net/ipv6/addrconf.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 Address [auto]configuration
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  */
  10
  11 /*
  12  *      Changes:
  13  *
  14  *      Janos Farkas                    :       delete timer on ifdown
  15  *      <chexum@bankinf.banki.hu>
  16  *      Andi Kleen                      :       kill double kfree on module
  17  *                                              unload.
  18  *      Maciej W. Rozycki               :       FDDI support
  19  *      sekiya@USAGI                    :       Don't send too many RS
  20  *                                              packets.
  21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  22  *                                              packets.
  23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  24  *                                              address validation timer.
  25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  26  *                                              support.
  27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  28  *                                              address on a same interface.
  29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  31  *                                              seq_file.
  32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  33  *                                              selection; consider scope,
  34  *                                              status etc.
  35  */
  36
  37 #define pr_fmt(fmt) "IPv6: " fmt
  38
  39 #include <linux/errno.h>
  40 #include <linux/types.h>
  41 #include <linux/kernel.h>
  42 #include <linux/sched/signal.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/inet.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64 #include <linux/hash.h>
  65
  66 #include <net/net_namespace.h>
  67 #include <net/sock.h>
  68 #include <net/snmp.h>
  69
  70 #include <net/6lowpan.h>
  71 #include <net/firewire.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <net/l3mdev.h>
  82 #include <linux/if_tunnel.h>
  83 #include <linux/rtnetlink.h>
  84 #include <linux/netconf.h>
  85 #include <linux/random.h>
  86 #include <linux/uaccess.h>
  87 #include <asm/unaligned.h>
  88
  89 #include <linux/proc_fs.h>
  90 #include <linux/seq_file.h>
  91 #include <linux/export.h>
  92
  93 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  94
  95 #define IPV6_MAX_STRLEN \
  96         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
  97
  98 static inline u32 cstamp_delta(unsigned long cstamp)
  99 {
 100         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 101 }
 102
 103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 104 {
 105         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 106         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 107         do_div(tmp, 1000000);
 108         return (s32)tmp;
 109 }
 110
 111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 112 {
 113         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 114         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 115         do_div(tmp, 1000000);
 116         if ((s32)tmp > mrt) {
 117                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 118                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 119                 do_div(tmp, 1000000);
 120         }
 121         return (s32)tmp;
 122 }
 123
 124 #ifdef CONFIG_SYSCTL
 125 static int addrconf_sysctl_register(struct inet6_dev *idev);
 126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 127 #else
 128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 129 {
 130         return 0;
 131 }
 132
 133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 134 {
 135 }
 136 #endif
 137
 138 static void ipv6_regen_rndid(struct inet6_dev *idev);
 139 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 140
 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 142 static int ipv6_count_addresses(const struct inet6_dev *idev);
 143 static int ipv6_generate_stable_address(struct in6_addr *addr,
 144                                         u8 dad_count,
 145                                         const struct inet6_dev *idev);
 146
 147 #define IN6_ADDR_HSIZE_SHIFT    8
 148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 149 /*
 150  *      Configured unicast address hash table
 151  */
 152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 153 static DEFINE_SPINLOCK(addrconf_hash_lock);
 154
 155 static void addrconf_verify(void);
 156 static void addrconf_verify_rtnl(void);
 157 static void addrconf_verify_work(struct work_struct *);
 158
 159 static struct workqueue_struct *addrconf_wq;
 160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 161
 162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 164
 165 static void addrconf_type_change(struct net_device *dev,
 166                                  unsigned long event);
 167 static int addrconf_ifdown(struct net_device *dev, int how);
 168
 169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 170                                                   int plen,
 171                                                   const struct net_device *dev,
 172                                                   u32 flags, u32 noflags,
 173                                                   bool no_gw);
 174
 175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 176 static void addrconf_dad_work(struct work_struct *w);
 177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 178                                    bool send_na);
 179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 180 static void addrconf_rs_timer(struct timer_list *t);
 181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 183
 184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 185                                 struct prefix_info *pinfo);
 186
 187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 188         .forwarding             = 0,
 189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 190         .mtu6                   = IPV6_MIN_MTU,
 191         .accept_ra              = 1,
 192         .accept_redirects       = 1,
 193         .autoconf               = 1,
 194         .force_mld_version      = 0,
 195         .mldv1_unsolicited_report_interval = 10 * HZ,
 196         .mldv2_unsolicited_report_interval = HZ,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .accept_ra_from_local   = 0,
 210         .accept_ra_min_hop_limit= 1,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_min_plen = 0,
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 0,
 224         .suppress_frag_ndisc    = 1,
 225         .accept_ra_mtu          = 1,
 226         .stable_secret          = {
 227                 .initialized = false,
 228         },
 229         .use_oif_addrs_only     = 0,
 230         .ignore_routes_with_linkdown = 0,
 231         .keep_addr_on_down      = 0,
 232         .seg6_enabled           = 0,
 233 #ifdef CONFIG_IPV6_SEG6_HMAC
 234         .seg6_require_hmac      = 0,
 235 #endif
 236         .enhanced_dad           = 1,
 237         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 238         .disable_policy         = 0,
 239 };
 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, bool del_rt)
1230 {
1231         struct fib6_info *f6i;
1232
1233         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
1234                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1235         if (f6i) {
1236                 if (del_rt)
1237                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1238                 else {
1239                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1240                                 fib6_set_expires(f6i, expires);
1241                         fib6_info_release(f6i);
1242                 }
1243         }
1244 }
1245
1246
1247 /* This function wants to get referenced ifp and releases it before return */
1248
1249 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1250 {
1251         int state;
1252         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1253         unsigned long expires;
1254
1255         ASSERT_RTNL();
1256
1257         spin_lock_bh(&ifp->lock);
1258         state = ifp->state;
1259         ifp->state = INET6_IFADDR_STATE_DEAD;
1260         spin_unlock_bh(&ifp->lock);
1261
1262         if (state == INET6_IFADDR_STATE_DEAD)
1263                 goto out;
1264
1265         spin_lock_bh(&addrconf_hash_lock);
1266         hlist_del_init_rcu(&ifp->addr_lst);
1267         spin_unlock_bh(&addrconf_hash_lock);
1268
1269         write_lock_bh(&ifp->idev->lock);
1270
1271         if (ifp->flags&IFA_F_TEMPORARY) {
1272                 list_del(&ifp->tmp_list);
1273                 if (ifp->ifpub) {
1274                         in6_ifa_put(ifp->ifpub);
1275                         ifp->ifpub = NULL;
1276                 }
1277                 __in6_ifa_put(ifp);
1278         }
1279
1280         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1281                 action = check_cleanup_prefix_route(ifp, &expires);
1282
1283         list_del_rcu(&ifp->if_list);
1284         __in6_ifa_put(ifp);
1285
1286         write_unlock_bh(&ifp->idev->lock);
1287
1288         addrconf_del_dad_work(ifp);
1289
1290         ipv6_ifa_notify(RTM_DELADDR, ifp);
1291
1292         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1293
1294         if (action != CLEANUP_PREFIX_RT_NOP) {
1295                 cleanup_prefix_route(ifp, expires,
1296                         action == CLEANUP_PREFIX_RT_DEL);
1297         }
1298
1299         /* clean up prefsrc entries */
1300         rt6_remove_prefsrc(ifp);
1301 out:
1302         in6_ifa_put(ifp);
1303 }
1304
1305 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1306                                 struct inet6_ifaddr *ift,
1307                                 bool block)
1308 {
1309         struct inet6_dev *idev = ifp->idev;
1310         struct in6_addr addr, *tmpaddr;
1311         unsigned long tmp_tstamp, age;
1312         unsigned long regen_advance;
1313         struct ifa6_config cfg;
1314         int ret = 0;
1315         unsigned long now = jiffies;
1316         long max_desync_factor;
1317         s32 cnf_temp_preferred_lft;
1318
1319         write_lock_bh(&idev->lock);
1320         if (ift) {
1321                 spin_lock_bh(&ift->lock);
1322                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1323                 spin_unlock_bh(&ift->lock);
1324                 tmpaddr = &addr;
1325         } else {
1326                 tmpaddr = NULL;
1327         }
1328 retry:
1329         in6_dev_hold(idev);
1330         if (idev->cnf.use_tempaddr <= 0) {
1331                 write_unlock_bh(&idev->lock);
1332                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1333                 in6_dev_put(idev);
1334                 ret = -1;
1335                 goto out;
1336         }
1337         spin_lock_bh(&ifp->lock);
1338         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1339                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1340                 spin_unlock_bh(&ifp->lock);
1341                 write_unlock_bh(&idev->lock);
1342                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1343                         __func__);
1344                 in6_dev_put(idev);
1345                 ret = -1;
1346                 goto out;
1347         }
1348         in6_ifa_hold(ifp);
1349         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1350         ipv6_try_regen_rndid(idev, tmpaddr);
1351         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1352         age = (now - ifp->tstamp) / HZ;
1353
1354         regen_advance = idev->cnf.regen_max_retry *
1355                         idev->cnf.dad_transmits *
1356                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1357
1358         /* recalculate max_desync_factor each time and update
1359          * idev->desync_factor if it's larger
1360          */
1361         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1362         max_desync_factor = min_t(__u32,
1363                                   idev->cnf.max_desync_factor,
1364                                   cnf_temp_preferred_lft - regen_advance);
1365
1366         if (unlikely(idev->desync_factor > max_desync_factor)) {
1367                 if (max_desync_factor > 0) {
1368                         get_random_bytes(&idev->desync_factor,
1369                                          sizeof(idev->desync_factor));
1370                         idev->desync_factor %= max_desync_factor;
1371                 } else {
1372                         idev->desync_factor = 0;
1373                 }
1374         }
1375
1376         memset(&cfg, 0, sizeof(cfg));
1377         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1378                               idev->cnf.temp_valid_lft + age);
1379         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1380         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1381
1382         cfg.plen = ifp->prefix_len;
1383         tmp_tstamp = ifp->tstamp;
1384         spin_unlock_bh(&ifp->lock);
1385
1386         write_unlock_bh(&idev->lock);
1387
1388         /* A temporary address is created only if this calculated Preferred
1389          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1390          * an implementation must not create a temporary address with a zero
1391          * Preferred Lifetime.
1392          * Use age calculation as in addrconf_verify to avoid unnecessary
1393          * temporary addresses being generated.
1394          */
1395         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1396         if (cfg.preferred_lft <= regen_advance + age) {
1397                 in6_ifa_put(ifp);
1398                 in6_dev_put(idev);
1399                 ret = -1;
1400                 goto out;
1401         }
1402
1403         cfg.ifa_flags = IFA_F_TEMPORARY;
1404         /* set in addrconf_prefix_rcv() */
1405         if (ifp->flags & IFA_F_OPTIMISTIC)
1406                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1407
1408         cfg.pfx = &addr;
1409         cfg.scope = ipv6_addr_scope(cfg.pfx);
1410
1411         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1412         if (IS_ERR(ift)) {
1413                 in6_ifa_put(ifp);
1414                 in6_dev_put(idev);
1415                 pr_info("%s: retry temporary address regeneration\n", __func__);
1416                 tmpaddr = &addr;
1417                 write_lock_bh(&idev->lock);
1418                 goto retry;
1419         }
1420
1421         spin_lock_bh(&ift->lock);
1422         ift->ifpub = ifp;
1423         ift->cstamp = now;
1424         ift->tstamp = tmp_tstamp;
1425         spin_unlock_bh(&ift->lock);
1426
1427         addrconf_dad_start(ift);
1428         in6_ifa_put(ift);
1429         in6_dev_put(idev);
1430 out:
1431         return ret;
1432 }
1433
1434 /*
1435  *      Choose an appropriate source address (RFC3484)
1436  */
1437 enum {
1438         IPV6_SADDR_RULE_INIT = 0,
1439         IPV6_SADDR_RULE_LOCAL,
1440         IPV6_SADDR_RULE_SCOPE,
1441         IPV6_SADDR_RULE_PREFERRED,
1442 #ifdef CONFIG_IPV6_MIP6
1443         IPV6_SADDR_RULE_HOA,
1444 #endif
1445         IPV6_SADDR_RULE_OIF,
1446         IPV6_SADDR_RULE_LABEL,
1447         IPV6_SADDR_RULE_PRIVACY,
1448         IPV6_SADDR_RULE_ORCHID,
1449         IPV6_SADDR_RULE_PREFIX,
1450 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1451         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1452 #endif
1453         IPV6_SADDR_RULE_MAX
1454 };
1455
1456 struct ipv6_saddr_score {
1457         int                     rule;
1458         int                     addr_type;
1459         struct inet6_ifaddr     *ifa;
1460         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1461         int                     scopedist;
1462         int                     matchlen;
1463 };
1464
1465 struct ipv6_saddr_dst {
1466         const struct in6_addr *addr;
1467         int ifindex;
1468         int scope;
1469         int label;
1470         unsigned int prefs;
1471 };
1472
1473 static inline int ipv6_saddr_preferred(int type)
1474 {
1475         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1476                 return 1;
1477         return 0;
1478 }
1479
1480 static bool ipv6_use_optimistic_addr(struct net *net,
1481                                      struct inet6_dev *idev)
1482 {
1483 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1484         if (!idev)
1485                 return false;
1486         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1487                 return false;
1488         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1489                 return false;
1490
1491         return true;
1492 #else
1493         return false;
1494 #endif
1495 }
1496
1497 static bool ipv6_allow_optimistic_dad(struct net *net,
1498                                       struct inet6_dev *idev)
1499 {
1500 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1501         if (!idev)
1502                 return false;
1503         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1504                 return false;
1505
1506         return true;
1507 #else
1508         return false;
1509 #endif
1510 }
1511
1512 static int ipv6_get_saddr_eval(struct net *net,
1513                                struct ipv6_saddr_score *score,
1514                                struct ipv6_saddr_dst *dst,
1515                                int i)
1516 {
1517         int ret;
1518
1519         if (i <= score->rule) {
1520                 switch (i) {
1521                 case IPV6_SADDR_RULE_SCOPE:
1522                         ret = score->scopedist;
1523                         break;
1524                 case IPV6_SADDR_RULE_PREFIX:
1525                         ret = score->matchlen;
1526                         break;
1527                 default:
1528                         ret = !!test_bit(i, score->scorebits);
1529                 }
1530                 goto out;
1531         }
1532
1533         switch (i) {
1534         case IPV6_SADDR_RULE_INIT:
1535                 /* Rule 0: remember if hiscore is not ready yet */
1536                 ret = !!score->ifa;
1537                 break;
1538         case IPV6_SADDR_RULE_LOCAL:
1539                 /* Rule 1: Prefer same address */
1540                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1541                 break;
1542         case IPV6_SADDR_RULE_SCOPE:
1543                 /* Rule 2: Prefer appropriate scope
1544                  *
1545                  *      ret
1546                  *       ^
1547                  *    -1 |  d 15
1548                  *    ---+--+-+---> scope
1549                  *       |
1550                  *       |             d is scope of the destination.
1551                  *  B-d  |  \
1552                  *       |   \      <- smaller scope is better if
1553                  *  B-15 |    \        if scope is enough for destination.
1554                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1555                  * d-C-1 | /
1556                  *       |/         <- greater is better
1557                  *   -C  /             if scope is not enough for destination.
1558                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1559                  *
1560                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1561                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1562                  * Assume B = 0 and we get C > 29.
1563                  */
1564                 ret = __ipv6_addr_src_scope(score->addr_type);
1565                 if (ret >= dst->scope)
1566                         ret = -ret;
1567                 else
1568                         ret -= 128;     /* 30 is enough */
1569                 score->scopedist = ret;
1570                 break;
1571         case IPV6_SADDR_RULE_PREFERRED:
1572             {
1573                 /* Rule 3: Avoid deprecated and optimistic addresses */
1574                 u8 avoid = IFA_F_DEPRECATED;
1575
1576                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1577                         avoid |= IFA_F_OPTIMISTIC;
1578                 ret = ipv6_saddr_preferred(score->addr_type) ||
1579                       !(score->ifa->flags & avoid);
1580                 break;
1581             }
1582 #ifdef CONFIG_IPV6_MIP6
1583         case IPV6_SADDR_RULE_HOA:
1584             {
1585                 /* Rule 4: Prefer home address */
1586                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1587                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1588                 break;
1589             }
1590 #endif
1591         case IPV6_SADDR_RULE_OIF:
1592                 /* Rule 5: Prefer outgoing interface */
1593                 ret = (!dst->ifindex ||
1594                        dst->ifindex == score->ifa->idev->dev->ifindex);
1595                 break;
1596         case IPV6_SADDR_RULE_LABEL:
1597                 /* Rule 6: Prefer matching label */
1598                 ret = ipv6_addr_label(net,
1599                                       &score->ifa->addr, score->addr_type,
1600                                       score->ifa->idev->dev->ifindex) == dst->label;
1601                 break;
1602         case IPV6_SADDR_RULE_PRIVACY:
1603             {
1604                 /* Rule 7: Prefer public address
1605                  * Note: prefer temporary address if use_tempaddr >= 2
1606                  */
1607                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1608                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1609                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1610                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1611                 break;
1612             }
1613         case IPV6_SADDR_RULE_ORCHID:
1614                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1615                  *          non-ORCHID vs non-ORCHID
1616                  */
1617                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1618                         ipv6_addr_orchid(dst->addr));
1619                 break;
1620         case IPV6_SADDR_RULE_PREFIX:
1621                 /* Rule 8: Use longest matching prefix */
1622                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1623                 if (ret > score->ifa->prefix_len)
1624                         ret = score->ifa->prefix_len;
1625                 score->matchlen = ret;
1626                 break;
1627 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1628         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1629                 /* Optimistic addresses still have lower precedence than other
1630                  * preferred addresses.
1631                  */
1632                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1633                 break;
1634 #endif
1635         default:
1636                 ret = 0;
1637         }
1638
1639         if (ret)
1640                 __set_bit(i, score->scorebits);
1641         score->rule = i;
1642 out:
1643         return ret;
1644 }
1645
1646 static int __ipv6_dev_get_saddr(struct net *net,
1647                                 struct ipv6_saddr_dst *dst,
1648                                 struct inet6_dev *idev,
1649                                 struct ipv6_saddr_score *scores,
1650                                 int hiscore_idx)
1651 {
1652         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1653
1654         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1655                 int i;
1656
1657                 /*
1658                  * - Tentative Address (RFC2462 section 5.4)
1659                  *  - A tentative address is not considered
1660                  *    "assigned to an interface" in the traditional
1661                  *    sense, unless it is also flagged as optimistic.
1662                  * - Candidate Source Address (section 4)
1663                  *  - In any case, anycast addresses, multicast
1664                  *    addresses, and the unspecified address MUST
1665                  *    NOT be included in a candidate set.
1666                  */
1667                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1668                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1669                         continue;
1670
1671                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1672
1673                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1674                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1675                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1676                                             idev->dev->name);
1677                         continue;
1678                 }
1679
1680                 score->rule = -1;
1681                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1682
1683                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1684                         int minihiscore, miniscore;
1685
1686                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1687                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1688
1689                         if (minihiscore > miniscore) {
1690                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1691                                     score->scopedist > 0) {
1692                                         /*
1693                                          * special case:
1694                                          * each remaining entry
1695                                          * has too small (not enough)
1696                                          * scope, because ifa entries
1697                                          * are sorted by their scope
1698                                          * values.
1699                                          */
1700                                         goto out;
1701                                 }
1702                                 break;
1703                         } else if (minihiscore < miniscore) {
1704                                 swap(hiscore, score);
1705                                 hiscore_idx = 1 - hiscore_idx;
1706
1707                                 /* restore our iterator */
1708                                 score->ifa = hiscore->ifa;
1709
1710                                 break;
1711                         }
1712                 }
1713         }
1714 out:
1715         return hiscore_idx;
1716 }
1717
1718 static int ipv6_get_saddr_master(struct net *net,
1719                                  const struct net_device *dst_dev,
1720                                  const struct net_device *master,
1721                                  struct ipv6_saddr_dst *dst,
1722                                  struct ipv6_saddr_score *scores,
1723                                  int hiscore_idx)
1724 {
1725         struct inet6_dev *idev;
1726
1727         idev = __in6_dev_get(dst_dev);
1728         if (idev)
1729                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1730                                                    scores, hiscore_idx);
1731
1732         idev = __in6_dev_get(master);
1733         if (idev)
1734                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1735                                                    scores, hiscore_idx);
1736
1737         return hiscore_idx;
1738 }
1739
1740 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1741                        const struct in6_addr *daddr, unsigned int prefs,
1742                        struct in6_addr *saddr)
1743 {
1744         struct ipv6_saddr_score scores[2], *hiscore;
1745         struct ipv6_saddr_dst dst;
1746         struct inet6_dev *idev;
1747         struct net_device *dev;
1748         int dst_type;
1749         bool use_oif_addr = false;
1750         int hiscore_idx = 0;
1751         int ret = 0;
1752
1753         dst_type = __ipv6_addr_type(daddr);
1754         dst.addr = daddr;
1755         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1756         dst.scope = __ipv6_addr_src_scope(dst_type);
1757         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1758         dst.prefs = prefs;
1759
1760         scores[hiscore_idx].rule = -1;
1761         scores[hiscore_idx].ifa = NULL;
1762
1763         rcu_read_lock();
1764
1765         /* Candidate Source Address (section 4)
1766          *  - multicast and link-local destination address,
1767          *    the set of candidate source address MUST only
1768          *    include addresses assigned to interfaces
1769          *    belonging to the same link as the outgoing
1770          *    interface.
1771          * (- For site-local destination addresses, the
1772          *    set of candidate source addresses MUST only
1773          *    include addresses assigned to interfaces
1774          *    belonging to the same site as the outgoing
1775          *    interface.)
1776          *  - "It is RECOMMENDED that the candidate source addresses
1777          *    be the set of unicast addresses assigned to the
1778          *    interface that will be used to send to the destination
1779          *    (the 'outgoing' interface)." (RFC 6724)
1780          */
1781         if (dst_dev) {
1782                 idev = __in6_dev_get(dst_dev);
1783                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1784                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1785                     (idev && idev->cnf.use_oif_addrs_only)) {
1786                         use_oif_addr = true;
1787                 }
1788         }
1789
1790         if (use_oif_addr) {
1791                 if (idev)
1792                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1793         } else {
1794                 const struct net_device *master;
1795                 int master_idx = 0;
1796
1797                 /* if dst_dev exists and is enslaved to an L3 device, then
1798                  * prefer addresses from dst_dev and then the master over
1799                  * any other enslaved devices in the L3 domain.
1800                  */
1801                 master = l3mdev_master_dev_rcu(dst_dev);
1802                 if (master) {
1803                         master_idx = master->ifindex;
1804
1805                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1806                                                             master, &dst,
1807                                                             scores, hiscore_idx);
1808
1809                         if (scores[hiscore_idx].ifa)
1810                                 goto out;
1811                 }
1812
1813                 for_each_netdev_rcu(net, dev) {
1814                         /* only consider addresses on devices in the
1815                          * same L3 domain
1816                          */
1817                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1818                                 continue;
1819                         idev = __in6_dev_get(dev);
1820                         if (!idev)
1821                                 continue;
1822                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1823                 }
1824         }
1825
1826 out:
1827         hiscore = &scores[hiscore_idx];
1828         if (!hiscore->ifa)
1829                 ret = -EADDRNOTAVAIL;
1830         else
1831                 *saddr = hiscore->ifa->addr;
1832
1833         rcu_read_unlock();
1834         return ret;
1835 }
1836 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1837
1838 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1839                       u32 banned_flags)
1840 {
1841         struct inet6_ifaddr *ifp;
1842         int err = -EADDRNOTAVAIL;
1843
1844         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1845                 if (ifp->scope > IFA_LINK)
1846                         break;
1847                 if (ifp->scope == IFA_LINK &&
1848                     !(ifp->flags & banned_flags)) {
1849                         *addr = ifp->addr;
1850                         err = 0;
1851                         break;
1852                 }
1853         }
1854         return err;
1855 }
1856
1857 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1858                     u32 banned_flags)
1859 {
1860         struct inet6_dev *idev;
1861         int err = -EADDRNOTAVAIL;
1862
1863         rcu_read_lock();
1864         idev = __in6_dev_get(dev);
1865         if (idev) {
1866                 read_lock_bh(&idev->lock);
1867                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1868                 read_unlock_bh(&idev->lock);
1869         }
1870         rcu_read_unlock();
1871         return err;
1872 }
1873
1874 static int ipv6_count_addresses(const struct inet6_dev *idev)
1875 {
1876         const struct inet6_ifaddr *ifp;
1877         int cnt = 0;
1878
1879         rcu_read_lock();
1880         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1881                 cnt++;
1882         rcu_read_unlock();
1883         return cnt;
1884 }
1885
1886 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1887                   const struct net_device *dev, int strict)
1888 {
1889         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1890                                        strict, IFA_F_TENTATIVE);
1891 }
1892 EXPORT_SYMBOL(ipv6_chk_addr);
1893
1894 /* device argument is used to find the L3 domain of interest. If
1895  * skip_dev_check is set, then the ifp device is not checked against
1896  * the passed in dev argument. So the 2 cases for addresses checks are:
1897  *   1. does the address exist in the L3 domain that dev is part of
1898  *      (skip_dev_check = true), or
1899  *
1900  *   2. does the address exist on the specific device
1901  *      (skip_dev_check = false)
1902  */
1903 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1904                             const struct net_device *dev, bool skip_dev_check,
1905                             int strict, u32 banned_flags)
1906 {
1907         unsigned int hash = inet6_addr_hash(net, addr);
1908         const struct net_device *l3mdev;
1909         struct inet6_ifaddr *ifp;
1910         u32 ifp_flags;
1911
1912         rcu_read_lock();
1913
1914         l3mdev = l3mdev_master_dev_rcu(dev);
1915         if (skip_dev_check)
1916                 dev = NULL;
1917
1918         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1919                 if (!net_eq(dev_net(ifp->idev->dev), net))
1920                         continue;
1921
1922                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1923                         continue;
1924
1925                 /* Decouple optimistic from tentative for evaluation here.
1926                  * Ban optimistic addresses explicitly, when required.
1927                  */
1928                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1929                             ? (ifp->flags&~IFA_F_TENTATIVE)
1930                             : ifp->flags;
1931                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1932                     !(ifp_flags&banned_flags) &&
1933                     (!dev || ifp->idev->dev == dev ||
1934                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1935                         rcu_read_unlock();
1936                         return 1;
1937                 }
1938         }
1939
1940         rcu_read_unlock();
1941         return 0;
1942 }
1943 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1944
1945
1946 /* Compares an address/prefix_len with addresses on device @dev.
1947  * If one is found it returns true.
1948  */
1949 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1950         const unsigned int prefix_len, struct net_device *dev)
1951 {
1952         const struct inet6_ifaddr *ifa;
1953         const struct inet6_dev *idev;
1954         bool ret = false;
1955
1956         rcu_read_lock();
1957         idev = __in6_dev_get(dev);
1958         if (idev) {
1959                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1960                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1961                         if (ret)
1962                                 break;
1963                 }
1964         }
1965         rcu_read_unlock();
1966
1967         return ret;
1968 }
1969 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1970
1971 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1972 {
1973         const struct inet6_ifaddr *ifa;
1974         const struct inet6_dev *idev;
1975         int     onlink;
1976
1977         onlink = 0;
1978         rcu_read_lock();
1979         idev = __in6_dev_get(dev);
1980         if (idev) {
1981                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1982                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1983                                                    ifa->prefix_len);
1984                         if (onlink)
1985                                 break;
1986                 }
1987         }
1988         rcu_read_unlock();
1989         return onlink;
1990 }
1991 EXPORT_SYMBOL(ipv6_chk_prefix);
1992
1993 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1994                                      struct net_device *dev, int strict)
1995 {
1996         unsigned int hash = inet6_addr_hash(net, addr);
1997         struct inet6_ifaddr *ifp, *result = NULL;
1998
1999         rcu_read_lock();
2000         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2001                 if (!net_eq(dev_net(ifp->idev->dev), net))
2002                         continue;
2003                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2004                         if (!dev || ifp->idev->dev == dev ||
2005                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2006                                 result = ifp;
2007                                 in6_ifa_hold(ifp);
2008                                 break;
2009                         }
2010                 }
2011         }
2012         rcu_read_unlock();
2013
2014         return result;
2015 }
2016
2017 /* Gets referenced address, destroys ifaddr */
2018
2019 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2020 {
2021         if (dad_failed)
2022                 ifp->flags |= IFA_F_DADFAILED;
2023
2024         if (ifp->flags&IFA_F_TEMPORARY) {
2025                 struct inet6_ifaddr *ifpub;
2026                 spin_lock_bh(&ifp->lock);
2027                 ifpub = ifp->ifpub;
2028                 if (ifpub) {
2029                         in6_ifa_hold(ifpub);
2030                         spin_unlock_bh(&ifp->lock);
2031                         ipv6_create_tempaddr(ifpub, ifp, true);
2032                         in6_ifa_put(ifpub);
2033                 } else {
2034                         spin_unlock_bh(&ifp->lock);
2035                 }
2036                 ipv6_del_addr(ifp);
2037         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2038                 spin_lock_bh(&ifp->lock);
2039                 addrconf_del_dad_work(ifp);
2040                 ifp->flags |= IFA_F_TENTATIVE;
2041                 if (dad_failed)
2042                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2043                 spin_unlock_bh(&ifp->lock);
2044                 if (dad_failed)
2045                         ipv6_ifa_notify(0, ifp);
2046                 in6_ifa_put(ifp);
2047         } else {
2048                 ipv6_del_addr(ifp);
2049         }
2050 }
2051
2052 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2053 {
2054         int err = -ENOENT;
2055
2056         spin_lock_bh(&ifp->lock);
2057         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2058                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2059                 err = 0;
2060         }
2061         spin_unlock_bh(&ifp->lock);
2062
2063         return err;
2064 }
2065
2066 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2067 {
2068         struct inet6_dev *idev = ifp->idev;
2069         struct net *net = dev_net(ifp->idev->dev);
2070
2071         if (addrconf_dad_end(ifp)) {
2072                 in6_ifa_put(ifp);
2073                 return;
2074         }
2075
2076         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2077                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2078
2079         spin_lock_bh(&ifp->lock);
2080
2081         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2082                 struct in6_addr new_addr;
2083                 struct inet6_ifaddr *ifp2;
2084                 int retries = ifp->stable_privacy_retry + 1;
2085                 struct ifa6_config cfg = {
2086                         .pfx = &new_addr,
2087                         .plen = ifp->prefix_len,
2088                         .ifa_flags = ifp->flags,
2089                         .valid_lft = ifp->valid_lft,
2090                         .preferred_lft = ifp->prefered_lft,
2091                         .scope = ifp->scope,
2092                 };
2093
2094                 if (retries > net->ipv6.sysctl.idgen_retries) {
2095                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2096                                              ifp->idev->dev->name);
2097                         goto errdad;
2098                 }
2099
2100                 new_addr = ifp->addr;
2101                 if (ipv6_generate_stable_address(&new_addr, retries,
2102                                                  idev))
2103                         goto errdad;
2104
2105                 spin_unlock_bh(&ifp->lock);
2106
2107                 if (idev->cnf.max_addresses &&
2108                     ipv6_count_addresses(idev) >=
2109                     idev->cnf.max_addresses)
2110                         goto lock_errdad;
2111
2112                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2113                                      ifp->idev->dev->name);
2114
2115                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2116                 if (IS_ERR(ifp2))
2117                         goto lock_errdad;
2118
2119                 spin_lock_bh(&ifp2->lock);
2120                 ifp2->stable_privacy_retry = retries;
2121                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2122                 spin_unlock_bh(&ifp2->lock);
2123
2124                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2125                 in6_ifa_put(ifp2);
2126 lock_errdad:
2127                 spin_lock_bh(&ifp->lock);
2128         }
2129
2130 errdad:
2131         /* transition from _POSTDAD to _ERRDAD */
2132         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2133         spin_unlock_bh(&ifp->lock);
2134
2135         addrconf_mod_dad_work(ifp, 0);
2136         in6_ifa_put(ifp);
2137 }
2138
2139 /* Join to solicited addr multicast group.
2140  * caller must hold RTNL */
2141 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2142 {
2143         struct in6_addr maddr;
2144
2145         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2146                 return;
2147
2148         addrconf_addr_solict_mult(addr, &maddr);
2149         ipv6_dev_mc_inc(dev, &maddr);
2150 }
2151
2152 /* caller must hold RTNL */
2153 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2154 {
2155         struct in6_addr maddr;
2156
2157         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2158                 return;
2159
2160         addrconf_addr_solict_mult(addr, &maddr);
2161         __ipv6_dev_mc_dec(idev, &maddr);
2162 }
2163
2164 /* caller must hold RTNL */
2165 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2166 {
2167         struct in6_addr addr;
2168
2169         if (ifp->prefix_len >= 127) /* RFC 6164 */
2170                 return;
2171         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2172         if (ipv6_addr_any(&addr))
2173                 return;
2174         __ipv6_dev_ac_inc(ifp->idev, &addr);
2175 }
2176
2177 /* caller must hold RTNL */
2178 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2179 {
2180         struct in6_addr addr;
2181
2182         if (ifp->prefix_len >= 127) /* RFC 6164 */
2183                 return;
2184         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2185         if (ipv6_addr_any(&addr))
2186                 return;
2187         __ipv6_dev_ac_dec(ifp->idev, &addr);
2188 }
2189
2190 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2191 {
2192         switch (dev->addr_len) {
2193         case ETH_ALEN:
2194                 memcpy(eui, dev->dev_addr, 3);
2195                 eui[3] = 0xFF;
2196                 eui[4] = 0xFE;
2197                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2198                 break;
2199         case EUI64_ADDR_LEN:
2200                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2201                 eui[0] ^= 2;
2202                 break;
2203         default:
2204                 return -1;
2205         }
2206
2207         return 0;
2208 }
2209
2210 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2211 {
2212         union fwnet_hwaddr *ha;
2213
2214         if (dev->addr_len != FWNET_ALEN)
2215                 return -1;
2216
2217         ha = (union fwnet_hwaddr *)dev->dev_addr;
2218
2219         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2220         eui[0] ^= 2;
2221         return 0;
2222 }
2223
2224 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2225 {
2226         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2227         if (dev->addr_len != ARCNET_ALEN)
2228                 return -1;
2229         memset(eui, 0, 7);
2230         eui[7] = *(u8 *)dev->dev_addr;
2231         return 0;
2232 }
2233
2234 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2235 {
2236         if (dev->addr_len != INFINIBAND_ALEN)
2237                 return -1;
2238         memcpy(eui, dev->dev_addr + 12, 8);
2239         eui[0] |= 2;
2240         return 0;
2241 }
2242
2243 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2244 {
2245         if (addr == 0)
2246                 return -1;
2247         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2248                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2249                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2250                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2251                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2252                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2253         eui[1] = 0;
2254         eui[2] = 0x5E;
2255         eui[3] = 0xFE;
2256         memcpy(eui + 4, &addr, 4);
2257         return 0;
2258 }
2259
2260 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2261 {
2262         if (dev->priv_flags & IFF_ISATAP)
2263                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2264         return -1;
2265 }
2266
2267 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2268 {
2269         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2270 }
2271
2272 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2273 {
2274         memcpy(eui, dev->perm_addr, 3);
2275         memcpy(eui + 5, dev->perm_addr + 3, 3);
2276         eui[3] = 0xFF;
2277         eui[4] = 0xFE;
2278         eui[0] ^= 2;
2279         return 0;
2280 }
2281
2282 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2283 {
2284         switch (dev->type) {
2285         case ARPHRD_ETHER:
2286         case ARPHRD_FDDI:
2287                 return addrconf_ifid_eui48(eui, dev);
2288         case ARPHRD_ARCNET:
2289                 return addrconf_ifid_arcnet(eui, dev);
2290         case ARPHRD_INFINIBAND:
2291                 return addrconf_ifid_infiniband(eui, dev);
2292         case ARPHRD_SIT:
2293                 return addrconf_ifid_sit(eui, dev);
2294         case ARPHRD_IPGRE:
2295         case ARPHRD_TUNNEL:
2296                 return addrconf_ifid_gre(eui, dev);
2297         case ARPHRD_6LOWPAN:
2298                 return addrconf_ifid_6lowpan(eui, dev);
2299         case ARPHRD_IEEE1394:
2300                 return addrconf_ifid_ieee1394(eui, dev);
2301         case ARPHRD_TUNNEL6:
2302         case ARPHRD_IP6GRE:
2303         case ARPHRD_RAWIP:
2304                 return addrconf_ifid_ip6tnl(eui, dev);
2305         }
2306         return -1;
2307 }
2308
2309 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2310 {
2311         int err = -1;
2312         struct inet6_ifaddr *ifp;
2313
2314         read_lock_bh(&idev->lock);
2315         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2316                 if (ifp->scope > IFA_LINK)
2317                         break;
2318                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2319                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2320                         err = 0;
2321                         break;
2322                 }
2323         }
2324         read_unlock_bh(&idev->lock);
2325         return err;
2326 }
2327
2328 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2329 static void ipv6_regen_rndid(struct inet6_dev *idev)
2330 {
2331 regen:
2332         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2333         idev->rndid[0] &= ~0x02;
2334
2335         /*
2336          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2337          * check if generated address is not inappropriate
2338          *
2339          *  - Reserved subnet anycast (RFC 2526)
2340          *      11111101 11....11 1xxxxxxx
2341          *  - ISATAP (RFC4214) 6.1
2342          *      00-00-5E-FE-xx-xx-xx-xx
2343          *  - value 0
2344          *  - XXX: already assigned to an address on the device
2345          */
2346         if (idev->rndid[0] == 0xfd &&
2347             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2348             (idev->rndid[7]&0x80))
2349                 goto regen;
2350         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2351                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2352                         goto regen;
2353                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2354                         goto regen;
2355         }
2356 }
2357
2358 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2359 {
2360         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2361                 ipv6_regen_rndid(idev);
2362 }
2363
2364 /*
2365  *      Add prefix route.
2366  */
2367
2368 static void
2369 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2370                       struct net_device *dev, unsigned long expires,
2371                       u32 flags, gfp_t gfp_flags)
2372 {
2373         struct fib6_config cfg = {
2374                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2375                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2376                 .fc_ifindex = dev->ifindex,
2377                 .fc_expires = expires,
2378                 .fc_dst_len = plen,
2379                 .fc_flags = RTF_UP | flags,
2380                 .fc_nlinfo.nl_net = dev_net(dev),
2381                 .fc_protocol = RTPROT_KERNEL,
2382                 .fc_type = RTN_UNICAST,
2383         };
2384
2385         cfg.fc_dst = *pfx;
2386
2387         /* Prevent useless cloning on PtP SIT.
2388            This thing is done here expecting that the whole
2389            class of non-broadcast devices need not cloning.
2390          */
2391 #if IS_ENABLED(CONFIG_IPV6_SIT)
2392         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2393                 cfg.fc_flags |= RTF_NONEXTHOP;
2394 #endif
2395
2396         ip6_route_add(&cfg, gfp_flags, NULL);
2397 }
2398
2399
2400 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2401                                                   int plen,
2402                                                   const struct net_device *dev,
2403                                                   u32 flags, u32 noflags,
2404                                                   bool no_gw)
2405 {
2406         struct fib6_node *fn;
2407         struct fib6_info *rt = NULL;
2408         struct fib6_table *table;
2409         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2410
2411         table = fib6_get_table(dev_net(dev), tb_id);
2412         if (!table)
2413                 return NULL;
2414
2415         rcu_read_lock();
2416         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2417         if (!fn)
2418                 goto out;
2419
2420         for_each_fib6_node_rt_rcu(fn) {
2421                 /* prefix routes only use builtin fib6_nh */
2422                 if (rt->nh)
2423                         continue;
2424
2425                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2426                         continue;
2427                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2428                         continue;
2429                 if ((rt->fib6_flags & flags) != flags)
2430                         continue;
2431                 if ((rt->fib6_flags & noflags) != 0)
2432                         continue;
2433                 if (!fib6_info_hold_safe(rt))
2434                         continue;
2435                 break;
2436         }
2437 out:
2438         rcu_read_unlock();
2439         return rt;
2440 }
2441
2442
2443 /* Create "default" multicast route to the interface */
2444
2445 static void addrconf_add_mroute(struct net_device *dev)
2446 {
2447         struct fib6_config cfg = {
2448                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2449                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2450                 .fc_ifindex = dev->ifindex,
2451                 .fc_dst_len = 8,
2452                 .fc_flags = RTF_UP,
2453                 .fc_type = RTN_UNICAST,
2454                 .fc_nlinfo.nl_net = dev_net(dev),
2455         };
2456
2457         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2458
2459         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2460 }
2461
2462 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2463 {
2464         struct inet6_dev *idev;
2465
2466         ASSERT_RTNL();
2467
2468         idev = ipv6_find_idev(dev);
2469         if (IS_ERR(idev))
2470                 return idev;
2471
2472         if (idev->cnf.disable_ipv6)
2473                 return ERR_PTR(-EACCES);
2474
2475         /* Add default multicast route */
2476         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2477                 addrconf_add_mroute(dev);
2478
2479         return idev;
2480 }
2481
2482 static void manage_tempaddrs(struct inet6_dev *idev,
2483                              struct inet6_ifaddr *ifp,
2484                              __u32 valid_lft, __u32 prefered_lft,
2485                              bool create, unsigned long now)
2486 {
2487         u32 flags;
2488         struct inet6_ifaddr *ift;
2489
2490         read_lock_bh(&idev->lock);
2491         /* update all temporary addresses in the list */
2492         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2493                 int age, max_valid, max_prefered;
2494
2495                 if (ifp != ift->ifpub)
2496                         continue;
2497
2498                 /* RFC 4941 section 3.3:
2499                  * If a received option will extend the lifetime of a public
2500                  * address, the lifetimes of temporary addresses should
2501                  * be extended, subject to the overall constraint that no
2502                  * temporary addresses should ever remain "valid" or "preferred"
2503                  * for a time longer than (TEMP_VALID_LIFETIME) or
2504                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2505                  */
2506                 age = (now - ift->cstamp) / HZ;
2507                 max_valid = idev->cnf.temp_valid_lft - age;
2508                 if (max_valid < 0)
2509                         max_valid = 0;
2510
2511                 max_prefered = idev->cnf.temp_prefered_lft -
2512                                idev->desync_factor - age;
2513                 if (max_prefered < 0)
2514                         max_prefered = 0;
2515
2516                 if (valid_lft > max_valid)
2517                         valid_lft = max_valid;
2518
2519                 if (prefered_lft > max_prefered)
2520                         prefered_lft = max_prefered;
2521
2522                 spin_lock(&ift->lock);
2523                 flags = ift->flags;
2524                 ift->valid_lft = valid_lft;
2525                 ift->prefered_lft = prefered_lft;
2526                 ift->tstamp = now;
2527                 if (prefered_lft > 0)
2528                         ift->flags &= ~IFA_F_DEPRECATED;
2529
2530                 spin_unlock(&ift->lock);
2531                 if (!(flags&IFA_F_TENTATIVE))
2532                         ipv6_ifa_notify(0, ift);
2533         }
2534
2535         if ((create || list_empty(&idev->tempaddr_list)) &&
2536             idev->cnf.use_tempaddr > 0) {
2537                 /* When a new public address is created as described
2538                  * in [ADDRCONF], also create a new temporary address.
2539                  * Also create a temporary address if it's enabled but
2540                  * no temporary address currently exists.
2541                  */
2542                 read_unlock_bh(&idev->lock);
2543                 ipv6_create_tempaddr(ifp, NULL, false);
2544         } else {
2545                 read_unlock_bh(&idev->lock);
2546         }
2547 }
2548
2549 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2550 {
2551         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2552                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2553 }
2554
2555 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2556                                  const struct prefix_info *pinfo,
2557                                  struct inet6_dev *in6_dev,
2558                                  const struct in6_addr *addr, int addr_type,
2559                                  u32 addr_flags, bool sllao, bool tokenized,
2560                                  __u32 valid_lft, u32 prefered_lft)
2561 {
2562         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2563         int create = 0, update_lft = 0;
2564
2565         if (!ifp && valid_lft) {
2566                 int max_addresses = in6_dev->cnf.max_addresses;
2567                 struct ifa6_config cfg = {
2568                         .pfx = addr,
2569                         .plen = pinfo->prefix_len,
2570                         .ifa_flags = addr_flags,
2571                         .valid_lft = valid_lft,
2572                         .preferred_lft = prefered_lft,
2573                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2574                 };
2575
2576 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2577                 if ((net->ipv6.devconf_all->optimistic_dad ||
2578                      in6_dev->cnf.optimistic_dad) &&
2579                     !net->ipv6.devconf_all->forwarding && sllao)
2580                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2581 #endif
2582
2583                 /* Do not allow to create too much of autoconfigured
2584                  * addresses; this would be too easy way to crash kernel.
2585                  */
2586                 if (!max_addresses ||
2587                     ipv6_count_addresses(in6_dev) < max_addresses)
2588                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2589
2590                 if (IS_ERR_OR_NULL(ifp))
2591                         return -1;
2592
2593                 create = 1;
2594                 spin_lock_bh(&ifp->lock);
2595                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2596                 ifp->cstamp = jiffies;
2597                 ifp->tokenized = tokenized;
2598                 spin_unlock_bh(&ifp->lock);
2599                 addrconf_dad_start(ifp);
2600         }
2601
2602         if (ifp) {
2603                 u32 flags;
2604                 unsigned long now;
2605                 u32 stored_lft;
2606
2607                 /* update lifetime (RFC2462 5.5.3 e) */
2608                 spin_lock_bh(&ifp->lock);
2609                 now = jiffies;
2610                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2611                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2612                 else
2613                         stored_lft = 0;
2614                 if (!create && stored_lft) {
2615                         const u32 minimum_lft = min_t(u32,
2616                                 stored_lft, MIN_VALID_LIFETIME);
2617                         valid_lft = max(valid_lft, minimum_lft);
2618
2619                         /* RFC4862 Section 5.5.3e:
2620                          * "Note that the preferred lifetime of the
2621                          *  corresponding address is always reset to
2622                          *  the Preferred Lifetime in the received
2623                          *  Prefix Information option, regardless of
2624                          *  whether the valid lifetime is also reset or
2625                          *  ignored."
2626                          *
2627                          * So we should always update prefered_lft here.
2628                          */
2629                         update_lft = 1;
2630                 }
2631
2632                 if (update_lft) {
2633                         ifp->valid_lft = valid_lft;
2634                         ifp->prefered_lft = prefered_lft;
2635                         ifp->tstamp = now;
2636                         flags = ifp->flags;
2637                         ifp->flags &= ~IFA_F_DEPRECATED;
2638                         spin_unlock_bh(&ifp->lock);
2639
2640                         if (!(flags&IFA_F_TENTATIVE))
2641                                 ipv6_ifa_notify(0, ifp);
2642                 } else
2643                         spin_unlock_bh(&ifp->lock);
2644
2645                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2646                                  create, now);
2647
2648                 in6_ifa_put(ifp);
2649                 addrconf_verify();
2650         }
2651
2652         return 0;
2653 }
2654 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2655
2656 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2657 {
2658         struct prefix_info *pinfo;
2659         __u32 valid_lft;
2660         __u32 prefered_lft;
2661         int addr_type, err;
2662         u32 addr_flags = 0;
2663         struct inet6_dev *in6_dev;
2664         struct net *net = dev_net(dev);
2665
2666         pinfo = (struct prefix_info *) opt;
2667
2668         if (len < sizeof(struct prefix_info)) {
2669                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2670                 return;
2671         }
2672
2673         /*
2674          *      Validation checks ([ADDRCONF], page 19)
2675          */
2676
2677         addr_type = ipv6_addr_type(&pinfo->prefix);
2678
2679         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2680                 return;
2681
2682         valid_lft = ntohl(pinfo->valid);
2683         prefered_lft = ntohl(pinfo->prefered);
2684
2685         if (prefered_lft > valid_lft) {
2686                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2687                 return;
2688         }
2689
2690         in6_dev = in6_dev_get(dev);
2691
2692         if (!in6_dev) {
2693                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2694                                     dev->name);
2695                 return;
2696         }
2697
2698         /*
2699          *      Two things going on here:
2700          *      1) Add routes for on-link prefixes
2701          *      2) Configure prefixes with the auto flag set
2702          */
2703
2704         if (pinfo->onlink) {
2705                 struct fib6_info *rt;
2706                 unsigned long rt_expires;
2707
2708                 /* Avoid arithmetic overflow. Really, we could
2709                  * save rt_expires in seconds, likely valid_lft,
2710                  * but it would require division in fib gc, that it
2711                  * not good.
2712                  */
2713                 if (HZ > USER_HZ)
2714                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2715                 else
2716                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2717
2718                 if (addrconf_finite_timeout(rt_expires))
2719                         rt_expires *= HZ;
2720
2721                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2722                                                pinfo->prefix_len,
2723                                                dev,
2724                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2725                                                RTF_DEFAULT, true);
2726
2727                 if (rt) {
2728                         /* Autoconf prefix route */
2729                         if (valid_lft == 0) {
2730                                 ip6_del_rt(net, rt);
2731                                 rt = NULL;
2732                         } else if (addrconf_finite_timeout(rt_expires)) {
2733                                 /* not infinity */
2734                                 fib6_set_expires(rt, jiffies + rt_expires);
2735                         } else {
2736                                 fib6_clean_expires(rt);
2737                         }
2738                 } else if (valid_lft) {
2739                         clock_t expires = 0;
2740                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2741                         if (addrconf_finite_timeout(rt_expires)) {
2742                                 /* not infinity */
2743                                 flags |= RTF_EXPIRES;
2744                                 expires = jiffies_to_clock_t(rt_expires);
2745                         }
2746                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2747                                               0, dev, expires, flags,
2748                                               GFP_ATOMIC);
2749                 }
2750                 fib6_info_release(rt);
2751         }
2752
2753         /* Try to figure out our local address for this prefix */
2754
2755         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2756                 struct in6_addr addr;
2757                 bool tokenized = false, dev_addr_generated = false;
2758
2759                 if (pinfo->prefix_len == 64) {
2760                         memcpy(&addr, &pinfo->prefix, 8);
2761
2762                         if (!ipv6_addr_any(&in6_dev->token)) {
2763                                 read_lock_bh(&in6_dev->lock);
2764                                 memcpy(addr.s6_addr + 8,
2765                                        in6_dev->token.s6_addr + 8, 8);
2766                                 read_unlock_bh(&in6_dev->lock);
2767                                 tokenized = true;
2768                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2769                                    !ipv6_generate_stable_address(&addr, 0,
2770                                                                  in6_dev)) {
2771                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2772                                 goto ok;
2773                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2774                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2775                                 goto put;
2776                         } else {
2777                                 dev_addr_generated = true;
2778                         }
2779                         goto ok;
2780                 }
2781                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2782                                     pinfo->prefix_len);
2783                 goto put;
2784
2785 ok:
2786                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2787                                                    &addr, addr_type,
2788                                                    addr_flags, sllao,
2789                                                    tokenized, valid_lft,
2790                                                    prefered_lft);
2791                 if (err)
2792                         goto put;
2793
2794                 /* Ignore error case here because previous prefix add addr was
2795                  * successful which will be notified.
2796                  */
2797                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2798                                               addr_type, addr_flags, sllao,
2799                                               tokenized, valid_lft,
2800                                               prefered_lft,
2801                                               dev_addr_generated);
2802         }
2803         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2804 put:
2805         in6_dev_put(in6_dev);
2806 }
2807
2808 /*
2809  *      Set destination address.
2810  *      Special case for SIT interfaces where we create a new "virtual"
2811  *      device.
2812  */
2813 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2814 {
2815         struct in6_ifreq ireq;
2816         struct net_device *dev;
2817         int err = -EINVAL;
2818
2819         rtnl_lock();
2820
2821         err = -EFAULT;
2822         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2823                 goto err_exit;
2824
2825         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2826
2827         err = -ENODEV;
2828         if (!dev)
2829                 goto err_exit;
2830
2831 #if IS_ENABLED(CONFIG_IPV6_SIT)
2832         if (dev->type == ARPHRD_SIT) {
2833                 const struct net_device_ops *ops = dev->netdev_ops;
2834                 struct ifreq ifr;
2835                 struct ip_tunnel_parm p;
2836
2837                 err = -EADDRNOTAVAIL;
2838                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2839                         goto err_exit;
2840
2841                 memset(&p, 0, sizeof(p));
2842                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2843                 p.iph.saddr = 0;
2844                 p.iph.version = 4;
2845                 p.iph.ihl = 5;
2846                 p.iph.protocol = IPPROTO_IPV6;
2847                 p.iph.ttl = 64;
2848                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2849
2850                 if (ops->ndo_do_ioctl) {
2851                         mm_segment_t oldfs = get_fs();
2852
2853                         set_fs(KERNEL_DS);
2854                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2855                         set_fs(oldfs);
2856                 } else
2857                         err = -EOPNOTSUPP;
2858
2859                 if (err == 0) {
2860                         err = -ENOBUFS;
2861                         dev = __dev_get_by_name(net, p.name);
2862                         if (!dev)
2863                                 goto err_exit;
2864                         err = dev_open(dev, NULL);
2865                 }
2866         }
2867 #endif
2868
2869 err_exit:
2870         rtnl_unlock();
2871         return err;
2872 }
2873
2874 static int ipv6_mc_config(struct sock *sk, bool join,
2875                           const struct in6_addr *addr, int ifindex)
2876 {
2877         int ret;
2878
2879         ASSERT_RTNL();
2880
2881         lock_sock(sk);
2882         if (join)
2883                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2884         else
2885                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2886         release_sock(sk);
2887
2888         return ret;
2889 }
2890
2891 /*
2892  *      Manual configuration of address on an interface
2893  */
2894 static int inet6_addr_add(struct net *net, int ifindex,
2895                           struct ifa6_config *cfg,
2896                           struct netlink_ext_ack *extack)
2897 {
2898         struct inet6_ifaddr *ifp;
2899         struct inet6_dev *idev;
2900         struct net_device *dev;
2901         unsigned long timeout;
2902         clock_t expires;
2903         u32 flags;
2904
2905         ASSERT_RTNL();
2906
2907         if (cfg->plen > 128)
2908                 return -EINVAL;
2909
2910         /* check the lifetime */
2911         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2912                 return -EINVAL;
2913
2914         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2915                 return -EINVAL;
2916
2917         dev = __dev_get_by_index(net, ifindex);
2918         if (!dev)
2919                 return -ENODEV;
2920
2921         idev = addrconf_add_dev(dev);
2922         if (IS_ERR(idev))
2923                 return PTR_ERR(idev);
2924
2925         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2926                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2927                                          true, cfg->pfx, ifindex);
2928
2929                 if (ret < 0)
2930                         return ret;
2931         }
2932
2933         cfg->scope = ipv6_addr_scope(cfg->pfx);
2934
2935         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2936         if (addrconf_finite_timeout(timeout)) {
2937                 expires = jiffies_to_clock_t(timeout * HZ);
2938                 cfg->valid_lft = timeout;
2939                 flags = RTF_EXPIRES;
2940         } else {
2941                 expires = 0;
2942                 flags = 0;
2943                 cfg->ifa_flags |= IFA_F_PERMANENT;
2944         }
2945
2946         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2947         if (addrconf_finite_timeout(timeout)) {
2948                 if (timeout == 0)
2949                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2950                 cfg->preferred_lft = timeout;
2951         }
2952
2953         ifp = ipv6_add_addr(idev, cfg, true, extack);
2954         if (!IS_ERR(ifp)) {
2955                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2956                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2957                                               ifp->rt_priority, dev, expires,
2958                                               flags, GFP_KERNEL);
2959                 }
2960
2961                 /* Send a netlink notification if DAD is enabled and
2962                  * optimistic flag is not set
2963                  */
2964                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2965                         ipv6_ifa_notify(0, ifp);
2966                 /*
2967                  * Note that section 3.1 of RFC 4429 indicates
2968                  * that the Optimistic flag should not be set for
2969                  * manually configured addresses
2970                  */
2971                 addrconf_dad_start(ifp);
2972                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2973                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2974                                          cfg->preferred_lft, true, jiffies);
2975                 in6_ifa_put(ifp);
2976                 addrconf_verify_rtnl();
2977                 return 0;
2978         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2979                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2980                                cfg->pfx, ifindex);
2981         }
2982
2983         return PTR_ERR(ifp);
2984 }
2985
2986 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2987                           const struct in6_addr *pfx, unsigned int plen)
2988 {
2989         struct inet6_ifaddr *ifp;
2990         struct inet6_dev *idev;
2991         struct net_device *dev;
2992
2993         if (plen > 128)
2994                 return -EINVAL;
2995
2996         dev = __dev_get_by_index(net, ifindex);
2997         if (!dev)
2998                 return -ENODEV;
2999
3000         idev = __in6_dev_get(dev);
3001         if (!idev)
3002                 return -ENXIO;
3003
3004         read_lock_bh(&idev->lock);
3005         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3006                 if (ifp->prefix_len == plen &&
3007                     ipv6_addr_equal(pfx, &ifp->addr)) {
3008                         in6_ifa_hold(ifp);
3009                         read_unlock_bh(&idev->lock);
3010
3011                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3012                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3013                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3014                                                  jiffies);
3015                         ipv6_del_addr(ifp);
3016                         addrconf_verify_rtnl();
3017                         if (ipv6_addr_is_multicast(pfx)) {
3018                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3019                                                false, pfx, dev->ifindex);
3020                         }
3021                         return 0;
3022                 }
3023         }
3024         read_unlock_bh(&idev->lock);
3025         return -EADDRNOTAVAIL;
3026 }
3027
3028
3029 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3030 {
3031         struct ifa6_config cfg = {
3032                 .ifa_flags = IFA_F_PERMANENT,
3033                 .preferred_lft = INFINITY_LIFE_TIME,
3034                 .valid_lft = INFINITY_LIFE_TIME,
3035         };
3036         struct in6_ifreq ireq;
3037         int err;
3038
3039         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3040                 return -EPERM;
3041
3042         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3043                 return -EFAULT;
3044
3045         cfg.pfx = &ireq.ifr6_addr;
3046         cfg.plen = ireq.ifr6_prefixlen;
3047
3048         rtnl_lock();
3049         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3050         rtnl_unlock();
3051         return err;
3052 }
3053
3054 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3055 {
3056         struct in6_ifreq ireq;
3057         int err;
3058
3059         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3060                 return -EPERM;
3061
3062         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3063                 return -EFAULT;
3064
3065         rtnl_lock();
3066         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3067                              ireq.ifr6_prefixlen);
3068         rtnl_unlock();
3069         return err;
3070 }
3071
3072 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3073                      int plen, int scope)
3074 {
3075         struct inet6_ifaddr *ifp;
3076         struct ifa6_config cfg = {
3077                 .pfx = addr,
3078                 .plen = plen,
3079                 .ifa_flags = IFA_F_PERMANENT,
3080                 .valid_lft = INFINITY_LIFE_TIME,
3081                 .preferred_lft = INFINITY_LIFE_TIME,
3082                 .scope = scope
3083         };
3084
3085         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3086         if (!IS_ERR(ifp)) {
3087                 spin_lock_bh(&ifp->lock);
3088                 ifp->flags &= ~IFA_F_TENTATIVE;
3089                 spin_unlock_bh(&ifp->lock);
3090                 rt_genid_bump_ipv6(dev_net(idev->dev));
3091                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3092                 in6_ifa_put(ifp);
3093         }
3094 }
3095
3096 #if IS_ENABLED(CONFIG_IPV6_SIT)
3097 static void sit_add_v4_addrs(struct inet6_dev *idev)
3098 {
3099         struct in6_addr addr;
3100         struct net_device *dev;
3101         struct net *net = dev_net(idev->dev);
3102         int scope, plen;
3103         u32 pflags = 0;
3104
3105         ASSERT_RTNL();
3106
3107         memset(&addr, 0, sizeof(struct in6_addr));
3108         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3109
3110         if (idev->dev->flags&IFF_POINTOPOINT) {
3111                 addr.s6_addr32[0] = htonl(0xfe800000);
3112                 scope = IFA_LINK;
3113                 plen = 64;
3114         } else {
3115                 scope = IPV6_ADDR_COMPATv4;
3116                 plen = 96;
3117                 pflags |= RTF_NONEXTHOP;
3118         }
3119
3120         if (addr.s6_addr32[3]) {
3121                 add_addr(idev, &addr, plen, scope);
3122                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3123                                       GFP_KERNEL);
3124                 return;
3125         }
3126
3127         for_each_netdev(net, dev) {
3128                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3129                 if (in_dev && (dev->flags & IFF_UP)) {
3130                         struct in_ifaddr *ifa;
3131                         int flag = scope;
3132
3133                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3134                                 addr.s6_addr32[3] = ifa->ifa_local;
3135
3136                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3137                                         continue;
3138                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3139                                         if (idev->dev->flags&IFF_POINTOPOINT)
3140                                                 continue;
3141                                         flag |= IFA_HOST;
3142                                 }
3143
3144                                 add_addr(idev, &addr, plen, flag);
3145                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3146                                                       0, pflags, GFP_KERNEL);
3147                         }
3148                 }
3149         }
3150 }
3151 #endif
3152
3153 static void init_loopback(struct net_device *dev)
3154 {
3155         struct inet6_dev  *idev;
3156
3157         /* ::1 */
3158
3159         ASSERT_RTNL();
3160
3161         idev = ipv6_find_idev(dev);
3162         if (IS_ERR(idev)) {
3163                 pr_debug("%s: add_dev failed\n", __func__);
3164                 return;
3165         }
3166
3167         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3168 }
3169
3170 void addrconf_add_linklocal(struct inet6_dev *idev,
3171                             const struct in6_addr *addr, u32 flags)
3172 {
3173         struct ifa6_config cfg = {
3174                 .pfx = addr,
3175                 .plen = 64,
3176                 .ifa_flags = flags | IFA_F_PERMANENT,
3177                 .valid_lft = INFINITY_LIFE_TIME,
3178                 .preferred_lft = INFINITY_LIFE_TIME,
3179                 .scope = IFA_LINK
3180         };
3181         struct inet6_ifaddr *ifp;
3182
3183 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3184         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3185              idev->cnf.optimistic_dad) &&
3186             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3187                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3188 #endif
3189
3190         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3191         if (!IS_ERR(ifp)) {
3192                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3193                                       0, 0, GFP_ATOMIC);
3194                 addrconf_dad_start(ifp);
3195                 in6_ifa_put(ifp);
3196         }
3197 }
3198 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3199
3200 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3201 {
3202         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3203                 return true;
3204
3205         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3206             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3207                 return true;
3208
3209         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3210             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3211                 return true;
3212
3213         return false;
3214 }
3215
3216 static int ipv6_generate_stable_address(struct in6_addr *address,
3217                                         u8 dad_count,
3218                                         const struct inet6_dev *idev)
3219 {
3220         static DEFINE_SPINLOCK(lock);
3221         static __u32 digest[SHA_DIGEST_WORDS];
3222         static __u32 workspace[SHA_WORKSPACE_WORDS];
3223
3224         static union {
3225                 char __data[SHA_MESSAGE_BYTES];
3226                 struct {
3227                         struct in6_addr secret;
3228                         __be32 prefix[2];
3229                         unsigned char hwaddr[MAX_ADDR_LEN];
3230                         u8 dad_count;
3231                 } __packed;
3232         } data;
3233
3234         struct in6_addr secret;
3235         struct in6_addr temp;
3236         struct net *net = dev_net(idev->dev);
3237
3238         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3239
3240         if (idev->cnf.stable_secret.initialized)
3241                 secret = idev->cnf.stable_secret.secret;
3242         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3243                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3244         else
3245                 return -1;
3246
3247 retry:
3248         spin_lock_bh(&lock);
3249
3250         sha_init(digest);
3251         memset(&data, 0, sizeof(data));
3252         memset(workspace, 0, sizeof(workspace));
3253         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3254         data.prefix[0] = address->s6_addr32[0];
3255         data.prefix[1] = address->s6_addr32[1];
3256         data.secret = secret;
3257         data.dad_count = dad_count;
3258
3259         sha_transform(digest, data.__data, workspace);
3260
3261         temp = *address;
3262         temp.s6_addr32[2] = (__force __be32)digest[0];
3263         temp.s6_addr32[3] = (__force __be32)digest[1];
3264
3265         spin_unlock_bh(&lock);
3266
3267         if (ipv6_reserved_interfaceid(temp)) {
3268                 dad_count++;
3269                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3270                         return -1;
3271                 goto retry;
3272         }
3273
3274         *address = temp;
3275         return 0;
3276 }
3277
3278 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3279 {
3280         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3281
3282         if (s->initialized)
3283                 return;
3284         s = &idev->cnf.stable_secret;
3285         get_random_bytes(&s->secret, sizeof(s->secret));
3286         s->initialized = true;
3287 }
3288
3289 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3290 {
3291         struct in6_addr addr;
3292
3293         /* no link local addresses on L3 master devices */
3294         if (netif_is_l3_master(idev->dev))
3295                 return;
3296
3297         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3298
3299         switch (idev->cnf.addr_gen_mode) {
3300         case IN6_ADDR_GEN_MODE_RANDOM:
3301                 ipv6_gen_mode_random_init(idev);
3302                 /* fallthrough */
3303         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3304                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3305                         addrconf_add_linklocal(idev, &addr,
3306                                                IFA_F_STABLE_PRIVACY);
3307                 else if (prefix_route)
3308                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3309                                               0, 0, GFP_KERNEL);
3310                 break;
3311         case IN6_ADDR_GEN_MODE_EUI64:
3312                 /* addrconf_add_linklocal also adds a prefix_route and we
3313                  * only need to care about prefix routes if ipv6_generate_eui64
3314                  * couldn't generate one.
3315                  */
3316                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3317                         addrconf_add_linklocal(idev, &addr, 0);
3318                 else if (prefix_route)
3319                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3320                                               0, 0, GFP_KERNEL);
3321                 break;
3322         case IN6_ADDR_GEN_MODE_NONE:
3323         default:
3324                 /* will not add any link local address */
3325                 break;
3326         }
3327 }
3328
3329 static void addrconf_dev_config(struct net_device *dev)
3330 {
3331         struct inet6_dev *idev;
3332
3333         ASSERT_RTNL();
3334
3335         if ((dev->type != ARPHRD_ETHER) &&
3336             (dev->type != ARPHRD_FDDI) &&
3337             (dev->type != ARPHRD_ARCNET) &&
3338             (dev->type != ARPHRD_INFINIBAND) &&
3339             (dev->type != ARPHRD_IEEE1394) &&
3340             (dev->type != ARPHRD_TUNNEL6) &&
3341             (dev->type != ARPHRD_6LOWPAN) &&
3342             (dev->type != ARPHRD_IP6GRE) &&
3343             (dev->type != ARPHRD_IPGRE) &&
3344             (dev->type != ARPHRD_TUNNEL) &&
3345             (dev->type != ARPHRD_NONE) &&
3346             (dev->type != ARPHRD_RAWIP)) {
3347                 /* Alas, we support only Ethernet autoconfiguration. */
3348                 return;
3349         }
3350
3351         idev = addrconf_add_dev(dev);
3352         if (IS_ERR(idev))
3353                 return;
3354
3355         /* this device type has no EUI support */
3356         if (dev->type == ARPHRD_NONE &&
3357             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3358                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3359
3360         addrconf_addr_gen(idev, false);
3361 }
3362
3363 #if IS_ENABLED(CONFIG_IPV6_SIT)
3364 static void addrconf_sit_config(struct net_device *dev)
3365 {
3366         struct inet6_dev *idev;
3367
3368         ASSERT_RTNL();
3369
3370         /*
3371          * Configure the tunnel with one of our IPv4
3372          * addresses... we should configure all of
3373          * our v4 addrs in the tunnel
3374          */
3375
3376         idev = ipv6_find_idev(dev);
3377         if (IS_ERR(idev)) {
3378                 pr_debug("%s: add_dev failed\n", __func__);
3379                 return;
3380         }
3381
3382         if (dev->priv_flags & IFF_ISATAP) {
3383                 addrconf_addr_gen(idev, false);
3384                 return;
3385         }
3386
3387         sit_add_v4_addrs(idev);
3388
3389         if (dev->flags&IFF_POINTOPOINT)
3390                 addrconf_add_mroute(dev);
3391 }
3392 #endif
3393
3394 #if IS_ENABLED(CONFIG_NET_IPGRE)
3395 static void addrconf_gre_config(struct net_device *dev)
3396 {
3397         struct inet6_dev *idev;
3398
3399         ASSERT_RTNL();
3400
3401         idev = ipv6_find_idev(dev);
3402         if (IS_ERR(idev)) {
3403                 pr_debug("%s: add_dev failed\n", __func__);
3404                 return;
3405         }
3406
3407         addrconf_addr_gen(idev, true);
3408         if (dev->flags & IFF_POINTOPOINT)
3409                 addrconf_add_mroute(dev);
3410 }
3411 #endif
3412
3413 static int fixup_permanent_addr(struct net *net,
3414                                 struct inet6_dev *idev,
3415                                 struct inet6_ifaddr *ifp)
3416 {
3417         /* !fib6_node means the host route was removed from the
3418          * FIB, for example, if 'lo' device is taken down. In that
3419          * case regenerate the host route.
3420          */
3421         if (!ifp->rt || !ifp->rt->fib6_node) {
3422                 struct fib6_info *f6i, *prev;
3423
3424                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3425                                          GFP_ATOMIC);
3426                 if (IS_ERR(f6i))
3427                         return PTR_ERR(f6i);
3428
3429                 /* ifp->rt can be accessed outside of rtnl */
3430                 spin_lock(&ifp->lock);
3431                 prev = ifp->rt;
3432                 ifp->rt = f6i;
3433                 spin_unlock(&ifp->lock);
3434
3435                 fib6_info_release(prev);
3436         }
3437
3438         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3439                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3440                                       ifp->rt_priority, idev->dev, 0, 0,
3441                                       GFP_ATOMIC);
3442         }
3443
3444         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3445                 addrconf_dad_start(ifp);
3446
3447         return 0;
3448 }
3449
3450 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3451 {
3452         struct inet6_ifaddr *ifp, *tmp;
3453         struct inet6_dev *idev;
3454
3455         idev = __in6_dev_get(dev);
3456         if (!idev)
3457                 return;
3458
3459         write_lock_bh(&idev->lock);
3460
3461         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3462                 if ((ifp->flags & IFA_F_PERMANENT) &&
3463                     fixup_permanent_addr(net, idev, ifp) < 0) {
3464                         write_unlock_bh(&idev->lock);
3465                         in6_ifa_hold(ifp);
3466                         ipv6_del_addr(ifp);
3467                         write_lock_bh(&idev->lock);
3468
3469                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3470                                              idev->dev->name, &ifp->addr);
3471                 }
3472         }
3473
3474         write_unlock_bh(&idev->lock);
3475 }
3476
3477 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3478                            void *ptr)
3479 {
3480         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3481         struct netdev_notifier_change_info *change_info;
3482         struct netdev_notifier_changeupper_info *info;
3483         struct inet6_dev *idev = __in6_dev_get(dev);
3484         struct net *net = dev_net(dev);
3485         int run_pending = 0;
3486         int err;
3487
3488         switch (event) {
3489         case NETDEV_REGISTER:
3490                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3491                         idev = ipv6_add_dev(dev);
3492                         if (IS_ERR(idev))
3493                                 return notifier_from_errno(PTR_ERR(idev));
3494                 }
3495                 break;
3496
3497         case NETDEV_CHANGEMTU:
3498                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3499                 if (dev->mtu < IPV6_MIN_MTU) {
3500                         addrconf_ifdown(dev, dev != net->loopback_dev);
3501                         break;
3502                 }
3503
3504                 if (idev) {
3505                         rt6_mtu_change(dev, dev->mtu);
3506                         idev->cnf.mtu6 = dev->mtu;
3507                         break;
3508                 }
3509
3510                 /* allocate new idev */
3511                 idev = ipv6_add_dev(dev);
3512                 if (IS_ERR(idev))
3513                         break;
3514
3515                 /* device is still not ready */
3516                 if (!(idev->if_flags & IF_READY))
3517                         break;
3518
3519                 run_pending = 1;
3520
3521                 /* fall through */
3522
3523         case NETDEV_UP:
3524         case NETDEV_CHANGE:
3525                 if (dev->flags & IFF_SLAVE)
3526                         break;
3527
3528                 if (idev && idev->cnf.disable_ipv6)
3529                         break;
3530
3531                 if (event == NETDEV_UP) {
3532                         /* restore routes for permanent addresses */
3533                         addrconf_permanent_addr(net, dev);
3534
3535                         if (!addrconf_link_ready(dev)) {
3536                                 /* device is not ready yet. */
3537                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3538                                          dev->name);
3539                                 break;
3540                         }
3541
3542                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3543                                 idev = ipv6_add_dev(dev);
3544
3545                         if (!IS_ERR_OR_NULL(idev)) {
3546                                 idev->if_flags |= IF_READY;
3547                                 run_pending = 1;
3548                         }
3549                 } else if (event == NETDEV_CHANGE) {
3550                         if (!addrconf_link_ready(dev)) {
3551                                 /* device is still not ready. */
3552                                 rt6_sync_down_dev(dev, event);
3553                                 break;
3554                         }
3555
3556                         if (!IS_ERR_OR_NULL(idev)) {
3557                                 if (idev->if_flags & IF_READY) {
3558                                         /* device is already configured -
3559                                          * but resend MLD reports, we might
3560                                          * have roamed and need to update
3561                                          * multicast snooping switches
3562                                          */
3563                                         ipv6_mc_up(idev);
3564                                         change_info = ptr;
3565                                         if (change_info->flags_changed & IFF_NOARP)
3566                                                 addrconf_dad_run(idev, true);
3567                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3568                                         break;
3569                                 }
3570                                 idev->if_flags |= IF_READY;
3571                         }
3572
3573                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3574                                 dev->name);
3575
3576                         run_pending = 1;
3577                 }
3578
3579                 switch (dev->type) {
3580 #if IS_ENABLED(CONFIG_IPV6_SIT)
3581                 case ARPHRD_SIT:
3582                         addrconf_sit_config(dev);
3583                         break;
3584 #endif
3585 #if IS_ENABLED(CONFIG_NET_IPGRE)
3586                 case ARPHRD_IPGRE:
3587                         addrconf_gre_config(dev);
3588                         break;
3589 #endif
3590                 case ARPHRD_LOOPBACK:
3591                         init_loopback(dev);
3592                         break;
3593
3594                 default:
3595                         addrconf_dev_config(dev);
3596                         break;
3597                 }
3598
3599                 if (!IS_ERR_OR_NULL(idev)) {
3600                         if (run_pending)
3601                                 addrconf_dad_run(idev, false);
3602
3603                         /* Device has an address by now */
3604                         rt6_sync_up(dev, RTNH_F_DEAD);
3605
3606                         /*
3607                          * If the MTU changed during the interface down,
3608                          * when the interface up, the changed MTU must be
3609                          * reflected in the idev as well as routers.
3610                          */
3611                         if (idev->cnf.mtu6 != dev->mtu &&
3612                             dev->mtu >= IPV6_MIN_MTU) {
3613                                 rt6_mtu_change(dev, dev->mtu);
3614                                 idev->cnf.mtu6 = dev->mtu;
3615                         }
3616                         idev->tstamp = jiffies;
3617                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3618
3619                         /*
3620                          * If the changed mtu during down is lower than
3621                          * IPV6_MIN_MTU stop IPv6 on this interface.
3622                          */
3623                         if (dev->mtu < IPV6_MIN_MTU)
3624                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3625                 }
3626                 break;
3627
3628         case NETDEV_DOWN:
3629         case NETDEV_UNREGISTER:
3630                 /*
3631                  *      Remove all addresses from this interface.
3632                  */
3633                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3634                 break;
3635
3636         case NETDEV_CHANGENAME:
3637                 if (idev) {
3638                         snmp6_unregister_dev(idev);
3639                         addrconf_sysctl_unregister(idev);
3640                         err = addrconf_sysctl_register(idev);
3641                         if (err)
3642                                 return notifier_from_errno(err);
3643                         err = snmp6_register_dev(idev);
3644                         if (err) {
3645                                 addrconf_sysctl_unregister(idev);
3646                                 return notifier_from_errno(err);
3647                         }
3648                 }
3649                 break;
3650
3651         case NETDEV_PRE_TYPE_CHANGE:
3652         case NETDEV_POST_TYPE_CHANGE:
3653                 if (idev)
3654                         addrconf_type_change(dev, event);
3655                 break;
3656
3657         case NETDEV_CHANGEUPPER:
3658                 info = ptr;
3659
3660                 /* flush all routes if dev is linked to or unlinked from
3661                  * an L3 master device (e.g., VRF)
3662                  */
3663                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3664                         addrconf_ifdown(dev, 0);
3665         }
3666
3667         return NOTIFY_OK;
3668 }
3669
3670 /*
3671  *      addrconf module should be notified of a device going up
3672  */
3673 static struct notifier_block ipv6_dev_notf = {
3674         .notifier_call = addrconf_notify,
3675         .priority = ADDRCONF_NOTIFY_PRIORITY,
3676 };
3677
3678 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3679 {
3680         struct inet6_dev *idev;
3681         ASSERT_RTNL();
3682
3683         idev = __in6_dev_get(dev);
3684
3685         if (event == NETDEV_POST_TYPE_CHANGE)
3686                 ipv6_mc_remap(idev);
3687         else if (event == NETDEV_PRE_TYPE_CHANGE)
3688                 ipv6_mc_unmap(idev);
3689 }
3690
3691 static bool addr_is_local(const struct in6_addr *addr)
3692 {
3693         return ipv6_addr_type(addr) &
3694                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3695 }
3696
3697 static int addrconf_ifdown(struct net_device *dev, int how)
3698 {
3699         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3700         struct net *net = dev_net(dev);
3701         struct inet6_dev *idev;
3702         struct inet6_ifaddr *ifa, *tmp;
3703         bool keep_addr = false;
3704         int state, i;
3705
3706         ASSERT_RTNL();
3707
3708         rt6_disable_ip(dev, event);
3709
3710         idev = __in6_dev_get(dev);
3711         if (!idev)
3712                 return -ENODEV;
3713
3714         /*
3715          * Step 1: remove reference to ipv6 device from parent device.
3716          *         Do not dev_put!
3717          */
3718         if (how) {
3719                 idev->dead = 1;
3720
3721                 /* protected by rtnl_lock */
3722                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3723
3724                 /* Step 1.5: remove snmp6 entry */
3725                 snmp6_unregister_dev(idev);
3726
3727         }
3728
3729         /* combine the user config with event to determine if permanent
3730          * addresses are to be removed from address hash table
3731          */
3732         if (!how && !idev->cnf.disable_ipv6) {
3733                 /* aggregate the system setting and interface setting */
3734                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3735
3736                 if (!_keep_addr)
3737                         _keep_addr = idev->cnf.keep_addr_on_down;
3738
3739                 keep_addr = (_keep_addr > 0);
3740         }
3741
3742         /* Step 2: clear hash table */
3743         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3744                 struct hlist_head *h = &inet6_addr_lst[i];
3745
3746                 spin_lock_bh(&addrconf_hash_lock);
3747 restart:
3748                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3749                         if (ifa->idev == idev) {
3750                                 addrconf_del_dad_work(ifa);
3751                                 /* combined flag + permanent flag decide if
3752                                  * address is retained on a down event
3753                                  */
3754                                 if (!keep_addr ||
3755                                     !(ifa->flags & IFA_F_PERMANENT) ||
3756                                     addr_is_local(&ifa->addr)) {
3757                                         hlist_del_init_rcu(&ifa->addr_lst);
3758                                         goto restart;
3759                                 }
3760                         }
3761                 }
3762                 spin_unlock_bh(&addrconf_hash_lock);
3763         }
3764
3765         write_lock_bh(&idev->lock);
3766
3767         addrconf_del_rs_timer(idev);
3768
3769         /* Step 2: clear flags for stateless addrconf */
3770         if (!how)
3771                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3772
3773         /* Step 3: clear tempaddr list */
3774         while (!list_empty(&idev->tempaddr_list)) {
3775                 ifa = list_first_entry(&idev->tempaddr_list,
3776                                        struct inet6_ifaddr, tmp_list);
3777                 list_del(&ifa->tmp_list);
3778                 write_unlock_bh(&idev->lock);
3779                 spin_lock_bh(&ifa->lock);
3780
3781                 if (ifa->ifpub) {
3782                         in6_ifa_put(ifa->ifpub);
3783                         ifa->ifpub = NULL;
3784                 }
3785                 spin_unlock_bh(&ifa->lock);
3786                 in6_ifa_put(ifa);
3787                 write_lock_bh(&idev->lock);
3788         }
3789
3790         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3791                 struct fib6_info *rt = NULL;
3792                 bool keep;
3793
3794                 addrconf_del_dad_work(ifa);
3795
3796                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3797                         !addr_is_local(&ifa->addr);
3798
3799                 write_unlock_bh(&idev->lock);
3800                 spin_lock_bh(&ifa->lock);
3801
3802                 if (keep) {
3803                         /* set state to skip the notifier below */
3804                         state = INET6_IFADDR_STATE_DEAD;
3805                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3806                         if (!(ifa->flags & IFA_F_NODAD))
3807                                 ifa->flags |= IFA_F_TENTATIVE;
3808
3809                         rt = ifa->rt;
3810                         ifa->rt = NULL;
3811                 } else {
3812                         state = ifa->state;
3813                         ifa->state = INET6_IFADDR_STATE_DEAD;
3814                 }
3815
3816                 spin_unlock_bh(&ifa->lock);
3817
3818                 if (rt)
3819                         ip6_del_rt(net, rt);
3820
3821                 if (state != INET6_IFADDR_STATE_DEAD) {
3822                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3823                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3824                 } else {
3825                         if (idev->cnf.forwarding)
3826                                 addrconf_leave_anycast(ifa);
3827                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3828                 }
3829
3830                 write_lock_bh(&idev->lock);
3831                 if (!keep) {
3832                         list_del_rcu(&ifa->if_list);
3833                         in6_ifa_put(ifa);
3834                 }
3835         }
3836
3837         write_unlock_bh(&idev->lock);
3838
3839         /* Step 5: Discard anycast and multicast list */
3840         if (how) {
3841                 ipv6_ac_destroy_dev(idev);
3842                 ipv6_mc_destroy_dev(idev);
3843         } else {
3844                 ipv6_mc_down(idev);
3845         }
3846
3847         idev->tstamp = jiffies;
3848
3849         /* Last: Shot the device (if unregistered) */
3850         if (how) {
3851                 addrconf_sysctl_unregister(idev);
3852                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3853                 neigh_ifdown(&nd_tbl, dev);
3854                 in6_dev_put(idev);
3855         }
3856         return 0;
3857 }
3858
3859 static void addrconf_rs_timer(struct timer_list *t)
3860 {
3861         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3862         struct net_device *dev = idev->dev;
3863         struct in6_addr lladdr;
3864
3865         write_lock(&idev->lock);
3866         if (idev->dead || !(idev->if_flags & IF_READY))
3867                 goto out;
3868
3869         if (!ipv6_accept_ra(idev))
3870                 goto out;
3871
3872         /* Announcement received after solicitation was sent */
3873         if (idev->if_flags & IF_RA_RCVD)
3874                 goto out;
3875
3876         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3877                 write_unlock(&idev->lock);
3878                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3879                         ndisc_send_rs(dev, &lladdr,
3880                                       &in6addr_linklocal_allrouters);
3881                 else
3882                         goto put;
3883
3884                 write_lock(&idev->lock);
3885                 idev->rs_interval = rfc3315_s14_backoff_update(
3886                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3887                 /* The wait after the last probe can be shorter */
3888                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3889                                              idev->cnf.rtr_solicits) ?
3890                                       idev->cnf.rtr_solicit_delay :
3891                                       idev->rs_interval);
3892         } else {
3893                 /*
3894                  * Note: we do not support deprecated "all on-link"
3895                  * assumption any longer.
3896                  */
3897                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3898         }
3899
3900 out:
3901         write_unlock(&idev->lock);
3902 put:
3903         in6_dev_put(idev);
3904 }
3905
3906 /*
3907  *      Duplicate Address Detection
3908  */
3909 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3910 {
3911         unsigned long rand_num;
3912         struct inet6_dev *idev = ifp->idev;
3913         u64 nonce;
3914
3915         if (ifp->flags & IFA_F_OPTIMISTIC)
3916                 rand_num = 0;
3917         else
3918                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3919
3920         nonce = 0;
3921         if (idev->cnf.enhanced_dad ||
3922             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3923                 do
3924                         get_random_bytes(&nonce, 6);
3925                 while (nonce == 0);
3926         }
3927         ifp->dad_nonce = nonce;
3928         ifp->dad_probes = idev->cnf.dad_transmits;
3929         addrconf_mod_dad_work(ifp, rand_num);
3930 }
3931
3932 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3933 {
3934         struct inet6_dev *idev = ifp->idev;
3935         struct net_device *dev = idev->dev;
3936         bool bump_id, notify = false;
3937         struct net *net;
3938
3939         addrconf_join_solict(dev, &ifp->addr);
3940
3941         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3942
3943         read_lock_bh(&idev->lock);
3944         spin_lock(&ifp->lock);
3945         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3946                 goto out;
3947
3948         net = dev_net(dev);
3949         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3950             (net->ipv6.devconf_all->accept_dad < 1 &&
3951              idev->cnf.accept_dad < 1) ||
3952             !(ifp->flags&IFA_F_TENTATIVE) ||
3953             ifp->flags & IFA_F_NODAD) {
3954                 bool send_na = false;
3955
3956                 if (ifp->flags & IFA_F_TENTATIVE &&
3957                     !(ifp->flags & IFA_F_OPTIMISTIC))
3958                         send_na = true;
3959                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3960                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3961                 spin_unlock(&ifp->lock);
3962                 read_unlock_bh(&idev->lock);
3963
3964                 addrconf_dad_completed(ifp, bump_id, send_na);
3965                 return;
3966         }
3967
3968         if (!(idev->if_flags & IF_READY)) {
3969                 spin_unlock(&ifp->lock);
3970                 read_unlock_bh(&idev->lock);
3971                 /*
3972                  * If the device is not ready:
3973                  * - keep it tentative if it is a permanent address.
3974                  * - otherwise, kill it.
3975                  */
3976                 in6_ifa_hold(ifp);
3977                 addrconf_dad_stop(ifp, 0);
3978                 return;
3979         }
3980
3981         /*
3982          * Optimistic nodes can start receiving
3983          * Frames right away
3984          */
3985         if (ifp->flags & IFA_F_OPTIMISTIC) {
3986                 ip6_ins_rt(net, ifp->rt);
3987                 if (ipv6_use_optimistic_addr(net, idev)) {
3988                         /* Because optimistic nodes can use this address,
3989                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3990                          */
3991                         notify = true;
3992                 }
3993         }
3994
3995         addrconf_dad_kick(ifp);
3996 out:
3997         spin_unlock(&ifp->lock);
3998         read_unlock_bh(&idev->lock);
3999         if (notify)
4000                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4001 }
4002
4003 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4004 {
4005         bool begin_dad = false;
4006
4007         spin_lock_bh(&ifp->lock);
4008         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4009                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4010                 begin_dad = true;
4011         }
4012         spin_unlock_bh(&ifp->lock);
4013
4014         if (begin_dad)
4015                 addrconf_mod_dad_work(ifp, 0);
4016 }
4017
4018 static void addrconf_dad_work(struct work_struct *w)
4019 {
4020         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4021                                                 struct inet6_ifaddr,
4022                                                 dad_work);
4023         struct inet6_dev *idev = ifp->idev;
4024         bool bump_id, disable_ipv6 = false;
4025         struct in6_addr mcaddr;
4026
4027         enum {
4028                 DAD_PROCESS,
4029                 DAD_BEGIN,
4030                 DAD_ABORT,
4031         } action = DAD_PROCESS;
4032
4033         rtnl_lock();
4034
4035         spin_lock_bh(&ifp->lock);
4036         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4037                 action = DAD_BEGIN;
4038                 ifp->state = INET6_IFADDR_STATE_DAD;
4039         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4040                 action = DAD_ABORT;
4041                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4042
4043                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4044                      idev->cnf.accept_dad > 1) &&
4045                     !idev->cnf.disable_ipv6 &&
4046                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4047                         struct in6_addr addr;
4048
4049                         addr.s6_addr32[0] = htonl(0xfe800000);
4050                         addr.s6_addr32[1] = 0;
4051
4052                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4053                             ipv6_addr_equal(&ifp->addr, &addr)) {
4054                                 /* DAD failed for link-local based on MAC */
4055                                 idev->cnf.disable_ipv6 = 1;
4056
4057                                 pr_info("%s: IPv6 being disabled!\n",
4058                                         ifp->idev->dev->name);
4059                                 disable_ipv6 = true;
4060                         }
4061                 }
4062         }
4063         spin_unlock_bh(&ifp->lock);
4064
4065         if (action == DAD_BEGIN) {
4066                 addrconf_dad_begin(ifp);
4067                 goto out;
4068         } else if (action == DAD_ABORT) {
4069                 in6_ifa_hold(ifp);
4070                 addrconf_dad_stop(ifp, 1);
4071                 if (disable_ipv6)
4072                         addrconf_ifdown(idev->dev, 0);
4073                 goto out;
4074         }
4075
4076         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4077                 goto out;
4078
4079         write_lock_bh(&idev->lock);
4080         if (idev->dead || !(idev->if_flags & IF_READY)) {
4081                 write_unlock_bh(&idev->lock);
4082                 goto out;
4083         }
4084
4085         spin_lock(&ifp->lock);
4086         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4087                 spin_unlock(&ifp->lock);
4088                 write_unlock_bh(&idev->lock);
4089                 goto out;
4090         }
4091
4092         if (ifp->dad_probes == 0) {
4093                 bool send_na = false;
4094
4095                 /*
4096                  * DAD was successful
4097                  */
4098
4099                 if (ifp->flags & IFA_F_TENTATIVE &&
4100                     !(ifp->flags & IFA_F_OPTIMISTIC))
4101                         send_na = true;
4102                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4103                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4104                 spin_unlock(&ifp->lock);
4105                 write_unlock_bh(&idev->lock);
4106
4107                 addrconf_dad_completed(ifp, bump_id, send_na);
4108
4109                 goto out;
4110         }
4111
4112         ifp->dad_probes--;
4113         addrconf_mod_dad_work(ifp,
4114                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4115         spin_unlock(&ifp->lock);
4116         write_unlock_bh(&idev->lock);
4117
4118         /* send a neighbour solicitation for our addr */
4119         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4120         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4121                       ifp->dad_nonce);
4122 out:
4123         in6_ifa_put(ifp);
4124         rtnl_unlock();
4125 }
4126
4127 /* ifp->idev must be at least read locked */
4128 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4129 {
4130         struct inet6_ifaddr *ifpiter;
4131         struct inet6_dev *idev = ifp->idev;
4132
4133         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4134                 if (ifpiter->scope > IFA_LINK)
4135                         break;
4136                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4137                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4138                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4139                     IFA_F_PERMANENT)
4140                         return false;
4141         }
4142         return true;
4143 }
4144
4145 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4146                                    bool send_na)
4147 {
4148         struct net_device *dev = ifp->idev->dev;
4149         struct in6_addr lladdr;
4150         bool send_rs, send_mld;
4151
4152         addrconf_del_dad_work(ifp);
4153
4154         /*
4155          *      Configure the address for reception. Now it is valid.
4156          */
4157
4158         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4159
4160         /* If added prefix is link local and we are prepared to process
4161            router advertisements, start sending router solicitations.
4162          */
4163
4164         read_lock_bh(&ifp->idev->lock);
4165         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4166         send_rs = send_mld &&
4167                   ipv6_accept_ra(ifp->idev) &&
4168                   ifp->idev->cnf.rtr_solicits != 0 &&
4169                   (dev->flags&IFF_LOOPBACK) == 0;
4170         read_unlock_bh(&ifp->idev->lock);
4171
4172         /* While dad is in progress mld report's source address is in6_addrany.
4173          * Resend with proper ll now.
4174          */
4175         if (send_mld)
4176                 ipv6_mc_dad_complete(ifp->idev);
4177
4178         /* send unsolicited NA if enabled */
4179         if (send_na &&
4180             (ifp->idev->cnf.ndisc_notify ||
4181              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4182                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4183                               /*router=*/ !!ifp->idev->cnf.forwarding,
4184                               /*solicited=*/ false, /*override=*/ true,
4185                               /*inc_opt=*/ true);
4186         }
4187
4188         if (send_rs) {
4189                 /*
4190                  *      If a host as already performed a random delay
4191                  *      [...] as part of DAD [...] there is no need
4192                  *      to delay again before sending the first RS
4193                  */
4194                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4195                         return;
4196                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4197
4198                 write_lock_bh(&ifp->idev->lock);
4199                 spin_lock(&ifp->lock);
4200                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4201                         ifp->idev->cnf.rtr_solicit_interval);
4202                 ifp->idev->rs_probes = 1;
4203                 ifp->idev->if_flags |= IF_RS_SENT;
4204                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4205                 spin_unlock(&ifp->lock);
4206                 write_unlock_bh(&ifp->idev->lock);
4207         }
4208
4209         if (bump_id)
4210                 rt_genid_bump_ipv6(dev_net(dev));
4211
4212         /* Make sure that a new temporary address will be created
4213          * before this temporary address becomes deprecated.
4214          */
4215         if (ifp->flags & IFA_F_TEMPORARY)
4216                 addrconf_verify_rtnl();
4217 }
4218
4219 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4220 {
4221         struct inet6_ifaddr *ifp;
4222
4223         read_lock_bh(&idev->lock);
4224         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4225                 spin_lock(&ifp->lock);
4226                 if ((ifp->flags & IFA_F_TENTATIVE &&
4227                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4228                         if (restart)
4229                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4230                         addrconf_dad_kick(ifp);
4231                 }
4232                 spin_unlock(&ifp->lock);
4233         }
4234         read_unlock_bh(&idev->lock);
4235 }
4236
4237 #ifdef CONFIG_PROC_FS
4238 struct if6_iter_state {
4239         struct seq_net_private p;
4240         int bucket;
4241         int offset;
4242 };
4243
4244 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4245 {
4246         struct if6_iter_state *state = seq->private;
4247         struct net *net = seq_file_net(seq);
4248         struct inet6_ifaddr *ifa = NULL;
4249         int p = 0;
4250
4251         /* initial bucket if pos is 0 */
4252         if (pos == 0) {
4253                 state->bucket = 0;
4254                 state->offset = 0;
4255         }
4256
4257         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4258                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4259                                          addr_lst) {
4260                         if (!net_eq(dev_net(ifa->idev->dev), net))
4261                                 continue;
4262                         /* sync with offset */
4263                         if (p < state->offset) {
4264                                 p++;
4265                                 continue;
4266                         }
4267                         return ifa;
4268                 }
4269
4270                 /* prepare for next bucket */
4271                 state->offset = 0;
4272                 p = 0;
4273         }
4274         return NULL;
4275 }
4276
4277 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4278                                          struct inet6_ifaddr *ifa)
4279 {
4280         struct if6_iter_state *state = seq->private;
4281         struct net *net = seq_file_net(seq);
4282
4283         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4284                 if (!net_eq(dev_net(ifa->idev->dev), net))
4285                         continue;
4286                 state->offset++;
4287                 return ifa;
4288         }
4289
4290         state->offset = 0;
4291         while (++state->bucket < IN6_ADDR_HSIZE) {
4292                 hlist_for_each_entry_rcu(ifa,
4293                                      &inet6_addr_lst[state->bucket], addr_lst) {
4294                         if (!net_eq(dev_net(ifa->idev->dev), net))
4295                                 continue;
4296                         return ifa;
4297                 }
4298         }
4299
4300         return NULL;
4301 }
4302
4303 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4304         __acquires(rcu)
4305 {
4306         rcu_read_lock();
4307         return if6_get_first(seq, *pos);
4308 }
4309
4310 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4311 {
4312         struct inet6_ifaddr *ifa;
4313
4314         ifa = if6_get_next(seq, v);
4315         ++*pos;
4316         return ifa;
4317 }
4318
4319 static void if6_seq_stop(struct seq_file *seq, void *v)
4320         __releases(rcu)
4321 {
4322         rcu_read_unlock();
4323 }
4324
4325 static int if6_seq_show(struct seq_file *seq, void *v)
4326 {
4327         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4328         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4329                    &ifp->addr,
4330                    ifp->idev->dev->ifindex,
4331                    ifp->prefix_len,
4332                    ifp->scope,
4333                    (u8) ifp->flags,
4334                    ifp->idev->dev->name);
4335         return 0;
4336 }
4337
4338 static const struct seq_operations if6_seq_ops = {
4339         .start  = if6_seq_start,
4340         .next   = if6_seq_next,
4341         .show   = if6_seq_show,
4342         .stop   = if6_seq_stop,
4343 };
4344
4345 static int __net_init if6_proc_net_init(struct net *net)
4346 {
4347         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4348                         sizeof(struct if6_iter_state)))
4349                 return -ENOMEM;
4350         return 0;
4351 }
4352
4353 static void __net_exit if6_proc_net_exit(struct net *net)
4354 {
4355         remove_proc_entry("if_inet6", net->proc_net);
4356 }
4357
4358 static struct pernet_operations if6_proc_net_ops = {
4359         .init = if6_proc_net_init,
4360         .exit = if6_proc_net_exit,
4361 };
4362
4363 int __init if6_proc_init(void)
4364 {
4365         return register_pernet_subsys(&if6_proc_net_ops);
4366 }
4367
4368 void if6_proc_exit(void)
4369 {
4370         unregister_pernet_subsys(&if6_proc_net_ops);
4371 }
4372 #endif  /* CONFIG_PROC_FS */
4373
4374 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4375 /* Check if address is a home address configured on any interface. */
4376 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4377 {
4378         unsigned int hash = inet6_addr_hash(net, addr);
4379         struct inet6_ifaddr *ifp = NULL;
4380         int ret = 0;
4381
4382         rcu_read_lock();
4383         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4384                 if (!net_eq(dev_net(ifp->idev->dev), net))
4385                         continue;
4386                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4387                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4388                         ret = 1;
4389                         break;
4390                 }
4391         }
4392         rcu_read_unlock();
4393         return ret;
4394 }
4395 #endif
4396
4397 /*
4398  *      Periodic address status verification
4399  */
4400
4401 static void addrconf_verify_rtnl(void)
4402 {
4403         unsigned long now, next, next_sec, next_sched;
4404         struct inet6_ifaddr *ifp;
4405         int i;
4406
4407         ASSERT_RTNL();
4408
4409         rcu_read_lock_bh();
4410         now = jiffies;
4411         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4412
4413         cancel_delayed_work(&addr_chk_work);
4414
4415         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4416 restart:
4417                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4418                         unsigned long age;
4419
4420                         /* When setting preferred_lft to a value not zero or
4421                          * infinity, while valid_lft is infinity
4422                          * IFA_F_PERMANENT has a non-infinity life time.
4423                          */
4424                         if ((ifp->flags & IFA_F_PERMANENT) &&
4425                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4426                                 continue;
4427
4428                         spin_lock(&ifp->lock);
4429                         /* We try to batch several events at once. */
4430                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4431
4432                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4433                             age >= ifp->valid_lft) {
4434                                 spin_unlock(&ifp->lock);
4435                                 in6_ifa_hold(ifp);
4436                                 ipv6_del_addr(ifp);
4437                                 goto restart;
4438                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4439                                 spin_unlock(&ifp->lock);
4440                                 continue;
4441                         } else if (age >= ifp->prefered_lft) {
4442                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4443                                 int deprecate = 0;
4444
4445                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4446                                         deprecate = 1;
4447                                         ifp->flags |= IFA_F_DEPRECATED;
4448                                 }
4449
4450                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4451                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4452                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4453
4454                                 spin_unlock(&ifp->lock);
4455
4456                                 if (deprecate) {
4457                                         in6_ifa_hold(ifp);
4458
4459                                         ipv6_ifa_notify(0, ifp);
4460                                         in6_ifa_put(ifp);
4461                                         goto restart;
4462                                 }
4463                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4464                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4465                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4466                                         ifp->idev->cnf.dad_transmits *
4467                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4468
4469                                 if (age >= ifp->prefered_lft - regen_advance) {
4470                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4471                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4472                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4473                                         if (!ifp->regen_count && ifpub) {
4474                                                 ifp->regen_count++;
4475                                                 in6_ifa_hold(ifp);
4476                                                 in6_ifa_hold(ifpub);
4477                                                 spin_unlock(&ifp->lock);
4478
4479                                                 spin_lock(&ifpub->lock);
4480                                                 ifpub->regen_count = 0;
4481                                                 spin_unlock(&ifpub->lock);
4482                                                 rcu_read_unlock_bh();
4483                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4484                                                 in6_ifa_put(ifpub);
4485                                                 in6_ifa_put(ifp);
4486                                                 rcu_read_lock_bh();
4487                                                 goto restart;
4488                                         }
4489                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4490                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4491                                 spin_unlock(&ifp->lock);
4492                         } else {
4493                                 /* ifp->prefered_lft <= ifp->valid_lft */
4494                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4495                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4496                                 spin_unlock(&ifp->lock);
4497                         }
4498                 }
4499         }
4500
4501         next_sec = round_jiffies_up(next);
4502         next_sched = next;
4503
4504         /* If rounded timeout is accurate enough, accept it. */
4505         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4506                 next_sched = next_sec;
4507
4508         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4509         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4510                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4511
4512         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4513                  now, next, next_sec, next_sched);
4514         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4515         rcu_read_unlock_bh();
4516 }
4517
4518 static void addrconf_verify_work(struct work_struct *w)
4519 {
4520         rtnl_lock();
4521         addrconf_verify_rtnl();
4522         rtnl_unlock();
4523 }
4524
4525 static void addrconf_verify(void)
4526 {
4527         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4528 }
4529
4530 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4531                                      struct in6_addr **peer_pfx)
4532 {
4533         struct in6_addr *pfx = NULL;
4534
4535         *peer_pfx = NULL;
4536
4537         if (addr)
4538                 pfx = nla_data(addr);
4539
4540         if (local) {
4541                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4542                         *peer_pfx = pfx;
4543                 pfx = nla_data(local);
4544         }
4545
4546         return pfx;
4547 }
4548
4549 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4550         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4551         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4552         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4553         [IFA_FLAGS]             = { .len = sizeof(u32) },
4554         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4555         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4556 };
4557
4558 static int
4559 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4560                   struct netlink_ext_ack *extack)
4561 {
4562         struct net *net = sock_net(skb->sk);
4563         struct ifaddrmsg *ifm;
4564         struct nlattr *tb[IFA_MAX+1];
4565         struct in6_addr *pfx, *peer_pfx;
4566         u32 ifa_flags;
4567         int err;
4568
4569         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4570                                      ifa_ipv6_policy, extack);
4571         if (err < 0)
4572                 return err;
4573
4574         ifm = nlmsg_data(nlh);
4575         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4576         if (!pfx)
4577                 return -EINVAL;
4578
4579         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4580
4581         /* We ignore other flags so far. */
4582         ifa_flags &= IFA_F_MANAGETEMPADDR;
4583
4584         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4585                               ifm->ifa_prefixlen);
4586 }
4587
4588 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4589                                unsigned long expires, u32 flags)
4590 {
4591         struct fib6_info *f6i;
4592         u32 prio;
4593
4594         f6i = addrconf_get_prefix_route(&ifp->addr, ifp->prefix_len,
4595                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4596         if (!f6i)
4597                 return -ENOENT;
4598
4599         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4600         if (f6i->fib6_metric != prio) {
4601                 /* delete old one */
4602                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4603
4604                 /* add new one */
4605                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4606                                       ifp->rt_priority, ifp->idev->dev,
4607                                       expires, flags, GFP_KERNEL);
4608         } else {
4609                 if (!expires)
4610                         fib6_clean_expires(f6i);
4611                 else
4612                         fib6_set_expires(f6i, expires);
4613
4614                 fib6_info_release(f6i);
4615         }
4616
4617         return 0;
4618 }
4619
4620 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4621 {
4622         u32 flags;
4623         clock_t expires;
4624         unsigned long timeout;
4625         bool was_managetempaddr;
4626         bool had_prefixroute;
4627
4628         ASSERT_RTNL();
4629
4630         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4631                 return -EINVAL;
4632
4633         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4634             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4635                 return -EINVAL;
4636
4637         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4638                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4639
4640         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4641         if (addrconf_finite_timeout(timeout)) {
4642                 expires = jiffies_to_clock_t(timeout * HZ);
4643                 cfg->valid_lft = timeout;
4644                 flags = RTF_EXPIRES;
4645         } else {
4646                 expires = 0;
4647                 flags = 0;
4648                 cfg->ifa_flags |= IFA_F_PERMANENT;
4649         }
4650
4651         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4652         if (addrconf_finite_timeout(timeout)) {
4653                 if (timeout == 0)
4654                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4655                 cfg->preferred_lft = timeout;
4656         }
4657
4658         spin_lock_bh(&ifp->lock);
4659         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4660         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4661                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4662         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4663                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4664                         IFA_F_NOPREFIXROUTE);
4665         ifp->flags |= cfg->ifa_flags;
4666         ifp->tstamp = jiffies;
4667         ifp->valid_lft = cfg->valid_lft;
4668         ifp->prefered_lft = cfg->preferred_lft;
4669
4670         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4671                 ifp->rt_priority = cfg->rt_priority;
4672
4673         spin_unlock_bh(&ifp->lock);
4674         if (!(ifp->flags&IFA_F_TENTATIVE))
4675                 ipv6_ifa_notify(0, ifp);
4676
4677         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4678                 int rc = -ENOENT;
4679
4680                 if (had_prefixroute)
4681                         rc = modify_prefix_route(ifp, expires, flags);
4682
4683                 /* prefix route could have been deleted; if so restore it */
4684                 if (rc == -ENOENT) {
4685                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4686                                               ifp->rt_priority, ifp->idev->dev,
4687                                               expires, flags, GFP_KERNEL);
4688                 }
4689         } else if (had_prefixroute) {
4690                 enum cleanup_prefix_rt_t action;
4691                 unsigned long rt_expires;
4692
4693                 write_lock_bh(&ifp->idev->lock);
4694                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4695                 write_unlock_bh(&ifp->idev->lock);
4696
4697                 if (action != CLEANUP_PREFIX_RT_NOP) {
4698                         cleanup_prefix_route(ifp, rt_expires,
4699                                 action == CLEANUP_PREFIX_RT_DEL);
4700                 }
4701         }
4702
4703         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4704                 if (was_managetempaddr &&
4705                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4706                         cfg->valid_lft = 0;
4707                         cfg->preferred_lft = 0;
4708                 }
4709                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4710                                  cfg->preferred_lft, !was_managetempaddr,
4711                                  jiffies);
4712         }
4713
4714         addrconf_verify_rtnl();
4715
4716         return 0;
4717 }
4718
4719 static int
4720 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4721                   struct netlink_ext_ack *extack)
4722 {
4723         struct net *net = sock_net(skb->sk);
4724         struct ifaddrmsg *ifm;
4725         struct nlattr *tb[IFA_MAX+1];
4726         struct in6_addr *peer_pfx;
4727         struct inet6_ifaddr *ifa;
4728         struct net_device *dev;
4729         struct inet6_dev *idev;
4730         struct ifa6_config cfg;
4731         int err;
4732
4733         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4734                                      ifa_ipv6_policy, extack);
4735         if (err < 0)
4736                 return err;
4737
4738         memset(&cfg, 0, sizeof(cfg));
4739
4740         ifm = nlmsg_data(nlh);
4741         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4742         if (!cfg.pfx)
4743                 return -EINVAL;
4744
4745         cfg.peer_pfx = peer_pfx;
4746         cfg.plen = ifm->ifa_prefixlen;
4747         if (tb[IFA_RT_PRIORITY])
4748                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4749
4750         cfg.valid_lft = INFINITY_LIFE_TIME;
4751         cfg.preferred_lft = INFINITY_LIFE_TIME;
4752
4753         if (tb[IFA_CACHEINFO]) {
4754                 struct ifa_cacheinfo *ci;
4755
4756                 ci = nla_data(tb[IFA_CACHEINFO]);
4757                 cfg.valid_lft = ci->ifa_valid;
4758                 cfg.preferred_lft = ci->ifa_prefered;
4759         }
4760
4761         dev =  __dev_get_by_index(net, ifm->ifa_index);
4762         if (!dev)
4763                 return -ENODEV;
4764
4765         if (tb[IFA_FLAGS])
4766                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4767         else
4768                 cfg.ifa_flags = ifm->ifa_flags;
4769
4770         /* We ignore other flags so far. */
4771         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4772                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4773                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4774
4775         idev = ipv6_find_idev(dev);
4776         if (IS_ERR(idev))
4777                 return PTR_ERR(idev);
4778
4779         if (!ipv6_allow_optimistic_dad(net, idev))
4780                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4781
4782         if (cfg.ifa_flags & IFA_F_NODAD &&
4783             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4784                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4785                 return -EINVAL;
4786         }
4787
4788         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4789         if (!ifa) {
4790                 /*
4791                  * It would be best to check for !NLM_F_CREATE here but
4792                  * userspace already relies on not having to provide this.
4793                  */
4794                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4795         }
4796
4797         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4798             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4799                 err = -EEXIST;
4800         else
4801                 err = inet6_addr_modify(ifa, &cfg);
4802
4803         in6_ifa_put(ifa);
4804
4805         return err;
4806 }
4807
4808 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4809                           u8 scope, int ifindex)
4810 {
4811         struct ifaddrmsg *ifm;
4812
4813         ifm = nlmsg_data(nlh);
4814         ifm->ifa_family = AF_INET6;
4815         ifm->ifa_prefixlen = prefixlen;
4816         ifm->ifa_flags = flags;
4817         ifm->ifa_scope = scope;
4818         ifm->ifa_index = ifindex;
4819 }
4820
4821 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4822                          unsigned long tstamp, u32 preferred, u32 valid)
4823 {
4824         struct ifa_cacheinfo ci;
4825
4826         ci.cstamp = cstamp_delta(cstamp);
4827         ci.tstamp = cstamp_delta(tstamp);
4828         ci.ifa_prefered = preferred;
4829         ci.ifa_valid = valid;
4830
4831         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4832 }
4833
4834 static inline int rt_scope(int ifa_scope)
4835 {
4836         if (ifa_scope & IFA_HOST)
4837                 return RT_SCOPE_HOST;
4838         else if (ifa_scope & IFA_LINK)
4839                 return RT_SCOPE_LINK;
4840         else if (ifa_scope & IFA_SITE)
4841                 return RT_SCOPE_SITE;
4842         else
4843                 return RT_SCOPE_UNIVERSE;
4844 }
4845
4846 static inline int inet6_ifaddr_msgsize(void)
4847 {
4848         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4849                + nla_total_size(16) /* IFA_LOCAL */
4850                + nla_total_size(16) /* IFA_ADDRESS */
4851                + nla_total_size(sizeof(struct ifa_cacheinfo))
4852                + nla_total_size(4)  /* IFA_FLAGS */
4853                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4854 }
4855
4856 enum addr_type_t {
4857         UNICAST_ADDR,
4858         MULTICAST_ADDR,
4859         ANYCAST_ADDR,
4860 };
4861
4862 struct inet6_fill_args {
4863         u32 portid;
4864         u32 seq;
4865         int event;
4866         unsigned int flags;
4867         int netnsid;
4868         int ifindex;
4869         enum addr_type_t type;
4870 };
4871
4872 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4873                              struct inet6_fill_args *args)
4874 {
4875         struct nlmsghdr  *nlh;
4876         u32 preferred, valid;
4877
4878         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4879                         sizeof(struct ifaddrmsg), args->flags);
4880         if (!nlh)
4881                 return -EMSGSIZE;
4882
4883         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4884                       ifa->idev->dev->ifindex);
4885
4886         if (args->netnsid >= 0 &&
4887             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4888                 goto error;
4889
4890         if (!((ifa->flags&IFA_F_PERMANENT) &&
4891               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4892                 preferred = ifa->prefered_lft;
4893                 valid = ifa->valid_lft;
4894                 if (preferred != INFINITY_LIFE_TIME) {
4895                         long tval = (jiffies - ifa->tstamp)/HZ;
4896                         if (preferred > tval)
4897                                 preferred -= tval;
4898                         else
4899                                 preferred = 0;
4900                         if (valid != INFINITY_LIFE_TIME) {
4901                                 if (valid > tval)
4902                                         valid -= tval;
4903                                 else
4904                                         valid = 0;
4905                         }
4906                 }
4907         } else {
4908                 preferred = INFINITY_LIFE_TIME;
4909                 valid = INFINITY_LIFE_TIME;
4910         }
4911
4912         if (!ipv6_addr_any(&ifa->peer_addr)) {
4913                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4914                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4915                         goto error;
4916         } else
4917                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4918                         goto error;
4919
4920         if (ifa->rt_priority &&
4921             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4922                 goto error;
4923
4924         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4925                 goto error;
4926
4927         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4928                 goto error;
4929
4930         nlmsg_end(skb, nlh);
4931         return 0;
4932
4933 error:
4934         nlmsg_cancel(skb, nlh);
4935         return -EMSGSIZE;
4936 }
4937
4938 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4939                                struct inet6_fill_args *args)
4940 {
4941         struct nlmsghdr  *nlh;
4942         u8 scope = RT_SCOPE_UNIVERSE;
4943         int ifindex = ifmca->idev->dev->ifindex;
4944
4945         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4946                 scope = RT_SCOPE_SITE;
4947
4948         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4949                         sizeof(struct ifaddrmsg), args->flags);
4950         if (!nlh)
4951                 return -EMSGSIZE;
4952
4953         if (args->netnsid >= 0 &&
4954             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4955                 return -EMSGSIZE;
4956
4957         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4958         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4959             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4960                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4961                 nlmsg_cancel(skb, nlh);
4962                 return -EMSGSIZE;
4963         }
4964
4965         nlmsg_end(skb, nlh);
4966         return 0;
4967 }
4968
4969 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4970                                struct inet6_fill_args *args)
4971 {
4972         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4973         int ifindex = dev ? dev->ifindex : 1;
4974         struct nlmsghdr  *nlh;
4975         u8 scope = RT_SCOPE_UNIVERSE;
4976
4977         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4978                 scope = RT_SCOPE_SITE;
4979
4980         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4981                         sizeof(struct ifaddrmsg), args->flags);
4982         if (!nlh)
4983                 return -EMSGSIZE;
4984
4985         if (args->netnsid >= 0 &&
4986             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4987                 return -EMSGSIZE;
4988
4989         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4990         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4991             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4992                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4993                 nlmsg_cancel(skb, nlh);
4994                 return -EMSGSIZE;
4995         }
4996
4997         nlmsg_end(skb, nlh);
4998         return 0;
4999 }
5000
5001 /* called with rcu_read_lock() */
5002 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5003                           struct netlink_callback *cb, int s_ip_idx,
5004                           struct inet6_fill_args *fillargs)
5005 {
5006         struct ifmcaddr6 *ifmca;
5007         struct ifacaddr6 *ifaca;
5008         int ip_idx = 0;
5009         int err = 1;
5010
5011         read_lock_bh(&idev->lock);
5012         switch (fillargs->type) {
5013         case UNICAST_ADDR: {
5014                 struct inet6_ifaddr *ifa;
5015                 fillargs->event = RTM_NEWADDR;
5016
5017                 /* unicast address incl. temp addr */
5018                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5019                         if (ip_idx < s_ip_idx)
5020                                 goto next;
5021                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5022                         if (err < 0)
5023                                 break;
5024                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5025 next:
5026                         ip_idx++;
5027                 }
5028                 break;
5029         }
5030         case MULTICAST_ADDR:
5031                 fillargs->event = RTM_GETMULTICAST;
5032
5033                 /* multicast address */
5034                 for (ifmca = idev->mc_list; ifmca;
5035                      ifmca = ifmca->next, ip_idx++) {
5036                         if (ip_idx < s_ip_idx)
5037                                 continue;
5038                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5039                         if (err < 0)
5040                                 break;
5041                 }
5042                 break;
5043         case ANYCAST_ADDR:
5044                 fillargs->event = RTM_GETANYCAST;
5045                 /* anycast address */
5046                 for (ifaca = idev->ac_list; ifaca;
5047                      ifaca = ifaca->aca_next, ip_idx++) {
5048                         if (ip_idx < s_ip_idx)
5049                                 continue;
5050                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5051                         if (err < 0)
5052                                 break;
5053                 }
5054                 break;
5055         default:
5056                 break;
5057         }
5058         read_unlock_bh(&idev->lock);
5059         cb->args[2] = ip_idx;
5060         return err;
5061 }
5062
5063 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5064                                        struct inet6_fill_args *fillargs,
5065                                        struct net **tgt_net, struct sock *sk,
5066                                        struct netlink_callback *cb)
5067 {
5068         struct netlink_ext_ack *extack = cb->extack;
5069         struct nlattr *tb[IFA_MAX+1];
5070         struct ifaddrmsg *ifm;
5071         int err, i;
5072
5073         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5074                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5075                 return -EINVAL;
5076         }
5077
5078         ifm = nlmsg_data(nlh);
5079         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5080                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5081                 return -EINVAL;
5082         }
5083
5084         fillargs->ifindex = ifm->ifa_index;
5085         if (fillargs->ifindex) {
5086                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5087                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5088         }
5089
5090         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5091                                             ifa_ipv6_policy, extack);
5092         if (err < 0)
5093                 return err;
5094
5095         for (i = 0; i <= IFA_MAX; ++i) {
5096                 if (!tb[i])
5097                         continue;
5098
5099                 if (i == IFA_TARGET_NETNSID) {
5100                         struct net *net;
5101
5102                         fillargs->netnsid = nla_get_s32(tb[i]);
5103                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5104                         if (IS_ERR(net)) {
5105                                 fillargs->netnsid = -1;
5106                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5107                                 return PTR_ERR(net);
5108                         }
5109                         *tgt_net = net;
5110                 } else {
5111                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5112                         return -EINVAL;
5113                 }
5114         }
5115
5116         return 0;
5117 }
5118
5119 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5120                            enum addr_type_t type)
5121 {
5122         const struct nlmsghdr *nlh = cb->nlh;
5123         struct inet6_fill_args fillargs = {
5124                 .portid = NETLINK_CB(cb->skb).portid,
5125                 .seq = cb->nlh->nlmsg_seq,
5126                 .flags = NLM_F_MULTI,
5127                 .netnsid = -1,
5128                 .type = type,
5129         };
5130         struct net *net = sock_net(skb->sk);
5131         struct net *tgt_net = net;
5132         int idx, s_idx, s_ip_idx;
5133         int h, s_h;
5134         struct net_device *dev;
5135         struct inet6_dev *idev;
5136         struct hlist_head *head;
5137         int err = 0;
5138
5139         s_h = cb->args[0];
5140         s_idx = idx = cb->args[1];
5141         s_ip_idx = cb->args[2];
5142
5143         if (cb->strict_check) {
5144                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5145                                                   skb->sk, cb);
5146                 if (err < 0)
5147                         goto put_tgt_net;
5148
5149                 err = 0;
5150                 if (fillargs.ifindex) {
5151                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5152                         if (!dev) {
5153                                 err = -ENODEV;
5154                                 goto put_tgt_net;
5155                         }
5156                         idev = __in6_dev_get(dev);
5157                         if (idev) {
5158                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5159                                                      &fillargs);
5160                                 if (err > 0)
5161                                         err = 0;
5162                         }
5163                         goto put_tgt_net;
5164                 }
5165         }
5166
5167         rcu_read_lock();
5168         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5169         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5170                 idx = 0;
5171                 head = &tgt_net->dev_index_head[h];
5172                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5173                         if (idx < s_idx)
5174                                 goto cont;
5175                         if (h > s_h || idx > s_idx)
5176                                 s_ip_idx = 0;
5177                         idev = __in6_dev_get(dev);
5178                         if (!idev)
5179                                 goto cont;
5180
5181                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5182                                            &fillargs) < 0)
5183                                 goto done;
5184 cont:
5185                         idx++;
5186                 }
5187         }
5188 done:
5189         rcu_read_unlock();
5190         cb->args[0] = h;
5191         cb->args[1] = idx;
5192 put_tgt_net:
5193         if (fillargs.netnsid >= 0)
5194                 put_net(tgt_net);
5195
5196         return skb->len ? : err;
5197 }
5198
5199 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5200 {
5201         enum addr_type_t type = UNICAST_ADDR;
5202
5203         return inet6_dump_addr(skb, cb, type);
5204 }
5205
5206 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5207 {
5208         enum addr_type_t type = MULTICAST_ADDR;
5209
5210         return inet6_dump_addr(skb, cb, type);
5211 }
5212
5213
5214 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5215 {
5216         enum addr_type_t type = ANYCAST_ADDR;
5217
5218         return inet6_dump_addr(skb, cb, type);
5219 }
5220
5221 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5222                                        const struct nlmsghdr *nlh,
5223                                        struct nlattr **tb,
5224                                        struct netlink_ext_ack *extack)
5225 {
5226         struct ifaddrmsg *ifm;
5227         int i, err;
5228
5229         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5230                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5231                 return -EINVAL;
5232         }
5233
5234         ifm = nlmsg_data(nlh);
5235         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5236                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5237                 return -EINVAL;
5238         }
5239
5240         if (!netlink_strict_get_check(skb))
5241                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5242                                               ifa_ipv6_policy, extack);
5243
5244         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5245                                             ifa_ipv6_policy, extack);
5246         if (err)
5247                 return err;
5248
5249         for (i = 0; i <= IFA_MAX; i++) {
5250                 if (!tb[i])
5251                         continue;
5252
5253                 switch (i) {
5254                 case IFA_TARGET_NETNSID:
5255                 case IFA_ADDRESS:
5256                 case IFA_LOCAL:
5257                         break;
5258                 default:
5259                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5260                         return -EINVAL;
5261                 }
5262         }
5263
5264         return 0;
5265 }
5266
5267 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5268                              struct netlink_ext_ack *extack)
5269 {
5270         struct net *net = sock_net(in_skb->sk);
5271         struct inet6_fill_args fillargs = {
5272                 .portid = NETLINK_CB(in_skb).portid,
5273                 .seq = nlh->nlmsg_seq,
5274                 .event = RTM_NEWADDR,
5275                 .flags = 0,
5276                 .netnsid = -1,
5277         };
5278         struct net *tgt_net = net;
5279         struct ifaddrmsg *ifm;
5280         struct nlattr *tb[IFA_MAX+1];
5281         struct in6_addr *addr = NULL, *peer;
5282         struct net_device *dev = NULL;
5283         struct inet6_ifaddr *ifa;
5284         struct sk_buff *skb;
5285         int err;
5286
5287         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5288         if (err < 0)
5289                 return err;
5290
5291         if (tb[IFA_TARGET_NETNSID]) {
5292                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5293
5294                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5295                                                   fillargs.netnsid);
5296                 if (IS_ERR(tgt_net))
5297                         return PTR_ERR(tgt_net);
5298         }
5299
5300         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5301         if (!addr)
5302                 return -EINVAL;
5303
5304         ifm = nlmsg_data(nlh);
5305         if (ifm->ifa_index)
5306                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5307
5308         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5309         if (!ifa) {
5310                 err = -EADDRNOTAVAIL;
5311                 goto errout;
5312         }
5313
5314         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5315         if (!skb) {
5316                 err = -ENOBUFS;
5317                 goto errout_ifa;
5318         }
5319
5320         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5321         if (err < 0) {
5322                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5323                 WARN_ON(err == -EMSGSIZE);
5324                 kfree_skb(skb);
5325                 goto errout_ifa;
5326         }
5327         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5328 errout_ifa:
5329         in6_ifa_put(ifa);
5330 errout:
5331         if (dev)
5332                 dev_put(dev);
5333         if (fillargs.netnsid >= 0)
5334                 put_net(tgt_net);
5335
5336         return err;
5337 }
5338
5339 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5340 {
5341         struct sk_buff *skb;
5342         struct net *net = dev_net(ifa->idev->dev);
5343         struct inet6_fill_args fillargs = {
5344                 .portid = 0,
5345                 .seq = 0,
5346                 .event = event,
5347                 .flags = 0,
5348                 .netnsid = -1,
5349         };
5350         int err = -ENOBUFS;
5351
5352         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5353         if (!skb)
5354                 goto errout;
5355
5356         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5357         if (err < 0) {
5358                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5359                 WARN_ON(err == -EMSGSIZE);
5360                 kfree_skb(skb);
5361                 goto errout;
5362         }
5363         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5364         return;
5365 errout:
5366         if (err < 0)
5367                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5368 }
5369
5370 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5371                                 __s32 *array, int bytes)
5372 {
5373         BUG_ON(bytes < (DEVCONF_MAX * 4));
5374
5375         memset(array, 0, bytes);
5376         array[DEVCONF_FORWARDING] = cnf->forwarding;
5377         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5378         array[DEVCONF_MTU6] = cnf->mtu6;
5379         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5380         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5381         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5382         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5383         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5384         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5385                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5386         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5387                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5388         array[DEVCONF_RTR_SOLICIT_DELAY] =
5389                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5390         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5391         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5392                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5393         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5394                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5395         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5396         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5397         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5398         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5399         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5400         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5401         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5402         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5403         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5404 #ifdef CONFIG_IPV6_ROUTER_PREF
5405         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5406         array[DEVCONF_RTR_PROBE_INTERVAL] =
5407                 jiffies_to_msecs(cnf->rtr_probe_interval);
5408 #ifdef CONFIG_IPV6_ROUTE_INFO
5409         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5410         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5411 #endif
5412 #endif
5413         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5414         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5415 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5416         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5417         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5418 #endif
5419 #ifdef CONFIG_IPV6_MROUTE
5420         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5421 #endif
5422         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5423         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5424         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5425         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5426         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5427         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5428         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5429         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5430         /* we omit DEVCONF_STABLE_SECRET for now */
5431         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5432         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5433         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5434         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5435         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5436 #ifdef CONFIG_IPV6_SEG6_HMAC
5437         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5438 #endif
5439         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5440         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5441         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5442         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5443 }
5444
5445 static inline size_t inet6_ifla6_size(void)
5446 {
5447         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5448              + nla_total_size(sizeof(struct ifla_cacheinfo))
5449              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5450              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5451              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5452              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5453              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5454              + 0;
5455 }
5456
5457 static inline size_t inet6_if_nlmsg_size(void)
5458 {
5459         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5460                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5461                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5462                + nla_total_size(4) /* IFLA_MTU */
5463                + nla_total_size(4) /* IFLA_LINK */
5464                + nla_total_size(1) /* IFLA_OPERSTATE */
5465                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5466 }
5467
5468 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5469                                         int bytes)
5470 {
5471         int i;
5472         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5473         BUG_ON(pad < 0);
5474
5475         /* Use put_unaligned() because stats may not be aligned for u64. */
5476         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5477         for (i = 1; i < ICMP6_MIB_MAX; i++)
5478                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5479
5480         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5481 }
5482
5483 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5484                                         int bytes, size_t syncpoff)
5485 {
5486         int i, c;
5487         u64 buff[IPSTATS_MIB_MAX];
5488         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5489
5490         BUG_ON(pad < 0);
5491
5492         memset(buff, 0, sizeof(buff));
5493         buff[0] = IPSTATS_MIB_MAX;
5494
5495         for_each_possible_cpu(c) {
5496                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5497                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5498         }
5499
5500         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5501         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5502 }
5503
5504 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5505                              int bytes)
5506 {
5507         switch (attrtype) {
5508         case IFLA_INET6_STATS:
5509                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5510                                      offsetof(struct ipstats_mib, syncp));
5511                 break;
5512         case IFLA_INET6_ICMP6STATS:
5513                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5514                 break;
5515         }
5516 }
5517
5518 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5519                                   u32 ext_filter_mask)
5520 {
5521         struct nlattr *nla;
5522         struct ifla_cacheinfo ci;
5523
5524         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5525                 goto nla_put_failure;
5526         ci.max_reasm_len = IPV6_MAXPLEN;
5527         ci.tstamp = cstamp_delta(idev->tstamp);
5528         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5529         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5530         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5531                 goto nla_put_failure;
5532         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5533         if (!nla)
5534                 goto nla_put_failure;
5535         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5536
5537         /* XXX - MC not implemented */
5538
5539         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5540                 return 0;
5541
5542         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5543         if (!nla)
5544                 goto nla_put_failure;
5545         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5546
5547         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5548         if (!nla)
5549                 goto nla_put_failure;
5550         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5551
5552         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5553         if (!nla)
5554                 goto nla_put_failure;
5555
5556         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5557                 goto nla_put_failure;
5558
5559         read_lock_bh(&idev->lock);
5560         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5561         read_unlock_bh(&idev->lock);
5562
5563         return 0;
5564
5565 nla_put_failure:
5566         return -EMSGSIZE;
5567 }
5568
5569 static size_t inet6_get_link_af_size(const struct net_device *dev,
5570                                      u32 ext_filter_mask)
5571 {
5572         if (!__in6_dev_get(dev))
5573                 return 0;
5574
5575         return inet6_ifla6_size();
5576 }
5577
5578 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5579                               u32 ext_filter_mask)
5580 {
5581         struct inet6_dev *idev = __in6_dev_get(dev);
5582
5583         if (!idev)
5584                 return -ENODATA;
5585
5586         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5587                 return -EMSGSIZE;
5588
5589         return 0;
5590 }
5591
5592 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5593 {
5594         struct inet6_ifaddr *ifp;
5595         struct net_device *dev = idev->dev;
5596         bool clear_token, update_rs = false;
5597         struct in6_addr ll_addr;
5598
5599         ASSERT_RTNL();
5600
5601         if (!token)
5602                 return -EINVAL;
5603         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5604                 return -EINVAL;
5605         if (!ipv6_accept_ra(idev))
5606                 return -EINVAL;
5607         if (idev->cnf.rtr_solicits == 0)
5608                 return -EINVAL;
5609
5610         write_lock_bh(&idev->lock);
5611
5612         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5613         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5614
5615         write_unlock_bh(&idev->lock);
5616
5617         clear_token = ipv6_addr_any(token);
5618         if (clear_token)
5619                 goto update_lft;
5620
5621         if (!idev->dead && (idev->if_flags & IF_READY) &&
5622             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5623                              IFA_F_OPTIMISTIC)) {
5624                 /* If we're not ready, then normal ifup will take care
5625                  * of this. Otherwise, we need to request our rs here.
5626                  */
5627                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5628                 update_rs = true;
5629         }
5630
5631 update_lft:
5632         write_lock_bh(&idev->lock);
5633
5634         if (update_rs) {
5635                 idev->if_flags |= IF_RS_SENT;
5636                 idev->rs_interval = rfc3315_s14_backoff_init(
5637                         idev->cnf.rtr_solicit_interval);
5638                 idev->rs_probes = 1;
5639                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5640         }
5641
5642         /* Well, that's kinda nasty ... */
5643         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5644                 spin_lock(&ifp->lock);
5645                 if (ifp->tokenized) {
5646                         ifp->valid_lft = 0;
5647                         ifp->prefered_lft = 0;
5648                 }
5649                 spin_unlock(&ifp->lock);
5650         }
5651
5652         write_unlock_bh(&idev->lock);
5653         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5654         addrconf_verify_rtnl();
5655         return 0;
5656 }
5657
5658 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5659         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5660         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5661 };
5662
5663 static int check_addr_gen_mode(int mode)
5664 {
5665         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5666             mode != IN6_ADDR_GEN_MODE_NONE &&
5667             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5668             mode != IN6_ADDR_GEN_MODE_RANDOM)
5669                 return -EINVAL;
5670         return 1;
5671 }
5672
5673 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5674                                 int mode)
5675 {
5676         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5677             !idev->cnf.stable_secret.initialized &&
5678             !net->ipv6.devconf_dflt->stable_secret.initialized)
5679                 return -EINVAL;
5680         return 1;
5681 }
5682
5683 static int inet6_validate_link_af(const struct net_device *dev,
5684                                   const struct nlattr *nla)
5685 {
5686         struct nlattr *tb[IFLA_INET6_MAX + 1];
5687         struct inet6_dev *idev = NULL;
5688         int err;
5689
5690         if (dev) {
5691                 idev = __in6_dev_get(dev);
5692                 if (!idev)
5693                         return -EAFNOSUPPORT;
5694         }
5695
5696         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5697                                           inet6_af_policy, NULL);
5698         if (err)
5699                 return err;
5700
5701         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5702                 return -EINVAL;
5703
5704         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5705                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5706
5707                 if (check_addr_gen_mode(mode) < 0)
5708                         return -EINVAL;
5709                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5710                         return -EINVAL;
5711         }
5712
5713         return 0;
5714 }
5715
5716 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5717 {
5718         struct inet6_dev *idev = __in6_dev_get(dev);
5719         struct nlattr *tb[IFLA_INET6_MAX + 1];
5720         int err;
5721
5722         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5723                 BUG();
5724
5725         if (tb[IFLA_INET6_TOKEN]) {
5726                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5727                 if (err)
5728                         return err;
5729         }
5730
5731         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5732                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5733
5734                 idev->cnf.addr_gen_mode = mode;
5735         }
5736
5737         return 0;
5738 }
5739
5740 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5741                              u32 portid, u32 seq, int event, unsigned int flags)
5742 {
5743         struct net_device *dev = idev->dev;
5744         struct ifinfomsg *hdr;
5745         struct nlmsghdr *nlh;
5746         void *protoinfo;
5747
5748         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5749         if (!nlh)
5750                 return -EMSGSIZE;
5751
5752         hdr = nlmsg_data(nlh);
5753         hdr->ifi_family = AF_INET6;
5754         hdr->__ifi_pad = 0;
5755         hdr->ifi_type = dev->type;
5756         hdr->ifi_index = dev->ifindex;
5757         hdr->ifi_flags = dev_get_flags(dev);
5758         hdr->ifi_change = 0;
5759
5760         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5761             (dev->addr_len &&
5762              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5763             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5764             (dev->ifindex != dev_get_iflink(dev) &&
5765              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5766             nla_put_u8(skb, IFLA_OPERSTATE,
5767                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5768                 goto nla_put_failure;
5769         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5770         if (!protoinfo)
5771                 goto nla_put_failure;
5772
5773         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5774                 goto nla_put_failure;
5775
5776         nla_nest_end(skb, protoinfo);
5777         nlmsg_end(skb, nlh);
5778         return 0;
5779
5780 nla_put_failure:
5781         nlmsg_cancel(skb, nlh);
5782         return -EMSGSIZE;
5783 }
5784
5785 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5786                                    struct netlink_ext_ack *extack)
5787 {
5788         struct ifinfomsg *ifm;
5789
5790         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5791                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5792                 return -EINVAL;
5793         }
5794
5795         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5796                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5797                 return -EINVAL;
5798         }
5799
5800         ifm = nlmsg_data(nlh);
5801         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5802             ifm->ifi_change || ifm->ifi_index) {
5803                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5804                 return -EINVAL;
5805         }
5806
5807         return 0;
5808 }
5809
5810 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5811 {
5812         struct net *net = sock_net(skb->sk);
5813         int h, s_h;
5814         int idx = 0, s_idx;
5815         struct net_device *dev;
5816         struct inet6_dev *idev;
5817         struct hlist_head *head;
5818
5819         /* only requests using strict checking can pass data to
5820          * influence the dump
5821          */
5822         if (cb->strict_check) {
5823                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5824
5825                 if (err < 0)
5826                         return err;
5827         }
5828
5829         s_h = cb->args[0];
5830         s_idx = cb->args[1];
5831
5832         rcu_read_lock();
5833         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5834                 idx = 0;
5835                 head = &net->dev_index_head[h];
5836                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5837                         if (idx < s_idx)
5838                                 goto cont;
5839                         idev = __in6_dev_get(dev);
5840                         if (!idev)
5841                                 goto cont;
5842                         if (inet6_fill_ifinfo(skb, idev,
5843                                               NETLINK_CB(cb->skb).portid,
5844                                               cb->nlh->nlmsg_seq,
5845                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5846                                 goto out;
5847 cont:
5848                         idx++;
5849                 }
5850         }
5851 out:
5852         rcu_read_unlock();
5853         cb->args[1] = idx;
5854         cb->args[0] = h;
5855
5856         return skb->len;
5857 }
5858
5859 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5860 {
5861         struct sk_buff *skb;
5862         struct net *net = dev_net(idev->dev);
5863         int err = -ENOBUFS;
5864
5865         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5866         if (!skb)
5867                 goto errout;
5868
5869         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5870         if (err < 0) {
5871                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5872                 WARN_ON(err == -EMSGSIZE);
5873                 kfree_skb(skb);
5874                 goto errout;
5875         }
5876         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5877         return;
5878 errout:
5879         if (err < 0)
5880                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5881 }
5882
5883 static inline size_t inet6_prefix_nlmsg_size(void)
5884 {
5885         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5886                + nla_total_size(sizeof(struct in6_addr))
5887                + nla_total_size(sizeof(struct prefix_cacheinfo));
5888 }
5889
5890 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5891                              struct prefix_info *pinfo, u32 portid, u32 seq,
5892                              int event, unsigned int flags)
5893 {
5894         struct prefixmsg *pmsg;
5895         struct nlmsghdr *nlh;
5896         struct prefix_cacheinfo ci;
5897
5898         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5899         if (!nlh)
5900                 return -EMSGSIZE;
5901
5902         pmsg = nlmsg_data(nlh);
5903         pmsg->prefix_family = AF_INET6;
5904         pmsg->prefix_pad1 = 0;
5905         pmsg->prefix_pad2 = 0;
5906         pmsg->prefix_ifindex = idev->dev->ifindex;
5907         pmsg->prefix_len = pinfo->prefix_len;
5908         pmsg->prefix_type = pinfo->type;
5909         pmsg->prefix_pad3 = 0;
5910         pmsg->prefix_flags = 0;
5911         if (pinfo->onlink)
5912                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5913         if (pinfo->autoconf)
5914                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5915
5916         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5917                 goto nla_put_failure;
5918         ci.preferred_time = ntohl(pinfo->prefered);
5919         ci.valid_time = ntohl(pinfo->valid);
5920         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5921                 goto nla_put_failure;
5922         nlmsg_end(skb, nlh);
5923         return 0;
5924
5925 nla_put_failure:
5926         nlmsg_cancel(skb, nlh);
5927         return -EMSGSIZE;
5928 }
5929
5930 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5931                          struct prefix_info *pinfo)
5932 {
5933         struct sk_buff *skb;
5934         struct net *net = dev_net(idev->dev);
5935         int err = -ENOBUFS;
5936
5937         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5938         if (!skb)
5939                 goto errout;
5940
5941         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5942         if (err < 0) {
5943                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5944                 WARN_ON(err == -EMSGSIZE);
5945                 kfree_skb(skb);
5946                 goto errout;
5947         }
5948         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5949         return;
5950 errout:
5951         if (err < 0)
5952                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5953 }
5954
5955 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5956 {
5957         struct net *net = dev_net(ifp->idev->dev);
5958
5959         if (event)
5960                 ASSERT_RTNL();
5961
5962         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5963
5964         switch (event) {
5965         case RTM_NEWADDR:
5966                 /*
5967                  * If the address was optimistic
5968                  * we inserted the route at the start of
5969                  * our DAD process, so we don't need
5970                  * to do it again
5971                  */
5972                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5973                         ip6_ins_rt(net, ifp->rt);
5974                 if (ifp->idev->cnf.forwarding)
5975                         addrconf_join_anycast(ifp);
5976                 if (!ipv6_addr_any(&ifp->peer_addr))
5977                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5978                                               ifp->idev->dev, 0, 0,
5979                                               GFP_ATOMIC);
5980                 break;
5981         case RTM_DELADDR:
5982                 if (ifp->idev->cnf.forwarding)
5983                         addrconf_leave_anycast(ifp);
5984                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5985                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5986                         struct fib6_info *rt;
5987
5988                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5989                                                        ifp->idev->dev, 0, 0,
5990                                                        false);
5991                         if (rt)
5992                                 ip6_del_rt(net, rt);
5993                 }
5994                 if (ifp->rt) {
5995                         ip6_del_rt(net, ifp->rt);
5996                         ifp->rt = NULL;
5997                 }
5998                 rt_genid_bump_ipv6(net);
5999                 break;
6000         }
6001         atomic_inc(&net->ipv6.dev_addr_genid);
6002 }
6003
6004 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6005 {
6006         rcu_read_lock_bh();
6007         if (likely(ifp->idev->dead == 0))
6008                 __ipv6_ifa_notify(event, ifp);
6009         rcu_read_unlock_bh();
6010 }
6011
6012 #ifdef CONFIG_SYSCTL
6013
6014 static
6015 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6016                            void __user *buffer, size_t *lenp, loff_t *ppos)
6017 {
6018         int *valp = ctl->data;
6019         int val = *valp;
6020         loff_t pos = *ppos;
6021         struct ctl_table lctl;
6022         int ret;
6023
6024         /*
6025          * ctl->data points to idev->cnf.forwarding, we should
6026          * not modify it until we get the rtnl lock.
6027          */
6028         lctl = *ctl;
6029         lctl.data = &val;
6030
6031         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6032
6033         if (write)
6034                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6035         if (ret)
6036                 *ppos = pos;
6037         return ret;
6038 }
6039
6040 static
6041 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6042                         void __user *buffer, size_t *lenp, loff_t *ppos)
6043 {
6044         struct inet6_dev *idev = ctl->extra1;
6045         int min_mtu = IPV6_MIN_MTU;
6046         struct ctl_table lctl;
6047
6048         lctl = *ctl;
6049         lctl.extra1 = &min_mtu;
6050         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6051
6052         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6053 }
6054
6055 static void dev_disable_change(struct inet6_dev *idev)
6056 {
6057         struct netdev_notifier_info info;
6058
6059         if (!idev || !idev->dev)
6060                 return;
6061
6062         netdev_notifier_info_init(&info, idev->dev);
6063         if (idev->cnf.disable_ipv6)
6064                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6065         else
6066                 addrconf_notify(NULL, NETDEV_UP, &info);
6067 }
6068
6069 static void addrconf_disable_change(struct net *net, __s32 newf)
6070 {
6071         struct net_device *dev;
6072         struct inet6_dev *idev;
6073
6074         for_each_netdev(net, dev) {
6075                 idev = __in6_dev_get(dev);
6076                 if (idev) {
6077                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6078                         idev->cnf.disable_ipv6 = newf;
6079                         if (changed)
6080                                 dev_disable_change(idev);
6081                 }
6082         }
6083 }
6084
6085 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6086 {
6087         struct net *net;
6088         int old;
6089
6090         if (!rtnl_trylock())
6091                 return restart_syscall();
6092
6093         net = (struct net *)table->extra2;
6094         old = *p;
6095         *p = newf;
6096
6097         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6098                 rtnl_unlock();
6099                 return 0;
6100         }
6101
6102         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6103                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6104                 addrconf_disable_change(net, newf);
6105         } else if ((!newf) ^ (!old))
6106                 dev_disable_change((struct inet6_dev *)table->extra1);
6107
6108         rtnl_unlock();
6109         return 0;
6110 }
6111
6112 static
6113 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6114                             void __user *buffer, size_t *lenp, loff_t *ppos)
6115 {
6116         int *valp = ctl->data;
6117         int val = *valp;
6118         loff_t pos = *ppos;
6119         struct ctl_table lctl;
6120         int ret;
6121
6122         /*
6123          * ctl->data points to idev->cnf.disable_ipv6, we should
6124          * not modify it until we get the rtnl lock.
6125          */
6126         lctl = *ctl;
6127         lctl.data = &val;
6128
6129         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6130
6131         if (write)
6132                 ret = addrconf_disable_ipv6(ctl, valp, val);
6133         if (ret)
6134                 *ppos = pos;
6135         return ret;
6136 }
6137
6138 static
6139 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6140                               void __user *buffer, size_t *lenp, loff_t *ppos)
6141 {
6142         int *valp = ctl->data;
6143         int ret;
6144         int old, new;
6145
6146         old = *valp;
6147         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6148         new = *valp;
6149
6150         if (write && old != new) {
6151                 struct net *net = ctl->extra2;
6152
6153                 if (!rtnl_trylock())
6154                         return restart_syscall();
6155
6156                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6157                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6158                                                      NETCONFA_PROXY_NEIGH,
6159                                                      NETCONFA_IFINDEX_DEFAULT,
6160                                                      net->ipv6.devconf_dflt);
6161                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6162                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6163                                                      NETCONFA_PROXY_NEIGH,
6164                                                      NETCONFA_IFINDEX_ALL,
6165                                                      net->ipv6.devconf_all);
6166                 else {
6167                         struct inet6_dev *idev = ctl->extra1;
6168
6169                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6170                                                      NETCONFA_PROXY_NEIGH,
6171                                                      idev->dev->ifindex,
6172                                                      &idev->cnf);
6173                 }
6174                 rtnl_unlock();
6175         }
6176
6177         return ret;
6178 }
6179
6180 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6181                                          void __user *buffer, size_t *lenp,
6182                                          loff_t *ppos)
6183 {
6184         int ret = 0;
6185         u32 new_val;
6186         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6187         struct net *net = (struct net *)ctl->extra2;
6188         struct ctl_table tmp = {
6189                 .data = &new_val,
6190                 .maxlen = sizeof(new_val),
6191                 .mode = ctl->mode,
6192         };
6193
6194         if (!rtnl_trylock())
6195                 return restart_syscall();
6196
6197         new_val = *((u32 *)ctl->data);
6198
6199         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6200         if (ret != 0)
6201                 goto out;
6202
6203         if (write) {
6204                 if (check_addr_gen_mode(new_val) < 0) {
6205                         ret = -EINVAL;
6206                         goto out;
6207                 }
6208
6209                 if (idev) {
6210                         if (check_stable_privacy(idev, net, new_val) < 0) {
6211                                 ret = -EINVAL;
6212                                 goto out;
6213                         }
6214
6215                         if (idev->cnf.addr_gen_mode != new_val) {
6216                                 idev->cnf.addr_gen_mode = new_val;
6217                                 addrconf_dev_config(idev->dev);
6218                         }
6219                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6220                         struct net_device *dev;
6221
6222                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6223                         for_each_netdev(net, dev) {
6224                                 idev = __in6_dev_get(dev);
6225                                 if (idev &&
6226                                     idev->cnf.addr_gen_mode != new_val) {
6227                                         idev->cnf.addr_gen_mode = new_val;
6228                                         addrconf_dev_config(idev->dev);
6229                                 }
6230                         }
6231                 }
6232
6233                 *((u32 *)ctl->data) = new_val;
6234         }
6235
6236 out:
6237         rtnl_unlock();
6238
6239         return ret;
6240 }
6241
6242 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6243                                          void __user *buffer, size_t *lenp,
6244                                          loff_t *ppos)
6245 {
6246         int err;
6247         struct in6_addr addr;
6248         char str[IPV6_MAX_STRLEN];
6249         struct ctl_table lctl = *ctl;
6250         struct net *net = ctl->extra2;
6251         struct ipv6_stable_secret *secret = ctl->data;
6252
6253         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6254                 return -EIO;
6255
6256         lctl.maxlen = IPV6_MAX_STRLEN;
6257         lctl.data = str;
6258
6259         if (!rtnl_trylock())
6260                 return restart_syscall();
6261
6262         if (!write && !secret->initialized) {
6263                 err = -EIO;
6264                 goto out;
6265         }
6266
6267         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6268         if (err >= sizeof(str)) {
6269                 err = -EIO;
6270                 goto out;
6271         }
6272
6273         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6274         if (err || !write)
6275                 goto out;
6276
6277         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6278                 err = -EIO;
6279                 goto out;
6280         }
6281
6282         secret->initialized = true;
6283         secret->secret = addr;
6284
6285         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6286                 struct net_device *dev;
6287
6288                 for_each_netdev(net, dev) {
6289                         struct inet6_dev *idev = __in6_dev_get(dev);
6290
6291                         if (idev) {
6292                                 idev->cnf.addr_gen_mode =
6293                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6294                         }
6295                 }
6296         } else {
6297                 struct inet6_dev *idev = ctl->extra1;
6298
6299                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6300         }
6301
6302 out:
6303         rtnl_unlock();
6304
6305         return err;
6306 }
6307
6308 static
6309 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6310                                                 int write,
6311                                                 void __user *buffer,
6312                                                 size_t *lenp,
6313                                                 loff_t *ppos)
6314 {
6315         int *valp = ctl->data;
6316         int val = *valp;
6317         loff_t pos = *ppos;
6318         struct ctl_table lctl;
6319         int ret;
6320
6321         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6322          * we should not modify it until we get the rtnl lock.
6323          */
6324         lctl = *ctl;
6325         lctl.data = &val;
6326
6327         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6328
6329         if (write)
6330                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6331         if (ret)
6332                 *ppos = pos;
6333         return ret;
6334 }
6335
6336 static
6337 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6338 {
6339         if (rt) {
6340                 if (action)
6341                         rt->dst.flags |= DST_NOPOLICY;
6342                 else
6343                         rt->dst.flags &= ~DST_NOPOLICY;
6344         }
6345 }
6346
6347 static
6348 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6349 {
6350         struct inet6_ifaddr *ifa;
6351
6352         read_lock_bh(&idev->lock);
6353         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6354                 spin_lock(&ifa->lock);
6355                 if (ifa->rt) {
6356                         /* host routes only use builtin fib6_nh */
6357                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6358                         int cpu;
6359
6360                         rcu_read_lock();
6361                         ifa->rt->dst_nopolicy = val ? true : false;
6362                         if (nh->rt6i_pcpu) {
6363                                 for_each_possible_cpu(cpu) {
6364                                         struct rt6_info **rtp;
6365
6366                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6367                                         addrconf_set_nopolicy(*rtp, val);
6368                                 }
6369                         }
6370                         rcu_read_unlock();
6371                 }
6372                 spin_unlock(&ifa->lock);
6373         }
6374         read_unlock_bh(&idev->lock);
6375 }
6376
6377 static
6378 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6379 {
6380         struct inet6_dev *idev;
6381         struct net *net;
6382
6383         if (!rtnl_trylock())
6384                 return restart_syscall();
6385
6386         *valp = val;
6387
6388         net = (struct net *)ctl->extra2;
6389         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6390                 rtnl_unlock();
6391                 return 0;
6392         }
6393
6394         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6395                 struct net_device *dev;
6396
6397                 for_each_netdev(net, dev) {
6398                         idev = __in6_dev_get(dev);
6399                         if (idev)
6400                                 addrconf_disable_policy_idev(idev, val);
6401                 }
6402         } else {
6403                 idev = (struct inet6_dev *)ctl->extra1;
6404                 addrconf_disable_policy_idev(idev, val);
6405         }
6406
6407         rtnl_unlock();
6408         return 0;
6409 }
6410
6411 static
6412 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6413                                    void __user *buffer, size_t *lenp,
6414                                    loff_t *ppos)
6415 {
6416         int *valp = ctl->data;
6417         int val = *valp;
6418         loff_t pos = *ppos;
6419         struct ctl_table lctl;
6420         int ret;
6421
6422         lctl = *ctl;
6423         lctl.data = &val;
6424         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6425
6426         if (write && (*valp != val))
6427                 ret = addrconf_disable_policy(ctl, valp, val);
6428
6429         if (ret)
6430                 *ppos = pos;
6431
6432         return ret;
6433 }
6434
6435 static int minus_one = -1;
6436 static const int two_five_five = 255;
6437
6438 static const struct ctl_table addrconf_sysctl[] = {
6439         {
6440                 .procname       = "forwarding",
6441                 .data           = &ipv6_devconf.forwarding,
6442                 .maxlen         = sizeof(int),
6443                 .mode           = 0644,
6444                 .proc_handler   = addrconf_sysctl_forward,
6445         },
6446         {
6447                 .procname       = "hop_limit",
6448                 .data           = &ipv6_devconf.hop_limit,
6449                 .maxlen         = sizeof(int),
6450                 .mode           = 0644,
6451                 .proc_handler   = proc_dointvec_minmax,
6452                 .extra1         = (void *)SYSCTL_ONE,
6453                 .extra2         = (void *)&two_five_five,
6454         },
6455         {
6456                 .procname       = "mtu",
6457                 .data           = &ipv6_devconf.mtu6,
6458                 .maxlen         = sizeof(int),
6459                 .mode           = 0644,
6460                 .proc_handler   = addrconf_sysctl_mtu,
6461         },
6462         {
6463                 .procname       = "accept_ra",
6464                 .data           = &ipv6_devconf.accept_ra,
6465                 .maxlen         = sizeof(int),
6466                 .mode           = 0644,
6467                 .proc_handler   = proc_dointvec,
6468         },
6469         {
6470                 .procname       = "accept_redirects",
6471                 .data           = &ipv6_devconf.accept_redirects,
6472                 .maxlen         = sizeof(int),
6473                 .mode           = 0644,
6474                 .proc_handler   = proc_dointvec,
6475         },
6476         {
6477                 .procname       = "autoconf",
6478                 .data           = &ipv6_devconf.autoconf,
6479                 .maxlen         = sizeof(int),
6480                 .mode           = 0644,
6481                 .proc_handler   = proc_dointvec,
6482         },
6483         {
6484                 .procname       = "dad_transmits",
6485                 .data           = &ipv6_devconf.dad_transmits,
6486                 .maxlen         = sizeof(int),
6487                 .mode           = 0644,
6488                 .proc_handler   = proc_dointvec,
6489         },
6490         {
6491                 .procname       = "router_solicitations",
6492                 .data           = &ipv6_devconf.rtr_solicits,
6493                 .maxlen         = sizeof(int),
6494                 .mode           = 0644,
6495                 .proc_handler   = proc_dointvec_minmax,
6496                 .extra1         = &minus_one,
6497         },
6498         {
6499                 .procname       = "router_solicitation_interval",
6500                 .data           = &ipv6_devconf.rtr_solicit_interval,
6501                 .maxlen         = sizeof(int),
6502                 .mode           = 0644,
6503                 .proc_handler   = proc_dointvec_jiffies,
6504         },
6505         {
6506                 .procname       = "router_solicitation_max_interval",
6507                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6508                 .maxlen         = sizeof(int),
6509                 .mode           = 0644,
6510                 .proc_handler   = proc_dointvec_jiffies,
6511         },
6512         {
6513                 .procname       = "router_solicitation_delay",
6514                 .data           = &ipv6_devconf.rtr_solicit_delay,
6515                 .maxlen         = sizeof(int),
6516                 .mode           = 0644,
6517                 .proc_handler   = proc_dointvec_jiffies,
6518         },
6519         {
6520                 .procname       = "force_mld_version",
6521                 .data           = &ipv6_devconf.force_mld_version,
6522                 .maxlen         = sizeof(int),
6523                 .mode           = 0644,
6524                 .proc_handler   = proc_dointvec,
6525         },
6526         {
6527                 .procname       = "mldv1_unsolicited_report_interval",
6528                 .data           =
6529                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6530                 .maxlen         = sizeof(int),
6531                 .mode           = 0644,
6532                 .proc_handler   = proc_dointvec_ms_jiffies,
6533         },
6534         {
6535                 .procname       = "mldv2_unsolicited_report_interval",
6536                 .data           =
6537                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6538                 .maxlen         = sizeof(int),
6539                 .mode           = 0644,
6540                 .proc_handler   = proc_dointvec_ms_jiffies,
6541         },
6542         {
6543                 .procname       = "use_tempaddr",
6544                 .data           = &ipv6_devconf.use_tempaddr,
6545                 .maxlen         = sizeof(int),
6546                 .mode           = 0644,
6547                 .proc_handler   = proc_dointvec,
6548         },
6549         {
6550                 .procname       = "temp_valid_lft",
6551                 .data           = &ipv6_devconf.temp_valid_lft,
6552                 .maxlen         = sizeof(int),
6553                 .mode           = 0644,
6554                 .proc_handler   = proc_dointvec,
6555         },
6556         {
6557                 .procname       = "temp_prefered_lft",
6558                 .data           = &ipv6_devconf.temp_prefered_lft,
6559                 .maxlen         = sizeof(int),
6560                 .mode           = 0644,
6561                 .proc_handler   = proc_dointvec,
6562         },
6563         {
6564                 .procname       = "regen_max_retry",
6565                 .data           = &ipv6_devconf.regen_max_retry,
6566                 .maxlen         = sizeof(int),
6567                 .mode           = 0644,
6568                 .proc_handler   = proc_dointvec,
6569         },
6570         {
6571                 .procname       = "max_desync_factor",
6572                 .data           = &ipv6_devconf.max_desync_factor,
6573                 .maxlen         = sizeof(int),
6574                 .mode           = 0644,
6575                 .proc_handler   = proc_dointvec,
6576         },
6577         {
6578                 .procname       = "max_addresses",
6579                 .data           = &ipv6_devconf.max_addresses,
6580                 .maxlen         = sizeof(int),
6581                 .mode           = 0644,
6582                 .proc_handler   = proc_dointvec,
6583         },
6584         {
6585                 .procname       = "accept_ra_defrtr",
6586                 .data           = &ipv6_devconf.accept_ra_defrtr,
6587                 .maxlen         = sizeof(int),
6588                 .mode           = 0644,
6589                 .proc_handler   = proc_dointvec,
6590         },
6591         {
6592                 .procname       = "accept_ra_min_hop_limit",
6593                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6594                 .maxlen         = sizeof(int),
6595                 .mode           = 0644,
6596                 .proc_handler   = proc_dointvec,
6597         },
6598         {
6599                 .procname       = "accept_ra_pinfo",
6600                 .data           = &ipv6_devconf.accept_ra_pinfo,
6601                 .maxlen         = sizeof(int),
6602                 .mode           = 0644,
6603                 .proc_handler   = proc_dointvec,
6604         },
6605 #ifdef CONFIG_IPV6_ROUTER_PREF
6606         {
6607                 .procname       = "accept_ra_rtr_pref",
6608                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6609                 .maxlen         = sizeof(int),
6610                 .mode           = 0644,
6611                 .proc_handler   = proc_dointvec,
6612         },
6613         {
6614                 .procname       = "router_probe_interval",
6615                 .data           = &ipv6_devconf.rtr_probe_interval,
6616                 .maxlen         = sizeof(int),
6617                 .mode           = 0644,
6618                 .proc_handler   = proc_dointvec_jiffies,
6619         },
6620 #ifdef CONFIG_IPV6_ROUTE_INFO
6621         {
6622                 .procname       = "accept_ra_rt_info_min_plen",
6623                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6624                 .maxlen         = sizeof(int),
6625                 .mode           = 0644,
6626                 .proc_handler   = proc_dointvec,
6627         },
6628         {
6629                 .procname       = "accept_ra_rt_info_max_plen",
6630                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6631                 .maxlen         = sizeof(int),
6632                 .mode           = 0644,
6633                 .proc_handler   = proc_dointvec,
6634         },
6635 #endif
6636 #endif
6637         {
6638                 .procname       = "proxy_ndp",
6639                 .data           = &ipv6_devconf.proxy_ndp,
6640                 .maxlen         = sizeof(int),
6641                 .mode           = 0644,
6642                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6643         },
6644         {
6645                 .procname       = "accept_source_route",
6646                 .data           = &ipv6_devconf.accept_source_route,
6647                 .maxlen         = sizeof(int),
6648                 .mode           = 0644,
6649                 .proc_handler   = proc_dointvec,
6650         },
6651 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6652         {
6653                 .procname       = "optimistic_dad",
6654                 .data           = &ipv6_devconf.optimistic_dad,
6655                 .maxlen         = sizeof(int),
6656                 .mode           = 0644,
6657                 .proc_handler   = proc_dointvec,
6658         },
6659         {
6660                 .procname       = "use_optimistic",
6661                 .data           = &ipv6_devconf.use_optimistic,
6662                 .maxlen         = sizeof(int),
6663                 .mode           = 0644,
6664                 .proc_handler   = proc_dointvec,
6665         },
6666 #endif
6667 #ifdef CONFIG_IPV6_MROUTE
6668         {
6669                 .procname       = "mc_forwarding",
6670                 .data           = &ipv6_devconf.mc_forwarding,
6671                 .maxlen         = sizeof(int),
6672                 .mode           = 0444,
6673                 .proc_handler   = proc_dointvec,
6674         },
6675 #endif
6676         {
6677                 .procname       = "disable_ipv6",
6678                 .data           = &ipv6_devconf.disable_ipv6,
6679                 .maxlen         = sizeof(int),
6680                 .mode           = 0644,
6681                 .proc_handler   = addrconf_sysctl_disable,
6682         },
6683         {
6684                 .procname       = "accept_dad",
6685                 .data           = &ipv6_devconf.accept_dad,
6686                 .maxlen         = sizeof(int),
6687                 .mode           = 0644,
6688                 .proc_handler   = proc_dointvec,
6689         },
6690         {
6691                 .procname       = "force_tllao",
6692                 .data           = &ipv6_devconf.force_tllao,
6693                 .maxlen         = sizeof(int),
6694                 .mode           = 0644,
6695                 .proc_handler   = proc_dointvec
6696         },
6697         {
6698                 .procname       = "ndisc_notify",
6699                 .data           = &ipv6_devconf.ndisc_notify,
6700                 .maxlen         = sizeof(int),
6701                 .mode           = 0644,
6702                 .proc_handler   = proc_dointvec
6703         },
6704         {
6705                 .procname       = "suppress_frag_ndisc",
6706                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6707                 .maxlen         = sizeof(int),
6708                 .mode           = 0644,
6709                 .proc_handler   = proc_dointvec
6710         },
6711         {
6712                 .procname       = "accept_ra_from_local",
6713                 .data           = &ipv6_devconf.accept_ra_from_local,
6714                 .maxlen         = sizeof(int),
6715                 .mode           = 0644,
6716                 .proc_handler   = proc_dointvec,
6717         },
6718         {
6719                 .procname       = "accept_ra_mtu",
6720                 .data           = &ipv6_devconf.accept_ra_mtu,
6721                 .maxlen         = sizeof(int),
6722                 .mode           = 0644,
6723                 .proc_handler   = proc_dointvec,
6724         },
6725         {
6726                 .procname       = "stable_secret",
6727                 .data           = &ipv6_devconf.stable_secret,
6728                 .maxlen         = IPV6_MAX_STRLEN,
6729                 .mode           = 0600,
6730                 .proc_handler   = addrconf_sysctl_stable_secret,
6731         },
6732         {
6733                 .procname       = "use_oif_addrs_only",
6734                 .data           = &ipv6_devconf.use_oif_addrs_only,
6735                 .maxlen         = sizeof(int),
6736                 .mode           = 0644,
6737                 .proc_handler   = proc_dointvec,
6738         },
6739         {
6740                 .procname       = "ignore_routes_with_linkdown",
6741                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6742                 .maxlen         = sizeof(int),
6743                 .mode           = 0644,
6744                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6745         },
6746         {
6747                 .procname       = "drop_unicast_in_l2_multicast",
6748                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6749                 .maxlen         = sizeof(int),
6750                 .mode           = 0644,
6751                 .proc_handler   = proc_dointvec,
6752         },
6753         {
6754                 .procname       = "drop_unsolicited_na",
6755                 .data           = &ipv6_devconf.drop_unsolicited_na,
6756                 .maxlen         = sizeof(int),
6757                 .mode           = 0644,
6758                 .proc_handler   = proc_dointvec,
6759         },
6760         {
6761                 .procname       = "keep_addr_on_down",
6762                 .data           = &ipv6_devconf.keep_addr_on_down,
6763                 .maxlen         = sizeof(int),
6764                 .mode           = 0644,
6765                 .proc_handler   = proc_dointvec,
6766
6767         },
6768         {
6769                 .procname       = "seg6_enabled",
6770                 .data           = &ipv6_devconf.seg6_enabled,
6771                 .maxlen         = sizeof(int),
6772                 .mode           = 0644,
6773                 .proc_handler   = proc_dointvec,
6774         },
6775 #ifdef CONFIG_IPV6_SEG6_HMAC
6776         {
6777                 .procname       = "seg6_require_hmac",
6778                 .data           = &ipv6_devconf.seg6_require_hmac,
6779                 .maxlen         = sizeof(int),
6780                 .mode           = 0644,
6781                 .proc_handler   = proc_dointvec,
6782         },
6783 #endif
6784         {
6785                 .procname       = "enhanced_dad",
6786                 .data           = &ipv6_devconf.enhanced_dad,
6787                 .maxlen         = sizeof(int),
6788                 .mode           = 0644,
6789                 .proc_handler   = proc_dointvec,
6790         },
6791         {
6792                 .procname               = "addr_gen_mode",
6793                 .data                   = &ipv6_devconf.addr_gen_mode,
6794                 .maxlen                 = sizeof(int),
6795                 .mode                   = 0644,
6796                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6797         },
6798         {
6799                 .procname       = "disable_policy",
6800                 .data           = &ipv6_devconf.disable_policy,
6801                 .maxlen         = sizeof(int),
6802                 .mode           = 0644,
6803                 .proc_handler   = addrconf_sysctl_disable_policy,
6804         },
6805         {
6806                 .procname       = "ndisc_tclass",
6807                 .data           = &ipv6_devconf.ndisc_tclass,
6808                 .maxlen         = sizeof(int),
6809                 .mode           = 0644,
6810                 .proc_handler   = proc_dointvec_minmax,
6811                 .extra1         = (void *)SYSCTL_ZERO,
6812                 .extra2         = (void *)&two_five_five,
6813         },
6814         {
6815                 /* sentinel */
6816         }
6817 };
6818
6819 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6820                 struct inet6_dev *idev, struct ipv6_devconf *p)
6821 {
6822         int i, ifindex;
6823         struct ctl_table *table;
6824         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6825
6826         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6827         if (!table)
6828                 goto out;
6829
6830         for (i = 0; table[i].data; i++) {
6831                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6832                 /* If one of these is already set, then it is not safe to
6833                  * overwrite either of them: this makes proc_dointvec_minmax
6834                  * usable.
6835                  */
6836                 if (!table[i].extra1 && !table[i].extra2) {
6837                         table[i].extra1 = idev; /* embedded; no ref */
6838                         table[i].extra2 = net;
6839                 }
6840         }
6841
6842         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6843
6844         p->sysctl_header = register_net_sysctl(net, path, table);
6845         if (!p->sysctl_header)
6846                 goto free;
6847
6848         if (!strcmp(dev_name, "all"))
6849                 ifindex = NETCONFA_IFINDEX_ALL;
6850         else if (!strcmp(dev_name, "default"))
6851                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6852         else
6853                 ifindex = idev->dev->ifindex;
6854         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6855                                      ifindex, p);
6856         return 0;
6857
6858 free:
6859         kfree(table);
6860 out:
6861         return -ENOBUFS;
6862 }
6863
6864 static void __addrconf_sysctl_unregister(struct net *net,
6865                                          struct ipv6_devconf *p, int ifindex)
6866 {
6867         struct ctl_table *table;
6868
6869         if (!p->sysctl_header)
6870                 return;
6871
6872         table = p->sysctl_header->ctl_table_arg;
6873         unregister_net_sysctl_table(p->sysctl_header);
6874         p->sysctl_header = NULL;
6875         kfree(table);
6876
6877         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6878 }
6879
6880 static int addrconf_sysctl_register(struct inet6_dev *idev)
6881 {
6882         int err;
6883
6884         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6885                 return -EINVAL;
6886
6887         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6888                                     &ndisc_ifinfo_sysctl_change);
6889         if (err)
6890                 return err;
6891         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6892                                          idev, &idev->cnf);
6893         if (err)
6894                 neigh_sysctl_unregister(idev->nd_parms);
6895
6896         return err;
6897 }
6898
6899 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6900 {
6901         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6902                                      idev->dev->ifindex);
6903         neigh_sysctl_unregister(idev->nd_parms);
6904 }
6905
6906
6907 #endif
6908
6909 static int __net_init addrconf_init_net(struct net *net)
6910 {
6911         int err = -ENOMEM;
6912         struct ipv6_devconf *all, *dflt;
6913
6914         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6915         if (!all)
6916                 goto err_alloc_all;
6917
6918         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6919         if (!dflt)
6920                 goto err_alloc_dflt;
6921
6922         if (IS_ENABLED(CONFIG_SYSCTL) &&
6923             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6924                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6925                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6926         }
6927
6928         /* these will be inherited by all namespaces */
6929         dflt->autoconf = ipv6_defaults.autoconf;
6930         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6931
6932         dflt->stable_secret.initialized = false;
6933         all->stable_secret.initialized = false;
6934
6935         net->ipv6.devconf_all = all;
6936         net->ipv6.devconf_dflt = dflt;
6937
6938 #ifdef CONFIG_SYSCTL
6939         err = __addrconf_sysctl_register(net, "all", NULL, all);
6940         if (err < 0)
6941                 goto err_reg_all;
6942
6943         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6944         if (err < 0)
6945                 goto err_reg_dflt;
6946 #endif
6947         return 0;
6948
6949 #ifdef CONFIG_SYSCTL
6950 err_reg_dflt:
6951         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6952 err_reg_all:
6953         kfree(dflt);
6954 #endif
6955 err_alloc_dflt:
6956         kfree(all);
6957 err_alloc_all:
6958         return err;
6959 }
6960
6961 static void __net_exit addrconf_exit_net(struct net *net)
6962 {
6963 #ifdef CONFIG_SYSCTL
6964         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6965                                      NETCONFA_IFINDEX_DEFAULT);
6966         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6967                                      NETCONFA_IFINDEX_ALL);
6968 #endif
6969         kfree(net->ipv6.devconf_dflt);
6970         kfree(net->ipv6.devconf_all);
6971 }
6972
6973 static struct pernet_operations addrconf_ops = {
6974         .init = addrconf_init_net,
6975         .exit = addrconf_exit_net,
6976 };
6977
6978 static struct rtnl_af_ops inet6_ops __read_mostly = {
6979         .family           = AF_INET6,
6980         .fill_link_af     = inet6_fill_link_af,
6981         .get_link_af_size = inet6_get_link_af_size,
6982         .validate_link_af = inet6_validate_link_af,
6983         .set_link_af      = inet6_set_link_af,
6984 };
6985
6986 /*
6987  *      Init / cleanup code
6988  */
6989
6990 int __init addrconf_init(void)
6991 {
6992         struct inet6_dev *idev;
6993         int i, err;
6994
6995         err = ipv6_addr_label_init();
6996         if (err < 0) {
6997                 pr_crit("%s: cannot initialize default policy table: %d\n",
6998                         __func__, err);
6999                 goto out;
7000         }
7001
7002         err = register_pernet_subsys(&addrconf_ops);
7003         if (err < 0)
7004                 goto out_addrlabel;
7005
7006         addrconf_wq = create_workqueue("ipv6_addrconf");
7007         if (!addrconf_wq) {
7008                 err = -ENOMEM;
7009                 goto out_nowq;
7010         }
7011
7012         /* The addrconf netdev notifier requires that loopback_dev
7013          * has it's ipv6 private information allocated and setup
7014          * before it can bring up and give link-local addresses
7015          * to other devices which are up.
7016          *
7017          * Unfortunately, loopback_dev is not necessarily the first
7018          * entry in the global dev_base list of net devices.  In fact,
7019          * it is likely to be the very last entry on that list.
7020          * So this causes the notifier registry below to try and
7021          * give link-local addresses to all devices besides loopback_dev
7022          * first, then loopback_dev, which cases all the non-loopback_dev
7023          * devices to fail to get a link-local address.
7024          *
7025          * So, as a temporary fix, allocate the ipv6 structure for
7026          * loopback_dev first by hand.
7027          * Longer term, all of the dependencies ipv6 has upon the loopback
7028          * device and it being up should be removed.
7029          */
7030         rtnl_lock();
7031         idev = ipv6_add_dev(init_net.loopback_dev);
7032         rtnl_unlock();
7033         if (IS_ERR(idev)) {
7034                 err = PTR_ERR(idev);
7035                 goto errlo;
7036         }
7037
7038         ip6_route_init_special_entries();
7039
7040         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7041                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7042
7043         register_netdevice_notifier(&ipv6_dev_notf);
7044
7045         addrconf_verify();
7046
7047         rtnl_af_register(&inet6_ops);
7048
7049         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7050                                    NULL, inet6_dump_ifinfo, 0);
7051         if (err < 0)
7052                 goto errout;
7053
7054         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7055                                    inet6_rtm_newaddr, NULL, 0);
7056         if (err < 0)
7057                 goto errout;
7058         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7059                                    inet6_rtm_deladdr, NULL, 0);
7060         if (err < 0)
7061                 goto errout;
7062         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7063                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7064                                    RTNL_FLAG_DOIT_UNLOCKED);
7065         if (err < 0)
7066                 goto errout;
7067         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7068                                    NULL, inet6_dump_ifmcaddr, 0);
7069         if (err < 0)
7070                 goto errout;
7071         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7072                                    NULL, inet6_dump_ifacaddr, 0);
7073         if (err < 0)
7074                 goto errout;
7075         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7076                                    inet6_netconf_get_devconf,
7077                                    inet6_netconf_dump_devconf,
7078                                    RTNL_FLAG_DOIT_UNLOCKED);
7079         if (err < 0)
7080                 goto errout;
7081         err = ipv6_addr_label_rtnl_register();
7082         if (err < 0)
7083                 goto errout;
7084
7085         return 0;
7086 errout:
7087         rtnl_unregister_all(PF_INET6);
7088         rtnl_af_unregister(&inet6_ops);
7089         unregister_netdevice_notifier(&ipv6_dev_notf);
7090 errlo:
7091         destroy_workqueue(addrconf_wq);
7092 out_nowq:
7093         unregister_pernet_subsys(&addrconf_ops);
7094 out_addrlabel:
7095         ipv6_addr_label_cleanup();
7096 out:
7097         return err;
7098 }
7099
7100 void addrconf_cleanup(void)
7101 {
7102         struct net_device *dev;
7103         int i;
7104
7105         unregister_netdevice_notifier(&ipv6_dev_notf);
7106         unregister_pernet_subsys(&addrconf_ops);
7107         ipv6_addr_label_cleanup();
7108
7109         rtnl_af_unregister(&inet6_ops);
7110
7111         rtnl_lock();
7112
7113         /* clean dev list */
7114         for_each_netdev(&init_net, dev) {
7115                 if (__in6_dev_get(dev) == NULL)
7116                         continue;
7117                 addrconf_ifdown(dev, 1);
7118         }
7119         addrconf_ifdown(init_net.loopback_dev, 2);
7120
7121         /*
7122          *      Check hash table.
7123          */
7124         spin_lock_bh(&addrconf_hash_lock);
7125         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7126                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7127         spin_unlock_bh(&addrconf_hash_lock);
7128         cancel_delayed_work(&addr_chk_work);
7129         rtnl_unlock();
7130
7131         destroy_workqueue(addrconf_wq);
7132 }