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,
1230                      bool del_rt, bool del_peer)
1231 {
1232         struct fib6_info *f6i;
1233
1234         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1235                                         ifp->prefix_len,
1236                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1237         if (f6i) {
1238                 if (del_rt)
1239                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1240                 else {
1241                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1242                                 fib6_set_expires(f6i, expires);
1243                         fib6_info_release(f6i);
1244                 }
1245         }
1246 }
1247
1248
1249 /* This function wants to get referenced ifp and releases it before return */
1250
1251 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1252 {
1253         int state;
1254         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1255         unsigned long expires;
1256
1257         ASSERT_RTNL();
1258
1259         spin_lock_bh(&ifp->lock);
1260         state = ifp->state;
1261         ifp->state = INET6_IFADDR_STATE_DEAD;
1262         spin_unlock_bh(&ifp->lock);
1263
1264         if (state == INET6_IFADDR_STATE_DEAD)
1265                 goto out;
1266
1267         spin_lock_bh(&addrconf_hash_lock);
1268         hlist_del_init_rcu(&ifp->addr_lst);
1269         spin_unlock_bh(&addrconf_hash_lock);
1270
1271         write_lock_bh(&ifp->idev->lock);
1272
1273         if (ifp->flags&IFA_F_TEMPORARY) {
1274                 list_del(&ifp->tmp_list);
1275                 if (ifp->ifpub) {
1276                         in6_ifa_put(ifp->ifpub);
1277                         ifp->ifpub = NULL;
1278                 }
1279                 __in6_ifa_put(ifp);
1280         }
1281
1282         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1283                 action = check_cleanup_prefix_route(ifp, &expires);
1284
1285         list_del_rcu(&ifp->if_list);
1286         __in6_ifa_put(ifp);
1287
1288         write_unlock_bh(&ifp->idev->lock);
1289
1290         addrconf_del_dad_work(ifp);
1291
1292         ipv6_ifa_notify(RTM_DELADDR, ifp);
1293
1294         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1295
1296         if (action != CLEANUP_PREFIX_RT_NOP) {
1297                 cleanup_prefix_route(ifp, expires,
1298                         action == CLEANUP_PREFIX_RT_DEL, false);
1299         }
1300
1301         /* clean up prefsrc entries */
1302         rt6_remove_prefsrc(ifp);
1303 out:
1304         in6_ifa_put(ifp);
1305 }
1306
1307 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1308                                 struct inet6_ifaddr *ift,
1309                                 bool block)
1310 {
1311         struct inet6_dev *idev = ifp->idev;
1312         struct in6_addr addr, *tmpaddr;
1313         unsigned long tmp_tstamp, age;
1314         unsigned long regen_advance;
1315         struct ifa6_config cfg;
1316         int ret = 0;
1317         unsigned long now = jiffies;
1318         long max_desync_factor;
1319         s32 cnf_temp_preferred_lft;
1320
1321         write_lock_bh(&idev->lock);
1322         if (ift) {
1323                 spin_lock_bh(&ift->lock);
1324                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1325                 spin_unlock_bh(&ift->lock);
1326                 tmpaddr = &addr;
1327         } else {
1328                 tmpaddr = NULL;
1329         }
1330 retry:
1331         in6_dev_hold(idev);
1332         if (idev->cnf.use_tempaddr <= 0) {
1333                 write_unlock_bh(&idev->lock);
1334                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1335                 in6_dev_put(idev);
1336                 ret = -1;
1337                 goto out;
1338         }
1339         spin_lock_bh(&ifp->lock);
1340         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1341                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1342                 spin_unlock_bh(&ifp->lock);
1343                 write_unlock_bh(&idev->lock);
1344                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1345                         __func__);
1346                 in6_dev_put(idev);
1347                 ret = -1;
1348                 goto out;
1349         }
1350         in6_ifa_hold(ifp);
1351         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1352         ipv6_try_regen_rndid(idev, tmpaddr);
1353         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1354         age = (now - ifp->tstamp) / HZ;
1355
1356         regen_advance = idev->cnf.regen_max_retry *
1357                         idev->cnf.dad_transmits *
1358                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1359
1360         /* recalculate max_desync_factor each time and update
1361          * idev->desync_factor if it's larger
1362          */
1363         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1364         max_desync_factor = min_t(__u32,
1365                                   idev->cnf.max_desync_factor,
1366                                   cnf_temp_preferred_lft - regen_advance);
1367
1368         if (unlikely(idev->desync_factor > max_desync_factor)) {
1369                 if (max_desync_factor > 0) {
1370                         get_random_bytes(&idev->desync_factor,
1371                                          sizeof(idev->desync_factor));
1372                         idev->desync_factor %= max_desync_factor;
1373                 } else {
1374                         idev->desync_factor = 0;
1375                 }
1376         }
1377
1378         memset(&cfg, 0, sizeof(cfg));
1379         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1380                               idev->cnf.temp_valid_lft + age);
1381         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1382         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1383
1384         cfg.plen = ifp->prefix_len;
1385         tmp_tstamp = ifp->tstamp;
1386         spin_unlock_bh(&ifp->lock);
1387
1388         write_unlock_bh(&idev->lock);
1389
1390         /* A temporary address is created only if this calculated Preferred
1391          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1392          * an implementation must not create a temporary address with a zero
1393          * Preferred Lifetime.
1394          * Use age calculation as in addrconf_verify to avoid unnecessary
1395          * temporary addresses being generated.
1396          */
1397         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1398         if (cfg.preferred_lft <= regen_advance + age) {
1399                 in6_ifa_put(ifp);
1400                 in6_dev_put(idev);
1401                 ret = -1;
1402                 goto out;
1403         }
1404
1405         cfg.ifa_flags = IFA_F_TEMPORARY;
1406         /* set in addrconf_prefix_rcv() */
1407         if (ifp->flags & IFA_F_OPTIMISTIC)
1408                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1409
1410         cfg.pfx = &addr;
1411         cfg.scope = ipv6_addr_scope(cfg.pfx);
1412
1413         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1414         if (IS_ERR(ift)) {
1415                 in6_ifa_put(ifp);
1416                 in6_dev_put(idev);
1417                 pr_info("%s: retry temporary address regeneration\n", __func__);
1418                 tmpaddr = &addr;
1419                 write_lock_bh(&idev->lock);
1420                 goto retry;
1421         }
1422
1423         spin_lock_bh(&ift->lock);
1424         ift->ifpub = ifp;
1425         ift->cstamp = now;
1426         ift->tstamp = tmp_tstamp;
1427         spin_unlock_bh(&ift->lock);
1428
1429         addrconf_dad_start(ift);
1430         in6_ifa_put(ift);
1431         in6_dev_put(idev);
1432 out:
1433         return ret;
1434 }
1435
1436 /*
1437  *      Choose an appropriate source address (RFC3484)
1438  */
1439 enum {
1440         IPV6_SADDR_RULE_INIT = 0,
1441         IPV6_SADDR_RULE_LOCAL,
1442         IPV6_SADDR_RULE_SCOPE,
1443         IPV6_SADDR_RULE_PREFERRED,
1444 #ifdef CONFIG_IPV6_MIP6
1445         IPV6_SADDR_RULE_HOA,
1446 #endif
1447         IPV6_SADDR_RULE_OIF,
1448         IPV6_SADDR_RULE_LABEL,
1449         IPV6_SADDR_RULE_PRIVACY,
1450         IPV6_SADDR_RULE_ORCHID,
1451         IPV6_SADDR_RULE_PREFIX,
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1454 #endif
1455         IPV6_SADDR_RULE_MAX
1456 };
1457
1458 struct ipv6_saddr_score {
1459         int                     rule;
1460         int                     addr_type;
1461         struct inet6_ifaddr     *ifa;
1462         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1463         int                     scopedist;
1464         int                     matchlen;
1465 };
1466
1467 struct ipv6_saddr_dst {
1468         const struct in6_addr *addr;
1469         int ifindex;
1470         int scope;
1471         int label;
1472         unsigned int prefs;
1473 };
1474
1475 static inline int ipv6_saddr_preferred(int type)
1476 {
1477         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1478                 return 1;
1479         return 0;
1480 }
1481
1482 static bool ipv6_use_optimistic_addr(struct net *net,
1483                                      struct inet6_dev *idev)
1484 {
1485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486         if (!idev)
1487                 return false;
1488         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1489                 return false;
1490         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1491                 return false;
1492
1493         return true;
1494 #else
1495         return false;
1496 #endif
1497 }
1498
1499 static bool ipv6_allow_optimistic_dad(struct net *net,
1500                                       struct inet6_dev *idev)
1501 {
1502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1503         if (!idev)
1504                 return false;
1505         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1506                 return false;
1507
1508         return true;
1509 #else
1510         return false;
1511 #endif
1512 }
1513
1514 static int ipv6_get_saddr_eval(struct net *net,
1515                                struct ipv6_saddr_score *score,
1516                                struct ipv6_saddr_dst *dst,
1517                                int i)
1518 {
1519         int ret;
1520
1521         if (i <= score->rule) {
1522                 switch (i) {
1523                 case IPV6_SADDR_RULE_SCOPE:
1524                         ret = score->scopedist;
1525                         break;
1526                 case IPV6_SADDR_RULE_PREFIX:
1527                         ret = score->matchlen;
1528                         break;
1529                 default:
1530                         ret = !!test_bit(i, score->scorebits);
1531                 }
1532                 goto out;
1533         }
1534
1535         switch (i) {
1536         case IPV6_SADDR_RULE_INIT:
1537                 /* Rule 0: remember if hiscore is not ready yet */
1538                 ret = !!score->ifa;
1539                 break;
1540         case IPV6_SADDR_RULE_LOCAL:
1541                 /* Rule 1: Prefer same address */
1542                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1543                 break;
1544         case IPV6_SADDR_RULE_SCOPE:
1545                 /* Rule 2: Prefer appropriate scope
1546                  *
1547                  *      ret
1548                  *       ^
1549                  *    -1 |  d 15
1550                  *    ---+--+-+---> scope
1551                  *       |
1552                  *       |             d is scope of the destination.
1553                  *  B-d  |  \
1554                  *       |   \      <- smaller scope is better if
1555                  *  B-15 |    \        if scope is enough for destination.
1556                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1557                  * d-C-1 | /
1558                  *       |/         <- greater is better
1559                  *   -C  /             if scope is not enough for destination.
1560                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1561                  *
1562                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1563                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1564                  * Assume B = 0 and we get C > 29.
1565                  */
1566                 ret = __ipv6_addr_src_scope(score->addr_type);
1567                 if (ret >= dst->scope)
1568                         ret = -ret;
1569                 else
1570                         ret -= 128;     /* 30 is enough */
1571                 score->scopedist = ret;
1572                 break;
1573         case IPV6_SADDR_RULE_PREFERRED:
1574             {
1575                 /* Rule 3: Avoid deprecated and optimistic addresses */
1576                 u8 avoid = IFA_F_DEPRECATED;
1577
1578                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1579                         avoid |= IFA_F_OPTIMISTIC;
1580                 ret = ipv6_saddr_preferred(score->addr_type) ||
1581                       !(score->ifa->flags & avoid);
1582                 break;
1583             }
1584 #ifdef CONFIG_IPV6_MIP6
1585         case IPV6_SADDR_RULE_HOA:
1586             {
1587                 /* Rule 4: Prefer home address */
1588                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1589                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1590                 break;
1591             }
1592 #endif
1593         case IPV6_SADDR_RULE_OIF:
1594                 /* Rule 5: Prefer outgoing interface */
1595                 ret = (!dst->ifindex ||
1596                        dst->ifindex == score->ifa->idev->dev->ifindex);
1597                 break;
1598         case IPV6_SADDR_RULE_LABEL:
1599                 /* Rule 6: Prefer matching label */
1600                 ret = ipv6_addr_label(net,
1601                                       &score->ifa->addr, score->addr_type,
1602                                       score->ifa->idev->dev->ifindex) == dst->label;
1603                 break;
1604         case IPV6_SADDR_RULE_PRIVACY:
1605             {
1606                 /* Rule 7: Prefer public address
1607                  * Note: prefer temporary address if use_tempaddr >= 2
1608                  */
1609                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1610                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1611                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1612                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1613                 break;
1614             }
1615         case IPV6_SADDR_RULE_ORCHID:
1616                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1617                  *          non-ORCHID vs non-ORCHID
1618                  */
1619                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1620                         ipv6_addr_orchid(dst->addr));
1621                 break;
1622         case IPV6_SADDR_RULE_PREFIX:
1623                 /* Rule 8: Use longest matching prefix */
1624                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1625                 if (ret > score->ifa->prefix_len)
1626                         ret = score->ifa->prefix_len;
1627                 score->matchlen = ret;
1628                 break;
1629 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1630         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1631                 /* Optimistic addresses still have lower precedence than other
1632                  * preferred addresses.
1633                  */
1634                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1635                 break;
1636 #endif
1637         default:
1638                 ret = 0;
1639         }
1640
1641         if (ret)
1642                 __set_bit(i, score->scorebits);
1643         score->rule = i;
1644 out:
1645         return ret;
1646 }
1647
1648 static int __ipv6_dev_get_saddr(struct net *net,
1649                                 struct ipv6_saddr_dst *dst,
1650                                 struct inet6_dev *idev,
1651                                 struct ipv6_saddr_score *scores,
1652                                 int hiscore_idx)
1653 {
1654         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1655
1656         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1657                 int i;
1658
1659                 /*
1660                  * - Tentative Address (RFC2462 section 5.4)
1661                  *  - A tentative address is not considered
1662                  *    "assigned to an interface" in the traditional
1663                  *    sense, unless it is also flagged as optimistic.
1664                  * - Candidate Source Address (section 4)
1665                  *  - In any case, anycast addresses, multicast
1666                  *    addresses, and the unspecified address MUST
1667                  *    NOT be included in a candidate set.
1668                  */
1669                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1670                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1671                         continue;
1672
1673                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1674
1675                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1676                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1677                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1678                                             idev->dev->name);
1679                         continue;
1680                 }
1681
1682                 score->rule = -1;
1683                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1684
1685                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1686                         int minihiscore, miniscore;
1687
1688                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1689                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1690
1691                         if (minihiscore > miniscore) {
1692                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1693                                     score->scopedist > 0) {
1694                                         /*
1695                                          * special case:
1696                                          * each remaining entry
1697                                          * has too small (not enough)
1698                                          * scope, because ifa entries
1699                                          * are sorted by their scope
1700                                          * values.
1701                                          */
1702                                         goto out;
1703                                 }
1704                                 break;
1705                         } else if (minihiscore < miniscore) {
1706                                 swap(hiscore, score);
1707                                 hiscore_idx = 1 - hiscore_idx;
1708
1709                                 /* restore our iterator */
1710                                 score->ifa = hiscore->ifa;
1711
1712                                 break;
1713                         }
1714                 }
1715         }
1716 out:
1717         return hiscore_idx;
1718 }
1719
1720 static int ipv6_get_saddr_master(struct net *net,
1721                                  const struct net_device *dst_dev,
1722                                  const struct net_device *master,
1723                                  struct ipv6_saddr_dst *dst,
1724                                  struct ipv6_saddr_score *scores,
1725                                  int hiscore_idx)
1726 {
1727         struct inet6_dev *idev;
1728
1729         idev = __in6_dev_get(dst_dev);
1730         if (idev)
1731                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1732                                                    scores, hiscore_idx);
1733
1734         idev = __in6_dev_get(master);
1735         if (idev)
1736                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1737                                                    scores, hiscore_idx);
1738
1739         return hiscore_idx;
1740 }
1741
1742 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1743                        const struct in6_addr *daddr, unsigned int prefs,
1744                        struct in6_addr *saddr)
1745 {
1746         struct ipv6_saddr_score scores[2], *hiscore;
1747         struct ipv6_saddr_dst dst;
1748         struct inet6_dev *idev;
1749         struct net_device *dev;
1750         int dst_type;
1751         bool use_oif_addr = false;
1752         int hiscore_idx = 0;
1753         int ret = 0;
1754
1755         dst_type = __ipv6_addr_type(daddr);
1756         dst.addr = daddr;
1757         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1758         dst.scope = __ipv6_addr_src_scope(dst_type);
1759         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1760         dst.prefs = prefs;
1761
1762         scores[hiscore_idx].rule = -1;
1763         scores[hiscore_idx].ifa = NULL;
1764
1765         rcu_read_lock();
1766
1767         /* Candidate Source Address (section 4)
1768          *  - multicast and link-local destination address,
1769          *    the set of candidate source address MUST only
1770          *    include addresses assigned to interfaces
1771          *    belonging to the same link as the outgoing
1772          *    interface.
1773          * (- For site-local destination addresses, the
1774          *    set of candidate source addresses MUST only
1775          *    include addresses assigned to interfaces
1776          *    belonging to the same site as the outgoing
1777          *    interface.)
1778          *  - "It is RECOMMENDED that the candidate source addresses
1779          *    be the set of unicast addresses assigned to the
1780          *    interface that will be used to send to the destination
1781          *    (the 'outgoing' interface)." (RFC 6724)
1782          */
1783         if (dst_dev) {
1784                 idev = __in6_dev_get(dst_dev);
1785                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1786                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1787                     (idev && idev->cnf.use_oif_addrs_only)) {
1788                         use_oif_addr = true;
1789                 }
1790         }
1791
1792         if (use_oif_addr) {
1793                 if (idev)
1794                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1795         } else {
1796                 const struct net_device *master;
1797                 int master_idx = 0;
1798
1799                 /* if dst_dev exists and is enslaved to an L3 device, then
1800                  * prefer addresses from dst_dev and then the master over
1801                  * any other enslaved devices in the L3 domain.
1802                  */
1803                 master = l3mdev_master_dev_rcu(dst_dev);
1804                 if (master) {
1805                         master_idx = master->ifindex;
1806
1807                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1808                                                             master, &dst,
1809                                                             scores, hiscore_idx);
1810
1811                         if (scores[hiscore_idx].ifa)
1812                                 goto out;
1813                 }
1814
1815                 for_each_netdev_rcu(net, dev) {
1816                         /* only consider addresses on devices in the
1817                          * same L3 domain
1818                          */
1819                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1820                                 continue;
1821                         idev = __in6_dev_get(dev);
1822                         if (!idev)
1823                                 continue;
1824                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1825                 }
1826         }
1827
1828 out:
1829         hiscore = &scores[hiscore_idx];
1830         if (!hiscore->ifa)
1831                 ret = -EADDRNOTAVAIL;
1832         else
1833                 *saddr = hiscore->ifa->addr;
1834
1835         rcu_read_unlock();
1836         return ret;
1837 }
1838 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1839
1840 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1841                       u32 banned_flags)
1842 {
1843         struct inet6_ifaddr *ifp;
1844         int err = -EADDRNOTAVAIL;
1845
1846         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1847                 if (ifp->scope > IFA_LINK)
1848                         break;
1849                 if (ifp->scope == IFA_LINK &&
1850                     !(ifp->flags & banned_flags)) {
1851                         *addr = ifp->addr;
1852                         err = 0;
1853                         break;
1854                 }
1855         }
1856         return err;
1857 }
1858
1859 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1860                     u32 banned_flags)
1861 {
1862         struct inet6_dev *idev;
1863         int err = -EADDRNOTAVAIL;
1864
1865         rcu_read_lock();
1866         idev = __in6_dev_get(dev);
1867         if (idev) {
1868                 read_lock_bh(&idev->lock);
1869                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1870                 read_unlock_bh(&idev->lock);
1871         }
1872         rcu_read_unlock();
1873         return err;
1874 }
1875
1876 static int ipv6_count_addresses(const struct inet6_dev *idev)
1877 {
1878         const struct inet6_ifaddr *ifp;
1879         int cnt = 0;
1880
1881         rcu_read_lock();
1882         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1883                 cnt++;
1884         rcu_read_unlock();
1885         return cnt;
1886 }
1887
1888 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1889                   const struct net_device *dev, int strict)
1890 {
1891         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1892                                        strict, IFA_F_TENTATIVE);
1893 }
1894 EXPORT_SYMBOL(ipv6_chk_addr);
1895
1896 /* device argument is used to find the L3 domain of interest. If
1897  * skip_dev_check is set, then the ifp device is not checked against
1898  * the passed in dev argument. So the 2 cases for addresses checks are:
1899  *   1. does the address exist in the L3 domain that dev is part of
1900  *      (skip_dev_check = true), or
1901  *
1902  *   2. does the address exist on the specific device
1903  *      (skip_dev_check = false)
1904  */
1905 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1906                             const struct net_device *dev, bool skip_dev_check,
1907                             int strict, u32 banned_flags)
1908 {
1909         unsigned int hash = inet6_addr_hash(net, addr);
1910         const struct net_device *l3mdev;
1911         struct inet6_ifaddr *ifp;
1912         u32 ifp_flags;
1913
1914         rcu_read_lock();
1915
1916         l3mdev = l3mdev_master_dev_rcu(dev);
1917         if (skip_dev_check)
1918                 dev = NULL;
1919
1920         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1921                 if (!net_eq(dev_net(ifp->idev->dev), net))
1922                         continue;
1923
1924                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1925                         continue;
1926
1927                 /* Decouple optimistic from tentative for evaluation here.
1928                  * Ban optimistic addresses explicitly, when required.
1929                  */
1930                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1931                             ? (ifp->flags&~IFA_F_TENTATIVE)
1932                             : ifp->flags;
1933                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1934                     !(ifp_flags&banned_flags) &&
1935                     (!dev || ifp->idev->dev == dev ||
1936                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1937                         rcu_read_unlock();
1938                         return 1;
1939                 }
1940         }
1941
1942         rcu_read_unlock();
1943         return 0;
1944 }
1945 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1946
1947
1948 /* Compares an address/prefix_len with addresses on device @dev.
1949  * If one is found it returns true.
1950  */
1951 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1952         const unsigned int prefix_len, struct net_device *dev)
1953 {
1954         const struct inet6_ifaddr *ifa;
1955         const struct inet6_dev *idev;
1956         bool ret = false;
1957
1958         rcu_read_lock();
1959         idev = __in6_dev_get(dev);
1960         if (idev) {
1961                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1962                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1963                         if (ret)
1964                                 break;
1965                 }
1966         }
1967         rcu_read_unlock();
1968
1969         return ret;
1970 }
1971 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1972
1973 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1974 {
1975         const struct inet6_ifaddr *ifa;
1976         const struct inet6_dev *idev;
1977         int     onlink;
1978
1979         onlink = 0;
1980         rcu_read_lock();
1981         idev = __in6_dev_get(dev);
1982         if (idev) {
1983                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1984                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1985                                                    ifa->prefix_len);
1986                         if (onlink)
1987                                 break;
1988                 }
1989         }
1990         rcu_read_unlock();
1991         return onlink;
1992 }
1993 EXPORT_SYMBOL(ipv6_chk_prefix);
1994
1995 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1996                                      struct net_device *dev, int strict)
1997 {
1998         unsigned int hash = inet6_addr_hash(net, addr);
1999         struct inet6_ifaddr *ifp, *result = NULL;
2000
2001         rcu_read_lock();
2002         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2003                 if (!net_eq(dev_net(ifp->idev->dev), net))
2004                         continue;
2005                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2006                         if (!dev || ifp->idev->dev == dev ||
2007                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2008                                 result = ifp;
2009                                 in6_ifa_hold(ifp);
2010                                 break;
2011                         }
2012                 }
2013         }
2014         rcu_read_unlock();
2015
2016         return result;
2017 }
2018
2019 /* Gets referenced address, destroys ifaddr */
2020
2021 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2022 {
2023         if (dad_failed)
2024                 ifp->flags |= IFA_F_DADFAILED;
2025
2026         if (ifp->flags&IFA_F_TEMPORARY) {
2027                 struct inet6_ifaddr *ifpub;
2028                 spin_lock_bh(&ifp->lock);
2029                 ifpub = ifp->ifpub;
2030                 if (ifpub) {
2031                         in6_ifa_hold(ifpub);
2032                         spin_unlock_bh(&ifp->lock);
2033                         ipv6_create_tempaddr(ifpub, ifp, true);
2034                         in6_ifa_put(ifpub);
2035                 } else {
2036                         spin_unlock_bh(&ifp->lock);
2037                 }
2038                 ipv6_del_addr(ifp);
2039         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2040                 spin_lock_bh(&ifp->lock);
2041                 addrconf_del_dad_work(ifp);
2042                 ifp->flags |= IFA_F_TENTATIVE;
2043                 if (dad_failed)
2044                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2045                 spin_unlock_bh(&ifp->lock);
2046                 if (dad_failed)
2047                         ipv6_ifa_notify(0, ifp);
2048                 in6_ifa_put(ifp);
2049         } else {
2050                 ipv6_del_addr(ifp);
2051         }
2052 }
2053
2054 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2055 {
2056         int err = -ENOENT;
2057
2058         spin_lock_bh(&ifp->lock);
2059         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2060                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2061                 err = 0;
2062         }
2063         spin_unlock_bh(&ifp->lock);
2064
2065         return err;
2066 }
2067
2068 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2069 {
2070         struct inet6_dev *idev = ifp->idev;
2071         struct net *net = dev_net(ifp->idev->dev);
2072
2073         if (addrconf_dad_end(ifp)) {
2074                 in6_ifa_put(ifp);
2075                 return;
2076         }
2077
2078         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2079                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2080
2081         spin_lock_bh(&ifp->lock);
2082
2083         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2084                 struct in6_addr new_addr;
2085                 struct inet6_ifaddr *ifp2;
2086                 int retries = ifp->stable_privacy_retry + 1;
2087                 struct ifa6_config cfg = {
2088                         .pfx = &new_addr,
2089                         .plen = ifp->prefix_len,
2090                         .ifa_flags = ifp->flags,
2091                         .valid_lft = ifp->valid_lft,
2092                         .preferred_lft = ifp->prefered_lft,
2093                         .scope = ifp->scope,
2094                 };
2095
2096                 if (retries > net->ipv6.sysctl.idgen_retries) {
2097                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2098                                              ifp->idev->dev->name);
2099                         goto errdad;
2100                 }
2101
2102                 new_addr = ifp->addr;
2103                 if (ipv6_generate_stable_address(&new_addr, retries,
2104                                                  idev))
2105                         goto errdad;
2106
2107                 spin_unlock_bh(&ifp->lock);
2108
2109                 if (idev->cnf.max_addresses &&
2110                     ipv6_count_addresses(idev) >=
2111                     idev->cnf.max_addresses)
2112                         goto lock_errdad;
2113
2114                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2115                                      ifp->idev->dev->name);
2116
2117                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2118                 if (IS_ERR(ifp2))
2119                         goto lock_errdad;
2120
2121                 spin_lock_bh(&ifp2->lock);
2122                 ifp2->stable_privacy_retry = retries;
2123                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2124                 spin_unlock_bh(&ifp2->lock);
2125
2126                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2127                 in6_ifa_put(ifp2);
2128 lock_errdad:
2129                 spin_lock_bh(&ifp->lock);
2130         }
2131
2132 errdad:
2133         /* transition from _POSTDAD to _ERRDAD */
2134         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2135         spin_unlock_bh(&ifp->lock);
2136
2137         addrconf_mod_dad_work(ifp, 0);
2138         in6_ifa_put(ifp);
2139 }
2140
2141 /* Join to solicited addr multicast group.
2142  * caller must hold RTNL */
2143 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2144 {
2145         struct in6_addr maddr;
2146
2147         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2148                 return;
2149
2150         addrconf_addr_solict_mult(addr, &maddr);
2151         ipv6_dev_mc_inc(dev, &maddr);
2152 }
2153
2154 /* caller must hold RTNL */
2155 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2156 {
2157         struct in6_addr maddr;
2158
2159         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2160                 return;
2161
2162         addrconf_addr_solict_mult(addr, &maddr);
2163         __ipv6_dev_mc_dec(idev, &maddr);
2164 }
2165
2166 /* caller must hold RTNL */
2167 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2168 {
2169         struct in6_addr addr;
2170
2171         if (ifp->prefix_len >= 127) /* RFC 6164 */
2172                 return;
2173         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2174         if (ipv6_addr_any(&addr))
2175                 return;
2176         __ipv6_dev_ac_inc(ifp->idev, &addr);
2177 }
2178
2179 /* caller must hold RTNL */
2180 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2181 {
2182         struct in6_addr addr;
2183
2184         if (ifp->prefix_len >= 127) /* RFC 6164 */
2185                 return;
2186         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2187         if (ipv6_addr_any(&addr))
2188                 return;
2189         __ipv6_dev_ac_dec(ifp->idev, &addr);
2190 }
2191
2192 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2193 {
2194         switch (dev->addr_len) {
2195         case ETH_ALEN:
2196                 memcpy(eui, dev->dev_addr, 3);
2197                 eui[3] = 0xFF;
2198                 eui[4] = 0xFE;
2199                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2200                 break;
2201         case EUI64_ADDR_LEN:
2202                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2203                 eui[0] ^= 2;
2204                 break;
2205         default:
2206                 return -1;
2207         }
2208
2209         return 0;
2210 }
2211
2212 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2213 {
2214         union fwnet_hwaddr *ha;
2215
2216         if (dev->addr_len != FWNET_ALEN)
2217                 return -1;
2218
2219         ha = (union fwnet_hwaddr *)dev->dev_addr;
2220
2221         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2222         eui[0] ^= 2;
2223         return 0;
2224 }
2225
2226 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2227 {
2228         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2229         if (dev->addr_len != ARCNET_ALEN)
2230                 return -1;
2231         memset(eui, 0, 7);
2232         eui[7] = *(u8 *)dev->dev_addr;
2233         return 0;
2234 }
2235
2236 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2237 {
2238         if (dev->addr_len != INFINIBAND_ALEN)
2239                 return -1;
2240         memcpy(eui, dev->dev_addr + 12, 8);
2241         eui[0] |= 2;
2242         return 0;
2243 }
2244
2245 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2246 {
2247         if (addr == 0)
2248                 return -1;
2249         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2250                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2251                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2252                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2253                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2254                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2255         eui[1] = 0;
2256         eui[2] = 0x5E;
2257         eui[3] = 0xFE;
2258         memcpy(eui + 4, &addr, 4);
2259         return 0;
2260 }
2261
2262 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2263 {
2264         if (dev->priv_flags & IFF_ISATAP)
2265                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2266         return -1;
2267 }
2268
2269 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2270 {
2271         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2272 }
2273
2274 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2275 {
2276         memcpy(eui, dev->perm_addr, 3);
2277         memcpy(eui + 5, dev->perm_addr + 3, 3);
2278         eui[3] = 0xFF;
2279         eui[4] = 0xFE;
2280         eui[0] ^= 2;
2281         return 0;
2282 }
2283
2284 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2285 {
2286         switch (dev->type) {
2287         case ARPHRD_ETHER:
2288         case ARPHRD_FDDI:
2289                 return addrconf_ifid_eui48(eui, dev);
2290         case ARPHRD_ARCNET:
2291                 return addrconf_ifid_arcnet(eui, dev);
2292         case ARPHRD_INFINIBAND:
2293                 return addrconf_ifid_infiniband(eui, dev);
2294         case ARPHRD_SIT:
2295                 return addrconf_ifid_sit(eui, dev);
2296         case ARPHRD_IPGRE:
2297         case ARPHRD_TUNNEL:
2298                 return addrconf_ifid_gre(eui, dev);
2299         case ARPHRD_6LOWPAN:
2300                 return addrconf_ifid_6lowpan(eui, dev);
2301         case ARPHRD_IEEE1394:
2302                 return addrconf_ifid_ieee1394(eui, dev);
2303         case ARPHRD_TUNNEL6:
2304         case ARPHRD_IP6GRE:
2305         case ARPHRD_RAWIP:
2306                 return addrconf_ifid_ip6tnl(eui, dev);
2307         }
2308         return -1;
2309 }
2310
2311 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2312 {
2313         int err = -1;
2314         struct inet6_ifaddr *ifp;
2315
2316         read_lock_bh(&idev->lock);
2317         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2318                 if (ifp->scope > IFA_LINK)
2319                         break;
2320                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2321                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2322                         err = 0;
2323                         break;
2324                 }
2325         }
2326         read_unlock_bh(&idev->lock);
2327         return err;
2328 }
2329
2330 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2331 static void ipv6_regen_rndid(struct inet6_dev *idev)
2332 {
2333 regen:
2334         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2335         idev->rndid[0] &= ~0x02;
2336
2337         /*
2338          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2339          * check if generated address is not inappropriate
2340          *
2341          *  - Reserved subnet anycast (RFC 2526)
2342          *      11111101 11....11 1xxxxxxx
2343          *  - ISATAP (RFC4214) 6.1
2344          *      00-00-5E-FE-xx-xx-xx-xx
2345          *  - value 0
2346          *  - XXX: already assigned to an address on the device
2347          */
2348         if (idev->rndid[0] == 0xfd &&
2349             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2350             (idev->rndid[7]&0x80))
2351                 goto regen;
2352         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2353                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2354                         goto regen;
2355                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2356                         goto regen;
2357         }
2358 }
2359
2360 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2361 {
2362         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2363                 ipv6_regen_rndid(idev);
2364 }
2365
2366 /*
2367  *      Add prefix route.
2368  */
2369
2370 static void
2371 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2372                       struct net_device *dev, unsigned long expires,
2373                       u32 flags, gfp_t gfp_flags)
2374 {
2375         struct fib6_config cfg = {
2376                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2377                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2378                 .fc_ifindex = dev->ifindex,
2379                 .fc_expires = expires,
2380                 .fc_dst_len = plen,
2381                 .fc_flags = RTF_UP | flags,
2382                 .fc_nlinfo.nl_net = dev_net(dev),
2383                 .fc_protocol = RTPROT_KERNEL,
2384                 .fc_type = RTN_UNICAST,
2385         };
2386
2387         cfg.fc_dst = *pfx;
2388
2389         /* Prevent useless cloning on PtP SIT.
2390            This thing is done here expecting that the whole
2391            class of non-broadcast devices need not cloning.
2392          */
2393 #if IS_ENABLED(CONFIG_IPV6_SIT)
2394         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2395                 cfg.fc_flags |= RTF_NONEXTHOP;
2396 #endif
2397
2398         ip6_route_add(&cfg, gfp_flags, NULL);
2399 }
2400
2401
2402 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2403                                                   int plen,
2404                                                   const struct net_device *dev,
2405                                                   u32 flags, u32 noflags,
2406                                                   bool no_gw)
2407 {
2408         struct fib6_node *fn;
2409         struct fib6_info *rt = NULL;
2410         struct fib6_table *table;
2411         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2412
2413         table = fib6_get_table(dev_net(dev), tb_id);
2414         if (!table)
2415                 return NULL;
2416
2417         rcu_read_lock();
2418         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2419         if (!fn)
2420                 goto out;
2421
2422         for_each_fib6_node_rt_rcu(fn) {
2423                 /* prefix routes only use builtin fib6_nh */
2424                 if (rt->nh)
2425                         continue;
2426
2427                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2428                         continue;
2429                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2430                         continue;
2431                 if ((rt->fib6_flags & flags) != flags)
2432                         continue;
2433                 if ((rt->fib6_flags & noflags) != 0)
2434                         continue;
2435                 if (!fib6_info_hold_safe(rt))
2436                         continue;
2437                 break;
2438         }
2439 out:
2440         rcu_read_unlock();
2441         return rt;
2442 }
2443
2444
2445 /* Create "default" multicast route to the interface */
2446
2447 static void addrconf_add_mroute(struct net_device *dev)
2448 {
2449         struct fib6_config cfg = {
2450                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2451                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2452                 .fc_ifindex = dev->ifindex,
2453                 .fc_dst_len = 8,
2454                 .fc_flags = RTF_UP,
2455                 .fc_type = RTN_UNICAST,
2456                 .fc_nlinfo.nl_net = dev_net(dev),
2457         };
2458
2459         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2460
2461         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2462 }
2463
2464 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2465 {
2466         struct inet6_dev *idev;
2467
2468         ASSERT_RTNL();
2469
2470         idev = ipv6_find_idev(dev);
2471         if (IS_ERR(idev))
2472                 return idev;
2473
2474         if (idev->cnf.disable_ipv6)
2475                 return ERR_PTR(-EACCES);
2476
2477         /* Add default multicast route */
2478         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2479                 addrconf_add_mroute(dev);
2480
2481         return idev;
2482 }
2483
2484 static void manage_tempaddrs(struct inet6_dev *idev,
2485                              struct inet6_ifaddr *ifp,
2486                              __u32 valid_lft, __u32 prefered_lft,
2487                              bool create, unsigned long now)
2488 {
2489         u32 flags;
2490         struct inet6_ifaddr *ift;
2491
2492         read_lock_bh(&idev->lock);
2493         /* update all temporary addresses in the list */
2494         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2495                 int age, max_valid, max_prefered;
2496
2497                 if (ifp != ift->ifpub)
2498                         continue;
2499
2500                 /* RFC 4941 section 3.3:
2501                  * If a received option will extend the lifetime of a public
2502                  * address, the lifetimes of temporary addresses should
2503                  * be extended, subject to the overall constraint that no
2504                  * temporary addresses should ever remain "valid" or "preferred"
2505                  * for a time longer than (TEMP_VALID_LIFETIME) or
2506                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2507                  */
2508                 age = (now - ift->cstamp) / HZ;
2509                 max_valid = idev->cnf.temp_valid_lft - age;
2510                 if (max_valid < 0)
2511                         max_valid = 0;
2512
2513                 max_prefered = idev->cnf.temp_prefered_lft -
2514                                idev->desync_factor - age;
2515                 if (max_prefered < 0)
2516                         max_prefered = 0;
2517
2518                 if (valid_lft > max_valid)
2519                         valid_lft = max_valid;
2520
2521                 if (prefered_lft > max_prefered)
2522                         prefered_lft = max_prefered;
2523
2524                 spin_lock(&ift->lock);
2525                 flags = ift->flags;
2526                 ift->valid_lft = valid_lft;
2527                 ift->prefered_lft = prefered_lft;
2528                 ift->tstamp = now;
2529                 if (prefered_lft > 0)
2530                         ift->flags &= ~IFA_F_DEPRECATED;
2531
2532                 spin_unlock(&ift->lock);
2533                 if (!(flags&IFA_F_TENTATIVE))
2534                         ipv6_ifa_notify(0, ift);
2535         }
2536
2537         if ((create || list_empty(&idev->tempaddr_list)) &&
2538             idev->cnf.use_tempaddr > 0) {
2539                 /* When a new public address is created as described
2540                  * in [ADDRCONF], also create a new temporary address.
2541                  * Also create a temporary address if it's enabled but
2542                  * no temporary address currently exists.
2543                  */
2544                 read_unlock_bh(&idev->lock);
2545                 ipv6_create_tempaddr(ifp, NULL, false);
2546         } else {
2547                 read_unlock_bh(&idev->lock);
2548         }
2549 }
2550
2551 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2552 {
2553         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2554                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2555 }
2556
2557 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2558                                  const struct prefix_info *pinfo,
2559                                  struct inet6_dev *in6_dev,
2560                                  const struct in6_addr *addr, int addr_type,
2561                                  u32 addr_flags, bool sllao, bool tokenized,
2562                                  __u32 valid_lft, u32 prefered_lft)
2563 {
2564         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2565         int create = 0, update_lft = 0;
2566
2567         if (!ifp && valid_lft) {
2568                 int max_addresses = in6_dev->cnf.max_addresses;
2569                 struct ifa6_config cfg = {
2570                         .pfx = addr,
2571                         .plen = pinfo->prefix_len,
2572                         .ifa_flags = addr_flags,
2573                         .valid_lft = valid_lft,
2574                         .preferred_lft = prefered_lft,
2575                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2576                 };
2577
2578 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2579                 if ((net->ipv6.devconf_all->optimistic_dad ||
2580                      in6_dev->cnf.optimistic_dad) &&
2581                     !net->ipv6.devconf_all->forwarding && sllao)
2582                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2583 #endif
2584
2585                 /* Do not allow to create too much of autoconfigured
2586                  * addresses; this would be too easy way to crash kernel.
2587                  */
2588                 if (!max_addresses ||
2589                     ipv6_count_addresses(in6_dev) < max_addresses)
2590                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2591
2592                 if (IS_ERR_OR_NULL(ifp))
2593                         return -1;
2594
2595                 create = 1;
2596                 spin_lock_bh(&ifp->lock);
2597                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2598                 ifp->cstamp = jiffies;
2599                 ifp->tokenized = tokenized;
2600                 spin_unlock_bh(&ifp->lock);
2601                 addrconf_dad_start(ifp);
2602         }
2603
2604         if (ifp) {
2605                 u32 flags;
2606                 unsigned long now;
2607                 u32 stored_lft;
2608
2609                 /* update lifetime (RFC2462 5.5.3 e) */
2610                 spin_lock_bh(&ifp->lock);
2611                 now = jiffies;
2612                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2613                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2614                 else
2615                         stored_lft = 0;
2616                 if (!create && stored_lft) {
2617                         const u32 minimum_lft = min_t(u32,
2618                                 stored_lft, MIN_VALID_LIFETIME);
2619                         valid_lft = max(valid_lft, minimum_lft);
2620
2621                         /* RFC4862 Section 5.5.3e:
2622                          * "Note that the preferred lifetime of the
2623                          *  corresponding address is always reset to
2624                          *  the Preferred Lifetime in the received
2625                          *  Prefix Information option, regardless of
2626                          *  whether the valid lifetime is also reset or
2627                          *  ignored."
2628                          *
2629                          * So we should always update prefered_lft here.
2630                          */
2631                         update_lft = 1;
2632                 }
2633
2634                 if (update_lft) {
2635                         ifp->valid_lft = valid_lft;
2636                         ifp->prefered_lft = prefered_lft;
2637                         ifp->tstamp = now;
2638                         flags = ifp->flags;
2639                         ifp->flags &= ~IFA_F_DEPRECATED;
2640                         spin_unlock_bh(&ifp->lock);
2641
2642                         if (!(flags&IFA_F_TENTATIVE))
2643                                 ipv6_ifa_notify(0, ifp);
2644                 } else
2645                         spin_unlock_bh(&ifp->lock);
2646
2647                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2648                                  create, now);
2649
2650                 in6_ifa_put(ifp);
2651                 addrconf_verify();
2652         }
2653
2654         return 0;
2655 }
2656 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2657
2658 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2659 {
2660         struct prefix_info *pinfo;
2661         __u32 valid_lft;
2662         __u32 prefered_lft;
2663         int addr_type, err;
2664         u32 addr_flags = 0;
2665         struct inet6_dev *in6_dev;
2666         struct net *net = dev_net(dev);
2667
2668         pinfo = (struct prefix_info *) opt;
2669
2670         if (len < sizeof(struct prefix_info)) {
2671                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2672                 return;
2673         }
2674
2675         /*
2676          *      Validation checks ([ADDRCONF], page 19)
2677          */
2678
2679         addr_type = ipv6_addr_type(&pinfo->prefix);
2680
2681         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2682                 return;
2683
2684         valid_lft = ntohl(pinfo->valid);
2685         prefered_lft = ntohl(pinfo->prefered);
2686
2687         if (prefered_lft > valid_lft) {
2688                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2689                 return;
2690         }
2691
2692         in6_dev = in6_dev_get(dev);
2693
2694         if (!in6_dev) {
2695                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2696                                     dev->name);
2697                 return;
2698         }
2699
2700         /*
2701          *      Two things going on here:
2702          *      1) Add routes for on-link prefixes
2703          *      2) Configure prefixes with the auto flag set
2704          */
2705
2706         if (pinfo->onlink) {
2707                 struct fib6_info *rt;
2708                 unsigned long rt_expires;
2709
2710                 /* Avoid arithmetic overflow. Really, we could
2711                  * save rt_expires in seconds, likely valid_lft,
2712                  * but it would require division in fib gc, that it
2713                  * not good.
2714                  */
2715                 if (HZ > USER_HZ)
2716                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2717                 else
2718                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2719
2720                 if (addrconf_finite_timeout(rt_expires))
2721                         rt_expires *= HZ;
2722
2723                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2724                                                pinfo->prefix_len,
2725                                                dev,
2726                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2727                                                RTF_DEFAULT, true);
2728
2729                 if (rt) {
2730                         /* Autoconf prefix route */
2731                         if (valid_lft == 0) {
2732                                 ip6_del_rt(net, rt);
2733                                 rt = NULL;
2734                         } else if (addrconf_finite_timeout(rt_expires)) {
2735                                 /* not infinity */
2736                                 fib6_set_expires(rt, jiffies + rt_expires);
2737                         } else {
2738                                 fib6_clean_expires(rt);
2739                         }
2740                 } else if (valid_lft) {
2741                         clock_t expires = 0;
2742                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2743                         if (addrconf_finite_timeout(rt_expires)) {
2744                                 /* not infinity */
2745                                 flags |= RTF_EXPIRES;
2746                                 expires = jiffies_to_clock_t(rt_expires);
2747                         }
2748                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2749                                               0, dev, expires, flags,
2750                                               GFP_ATOMIC);
2751                 }
2752                 fib6_info_release(rt);
2753         }
2754
2755         /* Try to figure out our local address for this prefix */
2756
2757         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2758                 struct in6_addr addr;
2759                 bool tokenized = false, dev_addr_generated = false;
2760
2761                 if (pinfo->prefix_len == 64) {
2762                         memcpy(&addr, &pinfo->prefix, 8);
2763
2764                         if (!ipv6_addr_any(&in6_dev->token)) {
2765                                 read_lock_bh(&in6_dev->lock);
2766                                 memcpy(addr.s6_addr + 8,
2767                                        in6_dev->token.s6_addr + 8, 8);
2768                                 read_unlock_bh(&in6_dev->lock);
2769                                 tokenized = true;
2770                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2771                                    !ipv6_generate_stable_address(&addr, 0,
2772                                                                  in6_dev)) {
2773                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2774                                 goto ok;
2775                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2776                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2777                                 goto put;
2778                         } else {
2779                                 dev_addr_generated = true;
2780                         }
2781                         goto ok;
2782                 }
2783                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2784                                     pinfo->prefix_len);
2785                 goto put;
2786
2787 ok:
2788                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2789                                                    &addr, addr_type,
2790                                                    addr_flags, sllao,
2791                                                    tokenized, valid_lft,
2792                                                    prefered_lft);
2793                 if (err)
2794                         goto put;
2795
2796                 /* Ignore error case here because previous prefix add addr was
2797                  * successful which will be notified.
2798                  */
2799                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2800                                               addr_type, addr_flags, sllao,
2801                                               tokenized, valid_lft,
2802                                               prefered_lft,
2803                                               dev_addr_generated);
2804         }
2805         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2806 put:
2807         in6_dev_put(in6_dev);
2808 }
2809
2810 /*
2811  *      Set destination address.
2812  *      Special case for SIT interfaces where we create a new "virtual"
2813  *      device.
2814  */
2815 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2816 {
2817         struct in6_ifreq ireq;
2818         struct net_device *dev;
2819         int err = -EINVAL;
2820
2821         rtnl_lock();
2822
2823         err = -EFAULT;
2824         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2825                 goto err_exit;
2826
2827         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2828
2829         err = -ENODEV;
2830         if (!dev)
2831                 goto err_exit;
2832
2833 #if IS_ENABLED(CONFIG_IPV6_SIT)
2834         if (dev->type == ARPHRD_SIT) {
2835                 const struct net_device_ops *ops = dev->netdev_ops;
2836                 struct ifreq ifr;
2837                 struct ip_tunnel_parm p;
2838
2839                 err = -EADDRNOTAVAIL;
2840                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2841                         goto err_exit;
2842
2843                 memset(&p, 0, sizeof(p));
2844                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2845                 p.iph.saddr = 0;
2846                 p.iph.version = 4;
2847                 p.iph.ihl = 5;
2848                 p.iph.protocol = IPPROTO_IPV6;
2849                 p.iph.ttl = 64;
2850                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2851
2852                 if (ops->ndo_do_ioctl) {
2853                         mm_segment_t oldfs = get_fs();
2854
2855                         set_fs(KERNEL_DS);
2856                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2857                         set_fs(oldfs);
2858                 } else
2859                         err = -EOPNOTSUPP;
2860
2861                 if (err == 0) {
2862                         err = -ENOBUFS;
2863                         dev = __dev_get_by_name(net, p.name);
2864                         if (!dev)
2865                                 goto err_exit;
2866                         err = dev_open(dev, NULL);
2867                 }
2868         }
2869 #endif
2870
2871 err_exit:
2872         rtnl_unlock();
2873         return err;
2874 }
2875
2876 static int ipv6_mc_config(struct sock *sk, bool join,
2877                           const struct in6_addr *addr, int ifindex)
2878 {
2879         int ret;
2880
2881         ASSERT_RTNL();
2882
2883         lock_sock(sk);
2884         if (join)
2885                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2886         else
2887                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2888         release_sock(sk);
2889
2890         return ret;
2891 }
2892
2893 /*
2894  *      Manual configuration of address on an interface
2895  */
2896 static int inet6_addr_add(struct net *net, int ifindex,
2897                           struct ifa6_config *cfg,
2898                           struct netlink_ext_ack *extack)
2899 {
2900         struct inet6_ifaddr *ifp;
2901         struct inet6_dev *idev;
2902         struct net_device *dev;
2903         unsigned long timeout;
2904         clock_t expires;
2905         u32 flags;
2906
2907         ASSERT_RTNL();
2908
2909         if (cfg->plen > 128)
2910                 return -EINVAL;
2911
2912         /* check the lifetime */
2913         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2914                 return -EINVAL;
2915
2916         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2917                 return -EINVAL;
2918
2919         dev = __dev_get_by_index(net, ifindex);
2920         if (!dev)
2921                 return -ENODEV;
2922
2923         idev = addrconf_add_dev(dev);
2924         if (IS_ERR(idev))
2925                 return PTR_ERR(idev);
2926
2927         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2928                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2929                                          true, cfg->pfx, ifindex);
2930
2931                 if (ret < 0)
2932                         return ret;
2933         }
2934
2935         cfg->scope = ipv6_addr_scope(cfg->pfx);
2936
2937         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2938         if (addrconf_finite_timeout(timeout)) {
2939                 expires = jiffies_to_clock_t(timeout * HZ);
2940                 cfg->valid_lft = timeout;
2941                 flags = RTF_EXPIRES;
2942         } else {
2943                 expires = 0;
2944                 flags = 0;
2945                 cfg->ifa_flags |= IFA_F_PERMANENT;
2946         }
2947
2948         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2949         if (addrconf_finite_timeout(timeout)) {
2950                 if (timeout == 0)
2951                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2952                 cfg->preferred_lft = timeout;
2953         }
2954
2955         ifp = ipv6_add_addr(idev, cfg, true, extack);
2956         if (!IS_ERR(ifp)) {
2957                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2958                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2959                                               ifp->rt_priority, dev, expires,
2960                                               flags, GFP_KERNEL);
2961                 }
2962
2963                 /* Send a netlink notification if DAD is enabled and
2964                  * optimistic flag is not set
2965                  */
2966                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2967                         ipv6_ifa_notify(0, ifp);
2968                 /*
2969                  * Note that section 3.1 of RFC 4429 indicates
2970                  * that the Optimistic flag should not be set for
2971                  * manually configured addresses
2972                  */
2973                 addrconf_dad_start(ifp);
2974                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2975                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2976                                          cfg->preferred_lft, true, jiffies);
2977                 in6_ifa_put(ifp);
2978                 addrconf_verify_rtnl();
2979                 return 0;
2980         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2981                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2982                                cfg->pfx, ifindex);
2983         }
2984
2985         return PTR_ERR(ifp);
2986 }
2987
2988 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2989                           const struct in6_addr *pfx, unsigned int plen)
2990 {
2991         struct inet6_ifaddr *ifp;
2992         struct inet6_dev *idev;
2993         struct net_device *dev;
2994
2995         if (plen > 128)
2996                 return -EINVAL;
2997
2998         dev = __dev_get_by_index(net, ifindex);
2999         if (!dev)
3000                 return -ENODEV;
3001
3002         idev = __in6_dev_get(dev);
3003         if (!idev)
3004                 return -ENXIO;
3005
3006         read_lock_bh(&idev->lock);
3007         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3008                 if (ifp->prefix_len == plen &&
3009                     ipv6_addr_equal(pfx, &ifp->addr)) {
3010                         in6_ifa_hold(ifp);
3011                         read_unlock_bh(&idev->lock);
3012
3013                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3014                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3015                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3016                                                  jiffies);
3017                         ipv6_del_addr(ifp);
3018                         addrconf_verify_rtnl();
3019                         if (ipv6_addr_is_multicast(pfx)) {
3020                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3021                                                false, pfx, dev->ifindex);
3022                         }
3023                         return 0;
3024                 }
3025         }
3026         read_unlock_bh(&idev->lock);
3027         return -EADDRNOTAVAIL;
3028 }
3029
3030
3031 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3032 {
3033         struct ifa6_config cfg = {
3034                 .ifa_flags = IFA_F_PERMANENT,
3035                 .preferred_lft = INFINITY_LIFE_TIME,
3036                 .valid_lft = INFINITY_LIFE_TIME,
3037         };
3038         struct in6_ifreq ireq;
3039         int err;
3040
3041         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3042                 return -EPERM;
3043
3044         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3045                 return -EFAULT;
3046
3047         cfg.pfx = &ireq.ifr6_addr;
3048         cfg.plen = ireq.ifr6_prefixlen;
3049
3050         rtnl_lock();
3051         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3052         rtnl_unlock();
3053         return err;
3054 }
3055
3056 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3057 {
3058         struct in6_ifreq ireq;
3059         int err;
3060
3061         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3062                 return -EPERM;
3063
3064         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3065                 return -EFAULT;
3066
3067         rtnl_lock();
3068         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3069                              ireq.ifr6_prefixlen);
3070         rtnl_unlock();
3071         return err;
3072 }
3073
3074 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3075                      int plen, int scope)
3076 {
3077         struct inet6_ifaddr *ifp;
3078         struct ifa6_config cfg = {
3079                 .pfx = addr,
3080                 .plen = plen,
3081                 .ifa_flags = IFA_F_PERMANENT,
3082                 .valid_lft = INFINITY_LIFE_TIME,
3083                 .preferred_lft = INFINITY_LIFE_TIME,
3084                 .scope = scope
3085         };
3086
3087         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3088         if (!IS_ERR(ifp)) {
3089                 spin_lock_bh(&ifp->lock);
3090                 ifp->flags &= ~IFA_F_TENTATIVE;
3091                 spin_unlock_bh(&ifp->lock);
3092                 rt_genid_bump_ipv6(dev_net(idev->dev));
3093                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3094                 in6_ifa_put(ifp);
3095         }
3096 }
3097
3098 #if IS_ENABLED(CONFIG_IPV6_SIT)
3099 static void sit_add_v4_addrs(struct inet6_dev *idev)
3100 {
3101         struct in6_addr addr;
3102         struct net_device *dev;
3103         struct net *net = dev_net(idev->dev);
3104         int scope, plen;
3105         u32 pflags = 0;
3106
3107         ASSERT_RTNL();
3108
3109         memset(&addr, 0, sizeof(struct in6_addr));
3110         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3111
3112         if (idev->dev->flags&IFF_POINTOPOINT) {
3113                 addr.s6_addr32[0] = htonl(0xfe800000);
3114                 scope = IFA_LINK;
3115                 plen = 64;
3116         } else {
3117                 scope = IPV6_ADDR_COMPATv4;
3118                 plen = 96;
3119                 pflags |= RTF_NONEXTHOP;
3120         }
3121
3122         if (addr.s6_addr32[3]) {
3123                 add_addr(idev, &addr, plen, scope);
3124                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3125                                       GFP_KERNEL);
3126                 return;
3127         }
3128
3129         for_each_netdev(net, dev) {
3130                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3131                 if (in_dev && (dev->flags & IFF_UP)) {
3132                         struct in_ifaddr *ifa;
3133                         int flag = scope;
3134
3135                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3136                                 addr.s6_addr32[3] = ifa->ifa_local;
3137
3138                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3139                                         continue;
3140                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3141                                         if (idev->dev->flags&IFF_POINTOPOINT)
3142                                                 continue;
3143                                         flag |= IFA_HOST;
3144                                 }
3145
3146                                 add_addr(idev, &addr, plen, flag);
3147                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3148                                                       0, pflags, GFP_KERNEL);
3149                         }
3150                 }
3151         }
3152 }
3153 #endif
3154
3155 static void init_loopback(struct net_device *dev)
3156 {
3157         struct inet6_dev  *idev;
3158
3159         /* ::1 */
3160
3161         ASSERT_RTNL();
3162
3163         idev = ipv6_find_idev(dev);
3164         if (IS_ERR(idev)) {
3165                 pr_debug("%s: add_dev failed\n", __func__);
3166                 return;
3167         }
3168
3169         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3170 }
3171
3172 void addrconf_add_linklocal(struct inet6_dev *idev,
3173                             const struct in6_addr *addr, u32 flags)
3174 {
3175         struct ifa6_config cfg = {
3176                 .pfx = addr,
3177                 .plen = 64,
3178                 .ifa_flags = flags | IFA_F_PERMANENT,
3179                 .valid_lft = INFINITY_LIFE_TIME,
3180                 .preferred_lft = INFINITY_LIFE_TIME,
3181                 .scope = IFA_LINK
3182         };
3183         struct inet6_ifaddr *ifp;
3184
3185 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3186         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3187              idev->cnf.optimistic_dad) &&
3188             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3189                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3190 #endif
3191
3192         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3193         if (!IS_ERR(ifp)) {
3194                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3195                                       0, 0, GFP_ATOMIC);
3196                 addrconf_dad_start(ifp);
3197                 in6_ifa_put(ifp);
3198         }
3199 }
3200 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3201
3202 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3203 {
3204         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3205                 return true;
3206
3207         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3208             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3209                 return true;
3210
3211         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3212             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3213                 return true;
3214
3215         return false;
3216 }
3217
3218 static int ipv6_generate_stable_address(struct in6_addr *address,
3219                                         u8 dad_count,
3220                                         const struct inet6_dev *idev)
3221 {
3222         static DEFINE_SPINLOCK(lock);
3223         static __u32 digest[SHA_DIGEST_WORDS];
3224         static __u32 workspace[SHA_WORKSPACE_WORDS];
3225
3226         static union {
3227                 char __data[SHA_MESSAGE_BYTES];
3228                 struct {
3229                         struct in6_addr secret;
3230                         __be32 prefix[2];
3231                         unsigned char hwaddr[MAX_ADDR_LEN];
3232                         u8 dad_count;
3233                 } __packed;
3234         } data;
3235
3236         struct in6_addr secret;
3237         struct in6_addr temp;
3238         struct net *net = dev_net(idev->dev);
3239
3240         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3241
3242         if (idev->cnf.stable_secret.initialized)
3243                 secret = idev->cnf.stable_secret.secret;
3244         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3245                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3246         else
3247                 return -1;
3248
3249 retry:
3250         spin_lock_bh(&lock);
3251
3252         sha_init(digest);
3253         memset(&data, 0, sizeof(data));
3254         memset(workspace, 0, sizeof(workspace));
3255         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3256         data.prefix[0] = address->s6_addr32[0];
3257         data.prefix[1] = address->s6_addr32[1];
3258         data.secret = secret;
3259         data.dad_count = dad_count;
3260
3261         sha_transform(digest, data.__data, workspace);
3262
3263         temp = *address;
3264         temp.s6_addr32[2] = (__force __be32)digest[0];
3265         temp.s6_addr32[3] = (__force __be32)digest[1];
3266
3267         spin_unlock_bh(&lock);
3268
3269         if (ipv6_reserved_interfaceid(temp)) {
3270                 dad_count++;
3271                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3272                         return -1;
3273                 goto retry;
3274         }
3275
3276         *address = temp;
3277         return 0;
3278 }
3279
3280 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3281 {
3282         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3283
3284         if (s->initialized)
3285                 return;
3286         s = &idev->cnf.stable_secret;
3287         get_random_bytes(&s->secret, sizeof(s->secret));
3288         s->initialized = true;
3289 }
3290
3291 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3292 {
3293         struct in6_addr addr;
3294
3295         /* no link local addresses on L3 master devices */
3296         if (netif_is_l3_master(idev->dev))
3297                 return;
3298
3299         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3300
3301         switch (idev->cnf.addr_gen_mode) {
3302         case IN6_ADDR_GEN_MODE_RANDOM:
3303                 ipv6_gen_mode_random_init(idev);
3304                 /* fallthrough */
3305         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3306                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3307                         addrconf_add_linklocal(idev, &addr,
3308                                                IFA_F_STABLE_PRIVACY);
3309                 else if (prefix_route)
3310                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3311                                               0, 0, GFP_KERNEL);
3312                 break;
3313         case IN6_ADDR_GEN_MODE_EUI64:
3314                 /* addrconf_add_linklocal also adds a prefix_route and we
3315                  * only need to care about prefix routes if ipv6_generate_eui64
3316                  * couldn't generate one.
3317                  */
3318                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3319                         addrconf_add_linklocal(idev, &addr, 0);
3320                 else if (prefix_route)
3321                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3322                                               0, 0, GFP_KERNEL);
3323                 break;
3324         case IN6_ADDR_GEN_MODE_NONE:
3325         default:
3326                 /* will not add any link local address */
3327                 break;
3328         }
3329 }
3330
3331 static void addrconf_dev_config(struct net_device *dev)
3332 {
3333         struct inet6_dev *idev;
3334
3335         ASSERT_RTNL();
3336
3337         if ((dev->type != ARPHRD_ETHER) &&
3338             (dev->type != ARPHRD_FDDI) &&
3339             (dev->type != ARPHRD_ARCNET) &&
3340             (dev->type != ARPHRD_INFINIBAND) &&
3341             (dev->type != ARPHRD_IEEE1394) &&
3342             (dev->type != ARPHRD_TUNNEL6) &&
3343             (dev->type != ARPHRD_6LOWPAN) &&
3344             (dev->type != ARPHRD_IP6GRE) &&
3345             (dev->type != ARPHRD_IPGRE) &&
3346             (dev->type != ARPHRD_TUNNEL) &&
3347             (dev->type != ARPHRD_NONE) &&
3348             (dev->type != ARPHRD_RAWIP)) {
3349                 /* Alas, we support only Ethernet autoconfiguration. */
3350                 idev = __in6_dev_get(dev);
3351                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3352                     dev->flags & IFF_MULTICAST)
3353                         ipv6_mc_up(idev);
3354                 return;
3355         }
3356
3357         idev = addrconf_add_dev(dev);
3358         if (IS_ERR(idev))
3359                 return;
3360
3361         /* this device type has no EUI support */
3362         if (dev->type == ARPHRD_NONE &&
3363             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3364                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3365
3366         addrconf_addr_gen(idev, false);
3367 }
3368
3369 #if IS_ENABLED(CONFIG_IPV6_SIT)
3370 static void addrconf_sit_config(struct net_device *dev)
3371 {
3372         struct inet6_dev *idev;
3373
3374         ASSERT_RTNL();
3375
3376         /*
3377          * Configure the tunnel with one of our IPv4
3378          * addresses... we should configure all of
3379          * our v4 addrs in the tunnel
3380          */
3381
3382         idev = ipv6_find_idev(dev);
3383         if (IS_ERR(idev)) {
3384                 pr_debug("%s: add_dev failed\n", __func__);
3385                 return;
3386         }
3387
3388         if (dev->priv_flags & IFF_ISATAP) {
3389                 addrconf_addr_gen(idev, false);
3390                 return;
3391         }
3392
3393         sit_add_v4_addrs(idev);
3394
3395         if (dev->flags&IFF_POINTOPOINT)
3396                 addrconf_add_mroute(dev);
3397 }
3398 #endif
3399
3400 #if IS_ENABLED(CONFIG_NET_IPGRE)
3401 static void addrconf_gre_config(struct net_device *dev)
3402 {
3403         struct inet6_dev *idev;
3404
3405         ASSERT_RTNL();
3406
3407         idev = ipv6_find_idev(dev);
3408         if (IS_ERR(idev)) {
3409                 pr_debug("%s: add_dev failed\n", __func__);
3410                 return;
3411         }
3412
3413         addrconf_addr_gen(idev, true);
3414         if (dev->flags & IFF_POINTOPOINT)
3415                 addrconf_add_mroute(dev);
3416 }
3417 #endif
3418
3419 static int fixup_permanent_addr(struct net *net,
3420                                 struct inet6_dev *idev,
3421                                 struct inet6_ifaddr *ifp)
3422 {
3423         /* !fib6_node means the host route was removed from the
3424          * FIB, for example, if 'lo' device is taken down. In that
3425          * case regenerate the host route.
3426          */
3427         if (!ifp->rt || !ifp->rt->fib6_node) {
3428                 struct fib6_info *f6i, *prev;
3429
3430                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3431                                          GFP_ATOMIC);
3432                 if (IS_ERR(f6i))
3433                         return PTR_ERR(f6i);
3434
3435                 /* ifp->rt can be accessed outside of rtnl */
3436                 spin_lock(&ifp->lock);
3437                 prev = ifp->rt;
3438                 ifp->rt = f6i;
3439                 spin_unlock(&ifp->lock);
3440
3441                 fib6_info_release(prev);
3442         }
3443
3444         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3445                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3446                                       ifp->rt_priority, idev->dev, 0, 0,
3447                                       GFP_ATOMIC);
3448         }
3449
3450         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3451                 addrconf_dad_start(ifp);
3452
3453         return 0;
3454 }
3455
3456 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3457 {
3458         struct inet6_ifaddr *ifp, *tmp;
3459         struct inet6_dev *idev;
3460
3461         idev = __in6_dev_get(dev);
3462         if (!idev)
3463                 return;
3464
3465         write_lock_bh(&idev->lock);
3466
3467         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3468                 if ((ifp->flags & IFA_F_PERMANENT) &&
3469                     fixup_permanent_addr(net, idev, ifp) < 0) {
3470                         write_unlock_bh(&idev->lock);
3471                         in6_ifa_hold(ifp);
3472                         ipv6_del_addr(ifp);
3473                         write_lock_bh(&idev->lock);
3474
3475                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3476                                              idev->dev->name, &ifp->addr);
3477                 }
3478         }
3479
3480         write_unlock_bh(&idev->lock);
3481 }
3482
3483 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3484                            void *ptr)
3485 {
3486         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3487         struct netdev_notifier_change_info *change_info;
3488         struct netdev_notifier_changeupper_info *info;
3489         struct inet6_dev *idev = __in6_dev_get(dev);
3490         struct net *net = dev_net(dev);
3491         int run_pending = 0;
3492         int err;
3493
3494         switch (event) {
3495         case NETDEV_REGISTER:
3496                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3497                         idev = ipv6_add_dev(dev);
3498                         if (IS_ERR(idev))
3499                                 return notifier_from_errno(PTR_ERR(idev));
3500                 }
3501                 break;
3502
3503         case NETDEV_CHANGEMTU:
3504                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3505                 if (dev->mtu < IPV6_MIN_MTU) {
3506                         addrconf_ifdown(dev, dev != net->loopback_dev);
3507                         break;
3508                 }
3509
3510                 if (idev) {
3511                         rt6_mtu_change(dev, dev->mtu);
3512                         idev->cnf.mtu6 = dev->mtu;
3513                         break;
3514                 }
3515
3516                 /* allocate new idev */
3517                 idev = ipv6_add_dev(dev);
3518                 if (IS_ERR(idev))
3519                         break;
3520
3521                 /* device is still not ready */
3522                 if (!(idev->if_flags & IF_READY))
3523                         break;
3524
3525                 run_pending = 1;
3526
3527                 /* fall through */
3528
3529         case NETDEV_UP:
3530         case NETDEV_CHANGE:
3531                 if (dev->flags & IFF_SLAVE)
3532                         break;
3533
3534                 if (idev && idev->cnf.disable_ipv6)
3535                         break;
3536
3537                 if (event == NETDEV_UP) {
3538                         /* restore routes for permanent addresses */
3539                         addrconf_permanent_addr(net, dev);
3540
3541                         if (!addrconf_link_ready(dev)) {
3542                                 /* device is not ready yet. */
3543                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3544                                          dev->name);
3545                                 break;
3546                         }
3547
3548                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3549                                 idev = ipv6_add_dev(dev);
3550
3551                         if (!IS_ERR_OR_NULL(idev)) {
3552                                 idev->if_flags |= IF_READY;
3553                                 run_pending = 1;
3554                         }
3555                 } else if (event == NETDEV_CHANGE) {
3556                         if (!addrconf_link_ready(dev)) {
3557                                 /* device is still not ready. */
3558                                 rt6_sync_down_dev(dev, event);
3559                                 break;
3560                         }
3561
3562                         if (!IS_ERR_OR_NULL(idev)) {
3563                                 if (idev->if_flags & IF_READY) {
3564                                         /* device is already configured -
3565                                          * but resend MLD reports, we might
3566                                          * have roamed and need to update
3567                                          * multicast snooping switches
3568                                          */
3569                                         ipv6_mc_up(idev);
3570                                         change_info = ptr;
3571                                         if (change_info->flags_changed & IFF_NOARP)
3572                                                 addrconf_dad_run(idev, true);
3573                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3574                                         break;
3575                                 }
3576                                 idev->if_flags |= IF_READY;
3577                         }
3578
3579                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3580                                 dev->name);
3581
3582                         run_pending = 1;
3583                 }
3584
3585                 switch (dev->type) {
3586 #if IS_ENABLED(CONFIG_IPV6_SIT)
3587                 case ARPHRD_SIT:
3588                         addrconf_sit_config(dev);
3589                         break;
3590 #endif
3591 #if IS_ENABLED(CONFIG_NET_IPGRE)
3592                 case ARPHRD_IPGRE:
3593                         addrconf_gre_config(dev);
3594                         break;
3595 #endif
3596                 case ARPHRD_LOOPBACK:
3597                         init_loopback(dev);
3598                         break;
3599
3600                 default:
3601                         addrconf_dev_config(dev);
3602                         break;
3603                 }
3604
3605                 if (!IS_ERR_OR_NULL(idev)) {
3606                         if (run_pending)
3607                                 addrconf_dad_run(idev, false);
3608
3609                         /* Device has an address by now */
3610                         rt6_sync_up(dev, RTNH_F_DEAD);
3611
3612                         /*
3613                          * If the MTU changed during the interface down,
3614                          * when the interface up, the changed MTU must be
3615                          * reflected in the idev as well as routers.
3616                          */
3617                         if (idev->cnf.mtu6 != dev->mtu &&
3618                             dev->mtu >= IPV6_MIN_MTU) {
3619                                 rt6_mtu_change(dev, dev->mtu);
3620                                 idev->cnf.mtu6 = dev->mtu;
3621                         }
3622                         idev->tstamp = jiffies;
3623                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3624
3625                         /*
3626                          * If the changed mtu during down is lower than
3627                          * IPV6_MIN_MTU stop IPv6 on this interface.
3628                          */
3629                         if (dev->mtu < IPV6_MIN_MTU)
3630                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3631                 }
3632                 break;
3633
3634         case NETDEV_DOWN:
3635         case NETDEV_UNREGISTER:
3636                 /*
3637                  *      Remove all addresses from this interface.
3638                  */
3639                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3640                 break;
3641
3642         case NETDEV_CHANGENAME:
3643                 if (idev) {
3644                         snmp6_unregister_dev(idev);
3645                         addrconf_sysctl_unregister(idev);
3646                         err = addrconf_sysctl_register(idev);
3647                         if (err)
3648                                 return notifier_from_errno(err);
3649                         err = snmp6_register_dev(idev);
3650                         if (err) {
3651                                 addrconf_sysctl_unregister(idev);
3652                                 return notifier_from_errno(err);
3653                         }
3654                 }
3655                 break;
3656
3657         case NETDEV_PRE_TYPE_CHANGE:
3658         case NETDEV_POST_TYPE_CHANGE:
3659                 if (idev)
3660                         addrconf_type_change(dev, event);
3661                 break;
3662
3663         case NETDEV_CHANGEUPPER:
3664                 info = ptr;
3665
3666                 /* flush all routes if dev is linked to or unlinked from
3667                  * an L3 master device (e.g., VRF)
3668                  */
3669                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3670                         addrconf_ifdown(dev, 0);
3671         }
3672
3673         return NOTIFY_OK;
3674 }
3675
3676 /*
3677  *      addrconf module should be notified of a device going up
3678  */
3679 static struct notifier_block ipv6_dev_notf = {
3680         .notifier_call = addrconf_notify,
3681         .priority = ADDRCONF_NOTIFY_PRIORITY,
3682 };
3683
3684 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3685 {
3686         struct inet6_dev *idev;
3687         ASSERT_RTNL();
3688
3689         idev = __in6_dev_get(dev);
3690
3691         if (event == NETDEV_POST_TYPE_CHANGE)
3692                 ipv6_mc_remap(idev);
3693         else if (event == NETDEV_PRE_TYPE_CHANGE)
3694                 ipv6_mc_unmap(idev);
3695 }
3696
3697 static bool addr_is_local(const struct in6_addr *addr)
3698 {
3699         return ipv6_addr_type(addr) &
3700                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3701 }
3702
3703 static int addrconf_ifdown(struct net_device *dev, int how)
3704 {
3705         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3706         struct net *net = dev_net(dev);
3707         struct inet6_dev *idev;
3708         struct inet6_ifaddr *ifa, *tmp;
3709         bool keep_addr = false;
3710         int state, i;
3711
3712         ASSERT_RTNL();
3713
3714         rt6_disable_ip(dev, event);
3715
3716         idev = __in6_dev_get(dev);
3717         if (!idev)
3718                 return -ENODEV;
3719
3720         /*
3721          * Step 1: remove reference to ipv6 device from parent device.
3722          *         Do not dev_put!
3723          */
3724         if (how) {
3725                 idev->dead = 1;
3726
3727                 /* protected by rtnl_lock */
3728                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3729
3730                 /* Step 1.5: remove snmp6 entry */
3731                 snmp6_unregister_dev(idev);
3732
3733         }
3734
3735         /* combine the user config with event to determine if permanent
3736          * addresses are to be removed from address hash table
3737          */
3738         if (!how && !idev->cnf.disable_ipv6) {
3739                 /* aggregate the system setting and interface setting */
3740                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3741
3742                 if (!_keep_addr)
3743                         _keep_addr = idev->cnf.keep_addr_on_down;
3744
3745                 keep_addr = (_keep_addr > 0);
3746         }
3747
3748         /* Step 2: clear hash table */
3749         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3750                 struct hlist_head *h = &inet6_addr_lst[i];
3751
3752                 spin_lock_bh(&addrconf_hash_lock);
3753 restart:
3754                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3755                         if (ifa->idev == idev) {
3756                                 addrconf_del_dad_work(ifa);
3757                                 /* combined flag + permanent flag decide if
3758                                  * address is retained on a down event
3759                                  */
3760                                 if (!keep_addr ||
3761                                     !(ifa->flags & IFA_F_PERMANENT) ||
3762                                     addr_is_local(&ifa->addr)) {
3763                                         hlist_del_init_rcu(&ifa->addr_lst);
3764                                         goto restart;
3765                                 }
3766                         }
3767                 }
3768                 spin_unlock_bh(&addrconf_hash_lock);
3769         }
3770
3771         write_lock_bh(&idev->lock);
3772
3773         addrconf_del_rs_timer(idev);
3774
3775         /* Step 2: clear flags for stateless addrconf */
3776         if (!how)
3777                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3778
3779         /* Step 3: clear tempaddr list */
3780         while (!list_empty(&idev->tempaddr_list)) {
3781                 ifa = list_first_entry(&idev->tempaddr_list,
3782                                        struct inet6_ifaddr, tmp_list);
3783                 list_del(&ifa->tmp_list);
3784                 write_unlock_bh(&idev->lock);
3785                 spin_lock_bh(&ifa->lock);
3786
3787                 if (ifa->ifpub) {
3788                         in6_ifa_put(ifa->ifpub);
3789                         ifa->ifpub = NULL;
3790                 }
3791                 spin_unlock_bh(&ifa->lock);
3792                 in6_ifa_put(ifa);
3793                 write_lock_bh(&idev->lock);
3794         }
3795
3796         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3797                 struct fib6_info *rt = NULL;
3798                 bool keep;
3799
3800                 addrconf_del_dad_work(ifa);
3801
3802                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3803                         !addr_is_local(&ifa->addr);
3804
3805                 write_unlock_bh(&idev->lock);
3806                 spin_lock_bh(&ifa->lock);
3807
3808                 if (keep) {
3809                         /* set state to skip the notifier below */
3810                         state = INET6_IFADDR_STATE_DEAD;
3811                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3812                         if (!(ifa->flags & IFA_F_NODAD))
3813                                 ifa->flags |= IFA_F_TENTATIVE;
3814
3815                         rt = ifa->rt;
3816                         ifa->rt = NULL;
3817                 } else {
3818                         state = ifa->state;
3819                         ifa->state = INET6_IFADDR_STATE_DEAD;
3820                 }
3821
3822                 spin_unlock_bh(&ifa->lock);
3823
3824                 if (rt)
3825                         ip6_del_rt(net, rt);
3826
3827                 if (state != INET6_IFADDR_STATE_DEAD) {
3828                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3829                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3830                 } else {
3831                         if (idev->cnf.forwarding)
3832                                 addrconf_leave_anycast(ifa);
3833                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3834                 }
3835
3836                 write_lock_bh(&idev->lock);
3837                 if (!keep) {
3838                         list_del_rcu(&ifa->if_list);
3839                         in6_ifa_put(ifa);
3840                 }
3841         }
3842
3843         write_unlock_bh(&idev->lock);
3844
3845         /* Step 5: Discard anycast and multicast list */
3846         if (how) {
3847                 ipv6_ac_destroy_dev(idev);
3848                 ipv6_mc_destroy_dev(idev);
3849         } else {
3850                 ipv6_mc_down(idev);
3851         }
3852
3853         idev->tstamp = jiffies;
3854
3855         /* Last: Shot the device (if unregistered) */
3856         if (how) {
3857                 addrconf_sysctl_unregister(idev);
3858                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3859                 neigh_ifdown(&nd_tbl, dev);
3860                 in6_dev_put(idev);
3861         }
3862         return 0;
3863 }
3864
3865 static void addrconf_rs_timer(struct timer_list *t)
3866 {
3867         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3868         struct net_device *dev = idev->dev;
3869         struct in6_addr lladdr;
3870
3871         write_lock(&idev->lock);
3872         if (idev->dead || !(idev->if_flags & IF_READY))
3873                 goto out;
3874
3875         if (!ipv6_accept_ra(idev))
3876                 goto out;
3877
3878         /* Announcement received after solicitation was sent */
3879         if (idev->if_flags & IF_RA_RCVD)
3880                 goto out;
3881
3882         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3883                 write_unlock(&idev->lock);
3884                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3885                         ndisc_send_rs(dev, &lladdr,
3886                                       &in6addr_linklocal_allrouters);
3887                 else
3888                         goto put;
3889
3890                 write_lock(&idev->lock);
3891                 idev->rs_interval = rfc3315_s14_backoff_update(
3892                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3893                 /* The wait after the last probe can be shorter */
3894                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3895                                              idev->cnf.rtr_solicits) ?
3896                                       idev->cnf.rtr_solicit_delay :
3897                                       idev->rs_interval);
3898         } else {
3899                 /*
3900                  * Note: we do not support deprecated "all on-link"
3901                  * assumption any longer.
3902                  */
3903                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3904         }
3905
3906 out:
3907         write_unlock(&idev->lock);
3908 put:
3909         in6_dev_put(idev);
3910 }
3911
3912 /*
3913  *      Duplicate Address Detection
3914  */
3915 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3916 {
3917         unsigned long rand_num;
3918         struct inet6_dev *idev = ifp->idev;
3919         u64 nonce;
3920
3921         if (ifp->flags & IFA_F_OPTIMISTIC)
3922                 rand_num = 0;
3923         else
3924                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3925
3926         nonce = 0;
3927         if (idev->cnf.enhanced_dad ||
3928             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3929                 do
3930                         get_random_bytes(&nonce, 6);
3931                 while (nonce == 0);
3932         }
3933         ifp->dad_nonce = nonce;
3934         ifp->dad_probes = idev->cnf.dad_transmits;
3935         addrconf_mod_dad_work(ifp, rand_num);
3936 }
3937
3938 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3939 {
3940         struct inet6_dev *idev = ifp->idev;
3941         struct net_device *dev = idev->dev;
3942         bool bump_id, notify = false;
3943         struct net *net;
3944
3945         addrconf_join_solict(dev, &ifp->addr);
3946
3947         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3948
3949         read_lock_bh(&idev->lock);
3950         spin_lock(&ifp->lock);
3951         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3952                 goto out;
3953
3954         net = dev_net(dev);
3955         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3956             (net->ipv6.devconf_all->accept_dad < 1 &&
3957              idev->cnf.accept_dad < 1) ||
3958             !(ifp->flags&IFA_F_TENTATIVE) ||
3959             ifp->flags & IFA_F_NODAD) {
3960                 bool send_na = false;
3961
3962                 if (ifp->flags & IFA_F_TENTATIVE &&
3963                     !(ifp->flags & IFA_F_OPTIMISTIC))
3964                         send_na = true;
3965                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3966                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3967                 spin_unlock(&ifp->lock);
3968                 read_unlock_bh(&idev->lock);
3969
3970                 addrconf_dad_completed(ifp, bump_id, send_na);
3971                 return;
3972         }
3973
3974         if (!(idev->if_flags & IF_READY)) {
3975                 spin_unlock(&ifp->lock);
3976                 read_unlock_bh(&idev->lock);
3977                 /*
3978                  * If the device is not ready:
3979                  * - keep it tentative if it is a permanent address.
3980                  * - otherwise, kill it.
3981                  */
3982                 in6_ifa_hold(ifp);
3983                 addrconf_dad_stop(ifp, 0);
3984                 return;
3985         }
3986
3987         /*
3988          * Optimistic nodes can start receiving
3989          * Frames right away
3990          */
3991         if (ifp->flags & IFA_F_OPTIMISTIC) {
3992                 ip6_ins_rt(net, ifp->rt);
3993                 if (ipv6_use_optimistic_addr(net, idev)) {
3994                         /* Because optimistic nodes can use this address,
3995                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3996                          */
3997                         notify = true;
3998                 }
3999         }
4000
4001         addrconf_dad_kick(ifp);
4002 out:
4003         spin_unlock(&ifp->lock);
4004         read_unlock_bh(&idev->lock);
4005         if (notify)
4006                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4007 }
4008
4009 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4010 {
4011         bool begin_dad = false;
4012
4013         spin_lock_bh(&ifp->lock);
4014         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4015                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4016                 begin_dad = true;
4017         }
4018         spin_unlock_bh(&ifp->lock);
4019
4020         if (begin_dad)
4021                 addrconf_mod_dad_work(ifp, 0);
4022 }
4023
4024 static void addrconf_dad_work(struct work_struct *w)
4025 {
4026         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4027                                                 struct inet6_ifaddr,
4028                                                 dad_work);
4029         struct inet6_dev *idev = ifp->idev;
4030         bool bump_id, disable_ipv6 = false;
4031         struct in6_addr mcaddr;
4032
4033         enum {
4034                 DAD_PROCESS,
4035                 DAD_BEGIN,
4036                 DAD_ABORT,
4037         } action = DAD_PROCESS;
4038
4039         rtnl_lock();
4040
4041         spin_lock_bh(&ifp->lock);
4042         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4043                 action = DAD_BEGIN;
4044                 ifp->state = INET6_IFADDR_STATE_DAD;
4045         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4046                 action = DAD_ABORT;
4047                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4048
4049                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4050                      idev->cnf.accept_dad > 1) &&
4051                     !idev->cnf.disable_ipv6 &&
4052                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4053                         struct in6_addr addr;
4054
4055                         addr.s6_addr32[0] = htonl(0xfe800000);
4056                         addr.s6_addr32[1] = 0;
4057
4058                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4059                             ipv6_addr_equal(&ifp->addr, &addr)) {
4060                                 /* DAD failed for link-local based on MAC */
4061                                 idev->cnf.disable_ipv6 = 1;
4062
4063                                 pr_info("%s: IPv6 being disabled!\n",
4064                                         ifp->idev->dev->name);
4065                                 disable_ipv6 = true;
4066                         }
4067                 }
4068         }
4069         spin_unlock_bh(&ifp->lock);
4070
4071         if (action == DAD_BEGIN) {
4072                 addrconf_dad_begin(ifp);
4073                 goto out;
4074         } else if (action == DAD_ABORT) {
4075                 in6_ifa_hold(ifp);
4076                 addrconf_dad_stop(ifp, 1);
4077                 if (disable_ipv6)
4078                         addrconf_ifdown(idev->dev, 0);
4079                 goto out;
4080         }
4081
4082         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4083                 goto out;
4084
4085         write_lock_bh(&idev->lock);
4086         if (idev->dead || !(idev->if_flags & IF_READY)) {
4087                 write_unlock_bh(&idev->lock);
4088                 goto out;
4089         }
4090
4091         spin_lock(&ifp->lock);
4092         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4093                 spin_unlock(&ifp->lock);
4094                 write_unlock_bh(&idev->lock);
4095                 goto out;
4096         }
4097
4098         if (ifp->dad_probes == 0) {
4099                 bool send_na = false;
4100
4101                 /*
4102                  * DAD was successful
4103                  */
4104
4105                 if (ifp->flags & IFA_F_TENTATIVE &&
4106                     !(ifp->flags & IFA_F_OPTIMISTIC))
4107                         send_na = true;
4108                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4109                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4110                 spin_unlock(&ifp->lock);
4111                 write_unlock_bh(&idev->lock);
4112
4113                 addrconf_dad_completed(ifp, bump_id, send_na);
4114
4115                 goto out;
4116         }
4117
4118         ifp->dad_probes--;
4119         addrconf_mod_dad_work(ifp,
4120                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4121         spin_unlock(&ifp->lock);
4122         write_unlock_bh(&idev->lock);
4123
4124         /* send a neighbour solicitation for our addr */
4125         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4126         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4127                       ifp->dad_nonce);
4128 out:
4129         in6_ifa_put(ifp);
4130         rtnl_unlock();
4131 }
4132
4133 /* ifp->idev must be at least read locked */
4134 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4135 {
4136         struct inet6_ifaddr *ifpiter;
4137         struct inet6_dev *idev = ifp->idev;
4138
4139         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4140                 if (ifpiter->scope > IFA_LINK)
4141                         break;
4142                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4143                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4144                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4145                     IFA_F_PERMANENT)
4146                         return false;
4147         }
4148         return true;
4149 }
4150
4151 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4152                                    bool send_na)
4153 {
4154         struct net_device *dev = ifp->idev->dev;
4155         struct in6_addr lladdr;
4156         bool send_rs, send_mld;
4157
4158         addrconf_del_dad_work(ifp);
4159
4160         /*
4161          *      Configure the address for reception. Now it is valid.
4162          */
4163
4164         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4165
4166         /* If added prefix is link local and we are prepared to process
4167            router advertisements, start sending router solicitations.
4168          */
4169
4170         read_lock_bh(&ifp->idev->lock);
4171         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4172         send_rs = send_mld &&
4173                   ipv6_accept_ra(ifp->idev) &&
4174                   ifp->idev->cnf.rtr_solicits != 0 &&
4175                   (dev->flags&IFF_LOOPBACK) == 0;
4176         read_unlock_bh(&ifp->idev->lock);
4177
4178         /* While dad is in progress mld report's source address is in6_addrany.
4179          * Resend with proper ll now.
4180          */
4181         if (send_mld)
4182                 ipv6_mc_dad_complete(ifp->idev);
4183
4184         /* send unsolicited NA if enabled */
4185         if (send_na &&
4186             (ifp->idev->cnf.ndisc_notify ||
4187              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4188                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4189                               /*router=*/ !!ifp->idev->cnf.forwarding,
4190                               /*solicited=*/ false, /*override=*/ true,
4191                               /*inc_opt=*/ true);
4192         }
4193
4194         if (send_rs) {
4195                 /*
4196                  *      If a host as already performed a random delay
4197                  *      [...] as part of DAD [...] there is no need
4198                  *      to delay again before sending the first RS
4199                  */
4200                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4201                         return;
4202                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4203
4204                 write_lock_bh(&ifp->idev->lock);
4205                 spin_lock(&ifp->lock);
4206                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4207                         ifp->idev->cnf.rtr_solicit_interval);
4208                 ifp->idev->rs_probes = 1;
4209                 ifp->idev->if_flags |= IF_RS_SENT;
4210                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4211                 spin_unlock(&ifp->lock);
4212                 write_unlock_bh(&ifp->idev->lock);
4213         }
4214
4215         if (bump_id)
4216                 rt_genid_bump_ipv6(dev_net(dev));
4217
4218         /* Make sure that a new temporary address will be created
4219          * before this temporary address becomes deprecated.
4220          */
4221         if (ifp->flags & IFA_F_TEMPORARY)
4222                 addrconf_verify_rtnl();
4223 }
4224
4225 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4226 {
4227         struct inet6_ifaddr *ifp;
4228
4229         read_lock_bh(&idev->lock);
4230         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4231                 spin_lock(&ifp->lock);
4232                 if ((ifp->flags & IFA_F_TENTATIVE &&
4233                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4234                         if (restart)
4235                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4236                         addrconf_dad_kick(ifp);
4237                 }
4238                 spin_unlock(&ifp->lock);
4239         }
4240         read_unlock_bh(&idev->lock);
4241 }
4242
4243 #ifdef CONFIG_PROC_FS
4244 struct if6_iter_state {
4245         struct seq_net_private p;
4246         int bucket;
4247         int offset;
4248 };
4249
4250 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4251 {
4252         struct if6_iter_state *state = seq->private;
4253         struct net *net = seq_file_net(seq);
4254         struct inet6_ifaddr *ifa = NULL;
4255         int p = 0;
4256
4257         /* initial bucket if pos is 0 */
4258         if (pos == 0) {
4259                 state->bucket = 0;
4260                 state->offset = 0;
4261         }
4262
4263         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4264                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4265                                          addr_lst) {
4266                         if (!net_eq(dev_net(ifa->idev->dev), net))
4267                                 continue;
4268                         /* sync with offset */
4269                         if (p < state->offset) {
4270                                 p++;
4271                                 continue;
4272                         }
4273                         return ifa;
4274                 }
4275
4276                 /* prepare for next bucket */
4277                 state->offset = 0;
4278                 p = 0;
4279         }
4280         return NULL;
4281 }
4282
4283 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4284                                          struct inet6_ifaddr *ifa)
4285 {
4286         struct if6_iter_state *state = seq->private;
4287         struct net *net = seq_file_net(seq);
4288
4289         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4290                 if (!net_eq(dev_net(ifa->idev->dev), net))
4291                         continue;
4292                 state->offset++;
4293                 return ifa;
4294         }
4295
4296         state->offset = 0;
4297         while (++state->bucket < IN6_ADDR_HSIZE) {
4298                 hlist_for_each_entry_rcu(ifa,
4299                                      &inet6_addr_lst[state->bucket], addr_lst) {
4300                         if (!net_eq(dev_net(ifa->idev->dev), net))
4301                                 continue;
4302                         return ifa;
4303                 }
4304         }
4305
4306         return NULL;
4307 }
4308
4309 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4310         __acquires(rcu)
4311 {
4312         rcu_read_lock();
4313         return if6_get_first(seq, *pos);
4314 }
4315
4316 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4317 {
4318         struct inet6_ifaddr *ifa;
4319
4320         ifa = if6_get_next(seq, v);
4321         ++*pos;
4322         return ifa;
4323 }
4324
4325 static void if6_seq_stop(struct seq_file *seq, void *v)
4326         __releases(rcu)
4327 {
4328         rcu_read_unlock();
4329 }
4330
4331 static int if6_seq_show(struct seq_file *seq, void *v)
4332 {
4333         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4334         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4335                    &ifp->addr,
4336                    ifp->idev->dev->ifindex,
4337                    ifp->prefix_len,
4338                    ifp->scope,
4339                    (u8) ifp->flags,
4340                    ifp->idev->dev->name);
4341         return 0;
4342 }
4343
4344 static const struct seq_operations if6_seq_ops = {
4345         .start  = if6_seq_start,
4346         .next   = if6_seq_next,
4347         .show   = if6_seq_show,
4348         .stop   = if6_seq_stop,
4349 };
4350
4351 static int __net_init if6_proc_net_init(struct net *net)
4352 {
4353         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4354                         sizeof(struct if6_iter_state)))
4355                 return -ENOMEM;
4356         return 0;
4357 }
4358
4359 static void __net_exit if6_proc_net_exit(struct net *net)
4360 {
4361         remove_proc_entry("if_inet6", net->proc_net);
4362 }
4363
4364 static struct pernet_operations if6_proc_net_ops = {
4365         .init = if6_proc_net_init,
4366         .exit = if6_proc_net_exit,
4367 };
4368
4369 int __init if6_proc_init(void)
4370 {
4371         return register_pernet_subsys(&if6_proc_net_ops);
4372 }
4373
4374 void if6_proc_exit(void)
4375 {
4376         unregister_pernet_subsys(&if6_proc_net_ops);
4377 }
4378 #endif  /* CONFIG_PROC_FS */
4379
4380 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4381 /* Check if address is a home address configured on any interface. */
4382 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4383 {
4384         unsigned int hash = inet6_addr_hash(net, addr);
4385         struct inet6_ifaddr *ifp = NULL;
4386         int ret = 0;
4387
4388         rcu_read_lock();
4389         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4390                 if (!net_eq(dev_net(ifp->idev->dev), net))
4391                         continue;
4392                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4393                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4394                         ret = 1;
4395                         break;
4396                 }
4397         }
4398         rcu_read_unlock();
4399         return ret;
4400 }
4401 #endif
4402
4403 /*
4404  *      Periodic address status verification
4405  */
4406
4407 static void addrconf_verify_rtnl(void)
4408 {
4409         unsigned long now, next, next_sec, next_sched;
4410         struct inet6_ifaddr *ifp;
4411         int i;
4412
4413         ASSERT_RTNL();
4414
4415         rcu_read_lock_bh();
4416         now = jiffies;
4417         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4418
4419         cancel_delayed_work(&addr_chk_work);
4420
4421         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4422 restart:
4423                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4424                         unsigned long age;
4425
4426                         /* When setting preferred_lft to a value not zero or
4427                          * infinity, while valid_lft is infinity
4428                          * IFA_F_PERMANENT has a non-infinity life time.
4429                          */
4430                         if ((ifp->flags & IFA_F_PERMANENT) &&
4431                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4432                                 continue;
4433
4434                         spin_lock(&ifp->lock);
4435                         /* We try to batch several events at once. */
4436                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4437
4438                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4439                             age >= ifp->valid_lft) {
4440                                 spin_unlock(&ifp->lock);
4441                                 in6_ifa_hold(ifp);
4442                                 ipv6_del_addr(ifp);
4443                                 goto restart;
4444                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4445                                 spin_unlock(&ifp->lock);
4446                                 continue;
4447                         } else if (age >= ifp->prefered_lft) {
4448                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4449                                 int deprecate = 0;
4450
4451                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4452                                         deprecate = 1;
4453                                         ifp->flags |= IFA_F_DEPRECATED;
4454                                 }
4455
4456                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4457                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4458                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4459
4460                                 spin_unlock(&ifp->lock);
4461
4462                                 if (deprecate) {
4463                                         in6_ifa_hold(ifp);
4464
4465                                         ipv6_ifa_notify(0, ifp);
4466                                         in6_ifa_put(ifp);
4467                                         goto restart;
4468                                 }
4469                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4470                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4471                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4472                                         ifp->idev->cnf.dad_transmits *
4473                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4474
4475                                 if (age >= ifp->prefered_lft - regen_advance) {
4476                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4477                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4478                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4479                                         if (!ifp->regen_count && ifpub) {
4480                                                 ifp->regen_count++;
4481                                                 in6_ifa_hold(ifp);
4482                                                 in6_ifa_hold(ifpub);
4483                                                 spin_unlock(&ifp->lock);
4484
4485                                                 spin_lock(&ifpub->lock);
4486                                                 ifpub->regen_count = 0;
4487                                                 spin_unlock(&ifpub->lock);
4488                                                 rcu_read_unlock_bh();
4489                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4490                                                 in6_ifa_put(ifpub);
4491                                                 in6_ifa_put(ifp);
4492                                                 rcu_read_lock_bh();
4493                                                 goto restart;
4494                                         }
4495                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4496                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4497                                 spin_unlock(&ifp->lock);
4498                         } else {
4499                                 /* ifp->prefered_lft <= ifp->valid_lft */
4500                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4501                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4502                                 spin_unlock(&ifp->lock);
4503                         }
4504                 }
4505         }
4506
4507         next_sec = round_jiffies_up(next);
4508         next_sched = next;
4509
4510         /* If rounded timeout is accurate enough, accept it. */
4511         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4512                 next_sched = next_sec;
4513
4514         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4515         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4516                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4517
4518         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4519                  now, next, next_sec, next_sched);
4520         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4521         rcu_read_unlock_bh();
4522 }
4523
4524 static void addrconf_verify_work(struct work_struct *w)
4525 {
4526         rtnl_lock();
4527         addrconf_verify_rtnl();
4528         rtnl_unlock();
4529 }
4530
4531 static void addrconf_verify(void)
4532 {
4533         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4534 }
4535
4536 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4537                                      struct in6_addr **peer_pfx)
4538 {
4539         struct in6_addr *pfx = NULL;
4540
4541         *peer_pfx = NULL;
4542
4543         if (addr)
4544                 pfx = nla_data(addr);
4545
4546         if (local) {
4547                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4548                         *peer_pfx = pfx;
4549                 pfx = nla_data(local);
4550         }
4551
4552         return pfx;
4553 }
4554
4555 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4556         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4557         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4558         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4559         [IFA_FLAGS]             = { .len = sizeof(u32) },
4560         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4561         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4562 };
4563
4564 static int
4565 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4566                   struct netlink_ext_ack *extack)
4567 {
4568         struct net *net = sock_net(skb->sk);
4569         struct ifaddrmsg *ifm;
4570         struct nlattr *tb[IFA_MAX+1];
4571         struct in6_addr *pfx, *peer_pfx;
4572         u32 ifa_flags;
4573         int err;
4574
4575         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4576                                      ifa_ipv6_policy, extack);
4577         if (err < 0)
4578                 return err;
4579
4580         ifm = nlmsg_data(nlh);
4581         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4582         if (!pfx)
4583                 return -EINVAL;
4584
4585         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4586
4587         /* We ignore other flags so far. */
4588         ifa_flags &= IFA_F_MANAGETEMPADDR;
4589
4590         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4591                               ifm->ifa_prefixlen);
4592 }
4593
4594 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4595                                unsigned long expires, u32 flags,
4596                                bool modify_peer)
4597 {
4598         struct fib6_info *f6i;
4599         u32 prio;
4600
4601         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4602                                         ifp->prefix_len,
4603                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4604         if (!f6i)
4605                 return -ENOENT;
4606
4607         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4608         if (f6i->fib6_metric != prio) {
4609                 /* delete old one */
4610                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4611
4612                 /* add new one */
4613                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4614                                       ifp->prefix_len,
4615                                       ifp->rt_priority, ifp->idev->dev,
4616                                       expires, flags, GFP_KERNEL);
4617         } else {
4618                 if (!expires)
4619                         fib6_clean_expires(f6i);
4620                 else
4621                         fib6_set_expires(f6i, expires);
4622
4623                 fib6_info_release(f6i);
4624         }
4625
4626         return 0;
4627 }
4628
4629 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4630 {
4631         u32 flags;
4632         clock_t expires;
4633         unsigned long timeout;
4634         bool was_managetempaddr;
4635         bool had_prefixroute;
4636         bool new_peer = false;
4637
4638         ASSERT_RTNL();
4639
4640         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4641                 return -EINVAL;
4642
4643         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4644             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4645                 return -EINVAL;
4646
4647         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4648                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4649
4650         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4651         if (addrconf_finite_timeout(timeout)) {
4652                 expires = jiffies_to_clock_t(timeout * HZ);
4653                 cfg->valid_lft = timeout;
4654                 flags = RTF_EXPIRES;
4655         } else {
4656                 expires = 0;
4657                 flags = 0;
4658                 cfg->ifa_flags |= IFA_F_PERMANENT;
4659         }
4660
4661         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4662         if (addrconf_finite_timeout(timeout)) {
4663                 if (timeout == 0)
4664                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4665                 cfg->preferred_lft = timeout;
4666         }
4667
4668         if (cfg->peer_pfx &&
4669             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4670                 if (!ipv6_addr_any(&ifp->peer_addr))
4671                         cleanup_prefix_route(ifp, expires, true, true);
4672                 new_peer = true;
4673         }
4674
4675         spin_lock_bh(&ifp->lock);
4676         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4677         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4678                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4679         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4680                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4681                         IFA_F_NOPREFIXROUTE);
4682         ifp->flags |= cfg->ifa_flags;
4683         ifp->tstamp = jiffies;
4684         ifp->valid_lft = cfg->valid_lft;
4685         ifp->prefered_lft = cfg->preferred_lft;
4686
4687         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4688                 ifp->rt_priority = cfg->rt_priority;
4689
4690         if (new_peer)
4691                 ifp->peer_addr = *cfg->peer_pfx;
4692
4693         spin_unlock_bh(&ifp->lock);
4694         if (!(ifp->flags&IFA_F_TENTATIVE))
4695                 ipv6_ifa_notify(0, ifp);
4696
4697         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4698                 int rc = -ENOENT;
4699
4700                 if (had_prefixroute)
4701                         rc = modify_prefix_route(ifp, expires, flags, false);
4702
4703                 /* prefix route could have been deleted; if so restore it */
4704                 if (rc == -ENOENT) {
4705                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4706                                               ifp->rt_priority, ifp->idev->dev,
4707                                               expires, flags, GFP_KERNEL);
4708                 }
4709
4710                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4711                         rc = modify_prefix_route(ifp, expires, flags, true);
4712
4713                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4714                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4715                                               ifp->rt_priority, ifp->idev->dev,
4716                                               expires, flags, GFP_KERNEL);
4717                 }
4718         } else if (had_prefixroute) {
4719                 enum cleanup_prefix_rt_t action;
4720                 unsigned long rt_expires;
4721
4722                 write_lock_bh(&ifp->idev->lock);
4723                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4724                 write_unlock_bh(&ifp->idev->lock);
4725
4726                 if (action != CLEANUP_PREFIX_RT_NOP) {
4727                         cleanup_prefix_route(ifp, rt_expires,
4728                                 action == CLEANUP_PREFIX_RT_DEL, false);
4729                 }
4730         }
4731
4732         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4733                 if (was_managetempaddr &&
4734                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4735                         cfg->valid_lft = 0;
4736                         cfg->preferred_lft = 0;
4737                 }
4738                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4739                                  cfg->preferred_lft, !was_managetempaddr,
4740                                  jiffies);
4741         }
4742
4743         addrconf_verify_rtnl();
4744
4745         return 0;
4746 }
4747
4748 static int
4749 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4750                   struct netlink_ext_ack *extack)
4751 {
4752         struct net *net = sock_net(skb->sk);
4753         struct ifaddrmsg *ifm;
4754         struct nlattr *tb[IFA_MAX+1];
4755         struct in6_addr *peer_pfx;
4756         struct inet6_ifaddr *ifa;
4757         struct net_device *dev;
4758         struct inet6_dev *idev;
4759         struct ifa6_config cfg;
4760         int err;
4761
4762         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4763                                      ifa_ipv6_policy, extack);
4764         if (err < 0)
4765                 return err;
4766
4767         memset(&cfg, 0, sizeof(cfg));
4768
4769         ifm = nlmsg_data(nlh);
4770         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4771         if (!cfg.pfx)
4772                 return -EINVAL;
4773
4774         cfg.peer_pfx = peer_pfx;
4775         cfg.plen = ifm->ifa_prefixlen;
4776         if (tb[IFA_RT_PRIORITY])
4777                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4778
4779         cfg.valid_lft = INFINITY_LIFE_TIME;
4780         cfg.preferred_lft = INFINITY_LIFE_TIME;
4781
4782         if (tb[IFA_CACHEINFO]) {
4783                 struct ifa_cacheinfo *ci;
4784
4785                 ci = nla_data(tb[IFA_CACHEINFO]);
4786                 cfg.valid_lft = ci->ifa_valid;
4787                 cfg.preferred_lft = ci->ifa_prefered;
4788         }
4789
4790         dev =  __dev_get_by_index(net, ifm->ifa_index);
4791         if (!dev)
4792                 return -ENODEV;
4793
4794         if (tb[IFA_FLAGS])
4795                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4796         else
4797                 cfg.ifa_flags = ifm->ifa_flags;
4798
4799         /* We ignore other flags so far. */
4800         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4801                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4802                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4803
4804         idev = ipv6_find_idev(dev);
4805         if (IS_ERR(idev))
4806                 return PTR_ERR(idev);
4807
4808         if (!ipv6_allow_optimistic_dad(net, idev))
4809                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4810
4811         if (cfg.ifa_flags & IFA_F_NODAD &&
4812             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4813                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4814                 return -EINVAL;
4815         }
4816
4817         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4818         if (!ifa) {
4819                 /*
4820                  * It would be best to check for !NLM_F_CREATE here but
4821                  * userspace already relies on not having to provide this.
4822                  */
4823                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4824         }
4825
4826         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4827             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4828                 err = -EEXIST;
4829         else
4830                 err = inet6_addr_modify(ifa, &cfg);
4831
4832         in6_ifa_put(ifa);
4833
4834         return err;
4835 }
4836
4837 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4838                           u8 scope, int ifindex)
4839 {
4840         struct ifaddrmsg *ifm;
4841
4842         ifm = nlmsg_data(nlh);
4843         ifm->ifa_family = AF_INET6;
4844         ifm->ifa_prefixlen = prefixlen;
4845         ifm->ifa_flags = flags;
4846         ifm->ifa_scope = scope;
4847         ifm->ifa_index = ifindex;
4848 }
4849
4850 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4851                          unsigned long tstamp, u32 preferred, u32 valid)
4852 {
4853         struct ifa_cacheinfo ci;
4854
4855         ci.cstamp = cstamp_delta(cstamp);
4856         ci.tstamp = cstamp_delta(tstamp);
4857         ci.ifa_prefered = preferred;
4858         ci.ifa_valid = valid;
4859
4860         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4861 }
4862
4863 static inline int rt_scope(int ifa_scope)
4864 {
4865         if (ifa_scope & IFA_HOST)
4866                 return RT_SCOPE_HOST;
4867         else if (ifa_scope & IFA_LINK)
4868                 return RT_SCOPE_LINK;
4869         else if (ifa_scope & IFA_SITE)
4870                 return RT_SCOPE_SITE;
4871         else
4872                 return RT_SCOPE_UNIVERSE;
4873 }
4874
4875 static inline int inet6_ifaddr_msgsize(void)
4876 {
4877         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4878                + nla_total_size(16) /* IFA_LOCAL */
4879                + nla_total_size(16) /* IFA_ADDRESS */
4880                + nla_total_size(sizeof(struct ifa_cacheinfo))
4881                + nla_total_size(4)  /* IFA_FLAGS */
4882                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4883 }
4884
4885 enum addr_type_t {
4886         UNICAST_ADDR,
4887         MULTICAST_ADDR,
4888         ANYCAST_ADDR,
4889 };
4890
4891 struct inet6_fill_args {
4892         u32 portid;
4893         u32 seq;
4894         int event;
4895         unsigned int flags;
4896         int netnsid;
4897         int ifindex;
4898         enum addr_type_t type;
4899 };
4900
4901 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4902                              struct inet6_fill_args *args)
4903 {
4904         struct nlmsghdr  *nlh;
4905         u32 preferred, valid;
4906
4907         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4908                         sizeof(struct ifaddrmsg), args->flags);
4909         if (!nlh)
4910                 return -EMSGSIZE;
4911
4912         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4913                       ifa->idev->dev->ifindex);
4914
4915         if (args->netnsid >= 0 &&
4916             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4917                 goto error;
4918
4919         if (!((ifa->flags&IFA_F_PERMANENT) &&
4920               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4921                 preferred = ifa->prefered_lft;
4922                 valid = ifa->valid_lft;
4923                 if (preferred != INFINITY_LIFE_TIME) {
4924                         long tval = (jiffies - ifa->tstamp)/HZ;
4925                         if (preferred > tval)
4926                                 preferred -= tval;
4927                         else
4928                                 preferred = 0;
4929                         if (valid != INFINITY_LIFE_TIME) {
4930                                 if (valid > tval)
4931                                         valid -= tval;
4932                                 else
4933                                         valid = 0;
4934                         }
4935                 }
4936         } else {
4937                 preferred = INFINITY_LIFE_TIME;
4938                 valid = INFINITY_LIFE_TIME;
4939         }
4940
4941         if (!ipv6_addr_any(&ifa->peer_addr)) {
4942                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4943                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4944                         goto error;
4945         } else
4946                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4947                         goto error;
4948
4949         if (ifa->rt_priority &&
4950             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4951                 goto error;
4952
4953         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4954                 goto error;
4955
4956         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4957                 goto error;
4958
4959         nlmsg_end(skb, nlh);
4960         return 0;
4961
4962 error:
4963         nlmsg_cancel(skb, nlh);
4964         return -EMSGSIZE;
4965 }
4966
4967 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4968                                struct inet6_fill_args *args)
4969 {
4970         struct nlmsghdr  *nlh;
4971         u8 scope = RT_SCOPE_UNIVERSE;
4972         int ifindex = ifmca->idev->dev->ifindex;
4973
4974         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4975                 scope = RT_SCOPE_SITE;
4976
4977         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4978                         sizeof(struct ifaddrmsg), args->flags);
4979         if (!nlh)
4980                 return -EMSGSIZE;
4981
4982         if (args->netnsid >= 0 &&
4983             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4984                 return -EMSGSIZE;
4985
4986         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4987         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4988             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4989                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4990                 nlmsg_cancel(skb, nlh);
4991                 return -EMSGSIZE;
4992         }
4993
4994         nlmsg_end(skb, nlh);
4995         return 0;
4996 }
4997
4998 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4999                                struct inet6_fill_args *args)
5000 {
5001         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5002         int ifindex = dev ? dev->ifindex : 1;
5003         struct nlmsghdr  *nlh;
5004         u8 scope = RT_SCOPE_UNIVERSE;
5005
5006         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5007                 scope = RT_SCOPE_SITE;
5008
5009         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5010                         sizeof(struct ifaddrmsg), args->flags);
5011         if (!nlh)
5012                 return -EMSGSIZE;
5013
5014         if (args->netnsid >= 0 &&
5015             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5016                 return -EMSGSIZE;
5017
5018         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5019         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5020             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5021                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5022                 nlmsg_cancel(skb, nlh);
5023                 return -EMSGSIZE;
5024         }
5025
5026         nlmsg_end(skb, nlh);
5027         return 0;
5028 }
5029
5030 /* called with rcu_read_lock() */
5031 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5032                           struct netlink_callback *cb, int s_ip_idx,
5033                           struct inet6_fill_args *fillargs)
5034 {
5035         struct ifmcaddr6 *ifmca;
5036         struct ifacaddr6 *ifaca;
5037         int ip_idx = 0;
5038         int err = 1;
5039
5040         read_lock_bh(&idev->lock);
5041         switch (fillargs->type) {
5042         case UNICAST_ADDR: {
5043                 struct inet6_ifaddr *ifa;
5044                 fillargs->event = RTM_NEWADDR;
5045
5046                 /* unicast address incl. temp addr */
5047                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5048                         if (ip_idx < s_ip_idx)
5049                                 goto next;
5050                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5051                         if (err < 0)
5052                                 break;
5053                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5054 next:
5055                         ip_idx++;
5056                 }
5057                 break;
5058         }
5059         case MULTICAST_ADDR:
5060                 fillargs->event = RTM_GETMULTICAST;
5061
5062                 /* multicast address */
5063                 for (ifmca = idev->mc_list; ifmca;
5064                      ifmca = ifmca->next, ip_idx++) {
5065                         if (ip_idx < s_ip_idx)
5066                                 continue;
5067                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5068                         if (err < 0)
5069                                 break;
5070                 }
5071                 break;
5072         case ANYCAST_ADDR:
5073                 fillargs->event = RTM_GETANYCAST;
5074                 /* anycast address */
5075                 for (ifaca = idev->ac_list; ifaca;
5076                      ifaca = ifaca->aca_next, ip_idx++) {
5077                         if (ip_idx < s_ip_idx)
5078                                 continue;
5079                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5080                         if (err < 0)
5081                                 break;
5082                 }
5083                 break;
5084         default:
5085                 break;
5086         }
5087         read_unlock_bh(&idev->lock);
5088         cb->args[2] = ip_idx;
5089         return err;
5090 }
5091
5092 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5093                                        struct inet6_fill_args *fillargs,
5094                                        struct net **tgt_net, struct sock *sk,
5095                                        struct netlink_callback *cb)
5096 {
5097         struct netlink_ext_ack *extack = cb->extack;
5098         struct nlattr *tb[IFA_MAX+1];
5099         struct ifaddrmsg *ifm;
5100         int err, i;
5101
5102         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5103                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5104                 return -EINVAL;
5105         }
5106
5107         ifm = nlmsg_data(nlh);
5108         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5109                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5110                 return -EINVAL;
5111         }
5112
5113         fillargs->ifindex = ifm->ifa_index;
5114         if (fillargs->ifindex) {
5115                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5116                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5117         }
5118
5119         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5120                                             ifa_ipv6_policy, extack);
5121         if (err < 0)
5122                 return err;
5123
5124         for (i = 0; i <= IFA_MAX; ++i) {
5125                 if (!tb[i])
5126                         continue;
5127
5128                 if (i == IFA_TARGET_NETNSID) {
5129                         struct net *net;
5130
5131                         fillargs->netnsid = nla_get_s32(tb[i]);
5132                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5133                         if (IS_ERR(net)) {
5134                                 fillargs->netnsid = -1;
5135                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5136                                 return PTR_ERR(net);
5137                         }
5138                         *tgt_net = net;
5139                 } else {
5140                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5141                         return -EINVAL;
5142                 }
5143         }
5144
5145         return 0;
5146 }
5147
5148 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5149                            enum addr_type_t type)
5150 {
5151         const struct nlmsghdr *nlh = cb->nlh;
5152         struct inet6_fill_args fillargs = {
5153                 .portid = NETLINK_CB(cb->skb).portid,
5154                 .seq = cb->nlh->nlmsg_seq,
5155                 .flags = NLM_F_MULTI,
5156                 .netnsid = -1,
5157                 .type = type,
5158         };
5159         struct net *net = sock_net(skb->sk);
5160         struct net *tgt_net = net;
5161         int idx, s_idx, s_ip_idx;
5162         int h, s_h;
5163         struct net_device *dev;
5164         struct inet6_dev *idev;
5165         struct hlist_head *head;
5166         int err = 0;
5167
5168         s_h = cb->args[0];
5169         s_idx = idx = cb->args[1];
5170         s_ip_idx = cb->args[2];
5171
5172         if (cb->strict_check) {
5173                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5174                                                   skb->sk, cb);
5175                 if (err < 0)
5176                         goto put_tgt_net;
5177
5178                 err = 0;
5179                 if (fillargs.ifindex) {
5180                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5181                         if (!dev) {
5182                                 err = -ENODEV;
5183                                 goto put_tgt_net;
5184                         }
5185                         idev = __in6_dev_get(dev);
5186                         if (idev) {
5187                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5188                                                      &fillargs);
5189                                 if (err > 0)
5190                                         err = 0;
5191                         }
5192                         goto put_tgt_net;
5193                 }
5194         }
5195
5196         rcu_read_lock();
5197         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5198         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5199                 idx = 0;
5200                 head = &tgt_net->dev_index_head[h];
5201                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5202                         if (idx < s_idx)
5203                                 goto cont;
5204                         if (h > s_h || idx > s_idx)
5205                                 s_ip_idx = 0;
5206                         idev = __in6_dev_get(dev);
5207                         if (!idev)
5208                                 goto cont;
5209
5210                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5211                                            &fillargs) < 0)
5212                                 goto done;
5213 cont:
5214                         idx++;
5215                 }
5216         }
5217 done:
5218         rcu_read_unlock();
5219         cb->args[0] = h;
5220         cb->args[1] = idx;
5221 put_tgt_net:
5222         if (fillargs.netnsid >= 0)
5223                 put_net(tgt_net);
5224
5225         return skb->len ? : err;
5226 }
5227
5228 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5229 {
5230         enum addr_type_t type = UNICAST_ADDR;
5231
5232         return inet6_dump_addr(skb, cb, type);
5233 }
5234
5235 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5236 {
5237         enum addr_type_t type = MULTICAST_ADDR;
5238
5239         return inet6_dump_addr(skb, cb, type);
5240 }
5241
5242
5243 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5244 {
5245         enum addr_type_t type = ANYCAST_ADDR;
5246
5247         return inet6_dump_addr(skb, cb, type);
5248 }
5249
5250 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5251                                        const struct nlmsghdr *nlh,
5252                                        struct nlattr **tb,
5253                                        struct netlink_ext_ack *extack)
5254 {
5255         struct ifaddrmsg *ifm;
5256         int i, err;
5257
5258         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5259                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5260                 return -EINVAL;
5261         }
5262
5263         if (!netlink_strict_get_check(skb))
5264                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5265                                               ifa_ipv6_policy, extack);
5266
5267         ifm = nlmsg_data(nlh);
5268         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5269                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5270                 return -EINVAL;
5271         }
5272
5273         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5274                                             ifa_ipv6_policy, extack);
5275         if (err)
5276                 return err;
5277
5278         for (i = 0; i <= IFA_MAX; i++) {
5279                 if (!tb[i])
5280                         continue;
5281
5282                 switch (i) {
5283                 case IFA_TARGET_NETNSID:
5284                 case IFA_ADDRESS:
5285                 case IFA_LOCAL:
5286                         break;
5287                 default:
5288                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5289                         return -EINVAL;
5290                 }
5291         }
5292
5293         return 0;
5294 }
5295
5296 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5297                              struct netlink_ext_ack *extack)
5298 {
5299         struct net *net = sock_net(in_skb->sk);
5300         struct inet6_fill_args fillargs = {
5301                 .portid = NETLINK_CB(in_skb).portid,
5302                 .seq = nlh->nlmsg_seq,
5303                 .event = RTM_NEWADDR,
5304                 .flags = 0,
5305                 .netnsid = -1,
5306         };
5307         struct net *tgt_net = net;
5308         struct ifaddrmsg *ifm;
5309         struct nlattr *tb[IFA_MAX+1];
5310         struct in6_addr *addr = NULL, *peer;
5311         struct net_device *dev = NULL;
5312         struct inet6_ifaddr *ifa;
5313         struct sk_buff *skb;
5314         int err;
5315
5316         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5317         if (err < 0)
5318                 return err;
5319
5320         if (tb[IFA_TARGET_NETNSID]) {
5321                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5322
5323                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5324                                                   fillargs.netnsid);
5325                 if (IS_ERR(tgt_net))
5326                         return PTR_ERR(tgt_net);
5327         }
5328
5329         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5330         if (!addr)
5331                 return -EINVAL;
5332
5333         ifm = nlmsg_data(nlh);
5334         if (ifm->ifa_index)
5335                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5336
5337         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5338         if (!ifa) {
5339                 err = -EADDRNOTAVAIL;
5340                 goto errout;
5341         }
5342
5343         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5344         if (!skb) {
5345                 err = -ENOBUFS;
5346                 goto errout_ifa;
5347         }
5348
5349         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5350         if (err < 0) {
5351                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5352                 WARN_ON(err == -EMSGSIZE);
5353                 kfree_skb(skb);
5354                 goto errout_ifa;
5355         }
5356         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5357 errout_ifa:
5358         in6_ifa_put(ifa);
5359 errout:
5360         if (dev)
5361                 dev_put(dev);
5362         if (fillargs.netnsid >= 0)
5363                 put_net(tgt_net);
5364
5365         return err;
5366 }
5367
5368 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5369 {
5370         struct sk_buff *skb;
5371         struct net *net = dev_net(ifa->idev->dev);
5372         struct inet6_fill_args fillargs = {
5373                 .portid = 0,
5374                 .seq = 0,
5375                 .event = event,
5376                 .flags = 0,
5377                 .netnsid = -1,
5378         };
5379         int err = -ENOBUFS;
5380
5381         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5382         if (!skb)
5383                 goto errout;
5384
5385         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5386         if (err < 0) {
5387                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5388                 WARN_ON(err == -EMSGSIZE);
5389                 kfree_skb(skb);
5390                 goto errout;
5391         }
5392         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5393         return;
5394 errout:
5395         if (err < 0)
5396                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5397 }
5398
5399 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5400                                 __s32 *array, int bytes)
5401 {
5402         BUG_ON(bytes < (DEVCONF_MAX * 4));
5403
5404         memset(array, 0, bytes);
5405         array[DEVCONF_FORWARDING] = cnf->forwarding;
5406         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5407         array[DEVCONF_MTU6] = cnf->mtu6;
5408         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5409         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5410         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5411         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5412         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5413         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5414                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5415         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5416                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5417         array[DEVCONF_RTR_SOLICIT_DELAY] =
5418                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5419         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5420         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5421                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5422         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5423                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5424         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5425         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5426         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5427         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5428         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5429         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5430         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5431         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5432         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5433 #ifdef CONFIG_IPV6_ROUTER_PREF
5434         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5435         array[DEVCONF_RTR_PROBE_INTERVAL] =
5436                 jiffies_to_msecs(cnf->rtr_probe_interval);
5437 #ifdef CONFIG_IPV6_ROUTE_INFO
5438         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5439         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5440 #endif
5441 #endif
5442         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5443         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5444 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5445         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5446         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5447 #endif
5448 #ifdef CONFIG_IPV6_MROUTE
5449         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5450 #endif
5451         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5452         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5453         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5454         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5455         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5456         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5457         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5458         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5459         /* we omit DEVCONF_STABLE_SECRET for now */
5460         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5461         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5462         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5463         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5464         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5465 #ifdef CONFIG_IPV6_SEG6_HMAC
5466         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5467 #endif
5468         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5469         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5470         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5471         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5472 }
5473
5474 static inline size_t inet6_ifla6_size(void)
5475 {
5476         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5477              + nla_total_size(sizeof(struct ifla_cacheinfo))
5478              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5479              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5480              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5481              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5482              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5483              + 0;
5484 }
5485
5486 static inline size_t inet6_if_nlmsg_size(void)
5487 {
5488         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5489                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5490                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5491                + nla_total_size(4) /* IFLA_MTU */
5492                + nla_total_size(4) /* IFLA_LINK */
5493                + nla_total_size(1) /* IFLA_OPERSTATE */
5494                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5495 }
5496
5497 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5498                                         int bytes)
5499 {
5500         int i;
5501         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5502         BUG_ON(pad < 0);
5503
5504         /* Use put_unaligned() because stats may not be aligned for u64. */
5505         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5506         for (i = 1; i < ICMP6_MIB_MAX; i++)
5507                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5508
5509         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5510 }
5511
5512 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5513                                         int bytes, size_t syncpoff)
5514 {
5515         int i, c;
5516         u64 buff[IPSTATS_MIB_MAX];
5517         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5518
5519         BUG_ON(pad < 0);
5520
5521         memset(buff, 0, sizeof(buff));
5522         buff[0] = IPSTATS_MIB_MAX;
5523
5524         for_each_possible_cpu(c) {
5525                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5526                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5527         }
5528
5529         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5530         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5531 }
5532
5533 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5534                              int bytes)
5535 {
5536         switch (attrtype) {
5537         case IFLA_INET6_STATS:
5538                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5539                                      offsetof(struct ipstats_mib, syncp));
5540                 break;
5541         case IFLA_INET6_ICMP6STATS:
5542                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5543                 break;
5544         }
5545 }
5546
5547 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5548                                   u32 ext_filter_mask)
5549 {
5550         struct nlattr *nla;
5551         struct ifla_cacheinfo ci;
5552
5553         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5554                 goto nla_put_failure;
5555         ci.max_reasm_len = IPV6_MAXPLEN;
5556         ci.tstamp = cstamp_delta(idev->tstamp);
5557         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5558         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5559         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5560                 goto nla_put_failure;
5561         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5562         if (!nla)
5563                 goto nla_put_failure;
5564         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5565
5566         /* XXX - MC not implemented */
5567
5568         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5569                 return 0;
5570
5571         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5572         if (!nla)
5573                 goto nla_put_failure;
5574         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5575
5576         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5577         if (!nla)
5578                 goto nla_put_failure;
5579         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5580
5581         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5582         if (!nla)
5583                 goto nla_put_failure;
5584         read_lock_bh(&idev->lock);
5585         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5586         read_unlock_bh(&idev->lock);
5587
5588         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5589                 goto nla_put_failure;
5590
5591         return 0;
5592
5593 nla_put_failure:
5594         return -EMSGSIZE;
5595 }
5596
5597 static size_t inet6_get_link_af_size(const struct net_device *dev,
5598                                      u32 ext_filter_mask)
5599 {
5600         if (!__in6_dev_get(dev))
5601                 return 0;
5602
5603         return inet6_ifla6_size();
5604 }
5605
5606 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5607                               u32 ext_filter_mask)
5608 {
5609         struct inet6_dev *idev = __in6_dev_get(dev);
5610
5611         if (!idev)
5612                 return -ENODATA;
5613
5614         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5615                 return -EMSGSIZE;
5616
5617         return 0;
5618 }
5619
5620 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5621 {
5622         struct inet6_ifaddr *ifp;
5623         struct net_device *dev = idev->dev;
5624         bool clear_token, update_rs = false;
5625         struct in6_addr ll_addr;
5626
5627         ASSERT_RTNL();
5628
5629         if (!token)
5630                 return -EINVAL;
5631         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5632                 return -EINVAL;
5633         if (!ipv6_accept_ra(idev))
5634                 return -EINVAL;
5635         if (idev->cnf.rtr_solicits == 0)
5636                 return -EINVAL;
5637
5638         write_lock_bh(&idev->lock);
5639
5640         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5641         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5642
5643         write_unlock_bh(&idev->lock);
5644
5645         clear_token = ipv6_addr_any(token);
5646         if (clear_token)
5647                 goto update_lft;
5648
5649         if (!idev->dead && (idev->if_flags & IF_READY) &&
5650             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5651                              IFA_F_OPTIMISTIC)) {
5652                 /* If we're not ready, then normal ifup will take care
5653                  * of this. Otherwise, we need to request our rs here.
5654                  */
5655                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5656                 update_rs = true;
5657         }
5658
5659 update_lft:
5660         write_lock_bh(&idev->lock);
5661
5662         if (update_rs) {
5663                 idev->if_flags |= IF_RS_SENT;
5664                 idev->rs_interval = rfc3315_s14_backoff_init(
5665                         idev->cnf.rtr_solicit_interval);
5666                 idev->rs_probes = 1;
5667                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5668         }
5669
5670         /* Well, that's kinda nasty ... */
5671         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5672                 spin_lock(&ifp->lock);
5673                 if (ifp->tokenized) {
5674                         ifp->valid_lft = 0;
5675                         ifp->prefered_lft = 0;
5676                 }
5677                 spin_unlock(&ifp->lock);
5678         }
5679
5680         write_unlock_bh(&idev->lock);
5681         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5682         addrconf_verify_rtnl();
5683         return 0;
5684 }
5685
5686 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5687         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5688         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5689 };
5690
5691 static int check_addr_gen_mode(int mode)
5692 {
5693         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5694             mode != IN6_ADDR_GEN_MODE_NONE &&
5695             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5696             mode != IN6_ADDR_GEN_MODE_RANDOM)
5697                 return -EINVAL;
5698         return 1;
5699 }
5700
5701 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5702                                 int mode)
5703 {
5704         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5705             !idev->cnf.stable_secret.initialized &&
5706             !net->ipv6.devconf_dflt->stable_secret.initialized)
5707                 return -EINVAL;
5708         return 1;
5709 }
5710
5711 static int inet6_validate_link_af(const struct net_device *dev,
5712                                   const struct nlattr *nla)
5713 {
5714         struct nlattr *tb[IFLA_INET6_MAX + 1];
5715         struct inet6_dev *idev = NULL;
5716         int err;
5717
5718         if (dev) {
5719                 idev = __in6_dev_get(dev);
5720                 if (!idev)
5721                         return -EAFNOSUPPORT;
5722         }
5723
5724         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5725                                           inet6_af_policy, NULL);
5726         if (err)
5727                 return err;
5728
5729         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5730                 return -EINVAL;
5731
5732         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5733                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5734
5735                 if (check_addr_gen_mode(mode) < 0)
5736                         return -EINVAL;
5737                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5738                         return -EINVAL;
5739         }
5740
5741         return 0;
5742 }
5743
5744 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5745 {
5746         struct inet6_dev *idev = __in6_dev_get(dev);
5747         struct nlattr *tb[IFLA_INET6_MAX + 1];
5748         int err;
5749
5750         if (!idev)
5751                 return -EAFNOSUPPORT;
5752
5753         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5754                 BUG();
5755
5756         if (tb[IFLA_INET6_TOKEN]) {
5757                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5758                 if (err)
5759                         return err;
5760         }
5761
5762         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5763                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5764
5765                 idev->cnf.addr_gen_mode = mode;
5766         }
5767
5768         return 0;
5769 }
5770
5771 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5772                              u32 portid, u32 seq, int event, unsigned int flags)
5773 {
5774         struct net_device *dev = idev->dev;
5775         struct ifinfomsg *hdr;
5776         struct nlmsghdr *nlh;
5777         void *protoinfo;
5778
5779         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5780         if (!nlh)
5781                 return -EMSGSIZE;
5782
5783         hdr = nlmsg_data(nlh);
5784         hdr->ifi_family = AF_INET6;
5785         hdr->__ifi_pad = 0;
5786         hdr->ifi_type = dev->type;
5787         hdr->ifi_index = dev->ifindex;
5788         hdr->ifi_flags = dev_get_flags(dev);
5789         hdr->ifi_change = 0;
5790
5791         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5792             (dev->addr_len &&
5793              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5794             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5795             (dev->ifindex != dev_get_iflink(dev) &&
5796              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5797             nla_put_u8(skb, IFLA_OPERSTATE,
5798                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5799                 goto nla_put_failure;
5800         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5801         if (!protoinfo)
5802                 goto nla_put_failure;
5803
5804         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5805                 goto nla_put_failure;
5806
5807         nla_nest_end(skb, protoinfo);
5808         nlmsg_end(skb, nlh);
5809         return 0;
5810
5811 nla_put_failure:
5812         nlmsg_cancel(skb, nlh);
5813         return -EMSGSIZE;
5814 }
5815
5816 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5817                                    struct netlink_ext_ack *extack)
5818 {
5819         struct ifinfomsg *ifm;
5820
5821         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5822                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5823                 return -EINVAL;
5824         }
5825
5826         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5827                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5828                 return -EINVAL;
5829         }
5830
5831         ifm = nlmsg_data(nlh);
5832         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5833             ifm->ifi_change || ifm->ifi_index) {
5834                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5835                 return -EINVAL;
5836         }
5837
5838         return 0;
5839 }
5840
5841 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5842 {
5843         struct net *net = sock_net(skb->sk);
5844         int h, s_h;
5845         int idx = 0, s_idx;
5846         struct net_device *dev;
5847         struct inet6_dev *idev;
5848         struct hlist_head *head;
5849
5850         /* only requests using strict checking can pass data to
5851          * influence the dump
5852          */
5853         if (cb->strict_check) {
5854                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5855
5856                 if (err < 0)
5857                         return err;
5858         }
5859
5860         s_h = cb->args[0];
5861         s_idx = cb->args[1];
5862
5863         rcu_read_lock();
5864         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5865                 idx = 0;
5866                 head = &net->dev_index_head[h];
5867                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5868                         if (idx < s_idx)
5869                                 goto cont;
5870                         idev = __in6_dev_get(dev);
5871                         if (!idev)
5872                                 goto cont;
5873                         if (inet6_fill_ifinfo(skb, idev,
5874                                               NETLINK_CB(cb->skb).portid,
5875                                               cb->nlh->nlmsg_seq,
5876                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5877                                 goto out;
5878 cont:
5879                         idx++;
5880                 }
5881         }
5882 out:
5883         rcu_read_unlock();
5884         cb->args[1] = idx;
5885         cb->args[0] = h;
5886
5887         return skb->len;
5888 }
5889
5890 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5891 {
5892         struct sk_buff *skb;
5893         struct net *net = dev_net(idev->dev);
5894         int err = -ENOBUFS;
5895
5896         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5897         if (!skb)
5898                 goto errout;
5899
5900         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5901         if (err < 0) {
5902                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5903                 WARN_ON(err == -EMSGSIZE);
5904                 kfree_skb(skb);
5905                 goto errout;
5906         }
5907         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5908         return;
5909 errout:
5910         if (err < 0)
5911                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5912 }
5913
5914 static inline size_t inet6_prefix_nlmsg_size(void)
5915 {
5916         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5917                + nla_total_size(sizeof(struct in6_addr))
5918                + nla_total_size(sizeof(struct prefix_cacheinfo));
5919 }
5920
5921 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5922                              struct prefix_info *pinfo, u32 portid, u32 seq,
5923                              int event, unsigned int flags)
5924 {
5925         struct prefixmsg *pmsg;
5926         struct nlmsghdr *nlh;
5927         struct prefix_cacheinfo ci;
5928
5929         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5930         if (!nlh)
5931                 return -EMSGSIZE;
5932
5933         pmsg = nlmsg_data(nlh);
5934         pmsg->prefix_family = AF_INET6;
5935         pmsg->prefix_pad1 = 0;
5936         pmsg->prefix_pad2 = 0;
5937         pmsg->prefix_ifindex = idev->dev->ifindex;
5938         pmsg->prefix_len = pinfo->prefix_len;
5939         pmsg->prefix_type = pinfo->type;
5940         pmsg->prefix_pad3 = 0;
5941         pmsg->prefix_flags = 0;
5942         if (pinfo->onlink)
5943                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5944         if (pinfo->autoconf)
5945                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5946
5947         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5948                 goto nla_put_failure;
5949         ci.preferred_time = ntohl(pinfo->prefered);
5950         ci.valid_time = ntohl(pinfo->valid);
5951         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5952                 goto nla_put_failure;
5953         nlmsg_end(skb, nlh);
5954         return 0;
5955
5956 nla_put_failure:
5957         nlmsg_cancel(skb, nlh);
5958         return -EMSGSIZE;
5959 }
5960
5961 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5962                          struct prefix_info *pinfo)
5963 {
5964         struct sk_buff *skb;
5965         struct net *net = dev_net(idev->dev);
5966         int err = -ENOBUFS;
5967
5968         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5969         if (!skb)
5970                 goto errout;
5971
5972         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5973         if (err < 0) {
5974                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5975                 WARN_ON(err == -EMSGSIZE);
5976                 kfree_skb(skb);
5977                 goto errout;
5978         }
5979         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5980         return;
5981 errout:
5982         if (err < 0)
5983                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5984 }
5985
5986 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5987 {
5988         struct net *net = dev_net(ifp->idev->dev);
5989
5990         if (event)
5991                 ASSERT_RTNL();
5992
5993         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5994
5995         switch (event) {
5996         case RTM_NEWADDR:
5997                 /*
5998                  * If the address was optimistic we inserted the route at the
5999                  * start of our DAD process, so we don't need to do it again.
6000                  * If the device was taken down in the middle of the DAD
6001                  * cycle there is a race where we could get here without a
6002                  * host route, so nothing to insert. That will be fixed when
6003                  * the device is brought up.
6004                  */
6005                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6006                         ip6_ins_rt(net, ifp->rt);
6007                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6008                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6009                                 &ifp->addr, ifp->idev->dev->name);
6010                 }
6011
6012                 if (ifp->idev->cnf.forwarding)
6013                         addrconf_join_anycast(ifp);
6014                 if (!ipv6_addr_any(&ifp->peer_addr))
6015                         addrconf_prefix_route(&ifp->peer_addr, 128,
6016                                               ifp->rt_priority, ifp->idev->dev,
6017                                               0, 0, GFP_ATOMIC);
6018                 break;
6019         case RTM_DELADDR:
6020                 if (ifp->idev->cnf.forwarding)
6021                         addrconf_leave_anycast(ifp);
6022                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6023                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6024                         struct fib6_info *rt;
6025
6026                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6027                                                        ifp->idev->dev, 0, 0,
6028                                                        false);
6029                         if (rt)
6030                                 ip6_del_rt(net, rt);
6031                 }
6032                 if (ifp->rt) {
6033                         ip6_del_rt(net, ifp->rt);
6034                         ifp->rt = NULL;
6035                 }
6036                 rt_genid_bump_ipv6(net);
6037                 break;
6038         }
6039         atomic_inc(&net->ipv6.dev_addr_genid);
6040 }
6041
6042 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6043 {
6044         rcu_read_lock_bh();
6045         if (likely(ifp->idev->dead == 0))
6046                 __ipv6_ifa_notify(event, ifp);
6047         rcu_read_unlock_bh();
6048 }
6049
6050 #ifdef CONFIG_SYSCTL
6051
6052 static
6053 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6054                            void __user *buffer, size_t *lenp, loff_t *ppos)
6055 {
6056         int *valp = ctl->data;
6057         int val = *valp;
6058         loff_t pos = *ppos;
6059         struct ctl_table lctl;
6060         int ret;
6061
6062         /*
6063          * ctl->data points to idev->cnf.forwarding, we should
6064          * not modify it until we get the rtnl lock.
6065          */
6066         lctl = *ctl;
6067         lctl.data = &val;
6068
6069         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6070
6071         if (write)
6072                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6073         if (ret)
6074                 *ppos = pos;
6075         return ret;
6076 }
6077
6078 static
6079 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6080                         void __user *buffer, size_t *lenp, loff_t *ppos)
6081 {
6082         struct inet6_dev *idev = ctl->extra1;
6083         int min_mtu = IPV6_MIN_MTU;
6084         struct ctl_table lctl;
6085
6086         lctl = *ctl;
6087         lctl.extra1 = &min_mtu;
6088         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6089
6090         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6091 }
6092
6093 static void dev_disable_change(struct inet6_dev *idev)
6094 {
6095         struct netdev_notifier_info info;
6096
6097         if (!idev || !idev->dev)
6098                 return;
6099
6100         netdev_notifier_info_init(&info, idev->dev);
6101         if (idev->cnf.disable_ipv6)
6102                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6103         else
6104                 addrconf_notify(NULL, NETDEV_UP, &info);
6105 }
6106
6107 static void addrconf_disable_change(struct net *net, __s32 newf)
6108 {
6109         struct net_device *dev;
6110         struct inet6_dev *idev;
6111
6112         for_each_netdev(net, dev) {
6113                 idev = __in6_dev_get(dev);
6114                 if (idev) {
6115                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6116                         idev->cnf.disable_ipv6 = newf;
6117                         if (changed)
6118                                 dev_disable_change(idev);
6119                 }
6120         }
6121 }
6122
6123 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6124 {
6125         struct net *net;
6126         int old;
6127
6128         if (!rtnl_trylock())
6129                 return restart_syscall();
6130
6131         net = (struct net *)table->extra2;
6132         old = *p;
6133         *p = newf;
6134
6135         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6136                 rtnl_unlock();
6137                 return 0;
6138         }
6139
6140         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6141                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6142                 addrconf_disable_change(net, newf);
6143         } else if ((!newf) ^ (!old))
6144                 dev_disable_change((struct inet6_dev *)table->extra1);
6145
6146         rtnl_unlock();
6147         return 0;
6148 }
6149
6150 static
6151 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6152                             void __user *buffer, size_t *lenp, loff_t *ppos)
6153 {
6154         int *valp = ctl->data;
6155         int val = *valp;
6156         loff_t pos = *ppos;
6157         struct ctl_table lctl;
6158         int ret;
6159
6160         /*
6161          * ctl->data points to idev->cnf.disable_ipv6, we should
6162          * not modify it until we get the rtnl lock.
6163          */
6164         lctl = *ctl;
6165         lctl.data = &val;
6166
6167         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6168
6169         if (write)
6170                 ret = addrconf_disable_ipv6(ctl, valp, val);
6171         if (ret)
6172                 *ppos = pos;
6173         return ret;
6174 }
6175
6176 static
6177 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6178                               void __user *buffer, size_t *lenp, loff_t *ppos)
6179 {
6180         int *valp = ctl->data;
6181         int ret;
6182         int old, new;
6183
6184         old = *valp;
6185         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6186         new = *valp;
6187
6188         if (write && old != new) {
6189                 struct net *net = ctl->extra2;
6190
6191                 if (!rtnl_trylock())
6192                         return restart_syscall();
6193
6194                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6195                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6196                                                      NETCONFA_PROXY_NEIGH,
6197                                                      NETCONFA_IFINDEX_DEFAULT,
6198                                                      net->ipv6.devconf_dflt);
6199                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6200                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6201                                                      NETCONFA_PROXY_NEIGH,
6202                                                      NETCONFA_IFINDEX_ALL,
6203                                                      net->ipv6.devconf_all);
6204                 else {
6205                         struct inet6_dev *idev = ctl->extra1;
6206
6207                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6208                                                      NETCONFA_PROXY_NEIGH,
6209                                                      idev->dev->ifindex,
6210                                                      &idev->cnf);
6211                 }
6212                 rtnl_unlock();
6213         }
6214
6215         return ret;
6216 }
6217
6218 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6219                                          void __user *buffer, size_t *lenp,
6220                                          loff_t *ppos)
6221 {
6222         int ret = 0;
6223         u32 new_val;
6224         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6225         struct net *net = (struct net *)ctl->extra2;
6226         struct ctl_table tmp = {
6227                 .data = &new_val,
6228                 .maxlen = sizeof(new_val),
6229                 .mode = ctl->mode,
6230         };
6231
6232         if (!rtnl_trylock())
6233                 return restart_syscall();
6234
6235         new_val = *((u32 *)ctl->data);
6236
6237         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6238         if (ret != 0)
6239                 goto out;
6240
6241         if (write) {
6242                 if (check_addr_gen_mode(new_val) < 0) {
6243                         ret = -EINVAL;
6244                         goto out;
6245                 }
6246
6247                 if (idev) {
6248                         if (check_stable_privacy(idev, net, new_val) < 0) {
6249                                 ret = -EINVAL;
6250                                 goto out;
6251                         }
6252
6253                         if (idev->cnf.addr_gen_mode != new_val) {
6254                                 idev->cnf.addr_gen_mode = new_val;
6255                                 addrconf_dev_config(idev->dev);
6256                         }
6257                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6258                         struct net_device *dev;
6259
6260                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6261                         for_each_netdev(net, dev) {
6262                                 idev = __in6_dev_get(dev);
6263                                 if (idev &&
6264                                     idev->cnf.addr_gen_mode != new_val) {
6265                                         idev->cnf.addr_gen_mode = new_val;
6266                                         addrconf_dev_config(idev->dev);
6267                                 }
6268                         }
6269                 }
6270
6271                 *((u32 *)ctl->data) = new_val;
6272         }
6273
6274 out:
6275         rtnl_unlock();
6276
6277         return ret;
6278 }
6279
6280 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6281                                          void __user *buffer, size_t *lenp,
6282                                          loff_t *ppos)
6283 {
6284         int err;
6285         struct in6_addr addr;
6286         char str[IPV6_MAX_STRLEN];
6287         struct ctl_table lctl = *ctl;
6288         struct net *net = ctl->extra2;
6289         struct ipv6_stable_secret *secret = ctl->data;
6290
6291         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6292                 return -EIO;
6293
6294         lctl.maxlen = IPV6_MAX_STRLEN;
6295         lctl.data = str;
6296
6297         if (!rtnl_trylock())
6298                 return restart_syscall();
6299
6300         if (!write && !secret->initialized) {
6301                 err = -EIO;
6302                 goto out;
6303         }
6304
6305         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6306         if (err >= sizeof(str)) {
6307                 err = -EIO;
6308                 goto out;
6309         }
6310
6311         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6312         if (err || !write)
6313                 goto out;
6314
6315         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6316                 err = -EIO;
6317                 goto out;
6318         }
6319
6320         secret->initialized = true;
6321         secret->secret = addr;
6322
6323         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6324                 struct net_device *dev;
6325
6326                 for_each_netdev(net, dev) {
6327                         struct inet6_dev *idev = __in6_dev_get(dev);
6328
6329                         if (idev) {
6330                                 idev->cnf.addr_gen_mode =
6331                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6332                         }
6333                 }
6334         } else {
6335                 struct inet6_dev *idev = ctl->extra1;
6336
6337                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6338         }
6339
6340 out:
6341         rtnl_unlock();
6342
6343         return err;
6344 }
6345
6346 static
6347 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6348                                                 int write,
6349                                                 void __user *buffer,
6350                                                 size_t *lenp,
6351                                                 loff_t *ppos)
6352 {
6353         int *valp = ctl->data;
6354         int val = *valp;
6355         loff_t pos = *ppos;
6356         struct ctl_table lctl;
6357         int ret;
6358
6359         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6360          * we should not modify it until we get the rtnl lock.
6361          */
6362         lctl = *ctl;
6363         lctl.data = &val;
6364
6365         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6366
6367         if (write)
6368                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6369         if (ret)
6370                 *ppos = pos;
6371         return ret;
6372 }
6373
6374 static
6375 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6376 {
6377         if (rt) {
6378                 if (action)
6379                         rt->dst.flags |= DST_NOPOLICY;
6380                 else
6381                         rt->dst.flags &= ~DST_NOPOLICY;
6382         }
6383 }
6384
6385 static
6386 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6387 {
6388         struct inet6_ifaddr *ifa;
6389
6390         read_lock_bh(&idev->lock);
6391         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6392                 spin_lock(&ifa->lock);
6393                 if (ifa->rt) {
6394                         /* host routes only use builtin fib6_nh */
6395                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6396                         int cpu;
6397
6398                         rcu_read_lock();
6399                         ifa->rt->dst_nopolicy = val ? true : false;
6400                         if (nh->rt6i_pcpu) {
6401                                 for_each_possible_cpu(cpu) {
6402                                         struct rt6_info **rtp;
6403
6404                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6405                                         addrconf_set_nopolicy(*rtp, val);
6406                                 }
6407                         }
6408                         rcu_read_unlock();
6409                 }
6410                 spin_unlock(&ifa->lock);
6411         }
6412         read_unlock_bh(&idev->lock);
6413 }
6414
6415 static
6416 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6417 {
6418         struct inet6_dev *idev;
6419         struct net *net;
6420
6421         if (!rtnl_trylock())
6422                 return restart_syscall();
6423
6424         *valp = val;
6425
6426         net = (struct net *)ctl->extra2;
6427         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6428                 rtnl_unlock();
6429                 return 0;
6430         }
6431
6432         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6433                 struct net_device *dev;
6434
6435                 for_each_netdev(net, dev) {
6436                         idev = __in6_dev_get(dev);
6437                         if (idev)
6438                                 addrconf_disable_policy_idev(idev, val);
6439                 }
6440         } else {
6441                 idev = (struct inet6_dev *)ctl->extra1;
6442                 addrconf_disable_policy_idev(idev, val);
6443         }
6444
6445         rtnl_unlock();
6446         return 0;
6447 }
6448
6449 static
6450 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6451                                    void __user *buffer, size_t *lenp,
6452                                    loff_t *ppos)
6453 {
6454         int *valp = ctl->data;
6455         int val = *valp;
6456         loff_t pos = *ppos;
6457         struct ctl_table lctl;
6458         int ret;
6459
6460         lctl = *ctl;
6461         lctl.data = &val;
6462         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6463
6464         if (write && (*valp != val))
6465                 ret = addrconf_disable_policy(ctl, valp, val);
6466
6467         if (ret)
6468                 *ppos = pos;
6469
6470         return ret;
6471 }
6472
6473 static int minus_one = -1;
6474 static const int two_five_five = 255;
6475
6476 static const struct ctl_table addrconf_sysctl[] = {
6477         {
6478                 .procname       = "forwarding",
6479                 .data           = &ipv6_devconf.forwarding,
6480                 .maxlen         = sizeof(int),
6481                 .mode           = 0644,
6482                 .proc_handler   = addrconf_sysctl_forward,
6483         },
6484         {
6485                 .procname       = "hop_limit",
6486                 .data           = &ipv6_devconf.hop_limit,
6487                 .maxlen         = sizeof(int),
6488                 .mode           = 0644,
6489                 .proc_handler   = proc_dointvec_minmax,
6490                 .extra1         = (void *)SYSCTL_ONE,
6491                 .extra2         = (void *)&two_five_five,
6492         },
6493         {
6494                 .procname       = "mtu",
6495                 .data           = &ipv6_devconf.mtu6,
6496                 .maxlen         = sizeof(int),
6497                 .mode           = 0644,
6498                 .proc_handler   = addrconf_sysctl_mtu,
6499         },
6500         {
6501                 .procname       = "accept_ra",
6502                 .data           = &ipv6_devconf.accept_ra,
6503                 .maxlen         = sizeof(int),
6504                 .mode           = 0644,
6505                 .proc_handler   = proc_dointvec,
6506         },
6507         {
6508                 .procname       = "accept_redirects",
6509                 .data           = &ipv6_devconf.accept_redirects,
6510                 .maxlen         = sizeof(int),
6511                 .mode           = 0644,
6512                 .proc_handler   = proc_dointvec,
6513         },
6514         {
6515                 .procname       = "autoconf",
6516                 .data           = &ipv6_devconf.autoconf,
6517                 .maxlen         = sizeof(int),
6518                 .mode           = 0644,
6519                 .proc_handler   = proc_dointvec,
6520         },
6521         {
6522                 .procname       = "dad_transmits",
6523                 .data           = &ipv6_devconf.dad_transmits,
6524                 .maxlen         = sizeof(int),
6525                 .mode           = 0644,
6526                 .proc_handler   = proc_dointvec,
6527         },
6528         {
6529                 .procname       = "router_solicitations",
6530                 .data           = &ipv6_devconf.rtr_solicits,
6531                 .maxlen         = sizeof(int),
6532                 .mode           = 0644,
6533                 .proc_handler   = proc_dointvec_minmax,
6534                 .extra1         = &minus_one,
6535         },
6536         {
6537                 .procname       = "router_solicitation_interval",
6538                 .data           = &ipv6_devconf.rtr_solicit_interval,
6539                 .maxlen         = sizeof(int),
6540                 .mode           = 0644,
6541                 .proc_handler   = proc_dointvec_jiffies,
6542         },
6543         {
6544                 .procname       = "router_solicitation_max_interval",
6545                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6546                 .maxlen         = sizeof(int),
6547                 .mode           = 0644,
6548                 .proc_handler   = proc_dointvec_jiffies,
6549         },
6550         {
6551                 .procname       = "router_solicitation_delay",
6552                 .data           = &ipv6_devconf.rtr_solicit_delay,
6553                 .maxlen         = sizeof(int),
6554                 .mode           = 0644,
6555                 .proc_handler   = proc_dointvec_jiffies,
6556         },
6557         {
6558                 .procname       = "force_mld_version",
6559                 .data           = &ipv6_devconf.force_mld_version,
6560                 .maxlen         = sizeof(int),
6561                 .mode           = 0644,
6562                 .proc_handler   = proc_dointvec,
6563         },
6564         {
6565                 .procname       = "mldv1_unsolicited_report_interval",
6566                 .data           =
6567                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6568                 .maxlen         = sizeof(int),
6569                 .mode           = 0644,
6570                 .proc_handler   = proc_dointvec_ms_jiffies,
6571         },
6572         {
6573                 .procname       = "mldv2_unsolicited_report_interval",
6574                 .data           =
6575                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6576                 .maxlen         = sizeof(int),
6577                 .mode           = 0644,
6578                 .proc_handler   = proc_dointvec_ms_jiffies,
6579         },
6580         {
6581                 .procname       = "use_tempaddr",
6582                 .data           = &ipv6_devconf.use_tempaddr,
6583                 .maxlen         = sizeof(int),
6584                 .mode           = 0644,
6585                 .proc_handler   = proc_dointvec,
6586         },
6587         {
6588                 .procname       = "temp_valid_lft",
6589                 .data           = &ipv6_devconf.temp_valid_lft,
6590                 .maxlen         = sizeof(int),
6591                 .mode           = 0644,
6592                 .proc_handler   = proc_dointvec,
6593         },
6594         {
6595                 .procname       = "temp_prefered_lft",
6596                 .data           = &ipv6_devconf.temp_prefered_lft,
6597                 .maxlen         = sizeof(int),
6598                 .mode           = 0644,
6599                 .proc_handler   = proc_dointvec,
6600         },
6601         {
6602                 .procname       = "regen_max_retry",
6603                 .data           = &ipv6_devconf.regen_max_retry,
6604                 .maxlen         = sizeof(int),
6605                 .mode           = 0644,
6606                 .proc_handler   = proc_dointvec,
6607         },
6608         {
6609                 .procname       = "max_desync_factor",
6610                 .data           = &ipv6_devconf.max_desync_factor,
6611                 .maxlen         = sizeof(int),
6612                 .mode           = 0644,
6613                 .proc_handler   = proc_dointvec,
6614         },
6615         {
6616                 .procname       = "max_addresses",
6617                 .data           = &ipv6_devconf.max_addresses,
6618                 .maxlen         = sizeof(int),
6619                 .mode           = 0644,
6620                 .proc_handler   = proc_dointvec,
6621         },
6622         {
6623                 .procname       = "accept_ra_defrtr",
6624                 .data           = &ipv6_devconf.accept_ra_defrtr,
6625                 .maxlen         = sizeof(int),
6626                 .mode           = 0644,
6627                 .proc_handler   = proc_dointvec,
6628         },
6629         {
6630                 .procname       = "accept_ra_min_hop_limit",
6631                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6632                 .maxlen         = sizeof(int),
6633                 .mode           = 0644,
6634                 .proc_handler   = proc_dointvec,
6635         },
6636         {
6637                 .procname       = "accept_ra_pinfo",
6638                 .data           = &ipv6_devconf.accept_ra_pinfo,
6639                 .maxlen         = sizeof(int),
6640                 .mode           = 0644,
6641                 .proc_handler   = proc_dointvec,
6642         },
6643 #ifdef CONFIG_IPV6_ROUTER_PREF
6644         {
6645                 .procname       = "accept_ra_rtr_pref",
6646                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6647                 .maxlen         = sizeof(int),
6648                 .mode           = 0644,
6649                 .proc_handler   = proc_dointvec,
6650         },
6651         {
6652                 .procname       = "router_probe_interval",
6653                 .data           = &ipv6_devconf.rtr_probe_interval,
6654                 .maxlen         = sizeof(int),
6655                 .mode           = 0644,
6656                 .proc_handler   = proc_dointvec_jiffies,
6657         },
6658 #ifdef CONFIG_IPV6_ROUTE_INFO
6659         {
6660                 .procname       = "accept_ra_rt_info_min_plen",
6661                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6662                 .maxlen         = sizeof(int),
6663                 .mode           = 0644,
6664                 .proc_handler   = proc_dointvec,
6665         },
6666         {
6667                 .procname       = "accept_ra_rt_info_max_plen",
6668                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6669                 .maxlen         = sizeof(int),
6670                 .mode           = 0644,
6671                 .proc_handler   = proc_dointvec,
6672         },
6673 #endif
6674 #endif
6675         {
6676                 .procname       = "proxy_ndp",
6677                 .data           = &ipv6_devconf.proxy_ndp,
6678                 .maxlen         = sizeof(int),
6679                 .mode           = 0644,
6680                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6681         },
6682         {
6683                 .procname       = "accept_source_route",
6684                 .data           = &ipv6_devconf.accept_source_route,
6685                 .maxlen         = sizeof(int),
6686                 .mode           = 0644,
6687                 .proc_handler   = proc_dointvec,
6688         },
6689 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6690         {
6691                 .procname       = "optimistic_dad",
6692                 .data           = &ipv6_devconf.optimistic_dad,
6693                 .maxlen         = sizeof(int),
6694                 .mode           = 0644,
6695                 .proc_handler   = proc_dointvec,
6696         },
6697         {
6698                 .procname       = "use_optimistic",
6699                 .data           = &ipv6_devconf.use_optimistic,
6700                 .maxlen         = sizeof(int),
6701                 .mode           = 0644,
6702                 .proc_handler   = proc_dointvec,
6703         },
6704 #endif
6705 #ifdef CONFIG_IPV6_MROUTE
6706         {
6707                 .procname       = "mc_forwarding",
6708                 .data           = &ipv6_devconf.mc_forwarding,
6709                 .maxlen         = sizeof(int),
6710                 .mode           = 0444,
6711                 .proc_handler   = proc_dointvec,
6712         },
6713 #endif
6714         {
6715                 .procname       = "disable_ipv6",
6716                 .data           = &ipv6_devconf.disable_ipv6,
6717                 .maxlen         = sizeof(int),
6718                 .mode           = 0644,
6719                 .proc_handler   = addrconf_sysctl_disable,
6720         },
6721         {
6722                 .procname       = "accept_dad",
6723                 .data           = &ipv6_devconf.accept_dad,
6724                 .maxlen         = sizeof(int),
6725                 .mode           = 0644,
6726                 .proc_handler   = proc_dointvec,
6727         },
6728         {
6729                 .procname       = "force_tllao",
6730                 .data           = &ipv6_devconf.force_tllao,
6731                 .maxlen         = sizeof(int),
6732                 .mode           = 0644,
6733                 .proc_handler   = proc_dointvec
6734         },
6735         {
6736                 .procname       = "ndisc_notify",
6737                 .data           = &ipv6_devconf.ndisc_notify,
6738                 .maxlen         = sizeof(int),
6739                 .mode           = 0644,
6740                 .proc_handler   = proc_dointvec
6741         },
6742         {
6743                 .procname       = "suppress_frag_ndisc",
6744                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6745                 .maxlen         = sizeof(int),
6746                 .mode           = 0644,
6747                 .proc_handler   = proc_dointvec
6748         },
6749         {
6750                 .procname       = "accept_ra_from_local",
6751                 .data           = &ipv6_devconf.accept_ra_from_local,
6752                 .maxlen         = sizeof(int),
6753                 .mode           = 0644,
6754                 .proc_handler   = proc_dointvec,
6755         },
6756         {
6757                 .procname       = "accept_ra_mtu",
6758                 .data           = &ipv6_devconf.accept_ra_mtu,
6759                 .maxlen         = sizeof(int),
6760                 .mode           = 0644,
6761                 .proc_handler   = proc_dointvec,
6762         },
6763         {
6764                 .procname       = "stable_secret",
6765                 .data           = &ipv6_devconf.stable_secret,
6766                 .maxlen         = IPV6_MAX_STRLEN,
6767                 .mode           = 0600,
6768                 .proc_handler   = addrconf_sysctl_stable_secret,
6769         },
6770         {
6771                 .procname       = "use_oif_addrs_only",
6772                 .data           = &ipv6_devconf.use_oif_addrs_only,
6773                 .maxlen         = sizeof(int),
6774                 .mode           = 0644,
6775                 .proc_handler   = proc_dointvec,
6776         },
6777         {
6778                 .procname       = "ignore_routes_with_linkdown",
6779                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6780                 .maxlen         = sizeof(int),
6781                 .mode           = 0644,
6782                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6783         },
6784         {
6785                 .procname       = "drop_unicast_in_l2_multicast",
6786                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6787                 .maxlen         = sizeof(int),
6788                 .mode           = 0644,
6789                 .proc_handler   = proc_dointvec,
6790         },
6791         {
6792                 .procname       = "drop_unsolicited_na",
6793                 .data           = &ipv6_devconf.drop_unsolicited_na,
6794                 .maxlen         = sizeof(int),
6795                 .mode           = 0644,
6796                 .proc_handler   = proc_dointvec,
6797         },
6798         {
6799                 .procname       = "keep_addr_on_down",
6800                 .data           = &ipv6_devconf.keep_addr_on_down,
6801                 .maxlen         = sizeof(int),
6802                 .mode           = 0644,
6803                 .proc_handler   = proc_dointvec,
6804
6805         },
6806         {
6807                 .procname       = "seg6_enabled",
6808                 .data           = &ipv6_devconf.seg6_enabled,
6809                 .maxlen         = sizeof(int),
6810                 .mode           = 0644,
6811                 .proc_handler   = proc_dointvec,
6812         },
6813 #ifdef CONFIG_IPV6_SEG6_HMAC
6814         {
6815                 .procname       = "seg6_require_hmac",
6816                 .data           = &ipv6_devconf.seg6_require_hmac,
6817                 .maxlen         = sizeof(int),
6818                 .mode           = 0644,
6819                 .proc_handler   = proc_dointvec,
6820         },
6821 #endif
6822         {
6823                 .procname       = "enhanced_dad",
6824                 .data           = &ipv6_devconf.enhanced_dad,
6825                 .maxlen         = sizeof(int),
6826                 .mode           = 0644,
6827                 .proc_handler   = proc_dointvec,
6828         },
6829         {
6830                 .procname               = "addr_gen_mode",
6831                 .data                   = &ipv6_devconf.addr_gen_mode,
6832                 .maxlen                 = sizeof(int),
6833                 .mode                   = 0644,
6834                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6835         },
6836         {
6837                 .procname       = "disable_policy",
6838                 .data           = &ipv6_devconf.disable_policy,
6839                 .maxlen         = sizeof(int),
6840                 .mode           = 0644,
6841                 .proc_handler   = addrconf_sysctl_disable_policy,
6842         },
6843         {
6844                 .procname       = "ndisc_tclass",
6845                 .data           = &ipv6_devconf.ndisc_tclass,
6846                 .maxlen         = sizeof(int),
6847                 .mode           = 0644,
6848                 .proc_handler   = proc_dointvec_minmax,
6849                 .extra1         = (void *)SYSCTL_ZERO,
6850                 .extra2         = (void *)&two_five_five,
6851         },
6852         {
6853                 /* sentinel */
6854         }
6855 };
6856
6857 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6858                 struct inet6_dev *idev, struct ipv6_devconf *p)
6859 {
6860         int i, ifindex;
6861         struct ctl_table *table;
6862         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6863
6864         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6865         if (!table)
6866                 goto out;
6867
6868         for (i = 0; table[i].data; i++) {
6869                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6870                 /* If one of these is already set, then it is not safe to
6871                  * overwrite either of them: this makes proc_dointvec_minmax
6872                  * usable.
6873                  */
6874                 if (!table[i].extra1 && !table[i].extra2) {
6875                         table[i].extra1 = idev; /* embedded; no ref */
6876                         table[i].extra2 = net;
6877                 }
6878         }
6879
6880         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6881
6882         p->sysctl_header = register_net_sysctl(net, path, table);
6883         if (!p->sysctl_header)
6884                 goto free;
6885
6886         if (!strcmp(dev_name, "all"))
6887                 ifindex = NETCONFA_IFINDEX_ALL;
6888         else if (!strcmp(dev_name, "default"))
6889                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6890         else
6891                 ifindex = idev->dev->ifindex;
6892         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6893                                      ifindex, p);
6894         return 0;
6895
6896 free:
6897         kfree(table);
6898 out:
6899         return -ENOBUFS;
6900 }
6901
6902 static void __addrconf_sysctl_unregister(struct net *net,
6903                                          struct ipv6_devconf *p, int ifindex)
6904 {
6905         struct ctl_table *table;
6906
6907         if (!p->sysctl_header)
6908                 return;
6909
6910         table = p->sysctl_header->ctl_table_arg;
6911         unregister_net_sysctl_table(p->sysctl_header);
6912         p->sysctl_header = NULL;
6913         kfree(table);
6914
6915         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6916 }
6917
6918 static int addrconf_sysctl_register(struct inet6_dev *idev)
6919 {
6920         int err;
6921
6922         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6923                 return -EINVAL;
6924
6925         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6926                                     &ndisc_ifinfo_sysctl_change);
6927         if (err)
6928                 return err;
6929         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6930                                          idev, &idev->cnf);
6931         if (err)
6932                 neigh_sysctl_unregister(idev->nd_parms);
6933
6934         return err;
6935 }
6936
6937 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6938 {
6939         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6940                                      idev->dev->ifindex);
6941         neigh_sysctl_unregister(idev->nd_parms);
6942 }
6943
6944
6945 #endif
6946
6947 static int __net_init addrconf_init_net(struct net *net)
6948 {
6949         int err = -ENOMEM;
6950         struct ipv6_devconf *all, *dflt;
6951
6952         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6953         if (!all)
6954                 goto err_alloc_all;
6955
6956         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6957         if (!dflt)
6958                 goto err_alloc_dflt;
6959
6960         if (IS_ENABLED(CONFIG_SYSCTL) &&
6961             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6962                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6963                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6964         }
6965
6966         /* these will be inherited by all namespaces */
6967         dflt->autoconf = ipv6_defaults.autoconf;
6968         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6969
6970         dflt->stable_secret.initialized = false;
6971         all->stable_secret.initialized = false;
6972
6973         net->ipv6.devconf_all = all;
6974         net->ipv6.devconf_dflt = dflt;
6975
6976 #ifdef CONFIG_SYSCTL
6977         err = __addrconf_sysctl_register(net, "all", NULL, all);
6978         if (err < 0)
6979                 goto err_reg_all;
6980
6981         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6982         if (err < 0)
6983                 goto err_reg_dflt;
6984 #endif
6985         return 0;
6986
6987 #ifdef CONFIG_SYSCTL
6988 err_reg_dflt:
6989         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6990 err_reg_all:
6991         kfree(dflt);
6992 #endif
6993 err_alloc_dflt:
6994         kfree(all);
6995 err_alloc_all:
6996         return err;
6997 }
6998
6999 static void __net_exit addrconf_exit_net(struct net *net)
7000 {
7001 #ifdef CONFIG_SYSCTL
7002         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7003                                      NETCONFA_IFINDEX_DEFAULT);
7004         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7005                                      NETCONFA_IFINDEX_ALL);
7006 #endif
7007         kfree(net->ipv6.devconf_dflt);
7008         kfree(net->ipv6.devconf_all);
7009 }
7010
7011 static struct pernet_operations addrconf_ops = {
7012         .init = addrconf_init_net,
7013         .exit = addrconf_exit_net,
7014 };
7015
7016 static struct rtnl_af_ops inet6_ops __read_mostly = {
7017         .family           = AF_INET6,
7018         .fill_link_af     = inet6_fill_link_af,
7019         .get_link_af_size = inet6_get_link_af_size,
7020         .validate_link_af = inet6_validate_link_af,
7021         .set_link_af      = inet6_set_link_af,
7022 };
7023
7024 /*
7025  *      Init / cleanup code
7026  */
7027
7028 int __init addrconf_init(void)
7029 {
7030         struct inet6_dev *idev;
7031         int i, err;
7032
7033         err = ipv6_addr_label_init();
7034         if (err < 0) {
7035                 pr_crit("%s: cannot initialize default policy table: %d\n",
7036                         __func__, err);
7037                 goto out;
7038         }
7039
7040         err = register_pernet_subsys(&addrconf_ops);
7041         if (err < 0)
7042                 goto out_addrlabel;
7043
7044         addrconf_wq = create_workqueue("ipv6_addrconf");
7045         if (!addrconf_wq) {
7046                 err = -ENOMEM;
7047                 goto out_nowq;
7048         }
7049
7050         /* The addrconf netdev notifier requires that loopback_dev
7051          * has it's ipv6 private information allocated and setup
7052          * before it can bring up and give link-local addresses
7053          * to other devices which are up.
7054          *
7055          * Unfortunately, loopback_dev is not necessarily the first
7056          * entry in the global dev_base list of net devices.  In fact,
7057          * it is likely to be the very last entry on that list.
7058          * So this causes the notifier registry below to try and
7059          * give link-local addresses to all devices besides loopback_dev
7060          * first, then loopback_dev, which cases all the non-loopback_dev
7061          * devices to fail to get a link-local address.
7062          *
7063          * So, as a temporary fix, allocate the ipv6 structure for
7064          * loopback_dev first by hand.
7065          * Longer term, all of the dependencies ipv6 has upon the loopback
7066          * device and it being up should be removed.
7067          */
7068         rtnl_lock();
7069         idev = ipv6_add_dev(init_net.loopback_dev);
7070         rtnl_unlock();
7071         if (IS_ERR(idev)) {
7072                 err = PTR_ERR(idev);
7073                 goto errlo;
7074         }
7075
7076         ip6_route_init_special_entries();
7077
7078         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7079                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7080
7081         register_netdevice_notifier(&ipv6_dev_notf);
7082
7083         addrconf_verify();
7084
7085         rtnl_af_register(&inet6_ops);
7086
7087         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7088                                    NULL, inet6_dump_ifinfo, 0);
7089         if (err < 0)
7090                 goto errout;
7091
7092         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7093                                    inet6_rtm_newaddr, NULL, 0);
7094         if (err < 0)
7095                 goto errout;
7096         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7097                                    inet6_rtm_deladdr, NULL, 0);
7098         if (err < 0)
7099                 goto errout;
7100         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7101                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7102                                    RTNL_FLAG_DOIT_UNLOCKED);
7103         if (err < 0)
7104                 goto errout;
7105         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7106                                    NULL, inet6_dump_ifmcaddr, 0);
7107         if (err < 0)
7108                 goto errout;
7109         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7110                                    NULL, inet6_dump_ifacaddr, 0);
7111         if (err < 0)
7112                 goto errout;
7113         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7114                                    inet6_netconf_get_devconf,
7115                                    inet6_netconf_dump_devconf,
7116                                    RTNL_FLAG_DOIT_UNLOCKED);
7117         if (err < 0)
7118                 goto errout;
7119         err = ipv6_addr_label_rtnl_register();
7120         if (err < 0)
7121                 goto errout;
7122
7123         return 0;
7124 errout:
7125         rtnl_unregister_all(PF_INET6);
7126         rtnl_af_unregister(&inet6_ops);
7127         unregister_netdevice_notifier(&ipv6_dev_notf);
7128 errlo:
7129         destroy_workqueue(addrconf_wq);
7130 out_nowq:
7131         unregister_pernet_subsys(&addrconf_ops);
7132 out_addrlabel:
7133         ipv6_addr_label_cleanup();
7134 out:
7135         return err;
7136 }
7137
7138 void addrconf_cleanup(void)
7139 {
7140         struct net_device *dev;
7141         int i;
7142
7143         unregister_netdevice_notifier(&ipv6_dev_notf);
7144         unregister_pernet_subsys(&addrconf_ops);
7145         ipv6_addr_label_cleanup();
7146
7147         rtnl_af_unregister(&inet6_ops);
7148
7149         rtnl_lock();
7150
7151         /* clean dev list */
7152         for_each_netdev(&init_net, dev) {
7153                 if (__in6_dev_get(dev) == NULL)
7154                         continue;
7155                 addrconf_ifdown(dev, 1);
7156         }
7157         addrconf_ifdown(init_net.loopback_dev, 2);
7158
7159         /*
7160          *      Check hash table.
7161          */
7162         spin_lock_bh(&addrconf_hash_lock);
7163         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7164                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7165         spin_unlock_bh(&addrconf_hash_lock);
7166         cancel_delayed_work(&addr_chk_work);
7167         rtnl_unlock();
7168
7169         destroy_workqueue(addrconf_wq);
7170 }