub)
2513                         continue;
2514
2515                 /* RFC 4941 section 3.3:
2516                  * If a received option will extend the lifetime of a public
2517                  * address, the lifetimes of temporary addresses should
2518                  * be extended, subject to the overall constraint that no
2519                  * temporary addresses should ever remain "valid" or "preferred"
2520                  * for a time longer than (TEMP_VALID_LIFETIME) or
2521                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2522                  */
2523                 age = (now - ift->cstamp) / HZ;
2524                 max_valid = idev->cnf.temp_valid_lft - age;
2525                 if (max_valid < 0)
2526                         max_valid = 0;
2527
2528                 max_prefered = idev->cnf.temp_prefered_lft -
2529                                idev->desync_factor - age;
2530                 if (max_prefered < 0)
2531                         max_prefered = 0;
2532
2533                 if (valid_lft > max_valid)
2534                         valid_lft = max_valid;
2535
2536                 if (prefered_lft > max_prefered)
2537                         prefered_lft = max_prefered;
2538
2539                 spin_lock(&ift->lock);
2540                 flags = ift->flags;
2541                 ift->valid_lft = valid_lft;
2542                 ift->prefered_lft = prefered_lft;
2543                 ift->tstamp = now;
2544                 if (prefered_lft > 0)
2545                         ift->flags &= ~IFA_F_DEPRECATED;
2546
2547                 spin_unlock(&ift->lock);
2548                 if (!(flags&IFA_F_TENTATIVE))
2549                         ipv6_ifa_notify(0, ift);
2550         }
2551
2552         if ((create || list_empty(&idev->tempaddr_list)) &&
2553             idev->cnf.use_tempaddr > 0) {
2554                 /* When a new public address is created as described
2555                  * in [ADDRCONF], also create a new temporary address.
2556                  * Also create a temporary address if it's enabled but
2557                  * no temporary address currently exists.
2558                  */
2559                 read_unlock_bh(&idev->lock);
2560                 ipv6_create_tempaddr(ifp, false);
2561         } else {
2562                 read_unlock_bh(&idev->lock);
2563         }
2564 }
2565
2566 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2567 {
2568         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2569                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2570 }
2571
2572 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2573                                  const struct prefix_info *pinfo,
2574                                  struct inet6_dev *in6_dev,
2575                                  const struct in6_addr *addr, int addr_type,
2576                                  u32 addr_flags, bool sllao, bool tokenized,
2577                                  __u32 valid_lft, u32 prefered_lft)
2578 {
2579         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2580         int create = 0;
2581
2582         if (!ifp && valid_lft) {
2583                 int max_addresses = in6_dev->cnf.max_addresses;
2584                 struct ifa6_config cfg = {
2585                         .pfx = addr,
2586                         .plen = pinfo->prefix_len,
2587                         .ifa_flags = addr_flags,
2588                         .valid_lft = valid_lft,
2589                         .preferred_lft = prefered_lft,
2590                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2591                 };
2592
2593 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2594                 if ((net->ipv6.devconf_all->optimistic_dad ||
2595                      in6_dev->cnf.optimistic_dad) &&
2596                     !net->ipv6.devconf_all->forwarding && sllao)
2597                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2598 #endif
2599
2600                 /* Do not allow to create too much of autoconfigured
2601                  * addresses; this would be too easy way to crash kernel.
2602                  */
2603                 if (!max_addresses ||
2604                     ipv6_count_addresses(in6_dev) < max_addresses)
2605                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2606
2607                 if (IS_ERR_OR_NULL(ifp))
2608                         return -1;
2609
2610                 create = 1;
2611                 spin_lock_bh(&ifp->lock);
2612                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2613                 ifp->cstamp = jiffies;
2614                 ifp->tokenized = tokenized;
2615                 spin_unlock_bh(&ifp->lock);
2616                 addrconf_dad_start(ifp);
2617         }
2618
2619         if (ifp) {
2620                 u32 flags;
2621                 unsigned long now;
2622                 u32 stored_lft;
2623
2624                 /* Update lifetime (RFC4862 5.5.3 e)
2625                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2626                  * improve the reaction of SLAAC to renumbering events
2627                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2628                  */
2629                 spin_lock_bh(&ifp->lock);
2630                 now = jiffies;
2631                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2632                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2633                 else
2634                         stored_lft = 0;
2635
2636                 if (!create && stored_lft) {
2637                         ifp->valid_lft = valid_lft;
2638                         ifp->prefered_lft = prefered_lft;
2639                         ifp->tstamp = now;
2640                         flags = ifp->flags;
2641                         ifp->flags &= ~IFA_F_DEPRECATED;
2642                         spin_unlock_bh(&ifp->lock);
2643
2644                         if (!(flags&IFA_F_TENTATIVE))
2645                                 ipv6_ifa_notify(0, ifp);
2646                 } else
2647                         spin_unlock_bh(&ifp->lock);
2648
2649                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2650                                  create, now);
2651
2652                 in6_ifa_put(ifp);
2653                 addrconf_verify();
2654         }
2655
2656         return 0;
2657 }
2658 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2659
2660 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2661 {
2662         struct prefix_info *pinfo;
2663         __u32 valid_lft;
2664         __u32 prefered_lft;
2665         int addr_type, err;
2666         u32 addr_flags = 0;
2667         struct inet6_dev *in6_dev;
2668         struct net *net = dev_net(dev);
2669
2670         pinfo = (struct prefix_info *) opt;
2671
2672         if (len < sizeof(struct prefix_info)) {
2673                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2674                 return;
2675         }
2676
2677         /*
2678          *      Validation checks ([ADDRCONF], page 19)
2679          */
2680
2681         addr_type = ipv6_addr_type(&pinfo->prefix);
2682
2683         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2684                 return;
2685
2686         valid_lft = ntohl(pinfo->valid);
2687         prefered_lft = ntohl(pinfo->prefered);
2688
2689         if (prefered_lft > valid_lft) {
2690                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2691                 return;
2692         }
2693
2694         in6_dev = in6_dev_get(dev);
2695
2696         if (!in6_dev) {
2697                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2698                                     dev->name);
2699                 return;
2700         }
2701
2702         /*
2703          *      Two things going on here:
2704          *      1) Add routes for on-link prefixes
2705          *      2) Configure prefixes with the auto flag set
2706          */
2707
2708         if (pinfo->onlink) {
2709                 struct fib6_info *rt;
2710                 unsigned long rt_expires;
2711
2712                 /* Avoid arithmetic overflow. Really, we could
2713                  * save rt_expires in seconds, likely valid_lft,
2714                  * but it would require division in fib gc, that it
2715                  * not good.
2716                  */
2717                 if (HZ > USER_HZ)
2718                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2719                 else
2720                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2721
2722                 if (addrconf_finite_timeout(rt_expires))
2723                         rt_expires *= HZ;
2724
2725                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2726                                                pinfo->prefix_len,
2727                                                dev,
2728                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2729                                                RTF_DEFAULT, true);
2730
2731                 if (rt) {
2732                         /* Autoconf prefix route */
2733                         if (valid_lft == 0) {
2734                                 ip6_del_rt(net, rt, false);
2735                                 rt = NULL;
2736                         } else if (addrconf_finite_timeout(rt_expires)) {
2737                                 /* not infinity */
2738                                 fib6_set_expires(rt, jiffies + rt_expires);
2739                         } else {
2740                                 fib6_clean_expires(rt);
2741                         }
2742                 } else if (valid_lft) {
2743                         clock_t expires = 0;
2744                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2745                         if (addrconf_finite_timeout(rt_expires)) {
2746                                 /* not infinity */
2747                                 flags |= RTF_EXPIRES;
2748                                 expires = jiffies_to_clock_t(rt_expires);
2749                         }
2750                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2751                                               0, dev, expires, flags,
2752                                               GFP_ATOMIC);
2753                 }
2754                 fib6_info_release(rt);
2755         }
2756
2757         /* Try to figure out our local address for this prefix */
2758
2759         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2760                 struct in6_addr addr;
2761                 bool tokenized = false, dev_addr_generated = false;
2762
2763                 if (pinfo->prefix_len == 64) {
2764                         memcpy(&addr, &pinfo->prefix, 8);
2765
2766                         if (!ipv6_addr_any(&in6_dev->token)) {
2767                                 read_lock_bh(&in6_dev->lock);
2768                                 memcpy(addr.s6_addr + 8,
2769                                        in6_dev->token.s6_addr + 8, 8);
2770                                 read_unlock_bh(&in6_dev->lock);
2771                                 tokenized = true;
2772                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2773                                    !ipv6_generate_stable_address(&addr, 0,
2774                                                                  in6_dev)) {
2775                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2776                                 goto ok;
2777                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2778                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2779                                 goto put;
2780                         } else {
2781                                 dev_addr_generated = true;
2782                         }
2783                         goto ok;
2784                 }
2785                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2786                                     pinfo->prefix_len);
2787                 goto put;
2788
2789 ok:
2790                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2791                                                    &addr, addr_type,
2792                                                    addr_flags, sllao,
2793                                                    tokenized, valid_lft,
2794                                                    prefered_lft);
2795                 if (err)
2796                         goto put;
2797
2798                 /* Ignore error case here because previous prefix add addr was
2799                  * successful which will be notified.
2800                  */
2801                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2802                                               addr_type, addr_flags, sllao,
2803                                               tokenized, valid_lft,
2804                                               prefered_lft,
2805                                               dev_addr_generated);
2806         }
2807         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2808 put:
2809         in6_dev_put(in6_dev);
2810 }
2811
2812 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2813                 struct in6_ifreq *ireq)
2814 {
2815         struct ip_tunnel_parm p = { };
2816         int err;
2817
2818         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2819                 return -EADDRNOTAVAIL;
2820
2821         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2822         p.iph.version = 4;
2823         p.iph.ihl = 5;
2824         p.iph.protocol = IPPROTO_IPV6;
2825         p.iph.ttl = 64;
2826
2827         if (!dev->netdev_ops->ndo_tunnel_ctl)
2828                 return -EOPNOTSUPP;
2829         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2830         if (err)
2831                 return err;
2832
2833         dev = __dev_get_by_name(net, p.name);
2834         if (!dev)
2835                 return -ENOBUFS;
2836         return dev_open(dev, NULL);
2837 }
2838
2839 /*
2840  *      Set destination address.
2841  *      Special case for SIT interfaces where we create a new "virtual"
2842  *      device.
2843  */
2844 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2845 {
2846         struct net_device *dev;
2847         struct in6_ifreq ireq;
2848         int err = -ENODEV;
2849
2850         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2851                 return -ENODEV;
2852         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2853                 return -EFAULT;
2854
2855         rtnl_lock();
2856         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2857         if (dev && dev->type == ARPHRD_SIT)
2858                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2859         rtnl_unlock();
2860         return err;
2861 }
2862
2863 static int ipv6_mc_config(struct sock *sk, bool join,
2864                           const struct in6_addr *addr, int ifindex)
2865 {
2866         int ret;
2867
2868         ASSERT_RTNL();
2869
2870         lock_sock(sk);
2871         if (join)
2872                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2873         else
2874                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2875         release_sock(sk);
2876
2877         return ret;
2878 }
2879
2880 /*
2881  *      Manual configuration of address on an interface
2882  */
2883 static int inet6_addr_add(struct net *net, int ifindex,
2884                           struct ifa6_config *cfg,
2885                           struct netlink_ext_ack *extack)
2886 {
2887         struct inet6_ifaddr *ifp;
2888         struct inet6_dev *idev;
2889         struct net_device *dev;
2890         unsigned long timeout;
2891         clock_t expires;
2892         u32 flags;
2893
2894         ASSERT_RTNL();
2895
2896         if (cfg->plen > 128)
2897                 return -EINVAL;
2898
2899         /* check the lifetime */
2900         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2901                 return -EINVAL;
2902
2903         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2904                 return -EINVAL;
2905
2906         dev = __dev_get_by_index(net, ifindex);
2907         if (!dev)
2908                 return -ENODEV;
2909
2910         idev = addrconf_add_dev(dev);
2911         if (IS_ERR(idev))
2912                 return PTR_ERR(idev);
2913
2914         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2915                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2916                                          true, cfg->pfx, ifindex);
2917
2918                 if (ret < 0)
2919                         return ret;
2920         }
2921
2922         cfg->scope = ipv6_addr_scope(cfg->pfx);
2923
2924         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2925         if (addrconf_finite_timeout(timeout)) {
2926                 expires = jiffies_to_clock_t(timeout * HZ);
2927                 cfg->valid_lft = timeout;
2928                 flags = RTF_EXPIRES;
2929         } else {
2930                 expires = 0;
2931                 flags = 0;
2932                 cfg->ifa_flags |= IFA_F_PERMANENT;
2933         }
2934
2935         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2936         if (addrconf_finite_timeout(timeout)) {
2937                 if (timeout == 0)
2938                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2939                 cfg->preferred_lft = timeout;
2940         }
2941
2942         ifp = ipv6_add_addr(idev, cfg, true, extack);
2943         if (!IS_ERR(ifp)) {
2944                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2945                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2946                                               ifp->rt_priority, dev, expires,
2947                                               flags, GFP_KERNEL);
2948                 }
2949
2950                 /* Send a netlink notification if DAD is enabled and
2951                  * optimistic flag is not set
2952                  */
2953                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2954                         ipv6_ifa_notify(0, ifp);
2955                 /*
2956                  * Note that section 3.1 of RFC 4429 indicates
2957                  * that the Optimistic flag should not be set for
2958                  * manually configured addresses
2959                  */
2960                 addrconf_dad_start(ifp);
2961                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2962                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2963                                          cfg->preferred_lft, true, jiffies);
2964                 in6_ifa_put(ifp);
2965                 addrconf_verify_rtnl();
2966                 return 0;
2967         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2968                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2969                                cfg->pfx, ifindex);
2970         }
2971
2972         return PTR_ERR(ifp);
2973 }
2974
2975 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2976                           const struct in6_addr *pfx, unsigned int plen)
2977 {
2978         struct inet6_ifaddr *ifp;
2979         struct inet6_dev *idev;
2980         struct net_device *dev;
2981
2982         if (plen > 128)
2983                 return -EINVAL;
2984
2985         dev = __dev_get_by_index(net, ifindex);
2986         if (!dev)
2987                 return -ENODEV;
2988
2989         idev = __in6_dev_get(dev);
2990         if (!idev)
2991                 return -ENXIO;
2992
2993         read_lock_bh(&idev->lock);
2994         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2995                 if (ifp->prefix_len == plen &&
2996                     ipv6_addr_equal(pfx, &ifp->addr)) {
2997                         in6_ifa_hold(ifp);
2998                         read_unlock_bh(&idev->lock);
2999
3000                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3001                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3002                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3003                                                  jiffies);
3004                         ipv6_del_addr(ifp);
3005                         addrconf_verify_rtnl();
3006                         if (ipv6_addr_is_multicast(pfx)) {
3007                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3008                                                false, pfx, dev->ifindex);
3009                         }
3010                         return 0;
3011                 }
3012         }
3013         read_unlock_bh(&idev->lock);
3014         return -EADDRNOTAVAIL;
3015 }
3016
3017
3018 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3019 {
3020         struct ifa6_config cfg = {
3021                 .ifa_flags = IFA_F_PERMANENT,
3022                 .preferred_lft = INFINITY_LIFE_TIME,
3023                 .valid_lft = INFINITY_LIFE_TIME,
3024         };
3025         struct in6_ifreq ireq;
3026         int err;
3027
3028         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3029                 return -EPERM;
3030
3031         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3032                 return -EFAULT;
3033
3034         cfg.pfx = &ireq.ifr6_addr;
3035         cfg.plen = ireq.ifr6_prefixlen;
3036
3037         rtnl_lock();
3038         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3039         rtnl_unlock();
3040         return err;
3041 }
3042
3043 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3044 {
3045         struct in6_ifreq ireq;
3046         int err;
3047
3048         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3049                 return -EPERM;
3050
3051         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3052                 return -EFAULT;
3053
3054         rtnl_lock();
3055         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3056                              ireq.ifr6_prefixlen);
3057         rtnl_unlock();
3058         return err;
3059 }
3060
3061 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3062                      int plen, int scope)
3063 {
3064         struct inet6_ifaddr *ifp;
3065         struct ifa6_config cfg = {
3066                 .pfx = addr,
3067                 .plen = plen,
3068                 .ifa_flags = IFA_F_PERMANENT,
3069                 .valid_lft = INFINITY_LIFE_TIME,
3070                 .preferred_lft = INFINITY_LIFE_TIME,
3071                 .scope = scope
3072         };
3073
3074         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3075         if (!IS_ERR(ifp)) {
3076                 spin_lock_bh(&ifp->lock);
3077                 ifp->flags &= ~IFA_F_TENTATIVE;
3078                 spin_unlock_bh(&ifp->lock);
3079                 rt_genid_bump_ipv6(dev_net(idev->dev));
3080                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3081                 in6_ifa_put(ifp);
3082         }
3083 }
3084
3085 #if IS_ENABLED(CONFIG_IPV6_SIT)
3086 static void sit_add_v4_addrs(struct inet6_dev *idev)
3087 {
3088         struct in6_addr addr;
3089         struct net_device *dev;
3090         struct net *net = dev_net(idev->dev);
3091         int scope, plen;
3092         u32 pflags = 0;
3093
3094         ASSERT_RTNL();
3095
3096         memset(&addr, 0, sizeof(struct in6_addr));
3097         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3098
3099         if (idev->dev->flags&IFF_POINTOPOINT) {
3100                 addr.s6_addr32[0] = htonl(0xfe800000);
3101                 scope = IFA_LINK;
3102                 plen = 64;
3103         } else {
3104                 scope = IPV6_ADDR_COMPATv4;
3105                 plen = 96;
3106                 pflags |= RTF_NONEXTHOP;
3107         }
3108
3109         if (addr.s6_addr32[3]) {
3110                 add_addr(idev, &addr, plen, scope);
3111                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3112                                       GFP_KERNEL);
3113                 return;
3114         }
3115
3116         for_each_netdev(net, dev) {
3117                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3118                 if (in_dev && (dev->flags & IFF_UP)) {
3119                         struct in_ifaddr *ifa;
3120                         int flag = scope;
3121
3122                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3123                                 addr.s6_addr32[3] = ifa->ifa_local;
3124
3125                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3126                                         continue;
3127                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3128                                         if (idev->dev->flags&IFF_POINTOPOINT)
3129                                                 continue;
3130                                         flag |= IFA_HOST;
3131                                 }
3132
3133                                 add_addr(idev, &addr, plen, flag);
3134                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3135                                                       0, pflags, GFP_KERNEL);
3136                         }
3137                 }
3138         }
3139 }
3140 #endif
3141
3142 static void init_loopback(struct net_device *dev)
3143 {
3144         struct inet6_dev  *idev;
3145
3146         /* ::1 */
3147
3148         ASSERT_RTNL();
3149
3150         idev = ipv6_find_idev(dev);
3151         if (IS_ERR(idev)) {
3152                 pr_debug("%s: add_dev failed\n", __func__);
3153                 return;
3154         }
3155
3156         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3157 }
3158
3159 void addrconf_add_linklocal(struct inet6_dev *idev,
3160                             const struct in6_addr *addr, u32 flags)
3161 {
3162         struct ifa6_config cfg = {
3163                 .pfx = addr,
3164                 .plen = 64,
3165                 .ifa_flags = flags | IFA_F_PERMANENT,
3166                 .valid_lft = INFINITY_LIFE_TIME,
3167                 .preferred_lft = INFINITY_LIFE_TIME,
3168                 .scope = IFA_LINK
3169         };
3170         struct inet6_ifaddr *ifp;
3171
3172 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3173         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3174              idev->cnf.optimistic_dad) &&
3175             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3176                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3177 #endif
3178
3179         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3180         if (!IS_ERR(ifp)) {
3181                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3182                                       0, 0, GFP_ATOMIC);
3183                 addrconf_dad_start(ifp);
3184                 in6_ifa_put(ifp);
3185         }
3186 }
3187 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3188
3189 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3190 {
3191         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3192                 return true;
3193
3194         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3195             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3196                 return true;
3197
3198         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3199             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3200                 return true;
3201
3202         return false;
3203 }
3204
3205 static int ipv6_generate_stable_address(struct in6_addr *address,
3206                                         u8 dad_count,
3207                                         const struct inet6_dev *idev)
3208 {
3209         static DEFINE_SPINLOCK(lock);
3210         static __u32 digest[SHA1_DIGEST_WORDS];
3211         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3212
3213         static union {
3214                 char __data[SHA1_BLOCK_SIZE];
3215                 struct {
3216                         struct in6_addr secret;
3217                         __be32 prefix[2];
3218                         unsigned char hwaddr[MAX_ADDR_LEN];
3219                         u8 dad_count;
3220                 } __packed;
3221         } data;
3222
3223         struct in6_addr secret;
3224         struct in6_addr temp;
3225         struct net *net = dev_net(idev->dev);
3226
3227         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3228
3229         if (idev->cnf.stable_secret.initialized)
3230                 secret = idev->cnf.stable_secret.secret;
3231         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3232                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3233         else
3234                 return -1;
3235
3236 retry:
3237         spin_lock_bh(&lock);
3238
3239         sha1_init(digest);
3240         memset(&data, 0, sizeof(data));
3241         memset(workspace, 0, sizeof(workspace));
3242         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3243         data.prefix[0] = address->s6_addr32[0];
3244         data.prefix[1] = address->s6_addr32[1];
3245         data.secret = secret;
3246         data.dad_count = dad_count;
3247
3248         sha1_transform(digest, data.__data, workspace);
3249
3250         temp = *address;
3251         temp.s6_addr32[2] = (__force __be32)digest[0];
3252         temp.s6_addr32[3] = (__force __be32)digest[1];
3253
3254         spin_unlock_bh(&lock);
3255
3256         if (ipv6_reserved_interfaceid(temp)) {
3257                 dad_count++;
3258                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3259                         return -1;
3260                 goto retry;
3261         }
3262
3263         *address = temp;
3264         return 0;
3265 }
3266
3267 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3268 {
3269         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3270
3271         if (s->initialized)
3272                 return;
3273         s = &idev->cnf.stable_secret;
3274         get_random_bytes(&s->secret, sizeof(s->secret));
3275         s->initialized = true;
3276 }
3277
3278 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3279 {
3280         struct in6_addr addr;
3281
3282         /* no link local addresses on L3 master devices */
3283         if (netif_is_l3_master(idev->dev))
3284                 return;
3285
3286         /* no link local addresses on devices flagged as slaves */
3287         if (idev->dev->flags & IFF_SLAVE)
3288                 return;
3289
3290         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3291
3292         switch (idev->cnf.addr_gen_mode) {
3293         case IN6_ADDR_GEN_MODE_RANDOM:
3294                 ipv6_gen_mode_random_init(idev);
3295                 fallthrough;
3296         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3297                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3298                         addrconf_add_linklocal(idev, &addr,
3299                                                IFA_F_STABLE_PRIVACY);
3300                 else if (prefix_route)
3301                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3302                                               0, 0, GFP_KERNEL);
3303                 break;
3304         case IN6_ADDR_GEN_MODE_EUI64:
3305                 /* addrconf_add_linklocal also adds a prefix_route and we
3306                  * only need to care about prefix routes if ipv6_generate_eui64
3307                  * couldn't generate one.
3308                  */
3309                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3310                         addrconf_add_linklocal(idev, &addr, 0);
3311                 else if (prefix_route)
3312                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3313                                               0, 0, GFP_KERNEL);
3314                 break;
3315         case IN6_ADDR_GEN_MODE_NONE:
3316         default:
3317                 /* will not add any link local address */
3318                 break;
3319         }
3320 }
3321
3322 static void addrconf_dev_config(struct net_device *dev)
3323 {
3324         struct inet6_dev *idev;
3325
3326         ASSERT_RTNL();
3327
3328         if ((dev->type != ARPHRD_ETHER) &&
3329             (dev->type != ARPHRD_FDDI) &&
3330             (dev->type != ARPHRD_ARCNET) &&
3331             (dev->type != ARPHRD_INFINIBAND) &&
3332             (dev->type != ARPHRD_IEEE1394) &&
3333             (dev->type != ARPHRD_TUNNEL6) &&
3334             (dev->type != ARPHRD_6LOWPAN) &&
3335             (dev->type != ARPHRD_IP6GRE) &&
3336             (dev->type != ARPHRD_IPGRE) &&
3337             (dev->type != ARPHRD_TUNNEL) &&
3338             (dev->type != ARPHRD_NONE) &&
3339             (dev->type != ARPHRD_RAWIP)) {
3340                 /* Alas, we support only Ethernet autoconfiguration. */
3341                 idev = __in6_dev_get(dev);
3342                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3343                     dev->flags & IFF_MULTICAST)
3344                         ipv6_mc_up(idev);
3345                 return;
3346         }
3347
3348         idev = addrconf_add_dev(dev);
3349         if (IS_ERR(idev))
3350                 return;
3351
3352         /* this device type has no EUI support */
3353         if (dev->type == ARPHRD_NONE &&
3354             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3355                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3356
3357         addrconf_addr_gen(idev, false);
3358 }
3359
3360 #if IS_ENABLED(CONFIG_IPV6_SIT)
3361 static void addrconf_sit_config(struct net_device *dev)
3362 {
3363         struct inet6_dev *idev;
3364
3365         ASSERT_RTNL();
3366
3367         /*
3368          * Configure the tunnel with one of our IPv4
3369          * addresses... we should configure all of
3370          * our v4 addrs in the tunnel
3371          */
3372
3373         idev = ipv6_find_idev(dev);
3374         if (IS_ERR(idev)) {
3375                 pr_debug("%s: add_dev failed\n", __func__);
3376                 return;
3377         }
3378
3379         if (dev->priv_flags & IFF_ISATAP) {
3380                 addrconf_addr_gen(idev, false);
3381                 return;
3382         }
3383
3384         sit_add_v4_addrs(idev);
3385
3386         if (dev->flags&IFF_POINTOPOINT)
3387                 addrconf_add_mroute(dev);
3388 }
3389 #endif
3390
3391 #if IS_ENABLED(CONFIG_NET_IPGRE)
3392 static void addrconf_gre_config(struct net_device *dev)
3393 {
3394         struct inet6_dev *idev;
3395
3396         ASSERT_RTNL();
3397
3398         idev = ipv6_find_idev(dev);
3399         if (IS_ERR(idev)) {
3400                 pr_debug("%s: add_dev failed\n", __func__);
3401                 return;
3402         }
3403
3404         addrconf_addr_gen(idev, true);
3405         if (dev->flags & IFF_POINTOPOINT)
3406                 addrconf_add_mroute(dev);
3407 }
3408 #endif
3409
3410 static int fixup_permanent_addr(struct net *net,
3411                                 struct inet6_dev *idev,
3412                                 struct inet6_ifaddr *ifp)
3413 {
3414         /* !fib6_node means the host route was removed from the
3415          * FIB, for example, if 'lo' device is taken down. In that
3416          * case regenerate the host route.
3417          */
3418         if (!ifp->rt || !ifp->rt->fib6_node) {
3419                 struct fib6_info *f6i, *prev;
3420
3421                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3422                                          GFP_ATOMIC);
3423                 if (IS_ERR(f6i))
3424                         return PTR_ERR(f6i);
3425
3426                 /* ifp->rt can be accessed outside of rtnl */
3427                 spin_lock(&ifp->lock);
3428                 prev = ifp->rt;
3429                 ifp->rt = f6i;
3430                 spin_unlock(&ifp->lock);
3431
3432                 fib6_info_release(prev);
3433         }
3434
3435         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3436                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3437                                       ifp->rt_priority, idev->dev, 0, 0,
3438                                       GFP_ATOMIC);
3439         }
3440
3441         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3442                 addrconf_dad_start(ifp);
3443
3444         return 0;
3445 }
3446
3447 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3448 {
3449         struct inet6_ifaddr *ifp, *tmp;
3450         struct inet6_dev *idev;
3451
3452         idev = __in6_dev_get(dev);
3453         if (!idev)
3454                 return;
3455
3456         write_lock_bh(&idev->lock);
3457
3458         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3459                 if ((ifp->flags & IFA_F_PERMANENT) &&
3460                     fixup_permanent_addr(net, idev, ifp) < 0) {
3461                         write_unlock_bh(&idev->lock);
3462                         in6_ifa_hold(ifp);
3463                         ipv6_del_addr(ifp);
3464                         write_lock_bh(&idev->lock);
3465
3466                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3467                                              idev->dev->name, &ifp->addr);
3468                 }
3469         }
3470
3471         write_unlock_bh(&idev->lock);
3472 }
3473
3474 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3475                            void *ptr)
3476 {
3477         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3478         struct netdev_notifier_change_info *change_info;
3479         struct netdev_notifier_changeupper_info *info;
3480         struct inet6_dev *idev = __in6_dev_get(dev);
3481         struct net *net = dev_net(dev);
3482         int run_pending = 0;
3483         int err;
3484
3485         switch (event) {
3486         case NETDEV_REGISTER:
3487                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3488                         idev = ipv6_add_dev(dev);
3489                         if (IS_ERR(idev))
3490                                 return notifier_from_errno(PTR_ERR(idev));
3491                 }
3492                 break;
3493
3494         case NETDEV_CHANGEMTU:
3495                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3496                 if (dev->mtu < IPV6_MIN_MTU) {
3497                         addrconf_ifdown(dev, dev != net->loopback_dev);
3498                         break;
3499                 }
3500
3501                 if (idev) {
3502                         rt6_mtu_change(dev, dev->mtu);
3503                         idev->cnf.mtu6 = dev->mtu;
3504                         break;
3505                 }
3506
3507                 /* allocate new idev */
3508                 idev = ipv6_add_dev(dev);
3509                 if (IS_ERR(idev))
3510                         break;
3511
3512                 /* device is still not ready */
3513                 if (!(idev->if_flags & IF_READY))
3514                         break;
3515
3516                 run_pending = 1;
3517                 fallthrough;
3518         case NETDEV_UP:
3519         case NETDEV_CHANGE:
3520                 if (dev->flags & IFF_SLAVE)
3521                         break;
3522
3523                 if (idev && idev->cnf.disable_ipv6)
3524                         break;
3525
3526                 if (event == NETDEV_UP) {
3527                         /* restore routes for permanent addresses */
3528                         addrconf_permanent_addr(net, dev);
3529
3530                         if (!addrconf_link_ready(dev)) {
3531                                 /* device is not ready yet. */
3532                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3533                                          dev->name);
3534                                 break;
3535                         }
3536
3537                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3538                                 idev = ipv6_add_dev(dev);
3539
3540                         if (!IS_ERR_OR_NULL(idev)) {
3541                                 idev->if_flags |= IF_READY;
3542                                 run_pending = 1;
3543                         }
3544                 } else if (event == NETDEV_CHANGE) {
3545                         if (!addrconf_link_ready(dev)) {
3546                                 /* device is still not ready. */
3547                                 rt6_sync_down_dev(dev, event);
3548                                 break;
3549                         }
3550
3551                         if (!IS_ERR_OR_NULL(idev)) {
3552                                 if (idev->if_flags & IF_READY) {
3553                                         /* device is already configured -
3554                                          * but resend MLD reports, we might
3555                                          * have roamed and need to update
3556                                          * multicast snooping switches
3557                                          */
3558                                         ipv6_mc_up(idev);
3559                                         change_info = ptr;
3560                                         if (change_info->flags_changed & IFF_NOARP)
3561                                                 addrconf_dad_run(idev, true);
3562                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3563                                         break;
3564                                 }
3565                                 idev->if_flags |= IF_READY;
3566                         }
3567
3568                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3569                                 dev->name);
3570
3571                         run_pending = 1;
3572                 }
3573
3574                 switch (dev->type) {
3575 #if IS_ENABLED(CONFIG_IPV6_SIT)
3576                 case ARPHRD_SIT:
3577                         addrconf_sit_config(dev);
3578                         break;
3579 #endif
3580 #if IS_ENABLED(CONFIG_NET_IPGRE)
3581                 case ARPHRD_IPGRE:
3582                         addrconf_gre_config(dev);
3583                         break;
3584 #endif
3585                 case ARPHRD_LOOPBACK:
3586                         init_loopback(dev);
3587                         break;
3588
3589                 default:
3590                         addrconf_dev_config(dev);
3591                         break;
3592                 }
3593
3594                 if (!IS_ERR_OR_NULL(idev)) {
3595                         if (run_pending)
3596                                 addrconf_dad_run(idev, false);
3597
3598                         /* Device has an address by now */
3599                         rt6_sync_up(dev, RTNH_F_DEAD);
3600
3601                         /*
3602                          * If the MTU changed during the interface down,
3603                          * when the interface up, the changed MTU must be
3604                          * reflected in the idev as well as routers.
3605                          */
3606                         if (idev->cnf.mtu6 != dev->mtu &&
3607                             dev->mtu >= IPV6_MIN_MTU) {
3608                                 rt6_mtu_change(dev, dev->mtu);
3609                                 idev->cnf.mtu6 = dev->mtu;
3610                         }
3611                         idev->tstamp = jiffies;
3612                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3613
3614                         /*
3615                          * If the changed mtu during down is lower than
3616                          * IPV6_MIN_MTU stop IPv6 on this interface.
3617                          */
3618                         if (dev->mtu < IPV6_MIN_MTU)
3619                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3620                 }
3621                 break;
3622
3623         case NETDEV_DOWN:
3624         case NETDEV_UNREGISTER:
3625                 /*
3626                  *      Remove all addresses from this interface.
3627                  */
3628                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3629                 break;
3630
3631         case NETDEV_CHANGENAME:
3632                 if (idev) {
3633                         snmp6_unregister_dev(idev);
3634                         addrconf_sysctl_unregister(idev);
3635                         err = addrconf_sysctl_register(idev);
3636                         if (err)
3637                                 return notifier_from_errno(err);
3638                         err = snmp6_register_dev(idev);
3639                         if (err) {
3640                                 addrconf_sysctl_unregister(idev);
3641                                 return notifier_from_errno(err);
3642                         }
3643                 }
3644                 break;
3645
3646         case NETDEV_PRE_TYPE_CHANGE:
3647         case NETDEV_POST_TYPE_CHANGE:
3648                 if (idev)
3649                         addrconf_type_change(dev, event);
3650                 break;
3651
3652         case NETDEV_CHANGEUPPER:
3653                 info = ptr;
3654
3655                 /* flush all routes if dev is linked to or unlinked from
3656                  * an L3 master device (e.g., VRF)
3657                  */
3658                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3659                         addrconf_ifdown(dev, false);
3660         }
3661
3662         return NOTIFY_OK;
3663 }
3664
3665 /*
3666  *      addrconf module should be notified of a device going up
3667  */
3668 static struct notifier_block ipv6_dev_notf = {
3669         .notifier_call = addrconf_notify,
3670         .priority = ADDRCONF_NOTIFY_PRIORITY,
3671 };
3672
3673 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3674 {
3675         struct inet6_dev *idev;
3676         ASSERT_RTNL();
3677
3678         idev = __in6_dev_get(dev);
3679
3680         if (event == NETDEV_POST_TYPE_CHANGE)
3681                 ipv6_mc_remap(idev);
3682         else if (event == NETDEV_PRE_TYPE_CHANGE)
3683                 ipv6_mc_unmap(idev);
3684 }
3685
3686 static bool addr_is_local(const struct in6_addr *addr)
3687 {
3688         return ipv6_addr_type(addr) &
3689                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3690 }
3691
3692 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3693 {
3694         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3695         struct net *net = dev_net(dev);
3696         struct inet6_dev *idev;
3697         struct inet6_ifaddr *ifa, *tmp;
3698         bool keep_addr = false;
3699         int state, i;
3700
3701         ASSERT_RTNL();
3702
3703         rt6_disable_ip(dev, event);
3704
3705         idev = __in6_dev_get(dev);
3706         if (!idev)
3707                 return -ENODEV;
3708
3709         /*
3710          * Step 1: remove reference to ipv6 device from parent device.
3711          *         Do not dev_put!
3712          */
3713         if (unregister) {
3714                 idev->dead = 1;
3715
3716                 /* protected by rtnl_lock */
3717                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3718
3719                 /* Step 1.5: remove snmp6 entry */
3720                 snmp6_unregister_dev(idev);
3721
3722         }
3723
3724         /* combine the user config with event to determine if permanent
3725          * addresses are to be removed from address hash table
3726          */
3727         if (!unregister && !idev->cnf.disable_ipv6) {
3728                 /* aggregate the system setting and interface setting */
3729                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3730
3731                 if (!_keep_addr)
3732                         _keep_addr = idev->cnf.keep_addr_on_down;
3733
3734                 keep_addr = (_keep_addr > 0);
3735         }
3736
3737         /* Step 2: clear hash table */
3738         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3739                 struct hlist_head *h = &inet6_addr_lst[i];
3740
3741                 spin_lock_bh(&addrconf_hash_lock);
3742 restart:
3743                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3744                         if (ifa->idev == idev) {
3745                                 addrconf_del_dad_work(ifa);
3746                                 /* combined flag + permanent flag decide if
3747                                  * address is retained on a down event
3748                                  */
3749                                 if (!keep_addr ||
3750                                     !(ifa->flags & IFA_F_PERMANENT) ||
3751                                     addr_is_local(&ifa->addr)) {
3752                                         hlist_del_init_rcu(&ifa->addr_lst);
3753                                         goto restart;
3754                                 }
3755                         }
3756                 }
3757                 spin_unlock_bh(&addrconf_hash_lock);
3758         }
3759
3760         write_lock_bh(&idev->lock);
3761
3762         addrconf_del_rs_timer(idev);
3763
3764         /* Step 2: clear flags for stateless addrconf */
3765         if (!unregister)
3766                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3767
3768         /* Step 3: clear tempaddr list */
3769         while (!list_empty(&idev->tempaddr_list)) {
3770                 ifa = list_first_entry(&idev->tempaddr_list,
3771                                        struct inet6_ifaddr, tmp_list);
3772                 list_del(&ifa->tmp_list);
3773                 write_unlock_bh(&idev->lock);
3774                 spin_lock_bh(&ifa->lock);
3775
3776                 if (ifa->ifpub) {
3777                         in6_ifa_put(ifa->ifpub);
3778                         ifa->ifpub = NULL;
3779                 }
3780                 spin_unlock_bh(&ifa->lock);
3781                 in6_ifa_put(ifa);
3782                 write_lock_bh(&idev->lock);
3783         }
3784
3785         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3786                 struct fib6_info *rt = NULL;
3787                 bool keep;
3788
3789                 addrconf_del_dad_work(ifa);
3790
3791                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3792                         !addr_is_local(&ifa->addr);
3793
3794                 write_unlock_bh(&idev->lock);
3795                 spin_lock_bh(&ifa->lock);
3796
3797                 if (keep) {
3798                         /* set state to skip the notifier below */
3799                         state = INET6_IFADDR_STATE_DEAD;
3800                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3801                         if (!(ifa->flags & IFA_F_NODAD))
3802                                 ifa->flags |= IFA_F_TENTATIVE;
3803
3804                         rt = ifa->rt;
3805                         ifa->rt = NULL;
3806                 } else {
3807                         state = ifa->state;
3808                         ifa->state = INET6_IFADDR_STATE_DEAD;
3809                 }
3810
3811                 spin_unlock_bh(&ifa->lock);
3812
3813                 if (rt)
3814                         ip6_del_rt(net, rt, false);
3815
3816                 if (state != INET6_IFADDR_STATE_DEAD) {
3817                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3818                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3819                 } else {
3820                         if (idev->cnf.forwarding)
3821                                 addrconf_leave_anycast(ifa);
3822                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3823                 }
3824
3825                 write_lock_bh(&idev->lock);
3826                 if (!keep) {
3827                         list_del_rcu(&ifa->if_list);
3828                         in6_ifa_put(ifa);
3829                 }
3830         }
3831
3832         write_unlock_bh(&idev->lock);
3833
3834         /* Step 5: Discard anycast and multicast list */
3835         if (unregister) {
3836                 ipv6_ac_destroy_dev(idev);
3837                 ipv6_mc_destroy_dev(idev);
3838         } else {
3839                 ipv6_mc_down(idev);
3840         }
3841
3842         idev->tstamp = jiffies;
3843
3844         /* Last: Shot the device (if unregistered) */
3845         if (unregister) {
3846                 addrconf_sysctl_unregister(idev);
3847                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3848                 neigh_ifdown(&nd_tbl, dev);
3849                 in6_dev_put(idev);
3850         }
3851         return 0;
3852 }
3853
3854 static void addrconf_rs_timer(struct timer_list *t)
3855 {
3856         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3857         struct net_device *dev = idev->dev;
3858         struct in6_addr lladdr;
3859
3860         write_lock(&idev->lock);
3861         if (idev->dead || !(idev->if_flags & IF_READY))
3862                 goto out;
3863
3864         if (!ipv6_accept_ra(idev))
3865                 goto out;
3866
3867         /* Announcement received after solicitation was sent */
3868         if (idev->if_flags & IF_RA_RCVD)
3869                 goto out;
3870
3871         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3872                 write_unlock(&idev->lock);
3873                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3874                         ndisc_send_rs(dev, &lladdr,
3875                                       &in6addr_linklocal_allrouters);
3876                 else
3877                         goto put;
3878
3879                 write_lock(&idev->lock);
3880                 idev->rs_interval = rfc3315_s14_backoff_update(
3881                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3882                 /* The wait after the last probe can be shorter */
3883                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3884                                              idev->cnf.rtr_solicits) ?
3885                                       idev->cnf.rtr_solicit_delay :
3886                                       idev->rs_interval);
3887         } else {
3888                 /*
3889                  * Note: we do not support deprecated "all on-link"
3890                  * assumption any longer.
3891                  */
3892                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3893         }
3894
3895 out:
3896         write_unlock(&idev->lock);
3897 put:
3898         in6_dev_put(idev);
3899 }
3900
3901 /*
3902  *      Duplicate Address Detection
3903  */
3904 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3905 {
3906         unsigned long rand_num;
3907         struct inet6_dev *idev = ifp->idev;
3908         u64 nonce;
3909
3910         if (ifp->flags & IFA_F_OPTIMISTIC)
3911                 rand_num = 0;
3912         else
3913                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3914
3915         nonce = 0;
3916         if (idev->cnf.enhanced_dad ||
3917             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3918                 do
3919                         get_random_bytes(&nonce, 6);
3920                 while (nonce == 0);
3921         }
3922         ifp->dad_nonce = nonce;
3923         ifp->dad_probes = idev->cnf.dad_transmits;
3924         addrconf_mod_dad_work(ifp, rand_num);
3925 }
3926
3927 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3928 {
3929         struct inet6_dev *idev = ifp->idev;
3930         struct net_device *dev = idev->dev;
3931         bool bump_id, notify = false;
3932         struct net *net;
3933
3934         addrconf_join_solict(dev, &ifp->addr);
3935
3936         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3937
3938         read_lock_bh(&idev->lock);
3939         spin_lock(&ifp->lock);
3940         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3941                 goto out;
3942
3943         net = dev_net(dev);
3944         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3945             (net->ipv6.devconf_all->accept_dad < 1 &&
3946              idev->cnf.accept_dad < 1) ||
3947             !(ifp->flags&IFA_F_TENTATIVE) ||
3948             ifp->flags & IFA_F_NODAD) {
3949                 bool send_na = false;
3950
3951                 if (ifp->flags & IFA_F_TENTATIVE &&
3952                     !(ifp->flags & IFA_F_OPTIMISTIC))
3953                         send_na = true;
3954                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3955                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3956                 spin_unlock(&ifp->lock);
3957                 read_unlock_bh(&idev->lock);
3958
3959                 addrconf_dad_completed(ifp, bump_id, send_na);
3960                 return;
3961         }
3962
3963         if (!(idev->if_flags & IF_READY)) {
3964                 spin_unlock(&ifp->lock);
3965                 read_unlock_bh(&idev->lock);
3966                 /*
3967                  * If the device is not ready:
3968                  * - keep it tentative if it is a permanent address.
3969                  * - otherwise, kill it.
3970                  */
3971                 in6_ifa_hold(ifp);
3972                 addrconf_dad_stop(ifp, 0);
3973                 return;
3974         }
3975
3976         /*
3977          * Optimistic nodes can start receiving
3978          * Frames right away
3979          */
3980         if (ifp->flags & IFA_F_OPTIMISTIC) {
3981                 ip6_ins_rt(net, ifp->rt);
3982                 if (ipv6_use_optimistic_addr(net, idev)) {
3983                         /* Because optimistic nodes can use this address,
3984                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3985                          */
3986                         notify = true;
3987                 }
3988         }
3989
3990         addrconf_dad_kick(ifp);
3991 out:
3992         spin_unlock(&ifp->lock);
3993         read_unlock_bh(&idev->lock);
3994         if (notify)
3995                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3996 }
3997
3998 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3999 {
4000         bool begin_dad = false;
4001
4002         spin_lock_bh(&ifp->lock);
4003         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4004                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4005                 begin_dad = true;
4006         }
4007         spin_unlock_bh(&ifp->lock);
4008
4009         if (begin_dad)
4010                 addrconf_mod_dad_work(ifp, 0);
4011 }
4012
4013 static void addrconf_dad_work(struct work_struct *w)
4014 {
4015         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4016                                                 struct inet6_ifaddr,
4017                                                 dad_work);
4018         struct inet6_dev *idev = ifp->idev;
4019         bool bump_id, disable_ipv6 = false;
4020         struct in6_addr mcaddr;
4021
4022         enum {
4023                 DAD_PROCESS,
4024                 DAD_BEGIN,
4025                 DAD_ABORT,
4026         } action = DAD_PROCESS;
4027
4028         rtnl_lock();
4029
4030         spin_lock_bh(&ifp->lock);
4031         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4032                 action = DAD_BEGIN;
4033                 ifp->state = INET6_IFADDR_STATE_DAD;
4034         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4035                 action = DAD_ABORT;
4036                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4037
4038                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4039                      idev->cnf.accept_dad > 1) &&
4040                     !idev->cnf.disable_ipv6 &&
4041                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4042                         struct in6_addr addr;
4043
4044                         addr.s6_addr32[0] = htonl(0xfe800000);
4045                         addr.s6_addr32[1] = 0;
4046
4047                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4048                             ipv6_addr_equal(&ifp->addr, &addr)) {
4049                                 /* DAD failed for link-local based on MAC */
4050                                 idev->cnf.disable_ipv6 = 1;
4051
4052                                 pr_info("%s: IPv6 being disabled!\n",
4053                                         ifp->idev->dev->name);
4054                                 disable_ipv6 = true;
4055                         }
4056                 }
4057         }
4058         spin_unlock_bh(&ifp->lock);
4059
4060         if (action == DAD_BEGIN) {
4061                 addrconf_dad_begin(ifp);
4062                 goto out;
4063         } else if (action == DAD_ABORT) {
4064                 in6_ifa_hold(ifp);
4065                 addrconf_dad_stop(ifp, 1);
4066                 if (disable_ipv6)
4067                         addrconf_ifdown(idev->dev, false);
4068                 goto out;
4069         }
4070
4071         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4072                 goto out;
4073
4074         write_lock_bh(&idev->lock);
4075         if (idev->dead || !(idev->if_flags & IF_READY)) {
4076                 write_unlock_bh(&idev->lock);
4077                 goto out;
4078         }
4079
4080         spin_lock(&ifp->lock);
4081         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4082                 spin_unlock(&ifp->lock);
4083                 write_unlock_bh(&idev->lock);
4084                 goto out;
4085         }
4086
4087         if (ifp->dad_probes == 0) {
4088                 bool send_na = false;
4089
4090                 /*
4091                  * DAD was successful
4092                  */
4093
4094                 if (ifp->flags & IFA_F_TENTATIVE &&
4095                     !(ifp->flags & IFA_F_OPTIMISTIC))
4096                         send_na = true;
4097                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4098                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4099                 spin_unlock(&ifp->lock);
4100                 write_unlock_bh(&idev->lock);
4101
4102                 addrconf_dad_completed(ifp, bump_id, send_na);
4103
4104                 goto out;
4105         }
4106
4107         ifp->dad_probes--;
4108         addrconf_mod_dad_work(ifp,
4109                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4110                                   HZ/100));
4111         spin_unlock(&ifp->lock);
4112         write_unlock_bh(&idev->lock);
4113
4114         /* send a neighbour solicitation for our addr */
4115         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4116         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4117                       ifp->dad_nonce);
4118 out:
4119         in6_ifa_put(ifp);
4120         rtnl_unlock();
4121 }
4122
4123 /* ifp->idev must be at least read locked */
4124 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4125 {
4126         struct inet6_ifaddr *ifpiter;
4127         struct inet6_dev *idev = ifp->idev;
4128
4129         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4130                 if (ifpiter->scope > IFA_LINK)
4131                         break;
4132                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4133                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4134                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4135                     IFA_F_PERMANENT)
4136                         return false;
4137         }
4138         return true;
4139 }
4140
4141 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4142                                    bool send_na)
4143 {
4144         struct net_device *dev = ifp->idev->dev;
4145         struct in6_addr lladdr;
4146         bool send_rs, send_mld;
4147
4148         addrconf_del_dad_work(ifp);
4149
4150         /*
4151          *      Configure the address for reception. Now it is valid.
4152          */
4153
4154         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4155
4156         /* If added prefix is link local and we are prepared to process
4157            router advertisements, start sending router solicitations.
4158          */
4159
4160         read_lock_bh(&ifp->idev->lock);
4161         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4162         send_rs = send_mld &&
4163                   ipv6_accept_ra(ifp->idev) &&
4164                   ifp->idev->cnf.rtr_solicits != 0 &&
4165                   (dev->flags&IFF_LOOPBACK) == 0;
4166         read_unlock_bh(&ifp->idev->lock);
4167
4168         /* While dad is in progress mld report's source address is in6_addrany.
4169          * Resend with proper ll now.
4170          */
4171         if (send_mld)
4172                 ipv6_mc_dad_complete(ifp->idev);
4173
4174         /* send unsolicited NA if enabled */
4175         if (send_na &&
4176             (ifp->idev->cnf.ndisc_notify ||
4177              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4178                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4179                               /*router=*/ !!ifp->idev->cnf.forwarding,
4180                               /*solicited=*/ false, /*override=*/ true,
4181                               /*inc_opt=*/ true);
4182         }
4183
4184         if (send_rs) {
4185                 /*
4186                  *      If a host as already performed a random delay
4187                  *      [...] as part of DAD [...] there is no need
4188                  *      to delay again before sending the first RS
4189                  */
4190                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4191                         return;
4192                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4193
4194                 write_lock_bh(&ifp->idev->lock);
4195                 spin_lock(&ifp->lock);
4196                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4197                         ifp->idev->cnf.rtr_solicit_interval);
4198                 ifp->idev->rs_probes = 1;
4199                 ifp->idev->if_flags |= IF_RS_SENT;
4200                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4201                 spin_unlock(&ifp->lock);
4202                 write_unlock_bh(&ifp->idev->lock);
4203         }
4204
4205         if (bump_id)
4206                 rt_genid_bump_ipv6(dev_net(dev));
4207
4208         /* Make sure that a new temporary address will be created
4209          * before this temporary address becomes deprecated.
4210          */
4211         if (ifp->flags & IFA_F_TEMPORARY)
4212                 addrconf_verify_rtnl();
4213 }
4214
4215 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4216 {
4217         struct inet6_ifaddr *ifp;
4218
4219         read_lock_bh(&idev->lock);
4220         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4221                 spin_lock(&ifp->lock);
4222                 if ((ifp->flags & IFA_F_TENTATIVE &&
4223                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4224                         if (restart)
4225                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4226                         addrconf_dad_kick(ifp);
4227                 }
4228                 spin_unlock(&ifp->lock);
4229         }
4230         read_unlock_bh(&idev->lock);
4231 }
4232
4233 #ifdef CONFIG_PROC_FS
4234 struct if6_iter_state {
4235         struct seq_net_private p;
4236         int bucket;
4237         int offset;
4238 };
4239
4240 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4241 {
4242         struct if6_iter_state *state = seq->private;
4243         struct net *net = seq_file_net(seq);
4244         struct inet6_ifaddr *ifa = NULL;
4245         int p = 0;
4246
4247         /* initial bucket if pos is 0 */
4248         if (pos == 0) {
4249                 state->bucket = 0;
4250                 state->offset = 0;
4251         }
4252
4253         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4254                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4255                                          addr_lst) {
4256                         if (!net_eq(dev_net(ifa->idev->dev), net))
4257                                 continue;
4258                         /* sync with offset */
4259                         if (p < state->offset) {
4260                                 p++;
4261                                 continue;
4262                         }
4263                         return ifa;
4264                 }
4265
4266                 /* prepare for next bucket */
4267                 state->offset = 0;
4268                 p = 0;
4269         }
4270         return NULL;
4271 }
4272
4273 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4274                                          struct inet6_ifaddr *ifa)
4275 {
4276         struct if6_iter_state *state = seq->private;
4277         struct net *net = seq_file_net(seq);
4278
4279         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4280                 if (!net_eq(dev_net(ifa->idev->dev), net))
4281                         continue;
4282                 state->offset++;
4283                 return ifa;
4284         }
4285
4286         state->offset = 0;
4287         while (++state->bucket < IN6_ADDR_HSIZE) {
4288                 hlist_for_each_entry_rcu(ifa,
4289                                      &inet6_addr_lst[state->bucket], addr_lst) {
4290                         if (!net_eq(dev_net(ifa->idev->dev), net))
4291                                 continue;
4292                         return ifa;
4293                 }
4294         }
4295
4296         return NULL;
4297 }
4298
4299 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4300         __acquires(rcu)
4301 {
4302         rcu_read_lock();
4303         return if6_get_first(seq, *pos);
4304 }
4305
4306 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4307 {
4308         struct inet6_ifaddr *ifa;
4309
4310         ifa = if6_get_next(seq, v);
4311         ++*pos;
4312         return ifa;
4313 }
4314
4315 static void if6_seq_stop(struct seq_file *seq, void *v)
4316         __releases(rcu)
4317 {
4318         rcu_read_unlock();
4319 }
4320
4321 static int if6_seq_show(struct seq_file *seq, void *v)
4322 {
4323         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4324         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4325                    &ifp->addr,
4326                    ifp->idev->dev->ifindex,
4327                    ifp->prefix_len,
4328                    ifp->scope,
4329                    (u8) ifp->flags,
4330                    ifp->idev->dev->name);
4331         return 0;
4332 }
4333
4334 static const struct seq_operations if6_seq_ops = {
4335         .start  = if6_seq_start,
4336         .next   = if6_seq_next,
4337         .show   = if6_seq_show,
4338         .stop   = if6_seq_stop,
4339 };
4340
4341 static int __net_init if6_proc_net_init(struct net *net)
4342 {
4343         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4344                         sizeof(struct if6_iter_state)))
4345                 return -ENOMEM;
4346         return 0;
4347 }
4348
4349 static void __net_exit if6_proc_net_exit(struct net *net)
4350 {
4351         remove_proc_entry("if_inet6", net->proc_net);
4352 }
4353
4354 static struct pernet_operations if6_proc_net_ops = {
4355         .init = if6_proc_net_init,
4356         .exit = if6_proc_net_exit,
4357 };
4358
4359 int __init if6_proc_init(void)
4360 {
4361         return register_pernet_subsys(&if6_proc_net_ops);
4362 }
4363
4364 void if6_proc_exit(void)
4365 {
4366         unregister_pernet_subsys(&if6_proc_net_ops);
4367 }
4368 #endif  /* CONFIG_PROC_FS */
4369
4370 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4371 /* Check if address is a home address configured on any interface. */
4372 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4373 {
4374         unsigned int hash = inet6_addr_hash(net, addr);
4375         struct inet6_ifaddr *ifp = NULL;
4376         int ret = 0;
4377
4378         rcu_read_lock();
4379         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4380                 if (!net_eq(dev_net(ifp->idev->dev), net))
4381                         continue;
4382                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4383                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4384                         ret = 1;
4385                         break;
4386                 }
4387         }
4388         rcu_read_unlock();
4389         return ret;
4390 }
4391 #endif
4392
4393 /* RFC6554 has some algorithm to avoid loops in segment routing by
4394  * checking if the segments contains any of a local interface address.
4395  *
4396  * Quote:
4397  *
4398  * To detect loops in the SRH, a router MUST determine if the SRH
4399  * includes multiple addresses assigned to any interface on that router.
4400  * If such addresses appear more than once and are separated by at least
4401  * one address not assigned to that router.
4402  */
4403 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4404                           unsigned char nsegs)
4405 {
4406         const struct in6_addr *addr;
4407         int i, ret = 0, found = 0;
4408         struct inet6_ifaddr *ifp;
4409         bool separated = false;
4410         unsigned int hash;
4411         bool hash_found;
4412
4413         rcu_read_lock();
4414         for (i = 0; i < nsegs; i++) {
4415                 addr = &segs[i];
4416                 hash = inet6_addr_hash(net, addr);
4417
4418                 hash_found = false;
4419                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4420                         if (!net_eq(dev_net(ifp->idev->dev), net))
4421                                 continue;
4422
4423                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4424                                 hash_found = true;
4425                                 break;
4426                         }
4427                 }
4428
4429                 if (hash_found) {
4430                         if (found > 1 && separated) {
4431                                 ret = 1;
4432                                 break;
4433                         }
4434
4435                         separated = false;
4436                         found++;
4437                 } else {
4438                         separated = true;
4439                 }
4440         }
4441         rcu_read_unlock();
4442
4443         return ret;
4444 }
4445
4446 /*
4447  *      Periodic address status verification
4448  */
4449
4450 static void addrconf_verify_rtnl(void)
4451 {
4452         unsigned long now, next, next_sec, next_sched;
4453         struct inet6_ifaddr *ifp;
4454         int i;
4455
4456         ASSERT_RTNL();
4457
4458         rcu_read_lock_bh();
4459         now = jiffies;
4460         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4461
4462         cancel_delayed_work(&addr_chk_work);
4463
4464         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4465 restart:
4466                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4467                         unsigned long age;
4468
4469                         /* When setting preferred_lft to a value not zero or
4470                          * infinity, while valid_lft is infinity
4471                          * IFA_F_PERMANENT has a non-infinity life time.
4472                          */
4473                         if ((ifp->flags & IFA_F_PERMANENT) &&
4474                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4475                                 continue;
4476
4477                         spin_lock(&ifp->lock);
4478                         /* We try to batch several events at once. */
4479                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4480
4481                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4482                             age >= ifp->valid_lft) {
4483                                 spin_unlock(&ifp->lock);
4484                                 in6_ifa_hold(ifp);
4485                                 ipv6_del_addr(ifp);
4486                                 goto restart;
4487                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4488                                 spin_unlock(&ifp->lock);
4489                                 continue;
4490                         } else if (age >= ifp->prefered_lft) {
4491                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4492                                 int deprecate = 0;
4493
4494                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4495                                         deprecate = 1;
4496                                         ifp->flags |= IFA_F_DEPRECATED;
4497                                 }
4498
4499                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4500                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4501                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4502
4503                                 spin_unlock(&ifp->lock);
4504
4505                                 if (deprecate) {
4506                                         in6_ifa_hold(ifp);
4507
4508                                         ipv6_ifa_notify(0, ifp);
4509                                         in6_ifa_put(ifp);
4510                                         goto restart;
4511                                 }
4512                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4513                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4514                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4515                                         ifp->idev->cnf.dad_transmits *
4516                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4517
4518                                 if (age >= ifp->prefered_lft - regen_advance) {
4519                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4520                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4521                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4522                                         if (!ifp->regen_count && ifpub) {
4523                                                 ifp->regen_count++;
4524                                                 in6_ifa_hold(ifp);
4525                                                 in6_ifa_hold(ifpub);
4526                                                 spin_unlock(&ifp->lock);
4527
4528                                                 spin_lock(&ifpub->lock);
4529                                                 ifpub->regen_count = 0;
4530                                                 spin_unlock(&ifpub->lock);
4531                                                 rcu_read_unlock_bh();
4532                                                 ipv6_create_tempaddr(ifpub, true);
4533                                                 in6_ifa_put(ifpub);
4534                                                 in6_ifa_put(ifp);
4535                                                 rcu_read_lock_bh();
4536                                                 goto restart;
4537                                         }
4538                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4539                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4540                                 spin_unlock(&ifp->lock);
4541                         } else {
4542                                 /* ifp->prefered_lft <= ifp->valid_lft */
4543                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4544                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4545                                 spin_unlock(&ifp->lock);
4546                         }
4547                 }
4548         }
4549
4550         next_sec = round_jiffies_up(next);
4551         next_sched = next;
4552
4553         /* If rounded timeout is accurate enough, accept it. */
4554         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4555                 next_sched = next_sec;
4556
4557         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4558         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4559                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4560
4561         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4562                  now, next, next_sec, next_sched);
4563         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4564         rcu_read_unlock_bh();
4565 }
4566
4567 static void addrconf_verify_work(struct work_struct *w)
4568 {
4569         rtnl_lock();
4570         addrconf_verify_rtnl();
4571         rtnl_unlock();
4572 }
4573
4574 static void addrconf_verify(void)
4575 {
4576         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4577 }
4578
4579 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4580                                      struct in6_addr **peer_pfx)
4581 {
4582         struct in6_addr *pfx = NULL;
4583
4584         *peer_pfx = NULL;
4585
4586         if (addr)
4587                 pfx = nla_data(addr);
4588
4589         if (local) {
4590                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4591                         *peer_pfx = pfx;
4592                 pfx = nla_data(local);
4593         }
4594
4595         return pfx;
4596 }
4597
4598 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4599         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4600         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4601         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4602         [IFA_FLAGS]             = { .len = sizeof(u32) },
4603         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4604         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4605 };
4606
4607 static int
4608 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4609                   struct netlink_ext_ack *extack)
4610 {
4611         struct net *net = sock_net(skb->sk);
4612         struct ifaddrmsg *ifm;
4613         struct nlattr *tb[IFA_MAX+1];
4614         struct in6_addr *pfx, *peer_pfx;
4615         u32 ifa_flags;
4616         int err;
4617
4618         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4619                                      ifa_ipv6_policy, extack);
4620         if (err < 0)
4621                 return err;
4622
4623         ifm = nlmsg_data(nlh);
4624         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4625         if (!pfx)
4626                 return -EINVAL;
4627
4628         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4629
4630         /* We ignore other flags so far. */
4631         ifa_flags &= IFA_F_MANAGETEMPADDR;
4632
4633         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4634                               ifm->ifa_prefixlen);
4635 }
4636
4637 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4638                                unsigned long expires, u32 flags,
4639                                bool modify_peer)
4640 {
4641         struct fib6_info *f6i;
4642         u32 prio;
4643
4644         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4645                                         ifp->prefix_len,
4646                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4647         if (!f6i)
4648                 return -ENOENT;
4649
4650         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4651         if (f6i->fib6_metric != prio) {
4652                 /* delete old one */
4653                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4654
4655                 /* add new one */
4656                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4657                                       ifp->prefix_len,
4658                                       ifp->rt_priority, ifp->idev->dev,
4659                                       expires, flags, GFP_KERNEL);
4660         } else {
4661                 if (!expires)
4662                         fib6_clean_expires(f6i);
4663                 else
4664                         fib6_set_expires(f6i, expires);
4665
4666                 fib6_info_release(f6i);
4667         }
4668
4669         return 0;
4670 }
4671
4672 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4673 {
4674         u32 flags;
4675         clock_t expires;
4676         unsigned long timeout;
4677         bool was_managetempaddr;
4678         bool had_prefixroute;
4679         bool new_peer = false;
4680
4681         ASSERT_RTNL();
4682
4683         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4684                 return -EINVAL;
4685
4686         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4687             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4688                 return -EINVAL;
4689
4690         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4691                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4692
4693         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4694         if (addrconf_finite_timeout(timeout)) {
4695                 expires = jiffies_to_clock_t(timeout * HZ);
4696                 cfg->valid_lft = timeout;
4697                 flags = RTF_EXPIRES;
4698         } else {
4699                 expires = 0;
4700                 flags = 0;
4701                 cfg->ifa_flags |= IFA_F_PERMANENT;
4702         }
4703
4704         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4705         if (addrconf_finite_timeout(timeout)) {
4706                 if (timeout == 0)
4707                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4708                 cfg->preferred_lft = timeout;
4709         }
4710
4711         if (cfg->peer_pfx &&
4712             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4713                 if (!ipv6_addr_any(&ifp->peer_addr))
4714                         cleanup_prefix_route(ifp, expires, true, true);
4715                 new_peer = true;
4716         }
4717
4718         spin_lock_bh(&ifp->lock);
4719         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4720         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4721                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4722         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4723                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4724                         IFA_F_NOPREFIXROUTE);
4725         ifp->flags |= cfg->ifa_flags;
4726         ifp->tstamp = jiffies;
4727         ifp->valid_lft = cfg->valid_lft;
4728         ifp->prefered_lft = cfg->preferred_lft;
4729
4730         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4731                 ifp->rt_priority = cfg->rt_priority;
4732
4733         if (new_peer)
4734                 ifp->peer_addr = *cfg->peer_pfx;
4735
4736         spin_unlock_bh(&ifp->lock);
4737         if (!(ifp->flags&IFA_F_TENTATIVE))
4738                 ipv6_ifa_notify(0, ifp);
4739
4740         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4741                 int rc = -ENOENT;
4742
4743                 if (had_prefixroute)
4744                         rc = modify_prefix_route(ifp, expires, flags, false);
4745
4746                 /* prefix route could have been deleted; if so restore it */
4747                 if (rc == -ENOENT) {
4748                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4749                                               ifp->rt_priority, ifp->idev->dev,
4750                                               expires, flags, GFP_KERNEL);
4751                 }
4752
4753                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4754                         rc = modify_prefix_route(ifp, expires, flags, true);
4755
4756                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4757                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4758                                               ifp->rt_priority, ifp->idev->dev,
4759                                               expires, flags, GFP_KERNEL);
4760                 }
4761         } else if (had_prefixroute) {
4762                 enum cleanup_prefix_rt_t action;
4763                 unsigned long rt_expires;
4764
4765                 write_lock_bh(&ifp->idev->lock);
4766                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4767                 write_unlock_bh(&ifp->idev->lock);
4768
4769                 if (action != CLEANUP_PREFIX_RT_NOP) {
4770                         cleanup_prefix_route(ifp, rt_expires,
4771                                 action == CLEANUP_PREFIX_RT_DEL, false);
4772                 }
4773         }
4774
4775         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4776                 if (was_managetempaddr &&
4777                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4778                         cfg->valid_lft = 0;
4779                         cfg->preferred_lft = 0;
4780                 }
4781                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4782                                  cfg->preferred_lft, !was_managetempaddr,
4783                                  jiffies);
4784         }
4785
4786         addrconf_verify_rtnl();
4787
4788         return 0;
4789 }
4790
4791 static int
4792 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4793                   struct netlink_ext_ack *extack)
4794 {
4795         struct net *net = sock_net(skb->sk);
4796         struct ifaddrmsg *ifm;
4797         struct nlattr *tb[IFA_MAX+1];
4798         struct in6_addr *peer_pfx;
4799         struct inet6_ifaddr *ifa;
4800         struct net_device *dev;
4801         struct inet6_dev *idev;
4802         struct ifa6_config cfg;
4803         int err;
4804
4805         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4806                                      ifa_ipv6_policy, extack);
4807         if (err < 0)
4808                 return err;
4809
4810         memset(&cfg, 0, sizeof(cfg));
4811
4812         ifm = nlmsg_data(nlh);
4813         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4814         if (!cfg.pfx)
4815                 return -EINVAL;
4816
4817         cfg.peer_pfx = peer_pfx;
4818         cfg.plen = ifm->ifa_prefixlen;
4819         if (tb[IFA_RT_PRIORITY])
4820                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4821
4822         cfg.valid_lft = INFINITY_LIFE_TIME;
4823         cfg.preferred_lft = INFINITY_LIFE_TIME;
4824
4825         if (tb[IFA_CACHEINFO]) {
4826                 struct ifa_cacheinfo *ci;
4827
4828                 ci = nla_data(tb[IFA_CACHEINFO]);
4829                 cfg.valid_lft = ci->ifa_valid;
4830                 cfg.preferred_lft = ci->ifa_prefered;
4831         }
4832
4833         dev =  __dev_get_by_index(net, ifm->ifa_index);
4834         if (!dev)
4835                 return -ENODEV;
4836
4837         if (tb[IFA_FLAGS])
4838                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4839         else
4840                 cfg.ifa_flags = ifm->ifa_flags;
4841
4842         /* We ignore other flags so far. */
4843         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4844                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4845                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4846
4847         idev = ipv6_find_idev(dev);
4848         if (IS_ERR(idev))
4849                 return PTR_ERR(idev);
4850
4851         if (!ipv6_allow_optimistic_dad(net, idev))
4852                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4853
4854         if (cfg.ifa_flags & IFA_F_NODAD &&
4855             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4856                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4857                 return -EINVAL;
4858         }
4859
4860         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4861         if (!ifa) {
4862                 /*
4863                  * It would be best to check for !NLM_F_CREATE here but
4864                  * userspace already relies on not having to provide this.
4865                  */
4866                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4867         }
4868
4869         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4870             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4871                 err = -EEXIST;
4872         else
4873                 err = inet6_addr_modify(ifa, &cfg);
4874
4875         in6_ifa_put(ifa);
4876
4877         return err;
4878 }
4879
4880 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4881                           u8 scope, int ifindex)
4882 {
4883         struct ifaddrmsg *ifm;
4884
4885         ifm = nlmsg_data(nlh);
4886         ifm->ifa_family = AF_INET6;
4887         ifm->ifa_prefixlen = prefixlen;
4888         ifm->ifa_flags = flags;
4889         ifm->ifa_scope = scope;
4890         ifm->ifa_index = ifindex;
4891 }
4892
4893 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4894                          unsigned long tstamp, u32 preferred, u32 valid)
4895 {
4896         struct ifa_cacheinfo ci;
4897
4898         ci.cstamp = cstamp_delta(cstamp);
4899         ci.tstamp = cstamp_delta(tstamp);
4900         ci.ifa_prefered = preferred;
4901         ci.ifa_valid = valid;
4902
4903         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4904 }
4905
4906 static inline int rt_scope(int ifa_scope)
4907 {
4908         if (ifa_scope & IFA_HOST)
4909                 return RT_SCOPE_HOST;
4910         else if (ifa_scope & IFA_LINK)
4911                 return RT_SCOPE_LINK;
4912         else if (ifa_scope & IFA_SITE)
4913                 return RT_SCOPE_SITE;
4914         else
4915                 return RT_SCOPE_UNIVERSE;
4916 }
4917
4918 static inline int inet6_ifaddr_msgsize(void)
4919 {
4920         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4921                + nla_total_size(16) /* IFA_LOCAL */
4922                + nla_total_size(16) /* IFA_ADDRESS */
4923                + nla_total_size(sizeof(struct ifa_cacheinfo))
4924                + nla_total_size(4)  /* IFA_FLAGS */
4925                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4926 }
4927
4928 enum addr_type_t {
4929         UNICAST_ADDR,
4930         MULTICAST_ADDR,
4931         ANYCAST_ADDR,
4932 };
4933
4934 struct inet6_fill_args {
4935         u32 portid;
4936         u32 seq;
4937         int event;
4938         unsigned int flags;
4939         int netnsid;
4940         int ifindex;
4941         enum addr_type_t type;
4942 };
4943
4944 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4945                              struct inet6_fill_args *args)
4946 {
4947         struct nlmsghdr  *nlh;
4948         u32 preferred, valid;
4949
4950         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4951                         sizeof(struct ifaddrmsg), args->flags);
4952         if (!nlh)
4953                 return -EMSGSIZE;
4954
4955         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4956                       ifa->idev->dev->ifindex);
4957
4958         if (args->netnsid >= 0 &&
4959             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4960                 goto error;
4961
4962         if (!((ifa->flags&IFA_F_PERMANENT) &&
4963               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4964                 preferred = ifa->prefered_lft;
4965                 valid = ifa->valid_lft;
4966                 if (preferred != INFINITY_LIFE_TIME) {
4967                         long tval = (jiffies - ifa->tstamp)/HZ;
4968                         if (preferred > tval)
4969                                 preferred -= tval;
4970                         else
4971                                 preferred = 0;
4972                         if (valid != INFINITY_LIFE_TIME) {
4973                                 if (valid > tval)
4974                                         valid -= tval;
4975                                 else
4976                                         valid = 0;
4977                         }
4978                 }
4979         } else {
4980                 preferred = INFINITY_LIFE_TIME;
4981                 valid = INFINITY_LIFE_TIME;
4982         }
4983
4984         if (!ipv6_addr_any(&ifa->peer_addr)) {
4985                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4986                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4987                         goto error;
4988         } else
4989                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4990                         goto error;
4991
4992         if (ifa->rt_priority &&
4993             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4994                 goto error;
4995
4996         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4997                 goto error;
4998
4999         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5000                 goto error;
5001
5002         nlmsg_end(skb, nlh);
5003         return 0;
5004
5005 error:
5006         nlmsg_cancel(skb, nlh);
5007         return -EMSGSIZE;
5008 }
5009
5010 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5011                                struct inet6_fill_args *args)
5012 {
5013         struct nlmsghdr  *nlh;
5014         u8 scope = RT_SCOPE_UNIVERSE;
5015         int ifindex = ifmca->idev->dev->ifindex;
5016
5017         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5018                 scope = RT_SCOPE_SITE;
5019
5020         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5021                         sizeof(struct ifaddrmsg), args->flags);
5022         if (!nlh)
5023                 return -EMSGSIZE;
5024
5025         if (args->netnsid >= 0 &&
5026             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5027                 nlmsg_cancel(skb, nlh);
5028                 return -EMSGSIZE;
5029         }
5030
5031         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5032         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5033             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5034                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5035                 nlmsg_cancel(skb, nlh);
5036                 return -EMSGSIZE;
5037         }
5038
5039         nlmsg_end(skb, nlh);
5040         return 0;
5041 }
5042
5043 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5044                                struct inet6_fill_args *args)
5045 {
5046         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5047         int ifindex = dev ? dev->ifindex : 1;
5048         struct nlmsghdr  *nlh;
5049         u8 scope = RT_SCOPE_UNIVERSE;
5050
5051         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5052                 scope = RT_SCOPE_SITE;
5053
5054         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5055                         sizeof(struct ifaddrmsg), args->flags);
5056         if (!nlh)
5057                 return -EMSGSIZE;
5058
5059         if (args->netnsid >= 0 &&
5060             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5061                 nlmsg_cancel(skb, nlh);
5062                 return -EMSGSIZE;
5063         }
5064
5065         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5066         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5067             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5068                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5069                 nlmsg_cancel(skb, nlh);
5070                 return -EMSGSIZE;
5071         }
5072
5073         nlmsg_end(skb, nlh);
5074         return 0;
5075 }
5076
5077 /* called with rcu_read_lock() */
5078 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5079                           struct netlink_callback *cb, int s_ip_idx,
5080                           struct inet6_fill_args *fillargs)
5081 {
5082         struct ifmcaddr6 *ifmca;
5083         struct ifacaddr6 *ifaca;
5084         int ip_idx = 0;
5085         int err = 1;
5086
5087         read_lock_bh(&idev->lock);
5088         switch (fillargs->type) {
5089         case UNICAST_ADDR: {
5090                 struct inet6_ifaddr *ifa;
5091                 fillargs->event = RTM_NEWADDR;
5092
5093                 /* unicast address incl. temp addr */
5094                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5095                         if (ip_idx < s_ip_idx)
5096                                 goto next;
5097                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5098                         if (err < 0)
5099                                 break;
5100                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5101 next:
5102                         ip_idx++;
5103                 }
5104                 break;
5105         }
5106         case MULTICAST_ADDR:
5107                 fillargs->event = RTM_GETMULTICAST;
5108
5109                 /* multicast address */
5110                 for (ifmca = idev->mc_list; ifmca;
5111                      ifmca = ifmca->next, ip_idx++) {
5112                         if (ip_idx < s_ip_idx)
5113                                 continue;
5114                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5115                         if (err < 0)
5116                                 break;
5117                 }
5118                 break;
5119         case ANYCAST_ADDR:
5120                 fillargs->event = RTM_GETANYCAST;
5121                 /* anycast address */
5122                 for (ifaca = idev->ac_list; ifaca;
5123                      ifaca = ifaca->aca_next, ip_idx++) {
5124                         if (ip_idx < s_ip_idx)
5125                                 continue;
5126                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5127                         if (err < 0)
5128                                 break;
5129                 }
5130                 break;
5131         default:
5132                 break;
5133         }
5134         read_unlock_bh(&idev->lock);
5135         cb->args[2] = ip_idx;
5136         return err;
5137 }
5138
5139 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5140                                        struct inet6_fill_args *fillargs,
5141                                        struct net **tgt_net, struct sock *sk,
5142                                        struct netlink_callback *cb)
5143 {
5144         struct netlink_ext_ack *extack = cb->extack;
5145         struct nlattr *tb[IFA_MAX+1];
5146         struct ifaddrmsg *ifm;
5147         int err, i;
5148
5149         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5150                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5151                 return -EINVAL;
5152         }
5153
5154         ifm = nlmsg_data(nlh);
5155         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5156                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5157                 return -EINVAL;
5158         }
5159
5160         fillargs->ifindex = ifm->ifa_index;
5161         if (fillargs->ifindex) {
5162                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5163                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5164         }
5165
5166         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5167                                             ifa_ipv6_policy, extack);
5168         if (err < 0)
5169                 return err;
5170
5171         for (i = 0; i <= IFA_MAX; ++i) {
5172                 if (!tb[i])
5173                         continue;
5174
5175                 if (i == IFA_TARGET_NETNSID) {
5176                         struct net *net;
5177
5178                         fillargs->netnsid = nla_get_s32(tb[i]);
5179                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5180                         if (IS_ERR(net)) {
5181                                 fillargs->netnsid = -1;
5182                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5183                                 return PTR_ERR(net);
5184                         }
5185                         *tgt_net = net;
5186                 } else {
5187                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5188                         return -EINVAL;
5189                 }
5190         }
5191
5192         return 0;
5193 }
5194
5195 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5196                            enum addr_type_t type)
5197 {
5198         const struct nlmsghdr *nlh = cb->nlh;
5199         struct inet6_fill_args fillargs = {
5200                 .portid = NETLINK_CB(cb->skb).portid,
5201                 .seq = cb->nlh->nlmsg_seq,
5202                 .flags = NLM_F_MULTI,
5203                 .netnsid = -1,
5204                 .type = type,
5205         };
5206         struct net *net = sock_net(skb->sk);
5207         struct net *tgt_net = net;
5208         int idx, s_idx, s_ip_idx;
5209         int h, s_h;
5210         struct net_device *dev;
5211         struct inet6_dev *idev;
5212         struct hlist_head *head;
5213         int err = 0;
5214
5215         s_h = cb->args[0];
5216         s_idx = idx = cb->args[1];
5217         s_ip_idx = cb->args[2];
5218
5219         if (cb->strict_check) {
5220                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5221                                                   skb->sk, cb);
5222                 if (err < 0)
5223                         goto put_tgt_net;
5224
5225                 err = 0;
5226                 if (fillargs.ifindex) {
5227                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5228                         if (!dev) {
5229                                 err = -ENODEV;
5230                                 goto put_tgt_net;
5231                         }
5232                         idev = __in6_dev_get(dev);
5233                         if (idev) {
5234                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5235                                                      &fillargs);
5236                                 if (err > 0)
5237                                         err = 0;
5238                         }
5239                         goto put_tgt_net;
5240                 }
5241         }
5242
5243         rcu_read_lock();
5244         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5245         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5246                 idx = 0;
5247                 head = &tgt_net->dev_index_head[h];
5248                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5249                         if (idx < s_idx)
5250                                 goto cont;
5251                         if (h > s_h || idx > s_idx)
5252                                 s_ip_idx = 0;
5253                         idev = __in6_dev_get(dev);
5254                         if (!idev)
5255                                 goto cont;
5256
5257                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5258                                            &fillargs) < 0)
5259                                 goto done;
5260 cont:
5261                         idx++;
5262                 }
5263         }
5264 done:
5265         rcu_read_unlock();
5266         cb->args[0] = h;
5267         cb->args[1] = idx;
5268 put_tgt_net:
5269         if (fillargs.netnsid >= 0)
5270                 put_net(tgt_net);
5271
5272         return skb->len ? : err;
5273 }
5274
5275 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5276 {
5277         enum addr_type_t type = UNICAST_ADDR;
5278
5279         return inet6_dump_addr(skb, cb, type);
5280 }
5281
5282 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5283 {
5284         enum addr_type_t type = MULTICAST_ADDR;
5285
5286         return inet6_dump_addr(skb, cb, type);
5287 }
5288
5289
5290 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5291 {
5292         enum addr_type_t type = ANYCAST_ADDR;
5293
5294         return inet6_dump_addr(skb, cb, type);
5295 }
5296
5297 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5298                                        const struct nlmsghdr *nlh,
5299                                        struct nlattr **tb,
5300                                        struct netlink_ext_ack *extack)
5301 {
5302         struct ifaddrmsg *ifm;
5303         int i, err;
5304
5305         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5306                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5307                 return -EINVAL;
5308         }
5309
5310         if (!netlink_strict_get_check(skb))
5311                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5312                                               ifa_ipv6_policy, extack);
5313
5314         ifm = nlmsg_data(nlh);
5315         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5316                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5317                 return -EINVAL;
5318         }
5319
5320         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5321                                             ifa_ipv6_policy, extack);
5322         if (err)
5323                 return err;
5324
5325         for (i = 0; i <= IFA_MAX; i++) {
5326                 if (!tb[i])
5327                         continue;
5328
5329                 switch (i) {
5330                 case IFA_TARGET_NETNSID:
5331                 case IFA_ADDRESS:
5332                 case IFA_LOCAL:
5333                         break;
5334                 default:
5335                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5336                         return -EINVAL;
5337                 }
5338         }
5339
5340         return 0;
5341 }
5342
5343 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5344                              struct netlink_ext_ack *extack)
5345 {
5346         struct net *net = sock_net(in_skb->sk);
5347         struct inet6_fill_args fillargs = {
5348                 .portid = NETLINK_CB(in_skb).portid,
5349                 .seq = nlh->nlmsg_seq,
5350                 .event = RTM_NEWADDR,
5351                 .flags = 0,
5352                 .netnsid = -1,
5353         };
5354         struct net *tgt_net = net;
5355         struct ifaddrmsg *ifm;
5356         struct nlattr *tb[IFA_MAX+1];
5357         struct in6_addr *addr = NULL, *peer;
5358         struct net_device *dev = NULL;
5359         struct inet6_ifaddr *ifa;
5360         struct sk_buff *skb;
5361         int err;
5362
5363         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5364         if (err < 0)
5365                 return err;
5366
5367         if (tb[IFA_TARGET_NETNSID]) {
5368                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5369
5370                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5371                                                   fillargs.netnsid);
5372                 if (IS_ERR(tgt_net))
5373                         return PTR_ERR(tgt_net);
5374         }
5375
5376         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5377         if (!addr)
5378                 return -EINVAL;
5379
5380         ifm = nlmsg_data(nlh);
5381         if (ifm->ifa_index)
5382                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5383
5384         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5385         if (!ifa) {
5386                 err = -EADDRNOTAVAIL;
5387                 goto errout;
5388         }
5389
5390         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5391         if (!skb) {
5392                 err = -ENOBUFS;
5393                 goto errout_ifa;
5394         }
5395
5396         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5397         if (err < 0) {
5398                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5399                 WARN_ON(err == -EMSGSIZE);
5400                 kfree_skb(skb);
5401                 goto errout_ifa;
5402         }
5403         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5404 errout_ifa:
5405         in6_ifa_put(ifa);
5406 errout:
5407         if (dev)
5408                 dev_put(dev);
5409         if (fillargs.netnsid >= 0)
5410                 put_net(tgt_net);
5411
5412         return err;
5413 }
5414
5415 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5416 {
5417         struct sk_buff *skb;
5418         struct net *net = dev_net(ifa->idev->dev);
5419         struct inet6_fill_args fillargs = {
5420                 .portid = 0,
5421                 .seq = 0,
5422                 .event = event,
5423                 .flags = 0,
5424                 .netnsid = -1,
5425         };
5426         int err = -ENOBUFS;
5427
5428         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5429         if (!skb)
5430                 goto errout;
5431
5432         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5433         if (err < 0) {
5434                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5435                 WARN_ON(err == -EMSGSIZE);
5436                 kfree_skb(skb);
5437                 goto errout;
5438         }
5439         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5440         return;
5441 errout:
5442         if (err < 0)
5443                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5444 }
5445
5446 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5447                                 __s32 *array, int bytes)
5448 {
5449         BUG_ON(bytes < (DEVCONF_MAX * 4));
5450
5451         memset(array, 0, bytes);
5452         array[DEVCONF_FORWARDING] = cnf->forwarding;
5453         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5454         array[DEVCONF_MTU6] = cnf->mtu6;
5455         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5456         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5457         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5458         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5459         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5460         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5461                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5462         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5463                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5464         array[DEVCONF_RTR_SOLICIT_DELAY] =
5465                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5466         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5467         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5468                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5469         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5470                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5471         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5472         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5473         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5474         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5475         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5476         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5477         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5478         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5479         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5480 #ifdef CONFIG_IPV6_ROUTER_PREF
5481         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5482         array[DEVCONF_RTR_PROBE_INTERVAL] =
5483                 jiffies_to_msecs(cnf->rtr_probe_interval);
5484 #ifdef CONFIG_IPV6_ROUTE_INFO
5485         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5486         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5487 #endif
5488 #endif
5489         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5490         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5491 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5492         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5493         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5494 #endif
5495 #ifdef CONFIG_IPV6_MROUTE
5496         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5497 #endif
5498         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5499         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5500         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5501         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5502         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5503         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5504         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5505         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5506         /* we omit DEVCONF_STABLE_SECRET for now */
5507         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5508         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5509         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5510         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5511         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5512 #ifdef CONFIG_IPV6_SEG6_HMAC
5513         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5514 #endif
5515         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5516         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5517         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5518         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5519         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5520 }
5521
5522 static inline size_t inet6_ifla6_size(void)
5523 {
5524         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5525              + nla_total_size(sizeof(struct ifla_cacheinfo))
5526              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5527              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5528              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5529              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5530              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5531              + 0;
5532 }
5533
5534 static inline size_t inet6_if_nlmsg_size(void)
5535 {
5536         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5537                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5538                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5539                + nla_total_size(4) /* IFLA_MTU */
5540                + nla_total_size(4) /* IFLA_LINK */
5541                + nla_total_size(1) /* IFLA_OPERSTATE */
5542                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5543 }
5544
5545 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5546                                         int bytes)
5547 {
5548         int i;
5549         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5550         BUG_ON(pad < 0);
5551
5552         /* Use put_unaligned() because stats may not be aligned for u64. */
5553         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5554         for (i = 1; i < ICMP6_MIB_MAX; i++)
5555                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5556
5557         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5558 }
5559
5560 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5561                                         int bytes, size_t syncpoff)
5562 {
5563         int i, c;
5564         u64 buff[IPSTATS_MIB_MAX];
5565         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5566
5567         BUG_ON(pad < 0);
5568
5569         memset(buff, 0, sizeof(buff));
5570         buff[0] = IPSTATS_MIB_MAX;
5571
5572         for_each_possible_cpu(c) {
5573                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5574                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5575         }
5576
5577         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5578         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5579 }
5580
5581 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5582                              int bytes)
5583 {
5584         switch (attrtype) {
5585         case IFLA_INET6_STATS:
5586                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5587                                      offsetof(struct ipstats_mib, syncp));
5588                 break;
5589         case IFLA_INET6_ICMP6STATS:
5590                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5591                 break;
5592         }
5593 }
5594
5595 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5596                                   u32 ext_filter_mask)
5597 {
5598         struct nlattr *nla;
5599         struct ifla_cacheinfo ci;
5600
5601         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5602                 goto nla_put_failure;
5603         ci.max_reasm_len = IPV6_MAXPLEN;
5604         ci.tstamp = cstamp_delta(idev->tstamp);
5605         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5606         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5607         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5608                 goto nla_put_failure;
5609         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5610         if (!nla)
5611                 goto nla_put_failure;
5612         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5613
5614         /* XXX - MC not implemented */
5615
5616         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5617                 return 0;
5618
5619         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5620         if (!nla)
5621                 goto nla_put_failure;
5622         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5623
5624         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5625         if (!nla)
5626                 goto nla_put_failure;
5627         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5628
5629         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5630         if (!nla)
5631                 goto nla_put_failure;
5632         read_lock_bh(&idev->lock);
5633         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5634         read_unlock_bh(&idev->lock);
5635
5636         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5637                 goto nla_put_failure;
5638
5639         return 0;
5640
5641 nla_put_failure:
5642         return -EMSGSIZE;
5643 }
5644
5645 static size_t inet6_get_link_af_size(const struct net_device *dev,
5646                                      u32 ext_filter_mask)
5647 {
5648         if (!__in6_dev_get(dev))
5649                 return 0;
5650
5651         return inet6_ifla6_size();
5652 }
5653
5654 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5655                               u32 ext_filter_mask)
5656 {
5657         struct inet6_dev *idev = __in6_dev_get(dev);
5658
5659         if (!idev)
5660                 return -ENODATA;
5661
5662         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5663                 return -EMSGSIZE;
5664
5665         return 0;
5666 }
5667
5668 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5669 {
5670         struct inet6_ifaddr *ifp;
5671         struct net_device *dev = idev->dev;
5672         bool clear_token, update_rs = false;
5673         struct in6_addr ll_addr;
5674
5675         ASSERT_RTNL();
5676
5677         if (!token)
5678                 return -EINVAL;
5679         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5680                 return -EINVAL;
5681         if (!ipv6_accept_ra(idev))
5682                 return -EINVAL;
5683         if (idev->cnf.rtr_solicits == 0)
5684                 return -EINVAL;
5685
5686         write_lock_bh(&idev->lock);
5687
5688         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5689         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5690
5691         write_unlock_bh(&idev->lock);
5692
5693         clear_token = ipv6_addr_any(token);
5694         if (clear_token)
5695                 goto update_lft;
5696
5697         if (!idev->dead && (idev->if_flags & IF_READY) &&
5698             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5699                              IFA_F_OPTIMISTIC)) {
5700                 /* If we're not ready, then normal ifup will take care
5701                  * of this. Otherwise, we need to request our rs here.
5702                  */
5703                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5704                 update_rs = true;
5705         }
5706
5707 update_lft:
5708         write_lock_bh(&idev->lock);
5709
5710         if (update_rs) {
5711                 idev->if_flags |= IF_RS_SENT;
5712                 idev->rs_interval = rfc3315_s14_backoff_init(
5713                         idev->cnf.rtr_solicit_interval);
5714                 idev->rs_probes = 1;
5715                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5716         }
5717
5718         /* Well, that's kinda nasty ... */
5719         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5720                 spin_lock(&ifp->lock);
5721                 if (ifp->tokenized) {
5722                         ifp->valid_lft = 0;
5723                         ifp->prefered_lft = 0;
5724                 }
5725                 spin_unlock(&ifp->lock);
5726         }
5727
5728         write_unlock_bh(&idev->lock);
5729         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5730         addrconf_verify_rtnl();
5731         return 0;
5732 }
5733
5734 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5735         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5736         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5737 };
5738
5739 static int check_addr_gen_mode(int mode)
5740 {
5741         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5742             mode != IN6_ADDR_GEN_MODE_NONE &&
5743             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5744             mode != IN6_ADDR_GEN_MODE_RANDOM)
5745                 return -EINVAL;
5746         return 1;
5747 }
5748
5749 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5750                                 int mode)
5751 {
5752         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5753             !idev->cnf.stable_secret.initialized &&
5754             !net->ipv6.devconf_dflt->stable_secret.initialized)
5755                 return -EINVAL;
5756         return 1;
5757 }
5758
5759 static int inet6_validate_link_af(const struct net_device *dev,
5760                                   const struct nlattr *nla)
5761 {
5762         struct nlattr *tb[IFLA_INET6_MAX + 1];
5763         struct inet6_dev *idev = NULL;
5764         int err;
5765
5766         if (dev) {
5767                 idev = __in6_dev_get(dev);
5768                 if (!idev)
5769                         return -EAFNOSUPPORT;
5770         }
5771
5772         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5773                                           inet6_af_policy, NULL);
5774         if (err)
5775                 return err;
5776
5777         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5778                 return -EINVAL;
5779
5780         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5781                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5782
5783                 if (check_addr_gen_mode(mode) < 0)
5784                         return -EINVAL;
5785                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5786                         return -EINVAL;
5787         }
5788
5789         return 0;
5790 }
5791
5792 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5793 {
5794         struct inet6_dev *idev = __in6_dev_get(dev);
5795         struct nlattr *tb[IFLA_INET6_MAX + 1];
5796         int err;
5797
5798         if (!idev)
5799                 return -EAFNOSUPPORT;
5800
5801         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5802                 BUG();
5803
5804         if (tb[IFLA_INET6_TOKEN]) {
5805                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5806                 if (err)
5807                         return err;
5808         }
5809
5810         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5811                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5812
5813                 idev->cnf.addr_gen_mode = mode;
5814         }
5815
5816         return 0;
5817 }
5818
5819 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5820                              u32 portid, u32 seq, int event, unsigned int flags)
5821 {
5822         struct net_device *dev = idev->dev;
5823         struct ifinfomsg *hdr;
5824         struct nlmsghdr *nlh;
5825         void *protoinfo;
5826
5827         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5828         if (!nlh)
5829                 return -EMSGSIZE;
5830
5831         hdr = nlmsg_data(nlh);
5832         hdr->ifi_family = AF_INET6;
5833         hdr->__ifi_pad = 0;
5834         hdr->ifi_type = dev->type;
5835         hdr->ifi_index = dev->ifindex;
5836         hdr->ifi_flags = dev_get_flags(dev);
5837         hdr->ifi_change = 0;
5838
5839         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5840             (dev->addr_len &&
5841              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5842             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5843             (dev->ifindex != dev_get_iflink(dev) &&
5844              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5845             nla_put_u8(skb, IFLA_OPERSTATE,
5846                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5847                 goto nla_put_failure;
5848         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5849         if (!protoinfo)
5850                 goto nla_put_failure;
5851
5852         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5853                 goto nla_put_failure;
5854
5855         nla_nest_end(skb, protoinfo);
5856         nlmsg_end(skb, nlh);
5857         return 0;
5858
5859 nla_put_failure:
5860         nlmsg_cancel(skb, nlh);
5861         return -EMSGSIZE;
5862 }
5863
5864 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5865                                    struct netlink_ext_ack *extack)
5866 {
5867         struct ifinfomsg *ifm;
5868
5869         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5870                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5871                 return -EINVAL;
5872         }
5873
5874         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5875                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5876                 return -EINVAL;
5877         }
5878
5879         ifm = nlmsg_data(nlh);
5880         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5881             ifm->ifi_change || ifm->ifi_index) {
5882                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5883                 return -EINVAL;
5884         }
5885
5886         return 0;
5887 }
5888
5889 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5890 {
5891         struct net *net = sock_net(skb->sk);
5892         int h, s_h;
5893         int idx = 0, s_idx;
5894         struct net_device *dev;
5895         struct inet6_dev *idev;
5896         struct hlist_head *head;
5897
5898         /* only requests using strict checking can pass data to
5899          * influence the dump
5900          */
5901         if (cb->strict_check) {
5902                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5903
5904                 if (err < 0)
5905                         return err;
5906         }
5907
5908         s_h = cb->args[0];
5909         s_idx = cb->args[1];
5910
5911         rcu_read_lock();
5912         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5913                 idx = 0;
5914                 head = &net->dev_index_head[h];
5915                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5916                         if (idx < s_idx)
5917                                 goto cont;
5918                         idev = __in6_dev_get(dev);
5919                         if (!idev)
5920                                 goto cont;
5921                         if (inet6_fill_ifinfo(skb, idev,
5922                                               NETLINK_CB(cb->skb).portid,
5923                                               cb->nlh->nlmsg_seq,
5924                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5925                                 goto out;
5926 cont:
5927                         idx++;
5928                 }
5929         }
5930 out:
5931         rcu_read_unlock();
5932         cb->args[1] = idx;
5933         cb->args[0] = h;
5934
5935         return skb->len;
5936 }
5937
5938 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5939 {
5940         struct sk_buff *skb;
5941         struct net *net = dev_net(idev->dev);
5942         int err = -ENOBUFS;
5943
5944         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5945         if (!skb)
5946                 goto errout;
5947
5948         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5949         if (err < 0) {
5950                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5951                 WARN_ON(err == -EMSGSIZE);
5952                 kfree_skb(skb);
5953                 goto errout;
5954         }
5955         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5956         return;
5957 errout:
5958         if (err < 0)
5959                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5960 }
5961
5962 static inline size_t inet6_prefix_nlmsg_size(void)
5963 {
5964         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5965                + nla_total_size(sizeof(struct in6_addr))
5966                + nla_total_size(sizeof(struct prefix_cacheinfo));
5967 }
5968
5969 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5970                              struct prefix_info *pinfo, u32 portid, u32 seq,
5971                              int event, unsigned int flags)
5972 {
5973         struct prefixmsg *pmsg;
5974         struct nlmsghdr *nlh;
5975         struct prefix_cacheinfo ci;
5976
5977         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5978         if (!nlh)
5979                 return -EMSGSIZE;
5980
5981         pmsg = nlmsg_data(nlh);
5982         pmsg->prefix_family = AF_INET6;
5983         pmsg->prefix_pad1 = 0;
5984         pmsg->prefix_pad2 = 0;
5985         pmsg->prefix_ifindex = idev->dev->ifindex;
5986         pmsg->prefix_len = pinfo->prefix_len;
5987         pmsg->prefix_type = pinfo->type;
5988         pmsg->prefix_pad3 = 0;
5989         pmsg->prefix_flags = 0;
5990         if (pinfo->onlink)
5991                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5992         if (pinfo->autoconf)
5993                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5994
5995         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5996                 goto nla_put_failure;
5997         ci.preferred_time = ntohl(pinfo->prefered);
5998         ci.valid_time = ntohl(pinfo->valid);
5999         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6000                 goto nla_put_failure;
6001         nlmsg_end(skb, nlh);
6002         return 0;
6003
6004 nla_put_failure:
6005         nlmsg_cancel(skb, nlh);
6006         return -EMSGSIZE;
6007 }
6008
6009 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6010                          struct prefix_info *pinfo)
6011 {
6012         struct sk_buff *skb;
6013         struct net *net = dev_net(idev->dev);
6014         int err = -ENOBUFS;
6015
6016         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6017         if (!skb)
6018                 goto errout;
6019
6020         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6021         if (err < 0) {
6022                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6023                 WARN_ON(err == -EMSGSIZE);
6024                 kfree_skb(skb);
6025                 goto errout;
6026         }
6027         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6028         return;
6029 errout:
6030         if (err < 0)
6031                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6032 }
6033
6034 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6035 {
6036         struct net *net = dev_net(ifp->idev->dev);
6037
6038         if (event)
6039                 ASSERT_RTNL();
6040
6041         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6042
6043         switch (event) {
6044         case RTM_NEWADDR:
6045                 /*
6046                  * If the address was optimistic we inserted the route at the
6047                  * start of our DAD process, so we don't need to do it again.
6048                  * If the device was taken down in the middle of the DAD
6049                  * cycle there is a race where we could get here without a
6050                  * host route, so nothing to insert. That will be fixed when
6051                  * the device is brought up.
6052                  */
6053                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6054                         ip6_ins_rt(net, ifp->rt);
6055                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6056                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6057                                 &ifp->addr, ifp->idev->dev->name);
6058                 }
6059
6060                 if (ifp->idev->cnf.forwarding)
6061                         addrconf_join_anycast(ifp);
6062                 if (!ipv6_addr_any(&ifp->peer_addr))
6063                         addrconf_prefix_route(&ifp->peer_addr, 128,
6064                                               ifp->rt_priority, ifp->idev->dev,
6065                                               0, 0, GFP_ATOMIC);
6066                 break;
6067         case RTM_DELADDR:
6068                 if (ifp->idev->cnf.forwarding)
6069                         addrconf_leave_anycast(ifp);
6070                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6071                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6072                         struct fib6_info *rt;
6073
6074                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6075                                                        ifp->idev->dev, 0, 0,
6076                                                        false);
6077                         if (rt)
6078                                 ip6_del_rt(net, rt, false);
6079                 }
6080                 if (ifp->rt) {
6081                         ip6_del_rt(net, ifp->rt, false);
6082                         ifp->rt = NULL;
6083                 }
6084                 rt_genid_bump_ipv6(net);
6085                 break;
6086         }
6087         atomic_inc(&net->ipv6.dev_addr_genid);
6088 }
6089
6090 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6091 {
6092         rcu_read_lock_bh();
6093         if (likely(ifp->idev->dead == 0))
6094                 __ipv6_ifa_notify(event, ifp);
6095         rcu_read_unlock_bh();
6096 }
6097
6098 #ifdef CONFIG_SYSCTL
6099
6100 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6101                 void *buffer, size_t *lenp, loff_t *ppos)
6102 {
6103         int *valp = ctl->data;
6104         int val = *valp;
6105         loff_t pos = *ppos;
6106         struct ctl_table lctl;
6107         int ret;
6108
6109         /*
6110          * ctl->data points to idev->cnf.forwarding, we should
6111          * not modify it until we get the rtnl lock.
6112          */
6113         lctl = *ctl;
6114         lctl.data = &val;
6115
6116         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6117
6118         if (write)
6119                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6120         if (ret)
6121                 *ppos = pos;
6122         return ret;
6123 }
6124
6125 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6126                 void *buffer, size_t *lenp, loff_t *ppos)
6127 {
6128         struct inet6_dev *idev = ctl->extra1;
6129         int min_mtu = IPV6_MIN_MTU;
6130         struct ctl_table lctl;
6131
6132         lctl = *ctl;
6133         lctl.extra1 = &min_mtu;
6134         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6135
6136         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6137 }
6138
6139 static void dev_disable_change(struct inet6_dev *idev)
6140 {
6141         struct netdev_notifier_info info;
6142
6143         if (!idev || !idev->dev)
6144                 return;
6145
6146         netdev_notifier_info_init(&info, idev->dev);
6147         if (idev->cnf.disable_ipv6)
6148                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6149         else
6150                 addrconf_notify(NULL, NETDEV_UP, &info);
6151 }
6152
6153 static void addrconf_disable_change(struct net *net, __s32 newf)
6154 {
6155         struct net_device *dev;
6156         struct inet6_dev *idev;
6157
6158         for_each_netdev(net, dev) {
6159                 idev = __in6_dev_get(dev);
6160                 if (idev) {
6161                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6162                         idev->cnf.disable_ipv6 = newf;
6163                         if (changed)
6164                                 dev_disable_change(idev);
6165                 }
6166         }
6167 }
6168
6169 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6170 {
6171         struct net *net;
6172         int old;
6173
6174         if (!rtnl_trylock())
6175                 return restart_syscall();
6176
6177         net = (struct net *)table->extra2;
6178         old = *p;
6179         *p = newf;
6180
6181         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6182                 rtnl_unlock();
6183                 return 0;
6184         }
6185
6186         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6187                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6188                 addrconf_disable_change(net, newf);
6189         } else if ((!newf) ^ (!old))
6190                 dev_disable_change((struct inet6_dev *)table->extra1);
6191
6192         rtnl_unlock();
6193         return 0;
6194 }
6195
6196 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6197                 void *buffer, size_t *lenp, loff_t *ppos)
6198 {
6199         int *valp = ctl->data;
6200         int val = *valp;
6201         loff_t pos = *ppos;
6202         struct ctl_table lctl;
6203         int ret;
6204
6205         /*
6206          * ctl->data points to idev->cnf.disable_ipv6, we should
6207          * not modify it until we get the rtnl lock.
6208          */
6209         lctl = *ctl;
6210         lctl.data = &val;
6211
6212         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6213
6214         if (write)
6215                 ret = addrconf_disable_ipv6(ctl, valp, val);
6216         if (ret)
6217                 *ppos = pos;
6218         return ret;
6219 }
6220
6221 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6222                 void *buffer, size_t *lenp, loff_t *ppos)
6223 {
6224         int *valp = ctl->data;
6225         int ret;
6226         int old, new;
6227
6228         old = *valp;
6229         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6230         new = *valp;
6231
6232         if (write && old != new) {
6233                 struct net *net = ctl->extra2;
6234
6235                 if (!rtnl_trylock())
6236                         return restart_syscall();
6237
6238                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6239                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6240                                                      NETCONFA_PROXY_NEIGH,
6241                                                      NETCONFA_IFINDEX_DEFAULT,
6242                                                      net->ipv6.devconf_dflt);
6243                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6244                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6245                                                      NETCONFA_PROXY_NEIGH,
6246                                                      NETCONFA_IFINDEX_ALL,
6247                                                      net->ipv6.devconf_all);
6248                 else {
6249                         struct inet6_dev *idev = ctl->extra1;
6250
6251                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6252                                                      NETCONFA_PROXY_NEIGH,
6253                                                      idev->dev->ifindex,
6254                                                      &idev->cnf);
6255                 }
6256                 rtnl_unlock();
6257         }
6258
6259         return ret;
6260 }
6261
6262 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6263                                          void *buffer, size_t *lenp,
6264                                          loff_t *ppos)
6265 {
6266         int ret = 0;
6267         u32 new_val;
6268         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6269         struct net *net = (struct net *)ctl->extra2;
6270         struct ctl_table tmp = {
6271                 .data = &new_val,
6272                 .maxlen = sizeof(new_val),
6273                 .mode = ctl->mode,
6274         };
6275
6276         if (!rtnl_trylock())
6277                 return restart_syscall();
6278
6279         new_val = *((u32 *)ctl->data);
6280
6281         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6282         if (ret != 0)
6283                 goto out;
6284
6285         if (write) {
6286                 if (check_addr_gen_mode(new_val) < 0) {
6287                         ret = -EINVAL;
6288                         goto out;
6289                 }
6290
6291                 if (idev) {
6292                         if (check_stable_privacy(idev, net, new_val) < 0) {
6293                                 ret = -EINVAL;
6294                                 goto out;
6295                         }
6296
6297                         if (idev->cnf.addr_gen_mode != new_val) {
6298                                 idev->cnf.addr_gen_mode = new_val;
6299                                 addrconf_dev_config(idev->dev);
6300                         }
6301                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6302                         struct net_device *dev;
6303
6304                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6305                         for_each_netdev(net, dev) {
6306                                 idev = __in6_dev_get(dev);
6307                                 if (idev &&
6308                                     idev->cnf.addr_gen_mode != new_val) {
6309                                         idev->cnf.addr_gen_mode = new_val;
6310                                         addrconf_dev_config(idev->dev);
6311                                 }
6312                         }
6313                 }
6314
6315                 *((u32 *)ctl->data) = new_val;
6316         }
6317
6318 out:
6319         rtnl_unlock();
6320
6321         return ret;
6322 }
6323
6324 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6325                                          void *buffer, size_t *lenp,
6326                                          loff_t *ppos)
6327 {
6328         int err;
6329         struct in6_addr addr;
6330         char str[IPV6_MAX_STRLEN];
6331         struct ctl_table lctl = *ctl;
6332         struct net *net = ctl->extra2;
6333         struct ipv6_stable_secret *secret = ctl->data;
6334
6335         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6336                 return -EIO;
6337
6338         lctl.maxlen = IPV6_MAX_STRLEN;
6339         lctl.data = str;
6340
6341         if (!rtnl_trylock())
6342                 return restart_syscall();
6343
6344         if (!write && !secret->initialized) {
6345                 err = -EIO;
6346                 goto out;
6347         }
6348
6349         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6350         if (err >= sizeof(str)) {
6351                 err = -EIO;
6352                 goto out;
6353         }
6354
6355         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6356         if (err || !write)
6357                 goto out;
6358
6359         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6360                 err = -EIO;
6361                 goto out;
6362         }
6363
6364         secret->initialized = true;
6365         secret->secret = addr;
6366
6367         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6368                 struct net_device *dev;
6369
6370                 for_each_netdev(net, dev) {
6371                         struct inet6_dev *idev = __in6_dev_get(dev);
6372
6373                         if (idev) {
6374                                 idev->cnf.addr_gen_mode =
6375                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6376                         }
6377                 }
6378         } else {
6379                 struct inet6_dev *idev = ctl->extra1;
6380
6381                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6382         }
6383
6384 out:
6385         rtnl_unlock();
6386
6387         return err;
6388 }
6389
6390 static
6391 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6392                                                 int write, void *buffer,
6393                                                 size_t *lenp,
6394                                                 loff_t *ppos)
6395 {
6396         int *valp = ctl->data;
6397         int val = *valp;
6398         loff_t pos = *ppos;
6399         struct ctl_table lctl;
6400         int ret;
6401
6402         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6403          * we should not modify it until we get the rtnl lock.
6404          */
6405         lctl = *ctl;
6406         lctl.data = &val;
6407
6408         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6409
6410         if (write)
6411                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6412         if (ret)
6413                 *ppos = pos;
6414         return ret;
6415 }
6416
6417 static
6418 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6419 {
6420         if (rt) {
6421                 if (action)
6422                         rt->dst.flags |= DST_NOPOLICY;
6423                 else
6424                         rt->dst.flags &= ~DST_NOPOLICY;
6425         }
6426 }
6427
6428 static
6429 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6430 {
6431         struct inet6_ifaddr *ifa;
6432
6433         read_lock_bh(&idev->lock);
6434         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6435                 spin_lock(&ifa->lock);
6436                 if (ifa->rt) {
6437                         /* host routes only use builtin fib6_nh */
6438                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6439                         int cpu;
6440
6441                         rcu_read_lock();
6442                         ifa->rt->dst_nopolicy = val ? true : false;
6443                         if (nh->rt6i_pcpu) {
6444                                 for_each_possible_cpu(cpu) {
6445                                         struct rt6_info **rtp;
6446
6447                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6448                                         addrconf_set_nopolicy(*rtp, val);
6449                                 }
6450                         }
6451                         rcu_read_unlock();
6452                 }
6453                 spin_unlock(&ifa->lock);
6454         }
6455         read_unlock_bh(&idev->lock);
6456 }
6457
6458 static
6459 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6460 {
6461         struct inet6_dev *idev;
6462         struct net *net;
6463
6464         if (!rtnl_trylock())
6465                 return restart_syscall();
6466
6467         *valp = val;
6468
6469         net = (struct net *)ctl->extra2;
6470         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6471                 rtnl_unlock();
6472                 return 0;
6473         }
6474
6475         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6476                 struct net_device *dev;
6477
6478                 for_each_netdev(net, dev) {
6479                         idev = __in6_dev_get(dev);
6480                         if (idev)
6481                                 addrconf_disable_policy_idev(idev, val);
6482                 }
6483         } else {
6484                 idev = (struct inet6_dev *)ctl->extra1;
6485                 addrconf_disable_policy_idev(idev, val);
6486         }
6487
6488         rtnl_unlock();
6489         return 0;
6490 }
6491
6492 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6493                                    void *buffer, size_t *lenp, loff_t *ppos)
6494 {
6495         int *valp = ctl->data;
6496         int val = *valp;
6497         loff_t pos = *ppos;
6498         struct ctl_table lctl;
6499         int ret;
6500
6501         lctl = *ctl;
6502         lctl.data = &val;
6503         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6504
6505         if (write && (*valp != val))
6506                 ret = addrconf_disable_policy(ctl, valp, val);
6507
6508         if (ret)
6509                 *ppos = pos;
6510
6511         return ret;
6512 }
6513
6514 static int minus_one = -1;
6515 static const int two_five_five = 255;
6516
6517 static const struct ctl_table addrconf_sysctl[] = {
6518         {
6519                 .procname       = "forwarding",
6520                 .data           = &ipv6_devconf.forwarding,
6521                 .maxlen         = sizeof(int),
6522                 .mode           = 0644,
6523                 .proc_handler   = addrconf_sysctl_forward,
6524         },
6525         {
6526                 .procname       = "hop_limit",
6527                 .data           = &ipv6_devconf.hop_limit,
6528                 .maxlen         = sizeof(int),
6529                 .mode           = 0644,
6530                 .proc_handler   = proc_dointvec_minmax,
6531                 .extra1         = (void *)SYSCTL_ONE,
6532                 .extra2         = (void *)&two_five_five,
6533         },
6534         {
6535                 .procname       = "mtu",
6536                 .data           = &ipv6_devconf.mtu6,
6537                 .maxlen         = sizeof(int),
6538                 .mode           = 0644,
6539                 .proc_handler   = addrconf_sysctl_mtu,
6540         },
6541         {
6542                 .procname       = "accept_ra",
6543                 .data           = &ipv6_devconf.accept_ra,
6544                 .maxlen         = sizeof(int),
6545                 .mode           = 0644,
6546                 .proc_handler   = proc_dointvec,
6547         },
6548         {
6549                 .procname       = "accept_redirects",
6550                 .data           = &ipv6_devconf.accept_redirects,
6551                 .maxlen         = sizeof(int),
6552                 .mode           = 0644,
6553                 .proc_handler   = proc_dointvec,
6554         },
6555         {
6556                 .procname       = "autoconf",
6557                 .data           = &ipv6_devconf.autoconf,
6558                 .maxlen         = sizeof(int),
6559                 .mode           = 0644,
6560                 .proc_handler   = proc_dointvec,
6561         },
6562         {
6563                 .procname       = "dad_transmits",
6564                 .data           = &ipv6_devconf.dad_transmits,
6565                 .maxlen         = sizeof(int),
6566                 .mode           = 0644,
6567                 .proc_handler   = proc_dointvec,
6568         },
6569         {
6570                 .procname       = "router_solicitations",
6571                 .data           = &ipv6_devconf.rtr_solicits,
6572                 .maxlen         = sizeof(int),
6573                 .mode           = 0644,
6574                 .proc_handler   = proc_dointvec_minmax,
6575                 .extra1         = &minus_one,
6576         },
6577         {
6578                 .procname       = "router_solicitation_interval",
6579                 .data           = &ipv6_devconf.rtr_solicit_interval,
6580                 .maxlen         = sizeof(int),
6581                 .mode           = 0644,
6582                 .proc_handler   = proc_dointvec_jiffies,
6583         },
6584         {
6585                 .procname       = "router_solicitation_max_interval",
6586                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6587                 .maxlen         = sizeof(int),
6588                 .mode           = 0644,
6589                 .proc_handler   = proc_dointvec_jiffies,
6590         },
6591         {
6592                 .procname       = "router_solicitation_delay",
6593                 .data           = &ipv6_devconf.rtr_solicit_delay,
6594                 .maxlen         = sizeof(int),
6595                 .mode           = 0644,
6596                 .proc_handler   = proc_dointvec_jiffies,
6597         },
6598         {
6599                 .procname       = "force_mld_version",
6600                 .data           = &ipv6_devconf.force_mld_version,
6601                 .maxlen         = sizeof(int),
6602                 .mode           = 0644,
6603                 .proc_handler   = proc_dointvec,
6604         },
6605         {
6606                 .procname       = "mldv1_unsolicited_report_interval",
6607                 .data           =
6608                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6609                 .maxlen         = sizeof(int),
6610                 .mode           = 0644,
6611                 .proc_handler   = proc_dointvec_ms_jiffies,
6612         },
6613         {
6614                 .procname       = "mldv2_unsolicited_report_interval",
6615                 .data           =
6616                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6617                 .maxlen         = sizeof(int),
6618                 .mode           = 0644,
6619                 .proc_handler   = proc_dointvec_ms_jiffies,
6620         },
6621         {
6622                 .procname       = "use_tempaddr",
6623                 .data           = &ipv6_devconf.use_tempaddr,
6624                 .maxlen         = sizeof(int),
6625                 .mode           = 0644,
6626                 .proc_handler   = proc_dointvec,
6627         },
6628         {
6629                 .procname       = "temp_valid_lft",
6630                 .data           = &ipv6_devconf.temp_valid_lft,
6631                 .maxlen         = sizeof(int),
6632                 .mode           = 0644,
6633                 .proc_handler   = proc_dointvec,
6634         },
6635         {
6636                 .procname       = "temp_prefered_lft",
6637                 .data           = &ipv6_devconf.temp_prefered_lft,
6638                 .maxlen         = sizeof(int),
6639                 .mode           = 0644,
6640                 .proc_handler   = proc_dointvec,
6641         },
6642         {
6643                 .procname       = "regen_max_retry",
6644                 .data           = &ipv6_devconf.regen_max_retry,
6645                 .maxlen         = sizeof(int),
6646                 .mode           = 0644,
6647                 .proc_handler   = proc_dointvec,
6648         },
6649         {
6650                 .procname       = "max_desync_factor",
6651                 .data           = &ipv6_devconf.max_desync_factor,
6652                 .maxlen         = sizeof(int),
6653                 .mode           = 0644,
6654                 .proc_handler   = proc_dointvec,
6655         },
6656         {
6657                 .procname       = "max_addresses",
6658                 .data           = &ipv6_devconf.max_addresses,
6659                 .maxlen         = sizeof(int),
6660                 .mode           = 0644,
6661                 .proc_handler   = proc_dointvec,
6662         },
6663         {
6664                 .procname       = "accept_ra_defrtr",
6665                 .data           = &ipv6_devconf.accept_ra_defrtr,
6666                 .maxlen         = sizeof(int),
6667                 .mode           = 0644,
6668                 .proc_handler   = proc_dointvec,
6669         },
6670         {
6671                 .procname       = "accept_ra_min_hop_limit",
6672                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6673                 .maxlen         = sizeof(int),
6674                 .mode           = 0644,
6675                 .proc_handler   = proc_dointvec,
6676         },
6677         {
6678                 .procname       = "accept_ra_pinfo",
6679                 .data           = &ipv6_devconf.accept_ra_pinfo,
6680                 .maxlen         = sizeof(int),
6681                 .mode           = 0644,
6682                 .proc_handler   = proc_dointvec,
6683         },
6684 #ifdef CONFIG_IPV6_ROUTER_PREF
6685         {
6686                 .procname       = "accept_ra_rtr_pref",
6687                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6688                 .maxlen         = sizeof(int),
6689                 .mode           = 0644,
6690                 .proc_handler   = proc_dointvec,
6691         },
6692         {
6693                 .procname       = "router_probe_interval",
6694                 .data           = &ipv6_devconf.rtr_probe_interval,
6695                 .maxlen         = sizeof(int),
6696                 .mode           = 0644,
6697                 .proc_handler   = proc_dointvec_jiffies,
6698         },
6699 #ifdef CONFIG_IPV6_ROUTE_INFO
6700         {
6701                 .procname       = "accept_ra_rt_info_min_plen",
6702                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6703                 .maxlen         = sizeof(int),
6704                 .mode           = 0644,
6705                 .proc_handler   = proc_dointvec,
6706         },
6707         {
6708                 .procname       = "accept_ra_rt_info_max_plen",
6709                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6710                 .maxlen         = sizeof(int),
6711                 .mode           = 0644,
6712                 .proc_handler   = proc_dointvec,
6713         },
6714 #endif
6715 #endif
6716         {
6717                 .procname       = "proxy_ndp",
6718                 .data           = &ipv6_devconf.proxy_ndp,
6719                 .maxlen         = sizeof(int),
6720                 .mode           = 0644,
6721                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6722         },
6723         {
6724                 .procname       = "accept_source_route",
6725                 .data           = &ipv6_devconf.accept_source_route,
6726                 .maxlen         = sizeof(int),
6727                 .mode           = 0644,
6728                 .proc_handler   = proc_dointvec,
6729         },
6730 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6731         {
6732                 .procname       = "optimistic_dad",
6733                 .data           = &ipv6_devconf.optimistic_dad,
6734                 .maxlen         = sizeof(int),
6735                 .mode           = 0644,
6736                 .proc_handler   = proc_dointvec,
6737         },
6738         {
6739                 .procname       = "use_optimistic",
6740                 .data           = &ipv6_devconf.use_optimistic,
6741                 .maxlen         = sizeof(int),
6742                 .mode           = 0644,
6743                 .proc_handler   = proc_dointvec,
6744         },
6745 #endif
6746 #ifdef CONFIG_IPV6_MROUTE
6747         {
6748                 .procname       = "mc_forwarding",
6749                 .data           = &ipv6_devconf.mc_forwarding,
6750                 .maxlen         = sizeof(int),
6751                 .mode           = 0444,
6752                 .proc_handler   = proc_dointvec,
6753         },
6754 #endif
6755         {
6756                 .procname       = "disable_ipv6",
6757                 .data           = &ipv6_devconf.disable_ipv6,
6758                 .maxlen         = sizeof(int),
6759                 .mode           = 0644,
6760                 .proc_handler   = addrconf_sysctl_disable,
6761         },
6762         {
6763                 .procname       = "accept_dad",
6764                 .data           = &ipv6_devconf.accept_dad,
6765                 .maxlen         = sizeof(int),
6766                 .mode           = 0644,
6767                 .proc_handler   = proc_dointvec,
6768         },
6769         {
6770                 .procname       = "force_tllao",
6771                 .data           = &ipv6_devconf.force_tllao,
6772                 .maxlen         = sizeof(int),
6773                 .mode           = 0644,
6774                 .proc_handler   = proc_dointvec
6775         },
6776         {
6777                 .procname       = "ndisc_notify",
6778                 .data           = &ipv6_devconf.ndisc_notify,
6779                 .maxlen         = sizeof(int),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_dointvec
6782         },
6783         {
6784                 .procname       = "suppress_frag_ndisc",
6785                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6786                 .maxlen         = sizeof(int),
6787                 .mode           = 0644,
6788                 .proc_handler   = proc_dointvec
6789         },
6790         {
6791                 .procname       = "accept_ra_from_local",
6792                 .data           = &ipv6_devconf.accept_ra_from_local,
6793                 .maxlen         = sizeof(int),
6794                 .mode           = 0644,
6795                 .proc_handler   = proc_dointvec,
6796         },
6797         {
6798                 .procname       = "accept_ra_mtu",
6799                 .data           = &ipv6_devconf.accept_ra_mtu,
6800                 .maxlen         = sizeof(int),
6801                 .mode           = 0644,
6802                 .proc_handler   = proc_dointvec,
6803         },
6804         {
6805                 .procname       = "stable_secret",
6806                 .data           = &ipv6_devconf.stable_secret,
6807                 .maxlen         = IPV6_MAX_STRLEN,
6808                 .mode           = 0600,
6809                 .proc_handler   = addrconf_sysctl_stable_secret,
6810         },
6811         {
6812                 .procname       = "use_oif_addrs_only",
6813                 .data           = &ipv6_devconf.use_oif_addrs_only,
6814                 .maxlen         = sizeof(int),
6815                 .mode           = 0644,
6816                 .proc_handler   = proc_dointvec,
6817         },
6818         {
6819                 .procname       = "ignore_routes_with_linkdown",
6820                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6821                 .maxlen         = sizeof(int),
6822                 .mode           = 0644,
6823                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6824         },
6825         {
6826                 .procname       = "drop_unicast_in_l2_multicast",
6827                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6828                 .maxlen         = sizeof(int),
6829                 .mode           = 0644,
6830                 .proc_handler   = proc_dointvec,
6831         },
6832         {
6833                 .procname       = "drop_unsolicited_na",
6834                 .data           = &ipv6_devconf.drop_unsolicited_na,
6835                 .maxlen         = sizeof(int),
6836                 .mode           = 0644,
6837                 .proc_handler   = proc_dointvec,
6838         },
6839         {
6840                 .procname       = "keep_addr_on_down",
6841                 .data           = &ipv6_devconf.keep_addr_on_down,
6842                 .maxlen         = sizeof(int),
6843                 .mode           = 0644,
6844                 .proc_handler   = proc_dointvec,
6845
6846         },
6847         {
6848                 .procname       = "seg6_enabled",
6849                 .data           = &ipv6_devconf.seg6_enabled,
6850                 .maxlen         = sizeof(int),
6851                 .mode           = 0644,
6852                 .proc_handler   = proc_dointvec,
6853         },
6854 #ifdef CONFIG_IPV6_SEG6_HMAC
6855         {
6856                 .procname       = "seg6_require_hmac",
6857                 .data           = &ipv6_devconf.seg6_require_hmac,
6858                 .maxlen         = sizeof(int),
6859                 .mode           = 0644,
6860                 .proc_handler   = proc_dointvec,
6861         },
6862 #endif
6863         {
6864                 .procname       = "enhanced_dad",
6865                 .data           = &ipv6_devconf.enhanced_dad,
6866                 .maxlen         = sizeof(int),
6867                 .mode           = 0644,
6868                 .proc_handler   = proc_dointvec,
6869         },
6870         {
6871                 .procname               = "addr_gen_mode",
6872                 .data                   = &ipv6_devconf.addr_gen_mode,
6873                 .maxlen                 = sizeof(int),
6874                 .mode                   = 0644,
6875                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6876         },
6877         {
6878                 .procname       = "disable_policy",
6879                 .data           = &ipv6_devconf.disable_policy,
6880                 .maxlen         = sizeof(int),
6881                 .mode           = 0644,
6882                 .proc_handler   = addrconf_sysctl_disable_policy,
6883         },
6884         {
6885                 .procname       = "ndisc_tclass",
6886                 .data           = &ipv6_devconf.ndisc_tclass,
6887                 .maxlen         = sizeof(int),
6888                 .mode           = 0644,
6889                 .proc_handler   = proc_dointvec_minmax,
6890                 .extra1         = (void *)SYSCTL_ZERO,
6891                 .extra2         = (void *)&two_five_five,
6892         },
6893         {
6894                 .procname       = "rpl_seg_enabled",
6895                 .data           = &ipv6_devconf.rpl_seg_enabled,
6896                 .maxlen         = sizeof(int),
6897                 .mode           = 0644,
6898                 .proc_handler   = proc_dointvec,
6899         },
6900         {
6901                 /* sentinel */
6902         }
6903 };
6904
6905 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6906                 struct inet6_dev *idev, struct ipv6_devconf *p)
6907 {
6908         int i, ifindex;
6909         struct ctl_table *table;
6910         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6911
6912         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6913         if (!table)
6914                 goto out;
6915
6916         for (i = 0; table[i].data; i++) {
6917                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6918                 /* If one of these is already set, then it is not safe to
6919                  * overwrite either of them: this makes proc_dointvec_minmax
6920                  * usable.
6921                  */
6922                 if (!table[i].extra1 && !table[i].extra2) {
6923                         table[i].extra1 = idev; /* embedded; no ref */
6924                         table[i].extra2 = net;
6925                 }
6926         }
6927
6928         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6929
6930         p->sysctl_header = register_net_sysctl(net, path, table);
6931         if (!p->sysctl_header)
6932                 goto free;
6933
6934         if (!strcmp(dev_name, "all"))
6935                 ifindex = NETCONFA_IFINDEX_ALL;
6936         else if (!strcmp(dev_name, "default"))
6937                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6938         else
6939                 ifindex = idev->dev->ifindex;
6940         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6941                                      ifindex, p);
6942         return 0;
6943
6944 free:
6945         kfree(table);
6946 out:
6947         return -ENOBUFS;
6948 }
6949
6950 static void __addrconf_sysctl_unregister(struct net *net,
6951                                          struct ipv6_devconf *p, int ifindex)
6952 {
6953         struct ctl_table *table;
6954
6955         if (!p->sysctl_header)
6956                 return;
6957
6958         table = p->sysctl_header->ctl_table_arg;
6959         unregister_net_sysctl_table(p->sysctl_header);
6960         p->sysctl_header = NULL;
6961         kfree(table);
6962
6963         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6964 }
6965
6966 static int addrconf_sysctl_register(struct inet6_dev *idev)
6967 {
6968         int err;
6969
6970         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6971                 return -EINVAL;
6972
6973         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6974                                     &ndisc_ifinfo_sysctl_change);
6975         if (err)
6976                 return err;
6977         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6978                                          idev, &idev->cnf);
6979         if (err)
6980                 neigh_sysctl_unregister(idev->nd_parms);
6981
6982         return err;
6983 }
6984
6985 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6986 {
6987         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6988                                      idev->dev->ifindex);
6989         neigh_sysctl_unregister(idev->nd_parms);
6990 }
6991
6992
6993 #endif
6994
6995 static int __net_init addrconf_init_net(struct net *net)
6996 {
6997         int err = -ENOMEM;
6998         struct ipv6_devconf *all, *dflt;
6999
7000         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7001         if (!all)
7002                 goto err_alloc_all;
7003
7004         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7005         if (!dflt)
7006                 goto err_alloc_dflt;
7007
7008         if (IS_ENABLED(CONFIG_SYSCTL) &&
7009             !net_eq(net, &init_net)) {
7010                 switch (sysctl_devconf_inherit_init_net) {
7011                 case 1:  /* copy from init_net */
7012                         memcpy(all, init_net.ipv6.devconf_all,
7013                                sizeof(ipv6_devconf));
7014                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7015                                sizeof(ipv6_devconf_dflt));
7016                         break;
7017                 case 3: /* copy from the current netns */
7018                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7019                                sizeof(ipv6_devconf));
7020                         memcpy(dflt,
7021                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7022                                sizeof(ipv6_devconf_dflt));
7023                         break;
7024                 case 0:
7025                 case 2:
7026                         /* use compiled values */
7027                         break;
7028                 }
7029         }
7030
7031         /* these will be inherited by all namespaces */
7032         dflt->autoconf = ipv6_defaults.autoconf;
7033         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7034
7035         dflt->stable_secret.initialized = false;
7036         all->stable_secret.initialized = false;
7037
7038         net->ipv6.devconf_all = all;
7039         net->ipv6.devconf_dflt = dflt;
7040
7041 #ifdef CONFIG_SYSCTL
7042         err = __addrconf_sysctl_register(net, "all", NULL, all);
7043         if (err < 0)
7044                 goto err_reg_all;
7045
7046         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7047         if (err < 0)
7048                 goto err_reg_dflt;
7049 #endif
7050         return 0;
7051
7052 #ifdef CONFIG_SYSCTL
7053 err_reg_dflt:
7054         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7055 err_reg_all:
7056         kfree(dflt);
7057 #endif
7058 err_alloc_dflt:
7059         kfree(all);
7060 err_alloc_all:
7061         return err;
7062 }
7063
7064 static void __net_exit addrconf_exit_net(struct net *net)
7065 {
7066 #ifdef CONFIG_SYSCTL
7067         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7068                                      NETCONFA_IFINDEX_DEFAULT);
7069         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7070                                      NETCONFA_IFINDEX_ALL);
7071 #endif
7072         kfree(net->ipv6.devconf_dflt);
7073         kfree(net->ipv6.devconf_all);
7074 }
7075
7076 static struct pernet_operations addrconf_ops = {
7077         .init = addrconf_init_net,
7078         .exit = addrconf_exit_net,
7079 };
7080
7081 static struct rtnl_af_ops inet6_ops __read_mostly = {
7082         .family           = AF_INET6,
7083         .fill_link_af     = inet6_fill_link_af,
7084         .get_link_af_size = inet6_get_link_af_size,
7085         .validate_link_af = inet6_validate_link_af,
7086         .set_link_af      = inet6_set_link_af,
7087 };
7088
7089 /*
7090  *      Init / cleanup code
7091  */
7092
7093 int __init addrconf_init(void)
7094 {
7095         struct inet6_dev *idev;
7096         int i, err;
7097
7098         err = ipv6_addr_label_init();
7099         if (err < 0) {
7100                 pr_crit("%s: cannot initialize default policy table: %d\n",
7101                         __func__, err);
7102                 goto out;
7103         }
7104
7105         err = register_pernet_subsys(&addrconf_ops);
7106         if (err < 0)
7107                 goto out_addrlabel;
7108
7109         addrconf_wq = create_workqueue("ipv6_addrconf");
7110         if (!addrconf_wq) {
7111                 err = -ENOMEM;
7112                 goto out_nowq;
7113         }
7114
7115         /* The addrconf netdev notifier requires that loopback_dev
7116          * has it's ipv6 private information allocated and setup
7117          * before it can bring up and give link-local addresses
7118          * to other devices which are up.
7119          *
7120          * Unfortunately, loopback_dev is not necessarily the first
7121          * entry in the global dev_base list of net devices.  In fact,
7122          * it is likely to be the very last entry on that list.
7123          * So this causes the notifier registry below to try and
7124          * give link-local addresses to all devices besides loopback_dev
7125          * first, then loopback_dev, which cases all the non-loopback_dev
7126          * devices to fail to get a link-local address.
7127          *
7128          * So, as a temporary fix, allocate the ipv6 structure for
7129          * loopback_dev first by hand.
7130          * Longer term, all of the dependencies ipv6 has upon the loopback
7131          * device and it being up should be removed.
7132          */
7133         rtnl_lock();
7134         idev = ipv6_add_dev(init_net.loopback_dev);
7135         rtnl_unlock();
7136         if (IS_ERR(idev)) {
7137                 err = PTR_ERR(idev);
7138                 goto errlo;
7139         }
7140
7141         ip6_route_init_special_entries();
7142
7143         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7144                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7145
7146         register_netdevice_notifier(&ipv6_dev_notf);
7147
7148         addrconf_verify();
7149
7150         rtnl_af_register(&inet6_ops);
7151
7152         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7153                                    NULL, inet6_dump_ifinfo, 0);
7154         if (err < 0)
7155                 goto errout;
7156
7157         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7158                                    inet6_rtm_newaddr, NULL, 0);
7159         if (err < 0)
7160                 goto errout;
7161         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7162                                    inet6_rtm_deladdr, NULL, 0);
7163         if (err < 0)
7164                 goto errout;
7165         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7166                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7167                                    RTNL_FLAG_DOIT_UNLOCKED);
7168         if (err < 0)
7169                 goto errout;
7170         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7171                                    NULL, inet6_dump_ifmcaddr, 0);
7172         if (err < 0)
7173                 goto errout;
7174         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7175                                    NULL, inet6_dump_ifacaddr, 0);
7176         if (err < 0)
7177                 goto errout;
7178         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7179                                    inet6_netconf_get_devconf,
7180                                    inet6_netconf_dump_devconf,
7181                                    RTNL_FLAG_DOIT_UNLOCKED);
7182         if (err < 0)
7183                 goto errout;
7184         err = ipv6_addr_label_rtnl_register();
7185         if (err < 0)
7186                 goto errout;
7187
7188         return 0;
7189 errout:
7190         rtnl_unregister_all(PF_INET6);
7191         rtnl_af_unregister(&inet6_ops);
7192         unregister_netdevice_notifier(&ipv6_dev_notf);
7193 errlo:
7194         destroy_workqueue(addrconf_wq);
7195 out_nowq:
7196         unregister_pernet_subsys(&addrconf_ops);
7197 out_addrlabel:
7198         ipv6_addr_label_cleanup();
7199 out:
7200         return err;
7201 }
7202
7203 void addrconf_cleanup(void)
7204 {
7205         struct net_device *dev;
7206         int i;
7207
7208         unregister_netdevice_notifier(&ipv6_dev_notf);
7209         unregister_pernet_subsys(&addrconf_ops);
7210         ipv6_addr_label_cleanup();
7211
7212         rtnl_af_unregister(&inet6_ops);
7213
7214         rtnl_lock();
7215
7216         /* clean dev list */
7217         for_each_netdev(&init_net, dev) {
7218                 if (__in6_dev_get(dev) == NULL)
7219                         continue;
7220                 addrconf_ifdown(dev, true);
7221         }
7222         addrconf_ifdown(init_net.loopback_dev, true);
7223
7224         /*
7225          *      Check hash table.
7226          */
7227         spin_lock_bh(&addrconf_hash_lock);
7228         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7229                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7230         spin_unlock_bh(&addrconf_hash_lock);
7231         cancel_delayed_work(&addr_chk_work);
7232         rtnl_unlock();
7233
7234         destroy_workqueue(addrconf_wq);
7235 }