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             <[email protected]>
   8  *      Alexey Kuznetsov        <[email protected]>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <[email protected]>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92
  93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  94
  95 #define IPV6_MAX_STRLEN \
  96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  97
  98 static inline u32 cstamp_delta(unsigned long cstamp)
  99 {
 100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 101 }
 102
 103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 104 {
 105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 107         do_div(tmp, 1000000);
 108         return (s32)tmp;
 109 }
 110
 111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 112 {
 113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 119                 do_div(tmp, 1000000);
 120         }
 121         return (s32)tmp;
 122 }
 123
 124 #ifdef CONFIG_SYSCTL
 125 static int addrconf_sysctl_register(struct inet6_dev *idev);
 126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 127 #else
 128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 129 {
 130         return 0;
 131 }
 132
 133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 134 {
 135 }
 136 #endif
 137
 138 static void ipv6_regen_rndid(struct inet6_dev *idev);
 139 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 140
 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 142 static int ipv6_count_addresses(const struct inet6_dev *idev);
 143 static int ipv6_generate_stable_address(struct in6_addr *addr,
 144                                         u8 dad_count,
 145                                         const struct inet6_dev *idev);
 146
 147 #define IN6_ADDR_HSIZE_SHIFT    8
 148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 149 /*
 150  *      Configured unicast address hash table
 151  */
 152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 153 static DEFINE_SPINLOCK(addrconf_hash_lock);
 154
 155 static void addrconf_verify(void);
 156 static void addrconf_verify_rtnl(void);
 157 static void addrconf_verify_work(struct work_struct *);
 158
 159 static struct workqueue_struct *addrconf_wq;
 160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 161
 162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 164
 165 static void addrconf_type_change(struct net_device *dev,
 166                                  unsigned long event);
 167 static int addrconf_ifdown(struct net_device *dev, int how);
 168
 169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 170                                                   int plen,
 171                                                   const struct net_device *dev,
 172                                                   u32 flags, u32 noflags,
 173                                                   bool no_gw);
 174
 175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 176 static void addrconf_dad_work(struct work_struct *w);
 177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 178                                    bool send_na);
 179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 180 static void addrconf_rs_timer(struct timer_list *t);
 181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 183
 184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 185                                 struct prefix_info *pinfo);
 186
 187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 188         .forwarding             = 0,
 189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 190         .mtu6                   = IPV6_MIN_MTU,
 191         .accept_ra              = 1,
 192         .accept_redirects       = 1,
 193         .autoconf               = 1,
 194         .force_mld_version      = 0,
 195         .mldv1_unsolicited_report_interval = 10 * HZ,
 196         .mldv2_unsolicited_report_interval = HZ,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .accept_ra_from_local   = 0,
 210         .accept_ra_min_hop_limit= 1,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_min_plen = 0,
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 0,
 224         .suppress_frag_ndisc    = 1,
 225         .accept_ra_mtu          = 1,
 226         .stable_secret          = {
 227                 .initialized = false,
 228         },
 229         .use_oif_addrs_only     = 0,
 230         .ignore_routes_with_linkdown = 0,
 231         .keep_addr_on_down      = 0,
 232         .seg6_enabled           = 0,
 233 #ifdef CONFIG_IPV6_SEG6_HMAC
 234         .seg6_require_hmac      = 0,
 235 #endif
 236         .enhanced_dad           = 1,
 237         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 238         .disable_policy         = 0,
 239 };
 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 };
 294
 295 /* Check if link is ready: is it up and is a valid qdisc available */
 296 static inline bool addrconf_link_ready(const struct net_device *dev)
 297 {
 298         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 299 }
 300
 301 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 302 {
 303         if (del_timer(&idev->rs_timer))
 304                 __in6_dev_put(idev);
 305 }
 306
 307 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 308 {
 309         if (cancel_delayed_work(&ifp->dad_work))
 310                 __in6_ifa_put(ifp);
 311 }
 312
 313 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 314                                   unsigned long when)
 315 {
 316         if (!timer_pending(&idev->rs_timer))
 317                 in6_dev_hold(idev);
 318         mod_timer(&idev->rs_timer, jiffies + when);
 319 }
 320
 321 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 322                                    unsigned long delay)
 323 {
 324         in6_ifa_hold(ifp);
 325         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 326                 in6_ifa_put(ifp);
 327 }
 328
 329 static int snmp6_alloc_dev(struct inet6_dev *idev)
 330 {
 331         int i;
 332
 333         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 334         if (!idev->stats.ipv6)
 335                 goto err_ip;
 336
 337         for_each_possible_cpu(i) {
 338                 struct ipstats_mib *addrconf_stats;
 339                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 340                 u64_stats_init(&addrconf_stats->syncp);
 341         }
 342
 343
 344         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 345                                         GFP_KERNEL);
 346         if (!idev->stats.icmpv6dev)
 347                 goto err_icmp;
 348         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 349                                            GFP_KERNEL);
 350         if (!idev->stats.icmpv6msgdev)
 351                 goto err_icmpmsg;
 352
 353         return 0;
 354
 355 err_icmpmsg:
 356         kfree(idev->stats.icmpv6dev);
 357 err_icmp:
 358         free_percpu(idev->stats.ipv6);
 359 err_ip:
 360         return -ENOMEM;
 361 }
 362
 363 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 364 {
 365         struct inet6_dev *ndev;
 366         int err = -ENOMEM;
 367
 368         ASSERT_RTNL();
 369
 370         if (dev->mtu < IPV6_MIN_MTU)
 371                 return ERR_PTR(-EINVAL);
 372
 373         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 374         if (!ndev)
 375                 return ERR_PTR(err);
 376
 377         rwlock_init(&ndev->lock);
 378         ndev->dev = dev;
 379         INIT_LIST_HEAD(&ndev->addr_list);
 380         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 381         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 382
 383         if (ndev->cnf.stable_secret.initialized)
 384                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 385
 386         ndev->cnf.mtu6 = dev->mtu;
 387         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 388         if (!ndev->nd_parms) {
 389                 kfree(ndev);
 390                 return ERR_PTR(err);
 391         }
 392         if (ndev->cnf.forwarding)
 393                 dev_disable_lro(dev);
 394         /* We refer to the device */
 395         dev_hold(dev);
 396
 397         if (snmp6_alloc_dev(ndev) < 0) {
 398                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 399                            __func__);
 400                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 401                 dev_put(dev);
 402                 kfree(ndev);
 403                 return ERR_PTR(err);
 404         }
 405
 406         if (snmp6_register_dev(ndev) < 0) {
 407                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 408                            __func__, dev->name);
 409                 goto err_release;
 410         }
 411
 412         /* One reference from device. */
 413         refcount_set(&ndev->refcnt, 1);
 414
 415         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 416                 ndev->cnf.accept_dad = -1;
 417
 418 #if IS_ENABLED(CONFIG_IPV6_SIT)
 419         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 420                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 421                 ndev->cnf.rtr_solicits = 0;
 422         }
 423 #endif
 424
 425         INIT_LIST_HEAD(&ndev->tempaddr_list);
 426         ndev->desync_factor = U32_MAX;
 427         if ((dev->flags&IFF_LOOPBACK) ||
 428             dev->type == ARPHRD_TUNNEL ||
 429             dev->type == ARPHRD_TUNNEL6 ||
 430             dev->type == ARPHRD_SIT ||
 431             dev->type == ARPHRD_NONE) {
 432                 ndev->cnf.use_tempaddr = -1;
 433         } else
 434                 ipv6_regen_rndid(ndev);
 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);
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,
1308                                 struct inet6_ifaddr *ift,
1309                                 bool block)
1310 {
1311         struct inet6_dev *idev = ifp->idev;
1312         struct in6_addr addr, *tmpaddr;
1313         unsigned long tmp_tstamp, age;
1314         unsigned long regen_advance;
1315         struct ifa6_config cfg;
1316         int ret = 0;
1317         unsigned long now = jiffies;
1318         long max_desync_factor;
1319         s32 cnf_temp_preferred_lft;
1320
1321         write_lock_bh(&idev->lock);
1322         if (ift) {
1323                 spin_lock_bh(&ift->lock);
1324                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1325                 spin_unlock_bh(&ift->lock);
1326                 tmpaddr = &addr;
1327         } else {
1328                 tmpaddr = NULL;
1329         }
1330 retry:
1331         in6_dev_hold(idev);
1332         if (idev->cnf.use_tempaddr <= 0) {
1333                 write_unlock_bh(&idev->lock);
1334                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1335                 in6_dev_put(idev);
1336                 ret = -1;
1337                 goto out;
1338         }
1339         spin_lock_bh(&ifp->lock);
1340         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1341                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1342                 spin_unlock_bh(&ifp->lock);
1343                 write_unlock_bh(&idev->lock);
1344                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1345                         __func__);
1346                 in6_dev_put(idev);
1347                 ret = -1;
1348                 goto out;
1349         }
1350         in6_ifa_hold(ifp);
1351         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1352         ipv6_try_regen_rndid(idev, tmpaddr);
1353         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1354         age = (now - ifp->tstamp) / HZ;
1355
1356         regen_advance = idev->cnf.regen_max_retry *
1357                         idev->cnf.dad_transmits *
1358                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1359
1360         /* recalculate max_desync_factor each time and update
1361          * idev->desync_factor if it's larger
1362          */
1363         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1364         max_desync_factor = min_t(__u32,
1365                                   idev->cnf.max_desync_factor,
1366                                   cnf_temp_preferred_lft - regen_advance);
1367
1368         if (unlikely(idev->desync_factor > max_desync_factor)) {
1369                 if (max_desync_factor > 0) {
1370                         get_random_bytes(&idev->desync_factor,
1371                                          sizeof(idev->desync_factor));
1372                         idev->desync_factor %= max_desync_factor;
1373                 } else {
1374                         idev->desync_factor = 0;
1375                 }
1376         }
1377
1378         memset(&cfg, 0, sizeof(cfg));
1379         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1380                               idev->cnf.temp_valid_lft + age);
1381         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1382         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1383
1384         cfg.plen = ifp->prefix_len;
1385         tmp_tstamp = ifp->tstamp;
1386         spin_unlock_bh(&ifp->lock);
1387
1388         write_unlock_bh(&idev->lock);
1389
1390         /* A temporary address is created only if this calculated Preferred
1391          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1392          * an implementation must not create a temporary address with a zero
1393          * Preferred Lifetime.
1394          * Use age calculation as in addrconf_verify to avoid unnecessary
1395          * temporary addresses being generated.
1396          */
1397         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1398         if (cfg.preferred_lft <= regen_advance + age) {
1399                 in6_ifa_put(ifp);
1400                 in6_dev_put(idev);
1401                 ret = -1;
1402                 goto out;
1403         }
1404
1405         cfg.ifa_flags = IFA_F_TEMPORARY;
1406         /* set in addrconf_prefix_rcv() */
1407         if (ifp->flags & IFA_F_OPTIMISTIC)
1408                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1409
1410         cfg.pfx = &addr;
1411         cfg.scope = ipv6_addr_scope(cfg.pfx);
1412
1413         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1414         if (IS_ERR(ift)) {
1415                 in6_ifa_put(ifp);
1416                 in6_dev_put(idev);
1417                 pr_info("%s: retry temporary address regeneration\n", __func__);
1418                 tmpaddr = &addr;
1419                 write_lock_bh(&idev->lock);
1420                 goto retry;
1421         }
1422
1423         spin_lock_bh(&ift->lock);
1424         ift->ifpub = ifp;
1425         ift->cstamp = now;
1426         ift->tstamp = tmp_tstamp;
1427         spin_unlock_bh(&ift->lock);
1428
1429         addrconf_dad_start(ift);
1430         in6_ifa_put(ift);
1431         in6_dev_put(idev);
1432 out:
1433         return ret;
1434 }
1435
1436 /*
1437  *      Choose an appropriate source address (RFC3484)
1438  */
1439 enum {
1440         IPV6_SADDR_RULE_INIT = 0,
1441         IPV6_SADDR_RULE_LOCAL,
1442         IPV6_SADDR_RULE_SCOPE,
1443         IPV6_SADDR_RULE_PREFERRED,
1444 #ifdef CONFIG_IPV6_MIP6
1445         IPV6_SADDR_RULE_HOA,
1446 #endif
1447         IPV6_SADDR_RULE_OIF,
1448         IPV6_SADDR_RULE_LABEL,
1449         IPV6_SADDR_RULE_PRIVACY,
1450         IPV6_SADDR_RULE_ORCHID,
1451         IPV6_SADDR_RULE_PREFIX,
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1454 #endif
1455         IPV6_SADDR_RULE_MAX
1456 };
1457
1458 struct ipv6_saddr_score {
1459         int                     rule;
1460         int                     addr_type;
1461         struct inet6_ifaddr     *ifa;
1462         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1463         int                     scopedist;
1464         int                     matchlen;
1465 };
1466
1467 struct ipv6_saddr_dst {
1468         const struct in6_addr *addr;
1469         int ifindex;
1470         int scope;
1471         int label;
1472         unsigned int prefs;
1473 };
1474
1475 static inline int ipv6_saddr_preferred(int type)
1476 {
1477         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1478                 return 1;
1479         return 0;
1480 }
1481
1482 static bool ipv6_use_optimistic_addr(struct net *net,
1483                                      struct inet6_dev *idev)
1484 {
1485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486         if (!idev)
1487                 return false;
1488         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1489                 return false;
1490         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1491                 return false;
1492
1493         return true;
1494 #else
1495         return false;
1496 #endif
1497 }
1498
1499 static bool ipv6_allow_optimistic_dad(struct net *net,
1500                                       struct inet6_dev *idev)
1501 {
1502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1503         if (!idev)
1504                 return false;
1505         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1506                 return false;
1507
1508         return true;
1509 #else
1510         return false;
1511 #endif
1512 }
1513
1514 static int ipv6_get_saddr_eval(struct net *net,
1515                                struct ipv6_saddr_score *score,
1516                                struct ipv6_saddr_dst *dst,
1517                                int i)
1518 {
1519         int ret;
1520
1521         if (i <= score->rule) {
1522                 switch (i) {
1523                 case IPV6_SADDR_RULE_SCOPE:
1524                         ret = score->scopedist;
1525                         break;
1526                 case IPV6_SADDR_RULE_PREFIX:
1527                         ret = score->matchlen;
1528                         break;
1529                 default:
1530                         ret = !!test_bit(i, score->scorebits);
1531                 }
1532                 goto out;
1533         }
1534
1535         switch (i) {
1536         case IPV6_SADDR_RULE_INIT:
1537                 /* Rule 0: remember if hiscore is not ready yet */
1538                 ret = !!score->ifa;
1539                 break;
1540         case IPV6_SADDR_RULE_LOCAL:
1541                 /* Rule 1: Prefer same address */
1542                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1543                 break;
1544         case IPV6_SADDR_RULE_SCOPE:
1545                 /* Rule 2: Prefer appropriate scope
1546                  *
1547                  *      ret
1548                  *       ^
1549                  *    -1 |  d 15
1550                  *    ---+--+-+---> scope
1551                  *       |
1552                  *       |             d is scope of the destination.
1553                  *  B-d  |  \
1554                  *       |   \      <- smaller scope is better if
1555                  *  B-15 |    \        if scope is enough for destination.
1556                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1557                  * d-C-1 | /
1558                  *       |/         <- greater is better
1559                  *   -C  /             if scope is not enough for destination.
1560                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1561                  *
1562                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1563                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1564                  * Assume B = 0 and we get C > 29.
1565                  */
1566                 ret = __ipv6_addr_src_scope(score->addr_type);
1567                 if (ret >= dst->scope)
1568                         ret = -ret;
1569                 else
1570                         ret -= 128;     /* 30 is enough */
1571                 score->scopedist = ret;
1572                 break;
1573         case IPV6_SADDR_RULE_PREFERRED:
1574             {
1575                 /* Rule 3: Avoid deprecated and optimistic addresses */
1576                 u8 avoid = IFA_F_DEPRECATED;
1577
1578                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1579                         avoid |= IFA_F_OPTIMISTIC;
1580                 ret = ipv6_saddr_preferred(score->addr_type) ||
1581                       !(score->ifa->flags & avoid);
1582                 break;
1583             }
1584 #ifdef CONFIG_IPV6_MIP6
1585         case IPV6_SADDR_RULE_HOA:
1586             {
1587                 /* Rule 4: Prefer home address */
1588                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1589                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1590                 break;
1591             }
1592 #endif
1593         case IPV6_SADDR_RULE_OIF:
1594                 /* Rule 5: Prefer outgoing interface */
1595                 ret = (!dst->ifindex ||
1596                        dst->ifindex == score->ifa->idev->dev->ifindex);
1597                 break;
1598         case IPV6_SADDR_RULE_LABEL:
1599                 /* Rule 6: Prefer matching label */
1600                 ret = ipv6_addr_label(net,
1601                                       &score->ifa->addr, score->addr_type,
1602                                       score->ifa->idev->dev->ifindex) == dst->label;
1603                 break;
1604         case IPV6_SADDR_RULE_PRIVACY:
1605             {
1606                 /* Rule 7: Prefer public address
1607                  * Note: prefer temporary address if use_tempaddr >= 2
1608                  */
1609                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1610                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1611                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1612                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1613                 break;
1614             }
1615         case IPV6_SADDR_RULE_ORCHID:
1616                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1617                  *          non-ORCHID vs non-ORCHID
1618                  */
1619                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1620                         ipv6_addr_orchid(dst->addr));
1621                 break;
1622         case IPV6_SADDR_RULE_PREFIX:
1623                 /* Rule 8: Use longest matching prefix */
1624                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1625                 if (ret > score->ifa->prefix_len)
1626                         ret = score->ifa->prefix_len;
1627                 score->matchlen = ret;
1628                 break;
1629 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1630         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1631                 /* Optimistic addresses still have lower precedence than other
1632                  * preferred addresses.
1633                  */
1634                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1635                 break;
1636 #endif
1637         default:
1638                 ret = 0;
1639         }
1640
1641         if (ret)
1642                 __set_bit(i, score->scorebits);
1643         score->rule = i;
1644 out:
1645         return ret;
1646 }
1647
1648 static int __ipv6_dev_get_saddr(struct net *net,
1649                                 struct ipv6_saddr_dst *dst,
1650                                 struct inet6_dev *idev,
1651                                 struct ipv6_saddr_score *scores,
1652                                 int hiscore_idx)
1653 {
1654         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1655
1656         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1657                 int i;
1658
1659                 /*
1660                  * - Tentative Address (RFC2462 section 5.4)
1661                  *  - A tentative address is not considered
1662                  *    "assigned to an interface" in the traditional
1663                  *    sense, unless it is also flagged as optimistic.
1664                  * - Candidate Source Address (section 4)
1665                  *  - In any case, anycast addresses, multicast
1666                  *    addresses, and the unspecified address MUST
1667                  *    NOT be included in a candidate set.
1668                  */
1669                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1670                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1671                         continue;
1672
1673                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1674
1675                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1676                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1677                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1678                                             idev->dev->name);
1679                         continue;
1680                 }
1681
1682                 score->rule = -1;
1683                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1684
1685                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1686                         int minihiscore, miniscore;
1687
1688                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1689                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1690
1691                         if (minihiscore > miniscore) {
1692                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1693                                     score->scopedist > 0) {
1694                                         /*
1695                                          * special case:
1696                                          * each remaining entry
1697                                          * has too small (not enough)
1698                                          * scope, because ifa entries
1699                                          * are sorted by their scope
1700                                          * values.
1701                                          */
1702                                         goto out;
1703                                 }
1704                                 break;
1705                         } else if (minihiscore < miniscore) {
1706                                 swap(hiscore, score);
1707                                 hiscore_idx = 1 - hiscore_idx;
1708
1709                                 /* restore our iterator */
1710                                 score->ifa = hiscore->ifa;
1711
1712                                 break;
1713                         }
1714                 }
1715         }
1716 out:
1717         return hiscore_idx;
1718 }
1719
1720 static int ipv6_get_saddr_master(struct net *net,
1721                                  const struct net_device *dst_dev,
1722                                  const struct net_device *master,
1723                                  struct ipv6_saddr_dst *dst,
1724                                  struct ipv6_saddr_score *scores,
1725                                  int hiscore_idx)
1726 {
1727         struct inet6_dev *idev;
1728
1729         idev = __in6_dev_get(dst_dev);
1730         if (idev)
1731                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1732                                                    scores, hiscore_idx);
1733
1734         idev = __in6_dev_get(master);
1735         if (idev)
1736                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1737                                                    scores, hiscore_idx);
1738
1739         return hiscore_idx;
1740 }
1741
1742 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1743                        const struct in6_addr *daddr, unsigned int prefs,
1744                        struct in6_addr *saddr)
1745 {
1746         struct ipv6_saddr_score scores[2], *hiscore;
1747         struct ipv6_saddr_dst dst;
1748         struct inet6_dev *idev;
1749         struct net_device *dev;
1750         int dst_type;
1751         bool use_oif_addr = false;
1752         int hiscore_idx = 0;
1753         int ret = 0;
1754
1755         dst_type = __ipv6_addr_type(daddr);
1756         dst.addr = daddr;
1757         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1758         dst.scope = __ipv6_addr_src_scope(dst_type);
1759         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1760         dst.prefs = prefs;
1761
1762         scores[hiscore_idx].rule = -1;
1763         scores[hiscore_idx].ifa = NULL;
1764
1765         rcu_read_lock();
1766
1767         /* Candidate Source Address (section 4)
1768          *  - multicast and link-local destination address,
1769          *    the set of candidate source address MUST only
1770          *    include addresses assigned to interfaces
1771          *    belonging to the same link as the outgoing
1772          *    interface.
1773          * (- For site-local destination addresses, the
1774          *    set of candidate source addresses MUST only
1775          *    include addresses assigned to interfaces
1776          *    belonging to the same site as the outgoing
1777          *    interface.)
1778          *  - "It is RECOMMENDED that the candidate source addresses
1779          *    be the set of unicast addresses assigned to the
1780          *    interface that will be used to send to the destination
1781          *    (the 'outgoing' interface)." (RFC 6724)
1782          */
1783         if (dst_dev) {
1784                 idev = __in6_dev_get(dst_dev);
1785                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1786                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1787                     (idev && idev->cnf.use_oif_addrs_only)) {
1788                         use_oif_addr = true;
1789                 }
1790         }
1791
1792         if (use_oif_addr) {
1793                 if (idev)
1794                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1795         } else {
1796                 const struct net_device *master;
1797                 int master_idx = 0;
1798
1799                 /* if dst_dev exists and is enslaved to an L3 device, then
1800                  * prefer addresses from dst_dev and then the master over
1801                  * any other enslaved devices in the L3 domain.
1802                  */
1803                 master = l3mdev_master_dev_rcu(dst_dev);
1804                 if (master) {
1805                         master_idx = master->ifindex;
1806
1807                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1808                                                             master, &dst,
1809                                                             scores, hiscore_idx);
1810
1811                         if (scores[hiscore_idx].ifa)
1812                                 goto out;
1813                 }
1814
1815                 for_each_netdev_rcu(net, dev) {
1816                         /* only consider addresses on devices in the
1817                          * same L3 domain
1818                          */
1819                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1820                                 continue;
1821                         idev = __in6_dev_get(dev);
1822                         if (!idev)
1823                                 continue;
1824                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1825                 }
1826         }
1827
1828 out:
1829         hiscore = &scores[hiscore_idx];
1830         if (!hiscore->ifa)
1831                 ret = -EADDRNOTAVAIL;
1832         else
1833                 *saddr = hiscore->ifa->addr;
1834
1835         rcu_read_unlock();
1836         return ret;
1837 }
1838 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1839
1840 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1841                       u32 banned_flags)
1842 {
1843         struct inet6_ifaddr *ifp;
1844         int err = -EADDRNOTAVAIL;
1845
1846         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1847                 if (ifp->scope > IFA_LINK)
1848                         break;
1849                 if (ifp->scope == IFA_LINK &&
1850                     !(ifp->flags & banned_flags)) {
1851                         *addr = ifp->addr;
1852                         err = 0;
1853                         break;
1854                 }
1855         }
1856         return err;
1857 }
1858
1859 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1860                     u32 banned_flags)
1861 {
1862         struct inet6_dev *idev;
1863         int err = -EADDRNOTAVAIL;
1864
1865         rcu_read_lock();
1866         idev = __in6_dev_get(dev);
1867         if (idev) {
1868                 read_lock_bh(&idev->lock);
1869                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1870                 read_unlock_bh(&idev->lock);
1871         }
1872         rcu_read_unlock();
1873         return err;
1874 }
1875
1876 static int ipv6_count_addresses(const struct inet6_dev *idev)
1877 {
1878         const struct inet6_ifaddr *ifp;
1879         int cnt = 0;
1880
1881         rcu_read_lock();
1882         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1883                 cnt++;
1884         rcu_read_unlock();
1885         return cnt;
1886 }
1887
1888 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1889                   const struct net_device *dev, int strict)
1890 {
1891         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1892                                        strict, IFA_F_TENTATIVE);
1893 }
1894 EXPORT_SYMBOL(ipv6_chk_addr);
1895
1896 /* device argument is used to find the L3 domain of interest. If
1897  * skip_dev_check is set, then the ifp device is not checked against
1898  * the passed in dev argument. So the 2 cases for addresses checks are:
1899  *   1. does the address exist in the L3 domain that dev is part of
1900  *      (skip_dev_check = true), or
1901  *
1902  *   2. does the address exist on the specific device
1903  *      (skip_dev_check = false)
1904  */
1905 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1906                             const struct net_device *dev, bool skip_dev_check,
1907                             int strict, u32 banned_flags)
1908 {
1909         unsigned int hash = inet6_addr_hash(net, addr);
1910         const struct net_device *l3mdev;
1911         struct inet6_ifaddr *ifp;
1912         u32 ifp_flags;
1913
1914         rcu_read_lock();
1915
1916         l3mdev = l3mdev_master_dev_rcu(dev);
1917         if (skip_dev_check)
1918                 dev = NULL;
1919
1920         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1921                 if (!net_eq(dev_net(ifp->idev->dev), net))
1922                         continue;
1923
1924                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1925                         continue;
1926
1927                 /* Decouple optimistic from tentative for evaluation here.
1928                  * Ban optimistic addresses explicitly, when required.
1929                  */
1930                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1931                             ? (ifp->flags&~IFA_F_TENTATIVE)
1932                             : ifp->flags;
1933                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1934                     !(ifp_flags&banned_flags) &&
1935                     (!dev || ifp->idev->dev == dev ||
1936                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1937                         rcu_read_unlock();
1938                         return 1;
1939                 }
1940         }
1941
1942         rcu_read_unlock();
1943         return 0;
1944 }
1945 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1946
1947
1948 /* Compares an address/prefix_len with addresses on device @dev.
1949  * If one is found it returns true.
1950  */
1951 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1952         const unsigned int prefix_len, struct net_device *dev)
1953 {
1954         const struct inet6_ifaddr *ifa;
1955         const struct inet6_dev *idev;
1956         bool ret = false;
1957
1958         rcu_read_lock();
1959         idev = __in6_dev_get(dev);
1960         if (idev) {
1961                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1962                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1963                         if (ret)
1964                                 break;
1965                 }
1966         }
1967         rcu_read_unlock();
1968
1969         return ret;
1970 }
1971 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1972
1973 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1974 {
1975         const struct inet6_ifaddr *ifa;
1976         const struct inet6_dev *idev;
1977         int     onlink;
1978
1979         onlink = 0;
1980         rcu_read_lock();
1981         idev = __in6_dev_get(dev);
1982         if (idev) {
1983                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1984                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1985                                                    ifa->prefix_len);
1986                         if (onlink)
1987                                 break;
1988                 }
1989         }
1990         rcu_read_unlock();
1991         return onlink;
1992 }
1993 EXPORT_SYMBOL(ipv6_chk_prefix);
1994
1995 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1996                                      struct net_device *dev, int strict)
1997 {
1998         unsigned int hash = inet6_addr_hash(net, addr);
1999         struct inet6_ifaddr *ifp, *result = NULL;
2000
2001         rcu_read_lock();
2002         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2003                 if (!net_eq(dev_net(ifp->idev->dev), net))
2004                         continue;
2005                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2006                         if (!dev || ifp->idev->dev == dev ||
2007                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2008                                 result = ifp;
2009                                 in6_ifa_hold(ifp);
2010                                 break;
2011                         }
2012                 }
2013         }
2014         rcu_read_unlock();
2015
2016         return result;
2017 }
2018
2019 /* Gets referenced address, destroys ifaddr */
2020
2021 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2022 {
2023         if (dad_failed)
2024                 ifp->flags |= IFA_F_DADFAILED;
2025
2026         if (ifp->flags&IFA_F_TEMPORARY) {
2027                 struct inet6_ifaddr *ifpub;
2028                 spin_lock_bh(&ifp->lock);
2029                 ifpub = ifp->ifpub;
2030                 if (ifpub) {
2031                         in6_ifa_hold(ifpub);
2032                         spin_unlock_bh(&ifp->lock);
2033                         ipv6_create_tempaddr(ifpub, ifp, true);
2034                         in6_ifa_put(ifpub);
2035                 } else {
2036                         spin_unlock_bh(&ifp->lock);
2037                 }
2038                 ipv6_del_addr(ifp);
2039         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2040                 spin_lock_bh(&ifp->lock);
2041                 addrconf_del_dad_work(ifp);
2042                 ifp->flags |= IFA_F_TENTATIVE;
2043                 if (dad_failed)
2044                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2045                 spin_unlock_bh(&ifp->lock);
2046                 if (dad_failed)
2047                         ipv6_ifa_notify(0, ifp);
2048                 in6_ifa_put(ifp);
2049         } else {
2050                 ipv6_del_addr(ifp);
2051         }
2052 }
2053
2054 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2055 {
2056         int err = -ENOENT;
2057
2058         spin_lock_bh(&ifp->lock);
2059         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2060                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2061                 err = 0;
2062         }
2063         spin_unlock_bh(&ifp->lock);
2064
2065         return err;
2066 }
2067
2068 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2069 {
2070         struct inet6_dev *idev = ifp->idev;
2071         struct net *net = dev_net(ifp->idev->dev);
2072
2073         if (addrconf_dad_end(ifp)) {
2074                 in6_ifa_put(ifp);
2075                 return;
2076         }
2077
2078         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2079                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2080
2081         spin_lock_bh(&ifp->lock);
2082
2083         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2084                 struct in6_addr new_addr;
2085                 struct inet6_ifaddr *ifp2;
2086                 int retries = ifp->stable_privacy_retry + 1;
2087                 struct ifa6_config cfg = {
2088                         .pfx = &new_addr,
2089                         .plen = ifp->prefix_len,
2090                         .ifa_flags = ifp->flags,
2091                         .valid_lft = ifp->valid_lft,
2092                         .preferred_lft = ifp->prefered_lft,
2093                         .scope = ifp->scope,
2094                 };
2095
2096                 if (retries > net->ipv6.sysctl.idgen_retries) {
2097                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2098                                              ifp->idev->dev->name);
2099                         goto errdad;
2100                 }
2101
2102                 new_addr = ifp->addr;
2103                 if (ipv6_generate_stable_address(&new_addr, retries,
2104                                                  idev))
2105                         goto errdad;
2106
2107                 spin_unlock_bh(&ifp->lock);
2108
2109                 if (idev->cnf.max_addresses &&
2110                     ipv6_count_addresses(idev) >=
2111                     idev->cnf.max_addresses)
2112                         goto lock_errdad;
2113
2114                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2115                                      ifp->idev->dev->name);
2116
2117                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2118                 if (IS_ERR(ifp2))
2119                         goto lock_errdad;
2120
2121                 spin_lock_bh(&ifp2->lock);
2122                 ifp2->stable_privacy_retry = retries;
2123                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2124                 spin_unlock_bh(&ifp2->lock);
2125
2126                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2127                 in6_ifa_put(ifp2);
2128 lock_errdad:
2129                 spin_lock_bh(&ifp->lock);
2130         }
2131
2132 errdad:
2133         /* transition from _POSTDAD to _ERRDAD */
2134         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2135         spin_unlock_bh(&ifp->lock);
2136
2137         addrconf_mod_dad_work(ifp, 0);
2138         in6_ifa_put(ifp);
2139 }
2140
2141 /* Join to solicited addr multicast group.
2142  * caller must hold RTNL */
2143 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2144 {
2145         struct in6_addr maddr;
2146
2147         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2148                 return;
2149
2150         addrconf_addr_solict_mult(addr, &maddr);
2151         ipv6_dev_mc_inc(dev, &maddr);
2152 }
2153
2154 /* caller must hold RTNL */
2155 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2156 {
2157         struct in6_addr maddr;
2158
2159         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2160                 return;
2161
2162         addrconf_addr_solict_mult(addr, &maddr);
2163         __ipv6_dev_mc_dec(idev, &maddr);
2164 }
2165
2166 /* caller must hold RTNL */
2167 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2168 {
2169         struct in6_addr addr;
2170
2171         if (ifp->prefix_len >= 127) /* RFC 6164 */
2172                 return;
2173         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2174         if (ipv6_addr_any(&addr))
2175                 return;
2176         __ipv6_dev_ac_inc(ifp->idev, &addr);
2177 }
2178
2179 /* caller must hold RTNL */
2180 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2181 {
2182         struct in6_addr addr;
2183
2184         if (ifp->prefix_len >= 127) /* RFC 6164 */
2185                 return;
2186         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2187         if (ipv6_addr_any(&addr))
2188                 return;
2189         __ipv6_dev_ac_dec(ifp->idev, &addr);
2190 }
2191
2192 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2193 {
2194         switch (dev->addr_len) {
2195         case ETH_ALEN:
2196                 memcpy(eui, dev->dev_addr, 3);
2197                 eui[3] = 0xFF;
2198                 eui[4] = 0xFE;
2199                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2200                 break;
2201         case EUI64_ADDR_LEN:
2202                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2203                 eui[0] ^= 2;
2204                 break;
2205         default:
2206                 return -1;
2207         }
2208
2209         return 0;
2210 }
2211
2212 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2213 {
2214         union fwnet_hwaddr *ha;
2215
2216         if (dev->addr_len != FWNET_ALEN)
2217                 return -1;
2218
2219         ha = (union fwnet_hwaddr *)dev->dev_addr;
2220
2221         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2222         eui[0] ^= 2;
2223         return 0;
2224 }
2225
2226 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2227 {
2228         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2229         if (dev->addr_len != ARCNET_ALEN)
2230                 return -1;
2231         memset(eui, 0, 7);
2232         eui[7] = *(u8 *)dev->dev_addr;
2233         return 0;
2234 }
2235
2236 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2237 {
2238         if (dev->addr_len != INFINIBAND_ALEN)
2239                 return -1;
2240         memcpy(eui, dev->dev_addr + 12, 8);
2241         eui[0] |= 2;
2242         return 0;
2243 }
2244
2245 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2246 {
2247         if (addr == 0)
2248                 return -1;
2249         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2250                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2251                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2252                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2253                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2254                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2255         eui[1] = 0;
2256         eui[2] = 0x5E;
2257         eui[3] = 0xFE;
2258         memcpy(eui + 4, &addr, 4);
2259         return 0;
2260 }
2261
2262 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2263 {
2264         if (dev->priv_flags & IFF_ISATAP)
2265                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2266         return -1;
2267 }
2268
2269 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2270 {
2271         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2272 }
2273
2274 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2275 {
2276         memcpy(eui, dev->perm_addr, 3);
2277         memcpy(eui + 5, dev->perm_addr + 3, 3);
2278         eui[3] = 0xFF;
2279         eui[4] = 0xFE;
2280         eui[0] ^= 2;
2281         return 0;
2282 }
2283
2284 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2285 {
2286         switch (dev->type) {
2287         case ARPHRD_ETHER:
2288         case ARPHRD_FDDI:
2289                 return addrconf_ifid_eui48(eui, dev);
2290         case ARPHRD_ARCNET:
2291                 return addrconf_ifid_arcnet(eui, dev);
2292         case ARPHRD_INFINIBAND:
2293                 return addrconf_ifid_infiniband(eui, dev);
2294         case ARPHRD_SIT:
2295                 return addrconf_ifid_sit(eui, dev);
2296         case ARPHRD_IPGRE:
2297         case ARPHRD_TUNNEL:
2298                 return addrconf_ifid_gre(eui, dev);
2299         case ARPHRD_6LOWPAN:
2300                 return addrconf_ifid_6lowpan(eui, dev);
2301         case ARPHRD_IEEE1394:
2302                 return addrconf_ifid_ieee1394(eui, dev);
2303         case ARPHRD_TUNNEL6:
2304         case ARPHRD_IP6GRE:
2305         case ARPHRD_RAWIP:
2306                 return addrconf_ifid_ip6tnl(eui, dev);
2307         }
2308         return -1;
2309 }
2310
2311 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2312 {
2313         int err = -1;
2314         struct inet6_ifaddr *ifp;
2315
2316         read_lock_bh(&idev->lock);
2317         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2318                 if (ifp->scope > IFA_LINK)
2319                         break;
2320                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2321                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2322                         err = 0;
2323                         break;
2324                 }
2325         }
2326         read_unlock_bh(&idev->lock);
2327         return err;
2328 }
2329
2330 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2331 static void ipv6_regen_rndid(struct inet6_dev *idev)
2332 {
2333 regen:
2334         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2335         idev->rndid[0] &= ~0x02;
2336
2337         /*
2338          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2339          * check if generated address is not inappropriate
2340          *
2341          *  - Reserved subnet anycast (RFC 2526)
2342          *      11111101 11....11 1xxxxxxx
2343          *  - ISATAP (RFC4214) 6.1
2344          *      00-00-5E-FE-xx-xx-xx-xx
2345          *  - value 0
2346          *  - XXX: already assigned to an address on the device
2347          */
2348         if (idev->rndid[0] == 0xfd &&
2349             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2350             (idev->rndid[7]&0x80))
2351                 goto regen;
2352         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2353                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2354                         goto regen;
2355                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2356                         goto regen;
2357         }
2358 }
2359
2360 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2361 {
2362         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2363                 ipv6_regen_rndid(idev);
2364 }
2365
2366 /*
2367  *      Add prefix route.
2368  */
2369
2370 static void
2371 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2372                       struct net_device *dev, unsigned long expires,
2373                       u32 flags, gfp_t gfp_flags)
2374 {
2375         struct fib6_config cfg = {
2376                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2377                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2378                 .fc_ifindex = dev->ifindex,
2379                 .fc_expires = expires,
2380                 .fc_dst_len = plen,
2381                 .fc_flags = RTF_UP | flags,
2382                 .fc_nlinfo.nl_net = dev_net(dev),
2383                 .fc_protocol = RTPROT_KERNEL,
2384                 .fc_type = RTN_UNICAST,
2385         };
2386
2387         cfg.fc_dst = *pfx;
2388
2389         /* Prevent useless cloning on PtP SIT.
2390            This thing is done here expecting that the whole
2391            class of non-broadcast devices need not cloning.
2392          */
2393 #if IS_ENABLED(CONFIG_IPV6_SIT)
2394         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2395                 cfg.fc_flags |= RTF_NONEXTHOP;
2396 #endif
2397
2398         ip6_route_add(&cfg, gfp_flags, NULL);
2399 }
2400
2401
2402 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2403                                                   int plen,
2404                                                   const struct net_device *dev,
2405                                                   u32 flags, u32 noflags,
2406                                                   bool no_gw)
2407 {
2408         struct fib6_node *fn;
2409         struct fib6_info *rt = NULL;
2410         struct fib6_table *table;
2411         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2412
2413         table = fib6_get_table(dev_net(dev), tb_id);
2414         if (!table)
2415                 return NULL;
2416
2417         rcu_read_lock();
2418         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2419         if (!fn)
2420                 goto out;
2421
2422         for_each_fib6_node_rt_rcu(fn) {
2423                 /* prefix routes only use builtin fib6_nh */
2424                 if (rt->nh)
2425                         continue;
2426
2427                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2428                         continue;
2429                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2430                         continue;
2431                 if ((rt->fib6_flags & flags) != flags)
2432                         continue;
2433                 if ((rt->fib6_flags & noflags) != 0)
2434                         continue;
2435                 if (!fib6_info_hold_safe(rt))
2436                         continue;
2437                 break;
2438         }
2439 out:
2440         rcu_read_unlock();
2441         return rt;
2442 }
2443
2444
2445 /* Create "default" multicast route to the interface */
2446
2447 static void addrconf_add_mroute(struct net_device *dev)
2448 {
2449         struct fib6_config cfg = {
2450                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2451                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2452                 .fc_ifindex = dev->ifindex,
2453                 .fc_dst_len = 8,
2454                 .fc_flags = RTF_UP,
2455                 .fc_type = RTN_UNICAST,
2456                 .fc_nlinfo.nl_net = dev_net(dev),
2457         };
2458
2459         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2460
2461         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2462 }
2463
2464 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2465 {
2466         struct inet6_dev *idev;
2467
2468         ASSERT_RTNL();
2469
2470         idev = ipv6_find_idev(dev);
2471         if (IS_ERR(idev))
2472                 return idev;
2473
2474         if (idev->cnf.disable_ipv6)
2475                 return ERR_PTR(-EACCES);
2476
2477         /* Add default multicast route */
2478         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2479                 addrconf_add_mroute(dev);
2480
2481         return idev;
2482 }
2483
2484 static void manage_tempaddrs(struct inet6_dev *idev,
2485                              struct inet6_ifaddr *ifp,
2486                              __u32 valid_lft, __u32 prefered_lft,
2487                              bool create, unsigned long now)
2488 {
2489         u32 flags;
2490         struct inet6_ifaddr *ift;
2491
2492         read_lock_bh(&idev->lock);
2493         /* update all temporary addresses in the list */
2494         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2495                 int age, max_valid, max_prefered;
2496
2497                 if (ifp != ift->ifpub)
2498                         continue;
2499
2500                 /* RFC 4941 section 3.3:
2501                  * If a received option will extend the lifetime of a public
2502                  * address, the lifetimes of temporary addresses should
2503                  * be extended, subject to the overall constraint that no
2504                  * temporary addresses should ever remain "valid" or "preferred"
2505                  * for a time longer than (TEMP_VALID_LIFETIME) or
2506                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2507                  */
2508                 age = (now - ift->cstamp) / HZ;
2509                 max_valid = idev->cnf.temp_valid_lft - age;
2510                 if (max_valid < 0)
2511                         max_valid = 0;
2512
2513                 max_prefered = idev->cnf.temp_prefered_lft -
2514                                idev->desync_factor - age;
2515                 if (max_prefered < 0)
2516                         max_prefered = 0;
2517
2518                 if (valid_lft > max_valid)
2519                         valid_lft = max_valid;
2520
2521                 if (prefered_lft > max_prefered)
2522                         prefered_lft = max_prefered;
2523
2524                 spin_lock(&ift->lock);
2525                 flags = ift->flags;
2526                 ift->valid_lft = valid_lft;
2527                 ift->prefered_lft = prefered_lft;
2528                 ift->tstamp = now;
2529                 if (prefered_lft > 0)
2530                         ift->flags &= ~IFA_F_DEPRECATED;
2531
2532                 spin_unlock(&ift->lock);
2533                 if (!(flags&IFA_F_TENTATIVE))
2534                         ipv6_ifa_notify(0, ift);
2535         }
2536
2537         if ((create || list_empty(&idev->tempaddr_list)) &&
2538             idev->cnf.use_tempaddr > 0) {
2539                 /* When a new public address is created as described
2540                  * in [ADDRCONF], also create a new temporary address.
2541                  * Also create a temporary address if it's enabled but
2542                  * no temporary address currently exists.
2543                  */
2544                 read_unlock_bh(&idev->lock);
2545                 ipv6_create_tempaddr(ifp, NULL, false);
2546         } else {
2547                 read_unlock_bh(&idev->lock);
2548         }
2549 }
2550
2551 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2552 {
2553         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2554                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2555 }
2556
2557 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2558                                  const struct prefix_info *pinfo,
2559                                  struct inet6_dev *in6_dev,
2560                                  const struct in6_addr *addr, int addr_type,
2561                                  u32 addr_flags, bool sllao, bool tokenized,
2562                                  __u32 valid_lft, u32 prefered_lft)
2563 {
2564         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2565         int create = 0, update_lft = 0;
2566
2567         if (!ifp && valid_lft) {
2568                 int max_addresses = in6_dev->cnf.max_addresses;
2569                 struct ifa6_config cfg = {
2570                         .pfx = addr,
2571                         .plen = pinfo->prefix_len,
2572                         .ifa_flags = addr_flags,
2573                         .valid_lft = valid_lft,
2574                         .preferred_lft = prefered_lft,
2575                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2576                 };
2577
2578 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2579                 if ((net->ipv6.devconf_all->optimistic_dad ||
2580                      in6_dev->cnf.optimistic_dad) &&
2581                     !net->ipv6.devconf_all->forwarding && sllao)
2582                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2583 #endif
2584
2585                 /* Do not allow to create too much of autoconfigured
2586                  * addresses; this would be too easy way to crash kernel.
2587                  */
2588                 if (!max_addresses ||
2589                     ipv6_count_addresses(in6_dev) < max_addresses)
2590                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2591
2592                 if (IS_ERR_OR_NULL(ifp))
2593                         return -1;
2594
2595                 create = 1;
2596                 spin_lock_bh(&ifp->lock);
2597                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2598                 ifp->cstamp = jiffies;
2599                 ifp->tokenized = tokenized;
2600                 spin_unlock_bh(&ifp->lock);
2601                 addrconf_dad_start(ifp);
2602         }
2603
2604         if (ifp) {
2605                 u32 flags;
2606                 unsigned long now;
2607                 u32 stored_lft;
2608
2609                 /* update lifetime (RFC2462 5.5.3 e) */
2610                 spin_lock_bh(&ifp->lock);
2611                 now = jiffies;
2612                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2613                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2614                 else
2615                         stored_lft = 0;
2616                 if (!create && stored_lft) {
2617                         const u32 minimum_lft = min_t(u32,
2618                                 stored_lft, MIN_VALID_LIFETIME);
2619                         valid_lft = max(valid_lft, minimum_lft);
2620
2621                         /* RFC4862 Section 5.5.3e:
2622                          * "Note that the preferred lifetime of the
2623                          *  corresponding address is always reset to
2624                          *  the Preferred Lifetime in the received
2625                          *  Prefix Information option, regardless of
2626                          *  whether the valid lifetime is also reset or
2627                          *  ignored."
2628                          *
2629                          * So we should always update prefered_lft here.
2630                          */
2631                         update_lft = 1;
2632                 }
2633
2634                 if (update_lft) {
2635                         ifp->valid_lft = valid_lft;
2636                         ifp->prefered_lft = prefered_lft;
2637                         ifp->tstamp = now;
2638                         flags = ifp->flags;
2639                         ifp->flags &= ~IFA_F_DEPRECATED;
2640                         spin_unlock_bh(&ifp->lock);
2641
2642                         if (!(flags&IFA_F_TENTATIVE))
2643                                 ipv6_ifa_notify(0, ifp);
2644                 } else
2645                         spin_unlock_bh(&ifp->lock);
2646
2647                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2648                                  create, now);
2649
2650                 in6_ifa_put(ifp);
2651                 addrconf_verify();
2652         }
2653
2654         return 0;
2655 }
2656 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2657
2658 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2659 {
2660         struct prefix_info *pinfo;
2661         __u32 valid_lft;
2662         __u32 prefered_lft;
2663         int addr_type, err;
2664         u32 addr_flags = 0;
2665         struct inet6_dev *in6_dev;
2666         struct net *net = dev_net(dev);
2667
2668         pinfo = (struct prefix_info *) opt;
2669
2670         if (len < sizeof(struct prefix_info)) {
2671                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2672                 return;
2673         }
2674
2675         /*
2676          *      Validation checks ([ADDRCONF], page 19)
2677          */
2678
2679         addr_type = ipv6_addr_type(&pinfo->prefix);
2680
2681         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2682                 return;
2683
2684         valid_lft = ntohl(pinfo->valid);
2685         prefered_lft = ntohl(pinfo->prefered);
2686
2687         if (prefered_lft > valid_lft) {
2688                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2689                 return;
2690         }
2691
2692         in6_dev = in6_dev_get(dev);
2693
2694         if (!in6_dev) {
2695                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2696                                     dev->name);
2697                 return;
2698         }
2699
2700         /*
2701          *      Two things going on here:
2702          *      1) Add routes for on-link prefixes
2703          *      2) Configure prefixes with the auto flag set
2704          */
2705
2706         if (pinfo->onlink) {
2707                 struct fib6_info *rt;
2708                 unsigned long rt_expires;
2709
2710                 /* Avoid arithmetic overflow. Really, we could
2711                  * save rt_expires in seconds, likely valid_lft,
2712                  * but it would require division in fib gc, that it
2713                  * not good.
2714                  */
2715                 if (HZ > USER_HZ)
2716                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2717                 else
2718                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2719
2720                 if (addrconf_finite_timeout(rt_expires))
2721                         rt_expires *= HZ;
2722
2723                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2724                                                pinfo->prefix_len,
2725                                                dev,
2726                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2727                                                RTF_DEFAULT, true);
2728
2729                 if (rt) {
2730                         /* Autoconf prefix route */
2731                         if (valid_lft == 0) {
2732                                 ip6_del_rt(net, rt);
2733                                 rt = NULL;
2734                         } else if (addrconf_finite_timeout(rt_expires)) {
2735                                 /* not infinity */
2736                                 fib6_set_expires(rt, jiffies + rt_expires);
2737                         } else {
2738                                 fib6_clean_expires(rt);
2739                         }
2740                 } else if (valid_lft) {
2741                         clock_t expires = 0;
2742                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2743                         if (addrconf_finite_timeout(rt_expires)) {
2744                                 /* not infinity */
2745                                 flags |= RTF_EXPIRES;
2746                                 expires = jiffies_to_clock_t(rt_expires);
2747                         }
2748                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2749                                               0, dev, expires, flags,
2750                                               GFP_ATOMIC);
2751                 }
2752                 fib6_info_release(rt);
2753         }
2754
2755         /* Try to figure out our local address for this prefix */
2756
2757         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2758                 struct in6_addr addr;
2759                 bool tokenized = false, dev_addr_generated = false;
2760
2761                 if (pinfo->prefix_len == 64) {
2762                         memcpy(&addr, &pinfo->prefix, 8);
2763
2764                         if (!ipv6_addr_any(&in6_dev->token)) {
2765                                 read_lock_bh(&in6_dev->lock);
2766                                 memcpy(addr.s6_addr + 8,
2767                                        in6_dev->token.s6_addr + 8, 8);
2768                                 read_unlock_bh(&in6_dev->lock);
2769                                 tokenized = true;
2770                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2771                                    !ipv6_generate_stable_address(&addr, 0,
2772                                                                  in6_dev)) {
2773                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2774                                 goto ok;
2775                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2776                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2777                                 goto put;
2778                         } else {
2779                                 dev_addr_generated = true;
2780                         }
2781                         goto ok;
2782                 }
2783                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2784                                     pinfo->prefix_len);
2785                 goto put;
2786
2787 ok:
2788                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2789                                                    &addr, addr_type,
2790                                                    addr_flags, sllao,
2791                                                    tokenized, valid_lft,
2792                                                    prefered_lft);
2793                 if (err)
2794                         goto put;
2795
2796                 /* Ignore error case here because previous prefix add addr was
2797                  * successful which will be notified.
2798                  */
2799                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2800                                               addr_type, addr_flags, sllao,
2801                                               tokenized, valid_lft,
2802                                               prefered_lft,
2803                                               dev_addr_generated);
2804         }
2805         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2806 put:
2807         in6_dev_put(in6_dev);
2808 }
2809
2810 /*
2811  *      Set destination address.
2812  *      Special case for SIT interfaces where we create a new "virtual"
2813  *      device.
2814  */
2815 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2816 {
2817         struct in6_ifreq ireq;
2818         struct net_device *dev;
2819         int err = -EINVAL;
2820
2821         rtnl_lock();
2822
2823         err = -EFAULT;
2824         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2825                 goto err_exit;
2826
2827         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2828
2829         err = -ENODEV;
2830         if (!dev)
2831                 goto err_exit;
2832
2833 #if IS_ENABLED(CONFIG_IPV6_SIT)
2834         if (dev->type == ARPHRD_SIT) {
2835                 const struct net_device_ops *ops = dev->netdev_ops;
2836                 struct ifreq ifr;
2837                 struct ip_tunnel_parm p;
2838
2839                 err = -EADDRNOTAVAIL;
2840                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2841                         goto err_exit;
2842
2843                 memset(&p, 0, sizeof(p));
2844                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2845                 p.iph.saddr = 0;
2846                 p.iph.version = 4;
2847                 p.iph.ihl = 5;
2848                 p.iph.protocol = IPPROTO_IPV6;
2849                 p.iph.ttl = 64;
2850                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2851
2852                 if (ops->ndo_do_ioctl) {
2853                         mm_segment_t oldfs = get_fs();
2854
2855                         set_fs(KERNEL_DS);
2856                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2857                         set_fs(oldfs);
2858                 } else
2859                         err = -EOPNOTSUPP;
2860
2861                 if (err == 0) {
2862                         err = -ENOBUFS;
2863                         dev = __dev_get_by_name(net, p.name);
2864                         if (!dev)
2865                                 goto err_exit;
2866                         err = dev_open(dev, NULL);
2867                 }
2868         }
2869 #endif
2870
2871 err_exit:
2872         rtnl_unlock();
2873         return err;
2874 }
2875
2876 static int ipv6_mc_config(struct sock *sk, bool join,
2877                           const struct in6_addr *addr, int ifindex)
2878 {
2879         int ret;
2880
2881         ASSERT_RTNL();
2882
2883         lock_sock(sk);
2884         if (join)
2885                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2886         else
2887                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2888         release_sock(sk);
2889
2890         return ret;
2891 }
2892
2893 /*
2894  *      Manual configuration of address on an interface
2895  */
2896 static int inet6_addr_add(struct net *net, int ifindex,
2897                           struct ifa6_config *cfg,
2898                           struct netlink_ext_ack *extack)
2899 {
2900         struct inet6_ifaddr *ifp;
2901         struct inet6_dev *idev;
2902         struct net_device *dev;
2903         unsigned long timeout;
2904         clock_t expires;
2905         u32 flags;
2906
2907         ASSERT_RTNL();
2908
2909         if (cfg->plen > 128)
2910                 return -EINVAL;
2911
2912         /* check the lifetime */
2913         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2914                 return -EINVAL;
2915
2916         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2917                 return -EINVAL;
2918
2919         dev = __dev_get_by_index(net, ifindex);
2920         if (!dev)
2921                 return -ENODEV;
2922
2923         idev = addrconf_add_dev(dev);
2924         if (IS_ERR(idev))
2925                 return PTR_ERR(idev);
2926
2927         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2928                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2929                                          true, cfg->pfx, ifindex);
2930
2931                 if (ret < 0)
2932                         return ret;
2933         }
2934
2935         cfg->scope = ipv6_addr_scope(cfg->pfx);
2936
2937         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2938         if (addrconf_finite_timeout(timeout)) {
2939                 expires = jiffies_to_clock_t(timeout * HZ);
2940                 cfg->valid_lft = timeout;
2941                 flags = RTF_EXPIRES;
2942         } else {
2943                 expires = 0;
2944                 flags = 0;
2945                 cfg->ifa_flags |= IFA_F_PERMANENT;
2946         }
2947
2948         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2949         if (addrconf_finite_timeout(timeout)) {
2950                 if (timeout == 0)
2951                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2952                 cfg->preferred_lft = timeout;
2953         }
2954
2955         ifp = ipv6_add_addr(idev, cfg, true, extack);
2956         if (!IS_ERR(ifp)) {
2957                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2958                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2959                                               ifp->rt_priority, dev, expires,
2960                                               flags, GFP_KERNEL);
2961                 }
2962
2963                 /* Send a netlink notification if DAD is enabled and
2964                  * optimistic flag is not set
2965                  */
2966                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2967                         ipv6_ifa_notify(0, ifp);
2968                 /*
2969                  * Note that section 3.1 of RFC 4429 indicates
2970                  * that the Optimistic flag should not be set for
2971                  * manually configured addresses
2972                  */
2973                 addrconf_dad_start(ifp);
2974                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2975                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2976                                          cfg->preferred_lft, true, jiffies);
2977                 in6_ifa_put(ifp);
2978                 addrconf_verify_rtnl();
2979                 return 0;
2980         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2981                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2982                                cfg->pfx, ifindex);
2983         }
2984
2985         return PTR_ERR(ifp);
2986 }
2987
2988 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2989                           const struct in6_addr *pfx, unsigned int plen)
2990 {
2991         struct inet6_ifaddr *ifp;
2992         struct inet6_dev *idev;
2993         struct net_device *dev;
2994
2995         if (plen > 128)
2996                 return -EINVAL;
2997
2998         dev = __dev_get_by_index(net, ifindex);
2999         if (!dev)
3000                 return -ENODEV;
3001
3002         idev = __in6_dev_get(dev);
3003         if (!idev)
3004                 return -ENXIO;
3005
3006         read_lock_bh(&idev->lock);
3007         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3008                 if (ifp->prefix_len == plen &&
3009                     ipv6_addr_equal(pfx, &ifp->addr)) {
3010                         in6_ifa_hold(ifp);
3011                         read_unlock_bh(&idev->lock);
3012
3013                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3014                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3015                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3016                                                  jiffies);
3017                         ipv6_del_addr(ifp);
3018                         addrconf_verify_rtnl();
3019                         if (ipv6_addr_is_multicast(pfx)) {
3020                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3021                                                false, pfx, dev->ifindex);
3022                         }
3023                         return 0;
3024                 }
3025         }
3026         read_unlock_bh(&idev->lock);
3027         return -EADDRNOTAVAIL;
3028 }
3029
3030
3031 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3032 {
3033         struct ifa6_config cfg = {
3034                 .ifa_flags = IFA_F_PERMANENT,
3035                 .preferred_lft = INFINITY_LIFE_TIME,
3036                 .valid_lft = INFINITY_LIFE_TIME,
3037         };
3038         struct in6_ifreq ireq;
3039         int err;
3040
3041         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3042                 return -EPERM;
3043
3044         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3045                 return -EFAULT;
3046
3047         cfg.pfx = &ireq.ifr6_addr;
3048         cfg.plen = ireq.ifr6_prefixlen;
3049
3050         rtnl_lock();
3051         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3052         rtnl_unlock();
3053         return err;
3054 }
3055
3056 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3057 {
3058         struct in6_ifreq ireq;
3059         int err;
3060
3061         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3062                 return -EPERM;
3063
3064         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3065                 return -EFAULT;
3066
3067         rtnl_lock();
3068         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3069                              ireq.ifr6_prefixlen);
3070         rtnl_unlock();
3071         return err;
3072 }
3073
3074 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3075                      int plen, int scope)
3076 {
3077         struct inet6_ifaddr *ifp;
3078         struct ifa6_config cfg = {
3079                 .pfx = addr,
3080                 .plen = plen,
3081                 .ifa_flags = IFA_F_PERMANENT,
3082                 .valid_lft = INFINITY_LIFE_TIME,
3083                 .preferred_lft = INFINITY_LIFE_TIME,
3084                 .scope = scope
3085         };
3086
3087         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3088         if (!IS_ERR(ifp)) {
3089                 spin_lock_bh(&ifp->lock);
3090                 ifp->flags &= ~IFA_F_TENTATIVE;
3091                 spin_unlock_bh(&ifp->lock);
3092                 rt_genid_bump_ipv6(dev_net(idev->dev));
3093                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3094                 in6_ifa_put(ifp);
3095         }
3096 }
3097
3098 #if IS_ENABLED(CONFIG_IPV6_SIT)
3099 static void sit_add_v4_addrs(struct inet6_dev *idev)
3100 {
3101         struct in6_addr addr;
3102         struct net_device *dev;
3103         struct net *net = dev_net(idev->dev);
3104         int scope, plen;
3105         u32 pflags = 0;
3106
3107         ASSERT_RTNL();
3108
3109         memset(&addr, 0, sizeof(struct in6_addr));
3110         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3111
3112         if (idev->dev->flags&IFF_POINTOPOINT) {
3113                 addr.s6_addr32[0] = htonl(0xfe800000);
3114                 scope = IFA_LINK;
3115                 plen = 64;
3116         } else {
3117                 scope = IPV6_ADDR_COMPATv4;
3118                 plen = 96;
3119                 pflags |= RTF_NONEXTHOP;
3120         }
3121
3122         if (addr.s6_addr32[3]) {
3123                 add_addr(idev, &addr, plen, scope);
3124                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3125                                       GFP_KERNEL);
3126                 return;
3127         }
3128
3129         for_each_netdev(net, dev) {
3130                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3131                 if (in_dev && (dev->flags & IFF_UP)) {
3132                         struct in_ifaddr *ifa;
3133                         int flag = scope;
3134
3135                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3136                                 addr.s6_addr32[3] = ifa->ifa_local;
3137
3138                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3139                                         continue;
3140                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3141                                         if (idev->dev->flags&IFF_POINTOPOINT)
3142                                                 continue;
3143                                         flag |= IFA_HOST;
3144                                 }
3145
3146                                 add_addr(idev, &addr, plen, flag);
3147                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3148                                                       0, pflags, GFP_KERNEL);
3149                         }
3150                 }
3151         }
3152 }
3153 #endif
3154
3155 static void init_loopback(struct net_device *dev)
3156 {
3157         struct inet6_dev  *idev;
3158
3159         /* ::1 */
3160
3161         ASSERT_RTNL();
3162
3163         idev = ipv6_find_idev(dev);
3164         if (IS_ERR(idev)) {
3165                 pr_debug("%s: add_dev failed\n", __func__);
3166                 return;
3167         }
3168
3169         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3170 }
3171
3172 void addrconf_add_linklocal(struct inet6_dev *idev,
3173                             const struct in6_addr *addr, u32 flags)
3174 {
3175         struct ifa6_config cfg = {
3176                 .pfx = addr,
3177                 .plen = 64,
3178                 .ifa_flags = flags | IFA_F_PERMANENT,
3179                 .valid_lft = INFINITY_LIFE_TIME,
3180                 .preferred_lft = INFINITY_LIFE_TIME,
3181                 .scope = IFA_LINK
3182         };
3183         struct inet6_ifaddr *ifp;
3184
3185 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3186         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3187              idev->cnf.optimistic_dad) &&
3188             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3189                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3190 #endif
3191
3192         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3193         if (!IS_ERR(ifp)) {
3194                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3195                                       0, 0, GFP_ATOMIC);
3196                 addrconf_dad_start(ifp);
3197                 in6_ifa_put(ifp);
3198         }
3199 }
3200 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3201
3202 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3203 {
3204         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3205                 return true;
3206
3207         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3208             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3209                 return true;
3210
3211         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3212             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3213                 return true;
3214
3215         return false;
3216 }
3217
3218 static int ipv6_generate_stable_address(struct in6_addr *address,
3219                                         u8 dad_count,
3220                                         const struct inet6_dev *idev)
3221 {
3222         static DEFINE_SPINLOCK(lock);
3223         static __u32 digest[SHA_DIGEST_WORDS];
3224         static __u32 workspace[SHA_WORKSPACE_WORDS];
3225
3226         static union {
3227                 char __data[SHA_MESSAGE_BYTES];
3228                 struct {
3229                         struct in6_addr secret;
3230                         __be32 prefix[2];
3231                         unsigned char hwaddr[MAX_ADDR_LEN];
3232                         u8 dad_count;
3233                 } __packed;
3234         } data;
3235
3236         struct in6_addr secret;
3237         struct in6_addr temp;
3238         struct net *net = dev_net(idev->dev);
3239
3240         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3241
3242         if (idev->cnf.stable_secret.initialized)
3243                 secret = idev->cnf.stable_secret.secret;
3244         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3245                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3246         else
3247                 return -1;
3248
3249 retry:
3250         spin_lock_bh(&lock);
3251
3252         sha_init(digest);
3253         memset(&data, 0, sizeof(data));
3254         memset(workspace, 0, sizeof(workspace));
3255         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3256         data.prefix[0] = address->s6_addr32[0];
3257         data.prefix[1] = address->s6_addr32[1];
3258         data.secret = secret;
3259         data.dad_count = dad_count;
3260
3261         sha_transform(digest, data.__data, workspace);
3262
3263         temp = *address;
3264         temp.s6_addr32[2] = (__force __be32)digest[0];
3265         temp.s6_addr32[3] = (__force __be32)digest[1];
3266
3267         spin_unlock_bh(&lock);
3268
3269         if (ipv6_reserved_interfaceid(temp)) {
3270                 dad_count++;
3271                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3272                         return -1;
3273                 goto retry;
3274         }
3275
3276         *address = temp;
3277         return 0;
3278 }
3279
3280 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3281 {
3282         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3283
3284         if (s->initialized)
3285                 return;
3286         s = &idev->cnf.stable_secret;
3287         get_random_bytes(&s->secret, sizeof(s->secret));
3288         s->initialized = true;
3289 }
3290
3291 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3292 {
3293         struct in6_addr addr;
3294
3295         /* no link local addresses on L3 master devices */
3296         if (netif_is_l3_master(idev->dev))
3297                 return;
3298
3299         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3300
3301         switch (idev->cnf.addr_gen_mode) {
3302         case IN6_ADDR_GEN_MODE_RANDOM:
3303                 ipv6_gen_mode_random_init(idev);
3304                 fallthrough;
3305         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3306                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3307                         addrconf_add_linklocal(idev, &addr,
3308                                                IFA_F_STABLE_PRIVACY);
3309                 else if (prefix_route)
3310                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3311                                               0, 0, GFP_KERNEL);
3312                 break;
3313         case IN6_ADDR_GEN_MODE_EUI64:
3314                 /* addrconf_add_linklocal also adds a prefix_route and we
3315                  * only need to care about prefix routes if ipv6_generate_eui64
3316                  * couldn't generate one.
3317                  */
3318                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3319                         addrconf_add_linklocal(idev, &addr, 0);
3320                 else if (prefix_route)
3321                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3322                                               0, 0, GFP_KERNEL);
3323                 break;
3324         case IN6_ADDR_GEN_MODE_NONE:
3325         default:
3326                 /* will not add any link local address */
3327                 break;
3328         }
3329 }
3330
3331 static void addrconf_dev_config(struct net_device *dev)
3332 {
3333         struct inet6_dev *idev;
3334
3335         ASSERT_RTNL();
3336
3337         if ((dev->type != ARPHRD_ETHER) &&
3338             (dev->type != ARPHRD_FDDI) &&
3339             (dev->type != ARPHRD_ARCNET) &&
3340             (dev->type != ARPHRD_INFINIBAND) &&
3341             (dev->type != ARPHRD_IEEE1394) &&
3342             (dev->type != ARPHRD_TUNNEL6) &&
3343             (dev->type != ARPHRD_6LOWPAN) &&
3344             (dev->type != ARPHRD_IP6GRE) &&
3345             (dev->type != ARPHRD_IPGRE) &&
3346             (dev->type != ARPHRD_TUNNEL) &&
3347             (dev->type != ARPHRD_NONE) &&
3348             (dev->type != ARPHRD_RAWIP)) {
3349                 /* Alas, we support only Ethernet autoconfiguration. */
3350                 idev = __in6_dev_get(dev);
3351                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3352                     dev->flags & IFF_MULTICAST)
3353                         ipv6_mc_up(idev);
3354                 return;
3355         }
3356
3357         idev = addrconf_add_dev(dev);
3358         if (IS_ERR(idev))
3359                 return;
3360
3361         /* this device type has no EUI support */
3362         if (dev->type == ARPHRD_NONE &&
3363             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3364                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3365
3366         addrconf_addr_gen(idev, false);
3367 }
3368
3369 #if IS_ENABLED(CONFIG_IPV6_SIT)
3370 static void addrconf_sit_config(struct net_device *dev)
3371 {
3372         struct inet6_dev *idev;
3373
3374         ASSERT_RTNL();
3375
3376         /*
3377          * Configure the tunnel with one of our IPv4
3378          * addresses... we should configure all of
3379          * our v4 addrs in the tunnel
3380          */
3381
3382         idev = ipv6_find_idev(dev);
3383         if (IS_ERR(idev)) {
3384                 pr_debug("%s: add_dev failed\n", __func__);
3385                 return;
3386         }
3387
3388         if (dev->priv_flags & IFF_ISATAP) {
3389                 addrconf_addr_gen(idev, false);
3390                 return;
3391         }
3392
3393         sit_add_v4_addrs(idev);
3394
3395         if (dev->flags&IFF_POINTOPOINT)
3396                 addrconf_add_mroute(dev);
3397 }
3398 #endif
3399
3400 #if IS_ENABLED(CONFIG_NET_IPGRE)
3401 static void addrconf_gre_config(struct net_device *dev)
3402 {
3403         struct inet6_dev *idev;
3404
3405         ASSERT_RTNL();
3406
3407         idev = ipv6_find_idev(dev);
3408         if (IS_ERR(idev)) {
3409                 pr_debug("%s: add_dev failed\n", __func__);
3410                 return;
3411         }
3412
3413         addrconf_addr_gen(idev, true);
3414         if (dev->flags & IFF_POINTOPOINT)
3415                 addrconf_add_mroute(dev);
3416 }
3417 #endif
3418
3419 static int fixup_permanent_addr(struct net *net,
3420                                 struct inet6_dev *idev,
3421                                 struct inet6_ifaddr *ifp)
3422 {
3423         /* !fib6_node means the host route was removed from the
3424          * FIB, for example, if 'lo' device is taken down. In that
3425          * case regenerate the host route.
3426          */
3427         if (!ifp->rt || !ifp->rt->fib6_node) {
3428                 struct fib6_info *f6i, *prev;
3429
3430                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3431                                          GFP_ATOMIC);
3432                 if (IS_ERR(f6i))
3433                         return PTR_ERR(f6i);
3434
3435                 /* ifp->rt can be accessed outside of rtnl */
3436                 spin_lock(&ifp->lock);
3437                 prev = ifp->rt;
3438                 ifp->rt = f6i;
3439                 spin_unlock(&ifp->lock);
3440
3441                 fib6_info_release(prev);
3442         }
3443
3444         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3445                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3446                                       ifp->rt_priority, idev->dev, 0, 0,
3447                                       GFP_ATOMIC);
3448         }
3449
3450         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3451                 addrconf_dad_start(ifp);
3452
3453         return 0;
3454 }
3455
3456 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3457 {
3458         struct inet6_ifaddr *ifp, *tmp;
3459         struct inet6_dev *idev;
3460
3461         idev = __in6_dev_get(dev);
3462         if (!idev)
3463                 return;
3464
3465         write_lock_bh(&idev->lock);
3466
3467         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3468                 if ((ifp->flags & IFA_F_PERMANENT) &&
3469                     fixup_permanent_addr(net, idev, ifp) < 0) {
3470                         write_unlock_bh(&idev->lock);
3471                         in6_ifa_hold(ifp);
3472                         ipv6_del_addr(ifp);
3473                         write_lock_bh(&idev->lock);
3474
3475                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3476                                              idev->dev->name, &ifp->addr);
3477                 }
3478         }
3479
3480         write_unlock_bh(&idev->lock);
3481 }
3482
3483 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3484                            void *ptr)
3485 {
3486         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3487         struct netdev_notifier_change_info *change_info;
3488         struct netdev_notifier_changeupper_info *info;
3489         struct inet6_dev *idev = __in6_dev_get(dev);
3490         struct net *net = dev_net(dev);
3491         int run_pending = 0;
3492         int err;
3493
3494         switch (event) {
3495         case NETDEV_REGISTER:
3496                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3497                         idev = ipv6_add_dev(dev);
3498                         if (IS_ERR(idev))
3499                                 return notifier_from_errno(PTR_ERR(idev));
3500                 }
3501                 break;
3502
3503         case NETDEV_CHANGEMTU:
3504                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3505                 if (dev->mtu < IPV6_MIN_MTU) {
3506                         addrconf_ifdown(dev, dev != net->loopback_dev);
3507                         break;
3508                 }
3509
3510                 if (idev) {
3511                         rt6_mtu_change(dev, dev->mtu);
3512                         idev->cnf.mtu6 = dev->mtu;
3513                         break;
3514                 }
3515
3516                 /* allocate new idev */
3517                 idev = ipv6_add_dev(dev);
3518                 if (IS_ERR(idev))
3519                         break;
3520
3521                 /* device is still not ready */
3522                 if (!(idev->if_flags & IF_READY))
3523                         break;
3524
3525                 run_pending = 1;
3526                 fallthrough;
3527         case NETDEV_UP:
3528         case NETDEV_CHANGE:
3529                 if (dev->flags & IFF_SLAVE)
3530                         break;
3531
3532                 if (idev && idev->cnf.disable_ipv6)
3533                         break;
3534
3535                 if (event == NETDEV_UP) {
3536                         /* restore routes for permanent addresses */
3537                         addrconf_permanent_addr(net, dev);
3538
3539                         if (!addrconf_link_ready(dev)) {
3540                                 /* device is not ready yet. */
3541                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3542                                          dev->name);
3543                                 break;
3544                         }
3545
3546                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3547                                 idev = ipv6_add_dev(dev);
3548
3549                         if (!IS_ERR_OR_NULL(idev)) {
3550                                 idev->if_flags |= IF_READY;
3551                                 run_pending = 1;
3552                         }
3553                 } else if (event == NETDEV_CHANGE) {
3554                         if (!addrconf_link_ready(dev)) {
3555                                 /* device is still not ready. */
3556                                 rt6_sync_down_dev(dev, event);
3557                                 break;
3558                         }
3559
3560                         if (!IS_ERR_OR_NULL(idev)) {
3561                                 if (idev->if_flags & IF_READY) {
3562                                         /* device is already configured -
3563                                          * but resend MLD reports, we might
3564                                          * have roamed and need to update
3565                                          * multicast snooping switches
3566                                          */
3567                                         ipv6_mc_up(idev);
3568                                         change_info = ptr;
3569                                         if (change_info->flags_changed & IFF_NOARP)
3570                                                 addrconf_dad_run(idev, true);
3571                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3572                                         break;
3573                                 }
3574                                 idev->if_flags |= IF_READY;
3575                         }
3576
3577                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3578                                 dev->name);
3579
3580                         run_pending = 1;
3581                 }
3582
3583                 switch (dev->type) {
3584 #if IS_ENABLED(CONFIG_IPV6_SIT)
3585                 case ARPHRD_SIT:
3586                         addrconf_sit_config(dev);
3587                         break;
3588 #endif
3589 #if IS_ENABLED(CONFIG_NET_IPGRE)
3590                 case ARPHRD_IPGRE:
3591                         addrconf_gre_config(dev);
3592                         break;
3593 #endif
3594                 case ARPHRD_LOOPBACK:
3595                         init_loopback(dev);
3596                         break;
3597
3598                 default:
3599                         addrconf_dev_config(dev);
3600                         break;
3601                 }
3602
3603                 if (!IS_ERR_OR_NULL(idev)) {
3604                         if (run_pending)
3605                                 addrconf_dad_run(idev, false);
3606
3607                         /* Device has an address by now */
3608                         rt6_sync_up(dev, RTNH_F_DEAD);
3609
3610                         /*
3611                          * If the MTU changed during the interface down,
3612                          * when the interface up, the changed MTU must be
3613                          * reflected in the idev as well as routers.
3614                          */
3615                         if (idev->cnf.mtu6 != dev->mtu &&
3616                             dev->mtu >= IPV6_MIN_MTU) {
3617                                 rt6_mtu_change(dev, dev->mtu);
3618                                 idev->cnf.mtu6 = dev->mtu;
3619                         }
3620                         idev->tstamp = jiffies;
3621                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3622
3623                         /*
3624                          * If the changed mtu during down is lower than
3625                          * IPV6_MIN_MTU stop IPv6 on this interface.
3626                          */
3627                         if (dev->mtu < IPV6_MIN_MTU)
3628                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3629                 }
3630                 break;
3631
3632         case NETDEV_DOWN:
3633         case NETDEV_UNREGISTER:
3634                 /*
3635                  *      Remove all addresses from this interface.
3636                  */
3637                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3638                 break;
3639
3640         case NETDEV_CHANGENAME:
3641                 if (idev) {
3642                         snmp6_unregister_dev(idev);
3643                         addrconf_sysctl_unregister(idev);
3644                         err = addrconf_sysctl_register(idev);
3645                         if (err)
3646                                 return notifier_from_errno(err);
3647                         err = snmp6_register_dev(idev);
3648                         if (err) {
3649                                 addrconf_sysctl_unregister(idev);
3650                                 return notifier_from_errno(err);
3651                         }
3652                 }
3653                 break;
3654
3655         case NETDEV_PRE_TYPE_CHANGE:
3656         case NETDEV_POST_TYPE_CHANGE:
3657                 if (idev)
3658                         addrconf_type_change(dev, event);
3659                 break;
3660
3661         case NETDEV_CHANGEUPPER:
3662                 info = ptr;
3663
3664                 /* flush all routes if dev is linked to or unlinked from
3665                  * an L3 master device (e.g., VRF)
3666                  */
3667                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3668                         addrconf_ifdown(dev, 0);
3669         }
3670
3671         return NOTIFY_OK;
3672 }
3673
3674 /*
3675  *      addrconf module should be notified of a device going up
3676  */
3677 static struct notifier_block ipv6_dev_notf = {
3678         .notifier_call = addrconf_notify,
3679         .priority = ADDRCONF_NOTIFY_PRIORITY,
3680 };
3681
3682 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3683 {
3684         struct inet6_dev *idev;
3685         ASSERT_RTNL();
3686
3687         idev = __in6_dev_get(dev);
3688
3689         if (event == NETDEV_POST_TYPE_CHANGE)
3690                 ipv6_mc_remap(idev);
3691         else if (event == NETDEV_PRE_TYPE_CHANGE)
3692                 ipv6_mc_unmap(idev);
3693 }
3694
3695 static bool addr_is_local(const struct in6_addr *addr)
3696 {
3697         return ipv6_addr_type(addr) &
3698                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3699 }
3700
3701 static int addrconf_ifdown(struct net_device *dev, int how)
3702 {
3703         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3704         struct net *net = dev_net(dev);
3705         struct inet6_dev *idev;
3706         struct inet6_ifaddr *ifa, *tmp;
3707         bool keep_addr = false;
3708         int state, i;
3709
3710         ASSERT_RTNL();
3711
3712         rt6_disable_ip(dev, event);
3713
3714         idev = __in6_dev_get(dev);
3715         if (!idev)
3716                 return -ENODEV;
3717
3718         /*
3719          * Step 1: remove reference to ipv6 device from parent device.
3720          *         Do not dev_put!
3721          */
3722         if (how) {
3723                 idev->dead = 1;
3724
3725                 /* protected by rtnl_lock */
3726                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3727
3728                 /* Step 1.5: remove snmp6 entry */
3729                 snmp6_unregister_dev(idev);
3730
3731         }
3732
3733         /* combine the user config with event to determine if permanent
3734          * addresses are to be removed from address hash table
3735          */
3736         if (!how && !idev->cnf.disable_ipv6) {
3737                 /* aggregate the system setting and interface setting */
3738                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3739
3740                 if (!_keep_addr)
3741                         _keep_addr = idev->cnf.keep_addr_on_down;
3742
3743                 keep_addr = (_keep_addr > 0);
3744         }
3745
3746         /* Step 2: clear hash table */
3747         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3748                 struct hlist_head *h = &inet6_addr_lst[i];
3749
3750                 spin_lock_bh(&addrconf_hash_lock);
3751 restart:
3752                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3753                         if (ifa->idev == idev) {
3754                                 addrconf_del_dad_work(ifa);
3755                                 /* combined flag + permanent flag decide if
3756                                  * address is retained on a down event
3757                                  */
3758                                 if (!keep_addr ||
3759                                     !(ifa->flags & IFA_F_PERMANENT) ||
3760                                     addr_is_local(&ifa->addr)) {
3761                                         hlist_del_init_rcu(&ifa->addr_lst);
3762                                         goto restart;
3763                                 }
3764                         }
3765                 }
3766                 spin_unlock_bh(&addrconf_hash_lock);
3767         }
3768
3769         write_lock_bh(&idev->lock);
3770
3771         addrconf_del_rs_timer(idev);
3772
3773         /* Step 2: clear flags for stateless addrconf */
3774         if (!how)
3775                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3776
3777         /* Step 3: clear tempaddr list */
3778         while (!list_empty(&idev->tempaddr_list)) {
3779                 ifa = list_first_entry(&idev->tempaddr_list,
3780                                        struct inet6_ifaddr, tmp_list);
3781                 list_del(&ifa->tmp_list);
3782                 write_unlock_bh(&idev->lock);
3783                 spin_lock_bh(&ifa->lock);
3784
3785                 if (ifa->ifpub) {
3786                         in6_ifa_put(ifa->ifpub);
3787                         ifa->ifpub = NULL;
3788                 }
3789                 spin_unlock_bh(&ifa->lock);
3790                 in6_ifa_put(ifa);
3791                 write_lock_bh(&idev->lock);
3792         }
3793
3794         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3795                 struct fib6_info *rt = NULL;
3796                 bool keep;
3797
3798                 addrconf_del_dad_work(ifa);
3799
3800                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3801                         !addr_is_local(&ifa->addr);
3802
3803                 write_unlock_bh(&idev->lock);
3804                 spin_lock_bh(&ifa->lock);
3805
3806                 if (keep) {
3807                         /* set state to skip the notifier below */
3808                         state = INET6_IFADDR_STATE_DEAD;
3809                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3810                         if (!(ifa->flags & IFA_F_NODAD))
3811                                 ifa->flags |= IFA_F_TENTATIVE;
3812
3813                         rt = ifa->rt;
3814                         ifa->rt = NULL;
3815                 } else {
3816                         state = ifa->state;
3817                         ifa->state = INET6_IFADDR_STATE_DEAD;
3818                 }
3819
3820                 spin_unlock_bh(&ifa->lock);
3821
3822                 if (rt)
3823                         ip6_del_rt(net, rt);
3824
3825                 if (state != INET6_IFADDR_STATE_DEAD) {
3826                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3827                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3828                 } else {
3829                         if (idev->cnf.forwarding)
3830                                 addrconf_leave_anycast(ifa);
3831                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3832                 }
3833
3834                 write_lock_bh(&idev->lock);
3835                 if (!keep) {
3836                         list_del_rcu(&ifa->if_list);
3837                         in6_ifa_put(ifa);
3838                 }
3839         }
3840
3841         write_unlock_bh(&idev->lock);
3842
3843         /* Step 5: Discard anycast and multicast list */
3844         if (how) {
3845                 ipv6_ac_destroy_dev(idev);
3846                 ipv6_mc_destroy_dev(idev);
3847         } else {
3848                 ipv6_mc_down(idev);
3849         }
3850
3851         idev->tstamp = jiffies;
3852
3853         /* Last: Shot the device (if unregistered) */
3854         if (how) {
3855                 addrconf_sysctl_unregister(idev);
3856                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3857                 neigh_ifdown(&nd_tbl, dev);
3858                 in6_dev_put(idev);
3859         }
3860         return 0;
3861 }
3862
3863 static void addrconf_rs_timer(struct timer_list *t)
3864 {
3865         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3866         struct net_device *dev = idev->dev;
3867         struct in6_addr lladdr;
3868
3869         write_lock(&idev->lock);
3870         if (idev->dead || !(idev->if_flags & IF_READY))
3871                 goto out;
3872
3873         if (!ipv6_accept_ra(idev))
3874                 goto out;
3875
3876         /* Announcement received after solicitation was sent */
3877         if (idev->if_flags & IF_RA_RCVD)
3878                 goto out;
3879
3880         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3881                 write_unlock(&idev->lock);
3882                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3883                         ndisc_send_rs(dev, &lladdr,
3884                                       &in6addr_linklocal_allrouters);
3885                 else
3886                         goto put;
3887
3888                 write_lock(&idev->lock);
3889                 idev->rs_interval = rfc3315_s14_backoff_update(
3890                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3891                 /* The wait after the last probe can be shorter */
3892                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3893                                              idev->cnf.rtr_solicits) ?
3894                                       idev->cnf.rtr_solicit_delay :
3895                                       idev->rs_interval);
3896         } else {
3897                 /*
3898                  * Note: we do not support deprecated "all on-link"
3899                  * assumption any longer.
3900                  */
3901                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3902         }
3903
3904 out:
3905         write_unlock(&idev->lock);
3906 put:
3907         in6_dev_put(idev);
3908 }
3909
3910 /*
3911  *      Duplicate Address Detection
3912  */
3913 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3914 {
3915         unsigned long rand_num;
3916         struct inet6_dev *idev = ifp->idev;
3917         u64 nonce;
3918
3919         if (ifp->flags & IFA_F_OPTIMISTIC)
3920                 rand_num = 0;
3921         else
3922                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3923
3924         nonce = 0;
3925         if (idev->cnf.enhanced_dad ||
3926             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3927                 do
3928                         get_random_bytes(&nonce, 6);
3929                 while (nonce == 0);
3930         }
3931         ifp->dad_nonce = nonce;
3932         ifp->dad_probes = idev->cnf.dad_transmits;
3933         addrconf_mod_dad_work(ifp, rand_num);
3934 }
3935
3936 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3937 {
3938         struct inet6_dev *idev = ifp->idev;
3939         struct net_device *dev = idev->dev;
3940         bool bump_id, notify = false;
3941         struct net *net;
3942
3943         addrconf_join_solict(dev, &ifp->addr);
3944
3945         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3946
3947         read_lock_bh(&idev->lock);
3948         spin_lock(&ifp->lock);
3949         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3950                 goto out;
3951
3952         net = dev_net(dev);
3953         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3954             (net->ipv6.devconf_all->accept_dad < 1 &&
3955              idev->cnf.accept_dad < 1) ||
3956             !(ifp->flags&IFA_F_TENTATIVE) ||
3957             ifp->flags & IFA_F_NODAD) {
3958                 bool send_na = false;
3959
3960                 if (ifp->flags & IFA_F_TENTATIVE &&
3961                     !(ifp->flags & IFA_F_OPTIMISTIC))
3962                         send_na = true;
3963                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3964                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3965                 spin_unlock(&ifp->lock);
3966                 read_unlock_bh(&idev->lock);
3967
3968                 addrconf_dad_completed(ifp, bump_id, send_na);
3969                 return;
3970         }
3971
3972         if (!(idev->if_flags & IF_READY)) {
3973                 spin_unlock(&ifp->lock);
3974                 read_unlock_bh(&idev->lock);
3975                 /*
3976                  * If the device is not ready:
3977                  * - keep it tentative if it is a permanent address.
3978                  * - otherwise, kill it.
3979                  */
3980                 in6_ifa_hold(ifp);
3981                 addrconf_dad_stop(ifp, 0);
3982                 return;
3983         }
3984
3985         /*
3986          * Optimistic nodes can start receiving
3987          * Frames right away
3988          */
3989         if (ifp->flags & IFA_F_OPTIMISTIC) {
3990                 ip6_ins_rt(net, ifp->rt);
3991                 if (ipv6_use_optimistic_addr(net, idev)) {
3992                         /* Because optimistic nodes can use this address,
3993                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3994                          */
3995                         notify = true;
3996                 }
3997         }
3998
3999         addrconf_dad_kick(ifp);
4000 out:
4001         spin_unlock(&ifp->lock);
4002         read_unlock_bh(&idev->lock);
4003         if (notify)
4004                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4005 }
4006
4007 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4008 {
4009         bool begin_dad = false;
4010
4011         spin_lock_bh(&ifp->lock);
4012         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4013                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4014                 begin_dad = true;
4015         }
4016         spin_unlock_bh(&ifp->lock);
4017
4018         if (begin_dad)
4019                 addrconf_mod_dad_work(ifp, 0);
4020 }
4021
4022 static void addrconf_dad_work(struct work_struct *w)
4023 {
4024         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4025                                                 struct inet6_ifaddr,
4026                                                 dad_work);
4027         struct inet6_dev *idev = ifp->idev;
4028         bool bump_id, disable_ipv6 = false;
4029         struct in6_addr mcaddr;
4030
4031         enum {
4032                 DAD_PROCESS,
4033                 DAD_BEGIN,
4034                 DAD_ABORT,
4035         } action = DAD_PROCESS;
4036
4037         rtnl_lock();
4038
4039         spin_lock_bh(&ifp->lock);
4040         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4041                 action = DAD_BEGIN;
4042                 ifp->state = INET6_IFADDR_STATE_DAD;
4043         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4044                 action = DAD_ABORT;
4045                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4046
4047                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4048                      idev->cnf.accept_dad > 1) &&
4049                     !idev->cnf.disable_ipv6 &&
4050                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4051                         struct in6_addr addr;
4052
4053                         addr.s6_addr32[0] = htonl(0xfe800000);
4054                         addr.s6_addr32[1] = 0;
4055
4056                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4057                             ipv6_addr_equal(&ifp->addr, &addr)) {
4058                                 /* DAD failed for link-local based on MAC */
4059                                 idev->cnf.disable_ipv6 = 1;
4060
4061                                 pr_info("%s: IPv6 being disabled!\n",
4062                                         ifp->idev->dev->name);
4063                                 disable_ipv6 = true;
4064                         }
4065                 }
4066         }
4067         spin_unlock_bh(&ifp->lock);
4068
4069         if (action == DAD_BEGIN) {
4070                 addrconf_dad_begin(ifp);
4071                 goto out;
4072         } else if (action == DAD_ABORT) {
4073                 in6_ifa_hold(ifp);
4074                 addrconf_dad_stop(ifp, 1);
4075                 if (disable_ipv6)
4076                         addrconf_ifdown(idev->dev, 0);
4077                 goto out;
4078         }
4079
4080         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4081                 goto out;
4082
4083         write_lock_bh(&idev->lock);
4084         if (idev->dead || !(idev->if_flags & IF_READY)) {
4085                 write_unlock_bh(&idev->lock);
4086                 goto out;
4087         }
4088
4089         spin_lock(&ifp->lock);
4090         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4091                 spin_unlock(&ifp->lock);
4092                 write_unlock_bh(&idev->lock);
4093                 goto out;
4094         }
4095
4096         if (ifp->dad_probes == 0) {
4097                 bool send_na = false;
4098
4099                 /*
4100                  * DAD was successful
4101                  */
4102
4103                 if (ifp->flags & IFA_F_TENTATIVE &&
4104                     !(ifp->flags & IFA_F_OPTIMISTIC))
4105                         send_na = true;
4106                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4107                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4108                 spin_unlock(&ifp->lock);
4109                 write_unlock_bh(&idev->lock);
4110
4111                 addrconf_dad_completed(ifp, bump_id, send_na);
4112
4113                 goto out;
4114         }
4115
4116         ifp->dad_probes--;
4117         addrconf_mod_dad_work(ifp,
4118                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4119         spin_unlock(&ifp->lock);
4120         write_unlock_bh(&idev->lock);
4121
4122         /* send a neighbour solicitation for our addr */
4123         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4124         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4125                       ifp->dad_nonce);
4126 out:
4127         in6_ifa_put(ifp);
4128         rtnl_unlock();
4129 }
4130
4131 /* ifp->idev must be at least read locked */
4132 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4133 {
4134         struct inet6_ifaddr *ifpiter;
4135         struct inet6_dev *idev = ifp->idev;
4136
4137         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4138                 if (ifpiter->scope > IFA_LINK)
4139                         break;
4140                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4141                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4142                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4143                     IFA_F_PERMANENT)
4144                         return false;
4145         }
4146         return true;
4147 }
4148
4149 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4150                                    bool send_na)
4151 {
4152         struct net_device *dev = ifp->idev->dev;
4153         struct in6_addr lladdr;
4154         bool send_rs, send_mld;
4155
4156         addrconf_del_dad_work(ifp);
4157
4158         /*
4159          *      Configure the address for reception. Now it is valid.
4160          */
4161
4162         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4163
4164         /* If added prefix is link local and we are prepared to process
4165            router advertisements, start sending router solicitations.
4166          */
4167
4168         read_lock_bh(&ifp->idev->lock);
4169         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4170         send_rs = send_mld &&
4171                   ipv6_accept_ra(ifp->idev) &&
4172                   ifp->idev->cnf.rtr_solicits != 0 &&
4173                   (dev->flags&IFF_LOOPBACK) == 0;
4174         read_unlock_bh(&ifp->idev->lock);
4175
4176         /* While dad is in progress mld report's source address is in6_addrany.
4177          * Resend with proper ll now.
4178          */
4179         if (send_mld)
4180                 ipv6_mc_dad_complete(ifp->idev);
4181
4182         /* send unsolicited NA if enabled */
4183         if (send_na &&
4184             (ifp->idev->cnf.ndisc_notify ||
4185              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4186                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4187                               /*router=*/ !!ifp->idev->cnf.forwarding,
4188                               /*solicited=*/ false, /*override=*/ true,
4189                               /*inc_opt=*/ true);
4190         }
4191
4192         if (send_rs) {
4193                 /*
4194                  *      If a host as already performed a random delay
4195                  *      [...] as part of DAD [...] there is no need
4196                  *      to delay again before sending the first RS
4197                  */
4198                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4199                         return;
4200                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4201
4202                 write_lock_bh(&ifp->idev->lock);
4203                 spin_lock(&ifp->lock);
4204                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4205                         ifp->idev->cnf.rtr_solicit_interval);
4206                 ifp->idev->rs_probes = 1;
4207                 ifp->idev->if_flags |= IF_RS_SENT;
4208                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4209                 spin_unlock(&ifp->lock);
4210                 write_unlock_bh(&ifp->idev->lock);
4211         }
4212
4213         if (bump_id)
4214                 rt_genid_bump_ipv6(dev_net(dev));
4215
4216         /* Make sure that a new temporary address will be created
4217          * before this temporary address becomes deprecated.
4218          */
4219         if (ifp->flags & IFA_F_TEMPORARY)
4220                 addrconf_verify_rtnl();
4221 }
4222
4223 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4224 {
4225         struct inet6_ifaddr *ifp;
4226
4227         read_lock_bh(&idev->lock);
4228         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4229                 spin_lock(&ifp->lock);
4230                 if ((ifp->flags & IFA_F_TENTATIVE &&
4231                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4232                         if (restart)
4233                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4234                         addrconf_dad_kick(ifp);
4235                 }
4236                 spin_unlock(&ifp->lock);
4237         }
4238         read_unlock_bh(&idev->lock);
4239 }
4240
4241 #ifdef CONFIG_PROC_FS
4242 struct if6_iter_state {
4243         struct seq_net_private p;
4244         int bucket;
4245         int offset;
4246 };
4247
4248 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4249 {
4250         struct if6_iter_state *state = seq->private;
4251         struct net *net = seq_file_net(seq);
4252         struct inet6_ifaddr *ifa = NULL;
4253         int p = 0;
4254
4255         /* initial bucket if pos is 0 */
4256         if (pos == 0) {
4257                 state->bucket = 0;
4258                 state->offset = 0;
4259         }
4260
4261         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4262                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4263                                          addr_lst) {
4264                         if (!net_eq(dev_net(ifa->idev->dev), net))
4265                                 continue;
4266                         /* sync with offset */
4267                         if (p < state->offset) {
4268                                 p++;
4269                                 continue;
4270                         }
4271                         return ifa;
4272                 }
4273
4274                 /* prepare for next bucket */
4275                 state->offset = 0;
4276                 p = 0;
4277         }
4278         return NULL;
4279 }
4280
4281 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4282                                          struct inet6_ifaddr *ifa)
4283 {
4284         struct if6_iter_state *state = seq->private;
4285         struct net *net = seq_file_net(seq);
4286
4287         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4288                 if (!net_eq(dev_net(ifa->idev->dev), net))
4289                         continue;
4290                 state->offset++;
4291                 return ifa;
4292         }
4293
4294         state->offset = 0;
4295         while (++state->bucket < IN6_ADDR_HSIZE) {
4296                 hlist_for_each_entry_rcu(ifa,
4297                                      &inet6_addr_lst[state->bucket], addr_lst) {
4298                         if (!net_eq(dev_net(ifa->idev->dev), net))
4299                                 continue;
4300                         return ifa;
4301                 }
4302         }
4303
4304         return NULL;
4305 }
4306
4307 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4308         __acquires(rcu)
4309 {
4310         rcu_read_lock();
4311         return if6_get_first(seq, *pos);
4312 }
4313
4314 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4315 {
4316         struct inet6_ifaddr *ifa;
4317
4318         ifa = if6_get_next(seq, v);
4319         ++*pos;
4320         return ifa;
4321 }
4322
4323 static void if6_seq_stop(struct seq_file *seq, void *v)
4324         __releases(rcu)
4325 {
4326         rcu_read_unlock();
4327 }
4328
4329 static int if6_seq_show(struct seq_file *seq, void *v)
4330 {
4331         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4332         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4333                    &ifp->addr,
4334                    ifp->idev->dev->ifindex,
4335                    ifp->prefix_len,
4336                    ifp->scope,
4337                    (u8) ifp->flags,
4338                    ifp->idev->dev->name);
4339         return 0;
4340 }
4341
4342 static const struct seq_operations if6_seq_ops = {
4343         .start  = if6_seq_start,
4344         .next   = if6_seq_next,
4345         .show   = if6_seq_show,
4346         .stop   = if6_seq_stop,
4347 };
4348
4349 static int __net_init if6_proc_net_init(struct net *net)
4350 {
4351         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4352                         sizeof(struct if6_iter_state)))
4353                 return -ENOMEM;
4354         return 0;
4355 }
4356
4357 static void __net_exit if6_proc_net_exit(struct net *net)
4358 {
4359         remove_proc_entry("if_inet6", net->proc_net);
4360 }
4361
4362 static struct pernet_operations if6_proc_net_ops = {
4363         .init = if6_proc_net_init,
4364         .exit = if6_proc_net_exit,
4365 };
4366
4367 int __init if6_proc_init(void)
4368 {
4369         return register_pernet_subsys(&if6_proc_net_ops);
4370 }
4371
4372 void if6_proc_exit(void)
4373 {
4374         unregister_pernet_subsys(&if6_proc_net_ops);
4375 }
4376 #endif  /* CONFIG_PROC_FS */
4377
4378 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4379 /* Check if address is a home address configured on any interface. */
4380 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4381 {
4382         unsigned int hash = inet6_addr_hash(net, addr);
4383         struct inet6_ifaddr *ifp = NULL;
4384         int ret = 0;
4385
4386         rcu_read_lock();
4387         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4388                 if (!net_eq(dev_net(ifp->idev->dev), net))
4389                         continue;
4390                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4391                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4392                         ret = 1;
4393                         break;
4394                 }
4395         }
4396         rcu_read_unlock();
4397         return ret;
4398 }
4399 #endif
4400
4401 /*
4402  *      Periodic address status verification
4403  */
4404
4405 static void addrconf_verify_rtnl(void)
4406 {
4407         unsigned long now, next, next_sec, next_sched;
4408         struct inet6_ifaddr *ifp;
4409         int i;
4410
4411         ASSERT_RTNL();
4412
4413         rcu_read_lock_bh();
4414         now = jiffies;
4415         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4416
4417         cancel_delayed_work(&addr_chk_work);
4418
4419         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4420 restart:
4421                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4422                         unsigned long age;
4423
4424                         /* When setting preferred_lft to a value not zero or
4425                          * infinity, while valid_lft is infinity
4426                          * IFA_F_PERMANENT has a non-infinity life time.
4427                          */
4428                         if ((ifp->flags & IFA_F_PERMANENT) &&
4429                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4430                                 continue;
4431
4432                         spin_lock(&ifp->lock);
4433                         /* We try to batch several events at once. */
4434                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4435
4436                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4437                             age >= ifp->valid_lft) {
4438                                 spin_unlock(&ifp->lock);
4439                                 in6_ifa_hold(ifp);
4440                                 ipv6_del_addr(ifp);
4441                                 goto restart;
4442                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4443                                 spin_unlock(&ifp->lock);
4444                                 continue;
4445                         } else if (age >= ifp->prefered_lft) {
4446                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4447                                 int deprecate = 0;
4448
4449                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4450                                         deprecate = 1;
4451                                         ifp->flags |= IFA_F_DEPRECATED;
4452                                 }
4453
4454                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4455                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4456                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4457
4458                                 spin_unlock(&ifp->lock);
4459
4460                                 if (deprecate) {
4461                                         in6_ifa_hold(ifp);
4462
4463                                         ipv6_ifa_notify(0, ifp);
4464                                         in6_ifa_put(ifp);
4465                                         goto restart;
4466                                 }
4467                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4468                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4469                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4470                                         ifp->idev->cnf.dad_transmits *
4471                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4472
4473                                 if (age >= ifp->prefered_lft - regen_advance) {
4474                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4475                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4476                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4477                                         if (!ifp->regen_count && ifpub) {
4478                                                 ifp->regen_count++;
4479                                                 in6_ifa_hold(ifp);
4480                                                 in6_ifa_hold(ifpub);
4481                                                 spin_unlock(&ifp->lock);
4482
4483                                                 spin_lock(&ifpub->lock);
4484                                                 ifpub->regen_count = 0;
4485                                                 spin_unlock(&ifpub->lock);
4486                                                 rcu_read_unlock_bh();
4487                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4488                                                 in6_ifa_put(ifpub);
4489                                                 in6_ifa_put(ifp);
4490                                                 rcu_read_lock_bh();
4491                                                 goto restart;
4492                                         }
4493                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4494                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4495                                 spin_unlock(&ifp->lock);
4496                         } else {
4497                                 /* ifp->prefered_lft <= ifp->valid_lft */
4498                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4499                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4500                                 spin_unlock(&ifp->lock);
4501                         }
4502                 }
4503         }
4504
4505         next_sec = round_jiffies_up(next);
4506         next_sched = next;
4507
4508         /* If rounded timeout is accurate enough, accept it. */
4509         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4510                 next_sched = next_sec;
4511
4512         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4513         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4514                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4515
4516         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4517                  now, next, next_sec, next_sched);
4518         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4519         rcu_read_unlock_bh();
4520 }
4521
4522 static void addrconf_verify_work(struct work_struct *w)
4523 {
4524         rtnl_lock();
4525         addrconf_verify_rtnl();
4526         rtnl_unlock();
4527 }
4528
4529 static void addrconf_verify(void)
4530 {
4531         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4532 }
4533
4534 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4535                                      struct in6_addr **peer_pfx)
4536 {
4537         struct in6_addr *pfx = NULL;
4538
4539         *peer_pfx = NULL;
4540
4541         if (addr)
4542                 pfx = nla_data(addr);
4543
4544         if (local) {
4545                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4546                         *peer_pfx = pfx;
4547                 pfx = nla_data(local);
4548         }
4549
4550         return pfx;
4551 }
4552
4553 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4554         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4555         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4556         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4557         [IFA_FLAGS]             = { .len = sizeof(u32) },
4558         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4559         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4560 };
4561
4562 static int
4563 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4564                   struct netlink_ext_ack *extack)
4565 {
4566         struct net *net = sock_net(skb->sk);
4567         struct ifaddrmsg *ifm;
4568         struct nlattr *tb[IFA_MAX+1];
4569         struct in6_addr *pfx, *peer_pfx;
4570         u32 ifa_flags;
4571         int err;
4572
4573         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4574                                      ifa_ipv6_policy, extack);
4575         if (err < 0)
4576                 return err;
4577
4578         ifm = nlmsg_data(nlh);
4579         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4580         if (!pfx)
4581                 return -EINVAL;
4582
4583         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4584
4585         /* We ignore other flags so far. */
4586         ifa_flags &= IFA_F_MANAGETEMPADDR;
4587
4588         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4589                               ifm->ifa_prefixlen);
4590 }
4591
4592 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4593                                unsigned long expires, u32 flags,
4594                                bool modify_peer)
4595 {
4596         struct fib6_info *f6i;
4597         u32 prio;
4598
4599         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4600                                         ifp->prefix_len,
4601                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4602         if (!f6i)
4603                 return -ENOENT;
4604
4605         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4606         if (f6i->fib6_metric != prio) {
4607                 /* delete old one */
4608                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4609
4610                 /* add new one */
4611                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4612                                       ifp->prefix_len,
4613                                       ifp->rt_priority, ifp->idev->dev,
4614                                       expires, flags, GFP_KERNEL);
4615         } else {
4616                 if (!expires)
4617                         fib6_clean_expires(f6i);
4618                 else
4619                         fib6_set_expires(f6i, expires);
4620
4621                 fib6_info_release(f6i);
4622         }
4623
4624         return 0;
4625 }
4626
4627 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4628 {
4629         u32 flags;
4630         clock_t expires;
4631         unsigned long timeout;
4632         bool was_managetempaddr;
4633         bool had_prefixroute;
4634         bool new_peer = false;
4635
4636         ASSERT_RTNL();
4637
4638         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4639                 return -EINVAL;
4640
4641         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4642             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4643                 return -EINVAL;
4644
4645         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4646                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4647
4648         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4649         if (addrconf_finite_timeout(timeout)) {
4650                 expires = jiffies_to_clock_t(timeout * HZ);
4651                 cfg->valid_lft = timeout;
4652                 flags = RTF_EXPIRES;
4653         } else {
4654                 expires = 0;
4655                 flags = 0;
4656                 cfg->ifa_flags |= IFA_F_PERMANENT;
4657         }
4658
4659         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4660         if (addrconf_finite_timeout(timeout)) {
4661                 if (timeout == 0)
4662                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4663                 cfg->preferred_lft = timeout;
4664         }
4665
4666         if (cfg->peer_pfx &&
4667             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4668                 if (!ipv6_addr_any(&ifp->peer_addr))
4669                         cleanup_prefix_route(ifp, expires, true, true);
4670                 new_peer = true;
4671         }
4672
4673         spin_lock_bh(&ifp->lock);
4674         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4675         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4676                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4677         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4678                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4679                         IFA_F_NOPREFIXROUTE);
4680         ifp->flags |= cfg->ifa_flags;
4681         ifp->tstamp = jiffies;
4682         ifp->valid_lft = cfg->valid_lft;
4683         ifp->prefered_lft = cfg->preferred_lft;
4684
4685         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4686                 ifp->rt_priority = cfg->rt_priority;
4687
4688         if (new_peer)
4689                 ifp->peer_addr = *cfg->peer_pfx;
4690
4691         spin_unlock_bh(&ifp->lock);
4692         if (!(ifp->flags&IFA_F_TENTATIVE))
4693                 ipv6_ifa_notify(0, ifp);
4694
4695         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4696                 int rc = -ENOENT;
4697
4698                 if (had_prefixroute)
4699                         rc = modify_prefix_route(ifp, expires, flags, false);
4700
4701                 /* prefix route could have been deleted; if so restore it */
4702                 if (rc == -ENOENT) {
4703                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4704                                               ifp->rt_priority, ifp->idev->dev,
4705                                               expires, flags, GFP_KERNEL);
4706                 }
4707
4708                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4709                         rc = modify_prefix_route(ifp, expires, flags, true);
4710
4711                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4712                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4713                                               ifp->rt_priority, ifp->idev->dev,
4714                                               expires, flags, GFP_KERNEL);
4715                 }
4716         } else if (had_prefixroute) {
4717                 enum cleanup_prefix_rt_t action;
4718                 unsigned long rt_expires;
4719
4720                 write_lock_bh(&ifp->idev->lock);
4721                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4722                 write_unlock_bh(&ifp->idev->lock);
4723
4724                 if (action != CLEANUP_PREFIX_RT_NOP) {
4725                         cleanup_prefix_route(ifp, rt_expires,
4726                                 action == CLEANUP_PREFIX_RT_DEL, false);
4727                 }
4728         }
4729
4730         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4731                 if (was_managetempaddr &&
4732                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4733                         cfg->valid_lft = 0;
4734                         cfg->preferred_lft = 0;
4735                 }
4736                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4737                                  cfg->preferred_lft, !was_managetempaddr,
4738                                  jiffies);
4739         }
4740
4741         addrconf_verify_rtnl();
4742
4743         return 0;
4744 }
4745
4746 static int
4747 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4748                   struct netlink_ext_ack *extack)
4749 {
4750         struct net *net = sock_net(skb->sk);
4751         struct ifaddrmsg *ifm;
4752         struct nlattr *tb[IFA_MAX+1];
4753         struct in6_addr *peer_pfx;
4754         struct inet6_ifaddr *ifa;
4755         struct net_device *dev;
4756         struct inet6_dev *idev;
4757         struct ifa6_config cfg;
4758         int err;
4759
4760         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4761                                      ifa_ipv6_policy, extack);
4762         if (err < 0)
4763                 return err;
4764
4765         memset(&cfg, 0, sizeof(cfg));
4766
4767         ifm = nlmsg_data(nlh);
4768         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4769         if (!cfg.pfx)
4770                 return -EINVAL;
4771
4772         cfg.peer_pfx = peer_pfx;
4773         cfg.plen = ifm->ifa_prefixlen;
4774         if (tb[IFA_RT_PRIORITY])
4775                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4776
4777         cfg.valid_lft = INFINITY_LIFE_TIME;
4778         cfg.preferred_lft = INFINITY_LIFE_TIME;
4779
4780         if (tb[IFA_CACHEINFO]) {
4781                 struct ifa_cacheinfo *ci;
4782
4783                 ci = nla_data(tb[IFA_CACHEINFO]);
4784                 cfg.valid_lft = ci->ifa_valid;
4785                 cfg.preferred_lft = ci->ifa_prefered;
4786         }
4787
4788         dev =  __dev_get_by_index(net, ifm->ifa_index);
4789         if (!dev)
4790                 return -ENODEV;
4791
4792         if (tb[IFA_FLAGS])
4793                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4794         else
4795                 cfg.ifa_flags = ifm->ifa_flags;
4796
4797         /* We ignore other flags so far. */
4798         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4799                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4800                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4801
4802         idev = ipv6_find_idev(dev);
4803         if (IS_ERR(idev))
4804                 return PTR_ERR(idev);
4805
4806         if (!ipv6_allow_optimistic_dad(net, idev))
4807                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4808
4809         if (cfg.ifa_flags & IFA_F_NODAD &&
4810             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4811                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4812                 return -EINVAL;
4813         }
4814
4815         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4816         if (!ifa) {
4817                 /*
4818                  * It would be best to check for !NLM_F_CREATE here but
4819                  * userspace already relies on not having to provide this.
4820                  */
4821                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4822         }
4823
4824         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4825             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4826                 err = -EEXIST;
4827         else
4828                 err = inet6_addr_modify(ifa, &cfg);
4829
4830         in6_ifa_put(ifa);
4831
4832         return err;
4833 }
4834
4835 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4836                           u8 scope, int ifindex)
4837 {
4838         struct ifaddrmsg *ifm;
4839
4840         ifm = nlmsg_data(nlh);
4841         ifm->ifa_family = AF_INET6;
4842         ifm->ifa_prefixlen = prefixlen;
4843         ifm->ifa_flags = flags;
4844         ifm->ifa_scope = scope;
4845         ifm->ifa_index = ifindex;
4846 }
4847
4848 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4849                          unsigned long tstamp, u32 preferred, u32 valid)
4850 {
4851         struct ifa_cacheinfo ci;
4852
4853         ci.cstamp = cstamp_delta(cstamp);
4854         ci.tstamp = cstamp_delta(tstamp);
4855         ci.ifa_prefered = preferred;
4856         ci.ifa_valid = valid;
4857
4858         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4859 }
4860
4861 static inline int rt_scope(int ifa_scope)
4862 {
4863         if (ifa_scope & IFA_HOST)
4864                 return RT_SCOPE_HOST;
4865         else if (ifa_scope & IFA_LINK)
4866                 return RT_SCOPE_LINK;
4867         else if (ifa_scope & IFA_SITE)
4868                 return RT_SCOPE_SITE;
4869         else
4870                 return RT_SCOPE_UNIVERSE;
4871 }
4872
4873 static inline int inet6_ifaddr_msgsize(void)
4874 {
4875         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4876                + nla_total_size(16) /* IFA_LOCAL */
4877                + nla_total_size(16) /* IFA_ADDRESS */
4878                + nla_total_size(sizeof(struct ifa_cacheinfo))
4879                + nla_total_size(4)  /* IFA_FLAGS */
4880                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4881 }
4882
4883 enum addr_type_t {
4884         UNICAST_ADDR,
4885         MULTICAST_ADDR,
4886         ANYCAST_ADDR,
4887 };
4888
4889 struct inet6_fill_args {
4890         u32 portid;
4891         u32 seq;
4892         int event;
4893         unsigned int flags;
4894         int netnsid;
4895         int ifindex;
4896         enum addr_type_t type;
4897 };
4898
4899 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4900                              struct inet6_fill_args *args)
4901 {
4902         struct nlmsghdr  *nlh;
4903         u32 preferred, valid;
4904
4905         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4906                         sizeof(struct ifaddrmsg), args->flags);
4907         if (!nlh)
4908                 return -EMSGSIZE;
4909
4910         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4911                       ifa->idev->dev->ifindex);
4912
4913         if (args->netnsid >= 0 &&
4914             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4915                 goto error;
4916
4917         if (!((ifa->flags&IFA_F_PERMANENT) &&
4918               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4919                 preferred = ifa->prefered_lft;
4920                 valid = ifa->valid_lft;
4921                 if (preferred != INFINITY_LIFE_TIME) {
4922                         long tval = (jiffies - ifa->tstamp)/HZ;
4923                         if (preferred > tval)
4924                                 preferred -= tval;
4925                         else
4926                                 preferred = 0;
4927                         if (valid != INFINITY_LIFE_TIME) {
4928                                 if (valid > tval)
4929                                         valid -= tval;
4930                                 else
4931                                         valid = 0;
4932                         }
4933                 }
4934         } else {
4935                 preferred = INFINITY_LIFE_TIME;
4936                 valid = INFINITY_LIFE_TIME;
4937         }
4938
4939         if (!ipv6_addr_any(&ifa->peer_addr)) {
4940                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4941                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4942                         goto error;
4943         } else
4944                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4945                         goto error;
4946
4947         if (ifa->rt_priority &&
4948             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4949                 goto error;
4950
4951         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4952                 goto error;
4953
4954         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4955                 goto error;
4956
4957         nlmsg_end(skb, nlh);
4958         return 0;
4959
4960 error:
4961         nlmsg_cancel(skb, nlh);
4962         return -EMSGSIZE;
4963 }
4964
4965 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4966                                struct inet6_fill_args *args)
4967 {
4968         struct nlmsghdr  *nlh;
4969         u8 scope = RT_SCOPE_UNIVERSE;
4970         int ifindex = ifmca->idev->dev->ifindex;
4971
4972         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4973                 scope = RT_SCOPE_SITE;
4974
4975         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4976                         sizeof(struct ifaddrmsg), args->flags);
4977         if (!nlh)
4978                 return -EMSGSIZE;
4979
4980         if (args->netnsid >= 0 &&
4981             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4982                 return -EMSGSIZE;
4983
4984         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4985         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4986             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4987                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4988                 nlmsg_cancel(skb, nlh);
4989                 return -EMSGSIZE;
4990         }
4991
4992         nlmsg_end(skb, nlh);
4993         return 0;
4994 }
4995
4996 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4997                                struct inet6_fill_args *args)
4998 {
4999         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5000         int ifindex = dev ? dev->ifindex : 1;
5001         struct nlmsghdr  *nlh;
5002         u8 scope = RT_SCOPE_UNIVERSE;
5003
5004         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5005                 scope = RT_SCOPE_SITE;
5006
5007         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5008                         sizeof(struct ifaddrmsg), args->flags);
5009         if (!nlh)
5010                 return -EMSGSIZE;
5011
5012         if (args->netnsid >= 0 &&
5013             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5014                 return -EMSGSIZE;
5015
5016         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5017         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5018             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5019                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5020                 nlmsg_cancel(skb, nlh);
5021                 return -EMSGSIZE;
5022         }
5023
5024         nlmsg_end(skb, nlh);
5025         return 0;
5026 }
5027
5028 /* called with rcu_read_lock() */
5029 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5030                           struct netlink_callback *cb, int s_ip_idx,
5031                           struct inet6_fill_args *fillargs)
5032 {
5033         struct ifmcaddr6 *ifmca;
5034         struct ifacaddr6 *ifaca;
5035         int ip_idx = 0;
5036         int err = 1;
5037
5038         read_lock_bh(&idev->lock);
5039         switch (fillargs->type) {
5040         case UNICAST_ADDR: {
5041                 struct inet6_ifaddr *ifa;
5042                 fillargs->event = RTM_NEWADDR;
5043
5044                 /* unicast address incl. temp addr */
5045                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5046                         if (ip_idx < s_ip_idx)
5047                                 goto next;
5048                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5049                         if (err < 0)
5050                                 break;
5051                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5052 next:
5053                         ip_idx++;
5054                 }
5055                 break;
5056         }
5057         case MULTICAST_ADDR:
5058                 fillargs->event = RTM_GETMULTICAST;
5059
5060                 /* multicast address */
5061                 for (ifmca = idev->mc_list; ifmca;
5062                      ifmca = ifmca->next, ip_idx++) {
5063                         if (ip_idx < s_ip_idx)
5064                                 continue;
5065                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5066                         if (err < 0)
5067                                 break;
5068                 }
5069                 break;
5070         case ANYCAST_ADDR:
5071                 fillargs->event = RTM_GETANYCAST;
5072                 /* anycast address */
5073                 for (ifaca = idev->ac_list; ifaca;
5074                      ifaca = ifaca->aca_next, ip_idx++) {
5075                         if (ip_idx < s_ip_idx)
5076                                 continue;
5077                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5078                         if (err < 0)
5079                                 break;
5080                 }
5081                 break;
5082         default:
5083                 break;
5084         }
5085         read_unlock_bh(&idev->lock);
5086         cb->args[2] = ip_idx;
5087         return err;
5088 }
5089
5090 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5091                                        struct inet6_fill_args *fillargs,
5092                                        struct net **tgt_net, struct sock *sk,
5093                                        struct netlink_callback *cb)
5094 {
5095         struct netlink_ext_ack *extack = cb->extack;
5096         struct nlattr *tb[IFA_MAX+1];
5097         struct ifaddrmsg *ifm;
5098         int err, i;
5099
5100         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5101                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5102                 return -EINVAL;
5103         }
5104
5105         ifm = nlmsg_data(nlh);
5106         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5107                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5108                 return -EINVAL;
5109         }
5110
5111         fillargs->ifindex = ifm->ifa_index;
5112         if (fillargs->ifindex) {
5113                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5114                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5115         }
5116
5117         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5118                                             ifa_ipv6_policy, extack);
5119         if (err < 0)
5120                 return err;
5121
5122         for (i = 0; i <= IFA_MAX; ++i) {
5123                 if (!tb[i])
5124                         continue;
5125
5126                 if (i == IFA_TARGET_NETNSID) {
5127                         struct net *net;
5128
5129                         fillargs->netnsid = nla_get_s32(tb[i]);
5130                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5131                         if (IS_ERR(net)) {
5132                                 fillargs->netnsid = -1;
5133                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5134                                 return PTR_ERR(net);
5135                         }
5136                         *tgt_net = net;
5137                 } else {
5138                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5139                         return -EINVAL;
5140                 }
5141         }
5142
5143         return 0;
5144 }
5145
5146 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5147                            enum addr_type_t type)
5148 {
5149         const struct nlmsghdr *nlh = cb->nlh;
5150         struct inet6_fill_args fillargs = {
5151                 .portid = NETLINK_CB(cb->skb).portid,
5152                 .seq = cb->nlh->nlmsg_seq,
5153                 .flags = NLM_F_MULTI,
5154                 .netnsid = -1,
5155                 .type = type,
5156         };
5157         struct net *net = sock_net(skb->sk);
5158         struct net *tgt_net = net;
5159         int idx, s_idx, s_ip_idx;
5160         int h, s_h;
5161         struct net_device *dev;
5162         struct inet6_dev *idev;
5163         struct hlist_head *head;
5164         int err = 0;
5165
5166         s_h = cb->args[0];
5167         s_idx = idx = cb->args[1];
5168         s_ip_idx = cb->args[2];
5169
5170         if (cb->strict_check) {
5171                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5172                                                   skb->sk, cb);
5173                 if (err < 0)
5174                         goto put_tgt_net;
5175
5176                 err = 0;
5177                 if (fillargs.ifindex) {
5178                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5179                         if (!dev) {
5180                                 err = -ENODEV;
5181                                 goto put_tgt_net;
5182                         }
5183                         idev = __in6_dev_get(dev);
5184                         if (idev) {
5185                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5186                                                      &fillargs);
5187                                 if (err > 0)
5188                                         err = 0;
5189                         }
5190                         goto put_tgt_net;
5191                 }
5192         }
5193
5194         rcu_read_lock();
5195         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5196         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5197                 idx = 0;
5198                 head = &tgt_net->dev_index_head[h];
5199                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5200                         if (idx < s_idx)
5201                                 goto cont;
5202                         if (h > s_h || idx > s_idx)
5203                                 s_ip_idx = 0;
5204                         idev = __in6_dev_get(dev);
5205                         if (!idev)
5206                                 goto cont;
5207
5208                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5209                                            &fillargs) < 0)
5210                                 goto done;
5211 cont:
5212                         idx++;
5213                 }
5214         }
5215 done:
5216         rcu_read_unlock();
5217         cb->args[0] = h;
5218         cb->args[1] = idx;
5219 put_tgt_net:
5220         if (fillargs.netnsid >= 0)
5221                 put_net(tgt_net);
5222
5223         return skb->len ? : err;
5224 }
5225
5226 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5227 {
5228         enum addr_type_t type = UNICAST_ADDR;
5229
5230         return inet6_dump_addr(skb, cb, type);
5231 }
5232
5233 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5234 {
5235         enum addr_type_t type = MULTICAST_ADDR;
5236
5237         return inet6_dump_addr(skb, cb, type);
5238 }
5239
5240
5241 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5242 {
5243         enum addr_type_t type = ANYCAST_ADDR;
5244
5245         return inet6_dump_addr(skb, cb, type);
5246 }
5247
5248 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5249                                        const struct nlmsghdr *nlh,
5250                                        struct nlattr **tb,
5251                                        struct netlink_ext_ack *extack)
5252 {
5253         struct ifaddrmsg *ifm;
5254         int i, err;
5255
5256         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5257                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5258                 return -EINVAL;
5259         }
5260
5261         if (!netlink_strict_get_check(skb))
5262                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5263                                               ifa_ipv6_policy, extack);
5264
5265         ifm = nlmsg_data(nlh);
5266         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5267                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5268                 return -EINVAL;
5269         }
5270
5271         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5272                                             ifa_ipv6_policy, extack);
5273         if (err)
5274                 return err;
5275
5276         for (i = 0; i <= IFA_MAX; i++) {
5277                 if (!tb[i])
5278                         continue;
5279
5280                 switch (i) {
5281                 case IFA_TARGET_NETNSID:
5282                 case IFA_ADDRESS:
5283                 case IFA_LOCAL:
5284                         break;
5285                 default:
5286                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5287                         return -EINVAL;
5288                 }
5289         }
5290
5291         return 0;
5292 }
5293
5294 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5295                              struct netlink_ext_ack *extack)
5296 {
5297         struct net *net = sock_net(in_skb->sk);
5298         struct inet6_fill_args fillargs = {
5299                 .portid = NETLINK_CB(in_skb).portid,
5300                 .seq = nlh->nlmsg_seq,
5301                 .event = RTM_NEWADDR,
5302                 .flags = 0,
5303                 .netnsid = -1,
5304         };
5305         struct net *tgt_net = net;
5306         struct ifaddrmsg *ifm;
5307         struct nlattr *tb[IFA_MAX+1];
5308         struct in6_addr *addr = NULL, *peer;
5309         struct net_device *dev = NULL;
5310         struct inet6_ifaddr *ifa;
5311         struct sk_buff *skb;
5312         int err;
5313
5314         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5315         if (err < 0)
5316                 return err;
5317
5318         if (tb[IFA_TARGET_NETNSID]) {
5319                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5320
5321                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5322                                                   fillargs.netnsid);
5323                 if (IS_ERR(tgt_net))
5324                         return PTR_ERR(tgt_net);
5325         }
5326
5327         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5328         if (!addr)
5329                 return -EINVAL;
5330
5331         ifm = nlmsg_data(nlh);
5332         if (ifm->ifa_index)
5333                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5334
5335         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5336         if (!ifa) {
5337                 err = -EADDRNOTAVAIL;
5338                 goto errout;
5339         }
5340
5341         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5342         if (!skb) {
5343                 err = -ENOBUFS;
5344                 goto errout_ifa;
5345         }
5346
5347         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5348         if (err < 0) {
5349                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5350                 WARN_ON(err == -EMSGSIZE);
5351                 kfree_skb(skb);
5352                 goto errout_ifa;
5353         }
5354         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5355 errout_ifa:
5356         in6_ifa_put(ifa);
5357 errout:
5358         if (dev)
5359                 dev_put(dev);
5360         if (fillargs.netnsid >= 0)
5361                 put_net(tgt_net);
5362
5363         return err;
5364 }
5365
5366 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5367 {
5368         struct sk_buff *skb;
5369         struct net *net = dev_net(ifa->idev->dev);
5370         struct inet6_fill_args fillargs = {
5371                 .portid = 0,
5372                 .seq = 0,
5373                 .event = event,
5374                 .flags = 0,
5375                 .netnsid = -1,
5376         };
5377         int err = -ENOBUFS;
5378
5379         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5380         if (!skb)
5381                 goto errout;
5382
5383         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5384         if (err < 0) {
5385                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5386                 WARN_ON(err == -EMSGSIZE);
5387                 kfree_skb(skb);
5388                 goto errout;
5389         }
5390         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5391         return;
5392 errout:
5393         if (err < 0)
5394                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5395 }
5396
5397 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5398                                 __s32 *array, int bytes)
5399 {
5400         BUG_ON(bytes < (DEVCONF_MAX * 4));
5401
5402         memset(array, 0, bytes);
5403         array[DEVCONF_FORWARDING] = cnf->forwarding;
5404         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5405         array[DEVCONF_MTU6] = cnf->mtu6;
5406         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5407         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5408         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5409         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5410         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5411         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5412                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5413         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5414                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5415         array[DEVCONF_RTR_SOLICIT_DELAY] =
5416                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5417         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5418         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5419                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5420         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5421                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5422         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5423         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5424         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5425         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5426         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5427         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5428         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5429         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5430         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5431 #ifdef CONFIG_IPV6_ROUTER_PREF
5432         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5433         array[DEVCONF_RTR_PROBE_INTERVAL] =
5434                 jiffies_to_msecs(cnf->rtr_probe_interval);
5435 #ifdef CONFIG_IPV6_ROUTE_INFO
5436         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5437         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5438 #endif
5439 #endif
5440         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5441         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5442 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5443         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5444         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5445 #endif
5446 #ifdef CONFIG_IPV6_MROUTE
5447         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5448 #endif
5449         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5450         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5451         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5452         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5453         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5454         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5455         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5456         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5457         /* we omit DEVCONF_STABLE_SECRET for now */
5458         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5459         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5460         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5461         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5462         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5463 #ifdef CONFIG_IPV6_SEG6_HMAC
5464         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5465 #endif
5466         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5467         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5468         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5469         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5470 }
5471
5472 static inline size_t inet6_ifla6_size(void)
5473 {
5474         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5475              + nla_total_size(sizeof(struct ifla_cacheinfo))
5476              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5477              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5478              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5479              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5480              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5481              + 0;
5482 }
5483
5484 static inline size_t inet6_if_nlmsg_size(void)
5485 {
5486         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5487                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5488                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5489                + nla_total_size(4) /* IFLA_MTU */
5490                + nla_total_size(4) /* IFLA_LINK */
5491                + nla_total_size(1) /* IFLA_OPERSTATE */
5492                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5493 }
5494
5495 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5496                                         int bytes)
5497 {
5498         int i;
5499         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5500         BUG_ON(pad < 0);
5501
5502         /* Use put_unaligned() because stats may not be aligned for u64. */
5503         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5504         for (i = 1; i < ICMP6_MIB_MAX; i++)
5505                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5506
5507         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5508 }
5509
5510 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5511                                         int bytes, size_t syncpoff)
5512 {
5513         int i, c;
5514         u64 buff[IPSTATS_MIB_MAX];
5515         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5516
5517         BUG_ON(pad < 0);
5518
5519         memset(buff, 0, sizeof(buff));
5520         buff[0] = IPSTATS_MIB_MAX;
5521
5522         for_each_possible_cpu(c) {
5523                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5524                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5525         }
5526
5527         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5528         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5529 }
5530
5531 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5532                              int bytes)
5533 {
5534         switch (attrtype) {
5535         case IFLA_INET6_STATS:
5536                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5537                                      offsetof(struct ipstats_mib, syncp));
5538                 break;
5539         case IFLA_INET6_ICMP6STATS:
5540                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5541                 break;
5542         }
5543 }
5544
5545 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5546                                   u32 ext_filter_mask)
5547 {
5548         struct nlattr *nla;
5549         struct ifla_cacheinfo ci;
5550
5551         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5552                 goto nla_put_failure;
5553         ci.max_reasm_len = IPV6_MAXPLEN;
5554         ci.tstamp = cstamp_delta(idev->tstamp);
5555         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5556         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5557         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5558                 goto nla_put_failure;
5559         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5560         if (!nla)
5561                 goto nla_put_failure;
5562         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5563
5564         /* XXX - MC not implemented */
5565
5566         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5567                 return 0;
5568
5569         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5570         if (!nla)
5571                 goto nla_put_failure;
5572         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5573
5574         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5575         if (!nla)
5576                 goto nla_put_failure;
5577         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5578
5579         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5580         if (!nla)
5581                 goto nla_put_failure;
5582         read_lock_bh(&idev->lock);
5583         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5584         read_unlock_bh(&idev->lock);
5585
5586         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5587                 goto nla_put_failure;
5588
5589         return 0;
5590
5591 nla_put_failure:
5592         return -EMSGSIZE;
5593 }
5594
5595 static size_t inet6_get_link_af_size(const struct net_device *dev,
5596                                      u32 ext_filter_mask)
5597 {
5598         if (!__in6_dev_get(dev))
5599                 return 0;
5600
5601         return inet6_ifla6_size();
5602 }
5603
5604 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5605                               u32 ext_filter_mask)
5606 {
5607         struct inet6_dev *idev = __in6_dev_get(dev);
5608
5609         if (!idev)
5610                 return -ENODATA;
5611
5612         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5613                 return -EMSGSIZE;
5614
5615         return 0;
5616 }
5617
5618 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5619 {
5620         struct inet6_ifaddr *ifp;
5621         struct net_device *dev = idev->dev;
5622         bool clear_token, update_rs = false;
5623         struct in6_addr ll_addr;
5624
5625         ASSERT_RTNL();
5626
5627         if (!token)
5628                 return -EINVAL;
5629         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5630                 return -EINVAL;
5631         if (!ipv6_accept_ra(idev))
5632                 return -EINVAL;
5633         if (idev->cnf.rtr_solicits == 0)
5634                 return -EINVAL;
5635
5636         write_lock_bh(&idev->lock);
5637
5638         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5639         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5640
5641         write_unlock_bh(&idev->lock);
5642
5643         clear_token = ipv6_addr_any(token);
5644         if (clear_token)
5645                 goto update_lft;
5646
5647         if (!idev->dead && (idev->if_flags & IF_READY) &&
5648             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5649                              IFA_F_OPTIMISTIC)) {
5650                 /* If we're not ready, then normal ifup will take care
5651                  * of this. Otherwise, we need to request our rs here.
5652                  */
5653                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5654                 update_rs = true;
5655         }
5656
5657 update_lft:
5658         write_lock_bh(&idev->lock);
5659
5660         if (update_rs) {
5661                 idev->if_flags |= IF_RS_SENT;
5662                 idev->rs_interval = rfc3315_s14_backoff_init(
5663                         idev->cnf.rtr_solicit_interval);
5664                 idev->rs_probes = 1;
5665                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5666         }
5667
5668         /* Well, that's kinda nasty ... */
5669         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5670                 spin_lock(&ifp->lock);
5671                 if (ifp->tokenized) {
5672                         ifp->valid_lft = 0;
5673                         ifp->prefered_lft = 0;
5674                 }
5675                 spin_unlock(&ifp->lock);
5676         }
5677
5678         write_unlock_bh(&idev->lock);
5679         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5680         addrconf_verify_rtnl();
5681         return 0;
5682 }
5683
5684 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5685         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5686         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5687 };
5688
5689 static int check_addr_gen_mode(int mode)
5690 {
5691         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5692             mode != IN6_ADDR_GEN_MODE_NONE &&
5693             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5694             mode != IN6_ADDR_GEN_MODE_RANDOM)
5695                 return -EINVAL;
5696         return 1;
5697 }
5698
5699 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5700                                 int mode)
5701 {
5702         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5703             !idev->cnf.stable_secret.initialized &&
5704             !net->ipv6.devconf_dflt->stable_secret.initialized)
5705                 return -EINVAL;
5706         return 1;
5707 }
5708
5709 static int inet6_validate_link_af(const struct net_device *dev,
5710                                   const struct nlattr *nla)
5711 {
5712         struct nlattr *tb[IFLA_INET6_MAX + 1];
5713         struct inet6_dev *idev = NULL;
5714         int err;
5715
5716         if (dev) {
5717                 idev = __in6_dev_get(dev);
5718                 if (!idev)
5719                         return -EAFNOSUPPORT;
5720         }
5721
5722         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5723                                           inet6_af_policy, NULL);
5724         if (err)
5725                 return err;
5726
5727         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5728                 return -EINVAL;
5729
5730         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5731                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5732
5733                 if (check_addr_gen_mode(mode) < 0)
5734                         return -EINVAL;
5735                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5736                         return -EINVAL;
5737         }
5738
5739         return 0;
5740 }
5741
5742 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5743 {
5744         struct inet6_dev *idev = __in6_dev_get(dev);
5745         struct nlattr *tb[IFLA_INET6_MAX + 1];
5746         int err;
5747
5748         if (!idev)
5749                 return -EAFNOSUPPORT;
5750
5751         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5752                 BUG();
5753
5754         if (tb[IFLA_INET6_TOKEN]) {
5755                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5756                 if (err)
5757                         return err;
5758         }
5759
5760         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5761                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5762
5763                 idev->cnf.addr_gen_mode = mode;
5764         }
5765
5766         return 0;
5767 }
5768
5769 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5770                              u32 portid, u32 seq, int event, unsigned int flags)
5771 {
5772         struct net_device *dev = idev->dev;
5773         struct ifinfomsg *hdr;
5774         struct nlmsghdr *nlh;
5775         void *protoinfo;
5776
5777         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5778         if (!nlh)
5779                 return -EMSGSIZE;
5780
5781         hdr = nlmsg_data(nlh);
5782         hdr->ifi_family = AF_INET6;
5783         hdr->__ifi_pad = 0;
5784         hdr->ifi_type = dev->type;
5785         hdr->ifi_index = dev->ifindex;
5786         hdr->ifi_flags = dev_get_flags(dev);
5787         hdr->ifi_change = 0;
5788
5789         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5790             (dev->addr_len &&
5791              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5792             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5793             (dev->ifindex != dev_get_iflink(dev) &&
5794              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5795             nla_put_u8(skb, IFLA_OPERSTATE,
5796                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5797                 goto nla_put_failure;
5798         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5799         if (!protoinfo)
5800                 goto nla_put_failure;
5801
5802         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5803                 goto nla_put_failure;
5804
5805         nla_nest_end(skb, protoinfo);
5806         nlmsg_end(skb, nlh);
5807         return 0;
5808
5809 nla_put_failure:
5810         nlmsg_cancel(skb, nlh);
5811         return -EMSGSIZE;
5812 }
5813
5814 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5815                                    struct netlink_ext_ack *extack)
5816 {
5817         struct ifinfomsg *ifm;
5818
5819         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5820                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5821                 return -EINVAL;
5822         }
5823
5824         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5825                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5826                 return -EINVAL;
5827         }
5828
5829         ifm = nlmsg_data(nlh);
5830         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5831             ifm->ifi_change || ifm->ifi_index) {
5832                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5833                 return -EINVAL;
5834         }
5835
5836         return 0;
5837 }
5838
5839 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5840 {
5841         struct net *net = sock_net(skb->sk);
5842         int h, s_h;
5843         int idx = 0, s_idx;
5844         struct net_device *dev;
5845         struct inet6_dev *idev;
5846         struct hlist_head *head;
5847
5848         /* only requests using strict checking can pass data to
5849          * influence the dump
5850          */
5851         if (cb->strict_check) {
5852                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5853
5854                 if (err < 0)
5855                         return err;
5856         }
5857
5858         s_h = cb->args[0];
5859         s_idx = cb->args[1];
5860
5861         rcu_read_lock();
5862         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5863                 idx = 0;
5864                 head = &net->dev_index_head[h];
5865                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5866                         if (idx < s_idx)
5867                                 goto cont;
5868                         idev = __in6_dev_get(dev);
5869                         if (!idev)
5870                                 goto cont;
5871                         if (inet6_fill_ifinfo(skb, idev,
5872                                               NETLINK_CB(cb->skb).portid,
5873                                               cb->nlh->nlmsg_seq,
5874                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5875                                 goto out;
5876 cont:
5877                         idx++;
5878                 }
5879         }
5880 out:
5881         rcu_read_unlock();
5882         cb->args[1] = idx;
5883         cb->args[0] = h;
5884
5885         return skb->len;
5886 }
5887
5888 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5889 {
5890         struct sk_buff *skb;
5891         struct net *net = dev_net(idev->dev);
5892         int err = -ENOBUFS;
5893
5894         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5895         if (!skb)
5896                 goto errout;
5897
5898         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5899         if (err < 0) {
5900                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5901                 WARN_ON(err == -EMSGSIZE);
5902                 kfree_skb(skb);
5903                 goto errout;
5904         }
5905         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5906         return;
5907 errout:
5908         if (err < 0)
5909                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5910 }
5911
5912 static inline size_t inet6_prefix_nlmsg_size(void)
5913 {
5914         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5915                + nla_total_size(sizeof(struct in6_addr))
5916                + nla_total_size(sizeof(struct prefix_cacheinfo));
5917 }
5918
5919 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5920                              struct prefix_info *pinfo, u32 portid, u32 seq,
5921                              int event, unsigned int flags)
5922 {
5923         struct prefixmsg *pmsg;
5924         struct nlmsghdr *nlh;
5925         struct prefix_cacheinfo ci;
5926
5927         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5928         if (!nlh)
5929                 return -EMSGSIZE;
5930
5931         pmsg = nlmsg_data(nlh);
5932         pmsg->prefix_family = AF_INET6;
5933         pmsg->prefix_pad1 = 0;
5934         pmsg->prefix_pad2 = 0;
5935         pmsg->prefix_ifindex = idev->dev->ifindex;
5936         pmsg->prefix_len = pinfo->prefix_len;
5937         pmsg->prefix_type = pinfo->type;
5938         pmsg->prefix_pad3 = 0;
5939         pmsg->prefix_flags = 0;
5940         if (pinfo->onlink)
5941                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5942         if (pinfo->autoconf)
5943                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5944
5945         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5946                 goto nla_put_failure;
5947         ci.preferred_time = ntohl(pinfo->prefered);
5948         ci.valid_time = ntohl(pinfo->valid);
5949         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5950                 goto nla_put_failure;
5951         nlmsg_end(skb, nlh);
5952         return 0;
5953
5954 nla_put_failure:
5955         nlmsg_cancel(skb, nlh);
5956         return -EMSGSIZE;
5957 }
5958
5959 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5960                          struct prefix_info *pinfo)
5961 {
5962         struct sk_buff *skb;
5963         struct net *net = dev_net(idev->dev);
5964         int err = -ENOBUFS;
5965
5966         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5967         if (!skb)
5968                 goto errout;
5969
5970         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5971         if (err < 0) {
5972                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5973                 WARN_ON(err == -EMSGSIZE);
5974                 kfree_skb(skb);
5975                 goto errout;
5976         }
5977         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5978         return;
5979 errout:
5980         if (err < 0)
5981                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5982 }
5983
5984 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5985 {
5986         struct net *net = dev_net(ifp->idev->dev);
5987
5988         if (event)
5989                 ASSERT_RTNL();
5990
5991         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5992
5993         switch (event) {
5994         case RTM_NEWADDR:
5995                 /*
5996                  * If the address was optimistic we inserted the route at the
5997                  * start of our DAD process, so we don't need to do it again.
5998                  * If the device was taken down in the middle of the DAD
5999                  * cycle there is a race where we could get here without a
6000                  * host route, so nothing to insert. That will be fixed when
6001                  * the device is brought up.
6002                  */
6003                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6004                         ip6_ins_rt(net, ifp->rt);
6005                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6006                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6007                                 &ifp->addr, ifp->idev->dev->name);
6008                 }
6009
6010                 if (ifp->idev->cnf.forwarding)
6011                         addrconf_join_anycast(ifp);
6012                 if (!ipv6_addr_any(&ifp->peer_addr))
6013                         addrconf_prefix_route(&ifp->peer_addr, 128,
6014                                               ifp->rt_priority, ifp->idev->dev,
6015                                               0, 0, GFP_ATOMIC);
6016                 break;
6017         case RTM_DELADDR:
6018                 if (ifp->idev->cnf.forwarding)
6019                         addrconf_leave_anycast(ifp);
6020                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6021                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6022                         struct fib6_info *rt;
6023
6024                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6025                                                        ifp->idev->dev, 0, 0,
6026                                                        false);
6027                         if (rt)
6028                                 ip6_del_rt(net, rt);
6029                 }
6030                 if (ifp->rt) {
6031                         ip6_del_rt(net, ifp->rt);
6032                         ifp->rt = NULL;
6033                 }
6034                 rt_genid_bump_ipv6(net);
6035                 break;
6036         }
6037         atomic_inc(&net->ipv6.dev_addr_genid);
6038 }
6039
6040 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6041 {
6042         rcu_read_lock_bh();
6043         if (likely(ifp->idev->dead == 0))
6044                 __ipv6_ifa_notify(event, ifp);
6045         rcu_read_unlock_bh();
6046 }
6047
6048 #ifdef CONFIG_SYSCTL
6049
6050 static
6051 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6052                            void __user *buffer, size_t *lenp, loff_t *ppos)
6053 {
6054         int *valp = ctl->data;
6055         int val = *valp;
6056         loff_t pos = *ppos;
6057         struct ctl_table lctl;
6058         int ret;
6059
6060         /*
6061          * ctl->data points to idev->cnf.forwarding, we should
6062          * not modify it until we get the rtnl lock.
6063          */
6064         lctl = *ctl;
6065         lctl.data = &val;
6066
6067         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6068
6069         if (write)
6070                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6071         if (ret)
6072                 *ppos = pos;
6073         return ret;
6074 }
6075
6076 static
6077 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6078                         void __user *buffer, size_t *lenp, loff_t *ppos)
6079 {
6080         struct inet6_dev *idev = ctl->extra1;
6081         int min_mtu = IPV6_MIN_MTU;
6082         struct ctl_table lctl;
6083
6084         lctl = *ctl;
6085         lctl.extra1 = &min_mtu;
6086         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6087
6088         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6089 }
6090
6091 static void dev_disable_change(struct inet6_dev *idev)
6092 {
6093         struct netdev_notifier_info info;
6094
6095         if (!idev || !idev->dev)
6096                 return;
6097
6098         netdev_notifier_info_init(&info, idev->dev);
6099         if (idev->cnf.disable_ipv6)
6100                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6101         else
6102                 addrconf_notify(NULL, NETDEV_UP, &info);
6103 }
6104
6105 static void addrconf_disable_change(struct net *net, __s32 newf)
6106 {
6107         struct net_device *dev;
6108         struct inet6_dev *idev;
6109
6110         for_each_netdev(net, dev) {
6111                 idev = __in6_dev_get(dev);
6112                 if (idev) {
6113                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6114                         idev->cnf.disable_ipv6 = newf;
6115                         if (changed)
6116                                 dev_disable_change(idev);
6117                 }
6118         }
6119 }
6120
6121 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6122 {
6123         struct net *net;
6124         int old;
6125
6126         if (!rtnl_trylock())
6127                 return restart_syscall();
6128
6129         net = (struct net *)table->extra2;
6130         old = *p;
6131         *p = newf;
6132
6133         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6134                 rtnl_unlock();
6135                 return 0;
6136         }
6137
6138         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6139                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6140                 addrconf_disable_change(net, newf);
6141         } else if ((!newf) ^ (!old))
6142                 dev_disable_change((struct inet6_dev *)table->extra1);
6143
6144         rtnl_unlock();
6145         return 0;
6146 }
6147
6148 static
6149 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6150                             void __user *buffer, size_t *lenp, loff_t *ppos)
6151 {
6152         int *valp = ctl->data;
6153         int val = *valp;
6154         loff_t pos = *ppos;
6155         struct ctl_table lctl;
6156         int ret;
6157
6158         /*
6159          * ctl->data points to idev->cnf.disable_ipv6, we should
6160          * not modify it until we get the rtnl lock.
6161          */
6162         lctl = *ctl;
6163         lctl.data = &val;
6164
6165         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6166
6167         if (write)
6168                 ret = addrconf_disable_ipv6(ctl, valp, val);
6169         if (ret)
6170                 *ppos = pos;
6171         return ret;
6172 }
6173
6174 static
6175 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6176                               void __user *buffer, size_t *lenp, loff_t *ppos)
6177 {
6178         int *valp = ctl->data;
6179         int ret;
6180         int old, new;
6181
6182         old = *valp;
6183         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6184         new = *valp;
6185
6186         if (write && old != new) {
6187                 struct net *net = ctl->extra2;
6188
6189                 if (!rtnl_trylock())
6190                         return restart_syscall();
6191
6192                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6193                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6194                                                      NETCONFA_PROXY_NEIGH,
6195                                                      NETCONFA_IFINDEX_DEFAULT,
6196                                                      net->ipv6.devconf_dflt);
6197                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6198                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6199                                                      NETCONFA_PROXY_NEIGH,
6200                                                      NETCONFA_IFINDEX_ALL,
6201                                                      net->ipv6.devconf_all);
6202                 else {
6203                         struct inet6_dev *idev = ctl->extra1;
6204
6205                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6206                                                      NETCONFA_PROXY_NEIGH,
6207                                                      idev->dev->ifindex,
6208                                                      &idev->cnf);
6209                 }
6210                 rtnl_unlock();
6211         }
6212
6213         return ret;
6214 }
6215
6216 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6217                                          void __user *buffer, size_t *lenp,
6218                                          loff_t *ppos)
6219 {
6220         int ret = 0;
6221         u32 new_val;
6222         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6223         struct net *net = (struct net *)ctl->extra2;
6224         struct ctl_table tmp = {
6225                 .data = &new_val,
6226                 .maxlen = sizeof(new_val),
6227                 .mode = ctl->mode,
6228         };
6229
6230         if (!rtnl_trylock())
6231                 return restart_syscall();
6232
6233         new_val = *((u32 *)ctl->data);
6234
6235         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6236         if (ret != 0)
6237                 goto out;
6238
6239         if (write) {
6240                 if (check_addr_gen_mode(new_val) < 0) {
6241                         ret = -EINVAL;
6242                         goto out;
6243                 }
6244
6245                 if (idev) {
6246                         if (check_stable_privacy(idev, net, new_val) < 0) {
6247                                 ret = -EINVAL;
6248                                 goto out;
6249                         }
6250
6251                         if (idev->cnf.addr_gen_mode != new_val) {
6252                                 idev->cnf.addr_gen_mode = new_val;
6253                                 addrconf_dev_config(idev->dev);
6254                         }
6255                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6256                         struct net_device *dev;
6257
6258                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6259                         for_each_netdev(net, dev) {
6260                                 idev = __in6_dev_get(dev);
6261                                 if (idev &&
6262                                     idev->cnf.addr_gen_mode != new_val) {
6263                                         idev->cnf.addr_gen_mode = new_val;
6264                                         addrconf_dev_config(idev->dev);
6265                                 }
6266                         }
6267                 }
6268
6269                 *((u32 *)ctl->data) = new_val;
6270         }
6271
6272 out:
6273         rtnl_unlock();
6274
6275         return ret;
6276 }
6277
6278 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6279                                          void __user *buffer, size_t *lenp,
6280                                          loff_t *ppos)
6281 {
6282         int err;
6283         struct in6_addr addr;
6284         char str[IPV6_MAX_STRLEN];
6285         struct ctl_table lctl = *ctl;
6286         struct net *net = ctl->extra2;
6287         struct ipv6_stable_secret *secret = ctl->data;
6288
6289         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6290                 return -EIO;
6291
6292         lctl.maxlen = IPV6_MAX_STRLEN;
6293         lctl.data = str;
6294
6295         if (!rtnl_trylock())
6296                 return restart_syscall();
6297
6298         if (!write && !secret->initialized) {
6299                 err = -EIO;
6300                 goto out;
6301         }
6302
6303         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6304         if (err >= sizeof(str)) {
6305                 err = -EIO;
6306                 goto out;
6307         }
6308
6309         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6310         if (err || !write)
6311                 goto out;
6312
6313         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6314                 err = -EIO;
6315                 goto out;
6316         }
6317
6318         secret->initialized = true;
6319         secret->secret = addr;
6320
6321         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6322                 struct net_device *dev;
6323
6324                 for_each_netdev(net, dev) {
6325                         struct inet6_dev *idev = __in6_dev_get(dev);
6326
6327                         if (idev) {
6328                                 idev->cnf.addr_gen_mode =
6329                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6330                         }
6331                 }
6332         } else {
6333                 struct inet6_dev *idev = ctl->extra1;
6334
6335                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6336         }
6337
6338 out:
6339         rtnl_unlock();
6340
6341         return err;
6342 }
6343
6344 static
6345 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6346                                                 int write,
6347                                                 void __user *buffer,
6348                                                 size_t *lenp,
6349                                                 loff_t *ppos)
6350 {
6351         int *valp = ctl->data;
6352         int val = *valp;
6353         loff_t pos = *ppos;
6354         struct ctl_table lctl;
6355         int ret;
6356
6357         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6358          * we should not modify it until we get the rtnl lock.
6359          */
6360         lctl = *ctl;
6361         lctl.data = &val;
6362
6363         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6364
6365         if (write)
6366                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6367         if (ret)
6368                 *ppos = pos;
6369         return ret;
6370 }
6371
6372 static
6373 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6374 {
6375         if (rt) {
6376                 if (action)
6377                         rt->dst.flags |= DST_NOPOLICY;
6378                 else
6379                         rt->dst.flags &= ~DST_NOPOLICY;
6380         }
6381 }
6382
6383 static
6384 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6385 {
6386         struct inet6_ifaddr *ifa;
6387
6388         read_lock_bh(&idev->lock);
6389         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6390                 spin_lock(&ifa->lock);
6391                 if (ifa->rt) {
6392                         /* host routes only use builtin fib6_nh */
6393                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6394                         int cpu;
6395
6396                         rcu_read_lock();
6397                         ifa->rt->dst_nopolicy = val ? true : false;
6398                         if (nh->rt6i_pcpu) {
6399                                 for_each_possible_cpu(cpu) {
6400                                         struct rt6_info **rtp;
6401
6402                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6403                                         addrconf_set_nopolicy(*rtp, val);
6404                                 }
6405                         }
6406                         rcu_read_unlock();
6407                 }
6408                 spin_unlock(&ifa->lock);
6409         }
6410         read_unlock_bh(&idev->lock);
6411 }
6412
6413 static
6414 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6415 {
6416         struct inet6_dev *idev;
6417         struct net *net;
6418
6419         if (!rtnl_trylock())
6420                 return restart_syscall();
6421
6422         *valp = val;
6423
6424         net = (struct net *)ctl->extra2;
6425         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6426                 rtnl_unlock();
6427                 return 0;
6428         }
6429
6430         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6431                 struct net_device *dev;
6432
6433                 for_each_netdev(net, dev) {
6434                         idev = __in6_dev_get(dev);
6435                         if (idev)
6436                                 addrconf_disable_policy_idev(idev, val);
6437                 }
6438         } else {
6439                 idev = (struct inet6_dev *)ctl->extra1;
6440                 addrconf_disable_policy_idev(idev, val);
6441         }
6442
6443         rtnl_unlock();
6444         return 0;
6445 }
6446
6447 static
6448 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6449                                    void __user *buffer, size_t *lenp,
6450                                    loff_t *ppos)
6451 {
6452         int *valp = ctl->data;
6453         int val = *valp;
6454         loff_t pos = *ppos;
6455         struct ctl_table lctl;
6456         int ret;
6457
6458         lctl = *ctl;
6459         lctl.data = &val;
6460         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6461
6462         if (write && (*valp != val))
6463                 ret = addrconf_disable_policy(ctl, valp, val);
6464
6465         if (ret)
6466                 *ppos = pos;
6467
6468         return ret;
6469 }
6470
6471 static int minus_one = -1;
6472 static const int two_five_five = 255;
6473
6474 static const struct ctl_table addrconf_sysctl[] = {
6475         {
6476                 .procname       = "forwarding",
6477                 .data           = &ipv6_devconf.forwarding,
6478                 .maxlen         = sizeof(int),
6479                 .mode           = 0644,
6480                 .proc_handler   = addrconf_sysctl_forward,
6481         },
6482         {
6483                 .procname       = "hop_limit",
6484                 .data           = &ipv6_devconf.hop_limit,
6485                 .maxlen         = sizeof(int),
6486                 .mode           = 0644,
6487                 .proc_handler   = proc_dointvec_minmax,
6488                 .extra1         = (void *)SYSCTL_ONE,
6489                 .extra2         = (void *)&two_five_five,
6490         },
6491         {
6492                 .procname       = "mtu",
6493                 .data           = &ipv6_devconf.mtu6,
6494                 .maxlen         = sizeof(int),
6495                 .mode           = 0644,
6496                 .proc_handler   = addrconf_sysctl_mtu,
6497         },
6498         {
6499                 .procname       = "accept_ra",
6500                 .data           = &ipv6_devconf.accept_ra,
6501                 .maxlen         = sizeof(int),
6502                 .mode           = 0644,
6503                 .proc_handler   = proc_dointvec,
6504         },
6505         {
6506                 .procname       = "accept_redirects",
6507                 .data           = &ipv6_devconf.accept_redirects,
6508                 .maxlen         = sizeof(int),
6509                 .mode           = 0644,
6510                 .proc_handler   = proc_dointvec,
6511         },
6512         {
6513                 .procname       = "autoconf",
6514                 .data           = &ipv6_devconf.autoconf,
6515                 .maxlen         = sizeof(int),
6516                 .mode           = 0644,
6517                 .proc_handler   = proc_dointvec,
6518         },
6519         {
6520                 .procname       = "dad_transmits",
6521                 .data           = &ipv6_devconf.dad_transmits,
6522                 .maxlen         = sizeof(int),
6523                 .mode           = 0644,
6524                 .proc_handler   = proc_dointvec,
6525         },
6526         {
6527                 .procname       = "router_solicitations",
6528                 .data           = &ipv6_devconf.rtr_solicits,
6529                 .maxlen         = sizeof(int),
6530                 .mode           = 0644,
6531                 .proc_handler   = proc_dointvec_minmax,
6532                 .extra1         = &minus_one,
6533         },
6534         {
6535                 .procname       = "router_solicitation_interval",
6536                 .data           = &ipv6_devconf.rtr_solicit_interval,
6537                 .maxlen         = sizeof(int),
6538                 .mode           = 0644,
6539                 .proc_handler   = proc_dointvec_jiffies,
6540         },
6541         {
6542                 .procname       = "router_solicitation_max_interval",
6543                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6544                 .maxlen         = sizeof(int),
6545                 .mode           = 0644,
6546                 .proc_handler   = proc_dointvec_jiffies,
6547         },
6548         {
6549                 .procname       = "router_solicitation_delay",
6550                 .data           = &ipv6_devconf.rtr_solicit_delay,
6551                 .maxlen         = sizeof(int),
6552                 .mode           = 0644,
6553                 .proc_handler   = proc_dointvec_jiffies,
6554         },
6555         {
6556                 .procname       = "force_mld_version",
6557                 .data           = &ipv6_devconf.force_mld_version,
6558                 .maxlen         = sizeof(int),
6559                 .mode           = 0644,
6560                 .proc_handler   = proc_dointvec,
6561         },
6562         {
6563                 .procname       = "mldv1_unsolicited_report_interval",
6564                 .data           =
6565                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6566                 .maxlen         = sizeof(int),
6567                 .mode           = 0644,
6568                 .proc_handler   = proc_dointvec_ms_jiffies,
6569         },
6570         {
6571                 .procname       = "mldv2_unsolicited_report_interval",
6572                 .data           =
6573                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6574                 .maxlen         = sizeof(int),
6575                 .mode           = 0644,
6576                 .proc_handler   = proc_dointvec_ms_jiffies,
6577         },
6578         {
6579                 .procname       = "use_tempaddr",
6580                 .data           = &ipv6_devconf.use_tempaddr,
6581                 .maxlen         = sizeof(int),
6582                 .mode           = 0644,
6583                 .proc_handler   = proc_dointvec,
6584         },
6585         {
6586                 .procname       = "temp_valid_lft",
6587                 .data           = &ipv6_devconf.temp_valid_lft,
6588                 .maxlen         = sizeof(int),
6589                 .mode           = 0644,
6590                 .proc_handler   = proc_dointvec,
6591         },
6592         {
6593                 .procname       = "temp_prefered_lft",
6594                 .data           = &ipv6_devconf.temp_prefered_lft,
6595                 .maxlen         = sizeof(int),
6596                 .mode           = 0644,
6597                 .proc_handler   = proc_dointvec,
6598         },
6599         {
6600                 .procname       = "regen_max_retry",
6601                 .data           = &ipv6_devconf.regen_max_retry,
6602                 .maxlen         = sizeof(int),
6603                 .mode           = 0644,
6604                 .proc_handler   = proc_dointvec,
6605         },
6606         {
6607                 .procname       = "max_desync_factor",
6608                 .data           = &ipv6_devconf.max_desync_factor,
6609                 .maxlen         = sizeof(int),
6610                 .mode           = 0644,
6611                 .proc_handler   = proc_dointvec,
6612         },
6613         {
6614                 .procname       = "max_addresses",
6615                 .data           = &ipv6_devconf.max_addresses,
6616                 .maxlen         = sizeof(int),
6617                 .mode           = 0644,
6618                 .proc_handler   = proc_dointvec,
6619         },
6620         {
6621                 .procname       = "accept_ra_defrtr",
6622                 .data           = &ipv6_devconf.accept_ra_defrtr,
6623                 .maxlen         = sizeof(int),
6624                 .mode           = 0644,
6625                 .proc_handler   = proc_dointvec,
6626         },
6627         {
6628                 .procname       = "accept_ra_min_hop_limit",
6629                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6630                 .maxlen         = sizeof(int),
6631                 .mode           = 0644,
6632                 .proc_handler   = proc_dointvec,
6633         },
6634         {
6635                 .procname       = "accept_ra_pinfo",
6636                 .data           = &ipv6_devconf.accept_ra_pinfo,
6637                 .maxlen         = sizeof(int),
6638                 .mode           = 0644,
6639                 .proc_handler   = proc_dointvec,
6640         },
6641 #ifdef CONFIG_IPV6_ROUTER_PREF
6642         {
6643                 .procname       = "accept_ra_rtr_pref",
6644                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6645                 .maxlen         = sizeof(int),
6646                 .mode           = 0644,
6647                 .proc_handler   = proc_dointvec,
6648         },
6649         {
6650                 .procname       = "router_probe_interval",
6651                 .data           = &ipv6_devconf.rtr_probe_interval,
6652                 .maxlen         = sizeof(int),
6653                 .mode           = 0644,
6654                 .proc_handler   = proc_dointvec_jiffies,
6655         },
6656 #ifdef CONFIG_IPV6_ROUTE_INFO
6657         {
6658                 .procname       = "accept_ra_rt_info_min_plen",
6659                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6660                 .maxlen         = sizeof(int),
6661                 .mode           = 0644,
6662                 .proc_handler   = proc_dointvec,
6663         },
6664         {
6665                 .procname       = "accept_ra_rt_info_max_plen",
6666                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6667                 .maxlen         = sizeof(int),
6668                 .mode           = 0644,
6669                 .proc_handler   = proc_dointvec,
6670         },
6671 #endif
6672 #endif
6673         {
6674                 .procname       = "proxy_ndp",
6675                 .data           = &ipv6_devconf.proxy_ndp,
6676                 .maxlen         = sizeof(int),
6677                 .mode           = 0644,
6678                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6679         },
6680         {
6681                 .procname       = "accept_source_route",
6682                 .data           = &ipv6_devconf.accept_source_route,
6683                 .maxlen         = sizeof(int),
6684                 .mode           = 0644,
6685                 .proc_handler   = proc_dointvec,
6686         },
6687 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6688         {
6689                 .procname       = "optimistic_dad",
6690                 .data           = &ipv6_devconf.optimistic_dad,
6691                 .maxlen         = sizeof(int),
6692                 .mode           = 0644,
6693                 .proc_handler   = proc_dointvec,
6694         },
6695         {
6696                 .procname       = "use_optimistic",
6697                 .data           = &ipv6_devconf.use_optimistic,
6698                 .maxlen         = sizeof(int),
6699                 .mode           = 0644,
6700                 .proc_handler   = proc_dointvec,
6701         },
6702 #endif
6703 #ifdef CONFIG_IPV6_MROUTE
6704         {
6705                 .procname       = "mc_forwarding",
6706                 .data           = &ipv6_devconf.mc_forwarding,
6707                 .maxlen         = sizeof(int),
6708                 .mode           = 0444,
6709                 .proc_handler   = proc_dointvec,
6710         },
6711 #endif
6712         {
6713                 .procname       = "disable_ipv6",
6714                 .data           = &ipv6_devconf.disable_ipv6,
6715                 .maxlen         = sizeof(int),
6716                 .mode           = 0644,
6717                 .proc_handler   = addrconf_sysctl_disable,
6718         },
6719         {
6720                 .procname       = "accept_dad",
6721                 .data           = &ipv6_devconf.accept_dad,
6722                 .maxlen         = sizeof(int),
6723                 .mode           = 0644,
6724                 .proc_handler   = proc_dointvec,
6725         },
6726         {
6727                 .procname       = "force_tllao",
6728                 .data           = &ipv6_devconf.force_tllao,
6729                 .maxlen         = sizeof(int),
6730                 .mode           = 0644,
6731                 .proc_handler   = proc_dointvec
6732         },
6733         {
6734                 .procname       = "ndisc_notify",
6735                 .data           = &ipv6_devconf.ndisc_notify,
6736                 .maxlen         = sizeof(int),
6737                 .mode           = 0644,
6738                 .proc_handler   = proc_dointvec
6739         },
6740         {
6741                 .procname       = "suppress_frag_ndisc",
6742                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6743                 .maxlen         = sizeof(int),
6744                 .mode           = 0644,
6745                 .proc_handler   = proc_dointvec
6746         },
6747         {
6748                 .procname       = "accept_ra_from_local",
6749                 .data           = &ipv6_devconf.accept_ra_from_local,
6750                 .maxlen         = sizeof(int),
6751                 .mode           = 0644,
6752                 .proc_handler   = proc_dointvec,
6753         },
6754         {
6755                 .procname       = "accept_ra_mtu",
6756                 .data           = &ipv6_devconf.accept_ra_mtu,
6757                 .maxlen         = sizeof(int),
6758                 .mode           = 0644,
6759                 .proc_handler   = proc_dointvec,
6760         },
6761         {
6762                 .procname       = "stable_secret",
6763                 .data           = &ipv6_devconf.stable_secret,
6764                 .maxlen         = IPV6_MAX_STRLEN,
6765                 .mode           = 0600,
6766                 .proc_handler   = addrconf_sysctl_stable_secret,
6767         },
6768         {
6769                 .procname       = "use_oif_addrs_only",
6770                 .data           = &ipv6_devconf.use_oif_addrs_only,
6771                 .maxlen         = sizeof(int),
6772                 .mode           = 0644,
6773                 .proc_handler   = proc_dointvec,
6774         },
6775         {
6776                 .procname       = "ignore_routes_with_linkdown",
6777                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6778                 .maxlen         = sizeof(int),
6779                 .mode           = 0644,
6780                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6781         },
6782         {
6783                 .procname       = "drop_unicast_in_l2_multicast",
6784                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6785                 .maxlen         = sizeof(int),
6786                 .mode           = 0644,
6787                 .proc_handler   = proc_dointvec,
6788         },
6789         {
6790                 .procname       = "drop_unsolicited_na",
6791                 .data           = &ipv6_devconf.drop_unsolicited_na,
6792                 .maxlen         = sizeof(int),
6793                 .mode           = 0644,
6794                 .proc_handler   = proc_dointvec,
6795         },
6796         {
6797                 .procname       = "keep_addr_on_down",
6798                 .data           = &ipv6_devconf.keep_addr_on_down,
6799                 .maxlen         = sizeof(int),
6800                 .mode           = 0644,
6801                 .proc_handler   = proc_dointvec,
6802
6803         },
6804         {
6805                 .procname       = "seg6_enabled",
6806                 .data           = &ipv6_devconf.seg6_enabled,
6807                 .maxlen         = sizeof(int),
6808                 .mode           = 0644,
6809                 .proc_handler   = proc_dointvec,
6810         },
6811 #ifdef CONFIG_IPV6_SEG6_HMAC
6812         {
6813                 .procname       = "seg6_require_hmac",
6814                 .data           = &ipv6_devconf.seg6_require_hmac,
6815                 .maxlen         = sizeof(int),
6816                 .mode           = 0644,
6817                 .proc_handler   = proc_dointvec,
6818         },
6819 #endif
6820         {
6821                 .procname       = "enhanced_dad",
6822                 .data           = &ipv6_devconf.enhanced_dad,
6823                 .maxlen         = sizeof(int),
6824                 .mode           = 0644,
6825                 .proc_handler   = proc_dointvec,
6826         },
6827         {
6828                 .procname               = "addr_gen_mode",
6829                 .data                   = &ipv6_devconf.addr_gen_mode,
6830                 .maxlen                 = sizeof(int),
6831                 .mode                   = 0644,
6832                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6833         },
6834         {
6835                 .procname       = "disable_policy",
6836                 .data           = &ipv6_devconf.disable_policy,
6837                 .maxlen         = sizeof(int),
6838                 .mode           = 0644,
6839                 .proc_handler   = addrconf_sysctl_disable_policy,
6840         },
6841         {
6842                 .procname       = "ndisc_tclass",
6843                 .data           = &ipv6_devconf.ndisc_tclass,
6844                 .maxlen         = sizeof(int),
6845                 .mode           = 0644,
6846                 .proc_handler   = proc_dointvec_minmax,
6847                 .extra1         = (void *)SYSCTL_ZERO,
6848                 .extra2         = (void *)&two_five_five,
6849         },
6850         {
6851                 /* sentinel */
6852         }
6853 };
6854
6855 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6856                 struct inet6_dev *idev, struct ipv6_devconf *p)
6857 {
6858         int i, ifindex;
6859         struct ctl_table *table;
6860         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6861
6862         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6863         if (!table)
6864                 goto out;
6865
6866         for (i = 0; table[i].data; i++) {
6867                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6868                 /* If one of these is already set, then it is not safe to
6869                  * overwrite either of them: this makes proc_dointvec_minmax
6870                  * usable.
6871                  */
6872                 if (!table[i].extra1 && !table[i].extra2) {
6873                         table[i].extra1 = idev; /* embedded; no ref */
6874                         table[i].extra2 = net;
6875                 }
6876         }
6877
6878         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6879
6880         p->sysctl_header = register_net_sysctl(net, path, table);
6881         if (!p->sysctl_header)
6882                 goto free;
6883
6884         if (!strcmp(dev_name, "all"))
6885                 ifindex = NETCONFA_IFINDEX_ALL;
6886         else if (!strcmp(dev_name, "default"))
6887                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6888         else
6889                 ifindex = idev->dev->ifindex;
6890         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6891                                      ifindex, p);
6892         return 0;
6893
6894 free:
6895         kfree(table);
6896 out:
6897         return -ENOBUFS;
6898 }
6899
6900 static void __addrconf_sysctl_unregister(struct net *net,
6901                                          struct ipv6_devconf *p, int ifindex)
6902 {
6903         struct ctl_table *table;
6904
6905         if (!p->sysctl_header)
6906                 return;
6907
6908         table = p->sysctl_header->ctl_table_arg;
6909         unregister_net_sysctl_table(p->sysctl_header);
6910         p->sysctl_header = NULL;
6911         kfree(table);
6912
6913         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6914 }
6915
6916 static int addrconf_sysctl_register(struct inet6_dev *idev)
6917 {
6918         int err;
6919
6920         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6921                 return -EINVAL;
6922
6923         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6924                                     &ndisc_ifinfo_sysctl_change);
6925         if (err)
6926                 return err;
6927         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6928                                          idev, &idev->cnf);
6929         if (err)
6930                 neigh_sysctl_unregister(idev->nd_parms);
6931
6932         return err;
6933 }
6934
6935 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6936 {
6937         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6938                                      idev->dev->ifindex);
6939         neigh_sysctl_unregister(idev->nd_parms);
6940 }
6941
6942
6943 #endif
6944
6945 static int __net_init addrconf_init_net(struct net *net)
6946 {
6947         int err = -ENOMEM;
6948         struct ipv6_devconf *all, *dflt;
6949
6950         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6951         if (!all)
6952                 goto err_alloc_all;
6953
6954         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6955         if (!dflt)
6956                 goto err_alloc_dflt;
6957
6958         if (IS_ENABLED(CONFIG_SYSCTL) &&
6959             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6960                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6961                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6962         }
6963
6964         /* these will be inherited by all namespaces */
6965         dflt->autoconf = ipv6_defaults.autoconf;
6966         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6967
6968         dflt->stable_secret.initialized = false;
6969         all->stable_secret.initialized = false;
6970
6971         net->ipv6.devconf_all = all;
6972         net->ipv6.devconf_dflt = dflt;
6973
6974 #ifdef CONFIG_SYSCTL
6975         err = __addrconf_sysctl_register(net, "all", NULL, all);
6976         if (err < 0)
6977                 goto err_reg_all;
6978
6979         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6980         if (err < 0)
6981                 goto err_reg_dflt;
6982 #endif
6983         return 0;
6984
6985 #ifdef CONFIG_SYSCTL
6986 err_reg_dflt:
6987         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6988 err_reg_all:
6989         kfree(dflt);
6990 #endif
6991 err_alloc_dflt:
6992         kfree(all);
6993 err_alloc_all:
6994         return err;
6995 }
6996
6997 static void __net_exit addrconf_exit_net(struct net *net)
6998 {
6999 #ifdef CONFIG_SYSCTL
7000         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7001                                      NETCONFA_IFINDEX_DEFAULT);
7002         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7003                                      NETCONFA_IFINDEX_ALL);
7004 #endif
7005         kfree(net->ipv6.devconf_dflt);
7006         kfree(net->ipv6.devconf_all);
7007 }
7008
7009 static struct pernet_operations addrconf_ops = {
7010         .init = addrconf_init_net,
7011         .exit = addrconf_exit_net,
7012 };
7013
7014 static struct rtnl_af_ops inet6_ops __read_mostly = {
7015         .family           = AF_INET6,
7016         .fill_link_af     = inet6_fill_link_af,
7017         .get_link_af_size = inet6_get_link_af_size,
7018         .validate_link_af = inet6_validate_link_af,
7019         .set_link_af      = inet6_set_link_af,
7020 };
7021
7022 /*
7023  *      Init / cleanup code
7024  */
7025
7026 int __init addrconf_init(void)
7027 {
7028         struct inet6_dev *idev;
7029         int i, err;
7030
7031         err = ipv6_addr_label_init();
7032         if (err < 0) {
7033                 pr_crit("%s: cannot initialize default policy table: %d\n",
7034                         __func__, err);
7035                 goto out;
7036         }
7037
7038         err = register_pernet_subsys(&addrconf_ops);
7039         if (err < 0)
7040                 goto out_addrlabel;
7041
7042         addrconf_wq = create_workqueue("ipv6_addrconf");
7043         if (!addrconf_wq) {
7044                 err = -ENOMEM;
7045                 goto out_nowq;
7046         }
7047
7048         /* The addrconf netdev notifier requires that loopback_dev
7049          * has it's ipv6 private information allocated and setup
7050          * before it can bring up and give link-local addresses
7051          * to other devices which are up.
7052          *
7053          * Unfortunately, loopback_dev is not necessarily the first
7054          * entry in the global dev_base list of net devices.  In fact,
7055          * it is likely to be the very last entry on that list.
7056          * So this causes the notifier registry below to try and
7057          * give link-local addresses to all devices besides loopback_dev
7058          * first, then loopback_dev, which cases all the non-loopback_dev
7059          * devices to fail to get a link-local address.
7060          *
7061          * So, as a temporary fix, allocate the ipv6 structure for
7062          * loopback_dev first by hand.
7063          * Longer term, all of the dependencies ipv6 has upon the loopback
7064          * device and it being up should be removed.
7065          */
7066         rtnl_lock();
7067         idev = ipv6_add_dev(init_net.loopback_dev);
7068         rtnl_unlock();
7069         if (IS_ERR(idev)) {
7070                 err = PTR_ERR(idev);
7071                 goto errlo;
7072         }
7073
7074         ip6_route_init_special_entries();
7075
7076         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7077                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7078
7079         register_netdevice_notifier(&ipv6_dev_notf);
7080
7081         addrconf_verify();
7082
7083         rtnl_af_register(&inet6_ops);
7084
7085         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7086                                    NULL, inet6_dump_ifinfo, 0);
7087         if (err < 0)
7088                 goto errout;
7089
7090         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7091                                    inet6_rtm_newaddr, NULL, 0);
7092         if (err < 0)
7093                 goto errout;
7094         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7095                                    inet6_rtm_deladdr, NULL, 0);
7096         if (err < 0)
7097                 goto errout;
7098         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7099                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7100                                    RTNL_FLAG_DOIT_UNLOCKED);
7101         if (err < 0)
7102                 goto errout;
7103         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7104                                    NULL, inet6_dump_ifmcaddr, 0);
7105         if (err < 0)
7106                 goto errout;
7107         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7108                                    NULL, inet6_dump_ifacaddr, 0);
7109         if (err < 0)
7110                 goto errout;
7111         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7112                                    inet6_netconf_get_devconf,
7113                                    inet6_netconf_dump_devconf,
7114                                    RTNL_FLAG_DOIT_UNLOCKED);
7115         if (err < 0)
7116                 goto errout;
7117         err = ipv6_addr_label_rtnl_register();
7118         if (err < 0)
7119                 goto errout;
7120
7121         return 0;
7122 errout:
7123         rtnl_unregister_all(PF_INET6);
7124         rtnl_af_unregister(&inet6_ops);
7125         unregister_netdevice_notifier(&ipv6_dev_notf);
7126 errlo:
7127         destroy_workqueue(addrconf_wq);
7128 out_nowq:
7129         unregister_pernet_subsys(&addrconf_ops);
7130 out_addrlabel:
7131         ipv6_addr_label_cleanup();
7132 out:
7133         return err;
7134 }
7135
7136 void addrconf_cleanup(void)
7137 {
7138         struct net_device *dev;
7139         int i;
7140
7141         unregister_netdevice_notifier(&ipv6_dev_notf);
7142         unregister_pernet_subsys(&addrconf_ops);
7143         ipv6_addr_label_cleanup();
7144
7145         rtnl_af_unregister(&inet6_ops);
7146
7147         rtnl_lock();
7148
7149         /* clean dev list */
7150         for_each_netdev(&init_net, dev) {
7151                 if (__in6_dev_get(dev) == NULL)
7152                         continue;
7153                 addrconf_ifdown(dev, 1);
7154         }
7155         addrconf_ifdown(init_net.loopback_dev, 2);
7156
7157         /*
7158          *      Check hash table.
7159          */
7160         spin_lock_bh(&addrconf_hash_lock);
7161         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7162                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7163         spin_unlock_bh(&addrconf_hash_lock);
7164         cancel_delayed_work(&addr_chk_work);
7165         rtnl_unlock();
7166
7167         destroy_workqueue(addrconf_wq);
7168 }