net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/sched/signal.h>
  47 #include <linux/socket.h>
  48 #include <linux/sockios.h>
  49 #include <linux/net.h>
  50 #include <linux/inet.h>
  51 #include <linux/in6.h>
  52 #include <linux/netdevice.h>
  53 #include <linux/if_addr.h>
  54 #include <linux/if_arp.h>
  55 #include <linux/if_arcnet.h>
  56 #include <linux/if_infiniband.h>
  57 #include <linux/route.h>
  58 #include <linux/inetdevice.h>
  59 #include <linux/init.h>
  60 #include <linux/slab.h>
  61 #ifdef CONFIG_SYSCTL
  62 #include <linux/sysctl.h>
  63 #endif
  64 #include <linux/capability.h>
  65 #include <linux/delay.h>
  66 #include <linux/notifier.h>
  67 #include <linux/string.h>
  68 #include <linux/hash.h>
  69
  70 #include <net/net_namespace.h>
  71 #include <net/sock.h>
  72 #include <net/snmp.h>
  73
  74 #include <net/6lowpan.h>
  75 #include <net/firewire.h>
  76 #include <net/ipv6.h>
  77 #include <net/protocol.h>
  78 #include <net/ndisc.h>
  79 #include <net/ip6_route.h>
  80 #include <net/addrconf.h>
  81 #include <net/tcp.h>
  82 #include <net/ip.h>
  83 #include <net/netlink.h>
  84 #include <net/pkt_sched.h>
  85 #include <net/l3mdev.h>
  86 #include <linux/if_tunnel.h>
  87 #include <linux/rtnetlink.h>
  88 #include <linux/netconf.h>
  89 #include <linux/random.h>
  90 #include <linux/uaccess.h>
  91 #include <asm/unaligned.h>
  92
  93 #include <linux/proc_fs.h>
  94 #include <linux/seq_file.h>
  95 #include <linux/export.h>
  96
  97 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  98
  99 #define IPV6_MAX_STRLEN \
 100         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
 108 {
 109         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
 110         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
 111         do_div(tmp, 1000000);
 112         return (s32)tmp;
 113 }
 114
 115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
 116 {
 117         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
 118         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
 119         do_div(tmp, 1000000);
 120         if ((s32)tmp > mrt) {
 121                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
 122                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
 123                 do_div(tmp, 1000000);
 124         }
 125         return (s32)tmp;
 126 }
 127
 128 #ifdef CONFIG_SYSCTL
 129 static int addrconf_sysctl_register(struct inet6_dev *idev);
 130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 131 #else
 132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 133 {
 134         return 0;
 135 }
 136
 137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 138 {
 139 }
 140 #endif
 141
 142 static void ipv6_regen_rndid(struct inet6_dev *idev);
 143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 144
 145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 146 static int ipv6_count_addresses(const struct inet6_dev *idev);
 147 static int ipv6_generate_stable_address(struct in6_addr *addr,
 148                                         u8 dad_count,
 149                                         const struct inet6_dev *idev);
 150
 151 #define IN6_ADDR_HSIZE_SHIFT    8
 152 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
 153 /*
 154  *      Configured unicast address hash table
 155  */
 156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 157 static DEFINE_SPINLOCK(addrconf_hash_lock);
 158
 159 static void addrconf_verify(void);
 160 static void addrconf_verify_rtnl(void);
 161 static void addrconf_verify_work(struct work_struct *);
 162
 163 static struct workqueue_struct *addrconf_wq;
 164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 165
 166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 168
 169 static void addrconf_type_change(struct net_device *dev,
 170                                  unsigned long event);
 171 static int addrconf_ifdown(struct net_device *dev, int how);
 172
 173 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 174                                                   int plen,
 175                                                   const struct net_device *dev,
 176                                                   u32 flags, u32 noflags);
 177
 178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 179 static void addrconf_dad_work(struct work_struct *w);
 180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
 181                                    bool send_na);
 182 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
 183 static void addrconf_rs_timer(struct timer_list *t);
 184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 186
 187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 188                                 struct prefix_info *pinfo);
 189
 190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 191         .forwarding             = 0,
 192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 193         .mtu6                   = IPV6_MIN_MTU,
 194         .accept_ra              = 1,
 195         .accept_redirects       = 1,
 196         .autoconf               = 1,
 197         .force_mld_version      = 0,
 198         .mldv1_unsolicited_report_interval = 10 * HZ,
 199         .mldv2_unsolicited_report_interval = HZ,
 200         .dad_transmits          = 1,
 201         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 202         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 203         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205         .use_tempaddr           = 0,
 206         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 207         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 208         .regen_max_retry        = REGEN_MAX_RETRY,
 209         .max_desync_factor      = MAX_DESYNC_FACTOR,
 210         .max_addresses          = IPV6_MAX_ADDRESSES,
 211         .accept_ra_defrtr       = 1,
 212         .accept_ra_from_local   = 0,
 213         .accept_ra_min_hop_limit= 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_min_plen = 0,
 220         .accept_ra_rt_info_max_plen = 0,
 221 #endif
 222 #endif
 223         .proxy_ndp              = 0,
 224         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 225         .disable_ipv6           = 0,
 226         .accept_dad             = 0,
 227         .suppress_frag_ndisc    = 1,
 228         .accept_ra_mtu          = 1,
 229         .stable_secret          = {
 230                 .initialized = false,
 231         },
 232         .use_oif_addrs_only     = 0,
 233         .ignore_routes_with_linkdown = 0,
 234         .keep_addr_on_down      = 0,
 235         .seg6_enabled           = 0,
 236 #ifdef CONFIG_IPV6_SEG6_HMAC
 237         .seg6_require_hmac      = 0,
 238 #endif
 239         .enhanced_dad           = 1,
 240         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 241         .disable_policy         = 0,
 242 };
 243
 244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 245         .forwarding             = 0,
 246         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 247         .mtu6                   = IPV6_MIN_MTU,
 248         .accept_ra              = 1,
 249         .accept_redirects       = 1,
 250         .autoconf               = 1,
 251         .force_mld_version      = 0,
 252         .mldv1_unsolicited_report_interval = 10 * HZ,
 253         .mldv2_unsolicited_report_interval = HZ,
 254         .dad_transmits          = 1,
 255         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 256         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 257         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
 258         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 259         .use_tempaddr           = 0,
 260         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 261         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 262         .regen_max_retry        = REGEN_MAX_RETRY,
 263         .max_desync_factor      = MAX_DESYNC_FACTOR,
 264         .max_addresses          = IPV6_MAX_ADDRESSES,
 265         .accept_ra_defrtr       = 1,
 266         .accept_ra_from_local   = 0,
 267         .accept_ra_min_hop_limit= 1,
 268         .accept_ra_pinfo        = 1,
 269 #ifdef CONFIG_IPV6_ROUTER_PREF
 270         .accept_ra_rtr_pref     = 1,
 271         .rtr_probe_interval     = 60 * HZ,
 272 #ifdef CONFIG_IPV6_ROUTE_INFO
 273         .accept_ra_rt_info_min_plen = 0,
 274         .accept_ra_rt_info_max_plen = 0,
 275 #endif
 276 #endif
 277         .proxy_ndp              = 0,
 278         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 279         .disable_ipv6           = 0,
 280         .accept_dad             = 1,
 281         .suppress_frag_ndisc    = 1,
 282         .accept_ra_mtu          = 1,
 283         .stable_secret          = {
 284                 .initialized = false,
 285         },
 286         .use_oif_addrs_only     = 0,
 287         .ignore_routes_with_linkdown = 0,
 288         .keep_addr_on_down      = 0,
 289         .seg6_enabled           = 0,
 290 #ifdef CONFIG_IPV6_SEG6_HMAC
 291         .seg6_require_hmac      = 0,
 292 #endif
 293         .enhanced_dad           = 1,
 294         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
 295         .disable_policy         = 0,
 296 };
 297
 298 /* Check if link is ready: is it up and is a valid qdisc available */
 299 static inline bool addrconf_link_ready(const struct net_device *dev)
 300 {
 301         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
 302 }
 303
 304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 305 {
 306         if (del_timer(&idev->rs_timer))
 307                 __in6_dev_put(idev);
 308 }
 309
 310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 311 {
 312         if (cancel_delayed_work(&ifp->dad_work))
 313                 __in6_ifa_put(ifp);
 314 }
 315
 316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 317                                   unsigned long when)
 318 {
 319         if (!timer_pending(&idev->rs_timer))
 320                 in6_dev_hold(idev);
 321         mod_timer(&idev->rs_timer, jiffies + when);
 322 }
 323
 324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 325                                    unsigned long delay)
 326 {
 327         in6_ifa_hold(ifp);
 328         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
 329                 in6_ifa_put(ifp);
 330 }
 331
 332 static int snmp6_alloc_dev(struct inet6_dev *idev)
 333 {
 334         int i;
 335
 336         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 337         if (!idev->stats.ipv6)
 338                 goto err_ip;
 339
 340         for_each_possible_cpu(i) {
 341                 struct ipstats_mib *addrconf_stats;
 342                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 343                 u64_stats_init(&addrconf_stats->syncp);
 344         }
 345
 346
 347         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 348                                         GFP_KERNEL);
 349         if (!idev->stats.icmpv6dev)
 350                 goto err_icmp;
 351         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 352                                            GFP_KERNEL);
 353         if (!idev->stats.icmpv6msgdev)
 354                 goto err_icmpmsg;
 355
 356         return 0;
 357
 358 err_icmpmsg:
 359         kfree(idev->stats.icmpv6dev);
 360 err_icmp:
 361         free_percpu(idev->stats.ipv6);
 362 err_ip:
 363         return -ENOMEM;
 364 }
 365
 366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 367 {
 368         struct inet6_dev *ndev;
 369         int err = -ENOMEM;
 370
 371         ASSERT_RTNL();
 372
 373         if (dev->mtu < IPV6_MIN_MTU)
 374                 return ERR_PTR(-EINVAL);
 375
 376         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 377         if (!ndev)
 378                 return ERR_PTR(err);
 379
 380         rwlock_init(&ndev->lock);
 381         ndev->dev = dev;
 382         INIT_LIST_HEAD(&ndev->addr_list);
 383         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
 384         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 385
 386         if (ndev->cnf.stable_secret.initialized)
 387                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
 388
 389         ndev->cnf.mtu6 = dev->mtu;
 390         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 391         if (!ndev->nd_parms) {
 392                 kfree(ndev);
 393                 return ERR_PTR(err);
 394         }
 395         if (ndev->cnf.forwarding)
 396                 dev_disable_lro(dev);
 397         /* We refer to the device */
 398         dev_hold(dev);
 399
 400         if (snmp6_alloc_dev(ndev) < 0) {
 401                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
 402                            __func__);
 403                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 404                 dev_put(dev);
 405                 kfree(ndev);
 406                 return ERR_PTR(err);
 407         }
 408
 409         if (snmp6_register_dev(ndev) < 0) {
 410                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
 411                            __func__, dev->name);
 412                 goto err_release;
 413         }
 414
 415         /* One reference from device. */
 416         refcount_set(&ndev->refcnt, 1);
 417
 418         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 419                 ndev->cnf.accept_dad = -1;
 420
 421 #if IS_ENABLED(CONFIG_IPV6_SIT)
 422         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 423                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 424                 ndev->cnf.rtr_solicits = 0;
 425         }
 426 #endif
 427
 428         INIT_LIST_HEAD(&ndev->tempaddr_list);
 429         ndev->desync_factor = U32_MAX;
 430         if ((dev->flags&IFF_LOOPBACK) ||
 431             dev->type == ARPHRD_TUNNEL ||
 432             dev->type == ARPHRD_TUNNEL6 ||
 433             dev->type == ARPHRD_SIT ||
 434             dev->type == ARPHRD_NONE) {
 435                 ndev->cnf.use_tempaddr = -1;
 436         } else
 437                 ipv6_regen_rndid(ndev);
 438
 439         ndev->token = in6addr_any;
 440
 441         if (netif_running(dev) && addrconf_link_ready(dev))
 442                 ndev->if_flags |= IF_READY;
 443
 444         ipv6_mc_init_dev(ndev);
 445         ndev->tstamp = jiffies;
 446         err = addrconf_sysctl_register(ndev);
 447         if (err) {
 448                 ipv6_mc_destroy_dev(ndev);
 449                 snmp6_unregister_dev(ndev);
 450                 goto err_release;
 451         }
 452         /* protected by rtnl_lock */
 453         rcu_assign_pointer(dev->ip6_ptr, ndev);
 454
 455         /* Join interface-local all-node multicast group */
 456         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 457
 458         /* Join all-node multicast group */
 459         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 460
 461         /* Join all-router multicast group if forwarding is set */
 462         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 463                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 464
 465         return ndev;
 466
 467 err_release:
 468         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 469         ndev->dead = 1;
 470         in6_dev_finish_destroy(ndev);
 471         return ERR_PTR(err);
 472 }
 473
 474 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 475 {
 476         struct inet6_dev *idev;
 477
 478         ASSERT_RTNL();
 479
 480         idev = __in6_dev_get(dev);
 481         if (!idev) {
 482                 idev = ipv6_add_dev(dev);
 483                 if (IS_ERR(idev))
 484                         return NULL;
 485         }
 486
 487         if (dev->flags&IFF_UP)
 488                 ipv6_mc_up(idev);
 489         return idev;
 490 }
 491
 492 static int inet6_netconf_msgsize_devconf(int type)
 493 {
 494         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 495                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 496         bool all = false;
 497
 498         if (type == NETCONFA_ALL)
 499                 all = true;
 500
 501         if (all || type == NETCONFA_FORWARDING)
 502                 size += nla_total_size(4);
 503 #ifdef CONFIG_IPV6_MROUTE
 504         if (all || type == NETCONFA_MC_FORWARDING)
 505                 size += nla_total_size(4);
 506 #endif
 507         if (all || type == NETCONFA_PROXY_NEIGH)
 508                 size += nla_total_size(4);
 509
 510         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
 511                 size += nla_total_size(4);
 512
 513         return size;
 514 }
 515
 516 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 517                                       struct ipv6_devconf *devconf, u32 portid,
 518                                       u32 seq, int event, unsigned int flags,
 519                                       int type)
 520 {
 521         struct nlmsghdr  *nlh;
 522         struct netconfmsg *ncm;
 523         bool all = false;
 524
 525         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 526                         flags);
 527         if (!nlh)
 528                 return -EMSGSIZE;
 529
 530         if (type == NETCONFA_ALL)
 531                 all = true;
 532
 533         ncm = nlmsg_data(nlh);
 534         ncm->ncm_family = AF_INET6;
 535
 536         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 537                 goto nla_put_failure;
 538
 539         if (!devconf)
 540                 goto out;
 541
 542         if ((all || type == NETCONFA_FORWARDING) &&
 543             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 544                 goto nla_put_failure;
 545 #ifdef CONFIG_IPV6_MROUTE
 546         if ((all || type == NETCONFA_MC_FORWARDING) &&
 547             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 548                         devconf->mc_forwarding) < 0)
 549                 goto nla_put_failure;
 550 #endif
 551         if ((all || type == NETCONFA_PROXY_NEIGH) &&
 552             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 553                 goto nla_put_failure;
 554
 555         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
 556             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 557                         devconf->ignore_routes_with_linkdown) < 0)
 558                 goto nla_put_failure;
 559
 560 out:
 561         nlmsg_end(skb, nlh);
 562         return 0;
 563
 564 nla_put_failure:
 565         nlmsg_cancel(skb, nlh);
 566         return -EMSGSIZE;
 567 }
 568
 569 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
 570                                   int ifindex, struct ipv6_devconf *devconf)
 571 {
 572         struct sk_buff *skb;
 573         int err = -ENOBUFS;
 574
 575         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
 576         if (!skb)
 577                 goto errout;
 578
 579         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 580                                          event, 0, type);
 581         if (err < 0) {
 582                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 583                 WARN_ON(err == -EMSGSIZE);
 584                 kfree_skb(skb);
 585                 goto errout;
 586         }
 587         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
 588         return;
 589 errout:
 590         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 591 }
 592
 593 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 594         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 595         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 596         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 597         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
 598 };
 599
 600 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
 601                                        const struct nlmsghdr *nlh,
 602                                        struct nlattr **tb,
 603                                        struct netlink_ext_ack *extack)
 604 {
 605         int i, err;
 606
 607         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
 608                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
 609                 return -EINVAL;
 610         }
 611
 612         if (!netlink_strict_get_check(skb))
 613                 return nlmsg_parse(nlh, sizeof(struct netconfmsg), tb,
 614                                    NETCONFA_MAX, devconf_ipv6_policy, extack);
 615
 616         err = nlmsg_parse_strict(nlh, sizeof(struct netconfmsg), tb,
 617                                  NETCONFA_MAX, devconf_ipv6_policy, extack);
 618         if (err)
 619                 return err;
 620
 621         for (i = 0; i <= NETCONFA_MAX; i++) {
 622                 if (!tb[i])
 623                         continue;
 624
 625                 switch (i) {
 626                 case NETCONFA_IFINDEX:
 627                         break;
 628                 default:
 629                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
 630                         return -EINVAL;
 631                 }
 632         }
 633
 634         return 0;
 635 }
 636
 637 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 638                                      struct nlmsghdr *nlh,
 639                                      struct netlink_ext_ack *extack)
 640 {
 641         struct net *net = sock_net(in_skb->sk);
 642         struct nlattr *tb[NETCONFA_MAX+1];
 643         struct inet6_dev *in6_dev = NULL;
 644         struct net_device *dev = NULL;
 645         struct sk_buff *skb;
 646         struct ipv6_devconf *devconf;
 647         int ifindex;
 648         int err;
 649
 650         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
 651         if (err < 0)
 652                 return err;
 653
 654         if (!tb[NETCONFA_IFINDEX])
 655                 return -EINVAL;
 656
 657         err = -EINVAL;
 658         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 659         switch (ifindex) {
 660         case NETCONFA_IFINDEX_ALL:
 661                 devconf = net->ipv6.devconf_all;
 662                 break;
 663         case NETCONFA_IFINDEX_DEFAULT:
 664                 devconf = net->ipv6.devconf_dflt;
 665                 break;
 666         default:
 667                 dev = dev_get_by_index(net, ifindex);
 668                 if (!dev)
 669                         return -EINVAL;
 670                 in6_dev = in6_dev_get(dev);
 671                 if (!in6_dev)
 672                         goto errout;
 673                 devconf = &in6_dev->cnf;
 674                 break;
 675         }
 676
 677         err = -ENOBUFS;
 678         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
 679         if (!skb)
 680                 goto errout;
 681
 682         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 683                                          NETLINK_CB(in_skb).portid,
 684                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 685                                          NETCONFA_ALL);
 686         if (err < 0) {
 687                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 688                 WARN_ON(err == -EMSGSIZE);
 689                 kfree_skb(skb);
 690                 goto errout;
 691         }
 692         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 693 errout:
 694         if (in6_dev)
 695                 in6_dev_put(in6_dev);
 696         if (dev)
 697                 dev_put(dev);
 698         return err;
 699 }
 700
 701 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 702                                       struct netlink_callback *cb)
 703 {
 704         const struct nlmsghdr *nlh = cb->nlh;
 705         struct net *net = sock_net(skb->sk);
 706         int h, s_h;
 707         int idx, s_idx;
 708         struct net_device *dev;
 709         struct inet6_dev *idev;
 710         struct hlist_head *head;
 711
 712         if (cb->strict_check) {
 713                 struct netlink_ext_ack *extack = cb->extack;
 714                 struct netconfmsg *ncm;
 715
 716                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
 717                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
 718                         return -EINVAL;
 719                 }
 720
 721                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
 722                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
 723                         return -EINVAL;
 724                 }
 725         }
 726
 727         s_h = cb->args[0];
 728         s_idx = idx = cb->args[1];
 729
 730         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 731                 idx = 0;
 732                 head = &net->dev_index_head[h];
 733                 rcu_read_lock();
 734                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 735                           net->dev_base_seq;
 736                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 737                         if (idx < s_idx)
 738                                 goto cont;
 739                         idev = __in6_dev_get(dev);
 740                         if (!idev)
 741                                 goto cont;
 742
 743                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 744                                                        &idev->cnf,
 745                                                        NETLINK_CB(cb->skb).portid,
 746                                                        nlh->nlmsg_seq,
 747                                                        RTM_NEWNETCONF,
 748                                                        NLM_F_MULTI,
 749                                                        NETCONFA_ALL) < 0) {
 750                                 rcu_read_unlock();
 751                                 goto done;
 752                         }
 753                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 754 cont:
 755                         idx++;
 756                 }
 757                 rcu_read_unlock();
 758         }
 759         if (h == NETDEV_HASHENTRIES) {
 760                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 761                                                net->ipv6.devconf_all,
 762                                                NETLINK_CB(cb->skb).portid,
 763                                                nlh->nlmsg_seq,
 764                                                RTM_NEWNETCONF, NLM_F_MULTI,
 765                                                NETCONFA_ALL) < 0)
 766                         goto done;
 767                 else
 768                         h++;
 769         }
 770         if (h == NETDEV_HASHENTRIES + 1) {
 771                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 772                                                net->ipv6.devconf_dflt,
 773                                                NETLINK_CB(cb->skb).portid,
 774                                                nlh->nlmsg_seq,
 775                                                RTM_NEWNETCONF, NLM_F_MULTI,
 776                                                NETCONFA_ALL) < 0)
 777                         goto done;
 778                 else
 779                         h++;
 780         }
 781 done:
 782         cb->args[0] = h;
 783         cb->args[1] = idx;
 784
 785         return skb->len;
 786 }
 787
 788 #ifdef CONFIG_SYSCTL
 789 static void dev_forward_change(struct inet6_dev *idev)
 790 {
 791         struct net_device *dev;
 792         struct inet6_ifaddr *ifa;
 793
 794         if (!idev)
 795                 return;
 796         dev = idev->dev;
 797         if (idev->cnf.forwarding)
 798                 dev_disable_lro(dev);
 799         if (dev->flags & IFF_MULTICAST) {
 800                 if (idev->cnf.forwarding) {
 801                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 802                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 803                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 804                 } else {
 805                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 806                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 807                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 808                 }
 809         }
 810
 811         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 812                 if (ifa->flags&IFA_F_TENTATIVE)
 813                         continue;
 814                 if (idev->cnf.forwarding)
 815                         addrconf_join_anycast(ifa);
 816                 else
 817                         addrconf_leave_anycast(ifa);
 818         }
 819         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
 820                                      NETCONFA_FORWARDING,
 821                                      dev->ifindex, &idev->cnf);
 822 }
 823
 824
 825 static void addrconf_forward_change(struct net *net, __s32 newf)
 826 {
 827         struct net_device *dev;
 828         struct inet6_dev *idev;
 829
 830         for_each_netdev(net, dev) {
 831                 idev = __in6_dev_get(dev);
 832                 if (idev) {
 833                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 834                         idev->cnf.forwarding = newf;
 835                         if (changed)
 836                                 dev_forward_change(idev);
 837                 }
 838         }
 839 }
 840
 841 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 842 {
 843         struct net *net;
 844         int old;
 845
 846         if (!rtnl_trylock())
 847                 return restart_syscall();
 848
 849         net = (struct net *)table->extra2;
 850         old = *p;
 851         *p = newf;
 852
 853         if (p == &net->ipv6.devconf_dflt->forwarding) {
 854                 if ((!newf) ^ (!old))
 855                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 856                                                      NETCONFA_FORWARDING,
 857                                                      NETCONFA_IFINDEX_DEFAULT,
 858                                                      net->ipv6.devconf_dflt);
 859                 rtnl_unlock();
 860                 return 0;
 861         }
 862
 863         if (p == &net->ipv6.devconf_all->forwarding) {
 864                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
 865
 866                 net->ipv6.devconf_dflt->forwarding = newf;
 867                 if ((!newf) ^ (!old_dflt))
 868                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 869                                                      NETCONFA_FORWARDING,
 870                                                      NETCONFA_IFINDEX_DEFAULT,
 871                                                      net->ipv6.devconf_dflt);
 872
 873                 addrconf_forward_change(net, newf);
 874                 if ((!newf) ^ (!old))
 875                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
 876                                                      NETCONFA_FORWARDING,
 877                                                      NETCONFA_IFINDEX_ALL,
 878                                                      net->ipv6.devconf_all);
 879         } else if ((!newf) ^ (!old))
 880                 dev_forward_change((struct inet6_dev *)table->extra1);
 881         rtnl_unlock();
 882
 883         if (newf)
 884                 rt6_purge_dflt_routers(net);
 885         return 1;
 886 }
 887
 888 static void addrconf_linkdown_change(struct net *net, __s32 newf)
 889 {
 890         struct net_device *dev;
 891         struct inet6_dev *idev;
 892
 893         for_each_netdev(net, dev) {
 894                 idev = __in6_dev_get(dev);
 895                 if (idev) {
 896                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
 897
 898                         idev->cnf.ignore_routes_with_linkdown = newf;
 899                         if (changed)
 900                                 inet6_netconf_notify_devconf(dev_net(dev),
 901                                                              RTM_NEWNETCONF,
 902                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 903                                                              dev->ifindex,
 904                                                              &idev->cnf);
 905                 }
 906         }
 907 }
 908
 909 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
 910 {
 911         struct net *net;
 912         int old;
 913
 914         if (!rtnl_trylock())
 915                 return restart_syscall();
 916
 917         net = (struct net *)table->extra2;
 918         old = *p;
 919         *p = newf;
 920
 921         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
 922                 if ((!newf) ^ (!old))
 923                         inet6_netconf_notify_devconf(net,
 924                                                      RTM_NEWNETCONF,
 925                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 926                                                      NETCONFA_IFINDEX_DEFAULT,
 927                                                      net->ipv6.devconf_dflt);
 928                 rtnl_unlock();
 929                 return 0;
 930         }
 931
 932         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
 933                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
 934                 addrconf_linkdown_change(net, newf);
 935                 if ((!newf) ^ (!old))
 936                         inet6_netconf_notify_devconf(net,
 937                                                      RTM_NEWNETCONF,
 938                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
 939                                                      NETCONFA_IFINDEX_ALL,
 940                                                      net->ipv6.devconf_all);
 941         }
 942         rtnl_unlock();
 943
 944         return 1;
 945 }
 946
 947 #endif
 948
 949 /* Nobody refers to this ifaddr, destroy it */
 950 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 951 {
 952         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 953
 954 #ifdef NET_REFCNT_DEBUG
 955         pr_debug("%s\n", __func__);
 956 #endif
 957
 958         in6_dev_put(ifp->idev);
 959
 960         if (cancel_delayed_work(&ifp->dad_work))
 961                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 962                           ifp);
 963
 964         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 965                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 966                 return;
 967         }
 968
 969         kfree_rcu(ifp, rcu);
 970 }
 971
 972 static void
 973 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 974 {
 975         struct list_head *p;
 976         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 977
 978         /*
 979          * Each device address list is sorted in order of scope -
 980          * global before linklocal.
 981          */
 982         list_for_each(p, &idev->addr_list) {
 983                 struct inet6_ifaddr *ifa
 984                         = list_entry(p, struct inet6_ifaddr, if_list);
 985                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 986                         break;
 987         }
 988
 989         list_add_tail_rcu(&ifp->if_list, p);
 990 }
 991
 992 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
 993 {
 994         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
 995
 996         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
 997 }
 998
 999 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1000                                struct net_device *dev, unsigned int hash)
1001 {
1002         struct inet6_ifaddr *ifp;
1003
1004         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1005                 if (!net_eq(dev_net(ifp->idev->dev), net))
1006                         continue;
1007                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1008                         if (!dev || ifp->idev->dev == dev)
1009                                 return true;
1010                 }
1011         }
1012         return false;
1013 }
1014
1015 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1016 {
1017         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1018         int err = 0;
1019
1020         spin_lock(&addrconf_hash_lock);
1021
1022         /* Ignore adding duplicate addresses on an interface */
1023         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1024                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1025                 err = -EEXIST;
1026         } else {
1027                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1028         }
1029
1030         spin_unlock(&addrconf_hash_lock);
1031
1032         return err;
1033 }
1034
1035 /* On success it returns ifp with increased reference count */
1036
1037 static struct inet6_ifaddr *
1038 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1039               bool can_block, struct netlink_ext_ack *extack)
1040 {
1041         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1042         int addr_type = ipv6_addr_type(cfg->pfx);
1043         struct net *net = dev_net(idev->dev);
1044         struct inet6_ifaddr *ifa = NULL;
1045         struct fib6_info *f6i = NULL;
1046         int err = 0;
1047
1048         if (addr_type == IPV6_ADDR_ANY ||
1049             addr_type & IPV6_ADDR_MULTICAST ||
1050             (!(idev->dev->flags & IFF_LOOPBACK) &&
1051              !netif_is_l3_master(idev->dev) &&
1052              addr_type & IPV6_ADDR_LOOPBACK))
1053                 return ERR_PTR(-EADDRNOTAVAIL);
1054
1055         if (idev->dead) {
1056                 err = -ENODEV;                  /*XXX*/
1057                 goto out;
1058         }
1059
1060         if (idev->cnf.disable_ipv6) {
1061                 err = -EACCES;
1062                 goto out;
1063         }
1064
1065         /* validator notifier needs to be blocking;
1066          * do not call in atomic context
1067          */
1068         if (can_block) {
1069                 struct in6_validator_info i6vi = {
1070                         .i6vi_addr = *cfg->pfx,
1071                         .i6vi_dev = idev,
1072                         .extack = extack,
1073                 };
1074
1075                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1076                 err = notifier_to_errno(err);
1077                 if (err < 0)
1078                         goto out;
1079         }
1080
1081         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1082         if (!ifa) {
1083                 err = -ENOBUFS;
1084                 goto out;
1085         }
1086
1087         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1088         if (IS_ERR(f6i)) {
1089                 err = PTR_ERR(f6i);
1090                 f6i = NULL;
1091                 goto out;
1092         }
1093
1094         if (net->ipv6.devconf_all->disable_policy ||
1095             idev->cnf.disable_policy)
1096                 f6i->dst_nopolicy = true;
1097
1098         neigh_parms_data_state_setall(idev->nd_parms);
1099
1100         ifa->addr = *cfg->pfx;
1101         if (cfg->peer_pfx)
1102                 ifa->peer_addr = *cfg->peer_pfx;
1103
1104         spin_lock_init(&ifa->lock);
1105         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1106         INIT_HLIST_NODE(&ifa->addr_lst);
1107         ifa->scope = cfg->scope;
1108         ifa->prefix_len = cfg->plen;
1109         ifa->rt_priority = cfg->rt_priority;
1110         ifa->flags = cfg->ifa_flags;
1111         /* No need to add the TENTATIVE flag for addresses with NODAD */
1112         if (!(cfg->ifa_flags & IFA_F_NODAD))
1113                 ifa->flags |= IFA_F_TENTATIVE;
1114         ifa->valid_lft = cfg->valid_lft;
1115         ifa->prefered_lft = cfg->preferred_lft;
1116         ifa->cstamp = ifa->tstamp = jiffies;
1117         ifa->tokenized = false;
1118
1119         ifa->rt = f6i;
1120
1121         ifa->idev = idev;
1122         in6_dev_hold(idev);
1123
1124         /* For caller */
1125         refcount_set(&ifa->refcnt, 1);
1126
1127         rcu_read_lock_bh();
1128
1129         err = ipv6_add_addr_hash(idev->dev, ifa);
1130         if (err < 0) {
1131                 rcu_read_unlock_bh();
1132                 goto out;
1133         }
1134
1135         write_lock(&idev->lock);
1136
1137         /* Add to inet6_dev unicast addr list. */
1138         ipv6_link_dev_addr(idev, ifa);
1139
1140         if (ifa->flags&IFA_F_TEMPORARY) {
1141                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1142                 in6_ifa_hold(ifa);
1143         }
1144
1145         in6_ifa_hold(ifa);
1146         write_unlock(&idev->lock);
1147
1148         rcu_read_unlock_bh();
1149
1150         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1151 out:
1152         if (unlikely(err < 0)) {
1153                 fib6_info_release(f6i);
1154
1155                 if (ifa) {
1156                         if (ifa->idev)
1157                                 in6_dev_put(ifa->idev);
1158                         kfree(ifa);
1159                 }
1160                 ifa = ERR_PTR(err);
1161         }
1162
1163         return ifa;
1164 }
1165
1166 enum cleanup_prefix_rt_t {
1167         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1168         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1169         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1170 };
1171
1172 /*
1173  * Check, whether the prefix for ifp would still need a prefix route
1174  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1175  * constants.
1176  *
1177  * 1) we don't purge prefix if address was not permanent.
1178  *    prefix is managed by its own lifetime.
1179  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1180  * 3) if there are no addresses, delete prefix.
1181  * 4) if there are still other permanent address(es),
1182  *    corresponding prefix is still permanent.
1183  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1184  *    don't purge the prefix, assume user space is managing it.
1185  * 6) otherwise, update prefix lifetime to the
1186  *    longest valid lifetime among the corresponding
1187  *    addresses on the device.
1188  *    Note: subsequent RA will update lifetime.
1189  **/
1190 static enum cleanup_prefix_rt_t
1191 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1192 {
1193         struct inet6_ifaddr *ifa;
1194         struct inet6_dev *idev = ifp->idev;
1195         unsigned long lifetime;
1196         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1197
1198         *expires = jiffies;
1199
1200         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1201                 if (ifa == ifp)
1202                         continue;
1203                 if (ifa->prefix_len != ifp->prefix_len ||
1204                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1205                                        ifp->prefix_len))
1206                         continue;
1207                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1208                         return CLEANUP_PREFIX_RT_NOP;
1209
1210                 action = CLEANUP_PREFIX_RT_EXPIRE;
1211
1212                 spin_lock(&ifa->lock);
1213
1214                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1215                 /*
1216                  * Note: Because this address is
1217                  * not permanent, lifetime <
1218                  * LONG_MAX / HZ here.
1219                  */
1220                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1221                         *expires = ifa->tstamp + lifetime * HZ;
1222                 spin_unlock(&ifa->lock);
1223         }
1224
1225         return action;
1226 }
1227
1228 static void
1229 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1230 {
1231         struct fib6_info *f6i;
1232
1233         f6i = addrconf_get_prefix_route(&ifp->addr,
1234                                        ifp->prefix_len,
1235                                        ifp->idev->dev,
1236                                        0, RTF_GATEWAY | RTF_DEFAULT);
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);
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 {
2407         struct fib6_node *fn;
2408         struct fib6_info *rt = NULL;
2409         struct fib6_table *table;
2410         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2411
2412         table = fib6_get_table(dev_net(dev), tb_id);
2413         if (!table)
2414                 return NULL;
2415
2416         rcu_read_lock();
2417         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2418         if (!fn)
2419                 goto out;
2420
2421         for_each_fib6_node_rt_rcu(fn) {
2422                 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2423                         continue;
2424                 if ((rt->fib6_flags & flags) != flags)
2425                         continue;
2426                 if ((rt->fib6_flags & noflags) != 0)
2427                         continue;
2428                 if (!fib6_info_hold_safe(rt))
2429                         continue;
2430                 break;
2431         }
2432 out:
2433         rcu_read_unlock();
2434         return rt;
2435 }
2436
2437
2438 /* Create "default" multicast route to the interface */
2439
2440 static void addrconf_add_mroute(struct net_device *dev)
2441 {
2442         struct fib6_config cfg = {
2443                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2444                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2445                 .fc_ifindex = dev->ifindex,
2446                 .fc_dst_len = 8,
2447                 .fc_flags = RTF_UP,
2448                 .fc_type = RTN_UNICAST,
2449                 .fc_nlinfo.nl_net = dev_net(dev),
2450         };
2451
2452         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2453
2454         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2455 }
2456
2457 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2458 {
2459         struct inet6_dev *idev;
2460
2461         ASSERT_RTNL();
2462
2463         idev = ipv6_find_idev(dev);
2464         if (!idev)
2465                 return ERR_PTR(-ENOBUFS);
2466
2467         if (idev->cnf.disable_ipv6)
2468                 return ERR_PTR(-EACCES);
2469
2470         /* Add default multicast route */
2471         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2472                 addrconf_add_mroute(dev);
2473
2474         return idev;
2475 }
2476
2477 static void manage_tempaddrs(struct inet6_dev *idev,
2478                              struct inet6_ifaddr *ifp,
2479                              __u32 valid_lft, __u32 prefered_lft,
2480                              bool create, unsigned long now)
2481 {
2482         u32 flags;
2483         struct inet6_ifaddr *ift;
2484
2485         read_lock_bh(&idev->lock);
2486         /* update all temporary addresses in the list */
2487         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2488                 int age, max_valid, max_prefered;
2489
2490                 if (ifp != ift->ifpub)
2491                         continue;
2492
2493                 /* RFC 4941 section 3.3:
2494                  * If a received option will extend the lifetime of a public
2495                  * address, the lifetimes of temporary addresses should
2496                  * be extended, subject to the overall constraint that no
2497                  * temporary addresses should ever remain "valid" or "preferred"
2498                  * for a time longer than (TEMP_VALID_LIFETIME) or
2499                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2500                  */
2501                 age = (now - ift->cstamp) / HZ;
2502                 max_valid = idev->cnf.temp_valid_lft - age;
2503                 if (max_valid < 0)
2504                         max_valid = 0;
2505
2506                 max_prefered = idev->cnf.temp_prefered_lft -
2507                                idev->desync_factor - age;
2508                 if (max_prefered < 0)
2509                         max_prefered = 0;
2510
2511                 if (valid_lft > max_valid)
2512                         valid_lft = max_valid;
2513
2514                 if (prefered_lft > max_prefered)
2515                         prefered_lft = max_prefered;
2516
2517                 spin_lock(&ift->lock);
2518                 flags = ift->flags;
2519                 ift->valid_lft = valid_lft;
2520                 ift->prefered_lft = prefered_lft;
2521                 ift->tstamp = now;
2522                 if (prefered_lft > 0)
2523                         ift->flags &= ~IFA_F_DEPRECATED;
2524
2525                 spin_unlock(&ift->lock);
2526                 if (!(flags&IFA_F_TENTATIVE))
2527                         ipv6_ifa_notify(0, ift);
2528         }
2529
2530         if ((create || list_empty(&idev->tempaddr_list)) &&
2531             idev->cnf.use_tempaddr > 0) {
2532                 /* When a new public address is created as described
2533                  * in [ADDRCONF], also create a new temporary address.
2534                  * Also create a temporary address if it's enabled but
2535                  * no temporary address currently exists.
2536                  */
2537                 read_unlock_bh(&idev->lock);
2538                 ipv6_create_tempaddr(ifp, NULL, false);
2539         } else {
2540                 read_unlock_bh(&idev->lock);
2541         }
2542 }
2543
2544 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2545 {
2546         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2547                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2548 }
2549
2550 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2551                                  const struct prefix_info *pinfo,
2552                                  struct inet6_dev *in6_dev,
2553                                  const struct in6_addr *addr, int addr_type,
2554                                  u32 addr_flags, bool sllao, bool tokenized,
2555                                  __u32 valid_lft, u32 prefered_lft)
2556 {
2557         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2558         int create = 0, update_lft = 0;
2559
2560         if (!ifp && valid_lft) {
2561                 int max_addresses = in6_dev->cnf.max_addresses;
2562                 struct ifa6_config cfg = {
2563                         .pfx = addr,
2564                         .plen = pinfo->prefix_len,
2565                         .ifa_flags = addr_flags,
2566                         .valid_lft = valid_lft,
2567                         .preferred_lft = prefered_lft,
2568                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2569                 };
2570
2571 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2572                 if ((net->ipv6.devconf_all->optimistic_dad ||
2573                      in6_dev->cnf.optimistic_dad) &&
2574                     !net->ipv6.devconf_all->forwarding && sllao)
2575                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2576 #endif
2577
2578                 /* Do not allow to create too much of autoconfigured
2579                  * addresses; this would be too easy way to crash kernel.
2580                  */
2581                 if (!max_addresses ||
2582                     ipv6_count_addresses(in6_dev) < max_addresses)
2583                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2584
2585                 if (IS_ERR_OR_NULL(ifp))
2586                         return -1;
2587
2588                 create = 1;
2589                 spin_lock_bh(&ifp->lock);
2590                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2591                 ifp->cstamp = jiffies;
2592                 ifp->tokenized = tokenized;
2593                 spin_unlock_bh(&ifp->lock);
2594                 addrconf_dad_start(ifp);
2595         }
2596
2597         if (ifp) {
2598                 u32 flags;
2599                 unsigned long now;
2600                 u32 stored_lft;
2601
2602                 /* update lifetime (RFC2462 5.5.3 e) */
2603                 spin_lock_bh(&ifp->lock);
2604                 now = jiffies;
2605                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2606                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2607                 else
2608                         stored_lft = 0;
2609                 if (!create && stored_lft) {
2610                         const u32 minimum_lft = min_t(u32,
2611                                 stored_lft, MIN_VALID_LIFETIME);
2612                         valid_lft = max(valid_lft, minimum_lft);
2613
2614                         /* RFC4862 Section 5.5.3e:
2615                          * "Note that the preferred lifetime of the
2616                          *  corresponding address is always reset to
2617                          *  the Preferred Lifetime in the received
2618                          *  Prefix Information option, regardless of
2619                          *  whether the valid lifetime is also reset or
2620                          *  ignored."
2621                          *
2622                          * So we should always update prefered_lft here.
2623                          */
2624                         update_lft = 1;
2625                 }
2626
2627                 if (update_lft) {
2628                         ifp->valid_lft = valid_lft;
2629                         ifp->prefered_lft = prefered_lft;
2630                         ifp->tstamp = now;
2631                         flags = ifp->flags;
2632                         ifp->flags &= ~IFA_F_DEPRECATED;
2633                         spin_unlock_bh(&ifp->lock);
2634
2635                         if (!(flags&IFA_F_TENTATIVE))
2636                                 ipv6_ifa_notify(0, ifp);
2637                 } else
2638                         spin_unlock_bh(&ifp->lock);
2639
2640                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2641                                  create, now);
2642
2643                 in6_ifa_put(ifp);
2644                 addrconf_verify();
2645         }
2646
2647         return 0;
2648 }
2649 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2650
2651 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2652 {
2653         struct prefix_info *pinfo;
2654         __u32 valid_lft;
2655         __u32 prefered_lft;
2656         int addr_type, err;
2657         u32 addr_flags = 0;
2658         struct inet6_dev *in6_dev;
2659         struct net *net = dev_net(dev);
2660
2661         pinfo = (struct prefix_info *) opt;
2662
2663         if (len < sizeof(struct prefix_info)) {
2664                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2665                 return;
2666         }
2667
2668         /*
2669          *      Validation checks ([ADDRCONF], page 19)
2670          */
2671
2672         addr_type = ipv6_addr_type(&pinfo->prefix);
2673
2674         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2675                 return;
2676
2677         valid_lft = ntohl(pinfo->valid);
2678         prefered_lft = ntohl(pinfo->prefered);
2679
2680         if (prefered_lft > valid_lft) {
2681                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2682                 return;
2683         }
2684
2685         in6_dev = in6_dev_get(dev);
2686
2687         if (!in6_dev) {
2688                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2689                                     dev->name);
2690                 return;
2691         }
2692
2693         /*
2694          *      Two things going on here:
2695          *      1) Add routes for on-link prefixes
2696          *      2) Configure prefixes with the auto flag set
2697          */
2698
2699         if (pinfo->onlink) {
2700                 struct fib6_info *rt;
2701                 unsigned long rt_expires;
2702
2703                 /* Avoid arithmetic overflow. Really, we could
2704                  * save rt_expires in seconds, likely valid_lft,
2705                  * but it would require division in fib gc, that it
2706                  * not good.
2707                  */
2708                 if (HZ > USER_HZ)
2709                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2710                 else
2711                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2712
2713                 if (addrconf_finite_timeout(rt_expires))
2714                         rt_expires *= HZ;
2715
2716                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2717                                                pinfo->prefix_len,
2718                                                dev,
2719                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2720                                                RTF_GATEWAY | RTF_DEFAULT);
2721
2722                 if (rt) {
2723                         /* Autoconf prefix route */
2724                         if (valid_lft == 0) {
2725                                 ip6_del_rt(net, rt);
2726                                 rt = NULL;
2727                         } else if (addrconf_finite_timeout(rt_expires)) {
2728                                 /* not infinity */
2729                                 fib6_set_expires(rt, jiffies + rt_expires);
2730                         } else {
2731                                 fib6_clean_expires(rt);
2732                         }
2733                 } else if (valid_lft) {
2734                         clock_t expires = 0;
2735                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2736                         if (addrconf_finite_timeout(rt_expires)) {
2737                                 /* not infinity */
2738                                 flags |= RTF_EXPIRES;
2739                                 expires = jiffies_to_clock_t(rt_expires);
2740                         }
2741                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2742                                               0, dev, expires, flags,
2743                                               GFP_ATOMIC);
2744                 }
2745                 fib6_info_release(rt);
2746         }
2747
2748         /* Try to figure out our local address for this prefix */
2749
2750         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2751                 struct in6_addr addr;
2752                 bool tokenized = false, dev_addr_generated = false;
2753
2754                 if (pinfo->prefix_len == 64) {
2755                         memcpy(&addr, &pinfo->prefix, 8);
2756
2757                         if (!ipv6_addr_any(&in6_dev->token)) {
2758                                 read_lock_bh(&in6_dev->lock);
2759                                 memcpy(addr.s6_addr + 8,
2760                                        in6_dev->token.s6_addr + 8, 8);
2761                                 read_unlock_bh(&in6_dev->lock);
2762                                 tokenized = true;
2763                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2764                                    !ipv6_generate_stable_address(&addr, 0,
2765                                                                  in6_dev)) {
2766                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2767                                 goto ok;
2768                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2769                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2770                                 goto put;
2771                         } else {
2772                                 dev_addr_generated = true;
2773                         }
2774                         goto ok;
2775                 }
2776                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2777                                     pinfo->prefix_len);
2778                 goto put;
2779
2780 ok:
2781                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2782                                                    &addr, addr_type,
2783                                                    addr_flags, sllao,
2784                                                    tokenized, valid_lft,
2785                                                    prefered_lft);
2786                 if (err)
2787                         goto put;
2788
2789                 /* Ignore error case here because previous prefix add addr was
2790                  * successful which will be notified.
2791                  */
2792                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2793                                               addr_type, addr_flags, sllao,
2794                                               tokenized, valid_lft,
2795                                               prefered_lft,
2796                                               dev_addr_generated);
2797         }
2798         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2799 put:
2800         in6_dev_put(in6_dev);
2801 }
2802
2803 /*
2804  *      Set destination address.
2805  *      Special case for SIT interfaces where we create a new "virtual"
2806  *      device.
2807  */
2808 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2809 {
2810         struct in6_ifreq ireq;
2811         struct net_device *dev;
2812         int err = -EINVAL;
2813
2814         rtnl_lock();
2815
2816         err = -EFAULT;
2817         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2818                 goto err_exit;
2819
2820         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2821
2822         err = -ENODEV;
2823         if (!dev)
2824                 goto err_exit;
2825
2826 #if IS_ENABLED(CONFIG_IPV6_SIT)
2827         if (dev->type == ARPHRD_SIT) {
2828                 const struct net_device_ops *ops = dev->netdev_ops;
2829                 struct ifreq ifr;
2830                 struct ip_tunnel_parm p;
2831
2832                 err = -EADDRNOTAVAIL;
2833                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2834                         goto err_exit;
2835
2836                 memset(&p, 0, sizeof(p));
2837                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2838                 p.iph.saddr = 0;
2839                 p.iph.version = 4;
2840                 p.iph.ihl = 5;
2841                 p.iph.protocol = IPPROTO_IPV6;
2842                 p.iph.ttl = 64;
2843                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2844
2845                 if (ops->ndo_do_ioctl) {
2846                         mm_segment_t oldfs = get_fs();
2847
2848                         set_fs(KERNEL_DS);
2849                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2850                         set_fs(oldfs);
2851                 } else
2852                         err = -EOPNOTSUPP;
2853
2854                 if (err == 0) {
2855                         err = -ENOBUFS;
2856                         dev = __dev_get_by_name(net, p.name);
2857                         if (!dev)
2858                                 goto err_exit;
2859                         err = dev_open(dev, NULL);
2860                 }
2861         }
2862 #endif
2863
2864 err_exit:
2865         rtnl_unlock();
2866         return err;
2867 }
2868
2869 static int ipv6_mc_config(struct sock *sk, bool join,
2870                           const struct in6_addr *addr, int ifindex)
2871 {
2872         int ret;
2873
2874         ASSERT_RTNL();
2875
2876         lock_sock(sk);
2877         if (join)
2878                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2879         else
2880                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2881         release_sock(sk);
2882
2883         return ret;
2884 }
2885
2886 /*
2887  *      Manual configuration of address on an interface
2888  */
2889 static int inet6_addr_add(struct net *net, int ifindex,
2890                           struct ifa6_config *cfg,
2891                           struct netlink_ext_ack *extack)
2892 {
2893         struct inet6_ifaddr *ifp;
2894         struct inet6_dev *idev;
2895         struct net_device *dev;
2896         unsigned long timeout;
2897         clock_t expires;
2898         u32 flags;
2899
2900         ASSERT_RTNL();
2901
2902         if (cfg->plen > 128)
2903                 return -EINVAL;
2904
2905         /* check the lifetime */
2906         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2907                 return -EINVAL;
2908
2909         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2910                 return -EINVAL;
2911
2912         dev = __dev_get_by_index(net, ifindex);
2913         if (!dev)
2914                 return -ENODEV;
2915
2916         idev = addrconf_add_dev(dev);
2917         if (IS_ERR(idev))
2918                 return PTR_ERR(idev);
2919
2920         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2921                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2922                                          true, cfg->pfx, ifindex);
2923
2924                 if (ret < 0)
2925                         return ret;
2926         }
2927
2928         cfg->scope = ipv6_addr_scope(cfg->pfx);
2929
2930         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2931         if (addrconf_finite_timeout(timeout)) {
2932                 expires = jiffies_to_clock_t(timeout * HZ);
2933                 cfg->valid_lft = timeout;
2934                 flags = RTF_EXPIRES;
2935         } else {
2936                 expires = 0;
2937                 flags = 0;
2938                 cfg->ifa_flags |= IFA_F_PERMANENT;
2939         }
2940
2941         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2942         if (addrconf_finite_timeout(timeout)) {
2943                 if (timeout == 0)
2944                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2945                 cfg->preferred_lft = timeout;
2946         }
2947
2948         ifp = ipv6_add_addr(idev, cfg, true, extack);
2949         if (!IS_ERR(ifp)) {
2950                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2951                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2952                                               ifp->rt_priority, dev, expires,
2953                                               flags, GFP_KERNEL);
2954                 }
2955
2956                 /* Send a netlink notification if DAD is enabled and
2957                  * optimistic flag is not set
2958                  */
2959                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2960                         ipv6_ifa_notify(0, ifp);
2961                 /*
2962                  * Note that section 3.1 of RFC 4429 indicates
2963                  * that the Optimistic flag should not be set for
2964                  * manually configured addresses
2965                  */
2966                 addrconf_dad_start(ifp);
2967                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2968                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2969                                          cfg->preferred_lft, true, jiffies);
2970                 in6_ifa_put(ifp);
2971                 addrconf_verify_rtnl();
2972                 return 0;
2973         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2974                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2975                                cfg->pfx, ifindex);
2976         }
2977
2978         return PTR_ERR(ifp);
2979 }
2980
2981 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2982                           const struct in6_addr *pfx, unsigned int plen)
2983 {
2984         struct inet6_ifaddr *ifp;
2985         struct inet6_dev *idev;
2986         struct net_device *dev;
2987
2988         if (plen > 128)
2989                 return -EINVAL;
2990
2991         dev = __dev_get_by_index(net, ifindex);
2992         if (!dev)
2993                 return -ENODEV;
2994
2995         idev = __in6_dev_get(dev);
2996         if (!idev)
2997                 return -ENXIO;
2998
2999         read_lock_bh(&idev->lock);
3000         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3001                 if (ifp->prefix_len == plen &&
3002                     ipv6_addr_equal(pfx, &ifp->addr)) {
3003                         in6_ifa_hold(ifp);
3004                         read_unlock_bh(&idev->lock);
3005
3006                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3007                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3008                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3009                                                  jiffies);
3010                         ipv6_del_addr(ifp);
3011                         addrconf_verify_rtnl();
3012                         if (ipv6_addr_is_multicast(pfx)) {
3013                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3014                                                false, pfx, dev->ifindex);
3015                         }
3016                         return 0;
3017                 }
3018         }
3019         read_unlock_bh(&idev->lock);
3020         return -EADDRNOTAVAIL;
3021 }
3022
3023
3024 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3025 {
3026         struct ifa6_config cfg = {
3027                 .ifa_flags = IFA_F_PERMANENT,
3028                 .preferred_lft = INFINITY_LIFE_TIME,
3029                 .valid_lft = INFINITY_LIFE_TIME,
3030         };
3031         struct in6_ifreq ireq;
3032         int err;
3033
3034         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3035                 return -EPERM;
3036
3037         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3038                 return -EFAULT;
3039
3040         cfg.pfx = &ireq.ifr6_addr;
3041         cfg.plen = ireq.ifr6_prefixlen;
3042
3043         rtnl_lock();
3044         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3045         rtnl_unlock();
3046         return err;
3047 }
3048
3049 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3050 {
3051         struct in6_ifreq ireq;
3052         int err;
3053
3054         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3055                 return -EPERM;
3056
3057         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3058                 return -EFAULT;
3059
3060         rtnl_lock();
3061         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3062                              ireq.ifr6_prefixlen);
3063         rtnl_unlock();
3064         return err;
3065 }
3066
3067 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3068                      int plen, int scope)
3069 {
3070         struct inet6_ifaddr *ifp;
3071         struct ifa6_config cfg = {
3072                 .pfx = addr,
3073                 .plen = plen,
3074                 .ifa_flags = IFA_F_PERMANENT,
3075                 .valid_lft = INFINITY_LIFE_TIME,
3076                 .preferred_lft = INFINITY_LIFE_TIME,
3077                 .scope = scope
3078         };
3079
3080         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3081         if (!IS_ERR(ifp)) {
3082                 spin_lock_bh(&ifp->lock);
3083                 ifp->flags &= ~IFA_F_TENTATIVE;
3084                 spin_unlock_bh(&ifp->lock);
3085                 rt_genid_bump_ipv6(dev_net(idev->dev));
3086                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3087                 in6_ifa_put(ifp);
3088         }
3089 }
3090
3091 #if IS_ENABLED(CONFIG_IPV6_SIT)
3092 static void sit_add_v4_addrs(struct inet6_dev *idev)
3093 {
3094         struct in6_addr addr;
3095         struct net_device *dev;
3096         struct net *net = dev_net(idev->dev);
3097         int scope, plen;
3098         u32 pflags = 0;
3099
3100         ASSERT_RTNL();
3101
3102         memset(&addr, 0, sizeof(struct in6_addr));
3103         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3104
3105         if (idev->dev->flags&IFF_POINTOPOINT) {
3106                 addr.s6_addr32[0] = htonl(0xfe800000);
3107                 scope = IFA_LINK;
3108                 plen = 64;
3109         } else {
3110                 scope = IPV6_ADDR_COMPATv4;
3111                 plen = 96;
3112                 pflags |= RTF_NONEXTHOP;
3113         }
3114
3115         if (addr.s6_addr32[3]) {
3116                 add_addr(idev, &addr, plen, scope);
3117                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3118                                       GFP_KERNEL);
3119                 return;
3120         }
3121
3122         for_each_netdev(net, dev) {
3123                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3124                 if (in_dev && (dev->flags & IFF_UP)) {
3125                         struct in_ifaddr *ifa;
3126
3127                         int flag = scope;
3128
3129                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3130
3131                                 addr.s6_addr32[3] = ifa->ifa_local;
3132
3133                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3134                                         continue;
3135                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3136                                         if (idev->dev->flags&IFF_POINTOPOINT)
3137                                                 continue;
3138                                         flag |= IFA_HOST;
3139                                 }
3140
3141                                 add_addr(idev, &addr, plen, flag);
3142                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3143                                                       0, pflags, GFP_KERNEL);
3144                         }
3145                 }
3146         }
3147 }
3148 #endif
3149
3150 static void init_loopback(struct net_device *dev)
3151 {
3152         struct inet6_dev  *idev;
3153
3154         /* ::1 */
3155
3156         ASSERT_RTNL();
3157
3158         idev = ipv6_find_idev(dev);
3159         if (!idev) {
3160                 pr_debug("%s: add_dev failed\n", __func__);
3161                 return;
3162         }
3163
3164         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3165 }
3166
3167 void addrconf_add_linklocal(struct inet6_dev *idev,
3168                             const struct in6_addr *addr, u32 flags)
3169 {
3170         struct ifa6_config cfg = {
3171                 .pfx = addr,
3172                 .plen = 64,
3173                 .ifa_flags = flags | IFA_F_PERMANENT,
3174                 .valid_lft = INFINITY_LIFE_TIME,
3175                 .preferred_lft = INFINITY_LIFE_TIME,
3176                 .scope = IFA_LINK
3177         };
3178         struct inet6_ifaddr *ifp;
3179
3180 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3181         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3182              idev->cnf.optimistic_dad) &&
3183             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3184                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3185 #endif
3186
3187         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3188         if (!IS_ERR(ifp)) {
3189                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3190                                       0, 0, GFP_ATOMIC);
3191                 addrconf_dad_start(ifp);
3192                 in6_ifa_put(ifp);
3193         }
3194 }
3195 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3196
3197 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3198 {
3199         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3200                 return true;
3201
3202         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3203             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3204                 return true;
3205
3206         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3207             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3208                 return true;
3209
3210         return false;
3211 }
3212
3213 static int ipv6_generate_stable_address(struct in6_addr *address,
3214                                         u8 dad_count,
3215                                         const struct inet6_dev *idev)
3216 {
3217         static DEFINE_SPINLOCK(lock);
3218         static __u32 digest[SHA_DIGEST_WORDS];
3219         static __u32 workspace[SHA_WORKSPACE_WORDS];
3220
3221         static union {
3222                 char __data[SHA_MESSAGE_BYTES];
3223                 struct {
3224                         struct in6_addr secret;
3225                         __be32 prefix[2];
3226                         unsigned char hwaddr[MAX_ADDR_LEN];
3227                         u8 dad_count;
3228                 } __packed;
3229         } data;
3230
3231         struct in6_addr secret;
3232         struct in6_addr temp;
3233         struct net *net = dev_net(idev->dev);
3234
3235         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3236
3237         if (idev->cnf.stable_secret.initialized)
3238                 secret = idev->cnf.stable_secret.secret;
3239         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3240                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3241         else
3242                 return -1;
3243
3244 retry:
3245         spin_lock_bh(&lock);
3246
3247         sha_init(digest);
3248         memset(&data, 0, sizeof(data));
3249         memset(workspace, 0, sizeof(workspace));
3250         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3251         data.prefix[0] = address->s6_addr32[0];
3252         data.prefix[1] = address->s6_addr32[1];
3253         data.secret = secret;
3254         data.dad_count = dad_count;
3255
3256         sha_transform(digest, data.__data, workspace);
3257
3258         temp = *address;
3259         temp.s6_addr32[2] = (__force __be32)digest[0];
3260         temp.s6_addr32[3] = (__force __be32)digest[1];
3261
3262         spin_unlock_bh(&lock);
3263
3264         if (ipv6_reserved_interfaceid(temp)) {
3265                 dad_count++;
3266                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3267                         return -1;
3268                 goto retry;
3269         }
3270
3271         *address = temp;
3272         return 0;
3273 }
3274
3275 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3276 {
3277         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3278
3279         if (s->initialized)
3280                 return;
3281         s = &idev->cnf.stable_secret;
3282         get_random_bytes(&s->secret, sizeof(s->secret));
3283         s->initialized = true;
3284 }
3285
3286 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3287 {
3288         struct in6_addr addr;
3289
3290         /* no link local addresses on L3 master devices */
3291         if (netif_is_l3_master(idev->dev))
3292                 return;
3293
3294         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3295
3296         switch (idev->cnf.addr_gen_mode) {
3297         case IN6_ADDR_GEN_MODE_RANDOM:
3298                 ipv6_gen_mode_random_init(idev);
3299                 /* fallthrough */
3300         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3301                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3302                         addrconf_add_linklocal(idev, &addr,
3303                                                IFA_F_STABLE_PRIVACY);
3304                 else if (prefix_route)
3305                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3306                                               0, 0, GFP_KERNEL);
3307                 break;
3308         case IN6_ADDR_GEN_MODE_EUI64:
3309                 /* addrconf_add_linklocal also adds a prefix_route and we
3310                  * only need to care about prefix routes if ipv6_generate_eui64
3311                  * couldn't generate one.
3312                  */
3313                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3314                         addrconf_add_linklocal(idev, &addr, 0);
3315                 else if (prefix_route)
3316                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3317                                               0, 0, GFP_KERNEL);
3318                 break;
3319         case IN6_ADDR_GEN_MODE_NONE:
3320         default:
3321                 /* will not add any link local address */
3322                 break;
3323         }
3324 }
3325
3326 static void addrconf_dev_config(struct net_device *dev)
3327 {
3328         struct inet6_dev *idev;
3329
3330         ASSERT_RTNL();
3331
3332         if ((dev->type != ARPHRD_ETHER) &&
3333             (dev->type != ARPHRD_FDDI) &&
3334             (dev->type != ARPHRD_ARCNET) &&
3335             (dev->type != ARPHRD_INFINIBAND) &&
3336             (dev->type != ARPHRD_IEEE1394) &&
3337             (dev->type != ARPHRD_TUNNEL6) &&
3338             (dev->type != ARPHRD_6LOWPAN) &&
3339             (dev->type != ARPHRD_IP6GRE) &&
3340             (dev->type != ARPHRD_IPGRE) &&
3341             (dev->type != ARPHRD_TUNNEL) &&
3342             (dev->type != ARPHRD_NONE) &&
3343             (dev->type != ARPHRD_RAWIP)) {
3344                 /* Alas, we support only Ethernet autoconfiguration. */
3345                 return;
3346         }
3347
3348         idev = addrconf_add_dev(dev);
3349         if (IS_ERR(idev))
3350                 return;
3351
3352         /* this device type has no EUI support */
3353         if (dev->type == ARPHRD_NONE &&
3354             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3355                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3356
3357         addrconf_addr_gen(idev, false);
3358 }
3359
3360 #if IS_ENABLED(CONFIG_IPV6_SIT)
3361 static void addrconf_sit_config(struct net_device *dev)
3362 {
3363         struct inet6_dev *idev;
3364
3365         ASSERT_RTNL();
3366
3367         /*
3368          * Configure the tunnel with one of our IPv4
3369          * addresses... we should configure all of
3370          * our v4 addrs in the tunnel
3371          */
3372
3373         idev = ipv6_find_idev(dev);
3374         if (!idev) {
3375                 pr_debug("%s: add_dev failed\n", __func__);
3376                 return;
3377         }
3378
3379         if (dev->priv_flags & IFF_ISATAP) {
3380                 addrconf_addr_gen(idev, false);
3381                 return;
3382         }
3383
3384         sit_add_v4_addrs(idev);
3385
3386         if (dev->flags&IFF_POINTOPOINT)
3387                 addrconf_add_mroute(dev);
3388 }
3389 #endif
3390
3391 #if IS_ENABLED(CONFIG_NET_IPGRE)
3392 static void addrconf_gre_config(struct net_device *dev)
3393 {
3394         struct inet6_dev *idev;
3395
3396         ASSERT_RTNL();
3397
3398         idev = ipv6_find_idev(dev);
3399         if (!idev) {
3400                 pr_debug("%s: add_dev failed\n", __func__);
3401                 return;
3402         }
3403
3404         addrconf_addr_gen(idev, true);
3405         if (dev->flags & IFF_POINTOPOINT)
3406                 addrconf_add_mroute(dev);
3407 }
3408 #endif
3409
3410 static int fixup_permanent_addr(struct net *net,
3411                                 struct inet6_dev *idev,
3412                                 struct inet6_ifaddr *ifp)
3413 {
3414         /* !fib6_node means the host route was removed from the
3415          * FIB, for example, if 'lo' device is taken down. In that
3416          * case regenerate the host route.
3417          */
3418         if (!ifp->rt || !ifp->rt->fib6_node) {
3419                 struct fib6_info *f6i, *prev;
3420
3421                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3422                                          GFP_ATOMIC);
3423                 if (IS_ERR(f6i))
3424                         return PTR_ERR(f6i);
3425
3426                 /* ifp->rt can be accessed outside of rtnl */
3427                 spin_lock(&ifp->lock);
3428                 prev = ifp->rt;
3429                 ifp->rt = f6i;
3430                 spin_unlock(&ifp->lock);
3431
3432                 fib6_info_release(prev);
3433         }
3434
3435         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3436                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3437                                       ifp->rt_priority, idev->dev, 0, 0,
3438                                       GFP_ATOMIC);
3439         }
3440
3441         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3442                 addrconf_dad_start(ifp);
3443
3444         return 0;
3445 }
3446
3447 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3448 {
3449         struct inet6_ifaddr *ifp, *tmp;
3450         struct inet6_dev *idev;
3451
3452         idev = __in6_dev_get(dev);
3453         if (!idev)
3454                 return;
3455
3456         write_lock_bh(&idev->lock);
3457
3458         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3459                 if ((ifp->flags & IFA_F_PERMANENT) &&
3460                     fixup_permanent_addr(net, idev, ifp) < 0) {
3461                         write_unlock_bh(&idev->lock);
3462                         in6_ifa_hold(ifp);
3463                         ipv6_del_addr(ifp);
3464                         write_lock_bh(&idev->lock);
3465
3466                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3467                                              idev->dev->name, &ifp->addr);
3468                 }
3469         }
3470
3471         write_unlock_bh(&idev->lock);
3472 }
3473
3474 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3475                            void *ptr)
3476 {
3477         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3478         struct netdev_notifier_change_info *change_info;
3479         struct netdev_notifier_changeupper_info *info;
3480         struct inet6_dev *idev = __in6_dev_get(dev);
3481         struct net *net = dev_net(dev);
3482         int run_pending = 0;
3483         int err;
3484
3485         switch (event) {
3486         case NETDEV_REGISTER:
3487                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3488                         idev = ipv6_add_dev(dev);
3489                         if (IS_ERR(idev))
3490                                 return notifier_from_errno(PTR_ERR(idev));
3491                 }
3492                 break;
3493
3494         case NETDEV_CHANGEMTU:
3495                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3496                 if (dev->mtu < IPV6_MIN_MTU) {
3497                         addrconf_ifdown(dev, dev != net->loopback_dev);
3498                         break;
3499                 }
3500
3501                 if (idev) {
3502                         rt6_mtu_change(dev, dev->mtu);
3503                         idev->cnf.mtu6 = dev->mtu;
3504                         break;
3505                 }
3506
3507                 /* allocate new idev */
3508                 idev = ipv6_add_dev(dev);
3509                 if (IS_ERR(idev))
3510                         break;
3511
3512                 /* device is still not ready */
3513                 if (!(idev->if_flags & IF_READY))
3514                         break;
3515
3516                 run_pending = 1;
3517
3518                 /* fall through */
3519
3520         case NETDEV_UP:
3521         case NETDEV_CHANGE:
3522                 if (dev->flags & IFF_SLAVE)
3523                         break;
3524
3525                 if (idev && idev->cnf.disable_ipv6)
3526                         break;
3527
3528                 if (event == NETDEV_UP) {
3529                         /* restore routes for permanent addresses */
3530                         addrconf_permanent_addr(net, dev);
3531
3532                         if (!addrconf_link_ready(dev)) {
3533                                 /* device is not ready yet. */
3534                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3535                                          dev->name);
3536                                 break;
3537                         }
3538
3539                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3540                                 idev = ipv6_add_dev(dev);
3541
3542                         if (!IS_ERR_OR_NULL(idev)) {
3543                                 idev->if_flags |= IF_READY;
3544                                 run_pending = 1;
3545                         }
3546                 } else if (event == NETDEV_CHANGE) {
3547                         if (!addrconf_link_ready(dev)) {
3548                                 /* device is still not ready. */
3549                                 rt6_sync_down_dev(dev, event);
3550                                 break;
3551                         }
3552
3553                         if (!IS_ERR_OR_NULL(idev)) {
3554                                 if (idev->if_flags & IF_READY) {
3555                                         /* device is already configured -
3556                                          * but resend MLD reports, we might
3557                                          * have roamed and need to update
3558                                          * multicast snooping switches
3559                                          */
3560                                         ipv6_mc_up(idev);
3561                                         change_info = ptr;
3562                                         if (change_info->flags_changed & IFF_NOARP)
3563                                                 addrconf_dad_run(idev, true);
3564                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3565                                         break;
3566                                 }
3567                                 idev->if_flags |= IF_READY;
3568                         }
3569
3570                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3571                                 dev->name);
3572
3573                         run_pending = 1;
3574                 }
3575
3576                 switch (dev->type) {
3577 #if IS_ENABLED(CONFIG_IPV6_SIT)
3578                 case ARPHRD_SIT:
3579                         addrconf_sit_config(dev);
3580                         break;
3581 #endif
3582 #if IS_ENABLED(CONFIG_NET_IPGRE)
3583                 case ARPHRD_IPGRE:
3584                         addrconf_gre_config(dev);
3585                         break;
3586 #endif
3587                 case ARPHRD_LOOPBACK:
3588                         init_loopback(dev);
3589                         break;
3590
3591                 default:
3592                         addrconf_dev_config(dev);
3593                         break;
3594                 }
3595
3596                 if (!IS_ERR_OR_NULL(idev)) {
3597                         if (run_pending)
3598                                 addrconf_dad_run(idev, false);
3599
3600                         /* Device has an address by now */
3601                         rt6_sync_up(dev, RTNH_F_DEAD);
3602
3603                         /*
3604                          * If the MTU changed during the interface down,
3605                          * when the interface up, the changed MTU must be
3606                          * reflected in the idev as well as routers.
3607                          */
3608                         if (idev->cnf.mtu6 != dev->mtu &&
3609                             dev->mtu >= IPV6_MIN_MTU) {
3610                                 rt6_mtu_change(dev, dev->mtu);
3611                                 idev->cnf.mtu6 = dev->mtu;
3612                         }
3613                         idev->tstamp = jiffies;
3614                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3615
3616                         /*
3617                          * If the changed mtu during down is lower than
3618                          * IPV6_MIN_MTU stop IPv6 on this interface.
3619                          */
3620                         if (dev->mtu < IPV6_MIN_MTU)
3621                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3622                 }
3623                 break;
3624
3625         case NETDEV_DOWN:
3626         case NETDEV_UNREGISTER:
3627                 /*
3628                  *      Remove all addresses from this interface.
3629                  */
3630                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3631                 break;
3632
3633         case NETDEV_CHANGENAME:
3634                 if (idev) {
3635                         snmp6_unregister_dev(idev);
3636                         addrconf_sysctl_unregister(idev);
3637                         err = addrconf_sysctl_register(idev);
3638                         if (err)
3639                                 return notifier_from_errno(err);
3640                         err = snmp6_register_dev(idev);
3641                         if (err) {
3642                                 addrconf_sysctl_unregister(idev);
3643                                 return notifier_from_errno(err);
3644                         }
3645                 }
3646                 break;
3647
3648         case NETDEV_PRE_TYPE_CHANGE:
3649         case NETDEV_POST_TYPE_CHANGE:
3650                 if (idev)
3651                         addrconf_type_change(dev, event);
3652                 break;
3653
3654         case NETDEV_CHANGEUPPER:
3655                 info = ptr;
3656
3657                 /* flush all routes if dev is linked to or unlinked from
3658                  * an L3 master device (e.g., VRF)
3659                  */
3660                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3661                         addrconf_ifdown(dev, 0);
3662         }
3663
3664         return NOTIFY_OK;
3665 }
3666
3667 /*
3668  *      addrconf module should be notified of a device going up
3669  */
3670 static struct notifier_block ipv6_dev_notf = {
3671         .notifier_call = addrconf_notify,
3672         .priority = ADDRCONF_NOTIFY_PRIORITY,
3673 };
3674
3675 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3676 {
3677         struct inet6_dev *idev;
3678         ASSERT_RTNL();
3679
3680         idev = __in6_dev_get(dev);
3681
3682         if (event == NETDEV_POST_TYPE_CHANGE)
3683                 ipv6_mc_remap(idev);
3684         else if (event == NETDEV_PRE_TYPE_CHANGE)
3685                 ipv6_mc_unmap(idev);
3686 }
3687
3688 static bool addr_is_local(const struct in6_addr *addr)
3689 {
3690         return ipv6_addr_type(addr) &
3691                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3692 }
3693
3694 static int addrconf_ifdown(struct net_device *dev, int how)
3695 {
3696         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3697         struct net *net = dev_net(dev);
3698         struct inet6_dev *idev;
3699         struct inet6_ifaddr *ifa, *tmp;
3700         bool keep_addr = false;
3701         int state, i;
3702
3703         ASSERT_RTNL();
3704
3705         rt6_disable_ip(dev, event);
3706
3707         idev = __in6_dev_get(dev);
3708         if (!idev)
3709                 return -ENODEV;
3710
3711         /*
3712          * Step 1: remove reference to ipv6 device from parent device.
3713          *         Do not dev_put!
3714          */
3715         if (how) {
3716                 idev->dead = 1;
3717
3718                 /* protected by rtnl_lock */
3719                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3720
3721                 /* Step 1.5: remove snmp6 entry */
3722                 snmp6_unregister_dev(idev);
3723
3724         }
3725
3726         /* combine the user config with event to determine if permanent
3727          * addresses are to be removed from address hash table
3728          */
3729         if (!how && !idev->cnf.disable_ipv6) {
3730                 /* aggregate the system setting and interface setting */
3731                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3732
3733                 if (!_keep_addr)
3734                         _keep_addr = idev->cnf.keep_addr_on_down;
3735
3736                 keep_addr = (_keep_addr > 0);
3737         }
3738
3739         /* Step 2: clear hash table */
3740         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3741                 struct hlist_head *h = &inet6_addr_lst[i];
3742
3743                 spin_lock_bh(&addrconf_hash_lock);
3744 restart:
3745                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3746                         if (ifa->idev == idev) {
3747                                 addrconf_del_dad_work(ifa);
3748                                 /* combined flag + permanent flag decide if
3749                                  * address is retained on a down event
3750                                  */
3751                                 if (!keep_addr ||
3752                                     !(ifa->flags & IFA_F_PERMANENT) ||
3753                                     addr_is_local(&ifa->addr)) {
3754                                         hlist_del_init_rcu(&ifa->addr_lst);
3755                                         goto restart;
3756                                 }
3757                         }
3758                 }
3759                 spin_unlock_bh(&addrconf_hash_lock);
3760         }
3761
3762         write_lock_bh(&idev->lock);
3763
3764         addrconf_del_rs_timer(idev);
3765
3766         /* Step 2: clear flags for stateless addrconf */
3767         if (!how)
3768                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3769
3770         /* Step 3: clear tempaddr list */
3771         while (!list_empty(&idev->tempaddr_list)) {
3772                 ifa = list_first_entry(&idev->tempaddr_list,
3773                                        struct inet6_ifaddr, tmp_list);
3774                 list_del(&ifa->tmp_list);
3775                 write_unlock_bh(&idev->lock);
3776                 spin_lock_bh(&ifa->lock);
3777
3778                 if (ifa->ifpub) {
3779                         in6_ifa_put(ifa->ifpub);
3780                         ifa->ifpub = NULL;
3781                 }
3782                 spin_unlock_bh(&ifa->lock);
3783                 in6_ifa_put(ifa);
3784                 write_lock_bh(&idev->lock);
3785         }
3786
3787         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3788                 struct fib6_info *rt = NULL;
3789                 bool keep;
3790
3791                 addrconf_del_dad_work(ifa);
3792
3793                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3794                         !addr_is_local(&ifa->addr);
3795
3796                 write_unlock_bh(&idev->lock);
3797                 spin_lock_bh(&ifa->lock);
3798
3799                 if (keep) {
3800                         /* set state to skip the notifier below */
3801                         state = INET6_IFADDR_STATE_DEAD;
3802                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3803                         if (!(ifa->flags & IFA_F_NODAD))
3804                                 ifa->flags |= IFA_F_TENTATIVE;
3805
3806                         rt = ifa->rt;
3807                         ifa->rt = NULL;
3808                 } else {
3809                         state = ifa->state;
3810                         ifa->state = INET6_IFADDR_STATE_DEAD;
3811                 }
3812
3813                 spin_unlock_bh(&ifa->lock);
3814
3815                 if (rt)
3816                         ip6_del_rt(net, rt);
3817
3818                 if (state != INET6_IFADDR_STATE_DEAD) {
3819                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3820                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3821                 } else {
3822                         if (idev->cnf.forwarding)
3823                                 addrconf_leave_anycast(ifa);
3824                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3825                 }
3826
3827                 write_lock_bh(&idev->lock);
3828                 if (!keep) {
3829                         list_del_rcu(&ifa->if_list);
3830                         in6_ifa_put(ifa);
3831                 }
3832         }
3833
3834         write_unlock_bh(&idev->lock);
3835
3836         /* Step 5: Discard anycast and multicast list */
3837         if (how) {
3838                 ipv6_ac_destroy_dev(idev);
3839                 ipv6_mc_destroy_dev(idev);
3840         } else {
3841                 ipv6_mc_down(idev);
3842         }
3843
3844         idev->tstamp = jiffies;
3845
3846         /* Last: Shot the device (if unregistered) */
3847         if (how) {
3848                 addrconf_sysctl_unregister(idev);
3849                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3850                 neigh_ifdown(&nd_tbl, dev);
3851                 in6_dev_put(idev);
3852         }
3853         return 0;
3854 }
3855
3856 static void addrconf_rs_timer(struct timer_list *t)
3857 {
3858         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3859         struct net_device *dev = idev->dev;
3860         struct in6_addr lladdr;
3861
3862         write_lock(&idev->lock);
3863         if (idev->dead || !(idev->if_flags & IF_READY))
3864                 goto out;
3865
3866         if (!ipv6_accept_ra(idev))
3867                 goto out;
3868
3869         /* Announcement received after solicitation was sent */
3870         if (idev->if_flags & IF_RA_RCVD)
3871                 goto out;
3872
3873         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3874                 write_unlock(&idev->lock);
3875                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3876                         ndisc_send_rs(dev, &lladdr,
3877                                       &in6addr_linklocal_allrouters);
3878                 else
3879                         goto put;
3880
3881                 write_lock(&idev->lock);
3882                 idev->rs_interval = rfc3315_s14_backoff_update(
3883                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3884                 /* The wait after the last probe can be shorter */
3885                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3886                                              idev->cnf.rtr_solicits) ?
3887                                       idev->cnf.rtr_solicit_delay :
3888                                       idev->rs_interval);
3889         } else {
3890                 /*
3891                  * Note: we do not support deprecated "all on-link"
3892                  * assumption any longer.
3893                  */
3894                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3895         }
3896
3897 out:
3898         write_unlock(&idev->lock);
3899 put:
3900         in6_dev_put(idev);
3901 }
3902
3903 /*
3904  *      Duplicate Address Detection
3905  */
3906 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3907 {
3908         unsigned long rand_num;
3909         struct inet6_dev *idev = ifp->idev;
3910         u64 nonce;
3911
3912         if (ifp->flags & IFA_F_OPTIMISTIC)
3913                 rand_num = 0;
3914         else
3915                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3916
3917         nonce = 0;
3918         if (idev->cnf.enhanced_dad ||
3919             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3920                 do
3921                         get_random_bytes(&nonce, 6);
3922                 while (nonce == 0);
3923         }
3924         ifp->dad_nonce = nonce;
3925         ifp->dad_probes = idev->cnf.dad_transmits;
3926         addrconf_mod_dad_work(ifp, rand_num);
3927 }
3928
3929 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3930 {
3931         struct inet6_dev *idev = ifp->idev;
3932         struct net_device *dev = idev->dev;
3933         bool bump_id, notify = false;
3934         struct net *net;
3935
3936         addrconf_join_solict(dev, &ifp->addr);
3937
3938         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3939
3940         read_lock_bh(&idev->lock);
3941         spin_lock(&ifp->lock);
3942         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3943                 goto out;
3944
3945         net = dev_net(dev);
3946         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3947             (net->ipv6.devconf_all->accept_dad < 1 &&
3948              idev->cnf.accept_dad < 1) ||
3949             !(ifp->flags&IFA_F_TENTATIVE) ||
3950             ifp->flags & IFA_F_NODAD) {
3951                 bool send_na = false;
3952
3953                 if (ifp->flags & IFA_F_TENTATIVE &&
3954                     !(ifp->flags & IFA_F_OPTIMISTIC))
3955                         send_na = true;
3956                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3957                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3958                 spin_unlock(&ifp->lock);
3959                 read_unlock_bh(&idev->lock);
3960
3961                 addrconf_dad_completed(ifp, bump_id, send_na);
3962                 return;
3963         }
3964
3965         if (!(idev->if_flags & IF_READY)) {
3966                 spin_unlock(&ifp->lock);
3967                 read_unlock_bh(&idev->lock);
3968                 /*
3969                  * If the device is not ready:
3970                  * - keep it tentative if it is a permanent address.
3971                  * - otherwise, kill it.
3972                  */
3973                 in6_ifa_hold(ifp);
3974                 addrconf_dad_stop(ifp, 0);
3975                 return;
3976         }
3977
3978         /*
3979          * Optimistic nodes can start receiving
3980          * Frames right away
3981          */
3982         if (ifp->flags & IFA_F_OPTIMISTIC) {
3983                 ip6_ins_rt(net, ifp->rt);
3984                 if (ipv6_use_optimistic_addr(net, idev)) {
3985                         /* Because optimistic nodes can use this address,
3986                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3987                          */
3988                         notify = true;
3989                 }
3990         }
3991
3992         addrconf_dad_kick(ifp);
3993 out:
3994         spin_unlock(&ifp->lock);
3995         read_unlock_bh(&idev->lock);
3996         if (notify)
3997                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3998 }
3999
4000 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4001 {
4002         bool begin_dad = false;
4003
4004         spin_lock_bh(&ifp->lock);
4005         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4006                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4007                 begin_dad = true;
4008         }
4009         spin_unlock_bh(&ifp->lock);
4010
4011         if (begin_dad)
4012                 addrconf_mod_dad_work(ifp, 0);
4013 }
4014
4015 static void addrconf_dad_work(struct work_struct *w)
4016 {
4017         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4018                                                 struct inet6_ifaddr,
4019                                                 dad_work);
4020         struct inet6_dev *idev = ifp->idev;
4021         bool bump_id, disable_ipv6 = false;
4022         struct in6_addr mcaddr;
4023
4024         enum {
4025                 DAD_PROCESS,
4026                 DAD_BEGIN,
4027                 DAD_ABORT,
4028         } action = DAD_PROCESS;
4029
4030         rtnl_lock();
4031
4032         spin_lock_bh(&ifp->lock);
4033         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4034                 action = DAD_BEGIN;
4035                 ifp->state = INET6_IFADDR_STATE_DAD;
4036         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4037                 action = DAD_ABORT;
4038                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4039
4040                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4041                      idev->cnf.accept_dad > 1) &&
4042                     !idev->cnf.disable_ipv6 &&
4043                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4044                         struct in6_addr addr;
4045
4046                         addr.s6_addr32[0] = htonl(0xfe800000);
4047                         addr.s6_addr32[1] = 0;
4048
4049                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4050                             ipv6_addr_equal(&ifp->addr, &addr)) {
4051                                 /* DAD failed for link-local based on MAC */
4052                                 idev->cnf.disable_ipv6 = 1;
4053
4054                                 pr_info("%s: IPv6 being disabled!\n",
4055                                         ifp->idev->dev->name);
4056                                 disable_ipv6 = true;
4057                         }
4058                 }
4059         }
4060         spin_unlock_bh(&ifp->lock);
4061
4062         if (action == DAD_BEGIN) {
4063                 addrconf_dad_begin(ifp);
4064                 goto out;
4065         } else if (action == DAD_ABORT) {
4066                 in6_ifa_hold(ifp);
4067                 addrconf_dad_stop(ifp, 1);
4068                 if (disable_ipv6)
4069                         addrconf_ifdown(idev->dev, 0);
4070                 goto out;
4071         }
4072
4073         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4074                 goto out;
4075
4076         write_lock_bh(&idev->lock);
4077         if (idev->dead || !(idev->if_flags & IF_READY)) {
4078                 write_unlock_bh(&idev->lock);
4079                 goto out;
4080         }
4081
4082         spin_lock(&ifp->lock);
4083         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4084                 spin_unlock(&ifp->lock);
4085                 write_unlock_bh(&idev->lock);
4086                 goto out;
4087         }
4088
4089         if (ifp->dad_probes == 0) {
4090                 bool send_na = false;
4091
4092                 /*
4093                  * DAD was successful
4094                  */
4095
4096                 if (ifp->flags & IFA_F_TENTATIVE &&
4097                     !(ifp->flags & IFA_F_OPTIMISTIC))
4098                         send_na = true;
4099                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4100                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4101                 spin_unlock(&ifp->lock);
4102                 write_unlock_bh(&idev->lock);
4103
4104                 addrconf_dad_completed(ifp, bump_id, send_na);
4105
4106                 goto out;
4107         }
4108
4109         ifp->dad_probes--;
4110         addrconf_mod_dad_work(ifp,
4111                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4112         spin_unlock(&ifp->lock);
4113         write_unlock_bh(&idev->lock);
4114
4115         /* send a neighbour solicitation for our addr */
4116         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4117         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4118                       ifp->dad_nonce);
4119 out:
4120         in6_ifa_put(ifp);
4121         rtnl_unlock();
4122 }
4123
4124 /* ifp->idev must be at least read locked */
4125 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4126 {
4127         struct inet6_ifaddr *ifpiter;
4128         struct inet6_dev *idev = ifp->idev;
4129
4130         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4131                 if (ifpiter->scope > IFA_LINK)
4132                         break;
4133                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4134                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4135                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4136                     IFA_F_PERMANENT)
4137                         return false;
4138         }
4139         return true;
4140 }
4141
4142 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4143                                    bool send_na)
4144 {
4145         struct net_device *dev = ifp->idev->dev;
4146         struct in6_addr lladdr;
4147         bool send_rs, send_mld;
4148
4149         addrconf_del_dad_work(ifp);
4150
4151         /*
4152          *      Configure the address for reception. Now it is valid.
4153          */
4154
4155         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4156
4157         /* If added prefix is link local and we are prepared to process
4158            router advertisements, start sending router solicitations.
4159          */
4160
4161         read_lock_bh(&ifp->idev->lock);
4162         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4163         send_rs = send_mld &&
4164                   ipv6_accept_ra(ifp->idev) &&
4165                   ifp->idev->cnf.rtr_solicits != 0 &&
4166                   (dev->flags&IFF_LOOPBACK) == 0;
4167         read_unlock_bh(&ifp->idev->lock);
4168
4169         /* While dad is in progress mld report's source address is in6_addrany.
4170          * Resend with proper ll now.
4171          */
4172         if (send_mld)
4173                 ipv6_mc_dad_complete(ifp->idev);
4174
4175         /* send unsolicited NA if enabled */
4176         if (send_na &&
4177             (ifp->idev->cnf.ndisc_notify ||
4178              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4179                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4180                               /*router=*/ !!ifp->idev->cnf.forwarding,
4181                               /*solicited=*/ false, /*override=*/ true,
4182                               /*inc_opt=*/ true);
4183         }
4184
4185         if (send_rs) {
4186                 /*
4187                  *      If a host as already performed a random delay
4188                  *      [...] as part of DAD [...] there is no need
4189                  *      to delay again before sending the first RS
4190                  */
4191                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4192                         return;
4193                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4194
4195                 write_lock_bh(&ifp->idev->lock);
4196                 spin_lock(&ifp->lock);
4197                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4198                         ifp->idev->cnf.rtr_solicit_interval);
4199                 ifp->idev->rs_probes = 1;
4200                 ifp->idev->if_flags |= IF_RS_SENT;
4201                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4202                 spin_unlock(&ifp->lock);
4203                 write_unlock_bh(&ifp->idev->lock);
4204         }
4205
4206         if (bump_id)
4207                 rt_genid_bump_ipv6(dev_net(dev));
4208
4209         /* Make sure that a new temporary address will be created
4210          * before this temporary address becomes deprecated.
4211          */
4212         if (ifp->flags & IFA_F_TEMPORARY)
4213                 addrconf_verify_rtnl();
4214 }
4215
4216 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4217 {
4218         struct inet6_ifaddr *ifp;
4219
4220         read_lock_bh(&idev->lock);
4221         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4222                 spin_lock(&ifp->lock);
4223                 if ((ifp->flags & IFA_F_TENTATIVE &&
4224                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4225                         if (restart)
4226                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4227                         addrconf_dad_kick(ifp);
4228                 }
4229                 spin_unlock(&ifp->lock);
4230         }
4231         read_unlock_bh(&idev->lock);
4232 }
4233
4234 #ifdef CONFIG_PROC_FS
4235 struct if6_iter_state {
4236         struct seq_net_private p;
4237         int bucket;
4238         int offset;
4239 };
4240
4241 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4242 {
4243         struct if6_iter_state *state = seq->private;
4244         struct net *net = seq_file_net(seq);
4245         struct inet6_ifaddr *ifa = NULL;
4246         int p = 0;
4247
4248         /* initial bucket if pos is 0 */
4249         if (pos == 0) {
4250                 state->bucket = 0;
4251                 state->offset = 0;
4252         }
4253
4254         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4255                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4256                                          addr_lst) {
4257                         if (!net_eq(dev_net(ifa->idev->dev), net))
4258                                 continue;
4259                         /* sync with offset */
4260                         if (p < state->offset) {
4261                                 p++;
4262                                 continue;
4263                         }
4264                         return ifa;
4265                 }
4266
4267                 /* prepare for next bucket */
4268                 state->offset = 0;
4269                 p = 0;
4270         }
4271         return NULL;
4272 }
4273
4274 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4275                                          struct inet6_ifaddr *ifa)
4276 {
4277         struct if6_iter_state *state = seq->private;
4278         struct net *net = seq_file_net(seq);
4279
4280         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4281                 if (!net_eq(dev_net(ifa->idev->dev), net))
4282                         continue;
4283                 state->offset++;
4284                 return ifa;
4285         }
4286
4287         state->offset = 0;
4288         while (++state->bucket < IN6_ADDR_HSIZE) {
4289                 hlist_for_each_entry_rcu(ifa,
4290                                      &inet6_addr_lst[state->bucket], addr_lst) {
4291                         if (!net_eq(dev_net(ifa->idev->dev), net))
4292                                 continue;
4293                         return ifa;
4294                 }
4295         }
4296
4297         return NULL;
4298 }
4299
4300 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4301         __acquires(rcu)
4302 {
4303         rcu_read_lock();
4304         return if6_get_first(seq, *pos);
4305 }
4306
4307 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4308 {
4309         struct inet6_ifaddr *ifa;
4310
4311         ifa = if6_get_next(seq, v);
4312         ++*pos;
4313         return ifa;
4314 }
4315
4316 static void if6_seq_stop(struct seq_file *seq, void *v)
4317         __releases(rcu)
4318 {
4319         rcu_read_unlock();
4320 }
4321
4322 static int if6_seq_show(struct seq_file *seq, void *v)
4323 {
4324         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4325         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4326                    &ifp->addr,
4327                    ifp->idev->dev->ifindex,
4328                    ifp->prefix_len,
4329                    ifp->scope,
4330                    (u8) ifp->flags,
4331                    ifp->idev->dev->name);
4332         return 0;
4333 }
4334
4335 static const struct seq_operations if6_seq_ops = {
4336         .start  = if6_seq_start,
4337         .next   = if6_seq_next,
4338         .show   = if6_seq_show,
4339         .stop   = if6_seq_stop,
4340 };
4341
4342 static int __net_init if6_proc_net_init(struct net *net)
4343 {
4344         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4345                         sizeof(struct if6_iter_state)))
4346                 return -ENOMEM;
4347         return 0;
4348 }
4349
4350 static void __net_exit if6_proc_net_exit(struct net *net)
4351 {
4352         remove_proc_entry("if_inet6", net->proc_net);
4353 }
4354
4355 static struct pernet_operations if6_proc_net_ops = {
4356         .init = if6_proc_net_init,
4357         .exit = if6_proc_net_exit,
4358 };
4359
4360 int __init if6_proc_init(void)
4361 {
4362         return register_pernet_subsys(&if6_proc_net_ops);
4363 }
4364
4365 void if6_proc_exit(void)
4366 {
4367         unregister_pernet_subsys(&if6_proc_net_ops);
4368 }
4369 #endif  /* CONFIG_PROC_FS */
4370
4371 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4372 /* Check if address is a home address configured on any interface. */
4373 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4374 {
4375         unsigned int hash = inet6_addr_hash(net, addr);
4376         struct inet6_ifaddr *ifp = NULL;
4377         int ret = 0;
4378
4379         rcu_read_lock();
4380         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4381                 if (!net_eq(dev_net(ifp->idev->dev), net))
4382                         continue;
4383                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4384                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4385                         ret = 1;
4386                         break;
4387                 }
4388         }
4389         rcu_read_unlock();
4390         return ret;
4391 }
4392 #endif
4393
4394 /*
4395  *      Periodic address status verification
4396  */
4397
4398 static void addrconf_verify_rtnl(void)
4399 {
4400         unsigned long now, next, next_sec, next_sched;
4401         struct inet6_ifaddr *ifp;
4402         int i;
4403
4404         ASSERT_RTNL();
4405
4406         rcu_read_lock_bh();
4407         now = jiffies;
4408         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4409
4410         cancel_delayed_work(&addr_chk_work);
4411
4412         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4413 restart:
4414                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4415                         unsigned long age;
4416
4417                         /* When setting preferred_lft to a value not zero or
4418                          * infinity, while valid_lft is infinity
4419                          * IFA_F_PERMANENT has a non-infinity life time.
4420                          */
4421                         if ((ifp->flags & IFA_F_PERMANENT) &&
4422                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4423                                 continue;
4424
4425                         spin_lock(&ifp->lock);
4426                         /* We try to batch several events at once. */
4427                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4428
4429                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4430                             age >= ifp->valid_lft) {
4431                                 spin_unlock(&ifp->lock);
4432                                 in6_ifa_hold(ifp);
4433                                 ipv6_del_addr(ifp);
4434                                 goto restart;
4435                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4436                                 spin_unlock(&ifp->lock);
4437                                 continue;
4438                         } else if (age >= ifp->prefered_lft) {
4439                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4440                                 int deprecate = 0;
4441
4442                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4443                                         deprecate = 1;
4444                                         ifp->flags |= IFA_F_DEPRECATED;
4445                                 }
4446
4447                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4448                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4449                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4450
4451                                 spin_unlock(&ifp->lock);
4452
4453                                 if (deprecate) {
4454                                         in6_ifa_hold(ifp);
4455
4456                                         ipv6_ifa_notify(0, ifp);
4457                                         in6_ifa_put(ifp);
4458                                         goto restart;
4459                                 }
4460                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4461                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4462                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4463                                         ifp->idev->cnf.dad_transmits *
4464                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4465
4466                                 if (age >= ifp->prefered_lft - regen_advance) {
4467                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4468                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4469                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4470                                         if (!ifp->regen_count && ifpub) {
4471                                                 ifp->regen_count++;
4472                                                 in6_ifa_hold(ifp);
4473                                                 in6_ifa_hold(ifpub);
4474                                                 spin_unlock(&ifp->lock);
4475
4476                                                 spin_lock(&ifpub->lock);
4477                                                 ifpub->regen_count = 0;
4478                                                 spin_unlock(&ifpub->lock);
4479                                                 rcu_read_unlock_bh();
4480                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4481                                                 in6_ifa_put(ifpub);
4482                                                 in6_ifa_put(ifp);
4483                                                 rcu_read_lock_bh();
4484                                                 goto restart;
4485                                         }
4486                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4487                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4488                                 spin_unlock(&ifp->lock);
4489                         } else {
4490                                 /* ifp->prefered_lft <= ifp->valid_lft */
4491                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4492                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4493                                 spin_unlock(&ifp->lock);
4494                         }
4495                 }
4496         }
4497
4498         next_sec = round_jiffies_up(next);
4499         next_sched = next;
4500
4501         /* If rounded timeout is accurate enough, accept it. */
4502         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4503                 next_sched = next_sec;
4504
4505         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4506         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4507                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4508
4509         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4510                  now, next, next_sec, next_sched);
4511         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4512         rcu_read_unlock_bh();
4513 }
4514
4515 static void addrconf_verify_work(struct work_struct *w)
4516 {
4517         rtnl_lock();
4518         addrconf_verify_rtnl();
4519         rtnl_unlock();
4520 }
4521
4522 static void addrconf_verify(void)
4523 {
4524         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4525 }
4526
4527 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4528                                      struct in6_addr **peer_pfx)
4529 {
4530         struct in6_addr *pfx = NULL;
4531
4532         *peer_pfx = NULL;
4533
4534         if (addr)
4535                 pfx = nla_data(addr);
4536
4537         if (local) {
4538                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4539                         *peer_pfx = pfx;
4540                 pfx = nla_data(local);
4541         }
4542
4543         return pfx;
4544 }
4545
4546 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4547         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4548         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4549         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4550         [IFA_FLAGS]             = { .len = sizeof(u32) },
4551         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4552         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4553 };
4554
4555 static int
4556 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4557                   struct netlink_ext_ack *extack)
4558 {
4559         struct net *net = sock_net(skb->sk);
4560         struct ifaddrmsg *ifm;
4561         struct nlattr *tb[IFA_MAX+1];
4562         struct in6_addr *pfx, *peer_pfx;
4563         u32 ifa_flags;
4564         int err;
4565
4566         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4567                           extack);
4568         if (err < 0)
4569                 return err;
4570
4571         ifm = nlmsg_data(nlh);
4572         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4573         if (!pfx)
4574                 return -EINVAL;
4575
4576         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4577
4578         /* We ignore other flags so far. */
4579         ifa_flags &= IFA_F_MANAGETEMPADDR;
4580
4581         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4582                               ifm->ifa_prefixlen);
4583 }
4584
4585 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4586                                unsigned long expires, u32 flags)
4587 {
4588         struct fib6_info *f6i;
4589         u32 prio;
4590
4591         f6i = addrconf_get_prefix_route(&ifp->addr,
4592                                         ifp->prefix_len,
4593                                         ifp->idev->dev,
4594                                         0, RTF_GATEWAY | RTF_DEFAULT);
4595         if (!f6i)
4596                 return -ENOENT;
4597
4598         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4599         if (f6i->fib6_metric != prio) {
4600                 /* delete old one */
4601                 ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4602
4603                 /* add new one */
4604                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4605                                       ifp->rt_priority, ifp->idev->dev,
4606                                       expires, flags, GFP_KERNEL);
4607         } else {
4608                 if (!expires)
4609                         fib6_clean_expires(f6i);
4610                 else
4611                         fib6_set_expires(f6i, expires);
4612
4613                 fib6_info_release(f6i);
4614         }
4615
4616         return 0;
4617 }
4618
4619 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4620 {
4621         u32 flags;
4622         clock_t expires;
4623         unsigned long timeout;
4624         bool was_managetempaddr;
4625         bool had_prefixroute;
4626
4627         ASSERT_RTNL();
4628
4629         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4630                 return -EINVAL;
4631
4632         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4633             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4634                 return -EINVAL;
4635
4636         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4637                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4638
4639         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4640         if (addrconf_finite_timeout(timeout)) {
4641                 expires = jiffies_to_clock_t(timeout * HZ);
4642                 cfg->valid_lft = timeout;
4643                 flags = RTF_EXPIRES;
4644         } else {
4645                 expires = 0;
4646                 flags = 0;
4647                 cfg->ifa_flags |= IFA_F_PERMANENT;
4648         }
4649
4650         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4651         if (addrconf_finite_timeout(timeout)) {
4652                 if (timeout == 0)
4653                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4654                 cfg->preferred_lft = timeout;
4655         }
4656
4657         spin_lock_bh(&ifp->lock);
4658         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4659         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4660                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4661         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4662                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4663                         IFA_F_NOPREFIXROUTE);
4664         ifp->flags |= cfg->ifa_flags;
4665         ifp->tstamp = jiffies;
4666         ifp->valid_lft = cfg->valid_lft;
4667         ifp->prefered_lft = cfg->preferred_lft;
4668
4669         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4670                 ifp->rt_priority = cfg->rt_priority;
4671
4672         spin_unlock_bh(&ifp->lock);
4673         if (!(ifp->flags&IFA_F_TENTATIVE))
4674                 ipv6_ifa_notify(0, ifp);
4675
4676         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4677                 int rc = -ENOENT;
4678
4679                 if (had_prefixroute)
4680                         rc = modify_prefix_route(ifp, expires, flags);
4681
4682                 /* prefix route could have been deleted; if so restore it */
4683                 if (rc == -ENOENT) {
4684                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4685                                               ifp->rt_priority, ifp->idev->dev,
4686                                               expires, flags, GFP_KERNEL);
4687                 }
4688         } else if (had_prefixroute) {
4689                 enum cleanup_prefix_rt_t action;
4690                 unsigned long rt_expires;
4691
4692                 write_lock_bh(&ifp->idev->lock);
4693                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4694                 write_unlock_bh(&ifp->idev->lock);
4695
4696                 if (action != CLEANUP_PREFIX_RT_NOP) {
4697                         cleanup_prefix_route(ifp, rt_expires,
4698                                 action == CLEANUP_PREFIX_RT_DEL);
4699                 }
4700         }
4701
4702         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4703                 if (was_managetempaddr &&
4704                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4705                         cfg->valid_lft = 0;
4706                         cfg->preferred_lft = 0;
4707                 }
4708                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4709                                  cfg->preferred_lft, !was_managetempaddr,
4710                                  jiffies);
4711         }
4712
4713         addrconf_verify_rtnl();
4714
4715         return 0;
4716 }
4717
4718 static int
4719 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4720                   struct netlink_ext_ack *extack)
4721 {
4722         struct net *net = sock_net(skb->sk);
4723         struct ifaddrmsg *ifm;
4724         struct nlattr *tb[IFA_MAX+1];
4725         struct in6_addr *peer_pfx;
4726         struct inet6_ifaddr *ifa;
4727         struct net_device *dev;
4728         struct inet6_dev *idev;
4729         struct ifa6_config cfg;
4730         int err;
4731
4732         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4733                           extack);
4734         if (err < 0)
4735                 return err;
4736
4737         memset(&cfg, 0, sizeof(cfg));
4738
4739         ifm = nlmsg_data(nlh);
4740         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4741         if (!cfg.pfx)
4742                 return -EINVAL;
4743
4744         cfg.peer_pfx = peer_pfx;
4745         cfg.plen = ifm->ifa_prefixlen;
4746         if (tb[IFA_RT_PRIORITY])
4747                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4748
4749         cfg.valid_lft = INFINITY_LIFE_TIME;
4750         cfg.preferred_lft = INFINITY_LIFE_TIME;
4751
4752         if (tb[IFA_CACHEINFO]) {
4753                 struct ifa_cacheinfo *ci;
4754
4755                 ci = nla_data(tb[IFA_CACHEINFO]);
4756                 cfg.valid_lft = ci->ifa_valid;
4757                 cfg.preferred_lft = ci->ifa_prefered;
4758         }
4759
4760         dev =  __dev_get_by_index(net, ifm->ifa_index);
4761         if (!dev)
4762                 return -ENODEV;
4763
4764         if (tb[IFA_FLAGS])
4765                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4766         else
4767                 cfg.ifa_flags = ifm->ifa_flags;
4768
4769         /* We ignore other flags so far. */
4770         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4771                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4772                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4773
4774         idev = ipv6_find_idev(dev);
4775         if (!idev)
4776                 return -ENOBUFS;
4777
4778         if (!ipv6_allow_optimistic_dad(net, idev))
4779                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4780
4781         if (cfg.ifa_flags & IFA_F_NODAD &&
4782             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4783                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4784                 return -EINVAL;
4785         }
4786
4787         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4788         if (!ifa) {
4789                 /*
4790                  * It would be best to check for !NLM_F_CREATE here but
4791                  * userspace already relies on not having to provide this.
4792                  */
4793                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4794         }
4795
4796         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4797             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4798                 err = -EEXIST;
4799         else
4800                 err = inet6_addr_modify(ifa, &cfg);
4801
4802         in6_ifa_put(ifa);
4803
4804         return err;
4805 }
4806
4807 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4808                           u8 scope, int ifindex)
4809 {
4810         struct ifaddrmsg *ifm;
4811
4812         ifm = nlmsg_data(nlh);
4813         ifm->ifa_family = AF_INET6;
4814         ifm->ifa_prefixlen = prefixlen;
4815         ifm->ifa_flags = flags;
4816         ifm->ifa_scope = scope;
4817         ifm->ifa_index = ifindex;
4818 }
4819
4820 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4821                          unsigned long tstamp, u32 preferred, u32 valid)
4822 {
4823         struct ifa_cacheinfo ci;
4824
4825         ci.cstamp = cstamp_delta(cstamp);
4826         ci.tstamp = cstamp_delta(tstamp);
4827         ci.ifa_prefered = preferred;
4828         ci.ifa_valid = valid;
4829
4830         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4831 }
4832
4833 static inline int rt_scope(int ifa_scope)
4834 {
4835         if (ifa_scope & IFA_HOST)
4836                 return RT_SCOPE_HOST;
4837         else if (ifa_scope & IFA_LINK)
4838                 return RT_SCOPE_LINK;
4839         else if (ifa_scope & IFA_SITE)
4840                 return RT_SCOPE_SITE;
4841         else
4842                 return RT_SCOPE_UNIVERSE;
4843 }
4844
4845 static inline int inet6_ifaddr_msgsize(void)
4846 {
4847         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4848                + nla_total_size(16) /* IFA_LOCAL */
4849                + nla_total_size(16) /* IFA_ADDRESS */
4850                + nla_total_size(sizeof(struct ifa_cacheinfo))
4851                + nla_total_size(4)  /* IFA_FLAGS */
4852                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4853 }
4854
4855 enum addr_type_t {
4856         UNICAST_ADDR,
4857         MULTICAST_ADDR,
4858         ANYCAST_ADDR,
4859 };
4860
4861 struct inet6_fill_args {
4862         u32 portid;
4863         u32 seq;
4864         int event;
4865         unsigned int flags;
4866         int netnsid;
4867         int ifindex;
4868         enum addr_type_t type;
4869 };
4870
4871 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4872                              struct inet6_fill_args *args)
4873 {
4874         struct nlmsghdr  *nlh;
4875         u32 preferred, valid;
4876
4877         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4878                         sizeof(struct ifaddrmsg), args->flags);
4879         if (!nlh)
4880                 return -EMSGSIZE;
4881
4882         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4883                       ifa->idev->dev->ifindex);
4884
4885         if (args->netnsid >= 0 &&
4886             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4887                 goto error;
4888
4889         if (!((ifa->flags&IFA_F_PERMANENT) &&
4890               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4891                 preferred = ifa->prefered_lft;
4892                 valid = ifa->valid_lft;
4893                 if (preferred != INFINITY_LIFE_TIME) {
4894                         long tval = (jiffies - ifa->tstamp)/HZ;
4895                         if (preferred > tval)
4896                                 preferred -= tval;
4897                         else
4898                                 preferred = 0;
4899                         if (valid != INFINITY_LIFE_TIME) {
4900                                 if (valid > tval)
4901                                         valid -= tval;
4902                                 else
4903                                         valid = 0;
4904                         }
4905                 }
4906         } else {
4907                 preferred = INFINITY_LIFE_TIME;
4908                 valid = INFINITY_LIFE_TIME;
4909         }
4910
4911         if (!ipv6_addr_any(&ifa->peer_addr)) {
4912                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4913                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4914                         goto error;
4915         } else
4916                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4917                         goto error;
4918
4919         if (ifa->rt_priority &&
4920             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4921                 goto error;
4922
4923         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4924                 goto error;
4925
4926         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4927                 goto error;
4928
4929         nlmsg_end(skb, nlh);
4930         return 0;
4931
4932 error:
4933         nlmsg_cancel(skb, nlh);
4934         return -EMSGSIZE;
4935 }
4936
4937 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4938                                struct inet6_fill_args *args)
4939 {
4940         struct nlmsghdr  *nlh;
4941         u8 scope = RT_SCOPE_UNIVERSE;
4942         int ifindex = ifmca->idev->dev->ifindex;
4943
4944         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4945                 scope = RT_SCOPE_SITE;
4946
4947         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4948                         sizeof(struct ifaddrmsg), args->flags);
4949         if (!nlh)
4950                 return -EMSGSIZE;
4951
4952         if (args->netnsid >= 0 &&
4953             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4954                 return -EMSGSIZE;
4955
4956         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4957         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4958             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4959                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4960                 nlmsg_cancel(skb, nlh);
4961                 return -EMSGSIZE;
4962         }
4963
4964         nlmsg_end(skb, nlh);
4965         return 0;
4966 }
4967
4968 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4969                                struct inet6_fill_args *args)
4970 {
4971         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4972         int ifindex = dev ? dev->ifindex : 1;
4973         struct nlmsghdr  *nlh;
4974         u8 scope = RT_SCOPE_UNIVERSE;
4975
4976         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4977                 scope = RT_SCOPE_SITE;
4978
4979         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4980                         sizeof(struct ifaddrmsg), args->flags);
4981         if (!nlh)
4982                 return -EMSGSIZE;
4983
4984         if (args->netnsid >= 0 &&
4985             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4986                 return -EMSGSIZE;
4987
4988         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4989         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4990             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4991                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4992                 nlmsg_cancel(skb, nlh);
4993                 return -EMSGSIZE;
4994         }
4995
4996         nlmsg_end(skb, nlh);
4997         return 0;
4998 }
4999
5000 /* called with rcu_read_lock() */
5001 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5002                           struct netlink_callback *cb, int s_ip_idx,
5003                           struct inet6_fill_args *fillargs)
5004 {
5005         struct ifmcaddr6 *ifmca;
5006         struct ifacaddr6 *ifaca;
5007         int ip_idx = 0;
5008         int err = 1;
5009
5010         read_lock_bh(&idev->lock);
5011         switch (fillargs->type) {
5012         case UNICAST_ADDR: {
5013                 struct inet6_ifaddr *ifa;
5014                 fillargs->event = RTM_NEWADDR;
5015
5016                 /* unicast address incl. temp addr */
5017                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5018                         if (ip_idx < s_ip_idx)
5019                                 goto next;
5020                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5021                         if (err < 0)
5022                                 break;
5023                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5024 next:
5025                         ip_idx++;
5026                 }
5027                 break;
5028         }
5029         case MULTICAST_ADDR:
5030                 fillargs->event = RTM_GETMULTICAST;
5031
5032                 /* multicast address */
5033                 for (ifmca = idev->mc_list; ifmca;
5034                      ifmca = ifmca->next, ip_idx++) {
5035                         if (ip_idx < s_ip_idx)
5036                                 continue;
5037                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5038                         if (err < 0)
5039                                 break;
5040                 }
5041                 break;
5042         case ANYCAST_ADDR:
5043                 fillargs->event = RTM_GETANYCAST;
5044                 /* anycast address */
5045                 for (ifaca = idev->ac_list; ifaca;
5046                      ifaca = ifaca->aca_next, ip_idx++) {
5047                         if (ip_idx < s_ip_idx)
5048                                 continue;
5049                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5050                         if (err < 0)
5051                                 break;
5052                 }
5053                 break;
5054         default:
5055                 break;
5056         }
5057         read_unlock_bh(&idev->lock);
5058         cb->args[2] = ip_idx;
5059         return err;
5060 }
5061
5062 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5063                                        struct inet6_fill_args *fillargs,
5064                                        struct net **tgt_net, struct sock *sk,
5065                                        struct netlink_callback *cb)
5066 {
5067         struct netlink_ext_ack *extack = cb->extack;
5068         struct nlattr *tb[IFA_MAX+1];
5069         struct ifaddrmsg *ifm;
5070         int err, i;
5071
5072         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5073                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5074                 return -EINVAL;
5075         }
5076
5077         ifm = nlmsg_data(nlh);
5078         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5079                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5080                 return -EINVAL;
5081         }
5082
5083         fillargs->ifindex = ifm->ifa_index;
5084         if (fillargs->ifindex) {
5085                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5086                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5087         }
5088
5089         err = nlmsg_parse_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5090                                  ifa_ipv6_policy, extack);
5091         if (err < 0)
5092                 return err;
5093
5094         for (i = 0; i <= IFA_MAX; ++i) {
5095                 if (!tb[i])
5096                         continue;
5097
5098                 if (i == IFA_TARGET_NETNSID) {
5099                         struct net *net;
5100
5101                         fillargs->netnsid = nla_get_s32(tb[i]);
5102                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5103                         if (IS_ERR(net)) {
5104                                 fillargs->netnsid = -1;
5105                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5106                                 return PTR_ERR(net);
5107                         }
5108                         *tgt_net = net;
5109                 } else {
5110                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5111                         return -EINVAL;
5112                 }
5113         }
5114
5115         return 0;
5116 }
5117
5118 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5119                            enum addr_type_t type)
5120 {
5121         const struct nlmsghdr *nlh = cb->nlh;
5122         struct inet6_fill_args fillargs = {
5123                 .portid = NETLINK_CB(cb->skb).portid,
5124                 .seq = cb->nlh->nlmsg_seq,
5125                 .flags = NLM_F_MULTI,
5126                 .netnsid = -1,
5127                 .type = type,
5128         };
5129         struct net *net = sock_net(skb->sk);
5130         struct net *tgt_net = net;
5131         int idx, s_idx, s_ip_idx;
5132         int h, s_h;
5133         struct net_device *dev;
5134         struct inet6_dev *idev;
5135         struct hlist_head *head;
5136         int err = 0;
5137
5138         s_h = cb->args[0];
5139         s_idx = idx = cb->args[1];
5140         s_ip_idx = cb->args[2];
5141
5142         if (cb->strict_check) {
5143                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5144                                                   skb->sk, cb);
5145                 if (err < 0)
5146                         goto put_tgt_net;
5147
5148                 err = 0;
5149                 if (fillargs.ifindex) {
5150                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5151                         if (!dev) {
5152                                 err = -ENODEV;
5153                                 goto put_tgt_net;
5154                         }
5155                         idev = __in6_dev_get(dev);
5156                         if (idev) {
5157                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5158                                                      &fillargs);
5159                                 if (err > 0)
5160                                         err = 0;
5161                         }
5162                         goto put_tgt_net;
5163                 }
5164         }
5165
5166         rcu_read_lock();
5167         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5168         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5169                 idx = 0;
5170                 head = &tgt_net->dev_index_head[h];
5171                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5172                         if (idx < s_idx)
5173                                 goto cont;
5174                         if (h > s_h || idx > s_idx)
5175                                 s_ip_idx = 0;
5176                         idev = __in6_dev_get(dev);
5177                         if (!idev)
5178                                 goto cont;
5179
5180                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5181                                            &fillargs) < 0)
5182                                 goto done;
5183 cont:
5184                         idx++;
5185                 }
5186         }
5187 done:
5188         rcu_read_unlock();
5189         cb->args[0] = h;
5190         cb->args[1] = idx;
5191 put_tgt_net:
5192         if (fillargs.netnsid >= 0)
5193                 put_net(tgt_net);
5194
5195         return skb->len ? : err;
5196 }
5197
5198 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5199 {
5200         enum addr_type_t type = UNICAST_ADDR;
5201
5202         return inet6_dump_addr(skb, cb, type);
5203 }
5204
5205 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5206 {
5207         enum addr_type_t type = MULTICAST_ADDR;
5208
5209         return inet6_dump_addr(skb, cb, type);
5210 }
5211
5212
5213 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5214 {
5215         enum addr_type_t type = ANYCAST_ADDR;
5216
5217         return inet6_dump_addr(skb, cb, type);
5218 }
5219
5220 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5221                                        const struct nlmsghdr *nlh,
5222                                        struct nlattr **tb,
5223                                        struct netlink_ext_ack *extack)
5224 {
5225         struct ifaddrmsg *ifm;
5226         int i, err;
5227
5228         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5229                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5230                 return -EINVAL;
5231         }
5232
5233         ifm = nlmsg_data(nlh);
5234         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5235                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5236                 return -EINVAL;
5237         }
5238
5239         if (!netlink_strict_get_check(skb))
5240                 return nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX,
5241                                    ifa_ipv6_policy, extack);
5242
5243         err = nlmsg_parse_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5244                                  ifa_ipv6_policy, extack);
5245         if (err)
5246                 return err;
5247
5248         for (i = 0; i <= IFA_MAX; i++) {
5249                 if (!tb[i])
5250                         continue;
5251
5252                 switch (i) {
5253                 case IFA_TARGET_NETNSID:
5254                 case IFA_ADDRESS:
5255                 case IFA_LOCAL:
5256                         break;
5257                 default:
5258                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5259                         return -EINVAL;
5260                 }
5261         }
5262
5263         return 0;
5264 }
5265
5266 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5267                              struct netlink_ext_ack *extack)
5268 {
5269         struct net *net = sock_net(in_skb->sk);
5270         struct inet6_fill_args fillargs = {
5271                 .portid = NETLINK_CB(in_skb).portid,
5272                 .seq = nlh->nlmsg_seq,
5273                 .event = RTM_NEWADDR,
5274                 .flags = 0,
5275                 .netnsid = -1,
5276         };
5277         struct net *tgt_net = net;
5278         struct ifaddrmsg *ifm;
5279         struct nlattr *tb[IFA_MAX+1];
5280         struct in6_addr *addr = NULL, *peer;
5281         struct net_device *dev = NULL;
5282         struct inet6_ifaddr *ifa;
5283         struct sk_buff *skb;
5284         int err;
5285
5286         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5287         if (err < 0)
5288                 return err;
5289
5290         if (tb[IFA_TARGET_NETNSID]) {
5291                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5292
5293                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5294                                                   fillargs.netnsid);
5295                 if (IS_ERR(tgt_net))
5296                         return PTR_ERR(tgt_net);
5297         }
5298
5299         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5300         if (!addr)
5301                 return -EINVAL;
5302
5303         ifm = nlmsg_data(nlh);
5304         if (ifm->ifa_index)
5305                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5306
5307         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5308         if (!ifa) {
5309                 err = -EADDRNOTAVAIL;
5310                 goto errout;
5311         }
5312
5313         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5314         if (!skb) {
5315                 err = -ENOBUFS;
5316                 goto errout_ifa;
5317         }
5318
5319         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5320         if (err < 0) {
5321                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5322                 WARN_ON(err == -EMSGSIZE);
5323                 kfree_skb(skb);
5324                 goto errout_ifa;
5325         }
5326         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5327 errout_ifa:
5328         in6_ifa_put(ifa);
5329 errout:
5330         if (dev)
5331                 dev_put(dev);
5332         if (fillargs.netnsid >= 0)
5333                 put_net(tgt_net);
5334
5335         return err;
5336 }
5337
5338 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5339 {
5340         struct sk_buff *skb;
5341         struct net *net = dev_net(ifa->idev->dev);
5342         struct inet6_fill_args fillargs = {
5343                 .portid = 0,
5344                 .seq = 0,
5345                 .event = event,
5346                 .flags = 0,
5347                 .netnsid = -1,
5348         };
5349         int err = -ENOBUFS;
5350
5351         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5352         if (!skb)
5353                 goto errout;
5354
5355         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5356         if (err < 0) {
5357                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5358                 WARN_ON(err == -EMSGSIZE);
5359                 kfree_skb(skb);
5360                 goto errout;
5361         }
5362         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5363         return;
5364 errout:
5365         if (err < 0)
5366                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5367 }
5368
5369 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5370                                 __s32 *array, int bytes)
5371 {
5372         BUG_ON(bytes < (DEVCONF_MAX * 4));
5373
5374         memset(array, 0, bytes);
5375         array[DEVCONF_FORWARDING] = cnf->forwarding;
5376         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5377         array[DEVCONF_MTU6] = cnf->mtu6;
5378         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5379         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5380         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5381         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5382         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5383         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5384                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5385         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5386                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5387         array[DEVCONF_RTR_SOLICIT_DELAY] =
5388                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5389         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5390         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5391                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5392         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5393                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5394         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5395         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5396         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5397         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5398         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5399         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5400         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5401         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5402         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5403 #ifdef CONFIG_IPV6_ROUTER_PREF
5404         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5405         array[DEVCONF_RTR_PROBE_INTERVAL] =
5406                 jiffies_to_msecs(cnf->rtr_probe_interval);
5407 #ifdef CONFIG_IPV6_ROUTE_INFO
5408         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5409         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5410 #endif
5411 #endif
5412         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5413         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5415         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5416         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5417 #endif
5418 #ifdef CONFIG_IPV6_MROUTE
5419         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5420 #endif
5421         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5422         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5423         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5424         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5425         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5426         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5427         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5428         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5429         /* we omit DEVCONF_STABLE_SECRET for now */
5430         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5431         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5432         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5433         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5434         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5435 #ifdef CONFIG_IPV6_SEG6_HMAC
5436         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5437 #endif
5438         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5439         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5440         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5441         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5442 }
5443
5444 static inline size_t inet6_ifla6_size(void)
5445 {
5446         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5447              + nla_total_size(sizeof(struct ifla_cacheinfo))
5448              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5449              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5450              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5451              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5452              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5453              + 0;
5454 }
5455
5456 static inline size_t inet6_if_nlmsg_size(void)
5457 {
5458         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5459                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5460                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5461                + nla_total_size(4) /* IFLA_MTU */
5462                + nla_total_size(4) /* IFLA_LINK */
5463                + nla_total_size(1) /* IFLA_OPERSTATE */
5464                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5465 }
5466
5467 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5468                                         int bytes)
5469 {
5470         int i;
5471         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5472         BUG_ON(pad < 0);
5473
5474         /* Use put_unaligned() because stats may not be aligned for u64. */
5475         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5476         for (i = 1; i < ICMP6_MIB_MAX; i++)
5477                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5478
5479         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5480 }
5481
5482 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5483                                         int bytes, size_t syncpoff)
5484 {
5485         int i, c;
5486         u64 buff[IPSTATS_MIB_MAX];
5487         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5488
5489         BUG_ON(pad < 0);
5490
5491         memset(buff, 0, sizeof(buff));
5492         buff[0] = IPSTATS_MIB_MAX;
5493
5494         for_each_possible_cpu(c) {
5495                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5496                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5497         }
5498
5499         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5500         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5501 }
5502
5503 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5504                              int bytes)
5505 {
5506         switch (attrtype) {
5507         case IFLA_INET6_STATS:
5508                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5509                                      offsetof(struct ipstats_mib, syncp));
5510                 break;
5511         case IFLA_INET6_ICMP6STATS:
5512                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5513                 break;
5514         }
5515 }
5516
5517 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5518                                   u32 ext_filter_mask)
5519 {
5520         struct nlattr *nla;
5521         struct ifla_cacheinfo ci;
5522
5523         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5524                 goto nla_put_failure;
5525         ci.max_reasm_len = IPV6_MAXPLEN;
5526         ci.tstamp = cstamp_delta(idev->tstamp);
5527         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5528         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5529         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5530                 goto nla_put_failure;
5531         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5532         if (!nla)
5533                 goto nla_put_failure;
5534         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5535
5536         /* XXX - MC not implemented */
5537
5538         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5539                 return 0;
5540
5541         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5542         if (!nla)
5543                 goto nla_put_failure;
5544         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5545
5546         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5547         if (!nla)
5548                 goto nla_put_failure;
5549         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5550
5551         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5552         if (!nla)
5553                 goto nla_put_failure;
5554
5555         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5556                 goto nla_put_failure;
5557
5558         read_lock_bh(&idev->lock);
5559         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5560         read_unlock_bh(&idev->lock);
5561
5562         return 0;
5563
5564 nla_put_failure:
5565         return -EMSGSIZE;
5566 }
5567
5568 static size_t inet6_get_link_af_size(const struct net_device *dev,
5569                                      u32 ext_filter_mask)
5570 {
5571         if (!__in6_dev_get(dev))
5572                 return 0;
5573
5574         return inet6_ifla6_size();
5575 }
5576
5577 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5578                               u32 ext_filter_mask)
5579 {
5580         struct inet6_dev *idev = __in6_dev_get(dev);
5581
5582         if (!idev)
5583                 return -ENODATA;
5584
5585         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5586                 return -EMSGSIZE;
5587
5588         return 0;
5589 }
5590
5591 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5592 {
5593         struct inet6_ifaddr *ifp;
5594         struct net_device *dev = idev->dev;
5595         bool clear_token, update_rs = false;
5596         struct in6_addr ll_addr;
5597
5598         ASSERT_RTNL();
5599
5600         if (!token)
5601                 return -EINVAL;
5602         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5603                 return -EINVAL;
5604         if (!ipv6_accept_ra(idev))
5605                 return -EINVAL;
5606         if (idev->cnf.rtr_solicits == 0)
5607                 return -EINVAL;
5608
5609         write_lock_bh(&idev->lock);
5610
5611         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5612         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5613
5614         write_unlock_bh(&idev->lock);
5615
5616         clear_token = ipv6_addr_any(token);
5617         if (clear_token)
5618                 goto update_lft;
5619
5620         if (!idev->dead && (idev->if_flags & IF_READY) &&
5621             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5622                              IFA_F_OPTIMISTIC)) {
5623                 /* If we're not ready, then normal ifup will take care
5624                  * of this. Otherwise, we need to request our rs here.
5625                  */
5626                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5627                 update_rs = true;
5628         }
5629
5630 update_lft:
5631         write_lock_bh(&idev->lock);
5632
5633         if (update_rs) {
5634                 idev->if_flags |= IF_RS_SENT;
5635                 idev->rs_interval = rfc3315_s14_backoff_init(
5636                         idev->cnf.rtr_solicit_interval);
5637                 idev->rs_probes = 1;
5638                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5639         }
5640
5641         /* Well, that's kinda nasty ... */
5642         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5643                 spin_lock(&ifp->lock);
5644                 if (ifp->tokenized) {
5645                         ifp->valid_lft = 0;
5646                         ifp->prefered_lft = 0;
5647                 }
5648                 spin_unlock(&ifp->lock);
5649         }
5650
5651         write_unlock_bh(&idev->lock);
5652         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5653         addrconf_verify_rtnl();
5654         return 0;
5655 }
5656
5657 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5658         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5659         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5660 };
5661
5662 static int inet6_validate_link_af(const struct net_device *dev,
5663                                   const struct nlattr *nla)
5664 {
5665         struct nlattr *tb[IFLA_INET6_MAX + 1];
5666
5667         if (dev && !__in6_dev_get(dev))
5668                 return -EAFNOSUPPORT;
5669
5670         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5671                                 NULL);
5672 }
5673
5674 static int check_addr_gen_mode(int mode)
5675 {
5676         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5677             mode != IN6_ADDR_GEN_MODE_NONE &&
5678             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5679             mode != IN6_ADDR_GEN_MODE_RANDOM)
5680                 return -EINVAL;
5681         return 1;
5682 }
5683
5684 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5685                                 int mode)
5686 {
5687         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5688             !idev->cnf.stable_secret.initialized &&
5689             !net->ipv6.devconf_dflt->stable_secret.initialized)
5690                 return -EINVAL;
5691         return 1;
5692 }
5693
5694 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5695 {
5696         int err = -EINVAL;
5697         struct inet6_dev *idev = __in6_dev_get(dev);
5698         struct nlattr *tb[IFLA_INET6_MAX + 1];
5699
5700         if (!idev)
5701                 return -EAFNOSUPPORT;
5702
5703         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5704                 BUG();
5705
5706         if (tb[IFLA_INET6_TOKEN]) {
5707                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5708                 if (err)
5709                         return err;
5710         }
5711
5712         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5713                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5714
5715                 if (check_addr_gen_mode(mode) < 0 ||
5716                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5717                         return -EINVAL;
5718
5719                 idev->cnf.addr_gen_mode = mode;
5720                 err = 0;
5721         }
5722
5723         return err;
5724 }
5725
5726 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5727                              u32 portid, u32 seq, int event, unsigned int flags)
5728 {
5729         struct net_device *dev = idev->dev;
5730         struct ifinfomsg *hdr;
5731         struct nlmsghdr *nlh;
5732         void *protoinfo;
5733
5734         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5735         if (!nlh)
5736                 return -EMSGSIZE;
5737
5738         hdr = nlmsg_data(nlh);
5739         hdr->ifi_family = AF_INET6;
5740         hdr->__ifi_pad = 0;
5741         hdr->ifi_type = dev->type;
5742         hdr->ifi_index = dev->ifindex;
5743         hdr->ifi_flags = dev_get_flags(dev);
5744         hdr->ifi_change = 0;
5745
5746         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5747             (dev->addr_len &&
5748              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5749             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5750             (dev->ifindex != dev_get_iflink(dev) &&
5751              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5752             nla_put_u8(skb, IFLA_OPERSTATE,
5753                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5754                 goto nla_put_failure;
5755         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5756         if (!protoinfo)
5757                 goto nla_put_failure;
5758
5759         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5760                 goto nla_put_failure;
5761
5762         nla_nest_end(skb, protoinfo);
5763         nlmsg_end(skb, nlh);
5764         return 0;
5765
5766 nla_put_failure:
5767         nlmsg_cancel(skb, nlh);
5768         return -EMSGSIZE;
5769 }
5770
5771 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5772                                    struct netlink_ext_ack *extack)
5773 {
5774         struct ifinfomsg *ifm;
5775
5776         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5777                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5778                 return -EINVAL;
5779         }
5780
5781         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5782                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5783                 return -EINVAL;
5784         }
5785
5786         ifm = nlmsg_data(nlh);
5787         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5788             ifm->ifi_change || ifm->ifi_index) {
5789                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5790                 return -EINVAL;
5791         }
5792
5793         return 0;
5794 }
5795
5796 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5797 {
5798         struct net *net = sock_net(skb->sk);
5799         int h, s_h;
5800         int idx = 0, s_idx;
5801         struct net_device *dev;
5802         struct inet6_dev *idev;
5803         struct hlist_head *head;
5804
5805         /* only requests using strict checking can pass data to
5806          * influence the dump
5807          */
5808         if (cb->strict_check) {
5809                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5810
5811                 if (err < 0)
5812                         return err;
5813         }
5814
5815         s_h = cb->args[0];
5816         s_idx = cb->args[1];
5817
5818         rcu_read_lock();
5819         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5820                 idx = 0;
5821                 head = &net->dev_index_head[h];
5822                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5823                         if (idx < s_idx)
5824                                 goto cont;
5825                         idev = __in6_dev_get(dev);
5826                         if (!idev)
5827                                 goto cont;
5828                         if (inet6_fill_ifinfo(skb, idev,
5829                                               NETLINK_CB(cb->skb).portid,
5830                                               cb->nlh->nlmsg_seq,
5831                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5832                                 goto out;
5833 cont:
5834                         idx++;
5835                 }
5836         }
5837 out:
5838         rcu_read_unlock();
5839         cb->args[1] = idx;
5840         cb->args[0] = h;
5841
5842         return skb->len;
5843 }
5844
5845 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5846 {
5847         struct sk_buff *skb;
5848         struct net *net = dev_net(idev->dev);
5849         int err = -ENOBUFS;
5850
5851         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5852         if (!skb)
5853                 goto errout;
5854
5855         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5856         if (err < 0) {
5857                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5858                 WARN_ON(err == -EMSGSIZE);
5859                 kfree_skb(skb);
5860                 goto errout;
5861         }
5862         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5863         return;
5864 errout:
5865         if (err < 0)
5866                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5867 }
5868
5869 static inline size_t inet6_prefix_nlmsg_size(void)
5870 {
5871         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5872                + nla_total_size(sizeof(struct in6_addr))
5873                + nla_total_size(sizeof(struct prefix_cacheinfo));
5874 }
5875
5876 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5877                              struct prefix_info *pinfo, u32 portid, u32 seq,
5878                              int event, unsigned int flags)
5879 {
5880         struct prefixmsg *pmsg;
5881         struct nlmsghdr *nlh;
5882         struct prefix_cacheinfo ci;
5883
5884         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5885         if (!nlh)
5886                 return -EMSGSIZE;
5887
5888         pmsg = nlmsg_data(nlh);
5889         pmsg->prefix_family = AF_INET6;
5890         pmsg->prefix_pad1 = 0;
5891         pmsg->prefix_pad2 = 0;
5892         pmsg->prefix_ifindex = idev->dev->ifindex;
5893         pmsg->prefix_len = pinfo->prefix_len;
5894         pmsg->prefix_type = pinfo->type;
5895         pmsg->prefix_pad3 = 0;
5896         pmsg->prefix_flags = 0;
5897         if (pinfo->onlink)
5898                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5899         if (pinfo->autoconf)
5900                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5901
5902         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5903                 goto nla_put_failure;
5904         ci.preferred_time = ntohl(pinfo->prefered);
5905         ci.valid_time = ntohl(pinfo->valid);
5906         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5907                 goto nla_put_failure;
5908         nlmsg_end(skb, nlh);
5909         return 0;
5910
5911 nla_put_failure:
5912         nlmsg_cancel(skb, nlh);
5913         return -EMSGSIZE;
5914 }
5915
5916 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5917                          struct prefix_info *pinfo)
5918 {
5919         struct sk_buff *skb;
5920         struct net *net = dev_net(idev->dev);
5921         int err = -ENOBUFS;
5922
5923         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5924         if (!skb)
5925                 goto errout;
5926
5927         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5928         if (err < 0) {
5929                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5930                 WARN_ON(err == -EMSGSIZE);
5931                 kfree_skb(skb);
5932                 goto errout;
5933         }
5934         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5935         return;
5936 errout:
5937         if (err < 0)
5938                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5939 }
5940
5941 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5942 {
5943         struct net *net = dev_net(ifp->idev->dev);
5944
5945         if (event)
5946                 ASSERT_RTNL();
5947
5948         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5949
5950         switch (event) {
5951         case RTM_NEWADDR:
5952                 /*
5953                  * If the address was optimistic
5954                  * we inserted the route at the start of
5955                  * our DAD process, so we don't need
5956                  * to do it again
5957                  */
5958                 if (!rcu_access_pointer(ifp->rt->fib6_node))
5959                         ip6_ins_rt(net, ifp->rt);
5960                 if (ifp->idev->cnf.forwarding)
5961                         addrconf_join_anycast(ifp);
5962                 if (!ipv6_addr_any(&ifp->peer_addr))
5963                         addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5964                                               ifp->idev->dev, 0, 0,
5965                                               GFP_ATOMIC);
5966                 break;
5967         case RTM_DELADDR:
5968                 if (ifp->idev->cnf.forwarding)
5969                         addrconf_leave_anycast(ifp);
5970                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5971                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5972                         struct fib6_info *rt;
5973
5974                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5975                                                        ifp->idev->dev, 0, 0);
5976                         if (rt)
5977                                 ip6_del_rt(net, rt);
5978                 }
5979                 if (ifp->rt) {
5980                         ip6_del_rt(net, ifp->rt);
5981                         ifp->rt = NULL;
5982                 }
5983                 rt_genid_bump_ipv6(net);
5984                 break;
5985         }
5986         atomic_inc(&net->ipv6.dev_addr_genid);
5987 }
5988
5989 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5990 {
5991         rcu_read_lock_bh();
5992         if (likely(ifp->idev->dead == 0))
5993                 __ipv6_ifa_notify(event, ifp);
5994         rcu_read_unlock_bh();
5995 }
5996
5997 #ifdef CONFIG_SYSCTL
5998
5999 static
6000 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6001                            void __user *buffer, size_t *lenp, loff_t *ppos)
6002 {
6003         int *valp = ctl->data;
6004         int val = *valp;
6005         loff_t pos = *ppos;
6006         struct ctl_table lctl;
6007         int ret;
6008
6009         /*
6010          * ctl->data points to idev->cnf.forwarding, we should
6011          * not modify it until we get the rtnl lock.
6012          */
6013         lctl = *ctl;
6014         lctl.data = &val;
6015
6016         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6017
6018         if (write)
6019                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6020         if (ret)
6021                 *ppos = pos;
6022         return ret;
6023 }
6024
6025 static
6026 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6027                         void __user *buffer, size_t *lenp, loff_t *ppos)
6028 {
6029         struct inet6_dev *idev = ctl->extra1;
6030         int min_mtu = IPV6_MIN_MTU;
6031         struct ctl_table lctl;
6032
6033         lctl = *ctl;
6034         lctl.extra1 = &min_mtu;
6035         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6036
6037         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6038 }
6039
6040 static void dev_disable_change(struct inet6_dev *idev)
6041 {
6042         struct netdev_notifier_info info;
6043
6044         if (!idev || !idev->dev)
6045                 return;
6046
6047         netdev_notifier_info_init(&info, idev->dev);
6048         if (idev->cnf.disable_ipv6)
6049                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6050         else
6051                 addrconf_notify(NULL, NETDEV_UP, &info);
6052 }
6053
6054 static void addrconf_disable_change(struct net *net, __s32 newf)
6055 {
6056         struct net_device *dev;
6057         struct inet6_dev *idev;
6058
6059         for_each_netdev(net, dev) {
6060                 idev = __in6_dev_get(dev);
6061                 if (idev) {
6062                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6063                         idev->cnf.disable_ipv6 = newf;
6064                         if (changed)
6065                                 dev_disable_change(idev);
6066                 }
6067         }
6068 }
6069
6070 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6071 {
6072         struct net *net;
6073         int old;
6074
6075         if (!rtnl_trylock())
6076                 return restart_syscall();
6077
6078         net = (struct net *)table->extra2;
6079         old = *p;
6080         *p = newf;
6081
6082         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6083                 rtnl_unlock();
6084                 return 0;
6085         }
6086
6087         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6088                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6089                 addrconf_disable_change(net, newf);
6090         } else if ((!newf) ^ (!old))
6091                 dev_disable_change((struct inet6_dev *)table->extra1);
6092
6093         rtnl_unlock();
6094         return 0;
6095 }
6096
6097 static
6098 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6099                             void __user *buffer, size_t *lenp, loff_t *ppos)
6100 {
6101         int *valp = ctl->data;
6102         int val = *valp;
6103         loff_t pos = *ppos;
6104         struct ctl_table lctl;
6105         int ret;
6106
6107         /*
6108          * ctl->data points to idev->cnf.disable_ipv6, we should
6109          * not modify it until we get the rtnl lock.
6110          */
6111         lctl = *ctl;
6112         lctl.data = &val;
6113
6114         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6115
6116         if (write)
6117                 ret = addrconf_disable_ipv6(ctl, valp, val);
6118         if (ret)
6119                 *ppos = pos;
6120         return ret;
6121 }
6122
6123 static
6124 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6125                               void __user *buffer, size_t *lenp, loff_t *ppos)
6126 {
6127         int *valp = ctl->data;
6128         int ret;
6129         int old, new;
6130
6131         old = *valp;
6132         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6133         new = *valp;
6134
6135         if (write && old != new) {
6136                 struct net *net = ctl->extra2;
6137
6138                 if (!rtnl_trylock())
6139                         return restart_syscall();
6140
6141                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6142                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6143                                                      NETCONFA_PROXY_NEIGH,
6144                                                      NETCONFA_IFINDEX_DEFAULT,
6145                                                      net->ipv6.devconf_dflt);
6146                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6147                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6148                                                      NETCONFA_PROXY_NEIGH,
6149                                                      NETCONFA_IFINDEX_ALL,
6150                                                      net->ipv6.devconf_all);
6151                 else {
6152                         struct inet6_dev *idev = ctl->extra1;
6153
6154                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6155                                                      NETCONFA_PROXY_NEIGH,
6156                                                      idev->dev->ifindex,
6157                                                      &idev->cnf);
6158                 }
6159                 rtnl_unlock();
6160         }
6161
6162         return ret;
6163 }
6164
6165 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6166                                          void __user *buffer, size_t *lenp,
6167                                          loff_t *ppos)
6168 {
6169         int ret = 0;
6170         u32 new_val;
6171         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6172         struct net *net = (struct net *)ctl->extra2;
6173         struct ctl_table tmp = {
6174                 .data = &new_val,
6175                 .maxlen = sizeof(new_val),
6176                 .mode = ctl->mode,
6177         };
6178
6179         if (!rtnl_trylock())
6180                 return restart_syscall();
6181
6182         new_val = *((u32 *)ctl->data);
6183
6184         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6185         if (ret != 0)
6186                 goto out;
6187
6188         if (write) {
6189                 if (check_addr_gen_mode(new_val) < 0) {
6190                         ret = -EINVAL;
6191                         goto out;
6192                 }
6193
6194                 if (idev) {
6195                         if (check_stable_privacy(idev, net, new_val) < 0) {
6196                                 ret = -EINVAL;
6197                                 goto out;
6198                         }
6199
6200                         if (idev->cnf.addr_gen_mode != new_val) {
6201                                 idev->cnf.addr_gen_mode = new_val;
6202                                 addrconf_dev_config(idev->dev);
6203                         }
6204                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6205                         struct net_device *dev;
6206
6207                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6208                         for_each_netdev(net, dev) {
6209                                 idev = __in6_dev_get(dev);
6210                                 if (idev &&
6211                                     idev->cnf.addr_gen_mode != new_val) {
6212                                         idev->cnf.addr_gen_mode = new_val;
6213                                         addrconf_dev_config(idev->dev);
6214                                 }
6215                         }
6216                 }
6217
6218                 *((u32 *)ctl->data) = new_val;
6219         }
6220
6221 out:
6222         rtnl_unlock();
6223
6224         return ret;
6225 }
6226
6227 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6228                                          void __user *buffer, size_t *lenp,
6229                                          loff_t *ppos)
6230 {
6231         int err;
6232         struct in6_addr addr;
6233         char str[IPV6_MAX_STRLEN];
6234         struct ctl_table lctl = *ctl;
6235         struct net *net = ctl->extra2;
6236         struct ipv6_stable_secret *secret = ctl->data;
6237
6238         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6239                 return -EIO;
6240
6241         lctl.maxlen = IPV6_MAX_STRLEN;
6242         lctl.data = str;
6243
6244         if (!rtnl_trylock())
6245                 return restart_syscall();
6246
6247         if (!write && !secret->initialized) {
6248                 err = -EIO;
6249                 goto out;
6250         }
6251
6252         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6253         if (err >= sizeof(str)) {
6254                 err = -EIO;
6255                 goto out;
6256         }
6257
6258         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6259         if (err || !write)
6260                 goto out;
6261
6262         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6263                 err = -EIO;
6264                 goto out;
6265         }
6266
6267         secret->initialized = true;
6268         secret->secret = addr;
6269
6270         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6271                 struct net_device *dev;
6272
6273                 for_each_netdev(net, dev) {
6274                         struct inet6_dev *idev = __in6_dev_get(dev);
6275
6276                         if (idev) {
6277                                 idev->cnf.addr_gen_mode =
6278                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6279                         }
6280                 }
6281         } else {
6282                 struct inet6_dev *idev = ctl->extra1;
6283
6284                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6285         }
6286
6287 out:
6288         rtnl_unlock();
6289
6290         return err;
6291 }
6292
6293 static
6294 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6295                                                 int write,
6296                                                 void __user *buffer,
6297                                                 size_t *lenp,
6298                                                 loff_t *ppos)
6299 {
6300         int *valp = ctl->data;
6301         int val = *valp;
6302         loff_t pos = *ppos;
6303         struct ctl_table lctl;
6304         int ret;
6305
6306         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6307          * we should not modify it until we get the rtnl lock.
6308          */
6309         lctl = *ctl;
6310         lctl.data = &val;
6311
6312         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6313
6314         if (write)
6315                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6316         if (ret)
6317                 *ppos = pos;
6318         return ret;
6319 }
6320
6321 static
6322 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6323 {
6324         if (rt) {
6325                 if (action)
6326                         rt->dst.flags |= DST_NOPOLICY;
6327                 else
6328                         rt->dst.flags &= ~DST_NOPOLICY;
6329         }
6330 }
6331
6332 static
6333 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6334 {
6335         struct inet6_ifaddr *ifa;
6336
6337         read_lock_bh(&idev->lock);
6338         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6339                 spin_lock(&ifa->lock);
6340                 if (ifa->rt) {
6341                         struct fib6_info *rt = ifa->rt;
6342                         int cpu;
6343
6344                         rcu_read_lock();
6345                         ifa->rt->dst_nopolicy = val ? true : false;
6346                         if (rt->rt6i_pcpu) {
6347                                 for_each_possible_cpu(cpu) {
6348                                         struct rt6_info **rtp;
6349
6350                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6351                                         addrconf_set_nopolicy(*rtp, val);
6352                                 }
6353                         }
6354                         rcu_read_unlock();
6355                 }
6356                 spin_unlock(&ifa->lock);
6357         }
6358         read_unlock_bh(&idev->lock);
6359 }
6360
6361 static
6362 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6363 {
6364         struct inet6_dev *idev;
6365         struct net *net;
6366
6367         if (!rtnl_trylock())
6368                 return restart_syscall();
6369
6370         *valp = val;
6371
6372         net = (struct net *)ctl->extra2;
6373         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6374                 rtnl_unlock();
6375                 return 0;
6376         }
6377
6378         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6379                 struct net_device *dev;
6380
6381                 for_each_netdev(net, dev) {
6382                         idev = __in6_dev_get(dev);
6383                         if (idev)
6384                                 addrconf_disable_policy_idev(idev, val);
6385                 }
6386         } else {
6387                 idev = (struct inet6_dev *)ctl->extra1;
6388                 addrconf_disable_policy_idev(idev, val);
6389         }
6390
6391         rtnl_unlock();
6392         return 0;
6393 }
6394
6395 static
6396 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6397                                    void __user *buffer, size_t *lenp,
6398                                    loff_t *ppos)
6399 {
6400         int *valp = ctl->data;
6401         int val = *valp;
6402         loff_t pos = *ppos;
6403         struct ctl_table lctl;
6404         int ret;
6405
6406         lctl = *ctl;
6407         lctl.data = &val;
6408         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6409
6410         if (write && (*valp != val))
6411                 ret = addrconf_disable_policy(ctl, valp, val);
6412
6413         if (ret)
6414                 *ppos = pos;
6415
6416         return ret;
6417 }
6418
6419 static int minus_one = -1;
6420 static const int zero = 0;
6421 static const int one = 1;
6422 static const int two_five_five = 255;
6423
6424 static const struct ctl_table addrconf_sysctl[] = {
6425         {
6426                 .procname       = "forwarding",
6427                 .data           = &ipv6_devconf.forwarding,
6428                 .maxlen         = sizeof(int),
6429                 .mode           = 0644,
6430                 .proc_handler   = addrconf_sysctl_forward,
6431         },
6432         {
6433                 .procname       = "hop_limit",
6434                 .data           = &ipv6_devconf.hop_limit,
6435                 .maxlen         = sizeof(int),
6436                 .mode           = 0644,
6437                 .proc_handler   = proc_dointvec_minmax,
6438                 .extra1         = (void *)&one,
6439                 .extra2         = (void *)&two_five_five,
6440         },
6441         {
6442                 .procname       = "mtu",
6443                 .data           = &ipv6_devconf.mtu6,
6444                 .maxlen         = sizeof(int),
6445                 .mode           = 0644,
6446                 .proc_handler   = addrconf_sysctl_mtu,
6447         },
6448         {
6449                 .procname       = "accept_ra",
6450                 .data           = &ipv6_devconf.accept_ra,
6451                 .maxlen         = sizeof(int),
6452                 .mode           = 0644,
6453                 .proc_handler   = proc_dointvec,
6454         },
6455         {
6456                 .procname       = "accept_redirects",
6457                 .data           = &ipv6_devconf.accept_redirects,
6458                 .maxlen         = sizeof(int),
6459                 .mode           = 0644,
6460                 .proc_handler   = proc_dointvec,
6461         },
6462         {
6463                 .procname       = "autoconf",
6464                 .data           = &ipv6_devconf.autoconf,
6465                 .maxlen         = sizeof(int),
6466                 .mode           = 0644,
6467                 .proc_handler   = proc_dointvec,
6468         },
6469         {
6470                 .procname       = "dad_transmits",
6471                 .data           = &ipv6_devconf.dad_transmits,
6472                 .maxlen         = sizeof(int),
6473                 .mode           = 0644,
6474                 .proc_handler   = proc_dointvec,
6475         },
6476         {
6477                 .procname       = "router_solicitations",
6478                 .data           = &ipv6_devconf.rtr_solicits,
6479                 .maxlen         = sizeof(int),
6480                 .mode           = 0644,
6481                 .proc_handler   = proc_dointvec_minmax,
6482                 .extra1         = &minus_one,
6483         },
6484         {
6485                 .procname       = "router_solicitation_interval",
6486                 .data           = &ipv6_devconf.rtr_solicit_interval,
6487                 .maxlen         = sizeof(int),
6488                 .mode           = 0644,
6489                 .proc_handler   = proc_dointvec_jiffies,
6490         },
6491         {
6492                 .procname       = "router_solicitation_max_interval",
6493                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6494                 .maxlen         = sizeof(int),
6495                 .mode           = 0644,
6496                 .proc_handler   = proc_dointvec_jiffies,
6497         },
6498         {
6499                 .procname       = "router_solicitation_delay",
6500                 .data           = &ipv6_devconf.rtr_solicit_delay,
6501                 .maxlen         = sizeof(int),
6502                 .mode           = 0644,
6503                 .proc_handler   = proc_dointvec_jiffies,
6504         },
6505         {
6506                 .procname       = "force_mld_version",
6507                 .data           = &ipv6_devconf.force_mld_version,
6508                 .maxlen         = sizeof(int),
6509                 .mode           = 0644,
6510                 .proc_handler   = proc_dointvec,
6511         },
6512         {
6513                 .procname       = "mldv1_unsolicited_report_interval",
6514                 .data           =
6515                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6516                 .maxlen         = sizeof(int),
6517                 .mode           = 0644,
6518                 .proc_handler   = proc_dointvec_ms_jiffies,
6519         },
6520         {
6521                 .procname       = "mldv2_unsolicited_report_interval",
6522                 .data           =
6523                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6524                 .maxlen         = sizeof(int),
6525                 .mode           = 0644,
6526                 .proc_handler   = proc_dointvec_ms_jiffies,
6527         },
6528         {
6529                 .procname       = "use_tempaddr",
6530                 .data           = &ipv6_devconf.use_tempaddr,
6531                 .maxlen         = sizeof(int),
6532                 .mode           = 0644,
6533                 .proc_handler   = proc_dointvec,
6534         },
6535         {
6536                 .procname       = "temp_valid_lft",
6537                 .data           = &ipv6_devconf.temp_valid_lft,
6538                 .maxlen         = sizeof(int),
6539                 .mode           = 0644,
6540                 .proc_handler   = proc_dointvec,
6541         },
6542         {
6543                 .procname       = "temp_prefered_lft",
6544                 .data           = &ipv6_devconf.temp_prefered_lft,
6545                 .maxlen         = sizeof(int),
6546                 .mode           = 0644,
6547                 .proc_handler   = proc_dointvec,
6548         },
6549         {
6550                 .procname       = "regen_max_retry",
6551                 .data           = &ipv6_devconf.regen_max_retry,
6552                 .maxlen         = sizeof(int),
6553                 .mode           = 0644,
6554                 .proc_handler   = proc_dointvec,
6555         },
6556         {
6557                 .procname       = "max_desync_factor",
6558                 .data           = &ipv6_devconf.max_desync_factor,
6559                 .maxlen         = sizeof(int),
6560                 .mode           = 0644,
6561                 .proc_handler   = proc_dointvec,
6562         },
6563         {
6564                 .procname       = "max_addresses",
6565                 .data           = &ipv6_devconf.max_addresses,
6566                 .maxlen         = sizeof(int),
6567                 .mode           = 0644,
6568                 .proc_handler   = proc_dointvec,
6569         },
6570         {
6571                 .procname       = "accept_ra_defrtr",
6572                 .data           = &ipv6_devconf.accept_ra_defrtr,
6573                 .maxlen         = sizeof(int),
6574                 .mode           = 0644,
6575                 .proc_handler   = proc_dointvec,
6576         },
6577         {
6578                 .procname       = "accept_ra_min_hop_limit",
6579                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6580                 .maxlen         = sizeof(int),
6581                 .mode           = 0644,
6582                 .proc_handler   = proc_dointvec,
6583         },
6584         {
6585                 .procname       = "accept_ra_pinfo",
6586                 .data           = &ipv6_devconf.accept_ra_pinfo,
6587                 .maxlen         = sizeof(int),
6588                 .mode           = 0644,
6589                 .proc_handler   = proc_dointvec,
6590         },
6591 #ifdef CONFIG_IPV6_ROUTER_PREF
6592         {
6593                 .procname       = "accept_ra_rtr_pref",
6594                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6595                 .maxlen         = sizeof(int),
6596                 .mode           = 0644,
6597                 .proc_handler   = proc_dointvec,
6598         },
6599         {
6600                 .procname       = "router_probe_interval",
6601                 .data           = &ipv6_devconf.rtr_probe_interval,
6602                 .maxlen         = sizeof(int),
6603                 .mode           = 0644,
6604                 .proc_handler   = proc_dointvec_jiffies,
6605         },
6606 #ifdef CONFIG_IPV6_ROUTE_INFO
6607         {
6608                 .procname       = "accept_ra_rt_info_min_plen",
6609                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6610                 .maxlen         = sizeof(int),
6611                 .mode           = 0644,
6612                 .proc_handler   = proc_dointvec,
6613         },
6614         {
6615                 .procname       = "accept_ra_rt_info_max_plen",
6616                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6617                 .maxlen         = sizeof(int),
6618                 .mode           = 0644,
6619                 .proc_handler   = proc_dointvec,
6620         },
6621 #endif
6622 #endif
6623         {
6624                 .procname       = "proxy_ndp",
6625                 .data           = &ipv6_devconf.proxy_ndp,
6626                 .maxlen         = sizeof(int),
6627                 .mode           = 0644,
6628                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6629         },
6630         {
6631                 .procname       = "accept_source_route",
6632                 .data           = &ipv6_devconf.accept_source_route,
6633                 .maxlen         = sizeof(int),
6634                 .mode           = 0644,
6635                 .proc_handler   = proc_dointvec,
6636         },
6637 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6638         {
6639                 .procname       = "optimistic_dad",
6640                 .data           = &ipv6_devconf.optimistic_dad,
6641                 .maxlen         = sizeof(int),
6642                 .mode           = 0644,
6643                 .proc_handler   = proc_dointvec,
6644         },
6645         {
6646                 .procname       = "use_optimistic",
6647                 .data           = &ipv6_devconf.use_optimistic,
6648                 .maxlen         = sizeof(int),
6649                 .mode           = 0644,
6650                 .proc_handler   = proc_dointvec,
6651         },
6652 #endif
6653 #ifdef CONFIG_IPV6_MROUTE
6654         {
6655                 .procname       = "mc_forwarding",
6656                 .data           = &ipv6_devconf.mc_forwarding,
6657                 .maxlen         = sizeof(int),
6658                 .mode           = 0444,
6659                 .proc_handler   = proc_dointvec,
6660         },
6661 #endif
6662         {
6663                 .procname       = "disable_ipv6",
6664                 .data           = &ipv6_devconf.disable_ipv6,
6665                 .maxlen         = sizeof(int),
6666                 .mode           = 0644,
6667                 .proc_handler   = addrconf_sysctl_disable,
6668         },
6669         {
6670                 .procname       = "accept_dad",
6671                 .data           = &ipv6_devconf.accept_dad,
6672                 .maxlen         = sizeof(int),
6673                 .mode           = 0644,
6674                 .proc_handler   = proc_dointvec,
6675         },
6676         {
6677                 .procname       = "force_tllao",
6678                 .data           = &ipv6_devconf.force_tllao,
6679                 .maxlen         = sizeof(int),
6680                 .mode           = 0644,
6681                 .proc_handler   = proc_dointvec
6682         },
6683         {
6684                 .procname       = "ndisc_notify",
6685                 .data           = &ipv6_devconf.ndisc_notify,
6686                 .maxlen         = sizeof(int),
6687                 .mode           = 0644,
6688                 .proc_handler   = proc_dointvec
6689         },
6690         {
6691                 .procname       = "suppress_frag_ndisc",
6692                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6693                 .maxlen         = sizeof(int),
6694                 .mode           = 0644,
6695                 .proc_handler   = proc_dointvec
6696         },
6697         {
6698                 .procname       = "accept_ra_from_local",
6699                 .data           = &ipv6_devconf.accept_ra_from_local,
6700                 .maxlen         = sizeof(int),
6701                 .mode           = 0644,
6702                 .proc_handler   = proc_dointvec,
6703         },
6704         {
6705                 .procname       = "accept_ra_mtu",
6706                 .data           = &ipv6_devconf.accept_ra_mtu,
6707                 .maxlen         = sizeof(int),
6708                 .mode           = 0644,
6709                 .proc_handler   = proc_dointvec,
6710         },
6711         {
6712                 .procname       = "stable_secret",
6713                 .data           = &ipv6_devconf.stable_secret,
6714                 .maxlen         = IPV6_MAX_STRLEN,
6715                 .mode           = 0600,
6716                 .proc_handler   = addrconf_sysctl_stable_secret,
6717         },
6718         {
6719                 .procname       = "use_oif_addrs_only",
6720                 .data           = &ipv6_devconf.use_oif_addrs_only,
6721                 .maxlen         = sizeof(int),
6722                 .mode           = 0644,
6723                 .proc_handler   = proc_dointvec,
6724         },
6725         {
6726                 .procname       = "ignore_routes_with_linkdown",
6727                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6728                 .maxlen         = sizeof(int),
6729                 .mode           = 0644,
6730                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6731         },
6732         {
6733                 .procname       = "drop_unicast_in_l2_multicast",
6734                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6735                 .maxlen         = sizeof(int),
6736                 .mode           = 0644,
6737                 .proc_handler   = proc_dointvec,
6738         },
6739         {
6740                 .procname       = "drop_unsolicited_na",
6741                 .data           = &ipv6_devconf.drop_unsolicited_na,
6742                 .maxlen         = sizeof(int),
6743                 .mode           = 0644,
6744                 .proc_handler   = proc_dointvec,
6745         },
6746         {
6747                 .procname       = "keep_addr_on_down",
6748                 .data           = &ipv6_devconf.keep_addr_on_down,
6749                 .maxlen         = sizeof(int),
6750                 .mode           = 0644,
6751                 .proc_handler   = proc_dointvec,
6752
6753         },
6754         {
6755                 .procname       = "seg6_enabled",
6756                 .data           = &ipv6_devconf.seg6_enabled,
6757                 .maxlen         = sizeof(int),
6758                 .mode           = 0644,
6759                 .proc_handler   = proc_dointvec,
6760         },
6761 #ifdef CONFIG_IPV6_SEG6_HMAC
6762         {
6763                 .procname       = "seg6_require_hmac",
6764                 .data           = &ipv6_devconf.seg6_require_hmac,
6765                 .maxlen         = sizeof(int),
6766                 .mode           = 0644,
6767                 .proc_handler   = proc_dointvec,
6768         },
6769 #endif
6770         {
6771                 .procname       = "enhanced_dad",
6772                 .data           = &ipv6_devconf.enhanced_dad,
6773                 .maxlen         = sizeof(int),
6774                 .mode           = 0644,
6775                 .proc_handler   = proc_dointvec,
6776         },
6777         {
6778                 .procname               = "addr_gen_mode",
6779                 .data                   = &ipv6_devconf.addr_gen_mode,
6780                 .maxlen                 = sizeof(int),
6781                 .mode                   = 0644,
6782                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6783         },
6784         {
6785                 .procname       = "disable_policy",
6786                 .data           = &ipv6_devconf.disable_policy,
6787                 .maxlen         = sizeof(int),
6788                 .mode           = 0644,
6789                 .proc_handler   = addrconf_sysctl_disable_policy,
6790         },
6791         {
6792                 .procname       = "ndisc_tclass",
6793                 .data           = &ipv6_devconf.ndisc_tclass,
6794                 .maxlen         = sizeof(int),
6795                 .mode           = 0644,
6796                 .proc_handler   = proc_dointvec_minmax,
6797                 .extra1         = (void *)&zero,
6798                 .extra2         = (void *)&two_five_five,
6799         },
6800         {
6801                 /* sentinel */
6802         }
6803 };
6804
6805 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6806                 struct inet6_dev *idev, struct ipv6_devconf *p)
6807 {
6808         int i, ifindex;
6809         struct ctl_table *table;
6810         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6811
6812         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6813         if (!table)
6814                 goto out;
6815
6816         for (i = 0; table[i].data; i++) {
6817                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6818                 /* If one of these is already set, then it is not safe to
6819                  * overwrite either of them: this makes proc_dointvec_minmax
6820                  * usable.
6821                  */
6822                 if (!table[i].extra1 && !table[i].extra2) {
6823                         table[i].extra1 = idev; /* embedded; no ref */
6824                         table[i].extra2 = net;
6825                 }
6826         }
6827
6828         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6829
6830         p->sysctl_header = register_net_sysctl(net, path, table);
6831         if (!p->sysctl_header)
6832                 goto free;
6833
6834         if (!strcmp(dev_name, "all"))
6835                 ifindex = NETCONFA_IFINDEX_ALL;
6836         else if (!strcmp(dev_name, "default"))
6837                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6838         else
6839                 ifindex = idev->dev->ifindex;
6840         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6841                                      ifindex, p);
6842         return 0;
6843
6844 free:
6845         kfree(table);
6846 out:
6847         return -ENOBUFS;
6848 }
6849
6850 static void __addrconf_sysctl_unregister(struct net *net,
6851                                          struct ipv6_devconf *p, int ifindex)
6852 {
6853         struct ctl_table *table;
6854
6855         if (!p->sysctl_header)
6856                 return;
6857
6858         table = p->sysctl_header->ctl_table_arg;
6859         unregister_net_sysctl_table(p->sysctl_header);
6860         p->sysctl_header = NULL;
6861         kfree(table);
6862
6863         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6864 }
6865
6866 static int addrconf_sysctl_register(struct inet6_dev *idev)
6867 {
6868         int err;
6869
6870         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6871                 return -EINVAL;
6872
6873         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6874                                     &ndisc_ifinfo_sysctl_change);
6875         if (err)
6876                 return err;
6877         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6878                                          idev, &idev->cnf);
6879         if (err)
6880                 neigh_sysctl_unregister(idev->nd_parms);
6881
6882         return err;
6883 }
6884
6885 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6886 {
6887         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6888                                      idev->dev->ifindex);
6889         neigh_sysctl_unregister(idev->nd_parms);
6890 }
6891
6892
6893 #endif
6894
6895 static int __net_init addrconf_init_net(struct net *net)
6896 {
6897         int err = -ENOMEM;
6898         struct ipv6_devconf *all, *dflt;
6899
6900         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6901         if (!all)
6902                 goto err_alloc_all;
6903
6904         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6905         if (!dflt)
6906                 goto err_alloc_dflt;
6907
6908         if (IS_ENABLED(CONFIG_SYSCTL) &&
6909             sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
6910                 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
6911                 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
6912         }
6913
6914         /* these will be inherited by all namespaces */
6915         dflt->autoconf = ipv6_defaults.autoconf;
6916         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6917
6918         dflt->stable_secret.initialized = false;
6919         all->stable_secret.initialized = false;
6920
6921         net->ipv6.devconf_all = all;
6922         net->ipv6.devconf_dflt = dflt;
6923
6924 #ifdef CONFIG_SYSCTL
6925         err = __addrconf_sysctl_register(net, "all", NULL, all);
6926         if (err < 0)
6927                 goto err_reg_all;
6928
6929         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6930         if (err < 0)
6931                 goto err_reg_dflt;
6932 #endif
6933         return 0;
6934
6935 #ifdef CONFIG_SYSCTL
6936 err_reg_dflt:
6937         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6938 err_reg_all:
6939         kfree(dflt);
6940 #endif
6941 err_alloc_dflt:
6942         kfree(all);
6943 err_alloc_all:
6944         return err;
6945 }
6946
6947 static void __net_exit addrconf_exit_net(struct net *net)
6948 {
6949 #ifdef CONFIG_SYSCTL
6950         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6951                                      NETCONFA_IFINDEX_DEFAULT);
6952         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6953                                      NETCONFA_IFINDEX_ALL);
6954 #endif
6955         kfree(net->ipv6.devconf_dflt);
6956         kfree(net->ipv6.devconf_all);
6957 }
6958
6959 static struct pernet_operations addrconf_ops = {
6960         .init = addrconf_init_net,
6961         .exit = addrconf_exit_net,
6962 };
6963
6964 static struct rtnl_af_ops inet6_ops __read_mostly = {
6965         .family           = AF_INET6,
6966         .fill_link_af     = inet6_fill_link_af,
6967         .get_link_af_size = inet6_get_link_af_size,
6968         .validate_link_af = inet6_validate_link_af,
6969         .set_link_af      = inet6_set_link_af,
6970 };
6971
6972 /*
6973  *      Init / cleanup code
6974  */
6975
6976 int __init addrconf_init(void)
6977 {
6978         struct inet6_dev *idev;
6979         int i, err;
6980
6981         err = ipv6_addr_label_init();
6982         if (err < 0) {
6983                 pr_crit("%s: cannot initialize default policy table: %d\n",
6984                         __func__, err);
6985                 goto out;
6986         }
6987
6988         err = register_pernet_subsys(&addrconf_ops);
6989         if (err < 0)
6990                 goto out_addrlabel;
6991
6992         addrconf_wq = create_workqueue("ipv6_addrconf");
6993         if (!addrconf_wq) {
6994                 err = -ENOMEM;
6995                 goto out_nowq;
6996         }
6997
6998         /* The addrconf netdev notifier requires that loopback_dev
6999          * has it's ipv6 private information allocated and setup
7000          * before it can bring up and give link-local addresses
7001          * to other devices which are up.
7002          *
7003          * Unfortunately, loopback_dev is not necessarily the first
7004          * entry in the global dev_base list of net devices.  In fact,
7005          * it is likely to be the very last entry on that list.
7006          * So this causes the notifier registry below to try and
7007          * give link-local addresses to all devices besides loopback_dev
7008          * first, then loopback_dev, which cases all the non-loopback_dev
7009          * devices to fail to get a link-local address.
7010          *
7011          * So, as a temporary fix, allocate the ipv6 structure for
7012          * loopback_dev first by hand.
7013          * Longer term, all of the dependencies ipv6 has upon the loopback
7014          * device and it being up should be removed.
7015          */
7016         rtnl_lock();
7017         idev = ipv6_add_dev(init_net.loopback_dev);
7018         rtnl_unlock();
7019         if (IS_ERR(idev)) {
7020                 err = PTR_ERR(idev);
7021                 goto errlo;
7022         }
7023
7024         ip6_route_init_special_entries();
7025
7026         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7027                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7028
7029         register_netdevice_notifier(&ipv6_dev_notf);
7030
7031         addrconf_verify();
7032
7033         rtnl_af_register(&inet6_ops);
7034
7035         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7036                                    NULL, inet6_dump_ifinfo, 0);
7037         if (err < 0)
7038                 goto errout;
7039
7040         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7041                                    inet6_rtm_newaddr, NULL, 0);
7042         if (err < 0)
7043                 goto errout;
7044         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7045                                    inet6_rtm_deladdr, NULL, 0);
7046         if (err < 0)
7047                 goto errout;
7048         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7049                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7050                                    RTNL_FLAG_DOIT_UNLOCKED);
7051         if (err < 0)
7052                 goto errout;
7053         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7054                                    NULL, inet6_dump_ifmcaddr, 0);
7055         if (err < 0)
7056                 goto errout;
7057         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7058                                    NULL, inet6_dump_ifacaddr, 0);
7059         if (err < 0)
7060                 goto errout;
7061         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7062                                    inet6_netconf_get_devconf,
7063                                    inet6_netconf_dump_devconf,
7064                                    RTNL_FLAG_DOIT_UNLOCKED);
7065         if (err < 0)
7066                 goto errout;
7067         err = ipv6_addr_label_rtnl_register();
7068         if (err < 0)
7069                 goto errout;
7070
7071         return 0;
7072 errout:
7073         rtnl_unregister_all(PF_INET6);
7074         rtnl_af_unregister(&inet6_ops);
7075         unregister_netdevice_notifier(&ipv6_dev_notf);
7076 errlo:
7077         destroy_workqueue(addrconf_wq);
7078 out_nowq:
7079         unregister_pernet_subsys(&addrconf_ops);
7080 out_addrlabel:
7081         ipv6_addr_label_cleanup();
7082 out:
7083         return err;
7084 }
7085
7086 void addrconf_cleanup(void)
7087 {
7088         struct net_device *dev;
7089         int i;
7090
7091         unregister_netdevice_notifier(&ipv6_dev_notf);
7092         unregister_pernet_subsys(&addrconf_ops);
7093         ipv6_addr_label_cleanup();
7094
7095         rtnl_af_unregister(&inet6_ops);
7096
7097         rtnl_lock();
7098
7099         /* clean dev list */
7100         for_each_netdev(&init_net, dev) {
7101                 if (__in6_dev_get(dev) == NULL)
7102                         continue;
7103                 addrconf_ifdown(dev, 1);
7104         }
7105         addrconf_ifdown(init_net.loopback_dev, 2);
7106
7107         /*
7108          *      Check hash table.
7109          */
7110         spin_lock_bh(&addrconf_hash_lock);
7111         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7112                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7113         spin_unlock_bh(&addrconf_hash_lock);
7114         cancel_delayed_work(&addr_chk_work);
7115         rtnl_unlock();
7116
7117         destroy_workqueue(addrconf_wq);
7118 }