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/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/inet.h>
  50 #include <linux/in6.h>
  51 #include <linux/netdevice.h>
  52 #include <linux/if_addr.h>
  53 #include <linux/if_arp.h>
  54 #include <linux/if_arcnet.h>
  55 #include <linux/if_infiniband.h>
  56 #include <linux/route.h>
  57 #include <linux/inetdevice.h>
  58 #include <linux/init.h>
  59 #include <linux/slab.h>
  60 #ifdef CONFIG_SYSCTL
  61 #include <linux/sysctl.h>
  62 #endif
  63 #include <linux/capability.h>
  64 #include <linux/delay.h>
  65 #include <linux/notifier.h>
  66 #include <linux/string.h>
  67 #include <linux/hash.h>
  68
  69 #include <net/net_namespace.h>
  70 #include <net/sock.h>
  71 #include <net/snmp.h>
  72
  73 #include <net/af_ieee802154.h>
  74 #include <net/firewire.h>
  75 #include <net/ipv6.h>
  76 #include <net/protocol.h>
  77 #include <net/ndisc.h>
  78 #include <net/ip6_route.h>
  79 #include <net/addrconf.h>
  80 #include <net/tcp.h>
  81 #include <net/ip.h>
  82 #include <net/netlink.h>
  83 #include <net/pkt_sched.h>
  84 #include <linux/if_tunnel.h>
  85 #include <linux/rtnetlink.h>
  86 #include <linux/netconf.h>
  87 #include <linux/random.h>
  88 #include <linux/uaccess.h>
  89 #include <asm/unaligned.h>
  90
  91 #include <linux/proc_fs.h>
  92 #include <linux/seq_file.h>
  93 #include <linux/export.h>
  94
  95 /* Set to 3 to get tracing... */
  96 #define ACONF_DEBUG 2
  97
  98 #if ACONF_DEBUG >= 3
  99 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
 100 #else
 101 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
 102 #endif
 103
 104 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 105
 106 #define IPV6_MAX_STRLEN \
 107         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
 108
 109 static inline u32 cstamp_delta(unsigned long cstamp)
 110 {
 111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 112 }
 113
 114 #ifdef CONFIG_SYSCTL
 115 static int addrconf_sysctl_register(struct inet6_dev *idev);
 116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 117 #else
 118 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
 119 {
 120         return 0;
 121 }
 122
 123 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 124 {
 125 }
 126 #endif
 127
 128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
 129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 130 static void ipv6_regen_rndid(unsigned long data);
 131
 132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 133 static int ipv6_count_addresses(struct inet6_dev *idev);
 134 static int ipv6_generate_stable_address(struct in6_addr *addr,
 135                                         u8 dad_count,
 136                                         const struct inet6_dev *idev);
 137
 138 /*
 139  *      Configured unicast address hash table
 140  */
 141 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 142 static DEFINE_SPINLOCK(addrconf_hash_lock);
 143
 144 static void addrconf_verify(void);
 145 static void addrconf_verify_rtnl(void);
 146 static void addrconf_verify_work(struct work_struct *);
 147
 148 static struct workqueue_struct *addrconf_wq;
 149 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
 150
 151 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 152 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 153
 154 static void addrconf_type_change(struct net_device *dev,
 155                                  unsigned long event);
 156 static int addrconf_ifdown(struct net_device *dev, int how);
 157
 158 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
 159                                                   int plen,
 160                                                   const struct net_device *dev,
 161                                                   u32 flags, u32 noflags);
 162
 163 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 164 static void addrconf_dad_work(struct work_struct *w);
 165 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 166 static void addrconf_dad_run(struct inet6_dev *idev);
 167 static void addrconf_rs_timer(unsigned long data);
 168 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 169 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 170
 171 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 172                                 struct prefix_info *pinfo);
 173 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 174                                struct net_device *dev);
 175
 176 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 177         .forwarding             = 0,
 178         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 179         .mtu6                   = IPV6_MIN_MTU,
 180         .accept_ra              = 1,
 181         .accept_redirects       = 1,
 182         .autoconf               = 1,
 183         .force_mld_version      = 0,
 184         .mldv1_unsolicited_report_interval = 10 * HZ,
 185         .mldv2_unsolicited_report_interval = HZ,
 186         .dad_transmits          = 1,
 187         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 188         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 189         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 190         .use_tempaddr           = 0,
 191         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 192         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 193         .regen_max_retry        = REGEN_MAX_RETRY,
 194         .max_desync_factor      = MAX_DESYNC_FACTOR,
 195         .max_addresses          = IPV6_MAX_ADDRESSES,
 196         .accept_ra_defrtr       = 1,
 197         .accept_ra_from_local   = 0,
 198         .accept_ra_min_hop_limit= 1,
 199         .accept_ra_pinfo        = 1,
 200 #ifdef CONFIG_IPV6_ROUTER_PREF
 201         .accept_ra_rtr_pref     = 1,
 202         .rtr_probe_interval     = 60 * HZ,
 203 #ifdef CONFIG_IPV6_ROUTE_INFO
 204         .accept_ra_rt_info_max_plen = 0,
 205 #endif
 206 #endif
 207         .proxy_ndp              = 0,
 208         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 209         .disable_ipv6           = 0,
 210         .accept_dad             = 1,
 211         .suppress_frag_ndisc    = 1,
 212         .accept_ra_mtu          = 1,
 213         .stable_secret          = {
 214                 .initialized = false,
 215         },
 216         .use_oif_addrs_only     = 0,
 217 };
 218
 219 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 220         .forwarding             = 0,
 221         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 222         .mtu6                   = IPV6_MIN_MTU,
 223         .accept_ra              = 1,
 224         .accept_redirects       = 1,
 225         .autoconf               = 1,
 226         .force_mld_version      = 0,
 227         .mldv1_unsolicited_report_interval = 10 * HZ,
 228         .mldv2_unsolicited_report_interval = HZ,
 229         .dad_transmits          = 1,
 230         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 231         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 232         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 233         .use_tempaddr           = 0,
 234         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 235         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 236         .regen_max_retry        = REGEN_MAX_RETRY,
 237         .max_desync_factor      = MAX_DESYNC_FACTOR,
 238         .max_addresses          = IPV6_MAX_ADDRESSES,
 239         .accept_ra_defrtr       = 1,
 240         .accept_ra_from_local   = 0,
 241         .accept_ra_min_hop_limit= 1,
 242         .accept_ra_pinfo        = 1,
 243 #ifdef CONFIG_IPV6_ROUTER_PREF
 244         .accept_ra_rtr_pref     = 1,
 245         .rtr_probe_interval     = 60 * HZ,
 246 #ifdef CONFIG_IPV6_ROUTE_INFO
 247         .accept_ra_rt_info_max_plen = 0,
 248 #endif
 249 #endif
 250         .proxy_ndp              = 0,
 251         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 252         .disable_ipv6           = 0,
 253         .accept_dad             = 1,
 254         .suppress_frag_ndisc    = 1,
 255         .accept_ra_mtu          = 1,
 256         .stable_secret          = {
 257                 .initialized = false,
 258         },
 259         .use_oif_addrs_only     = 0,
 260 };
 261
 262 /* Check if a valid qdisc is available */
 263 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 264 {
 265         return !qdisc_tx_is_noop(dev);
 266 }
 267
 268 static void addrconf_del_rs_timer(struct inet6_dev *idev)
 269 {
 270         if (del_timer(&idev->rs_timer))
 271                 __in6_dev_put(idev);
 272 }
 273
 274 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
 275 {
 276         if (cancel_delayed_work(&ifp->dad_work))
 277                 __in6_ifa_put(ifp);
 278 }
 279
 280 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
 281                                   unsigned long when)
 282 {
 283         if (!timer_pending(&idev->rs_timer))
 284                 in6_dev_hold(idev);
 285         mod_timer(&idev->rs_timer, jiffies + when);
 286 }
 287
 288 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
 289                                    unsigned long delay)
 290 {
 291         if (!delayed_work_pending(&ifp->dad_work))
 292                 in6_ifa_hold(ifp);
 293         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
 294 }
 295
 296 static int snmp6_alloc_dev(struct inet6_dev *idev)
 297 {
 298         int i;
 299
 300         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
 301         if (!idev->stats.ipv6)
 302                 goto err_ip;
 303
 304         for_each_possible_cpu(i) {
 305                 struct ipstats_mib *addrconf_stats;
 306                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
 307                 u64_stats_init(&addrconf_stats->syncp);
 308         }
 309
 310
 311         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 312                                         GFP_KERNEL);
 313         if (!idev->stats.icmpv6dev)
 314                 goto err_icmp;
 315         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 316                                            GFP_KERNEL);
 317         if (!idev->stats.icmpv6msgdev)
 318                 goto err_icmpmsg;
 319
 320         return 0;
 321
 322 err_icmpmsg:
 323         kfree(idev->stats.icmpv6dev);
 324 err_icmp:
 325         free_percpu(idev->stats.ipv6);
 326 err_ip:
 327         return -ENOMEM;
 328 }
 329
 330 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 331 {
 332         struct inet6_dev *ndev;
 333         int err = -ENOMEM;
 334
 335         ASSERT_RTNL();
 336
 337         if (dev->mtu < IPV6_MIN_MTU)
 338                 return ERR_PTR(-EINVAL);
 339
 340         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 341         if (!ndev)
 342                 return ERR_PTR(err);
 343
 344         rwlock_init(&ndev->lock);
 345         ndev->dev = dev;
 346         INIT_LIST_HEAD(&ndev->addr_list);
 347         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
 348                     (unsigned long)ndev);
 349         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 350         ndev->cnf.mtu6 = dev->mtu;
 351         ndev->cnf.sysctl = NULL;
 352         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 353         if (!ndev->nd_parms) {
 354                 kfree(ndev);
 355                 return ERR_PTR(err);
 356         }
 357         if (ndev->cnf.forwarding)
 358                 dev_disable_lro(dev);
 359         /* We refer to the device */
 360         dev_hold(dev);
 361
 362         if (snmp6_alloc_dev(ndev) < 0) {
 363                 ADBG(KERN_WARNING
 364                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 365                         __func__, dev->name);
 366                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 367                 dev_put(dev);
 368                 kfree(ndev);
 369                 return ERR_PTR(err);
 370         }
 371
 372         if (snmp6_register_dev(ndev) < 0) {
 373                 ADBG(KERN_WARNING
 374                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 375                         __func__, dev->name);
 376                 goto err_release;
 377         }
 378
 379         /* One reference from device.  We must do this before
 380          * we invoke __ipv6_regen_rndid().
 381          */
 382         in6_dev_hold(ndev);
 383
 384         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 385                 ndev->cnf.accept_dad = -1;
 386
 387 #if IS_ENABLED(CONFIG_IPV6_SIT)
 388         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 389                 pr_info("%s: Disabled Multicast RS\n", dev->name);
 390                 ndev->cnf.rtr_solicits = 0;
 391         }
 392 #endif
 393
 394         INIT_LIST_HEAD(&ndev->tempaddr_list);
 395         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 396         if ((dev->flags&IFF_LOOPBACK) ||
 397             dev->type == ARPHRD_TUNNEL ||
 398             dev->type == ARPHRD_TUNNEL6 ||
 399             dev->type == ARPHRD_SIT ||
 400             dev->type == ARPHRD_NONE) {
 401                 ndev->cnf.use_tempaddr = -1;
 402         } else {
 403                 in6_dev_hold(ndev);
 404                 ipv6_regen_rndid((unsigned long) ndev);
 405         }
 406
 407         ndev->token = in6addr_any;
 408
 409         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 410                 ndev->if_flags |= IF_READY;
 411
 412         ipv6_mc_init_dev(ndev);
 413         ndev->tstamp = jiffies;
 414         err = addrconf_sysctl_register(ndev);
 415         if (err) {
 416                 ipv6_mc_destroy_dev(ndev);
 417                 del_timer(&ndev->regen_timer);
 418                 goto err_release;
 419         }
 420         /* protected by rtnl_lock */
 421         rcu_assign_pointer(dev->ip6_ptr, ndev);
 422
 423         /* Join interface-local all-node multicast group */
 424         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
 425
 426         /* Join all-node multicast group */
 427         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 428
 429         /* Join all-router multicast group if forwarding is set */
 430         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 431                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 432
 433         return ndev;
 434
 435 err_release:
 436         neigh_parms_release(&nd_tbl, ndev->nd_parms);
 437         ndev->dead = 1;
 438         in6_dev_finish_destroy(ndev);
 439         return ERR_PTR(err);
 440 }
 441
 442 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 443 {
 444         struct inet6_dev *idev;
 445
 446         ASSERT_RTNL();
 447
 448         idev = __in6_dev_get(dev);
 449         if (!idev) {
 450                 idev = ipv6_add_dev(dev);
 451                 if (IS_ERR(idev))
 452                         return NULL;
 453         }
 454
 455         if (dev->flags&IFF_UP)
 456                 ipv6_mc_up(idev);
 457         return idev;
 458 }
 459
 460 static int inet6_netconf_msgsize_devconf(int type)
 461 {
 462         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
 463                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
 464
 465         /* type -1 is used for ALL */
 466         if (type == -1 || type == NETCONFA_FORWARDING)
 467                 size += nla_total_size(4);
 468 #ifdef CONFIG_IPV6_MROUTE
 469         if (type == -1 || type == NETCONFA_MC_FORWARDING)
 470                 size += nla_total_size(4);
 471 #endif
 472         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
 473                 size += nla_total_size(4);
 474
 475         return size;
 476 }
 477
 478 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
 479                                       struct ipv6_devconf *devconf, u32 portid,
 480                                       u32 seq, int event, unsigned int flags,
 481                                       int type)
 482 {
 483         struct nlmsghdr  *nlh;
 484         struct netconfmsg *ncm;
 485
 486         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
 487                         flags);
 488         if (!nlh)
 489                 return -EMSGSIZE;
 490
 491         ncm = nlmsg_data(nlh);
 492         ncm->ncm_family = AF_INET6;
 493
 494         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
 495                 goto nla_put_failure;
 496
 497         /* type -1 is used for ALL */
 498         if ((type == -1 || type == NETCONFA_FORWARDING) &&
 499             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
 500                 goto nla_put_failure;
 501 #ifdef CONFIG_IPV6_MROUTE
 502         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
 503             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
 504                         devconf->mc_forwarding) < 0)
 505                 goto nla_put_failure;
 506 #endif
 507         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
 508             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
 509                 goto nla_put_failure;
 510
 511         nlmsg_end(skb, nlh);
 512         return 0;
 513
 514 nla_put_failure:
 515         nlmsg_cancel(skb, nlh);
 516         return -EMSGSIZE;
 517 }
 518
 519 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
 520                                   struct ipv6_devconf *devconf)
 521 {
 522         struct sk_buff *skb;
 523         int err = -ENOBUFS;
 524
 525         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
 526         if (!skb)
 527                 goto errout;
 528
 529         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
 530                                          RTM_NEWNETCONF, 0, type);
 531         if (err < 0) {
 532                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 533                 WARN_ON(err == -EMSGSIZE);
 534                 kfree_skb(skb);
 535                 goto errout;
 536         }
 537         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
 538         return;
 539 errout:
 540         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
 541 }
 542
 543 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
 544         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
 545         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
 546         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
 547 };
 548
 549 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
 550                                      struct nlmsghdr *nlh)
 551 {
 552         struct net *net = sock_net(in_skb->sk);
 553         struct nlattr *tb[NETCONFA_MAX+1];
 554         struct netconfmsg *ncm;
 555         struct sk_buff *skb;
 556         struct ipv6_devconf *devconf;
 557         struct inet6_dev *in6_dev;
 558         struct net_device *dev;
 559         int ifindex;
 560         int err;
 561
 562         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
 563                           devconf_ipv6_policy);
 564         if (err < 0)
 565                 goto errout;
 566
 567         err = EINVAL;
 568         if (!tb[NETCONFA_IFINDEX])
 569                 goto errout;
 570
 571         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
 572         switch (ifindex) {
 573         case NETCONFA_IFINDEX_ALL:
 574                 devconf = net->ipv6.devconf_all;
 575                 break;
 576         case NETCONFA_IFINDEX_DEFAULT:
 577                 devconf = net->ipv6.devconf_dflt;
 578                 break;
 579         default:
 580                 dev = __dev_get_by_index(net, ifindex);
 581                 if (!dev)
 582                         goto errout;
 583                 in6_dev = __in6_dev_get(dev);
 584                 if (!in6_dev)
 585                         goto errout;
 586                 devconf = &in6_dev->cnf;
 587                 break;
 588         }
 589
 590         err = -ENOBUFS;
 591         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
 592         if (!skb)
 593                 goto errout;
 594
 595         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
 596                                          NETLINK_CB(in_skb).portid,
 597                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
 598                                          -1);
 599         if (err < 0) {
 600                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
 601                 WARN_ON(err == -EMSGSIZE);
 602                 kfree_skb(skb);
 603                 goto errout;
 604         }
 605         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
 606 errout:
 607         return err;
 608 }
 609
 610 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
 611                                       struct netlink_callback *cb)
 612 {
 613         struct net *net = sock_net(skb->sk);
 614         int h, s_h;
 615         int idx, s_idx;
 616         struct net_device *dev;
 617         struct inet6_dev *idev;
 618         struct hlist_head *head;
 619
 620         s_h = cb->args[0];
 621         s_idx = idx = cb->args[1];
 622
 623         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 624                 idx = 0;
 625                 head = &net->dev_index_head[h];
 626                 rcu_read_lock();
 627                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
 628                           net->dev_base_seq;
 629                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
 630                         if (idx < s_idx)
 631                                 goto cont;
 632                         idev = __in6_dev_get(dev);
 633                         if (!idev)
 634                                 goto cont;
 635
 636                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
 637                                                        &idev->cnf,
 638                                                        NETLINK_CB(cb->skb).portid,
 639                                                        cb->nlh->nlmsg_seq,
 640                                                        RTM_NEWNETCONF,
 641                                                        NLM_F_MULTI,
 642                                                        -1) < 0) {
 643                                 rcu_read_unlock();
 644                                 goto done;
 645                         }
 646                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 647 cont:
 648                         idx++;
 649                 }
 650                 rcu_read_unlock();
 651         }
 652         if (h == NETDEV_HASHENTRIES) {
 653                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
 654                                                net->ipv6.devconf_all,
 655                                                NETLINK_CB(cb->skb).portid,
 656                                                cb->nlh->nlmsg_seq,
 657                                                RTM_NEWNETCONF, NLM_F_MULTI,
 658                                                -1) < 0)
 659                         goto done;
 660                 else
 661                         h++;
 662         }
 663         if (h == NETDEV_HASHENTRIES + 1) {
 664                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
 665                                                net->ipv6.devconf_dflt,
 666                                                NETLINK_CB(cb->skb).portid,
 667                                                cb->nlh->nlmsg_seq,
 668                                                RTM_NEWNETCONF, NLM_F_MULTI,
 669                                                -1) < 0)
 670                         goto done;
 671                 else
 672                         h++;
 673         }
 674 done:
 675         cb->args[0] = h;
 676         cb->args[1] = idx;
 677
 678         return skb->len;
 679 }
 680
 681 #ifdef CONFIG_SYSCTL
 682 static void dev_forward_change(struct inet6_dev *idev)
 683 {
 684         struct net_device *dev;
 685         struct inet6_ifaddr *ifa;
 686
 687         if (!idev)
 688                 return;
 689         dev = idev->dev;
 690         if (idev->cnf.forwarding)
 691                 dev_disable_lro(dev);
 692         if (dev->flags & IFF_MULTICAST) {
 693                 if (idev->cnf.forwarding) {
 694                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 695                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
 696                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
 697                 } else {
 698                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 699                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
 700                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
 701                 }
 702         }
 703
 704         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 705                 if (ifa->flags&IFA_F_TENTATIVE)
 706                         continue;
 707                 if (idev->cnf.forwarding)
 708                         addrconf_join_anycast(ifa);
 709                 else
 710                         addrconf_leave_anycast(ifa);
 711         }
 712         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
 713                                      dev->ifindex, &idev->cnf);
 714 }
 715
 716
 717 static void addrconf_forward_change(struct net *net, __s32 newf)
 718 {
 719         struct net_device *dev;
 720         struct inet6_dev *idev;
 721
 722         for_each_netdev(net, dev) {
 723                 idev = __in6_dev_get(dev);
 724                 if (idev) {
 725                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 726                         idev->cnf.forwarding = newf;
 727                         if (changed)
 728                                 dev_forward_change(idev);
 729                 }
 730         }
 731 }
 732
 733 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 734 {
 735         struct net *net;
 736         int old;
 737
 738         if (!rtnl_trylock())
 739                 return restart_syscall();
 740
 741         net = (struct net *)table->extra2;
 742         old = *p;
 743         *p = newf;
 744
 745         if (p == &net->ipv6.devconf_dflt->forwarding) {
 746                 if ((!newf) ^ (!old))
 747                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 748                                                      NETCONFA_IFINDEX_DEFAULT,
 749                                                      net->ipv6.devconf_dflt);
 750                 rtnl_unlock();
 751                 return 0;
 752         }
 753
 754         if (p == &net->ipv6.devconf_all->forwarding) {
 755                 net->ipv6.devconf_dflt->forwarding = newf;
 756                 addrconf_forward_change(net, newf);
 757                 if ((!newf) ^ (!old))
 758                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
 759                                                      NETCONFA_IFINDEX_ALL,
 760                                                      net->ipv6.devconf_all);
 761         } else if ((!newf) ^ (!old))
 762                 dev_forward_change((struct inet6_dev *)table->extra1);
 763         rtnl_unlock();
 764
 765         if (newf)
 766                 rt6_purge_dflt_routers(net);
 767         return 1;
 768 }
 769 #endif
 770
 771 /* Nobody refers to this ifaddr, destroy it */
 772 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 773 {
 774         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 775
 776 #ifdef NET_REFCNT_DEBUG
 777         pr_debug("%s\n", __func__);
 778 #endif
 779
 780         in6_dev_put(ifp->idev);
 781
 782         if (cancel_delayed_work(&ifp->dad_work))
 783                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
 784                           ifp);
 785
 786         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 787                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 788                 return;
 789         }
 790         ip6_rt_put(ifp->rt);
 791
 792         kfree_rcu(ifp, rcu);
 793 }
 794
 795 static void
 796 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 797 {
 798         struct list_head *p;
 799         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 800
 801         /*
 802          * Each device address list is sorted in order of scope -
 803          * global before linklocal.
 804          */
 805         list_for_each(p, &idev->addr_list) {
 806                 struct inet6_ifaddr *ifa
 807                         = list_entry(p, struct inet6_ifaddr, if_list);
 808                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 809                         break;
 810         }
 811
 812         list_add_tail(&ifp->if_list, p);
 813 }
 814
 815 static u32 inet6_addr_hash(const struct in6_addr *addr)
 816 {
 817         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
 818 }
 819
 820 /* On success it returns ifp with increased reference count */
 821
 822 static struct inet6_ifaddr *
 823 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
 824               const struct in6_addr *peer_addr, int pfxlen,
 825               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
 826 {
 827         struct inet6_ifaddr *ifa = NULL;
 828         struct rt6_info *rt;
 829         unsigned int hash;
 830         int err = 0;
 831         int addr_type = ipv6_addr_type(addr);
 832
 833         if (addr_type == IPV6_ADDR_ANY ||
 834             addr_type & IPV6_ADDR_MULTICAST ||
 835             (!(idev->dev->flags & IFF_LOOPBACK) &&
 836              addr_type & IPV6_ADDR_LOOPBACK))
 837                 return ERR_PTR(-EADDRNOTAVAIL);
 838
 839         rcu_read_lock_bh();
 840         if (idev->dead) {
 841                 err = -ENODEV;                  /*XXX*/
 842                 goto out2;
 843         }
 844
 845         if (idev->cnf.disable_ipv6) {
 846                 err = -EACCES;
 847                 goto out2;
 848         }
 849
 850         spin_lock(&addrconf_hash_lock);
 851
 852         /* Ignore adding duplicate addresses on an interface */
 853         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 854                 ADBG("ipv6_add_addr: already assigned\n");
 855                 err = -EEXIST;
 856                 goto out;
 857         }
 858
 859         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 860
 861         if (!ifa) {
 862                 ADBG("ipv6_add_addr: malloc failed\n");
 863                 err = -ENOBUFS;
 864                 goto out;
 865         }
 866
 867         rt = addrconf_dst_alloc(idev, addr, false);
 868         if (IS_ERR(rt)) {
 869                 err = PTR_ERR(rt);
 870                 goto out;
 871         }
 872
 873         neigh_parms_data_state_setall(idev->nd_parms);
 874
 875         ifa->addr = *addr;
 876         if (peer_addr)
 877                 ifa->peer_addr = *peer_addr;
 878
 879         spin_lock_init(&ifa->lock);
 880         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
 881         INIT_HLIST_NODE(&ifa->addr_lst);
 882         ifa->scope = scope;
 883         ifa->prefix_len = pfxlen;
 884         ifa->flags = flags | IFA_F_TENTATIVE;
 885         ifa->valid_lft = valid_lft;
 886         ifa->prefered_lft = prefered_lft;
 887         ifa->cstamp = ifa->tstamp = jiffies;
 888         ifa->tokenized = false;
 889
 890         ifa->rt = rt;
 891
 892         ifa->idev = idev;
 893         in6_dev_hold(idev);
 894         /* For caller */
 895         in6_ifa_hold(ifa);
 896
 897         /* Add to big hash table */
 898         hash = inet6_addr_hash(addr);
 899
 900         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 901         spin_unlock(&addrconf_hash_lock);
 902
 903         write_lock(&idev->lock);
 904         /* Add to inet6_dev unicast addr list. */
 905         ipv6_link_dev_addr(idev, ifa);
 906
 907         if (ifa->flags&IFA_F_TEMPORARY) {
 908                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 909                 in6_ifa_hold(ifa);
 910         }
 911
 912         in6_ifa_hold(ifa);
 913         write_unlock(&idev->lock);
 914 out2:
 915         rcu_read_unlock_bh();
 916
 917         if (likely(err == 0))
 918                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
 919         else {
 920                 kfree(ifa);
 921                 ifa = ERR_PTR(err);
 922         }
 923
 924         return ifa;
 925 out:
 926         spin_unlock(&addrconf_hash_lock);
 927         goto out2;
 928 }
 929
 930 enum cleanup_prefix_rt_t {
 931         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
 932         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
 933         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
 934 };
 935
 936 /*
 937  * Check, whether the prefix for ifp would still need a prefix route
 938  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
 939  * constants.
 940  *
 941  * 1) we don't purge prefix if address was not permanent.
 942  *    prefix is managed by its own lifetime.
 943  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
 944  * 3) if there are no addresses, delete prefix.
 945  * 4) if there are still other permanent address(es),
 946  *    corresponding prefix is still permanent.
 947  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
 948  *    don't purge the prefix, assume user space is managing it.
 949  * 6) otherwise, update prefix lifetime to the
 950  *    longest valid lifetime among the corresponding
 951  *    addresses on the device.
 952  *    Note: subsequent RA will update lifetime.
 953  **/
 954 static enum cleanup_prefix_rt_t
 955 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
 956 {
 957         struct inet6_ifaddr *ifa;
 958         struct inet6_dev *idev = ifp->idev;
 959         unsigned long lifetime;
 960         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
 961
 962         *expires = jiffies;
 963
 964         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 965                 if (ifa == ifp)
 966                         continue;
 967                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 968                                        ifp->prefix_len))
 969                         continue;
 970                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
 971                         return CLEANUP_PREFIX_RT_NOP;
 972
 973                 action = CLEANUP_PREFIX_RT_EXPIRE;
 974
 975                 spin_lock(&ifa->lock);
 976
 977                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 978                 /*
 979                  * Note: Because this address is
 980                  * not permanent, lifetime <
 981                  * LONG_MAX / HZ here.
 982                  */
 983                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
 984                         *expires = ifa->tstamp + lifetime * HZ;
 985                 spin_unlock(&ifa->lock);
 986         }
 987
 988         return action;
 989 }
 990
 991 static void
 992 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
 993 {
 994         struct rt6_info *rt;
 995
 996         rt = addrconf_get_prefix_route(&ifp->addr,
 997                                        ifp->prefix_len,
 998                                        ifp->idev->dev,
 999                                        0, RTF_GATEWAY | RTF_DEFAULT);
1000         if (rt) {
1001                 if (del_rt)
1002                         ip6_del_rt(rt);
1003                 else {
1004                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1005                                 rt6_set_expires(rt, expires);
1006                         ip6_rt_put(rt);
1007                 }
1008         }
1009 }
1010
1011
1012 /* This function wants to get referenced ifp and releases it before return */
1013
1014 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1015 {
1016         int state;
1017         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1018         unsigned long expires;
1019
1020         ASSERT_RTNL();
1021
1022         spin_lock_bh(&ifp->lock);
1023         state = ifp->state;
1024         ifp->state = INET6_IFADDR_STATE_DEAD;
1025         spin_unlock_bh(&ifp->lock);
1026
1027         if (state == INET6_IFADDR_STATE_DEAD)
1028                 goto out;
1029
1030         spin_lock_bh(&addrconf_hash_lock);
1031         hlist_del_init_rcu(&ifp->addr_lst);
1032         spin_unlock_bh(&addrconf_hash_lock);
1033
1034         write_lock_bh(&ifp->idev->lock);
1035
1036         if (ifp->flags&IFA_F_TEMPORARY) {
1037                 list_del(&ifp->tmp_list);
1038                 if (ifp->ifpub) {
1039                         in6_ifa_put(ifp->ifpub);
1040                         ifp->ifpub = NULL;
1041                 }
1042                 __in6_ifa_put(ifp);
1043         }
1044
1045         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1046                 action = check_cleanup_prefix_route(ifp, &expires);
1047
1048         list_del_init(&ifp->if_list);
1049         __in6_ifa_put(ifp);
1050
1051         write_unlock_bh(&ifp->idev->lock);
1052
1053         addrconf_del_dad_work(ifp);
1054
1055         ipv6_ifa_notify(RTM_DELADDR, ifp);
1056
1057         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1058
1059         if (action != CLEANUP_PREFIX_RT_NOP) {
1060                 cleanup_prefix_route(ifp, expires,
1061                         action == CLEANUP_PREFIX_RT_DEL);
1062         }
1063
1064         /* clean up prefsrc entries */
1065         rt6_remove_prefsrc(ifp);
1066 out:
1067         in6_ifa_put(ifp);
1068 }
1069
1070 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1071 {
1072         struct inet6_dev *idev = ifp->idev;
1073         struct in6_addr addr, *tmpaddr;
1074         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1075         unsigned long regen_advance;
1076         int tmp_plen;
1077         int ret = 0;
1078         u32 addr_flags;
1079         unsigned long now = jiffies;
1080
1081         write_lock_bh(&idev->lock);
1082         if (ift) {
1083                 spin_lock_bh(&ift->lock);
1084                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1085                 spin_unlock_bh(&ift->lock);
1086                 tmpaddr = &addr;
1087         } else {
1088                 tmpaddr = NULL;
1089         }
1090 retry:
1091         in6_dev_hold(idev);
1092         if (idev->cnf.use_tempaddr <= 0) {
1093                 write_unlock_bh(&idev->lock);
1094                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1095                 in6_dev_put(idev);
1096                 ret = -1;
1097                 goto out;
1098         }
1099         spin_lock_bh(&ifp->lock);
1100         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1101                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1102                 spin_unlock_bh(&ifp->lock);
1103                 write_unlock_bh(&idev->lock);
1104                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1105                         __func__);
1106                 in6_dev_put(idev);
1107                 ret = -1;
1108                 goto out;
1109         }
1110         in6_ifa_hold(ifp);
1111         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1112         __ipv6_try_regen_rndid(idev, tmpaddr);
1113         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1114         age = (now - ifp->tstamp) / HZ;
1115         tmp_valid_lft = min_t(__u32,
1116                               ifp->valid_lft,
1117                               idev->cnf.temp_valid_lft + age);
1118         tmp_prefered_lft = min_t(__u32,
1119                                  ifp->prefered_lft,
1120                                  idev->cnf.temp_prefered_lft + age -
1121                                  idev->cnf.max_desync_factor);
1122         tmp_plen = ifp->prefix_len;
1123         tmp_tstamp = ifp->tstamp;
1124         spin_unlock_bh(&ifp->lock);
1125
1126         regen_advance = idev->cnf.regen_max_retry *
1127                         idev->cnf.dad_transmits *
1128                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1129         write_unlock_bh(&idev->lock);
1130
1131         /* A temporary address is created only if this calculated Preferred
1132          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1133          * an implementation must not create a temporary address with a zero
1134          * Preferred Lifetime.
1135          * Use age calculation as in addrconf_verify to avoid unnecessary
1136          * temporary addresses being generated.
1137          */
1138         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1139         if (tmp_prefered_lft <= regen_advance + age) {
1140                 in6_ifa_put(ifp);
1141                 in6_dev_put(idev);
1142                 ret = -1;
1143                 goto out;
1144         }
1145
1146         addr_flags = IFA_F_TEMPORARY;
1147         /* set in addrconf_prefix_rcv() */
1148         if (ifp->flags & IFA_F_OPTIMISTIC)
1149                 addr_flags |= IFA_F_OPTIMISTIC;
1150
1151         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1152                             ipv6_addr_scope(&addr), addr_flags,
1153                             tmp_valid_lft, tmp_prefered_lft);
1154         if (IS_ERR(ift)) {
1155                 in6_ifa_put(ifp);
1156                 in6_dev_put(idev);
1157                 pr_info("%s: retry temporary address regeneration\n", __func__);
1158                 tmpaddr = &addr;
1159                 write_lock_bh(&idev->lock);
1160                 goto retry;
1161         }
1162
1163         spin_lock_bh(&ift->lock);
1164         ift->ifpub = ifp;
1165         ift->cstamp = now;
1166         ift->tstamp = tmp_tstamp;
1167         spin_unlock_bh(&ift->lock);
1168
1169         addrconf_dad_start(ift);
1170         in6_ifa_put(ift);
1171         in6_dev_put(idev);
1172 out:
1173         return ret;
1174 }
1175
1176 /*
1177  *      Choose an appropriate source address (RFC3484)
1178  */
1179 enum {
1180         IPV6_SADDR_RULE_INIT = 0,
1181         IPV6_SADDR_RULE_LOCAL,
1182         IPV6_SADDR_RULE_SCOPE,
1183         IPV6_SADDR_RULE_PREFERRED,
1184 #ifdef CONFIG_IPV6_MIP6
1185         IPV6_SADDR_RULE_HOA,
1186 #endif
1187         IPV6_SADDR_RULE_OIF,
1188         IPV6_SADDR_RULE_LABEL,
1189         IPV6_SADDR_RULE_PRIVACY,
1190         IPV6_SADDR_RULE_ORCHID,
1191         IPV6_SADDR_RULE_PREFIX,
1192 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1193         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1194 #endif
1195         IPV6_SADDR_RULE_MAX
1196 };
1197
1198 struct ipv6_saddr_score {
1199         int                     rule;
1200         int                     addr_type;
1201         struct inet6_ifaddr     *ifa;
1202         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1203         int                     scopedist;
1204         int                     matchlen;
1205 };
1206
1207 struct ipv6_saddr_dst {
1208         const struct in6_addr *addr;
1209         int ifindex;
1210         int scope;
1211         int label;
1212         unsigned int prefs;
1213 };
1214
1215 static inline int ipv6_saddr_preferred(int type)
1216 {
1217         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1218                 return 1;
1219         return 0;
1220 }
1221
1222 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1223 {
1224 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1225         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1226 #else
1227         return false;
1228 #endif
1229 }
1230
1231 static int ipv6_get_saddr_eval(struct net *net,
1232                                struct ipv6_saddr_score *score,
1233                                struct ipv6_saddr_dst *dst,
1234                                int i)
1235 {
1236         int ret;
1237
1238         if (i <= score->rule) {
1239                 switch (i) {
1240                 case IPV6_SADDR_RULE_SCOPE:
1241                         ret = score->scopedist;
1242                         break;
1243                 case IPV6_SADDR_RULE_PREFIX:
1244                         ret = score->matchlen;
1245                         break;
1246                 default:
1247                         ret = !!test_bit(i, score->scorebits);
1248                 }
1249                 goto out;
1250         }
1251
1252         switch (i) {
1253         case IPV6_SADDR_RULE_INIT:
1254                 /* Rule 0: remember if hiscore is not ready yet */
1255                 ret = !!score->ifa;
1256                 break;
1257         case IPV6_SADDR_RULE_LOCAL:
1258                 /* Rule 1: Prefer same address */
1259                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1260                 break;
1261         case IPV6_SADDR_RULE_SCOPE:
1262                 /* Rule 2: Prefer appropriate scope
1263                  *
1264                  *      ret
1265                  *       ^
1266                  *    -1 |  d 15
1267                  *    ---+--+-+---> scope
1268                  *       |
1269                  *       |             d is scope of the destination.
1270                  *  B-d  |  \
1271                  *       |   \      <- smaller scope is better if
1272                  *  B-15 |    \        if scope is enough for destination.
1273                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1274                  * d-C-1 | /
1275                  *       |/         <- greater is better
1276                  *   -C  /             if scope is not enough for destination.
1277                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1278                  *
1279                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1280                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1281                  * Assume B = 0 and we get C > 29.
1282                  */
1283                 ret = __ipv6_addr_src_scope(score->addr_type);
1284                 if (ret >= dst->scope)
1285                         ret = -ret;
1286                 else
1287                         ret -= 128;     /* 30 is enough */
1288                 score->scopedist = ret;
1289                 break;
1290         case IPV6_SADDR_RULE_PREFERRED:
1291             {
1292                 /* Rule 3: Avoid deprecated and optimistic addresses */
1293                 u8 avoid = IFA_F_DEPRECATED;
1294
1295                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1296                         avoid |= IFA_F_OPTIMISTIC;
1297                 ret = ipv6_saddr_preferred(score->addr_type) ||
1298                       !(score->ifa->flags & avoid);
1299                 break;
1300             }
1301 #ifdef CONFIG_IPV6_MIP6
1302         case IPV6_SADDR_RULE_HOA:
1303             {
1304                 /* Rule 4: Prefer home address */
1305                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1306                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1307                 break;
1308             }
1309 #endif
1310         case IPV6_SADDR_RULE_OIF:
1311                 /* Rule 5: Prefer outgoing interface */
1312                 ret = (!dst->ifindex ||
1313                        dst->ifindex == score->ifa->idev->dev->ifindex);
1314                 break;
1315         case IPV6_SADDR_RULE_LABEL:
1316                 /* Rule 6: Prefer matching label */
1317                 ret = ipv6_addr_label(net,
1318                                       &score->ifa->addr, score->addr_type,
1319                                       score->ifa->idev->dev->ifindex) == dst->label;
1320                 break;
1321         case IPV6_SADDR_RULE_PRIVACY:
1322             {
1323                 /* Rule 7: Prefer public address
1324                  * Note: prefer temporary address if use_tempaddr >= 2
1325                  */
1326                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1327                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1328                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1329                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1330                 break;
1331             }
1332         case IPV6_SADDR_RULE_ORCHID:
1333                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1334                  *          non-ORCHID vs non-ORCHID
1335                  */
1336                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1337                         ipv6_addr_orchid(dst->addr));
1338                 break;
1339         case IPV6_SADDR_RULE_PREFIX:
1340                 /* Rule 8: Use longest matching prefix */
1341                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1342                 if (ret > score->ifa->prefix_len)
1343                         ret = score->ifa->prefix_len;
1344                 score->matchlen = ret;
1345                 break;
1346 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1347         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1348                 /* Optimistic addresses still have lower precedence than other
1349                  * preferred addresses.
1350                  */
1351                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1352                 break;
1353 #endif
1354         default:
1355                 ret = 0;
1356         }
1357
1358         if (ret)
1359                 __set_bit(i, score->scorebits);
1360         score->rule = i;
1361 out:
1362         return ret;
1363 }
1364
1365 static int __ipv6_dev_get_saddr(struct net *net,
1366                                 struct ipv6_saddr_dst *dst,
1367                                 struct inet6_dev *idev,
1368                                 struct ipv6_saddr_score *scores,
1369                                 int hiscore_idx)
1370 {
1371         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1372
1373         read_lock_bh(&idev->lock);
1374         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1375                 int i;
1376
1377                 /*
1378                  * - Tentative Address (RFC2462 section 5.4)
1379                  *  - A tentative address is not considered
1380                  *    "assigned to an interface" in the traditional
1381                  *    sense, unless it is also flagged as optimistic.
1382                  * - Candidate Source Address (section 4)
1383                  *  - In any case, anycast addresses, multicast
1384                  *    addresses, and the unspecified address MUST
1385                  *    NOT be included in a candidate set.
1386                  */
1387                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1388                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1389                         continue;
1390
1391                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1392
1393                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1394                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1395                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1396                                             idev->dev->name);
1397                         continue;
1398                 }
1399
1400                 score->rule = -1;
1401                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1402
1403                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1404                         int minihiscore, miniscore;
1405
1406                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1407                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1408
1409                         if (minihiscore > miniscore) {
1410                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1411                                     score->scopedist > 0) {
1412                                         /*
1413                                          * special case:
1414                                          * each remaining entry
1415                                          * has too small (not enough)
1416                                          * scope, because ifa entries
1417                                          * are sorted by their scope
1418                                          * values.
1419                                          */
1420                                         goto out;
1421                                 }
1422                                 break;
1423                         } else if (minihiscore < miniscore) {
1424                                 if (hiscore->ifa)
1425                                         in6_ifa_put(hiscore->ifa);
1426
1427                                 in6_ifa_hold(score->ifa);
1428
1429                                 swap(hiscore, score);
1430                                 hiscore_idx = 1 - hiscore_idx;
1431
1432                                 /* restore our iterator */
1433                                 score->ifa = hiscore->ifa;
1434
1435                                 break;
1436                         }
1437                 }
1438         }
1439 out:
1440         read_unlock_bh(&idev->lock);
1441         return hiscore_idx;
1442 }
1443
1444 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1445                        const struct in6_addr *daddr, unsigned int prefs,
1446                        struct in6_addr *saddr)
1447 {
1448         struct ipv6_saddr_score scores[2], *hiscore;
1449         struct ipv6_saddr_dst dst;
1450         struct inet6_dev *idev;
1451         struct net_device *dev;
1452         int dst_type;
1453         bool use_oif_addr = false;
1454         int hiscore_idx = 0;
1455
1456         dst_type = __ipv6_addr_type(daddr);
1457         dst.addr = daddr;
1458         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1459         dst.scope = __ipv6_addr_src_scope(dst_type);
1460         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1461         dst.prefs = prefs;
1462
1463         scores[hiscore_idx].rule = -1;
1464         scores[hiscore_idx].ifa = NULL;
1465
1466         rcu_read_lock();
1467
1468         /* Candidate Source Address (section 4)
1469          *  - multicast and link-local destination address,
1470          *    the set of candidate source address MUST only
1471          *    include addresses assigned to interfaces
1472          *    belonging to the same link as the outgoing
1473          *    interface.
1474          * (- For site-local destination addresses, the
1475          *    set of candidate source addresses MUST only
1476          *    include addresses assigned to interfaces
1477          *    belonging to the same site as the outgoing
1478          *    interface.)
1479          *  - "It is RECOMMENDED that the candidate source addresses
1480          *    be the set of unicast addresses assigned to the
1481          *    interface that will be used to send to the destination
1482          *    (the 'outgoing' interface)." (RFC 6724)
1483          */
1484         if (dst_dev) {
1485                 idev = __in6_dev_get(dst_dev);
1486                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1487                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1488                     (idev && idev->cnf.use_oif_addrs_only)) {
1489                         use_oif_addr = true;
1490                 }
1491         }
1492
1493         if (use_oif_addr) {
1494                 if (idev)
1495                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1496         } else {
1497                 for_each_netdev_rcu(net, dev) {
1498                         idev = __in6_dev_get(dev);
1499                         if (!idev)
1500                                 continue;
1501                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1502                 }
1503         }
1504         rcu_read_unlock();
1505
1506         hiscore = &scores[hiscore_idx];
1507         if (!hiscore->ifa)
1508                 return -EADDRNOTAVAIL;
1509
1510         *saddr = hiscore->ifa->addr;
1511         in6_ifa_put(hiscore->ifa);
1512         return 0;
1513 }
1514 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1515
1516 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1517                       u32 banned_flags)
1518 {
1519         struct inet6_ifaddr *ifp;
1520         int err = -EADDRNOTAVAIL;
1521
1522         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1523                 if (ifp->scope > IFA_LINK)
1524                         break;
1525                 if (ifp->scope == IFA_LINK &&
1526                     !(ifp->flags & banned_flags)) {
1527                         *addr = ifp->addr;
1528                         err = 0;
1529                         break;
1530                 }
1531         }
1532         return err;
1533 }
1534
1535 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1536                     u32 banned_flags)
1537 {
1538         struct inet6_dev *idev;
1539         int err = -EADDRNOTAVAIL;
1540
1541         rcu_read_lock();
1542         idev = __in6_dev_get(dev);
1543         if (idev) {
1544                 read_lock_bh(&idev->lock);
1545                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1546                 read_unlock_bh(&idev->lock);
1547         }
1548         rcu_read_unlock();
1549         return err;
1550 }
1551
1552 static int ipv6_count_addresses(struct inet6_dev *idev)
1553 {
1554         int cnt = 0;
1555         struct inet6_ifaddr *ifp;
1556
1557         read_lock_bh(&idev->lock);
1558         list_for_each_entry(ifp, &idev->addr_list, if_list)
1559                 cnt++;
1560         read_unlock_bh(&idev->lock);
1561         return cnt;
1562 }
1563
1564 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1565                   const struct net_device *dev, int strict)
1566 {
1567         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1568 }
1569 EXPORT_SYMBOL(ipv6_chk_addr);
1570
1571 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1572                             const struct net_device *dev, int strict,
1573                             u32 banned_flags)
1574 {
1575         struct inet6_ifaddr *ifp;
1576         unsigned int hash = inet6_addr_hash(addr);
1577         u32 ifp_flags;
1578
1579         rcu_read_lock_bh();
1580         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1581                 if (!net_eq(dev_net(ifp->idev->dev), net))
1582                         continue;
1583                 /* Decouple optimistic from tentative for evaluation here.
1584                  * Ban optimistic addresses explicitly, when required.
1585                  */
1586                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1587                             ? (ifp->flags&~IFA_F_TENTATIVE)
1588                             : ifp->flags;
1589                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1590                     !(ifp_flags&banned_flags) &&
1591                     (!dev || ifp->idev->dev == dev ||
1592                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1593                         rcu_read_unlock_bh();
1594                         return 1;
1595                 }
1596         }
1597
1598         rcu_read_unlock_bh();
1599         return 0;
1600 }
1601 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1602
1603 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1604                                struct net_device *dev)
1605 {
1606         unsigned int hash = inet6_addr_hash(addr);
1607         struct inet6_ifaddr *ifp;
1608
1609         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1610                 if (!net_eq(dev_net(ifp->idev->dev), net))
1611                         continue;
1612                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1613                         if (!dev || ifp->idev->dev == dev)
1614                                 return true;
1615                 }
1616         }
1617         return false;
1618 }
1619
1620 /* Compares an address/prefix_len with addresses on device @dev.
1621  * If one is found it returns true.
1622  */
1623 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1624         const unsigned int prefix_len, struct net_device *dev)
1625 {
1626         struct inet6_dev *idev;
1627         struct inet6_ifaddr *ifa;
1628         bool ret = false;
1629
1630         rcu_read_lock();
1631         idev = __in6_dev_get(dev);
1632         if (idev) {
1633                 read_lock_bh(&idev->lock);
1634                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1635                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1636                         if (ret)
1637                                 break;
1638                 }
1639                 read_unlock_bh(&idev->lock);
1640         }
1641         rcu_read_unlock();
1642
1643         return ret;
1644 }
1645 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1646
1647 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1648 {
1649         struct inet6_dev *idev;
1650         struct inet6_ifaddr *ifa;
1651         int     onlink;
1652
1653         onlink = 0;
1654         rcu_read_lock();
1655         idev = __in6_dev_get(dev);
1656         if (idev) {
1657                 read_lock_bh(&idev->lock);
1658                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1659                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1660                                                    ifa->prefix_len);
1661                         if (onlink)
1662                                 break;
1663                 }
1664                 read_unlock_bh(&idev->lock);
1665         }
1666         rcu_read_unlock();
1667         return onlink;
1668 }
1669 EXPORT_SYMBOL(ipv6_chk_prefix);
1670
1671 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1672                                      struct net_device *dev, int strict)
1673 {
1674         struct inet6_ifaddr *ifp, *result = NULL;
1675         unsigned int hash = inet6_addr_hash(addr);
1676
1677         rcu_read_lock_bh();
1678         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1679                 if (!net_eq(dev_net(ifp->idev->dev), net))
1680                         continue;
1681                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1682                         if (!dev || ifp->idev->dev == dev ||
1683                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1684                                 result = ifp;
1685                                 in6_ifa_hold(ifp);
1686                                 break;
1687                         }
1688                 }
1689         }
1690         rcu_read_unlock_bh();
1691
1692         return result;
1693 }
1694
1695 /* Gets referenced address, destroys ifaddr */
1696
1697 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1698 {
1699         if (ifp->flags&IFA_F_PERMANENT) {
1700                 spin_lock_bh(&ifp->lock);
1701                 addrconf_del_dad_work(ifp);
1702                 ifp->flags |= IFA_F_TENTATIVE;
1703                 if (dad_failed)
1704                         ifp->flags |= IFA_F_DADFAILED;
1705                 spin_unlock_bh(&ifp->lock);
1706                 if (dad_failed)
1707                         ipv6_ifa_notify(0, ifp);
1708                 in6_ifa_put(ifp);
1709         } else if (ifp->flags&IFA_F_TEMPORARY) {
1710                 struct inet6_ifaddr *ifpub;
1711                 spin_lock_bh(&ifp->lock);
1712                 ifpub = ifp->ifpub;
1713                 if (ifpub) {
1714                         in6_ifa_hold(ifpub);
1715                         spin_unlock_bh(&ifp->lock);
1716                         ipv6_create_tempaddr(ifpub, ifp);
1717                         in6_ifa_put(ifpub);
1718                 } else {
1719                         spin_unlock_bh(&ifp->lock);
1720                 }
1721                 ipv6_del_addr(ifp);
1722         } else {
1723                 ipv6_del_addr(ifp);
1724         }
1725 }
1726
1727 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1728 {
1729         int err = -ENOENT;
1730
1731         spin_lock_bh(&ifp->lock);
1732         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1733                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1734                 err = 0;
1735         }
1736         spin_unlock_bh(&ifp->lock);
1737
1738         return err;
1739 }
1740
1741 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1742 {
1743         struct in6_addr addr;
1744         struct inet6_dev *idev = ifp->idev;
1745         struct net *net = dev_net(ifp->idev->dev);
1746
1747         if (addrconf_dad_end(ifp)) {
1748                 in6_ifa_put(ifp);
1749                 return;
1750         }
1751
1752         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1753                              ifp->idev->dev->name, &ifp->addr);
1754
1755         spin_lock_bh(&ifp->lock);
1756
1757         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1758                 int scope = ifp->scope;
1759                 u32 flags = ifp->flags;
1760                 struct in6_addr new_addr;
1761                 struct inet6_ifaddr *ifp2;
1762                 u32 valid_lft, preferred_lft;
1763                 int pfxlen = ifp->prefix_len;
1764                 int retries = ifp->stable_privacy_retry + 1;
1765
1766                 if (retries > net->ipv6.sysctl.idgen_retries) {
1767                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1768                                              ifp->idev->dev->name);
1769                         goto errdad;
1770                 }
1771
1772                 new_addr = ifp->addr;
1773                 if (ipv6_generate_stable_address(&new_addr, retries,
1774                                                  idev))
1775                         goto errdad;
1776
1777                 valid_lft = ifp->valid_lft;
1778                 preferred_lft = ifp->prefered_lft;
1779
1780                 spin_unlock_bh(&ifp->lock);
1781
1782                 if (idev->cnf.max_addresses &&
1783                     ipv6_count_addresses(idev) >=
1784                     idev->cnf.max_addresses)
1785                         goto lock_errdad;
1786
1787                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1788                                      ifp->idev->dev->name);
1789
1790                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1791                                      scope, flags, valid_lft,
1792                                      preferred_lft);
1793                 if (IS_ERR(ifp2))
1794                         goto lock_errdad;
1795
1796                 spin_lock_bh(&ifp2->lock);
1797                 ifp2->stable_privacy_retry = retries;
1798                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1799                 spin_unlock_bh(&ifp2->lock);
1800
1801                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1802                 in6_ifa_put(ifp2);
1803 lock_errdad:
1804                 spin_lock_bh(&ifp->lock);
1805         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1806                 addr.s6_addr32[0] = htonl(0xfe800000);
1807                 addr.s6_addr32[1] = 0;
1808
1809                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1810                     ipv6_addr_equal(&ifp->addr, &addr)) {
1811                         /* DAD failed for link-local based on MAC address */
1812                         idev->cnf.disable_ipv6 = 1;
1813
1814                         pr_info("%s: IPv6 being disabled!\n",
1815                                 ifp->idev->dev->name);
1816                 }
1817         }
1818
1819 errdad:
1820         /* transition from _POSTDAD to _ERRDAD */
1821         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1822         spin_unlock_bh(&ifp->lock);
1823
1824         addrconf_mod_dad_work(ifp, 0);
1825 }
1826
1827 /* Join to solicited addr multicast group.
1828  * caller must hold RTNL */
1829 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1830 {
1831         struct in6_addr maddr;
1832
1833         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1834                 return;
1835
1836         addrconf_addr_solict_mult(addr, &maddr);
1837         ipv6_dev_mc_inc(dev, &maddr);
1838 }
1839
1840 /* caller must hold RTNL */
1841 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1842 {
1843         struct in6_addr maddr;
1844
1845         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1846                 return;
1847
1848         addrconf_addr_solict_mult(addr, &maddr);
1849         __ipv6_dev_mc_dec(idev, &maddr);
1850 }
1851
1852 /* caller must hold RTNL */
1853 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1854 {
1855         struct in6_addr addr;
1856
1857         if (ifp->prefix_len >= 127) /* RFC 6164 */
1858                 return;
1859         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1860         if (ipv6_addr_any(&addr))
1861                 return;
1862         __ipv6_dev_ac_inc(ifp->idev, &addr);
1863 }
1864
1865 /* caller must hold RTNL */
1866 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1867 {
1868         struct in6_addr addr;
1869
1870         if (ifp->prefix_len >= 127) /* RFC 6164 */
1871                 return;
1872         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1873         if (ipv6_addr_any(&addr))
1874                 return;
1875         __ipv6_dev_ac_dec(ifp->idev, &addr);
1876 }
1877
1878 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1879 {
1880         if (dev->addr_len != ETH_ALEN)
1881                 return -1;
1882         memcpy(eui, dev->dev_addr, 3);
1883         memcpy(eui + 5, dev->dev_addr + 3, 3);
1884
1885         /*
1886          * The zSeries OSA network cards can be shared among various
1887          * OS instances, but the OSA cards have only one MAC address.
1888          * This leads to duplicate address conflicts in conjunction
1889          * with IPv6 if more than one instance uses the same card.
1890          *
1891          * The driver for these cards can deliver a unique 16-bit
1892          * identifier for each instance sharing the same card.  It is
1893          * placed instead of 0xFFFE in the interface identifier.  The
1894          * "u" bit of the interface identifier is not inverted in this
1895          * case.  Hence the resulting interface identifier has local
1896          * scope according to RFC2373.
1897          */
1898         if (dev->dev_id) {
1899                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1900                 eui[4] = dev->dev_id & 0xFF;
1901         } else {
1902                 eui[3] = 0xFF;
1903                 eui[4] = 0xFE;
1904                 eui[0] ^= 2;
1905         }
1906         return 0;
1907 }
1908
1909 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1910 {
1911         if (dev->addr_len != IEEE802154_ADDR_LEN)
1912                 return -1;
1913         memcpy(eui, dev->dev_addr, 8);
1914         eui[0] ^= 2;
1915         return 0;
1916 }
1917
1918 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1919 {
1920         union fwnet_hwaddr *ha;
1921
1922         if (dev->addr_len != FWNET_ALEN)
1923                 return -1;
1924
1925         ha = (union fwnet_hwaddr *)dev->dev_addr;
1926
1927         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1928         eui[0] ^= 2;
1929         return 0;
1930 }
1931
1932 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1933 {
1934         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1935         if (dev->addr_len != ARCNET_ALEN)
1936                 return -1;
1937         memset(eui, 0, 7);
1938         eui[7] = *(u8 *)dev->dev_addr;
1939         return 0;
1940 }
1941
1942 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1943 {
1944         if (dev->addr_len != INFINIBAND_ALEN)
1945                 return -1;
1946         memcpy(eui, dev->dev_addr + 12, 8);
1947         eui[0] |= 2;
1948         return 0;
1949 }
1950
1951 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1952 {
1953         if (addr == 0)
1954                 return -1;
1955         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1956                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1957                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1958                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1959                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1960                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1961         eui[1] = 0;
1962         eui[2] = 0x5E;
1963         eui[3] = 0xFE;
1964         memcpy(eui + 4, &addr, 4);
1965         return 0;
1966 }
1967
1968 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1969 {
1970         if (dev->priv_flags & IFF_ISATAP)
1971                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1972         return -1;
1973 }
1974
1975 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1976 {
1977         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1978 }
1979
1980 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1981 {
1982         memcpy(eui, dev->perm_addr, 3);
1983         memcpy(eui + 5, dev->perm_addr + 3, 3);
1984         eui[3] = 0xFF;
1985         eui[4] = 0xFE;
1986         eui[0] ^= 2;
1987         return 0;
1988 }
1989
1990 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1991 {
1992         switch (dev->type) {
1993         case ARPHRD_ETHER:
1994         case ARPHRD_FDDI:
1995                 return addrconf_ifid_eui48(eui, dev);
1996         case ARPHRD_ARCNET:
1997                 return addrconf_ifid_arcnet(eui, dev);
1998         case ARPHRD_INFINIBAND:
1999                 return addrconf_ifid_infiniband(eui, dev);
2000         case ARPHRD_SIT:
2001                 return addrconf_ifid_sit(eui, dev);
2002         case ARPHRD_IPGRE:
2003                 return addrconf_ifid_gre(eui, dev);
2004         case ARPHRD_6LOWPAN:
2005         case ARPHRD_IEEE802154:
2006                 return addrconf_ifid_eui64(eui, dev);
2007         case ARPHRD_IEEE1394:
2008                 return addrconf_ifid_ieee1394(eui, dev);
2009         case ARPHRD_TUNNEL6:
2010                 return addrconf_ifid_ip6tnl(eui, dev);
2011         }
2012         return -1;
2013 }
2014
2015 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2016 {
2017         int err = -1;
2018         struct inet6_ifaddr *ifp;
2019
2020         read_lock_bh(&idev->lock);
2021         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2022                 if (ifp->scope > IFA_LINK)
2023                         break;
2024                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2025                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2026                         err = 0;
2027                         break;
2028                 }
2029         }
2030         read_unlock_bh(&idev->lock);
2031         return err;
2032 }
2033
2034 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2035 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2036 {
2037 regen:
2038         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2039         idev->rndid[0] &= ~0x02;
2040
2041         /*
2042          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2043          * check if generated address is not inappropriate
2044          *
2045          *  - Reserved subnet anycast (RFC 2526)
2046          *      11111101 11....11 1xxxxxxx
2047          *  - ISATAP (RFC4214) 6.1
2048          *      00-00-5E-FE-xx-xx-xx-xx
2049          *  - value 0
2050          *  - XXX: already assigned to an address on the device
2051          */
2052         if (idev->rndid[0] == 0xfd &&
2053             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2054             (idev->rndid[7]&0x80))
2055                 goto regen;
2056         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2057                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2058                         goto regen;
2059                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2060                         goto regen;
2061         }
2062 }
2063
2064 static void ipv6_regen_rndid(unsigned long data)
2065 {
2066         struct inet6_dev *idev = (struct inet6_dev *) data;
2067         unsigned long expires;
2068
2069         rcu_read_lock_bh();
2070         write_lock_bh(&idev->lock);
2071
2072         if (idev->dead)
2073                 goto out;
2074
2075         __ipv6_regen_rndid(idev);
2076
2077         expires = jiffies +
2078                 idev->cnf.temp_prefered_lft * HZ -
2079                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2080                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2081                 idev->cnf.max_desync_factor * HZ;
2082         if (time_before(expires, jiffies)) {
2083                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2084                         __func__, idev->dev->name);
2085                 goto out;
2086         }
2087
2088         if (!mod_timer(&idev->regen_timer, expires))
2089                 in6_dev_hold(idev);
2090
2091 out:
2092         write_unlock_bh(&idev->lock);
2093         rcu_read_unlock_bh();
2094         in6_dev_put(idev);
2095 }
2096
2097 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2098 {
2099         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2100                 __ipv6_regen_rndid(idev);
2101 }
2102
2103 /*
2104  *      Add prefix route.
2105  */
2106
2107 static void
2108 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2109                       unsigned long expires, u32 flags)
2110 {
2111         struct fib6_config cfg = {
2112                 .fc_table = RT6_TABLE_PREFIX,
2113                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2114                 .fc_ifindex = dev->ifindex,
2115                 .fc_expires = expires,
2116                 .fc_dst_len = plen,
2117                 .fc_flags = RTF_UP | flags,
2118                 .fc_nlinfo.nl_net = dev_net(dev),
2119                 .fc_protocol = RTPROT_KERNEL,
2120         };
2121
2122         cfg.fc_dst = *pfx;
2123
2124         /* Prevent useless cloning on PtP SIT.
2125            This thing is done here expecting that the whole
2126            class of non-broadcast devices need not cloning.
2127          */
2128 #if IS_ENABLED(CONFIG_IPV6_SIT)
2129         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2130                 cfg.fc_flags |= RTF_NONEXTHOP;
2131 #endif
2132
2133         ip6_route_add(&cfg);
2134 }
2135
2136
2137 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2138                                                   int plen,
2139                                                   const struct net_device *dev,
2140                                                   u32 flags, u32 noflags)
2141 {
2142         struct fib6_node *fn;
2143         struct rt6_info *rt = NULL;
2144         struct fib6_table *table;
2145
2146         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2147         if (!table)
2148                 return NULL;
2149
2150         read_lock_bh(&table->tb6_lock);
2151         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2152         if (!fn)
2153                 goto out;
2154
2155         noflags |= RTF_CACHE;
2156         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2157                 if (rt->dst.dev->ifindex != dev->ifindex)
2158                         continue;
2159                 if ((rt->rt6i_flags & flags) != flags)
2160                         continue;
2161                 if ((rt->rt6i_flags & noflags) != 0)
2162                         continue;
2163                 dst_hold(&rt->dst);
2164                 break;
2165         }
2166 out:
2167         read_unlock_bh(&table->tb6_lock);
2168         return rt;
2169 }
2170
2171
2172 /* Create "default" multicast route to the interface */
2173
2174 static void addrconf_add_mroute(struct net_device *dev)
2175 {
2176         struct fib6_config cfg = {
2177                 .fc_table = RT6_TABLE_LOCAL,
2178                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2179                 .fc_ifindex = dev->ifindex,
2180                 .fc_dst_len = 8,
2181                 .fc_flags = RTF_UP,
2182                 .fc_nlinfo.nl_net = dev_net(dev),
2183         };
2184
2185         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2186
2187         ip6_route_add(&cfg);
2188 }
2189
2190 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2191 {
2192         struct inet6_dev *idev;
2193
2194         ASSERT_RTNL();
2195
2196         idev = ipv6_find_idev(dev);
2197         if (!idev)
2198                 return ERR_PTR(-ENOBUFS);
2199
2200         if (idev->cnf.disable_ipv6)
2201                 return ERR_PTR(-EACCES);
2202
2203         /* Add default multicast route */
2204         if (!(dev->flags & IFF_LOOPBACK))
2205                 addrconf_add_mroute(dev);
2206
2207         return idev;
2208 }
2209
2210 static void manage_tempaddrs(struct inet6_dev *idev,
2211                              struct inet6_ifaddr *ifp,
2212                              __u32 valid_lft, __u32 prefered_lft,
2213                              bool create, unsigned long now)
2214 {
2215         u32 flags;
2216         struct inet6_ifaddr *ift;
2217
2218         read_lock_bh(&idev->lock);
2219         /* update all temporary addresses in the list */
2220         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2221                 int age, max_valid, max_prefered;
2222
2223                 if (ifp != ift->ifpub)
2224                         continue;
2225
2226                 /* RFC 4941 section 3.3:
2227                  * If a received option will extend the lifetime of a public
2228                  * address, the lifetimes of temporary addresses should
2229                  * be extended, subject to the overall constraint that no
2230                  * temporary addresses should ever remain "valid" or "preferred"
2231                  * for a time longer than (TEMP_VALID_LIFETIME) or
2232                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2233                  */
2234                 age = (now - ift->cstamp) / HZ;
2235                 max_valid = idev->cnf.temp_valid_lft - age;
2236                 if (max_valid < 0)
2237                         max_valid = 0;
2238
2239                 max_prefered = idev->cnf.temp_prefered_lft -
2240                                idev->cnf.max_desync_factor - age;
2241                 if (max_prefered < 0)
2242                         max_prefered = 0;
2243
2244                 if (valid_lft > max_valid)
2245                         valid_lft = max_valid;
2246
2247                 if (prefered_lft > max_prefered)
2248                         prefered_lft = max_prefered;
2249
2250                 spin_lock(&ift->lock);
2251                 flags = ift->flags;
2252                 ift->valid_lft = valid_lft;
2253                 ift->prefered_lft = prefered_lft;
2254                 ift->tstamp = now;
2255                 if (prefered_lft > 0)
2256                         ift->flags &= ~IFA_F_DEPRECATED;
2257
2258                 spin_unlock(&ift->lock);
2259                 if (!(flags&IFA_F_TENTATIVE))
2260                         ipv6_ifa_notify(0, ift);
2261         }
2262
2263         if ((create || list_empty(&idev->tempaddr_list)) &&
2264             idev->cnf.use_tempaddr > 0) {
2265                 /* When a new public address is created as described
2266                  * in [ADDRCONF], also create a new temporary address.
2267                  * Also create a temporary address if it's enabled but
2268                  * no temporary address currently exists.
2269                  */
2270                 read_unlock_bh(&idev->lock);
2271                 ipv6_create_tempaddr(ifp, NULL);
2272         } else {
2273                 read_unlock_bh(&idev->lock);
2274         }
2275 }
2276
2277 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2278 {
2279         struct prefix_info *pinfo;
2280         __u32 valid_lft;
2281         __u32 prefered_lft;
2282         int addr_type;
2283         u32 addr_flags = 0;
2284         struct inet6_dev *in6_dev;
2285         struct net *net = dev_net(dev);
2286
2287         pinfo = (struct prefix_info *) opt;
2288
2289         if (len < sizeof(struct prefix_info)) {
2290                 ADBG("addrconf: prefix option too short\n");
2291                 return;
2292         }
2293
2294         /*
2295          *      Validation checks ([ADDRCONF], page 19)
2296          */
2297
2298         addr_type = ipv6_addr_type(&pinfo->prefix);
2299
2300         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2301                 return;
2302
2303         valid_lft = ntohl(pinfo->valid);
2304         prefered_lft = ntohl(pinfo->prefered);
2305
2306         if (prefered_lft > valid_lft) {
2307                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2308                 return;
2309         }
2310
2311         in6_dev = in6_dev_get(dev);
2312
2313         if (!in6_dev) {
2314                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2315                                     dev->name);
2316                 return;
2317         }
2318
2319         /*
2320          *      Two things going on here:
2321          *      1) Add routes for on-link prefixes
2322          *      2) Configure prefixes with the auto flag set
2323          */
2324
2325         if (pinfo->onlink) {
2326                 struct rt6_info *rt;
2327                 unsigned long rt_expires;
2328
2329                 /* Avoid arithmetic overflow. Really, we could
2330                  * save rt_expires in seconds, likely valid_lft,
2331                  * but it would require division in fib gc, that it
2332                  * not good.
2333                  */
2334                 if (HZ > USER_HZ)
2335                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2336                 else
2337                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2338
2339                 if (addrconf_finite_timeout(rt_expires))
2340                         rt_expires *= HZ;
2341
2342                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2343                                                pinfo->prefix_len,
2344                                                dev,
2345                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2346                                                RTF_GATEWAY | RTF_DEFAULT);
2347
2348                 if (rt) {
2349                         /* Autoconf prefix route */
2350                         if (valid_lft == 0) {
2351                                 ip6_del_rt(rt);
2352                                 rt = NULL;
2353                         } else if (addrconf_finite_timeout(rt_expires)) {
2354                                 /* not infinity */
2355                                 rt6_set_expires(rt, jiffies + rt_expires);
2356                         } else {
2357                                 rt6_clean_expires(rt);
2358                         }
2359                 } else if (valid_lft) {
2360                         clock_t expires = 0;
2361                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2362                         if (addrconf_finite_timeout(rt_expires)) {
2363                                 /* not infinity */
2364                                 flags |= RTF_EXPIRES;
2365                                 expires = jiffies_to_clock_t(rt_expires);
2366                         }
2367                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2368                                               dev, expires, flags);
2369                 }
2370                 ip6_rt_put(rt);
2371         }
2372
2373         /* Try to figure out our local address for this prefix */
2374
2375         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2376                 struct inet6_ifaddr *ifp;
2377                 struct in6_addr addr;
2378                 int create = 0, update_lft = 0;
2379                 bool tokenized = false;
2380
2381                 if (pinfo->prefix_len == 64) {
2382                         memcpy(&addr, &pinfo->prefix, 8);
2383
2384                         if (!ipv6_addr_any(&in6_dev->token)) {
2385                                 read_lock_bh(&in6_dev->lock);
2386                                 memcpy(addr.s6_addr + 8,
2387                                        in6_dev->token.s6_addr + 8, 8);
2388                                 read_unlock_bh(&in6_dev->lock);
2389                                 tokenized = true;
2390                         } else if (in6_dev->addr_gen_mode ==
2391                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2392                                    !ipv6_generate_stable_address(&addr, 0,
2393                                                                  in6_dev)) {
2394                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2395                                 goto ok;
2396                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2397                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2398                                 in6_dev_put(in6_dev);
2399                                 return;
2400                         }
2401                         goto ok;
2402                 }
2403                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2404                                     pinfo->prefix_len);
2405                 in6_dev_put(in6_dev);
2406                 return;
2407
2408 ok:
2409
2410                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2411
2412                 if (!ifp && valid_lft) {
2413                         int max_addresses = in6_dev->cnf.max_addresses;
2414
2415 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2416                         if (in6_dev->cnf.optimistic_dad &&
2417                             !net->ipv6.devconf_all->forwarding && sllao)
2418                                 addr_flags = IFA_F_OPTIMISTIC;
2419 #endif
2420
2421                         /* Do not allow to create too much of autoconfigured
2422                          * addresses; this would be too easy way to crash kernel.
2423                          */
2424                         if (!max_addresses ||
2425                             ipv6_count_addresses(in6_dev) < max_addresses)
2426                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2427                                                     pinfo->prefix_len,
2428                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2429                                                     addr_flags, valid_lft,
2430                                                     prefered_lft);
2431
2432                         if (IS_ERR_OR_NULL(ifp)) {
2433                                 in6_dev_put(in6_dev);
2434                                 return;
2435                         }
2436
2437                         update_lft = 0;
2438                         create = 1;
2439                         spin_lock_bh(&ifp->lock);
2440                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2441                         ifp->cstamp = jiffies;
2442                         ifp->tokenized = tokenized;
2443                         spin_unlock_bh(&ifp->lock);
2444                         addrconf_dad_start(ifp);
2445                 }
2446
2447                 if (ifp) {
2448                         u32 flags;
2449                         unsigned long now;
2450                         u32 stored_lft;
2451
2452                         /* update lifetime (RFC2462 5.5.3 e) */
2453                         spin_lock_bh(&ifp->lock);
2454                         now = jiffies;
2455                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2456                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2457                         else
2458                                 stored_lft = 0;
2459                         if (!update_lft && !create && stored_lft) {
2460                                 const u32 minimum_lft = min_t(u32,
2461                                         stored_lft, MIN_VALID_LIFETIME);
2462                                 valid_lft = max(valid_lft, minimum_lft);
2463
2464                                 /* RFC4862 Section 5.5.3e:
2465                                  * "Note that the preferred lifetime of the
2466                                  *  corresponding address is always reset to
2467                                  *  the Preferred Lifetime in the received
2468                                  *  Prefix Information option, regardless of
2469                                  *  whether the valid lifetime is also reset or
2470                                  *  ignored."
2471                                  *
2472                                  * So we should always update prefered_lft here.
2473                                  */
2474                                 update_lft = 1;
2475                         }
2476
2477                         if (update_lft) {
2478                                 ifp->valid_lft = valid_lft;
2479                                 ifp->prefered_lft = prefered_lft;
2480                                 ifp->tstamp = now;
2481                                 flags = ifp->flags;
2482                                 ifp->flags &= ~IFA_F_DEPRECATED;
2483                                 spin_unlock_bh(&ifp->lock);
2484
2485                                 if (!(flags&IFA_F_TENTATIVE))
2486                                         ipv6_ifa_notify(0, ifp);
2487                         } else
2488                                 spin_unlock_bh(&ifp->lock);
2489
2490                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2491                                          create, now);
2492
2493                         in6_ifa_put(ifp);
2494                         addrconf_verify();
2495                 }
2496         }
2497         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2498         in6_dev_put(in6_dev);
2499 }
2500
2501 /*
2502  *      Set destination address.
2503  *      Special case for SIT interfaces where we create a new "virtual"
2504  *      device.
2505  */
2506 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2507 {
2508         struct in6_ifreq ireq;
2509         struct net_device *dev;
2510         int err = -EINVAL;
2511
2512         rtnl_lock();
2513
2514         err = -EFAULT;
2515         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2516                 goto err_exit;
2517
2518         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2519
2520         err = -ENODEV;
2521         if (!dev)
2522                 goto err_exit;
2523
2524 #if IS_ENABLED(CONFIG_IPV6_SIT)
2525         if (dev->type == ARPHRD_SIT) {
2526                 const struct net_device_ops *ops = dev->netdev_ops;
2527                 struct ifreq ifr;
2528                 struct ip_tunnel_parm p;
2529
2530                 err = -EADDRNOTAVAIL;
2531                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2532                         goto err_exit;
2533
2534                 memset(&p, 0, sizeof(p));
2535                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2536                 p.iph.saddr = 0;
2537                 p.iph.version = 4;
2538                 p.iph.ihl = 5;
2539                 p.iph.protocol = IPPROTO_IPV6;
2540                 p.iph.ttl = 64;
2541                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2542
2543                 if (ops->ndo_do_ioctl) {
2544                         mm_segment_t oldfs = get_fs();
2545
2546                         set_fs(KERNEL_DS);
2547                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2548                         set_fs(oldfs);
2549                 } else
2550                         err = -EOPNOTSUPP;
2551
2552                 if (err == 0) {
2553                         err = -ENOBUFS;
2554                         dev = __dev_get_by_name(net, p.name);
2555                         if (!dev)
2556                                 goto err_exit;
2557                         err = dev_open(dev);
2558                 }
2559         }
2560 #endif
2561
2562 err_exit:
2563         rtnl_unlock();
2564         return err;
2565 }
2566
2567 static int ipv6_mc_config(struct sock *sk, bool join,
2568                           const struct in6_addr *addr, int ifindex)
2569 {
2570         int ret;
2571
2572         ASSERT_RTNL();
2573
2574         lock_sock(sk);
2575         if (join)
2576                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2577         else
2578                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2579         release_sock(sk);
2580
2581         return ret;
2582 }
2583
2584 /*
2585  *      Manual configuration of address on an interface
2586  */
2587 static int inet6_addr_add(struct net *net, int ifindex,
2588                           const struct in6_addr *pfx,
2589                           const struct in6_addr *peer_pfx,
2590                           unsigned int plen, __u32 ifa_flags,
2591                           __u32 prefered_lft, __u32 valid_lft)
2592 {
2593         struct inet6_ifaddr *ifp;
2594         struct inet6_dev *idev;
2595         struct net_device *dev;
2596         unsigned long timeout;
2597         clock_t expires;
2598         int scope;
2599         u32 flags;
2600
2601         ASSERT_RTNL();
2602
2603         if (plen > 128)
2604                 return -EINVAL;
2605
2606         /* check the lifetime */
2607         if (!valid_lft || prefered_lft > valid_lft)
2608                 return -EINVAL;
2609
2610         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2611                 return -EINVAL;
2612
2613         dev = __dev_get_by_index(net, ifindex);
2614         if (!dev)
2615                 return -ENODEV;
2616
2617         idev = addrconf_add_dev(dev);
2618         if (IS_ERR(idev))
2619                 return PTR_ERR(idev);
2620
2621         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2622                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2623                                          true, pfx, ifindex);
2624
2625                 if (ret < 0)
2626                         return ret;
2627         }
2628
2629         scope = ipv6_addr_scope(pfx);
2630
2631         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2632         if (addrconf_finite_timeout(timeout)) {
2633                 expires = jiffies_to_clock_t(timeout * HZ);
2634                 valid_lft = timeout;
2635                 flags = RTF_EXPIRES;
2636         } else {
2637                 expires = 0;
2638                 flags = 0;
2639                 ifa_flags |= IFA_F_PERMANENT;
2640         }
2641
2642         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2643         if (addrconf_finite_timeout(timeout)) {
2644                 if (timeout == 0)
2645                         ifa_flags |= IFA_F_DEPRECATED;
2646                 prefered_lft = timeout;
2647         }
2648
2649         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2650                             valid_lft, prefered_lft);
2651
2652         if (!IS_ERR(ifp)) {
2653                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2654                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2655                                               expires, flags);
2656                 }
2657
2658                 /*
2659                  * Note that section 3.1 of RFC 4429 indicates
2660                  * that the Optimistic flag should not be set for
2661                  * manually configured addresses
2662                  */
2663                 addrconf_dad_start(ifp);
2664                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2665                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2666                                          true, jiffies);
2667                 in6_ifa_put(ifp);
2668                 addrconf_verify_rtnl();
2669                 return 0;
2670         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2671                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2672                                false, pfx, ifindex);
2673         }
2674
2675         return PTR_ERR(ifp);
2676 }
2677
2678 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2679                           const struct in6_addr *pfx, unsigned int plen)
2680 {
2681         struct inet6_ifaddr *ifp;
2682         struct inet6_dev *idev;
2683         struct net_device *dev;
2684
2685         if (plen > 128)
2686                 return -EINVAL;
2687
2688         dev = __dev_get_by_index(net, ifindex);
2689         if (!dev)
2690                 return -ENODEV;
2691
2692         idev = __in6_dev_get(dev);
2693         if (!idev)
2694                 return -ENXIO;
2695
2696         read_lock_bh(&idev->lock);
2697         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2698                 if (ifp->prefix_len == plen &&
2699                     ipv6_addr_equal(pfx, &ifp->addr)) {
2700                         in6_ifa_hold(ifp);
2701                         read_unlock_bh(&idev->lock);
2702
2703                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2704                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2705                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2706                                                  jiffies);
2707                         ipv6_del_addr(ifp);
2708                         addrconf_verify_rtnl();
2709                         if (ipv6_addr_is_multicast(pfx)) {
2710                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2711                                                false, pfx, dev->ifindex);
2712                         }
2713                         return 0;
2714                 }
2715         }
2716         read_unlock_bh(&idev->lock);
2717         return -EADDRNOTAVAIL;
2718 }
2719
2720
2721 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2722 {
2723         struct in6_ifreq ireq;
2724         int err;
2725
2726         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2727                 return -EPERM;
2728
2729         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2730                 return -EFAULT;
2731
2732         rtnl_lock();
2733         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2734                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2735                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2736         rtnl_unlock();
2737         return err;
2738 }
2739
2740 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2741 {
2742         struct in6_ifreq ireq;
2743         int err;
2744
2745         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2746                 return -EPERM;
2747
2748         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2749                 return -EFAULT;
2750
2751         rtnl_lock();
2752         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2753                              ireq.ifr6_prefixlen);
2754         rtnl_unlock();
2755         return err;
2756 }
2757
2758 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2759                      int plen, int scope)
2760 {
2761         struct inet6_ifaddr *ifp;
2762
2763         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2764                             scope, IFA_F_PERMANENT,
2765                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2766         if (!IS_ERR(ifp)) {
2767                 spin_lock_bh(&ifp->lock);
2768                 ifp->flags &= ~IFA_F_TENTATIVE;
2769                 spin_unlock_bh(&ifp->lock);
2770                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2771                 in6_ifa_put(ifp);
2772         }
2773 }
2774
2775 #if IS_ENABLED(CONFIG_IPV6_SIT)
2776 static void sit_add_v4_addrs(struct inet6_dev *idev)
2777 {
2778         struct in6_addr addr;
2779         struct net_device *dev;
2780         struct net *net = dev_net(idev->dev);
2781         int scope, plen;
2782         u32 pflags = 0;
2783
2784         ASSERT_RTNL();
2785
2786         memset(&addr, 0, sizeof(struct in6_addr));
2787         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2788
2789         if (idev->dev->flags&IFF_POINTOPOINT) {
2790                 addr.s6_addr32[0] = htonl(0xfe800000);
2791                 scope = IFA_LINK;
2792                 plen = 64;
2793         } else {
2794                 scope = IPV6_ADDR_COMPATv4;
2795                 plen = 96;
2796                 pflags |= RTF_NONEXTHOP;
2797         }
2798
2799         if (addr.s6_addr32[3]) {
2800                 add_addr(idev, &addr, plen, scope);
2801                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2802                 return;
2803         }
2804
2805         for_each_netdev(net, dev) {
2806                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2807                 if (in_dev && (dev->flags & IFF_UP)) {
2808                         struct in_ifaddr *ifa;
2809
2810                         int flag = scope;
2811
2812                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2813
2814                                 addr.s6_addr32[3] = ifa->ifa_local;
2815
2816                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2817                                         continue;
2818                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2819                                         if (idev->dev->flags&IFF_POINTOPOINT)
2820                                                 continue;
2821                                         flag |= IFA_HOST;
2822                                 }
2823
2824                                 add_addr(idev, &addr, plen, flag);
2825                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2826                                                       pflags);
2827                         }
2828                 }
2829         }
2830 }
2831 #endif
2832
2833 static void init_loopback(struct net_device *dev)
2834 {
2835         struct inet6_dev  *idev;
2836         struct net_device *sp_dev;
2837         struct inet6_ifaddr *sp_ifa;
2838         struct rt6_info *sp_rt;
2839
2840         /* ::1 */
2841
2842         ASSERT_RTNL();
2843
2844         idev = ipv6_find_idev(dev);
2845         if (!idev) {
2846                 pr_debug("%s: add_dev failed\n", __func__);
2847                 return;
2848         }
2849
2850         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2851
2852         /* Add routes to other interface's IPv6 addresses */
2853         for_each_netdev(dev_net(dev), sp_dev) {
2854                 if (!strcmp(sp_dev->name, dev->name))
2855                         continue;
2856
2857                 idev = __in6_dev_get(sp_dev);
2858                 if (!idev)
2859                         continue;
2860
2861                 read_lock_bh(&idev->lock);
2862                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2863
2864                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2865                                 continue;
2866
2867                         if (sp_ifa->rt) {
2868                                 /* This dst has been added to garbage list when
2869                                  * lo device down, release this obsolete dst and
2870                                  * reallocate a new router for ifa.
2871                                  */
2872                                 if (sp_ifa->rt->dst.obsolete > 0) {
2873                                         ip6_rt_put(sp_ifa->rt);
2874                                         sp_ifa->rt = NULL;
2875                                 } else {
2876                                         continue;
2877                                 }
2878                         }
2879
2880                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2881
2882                         /* Failure cases are ignored */
2883                         if (!IS_ERR(sp_rt)) {
2884                                 sp_ifa->rt = sp_rt;
2885                                 ip6_ins_rt(sp_rt);
2886                         }
2887                 }
2888                 read_unlock_bh(&idev->lock);
2889         }
2890 }
2891
2892 static void addrconf_add_linklocal(struct inet6_dev *idev,
2893                                    const struct in6_addr *addr, u32 flags)
2894 {
2895         struct inet6_ifaddr *ifp;
2896         u32 addr_flags = flags | IFA_F_PERMANENT;
2897
2898 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2899         if (idev->cnf.optimistic_dad &&
2900             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2901                 addr_flags |= IFA_F_OPTIMISTIC;
2902 #endif
2903
2904         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2905                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2906         if (!IS_ERR(ifp)) {
2907                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2908                 addrconf_dad_start(ifp);
2909                 in6_ifa_put(ifp);
2910         }
2911 }
2912
2913 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2914 {
2915         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2916                 return true;
2917
2918         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2919             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2920                 return true;
2921
2922         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2923             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2924                 return true;
2925
2926         return false;
2927 }
2928
2929 static int ipv6_generate_stable_address(struct in6_addr *address,
2930                                         u8 dad_count,
2931                                         const struct inet6_dev *idev)
2932 {
2933         static DEFINE_SPINLOCK(lock);
2934         static __u32 digest[SHA_DIGEST_WORDS];
2935         static __u32 workspace[SHA_WORKSPACE_WORDS];
2936
2937         static union {
2938                 char __data[SHA_MESSAGE_BYTES];
2939                 struct {
2940                         struct in6_addr secret;
2941                         __be32 prefix[2];
2942                         unsigned char hwaddr[MAX_ADDR_LEN];
2943                         u8 dad_count;
2944                 } __packed;
2945         } data;
2946
2947         struct in6_addr secret;
2948         struct in6_addr temp;
2949         struct net *net = dev_net(idev->dev);
2950
2951         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2952
2953         if (idev->cnf.stable_secret.initialized)
2954                 secret = idev->cnf.stable_secret.secret;
2955         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
2956                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
2957         else
2958                 return -1;
2959
2960 retry:
2961         spin_lock_bh(&lock);
2962
2963         sha_init(digest);
2964         memset(&data, 0, sizeof(data));
2965         memset(workspace, 0, sizeof(workspace));
2966         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
2967         data.prefix[0] = address->s6_addr32[0];
2968         data.prefix[1] = address->s6_addr32[1];
2969         data.secret = secret;
2970         data.dad_count = dad_count;
2971
2972         sha_transform(digest, data.__data, workspace);
2973
2974         temp = *address;
2975         temp.s6_addr32[2] = (__force __be32)digest[0];
2976         temp.s6_addr32[3] = (__force __be32)digest[1];
2977
2978         spin_unlock_bh(&lock);
2979
2980         if (ipv6_reserved_interfaceid(temp)) {
2981                 dad_count++;
2982                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
2983                         return -1;
2984                 goto retry;
2985         }
2986
2987         *address = temp;
2988         return 0;
2989 }
2990
2991 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2992 {
2993         struct in6_addr addr;
2994
2995         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2996
2997         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
2998                 if (!ipv6_generate_stable_address(&addr, 0, idev))
2999                         addrconf_add_linklocal(idev, &addr,
3000                                                IFA_F_STABLE_PRIVACY);
3001                 else if (prefix_route)
3002                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3003         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3004                 /* addrconf_add_linklocal also adds a prefix_route and we
3005                  * only need to care about prefix routes if ipv6_generate_eui64
3006                  * couldn't generate one.
3007                  */
3008                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3009                         addrconf_add_linklocal(idev, &addr, 0);
3010                 else if (prefix_route)
3011                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3012         }
3013 }
3014
3015 static void addrconf_dev_config(struct net_device *dev)
3016 {
3017         struct inet6_dev *idev;
3018
3019         ASSERT_RTNL();
3020
3021         if ((dev->type != ARPHRD_ETHER) &&
3022             (dev->type != ARPHRD_FDDI) &&
3023             (dev->type != ARPHRD_ARCNET) &&
3024             (dev->type != ARPHRD_INFINIBAND) &&
3025             (dev->type != ARPHRD_IEEE802154) &&
3026             (dev->type != ARPHRD_IEEE1394) &&
3027             (dev->type != ARPHRD_TUNNEL6) &&
3028             (dev->type != ARPHRD_6LOWPAN)) {
3029                 /* Alas, we support only Ethernet autoconfiguration. */
3030                 return;
3031         }
3032
3033         idev = addrconf_add_dev(dev);
3034         if (IS_ERR(idev))
3035                 return;
3036
3037         addrconf_addr_gen(idev, false);
3038 }
3039
3040 #if IS_ENABLED(CONFIG_IPV6_SIT)
3041 static void addrconf_sit_config(struct net_device *dev)
3042 {
3043         struct inet6_dev *idev;
3044
3045         ASSERT_RTNL();
3046
3047         /*
3048          * Configure the tunnel with one of our IPv4
3049          * addresses... we should configure all of
3050          * our v4 addrs in the tunnel
3051          */
3052
3053         idev = ipv6_find_idev(dev);
3054         if (!idev) {
3055                 pr_debug("%s: add_dev failed\n", __func__);
3056                 return;
3057         }
3058
3059         if (dev->priv_flags & IFF_ISATAP) {
3060                 addrconf_addr_gen(idev, false);
3061                 return;
3062         }
3063
3064         sit_add_v4_addrs(idev);
3065
3066         if (dev->flags&IFF_POINTOPOINT)
3067                 addrconf_add_mroute(dev);
3068 }
3069 #endif
3070
3071 #if IS_ENABLED(CONFIG_NET_IPGRE)
3072 static void addrconf_gre_config(struct net_device *dev)
3073 {
3074         struct inet6_dev *idev;
3075
3076         ASSERT_RTNL();
3077
3078         idev = ipv6_find_idev(dev);
3079         if (!idev) {
3080                 pr_debug("%s: add_dev failed\n", __func__);
3081                 return;
3082         }
3083
3084         addrconf_addr_gen(idev, true);
3085 }
3086 #endif
3087
3088 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3089                            void *ptr)
3090 {
3091         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3092         struct inet6_dev *idev = __in6_dev_get(dev);
3093         int run_pending = 0;
3094         int err;
3095
3096         switch (event) {
3097         case NETDEV_REGISTER:
3098                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3099                         idev = ipv6_add_dev(dev);
3100                         if (IS_ERR(idev))
3101                                 return notifier_from_errno(PTR_ERR(idev));
3102                 }
3103                 break;
3104
3105         case NETDEV_UP:
3106         case NETDEV_CHANGE:
3107                 if (dev->flags & IFF_SLAVE)
3108                         break;
3109
3110                 if (idev && idev->cnf.disable_ipv6)
3111                         break;
3112
3113                 if (event == NETDEV_UP) {
3114                         if (!addrconf_qdisc_ok(dev)) {
3115                                 /* device is not ready yet. */
3116                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3117                                         dev->name);
3118                                 break;
3119                         }
3120
3121                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3122                                 idev = ipv6_add_dev(dev);
3123
3124                         if (!IS_ERR_OR_NULL(idev)) {
3125                                 idev->if_flags |= IF_READY;
3126                                 run_pending = 1;
3127                         }
3128                 } else {
3129                         if (!addrconf_qdisc_ok(dev)) {
3130                                 /* device is still not ready. */
3131                                 break;
3132                         }
3133
3134                         if (idev) {
3135                                 if (idev->if_flags & IF_READY)
3136                                         /* device is already configured. */
3137                                         break;
3138                                 idev->if_flags |= IF_READY;
3139                         }
3140
3141                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3142                                 dev->name);
3143
3144                         run_pending = 1;
3145                 }
3146
3147                 switch (dev->type) {
3148 #if IS_ENABLED(CONFIG_IPV6_SIT)
3149                 case ARPHRD_SIT:
3150                         addrconf_sit_config(dev);
3151                         break;
3152 #endif
3153 #if IS_ENABLED(CONFIG_NET_IPGRE)
3154                 case ARPHRD_IPGRE:
3155                         addrconf_gre_config(dev);
3156                         break;
3157 #endif
3158                 case ARPHRD_LOOPBACK:
3159                         init_loopback(dev);
3160                         break;
3161
3162                 default:
3163                         addrconf_dev_config(dev);
3164                         break;
3165                 }
3166
3167                 if (!IS_ERR_OR_NULL(idev)) {
3168                         if (run_pending)
3169                                 addrconf_dad_run(idev);
3170
3171                         /*
3172                          * If the MTU changed during the interface down,
3173                          * when the interface up, the changed MTU must be
3174                          * reflected in the idev as well as routers.
3175                          */
3176                         if (idev->cnf.mtu6 != dev->mtu &&
3177                             dev->mtu >= IPV6_MIN_MTU) {
3178                                 rt6_mtu_change(dev, dev->mtu);
3179                                 idev->cnf.mtu6 = dev->mtu;
3180                         }
3181                         idev->tstamp = jiffies;
3182                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3183
3184                         /*
3185                          * If the changed mtu during down is lower than
3186                          * IPV6_MIN_MTU stop IPv6 on this interface.
3187                          */
3188                         if (dev->mtu < IPV6_MIN_MTU)
3189                                 addrconf_ifdown(dev, 1);
3190                 }
3191                 break;
3192
3193         case NETDEV_CHANGEMTU:
3194                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3195                         rt6_mtu_change(dev, dev->mtu);
3196                         idev->cnf.mtu6 = dev->mtu;
3197                         break;
3198                 }
3199
3200                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3201                         idev = ipv6_add_dev(dev);
3202                         if (!IS_ERR(idev))
3203                                 break;
3204                 }
3205
3206                 /*
3207                  * if MTU under IPV6_MIN_MTU.
3208                  * Stop IPv6 on this interface.
3209                  */
3210
3211         case NETDEV_DOWN:
3212         case NETDEV_UNREGISTER:
3213                 /*
3214                  *      Remove all addresses from this interface.
3215                  */
3216                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3217                 break;
3218
3219         case NETDEV_CHANGENAME:
3220                 if (idev) {
3221                         snmp6_unregister_dev(idev);
3222                         addrconf_sysctl_unregister(idev);
3223                         err = addrconf_sysctl_register(idev);
3224                         if (err)
3225                                 return notifier_from_errno(err);
3226                         err = snmp6_register_dev(idev);
3227                         if (err) {
3228                                 addrconf_sysctl_unregister(idev);
3229                                 return notifier_from_errno(err);
3230                         }
3231                 }
3232                 break;
3233
3234         case NETDEV_PRE_TYPE_CHANGE:
3235         case NETDEV_POST_TYPE_CHANGE:
3236                 addrconf_type_change(dev, event);
3237                 break;
3238         }
3239
3240         return NOTIFY_OK;
3241 }
3242
3243 /*
3244  *      addrconf module should be notified of a device going up
3245  */
3246 static struct notifier_block ipv6_dev_notf = {
3247         .notifier_call = addrconf_notify,
3248 };
3249
3250 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3251 {
3252         struct inet6_dev *idev;
3253         ASSERT_RTNL();
3254
3255         idev = __in6_dev_get(dev);
3256
3257         if (event == NETDEV_POST_TYPE_CHANGE)
3258                 ipv6_mc_remap(idev);
3259         else if (event == NETDEV_PRE_TYPE_CHANGE)
3260                 ipv6_mc_unmap(idev);
3261 }
3262
3263 static int addrconf_ifdown(struct net_device *dev, int how)
3264 {
3265         struct net *net = dev_net(dev);
3266         struct inet6_dev *idev;
3267         struct inet6_ifaddr *ifa;
3268         int state, i;
3269
3270         ASSERT_RTNL();
3271
3272         rt6_ifdown(net, dev);
3273         neigh_ifdown(&nd_tbl, dev);
3274
3275         idev = __in6_dev_get(dev);
3276         if (!idev)
3277                 return -ENODEV;
3278
3279         /*
3280          * Step 1: remove reference to ipv6 device from parent device.
3281          *         Do not dev_put!
3282          */
3283         if (how) {
3284                 idev->dead = 1;
3285
3286                 /* protected by rtnl_lock */
3287                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3288
3289                 /* Step 1.5: remove snmp6 entry */
3290                 snmp6_unregister_dev(idev);
3291
3292         }
3293
3294         /* Step 2: clear hash table */
3295         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3296                 struct hlist_head *h = &inet6_addr_lst[i];
3297
3298                 spin_lock_bh(&addrconf_hash_lock);
3299 restart:
3300                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3301                         if (ifa->idev == idev) {
3302                                 hlist_del_init_rcu(&ifa->addr_lst);
3303                                 addrconf_del_dad_work(ifa);
3304                                 goto restart;
3305                         }
3306                 }
3307                 spin_unlock_bh(&addrconf_hash_lock);
3308         }
3309
3310         write_lock_bh(&idev->lock);
3311
3312         addrconf_del_rs_timer(idev);
3313
3314         /* Step 2: clear flags for stateless addrconf */
3315         if (!how)
3316                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3317
3318         if (how && del_timer(&idev->regen_timer))
3319                 in6_dev_put(idev);
3320
3321         /* Step 3: clear tempaddr list */
3322         while (!list_empty(&idev->tempaddr_list)) {
3323                 ifa = list_first_entry(&idev->tempaddr_list,
3324                                        struct inet6_ifaddr, tmp_list);
3325                 list_del(&ifa->tmp_list);
3326                 write_unlock_bh(&idev->lock);
3327                 spin_lock_bh(&ifa->lock);
3328
3329                 if (ifa->ifpub) {
3330                         in6_ifa_put(ifa->ifpub);
3331                         ifa->ifpub = NULL;
3332                 }
3333                 spin_unlock_bh(&ifa->lock);
3334                 in6_ifa_put(ifa);
3335                 write_lock_bh(&idev->lock);
3336         }
3337
3338         while (!list_empty(&idev->addr_list)) {
3339                 ifa = list_first_entry(&idev->addr_list,
3340                                        struct inet6_ifaddr, if_list);
3341                 addrconf_del_dad_work(ifa);
3342
3343                 list_del(&ifa->if_list);
3344
3345                 write_unlock_bh(&idev->lock);
3346
3347                 spin_lock_bh(&ifa->lock);
3348                 state = ifa->state;
3349                 ifa->state = INET6_IFADDR_STATE_DEAD;
3350                 spin_unlock_bh(&ifa->lock);
3351
3352                 if (state != INET6_IFADDR_STATE_DEAD) {
3353                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3354                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3355                 }
3356                 in6_ifa_put(ifa);
3357
3358                 write_lock_bh(&idev->lock);
3359         }
3360
3361         write_unlock_bh(&idev->lock);
3362
3363         /* Step 5: Discard anycast and multicast list */
3364         if (how) {
3365                 ipv6_ac_destroy_dev(idev);
3366                 ipv6_mc_destroy_dev(idev);
3367         } else {
3368                 ipv6_mc_down(idev);
3369         }
3370
3371         idev->tstamp = jiffies;
3372
3373         /* Last: Shot the device (if unregistered) */
3374         if (how) {
3375                 addrconf_sysctl_unregister(idev);
3376                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3377                 neigh_ifdown(&nd_tbl, dev);
3378                 in6_dev_put(idev);
3379         }
3380         return 0;
3381 }
3382
3383 static void addrconf_rs_timer(unsigned long data)
3384 {
3385         struct inet6_dev *idev = (struct inet6_dev *)data;
3386         struct net_device *dev = idev->dev;
3387         struct in6_addr lladdr;
3388
3389         write_lock(&idev->lock);
3390         if (idev->dead || !(idev->if_flags & IF_READY))
3391                 goto out;
3392
3393         if (!ipv6_accept_ra(idev))
3394                 goto out;
3395
3396         /* Announcement received after solicitation was sent */
3397         if (idev->if_flags & IF_RA_RCVD)
3398                 goto out;
3399
3400         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3401                 write_unlock(&idev->lock);
3402                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3403                         ndisc_send_rs(dev, &lladdr,
3404                                       &in6addr_linklocal_allrouters);
3405                 else
3406                         goto put;
3407
3408                 write_lock(&idev->lock);
3409                 /* The wait after the last probe can be shorter */
3410                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3411                                              idev->cnf.rtr_solicits) ?
3412                                       idev->cnf.rtr_solicit_delay :
3413                                       idev->cnf.rtr_solicit_interval);
3414         } else {
3415                 /*
3416                  * Note: we do not support deprecated "all on-link"
3417                  * assumption any longer.
3418                  */
3419                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3420         }
3421
3422 out:
3423         write_unlock(&idev->lock);
3424 put:
3425         in6_dev_put(idev);
3426 }
3427
3428 /*
3429  *      Duplicate Address Detection
3430  */
3431 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3432 {
3433         unsigned long rand_num;
3434         struct inet6_dev *idev = ifp->idev;
3435
3436         if (ifp->flags & IFA_F_OPTIMISTIC)
3437                 rand_num = 0;
3438         else
3439                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3440
3441         ifp->dad_probes = idev->cnf.dad_transmits;
3442         addrconf_mod_dad_work(ifp, rand_num);
3443 }
3444
3445 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3446 {
3447         struct inet6_dev *idev = ifp->idev;
3448         struct net_device *dev = idev->dev;
3449
3450         addrconf_join_solict(dev, &ifp->addr);
3451
3452         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3453
3454         read_lock_bh(&idev->lock);
3455         spin_lock(&ifp->lock);
3456         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3457                 goto out;
3458
3459         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3460             idev->cnf.accept_dad < 1 ||
3461             !(ifp->flags&IFA_F_TENTATIVE) ||
3462             ifp->flags & IFA_F_NODAD) {
3463                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3464                 spin_unlock(&ifp->lock);
3465                 read_unlock_bh(&idev->lock);
3466
3467                 addrconf_dad_completed(ifp);
3468                 return;
3469         }
3470
3471         if (!(idev->if_flags & IF_READY)) {
3472                 spin_unlock(&ifp->lock);
3473                 read_unlock_bh(&idev->lock);
3474                 /*
3475                  * If the device is not ready:
3476                  * - keep it tentative if it is a permanent address.
3477                  * - otherwise, kill it.
3478                  */
3479                 in6_ifa_hold(ifp);
3480                 addrconf_dad_stop(ifp, 0);
3481                 return;
3482         }
3483
3484         /*
3485          * Optimistic nodes can start receiving
3486          * Frames right away
3487          */
3488         if (ifp->flags & IFA_F_OPTIMISTIC) {
3489                 ip6_ins_rt(ifp->rt);
3490                 if (ipv6_use_optimistic_addr(idev)) {
3491                         /* Because optimistic nodes can use this address,
3492                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3493                          */
3494                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3495                 }
3496         }
3497
3498         addrconf_dad_kick(ifp);
3499 out:
3500         spin_unlock(&ifp->lock);
3501         read_unlock_bh(&idev->lock);
3502 }
3503
3504 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3505 {
3506         bool begin_dad = false;
3507
3508         spin_lock_bh(&ifp->lock);
3509         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3510                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3511                 begin_dad = true;
3512         }
3513         spin_unlock_bh(&ifp->lock);
3514
3515         if (begin_dad)
3516                 addrconf_mod_dad_work(ifp, 0);
3517 }
3518
3519 static void addrconf_dad_work(struct work_struct *w)
3520 {
3521         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3522                                                 struct inet6_ifaddr,
3523                                                 dad_work);
3524         struct inet6_dev *idev = ifp->idev;
3525         struct in6_addr mcaddr;
3526
3527         enum {
3528                 DAD_PROCESS,
3529                 DAD_BEGIN,
3530                 DAD_ABORT,
3531         } action = DAD_PROCESS;
3532
3533         rtnl_lock();
3534
3535         spin_lock_bh(&ifp->lock);
3536         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3537                 action = DAD_BEGIN;
3538                 ifp->state = INET6_IFADDR_STATE_DAD;
3539         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3540                 action = DAD_ABORT;
3541                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3542         }
3543         spin_unlock_bh(&ifp->lock);
3544
3545         if (action == DAD_BEGIN) {
3546                 addrconf_dad_begin(ifp);
3547                 goto out;
3548         } else if (action == DAD_ABORT) {
3549                 addrconf_dad_stop(ifp, 1);
3550                 goto out;
3551         }
3552
3553         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3554                 goto out;
3555
3556         write_lock_bh(&idev->lock);
3557         if (idev->dead || !(idev->if_flags & IF_READY)) {
3558                 write_unlock_bh(&idev->lock);
3559                 goto out;
3560         }
3561
3562         spin_lock(&ifp->lock);
3563         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3564                 spin_unlock(&ifp->lock);
3565                 write_unlock_bh(&idev->lock);
3566                 goto out;
3567         }
3568
3569         if (ifp->dad_probes == 0) {
3570                 /*
3571                  * DAD was successful
3572                  */
3573
3574                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3575                 spin_unlock(&ifp->lock);
3576                 write_unlock_bh(&idev->lock);
3577
3578                 addrconf_dad_completed(ifp);
3579
3580                 goto out;
3581         }
3582
3583         ifp->dad_probes--;
3584         addrconf_mod_dad_work(ifp,
3585                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3586         spin_unlock(&ifp->lock);
3587         write_unlock_bh(&idev->lock);
3588
3589         /* send a neighbour solicitation for our addr */
3590         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3591         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3592 out:
3593         in6_ifa_put(ifp);
3594         rtnl_unlock();
3595 }
3596
3597 /* ifp->idev must be at least read locked */
3598 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3599 {
3600         struct inet6_ifaddr *ifpiter;
3601         struct inet6_dev *idev = ifp->idev;
3602
3603         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3604                 if (ifpiter->scope > IFA_LINK)
3605                         break;
3606                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3607                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3608                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3609                     IFA_F_PERMANENT)
3610                         return false;
3611         }
3612         return true;
3613 }
3614
3615 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3616 {
3617         struct net_device *dev = ifp->idev->dev;
3618         struct in6_addr lladdr;
3619         bool send_rs, send_mld;
3620
3621         addrconf_del_dad_work(ifp);
3622
3623         /*
3624          *      Configure the address for reception. Now it is valid.
3625          */
3626
3627         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3628
3629         /* If added prefix is link local and we are prepared to process
3630            router advertisements, start sending router solicitations.
3631          */
3632
3633         read_lock_bh(&ifp->idev->lock);
3634         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3635         send_rs = send_mld &&
3636                   ipv6_accept_ra(ifp->idev) &&
3637                   ifp->idev->cnf.rtr_solicits > 0 &&
3638                   (dev->flags&IFF_LOOPBACK) == 0;
3639         read_unlock_bh(&ifp->idev->lock);
3640
3641         /* While dad is in progress mld report's source address is in6_addrany.
3642          * Resend with proper ll now.
3643          */
3644         if (send_mld)
3645                 ipv6_mc_dad_complete(ifp->idev);
3646
3647         if (send_rs) {
3648                 /*
3649                  *      If a host as already performed a random delay
3650                  *      [...] as part of DAD [...] there is no need
3651                  *      to delay again before sending the first RS
3652                  */
3653                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3654                         return;
3655                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3656
3657                 write_lock_bh(&ifp->idev->lock);
3658                 spin_lock(&ifp->lock);
3659                 ifp->idev->rs_probes = 1;
3660                 ifp->idev->if_flags |= IF_RS_SENT;
3661                 addrconf_mod_rs_timer(ifp->idev,
3662                                       ifp->idev->cnf.rtr_solicit_interval);
3663                 spin_unlock(&ifp->lock);
3664                 write_unlock_bh(&ifp->idev->lock);
3665         }
3666 }
3667
3668 static void addrconf_dad_run(struct inet6_dev *idev)
3669 {
3670         struct inet6_ifaddr *ifp;
3671
3672         read_lock_bh(&idev->lock);
3673         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3674                 spin_lock(&ifp->lock);
3675                 if (ifp->flags & IFA_F_TENTATIVE &&
3676                     ifp->state == INET6_IFADDR_STATE_DAD)
3677                         addrconf_dad_kick(ifp);
3678                 spin_unlock(&ifp->lock);
3679         }
3680         read_unlock_bh(&idev->lock);
3681 }
3682
3683 #ifdef CONFIG_PROC_FS
3684 struct if6_iter_state {
3685         struct seq_net_private p;
3686         int bucket;
3687         int offset;
3688 };
3689
3690 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3691 {
3692         struct inet6_ifaddr *ifa = NULL;
3693         struct if6_iter_state *state = seq->private;
3694         struct net *net = seq_file_net(seq);
3695         int p = 0;
3696
3697         /* initial bucket if pos is 0 */
3698         if (pos == 0) {
3699                 state->bucket = 0;
3700                 state->offset = 0;
3701         }
3702
3703         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3704                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3705                                          addr_lst) {
3706                         if (!net_eq(dev_net(ifa->idev->dev), net))
3707                                 continue;
3708                         /* sync with offset */
3709                         if (p < state->offset) {
3710                                 p++;
3711                                 continue;
3712                         }
3713                         state->offset++;
3714                         return ifa;
3715                 }
3716
3717                 /* prepare for next bucket */
3718                 state->offset = 0;
3719                 p = 0;
3720         }
3721         return NULL;
3722 }
3723
3724 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3725                                          struct inet6_ifaddr *ifa)
3726 {
3727         struct if6_iter_state *state = seq->private;
3728         struct net *net = seq_file_net(seq);
3729
3730         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3731                 if (!net_eq(dev_net(ifa->idev->dev), net))
3732                         continue;
3733                 state->offset++;
3734                 return ifa;
3735         }
3736
3737         while (++state->bucket < IN6_ADDR_HSIZE) {
3738                 state->offset = 0;
3739                 hlist_for_each_entry_rcu_bh(ifa,
3740                                      &inet6_addr_lst[state->bucket], addr_lst) {
3741                         if (!net_eq(dev_net(ifa->idev->dev), net))
3742                                 continue;
3743                         state->offset++;
3744                         return ifa;
3745                 }
3746         }
3747
3748         return NULL;
3749 }
3750
3751 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3752         __acquires(rcu_bh)
3753 {
3754         rcu_read_lock_bh();
3755         return if6_get_first(seq, *pos);
3756 }
3757
3758 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3759 {
3760         struct inet6_ifaddr *ifa;
3761
3762         ifa = if6_get_next(seq, v);
3763         ++*pos;
3764         return ifa;
3765 }
3766
3767 static void if6_seq_stop(struct seq_file *seq, void *v)
3768         __releases(rcu_bh)
3769 {
3770         rcu_read_unlock_bh();
3771 }
3772
3773 static int if6_seq_show(struct seq_file *seq, void *v)
3774 {
3775         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3776         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3777                    &ifp->addr,
3778                    ifp->idev->dev->ifindex,
3779                    ifp->prefix_len,
3780                    ifp->scope,
3781                    (u8) ifp->flags,
3782                    ifp->idev->dev->name);
3783         return 0;
3784 }
3785
3786 static const struct seq_operations if6_seq_ops = {
3787         .start  = if6_seq_start,
3788         .next   = if6_seq_next,
3789         .show   = if6_seq_show,
3790         .stop   = if6_seq_stop,
3791 };
3792
3793 static int if6_seq_open(struct inode *inode, struct file *file)
3794 {
3795         return seq_open_net(inode, file, &if6_seq_ops,
3796                             sizeof(struct if6_iter_state));
3797 }
3798
3799 static const struct file_operations if6_fops = {
3800         .owner          = THIS_MODULE,
3801         .open           = if6_seq_open,
3802         .read           = seq_read,
3803         .llseek         = seq_lseek,
3804         .release        = seq_release_net,
3805 };
3806
3807 static int __net_init if6_proc_net_init(struct net *net)
3808 {
3809         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3810                 return -ENOMEM;
3811         return 0;
3812 }
3813
3814 static void __net_exit if6_proc_net_exit(struct net *net)
3815 {
3816         remove_proc_entry("if_inet6", net->proc_net);
3817 }
3818
3819 static struct pernet_operations if6_proc_net_ops = {
3820         .init = if6_proc_net_init,
3821         .exit = if6_proc_net_exit,
3822 };
3823
3824 int __init if6_proc_init(void)
3825 {
3826         return register_pernet_subsys(&if6_proc_net_ops);
3827 }
3828
3829 void if6_proc_exit(void)
3830 {
3831         unregister_pernet_subsys(&if6_proc_net_ops);
3832 }
3833 #endif  /* CONFIG_PROC_FS */
3834
3835 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3836 /* Check if address is a home address configured on any interface. */
3837 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3838 {
3839         int ret = 0;
3840         struct inet6_ifaddr *ifp = NULL;
3841         unsigned int hash = inet6_addr_hash(addr);
3842
3843         rcu_read_lock_bh();
3844         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3845                 if (!net_eq(dev_net(ifp->idev->dev), net))
3846                         continue;
3847                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3848                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3849                         ret = 1;
3850                         break;
3851                 }
3852         }
3853         rcu_read_unlock_bh();
3854         return ret;
3855 }
3856 #endif
3857
3858 /*
3859  *      Periodic address status verification
3860  */
3861
3862 static void addrconf_verify_rtnl(void)
3863 {
3864         unsigned long now, next, next_sec, next_sched;
3865         struct inet6_ifaddr *ifp;
3866         int i;
3867
3868         ASSERT_RTNL();
3869
3870         rcu_read_lock_bh();
3871         now = jiffies;
3872         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3873
3874         cancel_delayed_work(&addr_chk_work);
3875
3876         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3877 restart:
3878                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3879                         unsigned long age;
3880
3881                         /* When setting preferred_lft to a value not zero or
3882                          * infinity, while valid_lft is infinity
3883                          * IFA_F_PERMANENT has a non-infinity life time.
3884                          */
3885                         if ((ifp->flags & IFA_F_PERMANENT) &&
3886                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3887                                 continue;
3888
3889                         spin_lock(&ifp->lock);
3890                         /* We try to batch several events at once. */
3891                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3892
3893                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3894                             age >= ifp->valid_lft) {
3895                                 spin_unlock(&ifp->lock);
3896                                 in6_ifa_hold(ifp);
3897                                 ipv6_del_addr(ifp);
3898                                 goto restart;
3899                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3900                                 spin_unlock(&ifp->lock);
3901                                 continue;
3902                         } else if (age >= ifp->prefered_lft) {
3903                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3904                                 int deprecate = 0;
3905
3906                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3907                                         deprecate = 1;
3908                                         ifp->flags |= IFA_F_DEPRECATED;
3909                                 }
3910
3911                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3912                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3913                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3914
3915                                 spin_unlock(&ifp->lock);
3916
3917                                 if (deprecate) {
3918                                         in6_ifa_hold(ifp);
3919
3920                                         ipv6_ifa_notify(0, ifp);
3921                                         in6_ifa_put(ifp);
3922                                         goto restart;
3923                                 }
3924                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3925                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3926                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3927                                         ifp->idev->cnf.dad_transmits *
3928                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3929
3930                                 if (age >= ifp->prefered_lft - regen_advance) {
3931                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3932                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3933                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3934                                         if (!ifp->regen_count && ifpub) {
3935                                                 ifp->regen_count++;
3936                                                 in6_ifa_hold(ifp);
3937                                                 in6_ifa_hold(ifpub);
3938                                                 spin_unlock(&ifp->lock);
3939
3940                                                 spin_lock(&ifpub->lock);
3941                                                 ifpub->regen_count = 0;
3942                                                 spin_unlock(&ifpub->lock);
3943                                                 ipv6_create_tempaddr(ifpub, ifp);
3944                                                 in6_ifa_put(ifpub);
3945                                                 in6_ifa_put(ifp);
3946                                                 goto restart;
3947                                         }
3948                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3949                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3950                                 spin_unlock(&ifp->lock);
3951                         } else {
3952                                 /* ifp->prefered_lft <= ifp->valid_lft */
3953                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3954                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3955                                 spin_unlock(&ifp->lock);
3956                         }
3957                 }
3958         }
3959
3960         next_sec = round_jiffies_up(next);
3961         next_sched = next;
3962
3963         /* If rounded timeout is accurate enough, accept it. */
3964         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3965                 next_sched = next_sec;
3966
3967         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3968         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3969                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3970
3971         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3972               now, next, next_sec, next_sched);
3973         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3974         rcu_read_unlock_bh();
3975 }
3976
3977 static void addrconf_verify_work(struct work_struct *w)
3978 {
3979         rtnl_lock();
3980         addrconf_verify_rtnl();
3981         rtnl_unlock();
3982 }
3983
3984 static void addrconf_verify(void)
3985 {
3986         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3987 }
3988
3989 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3990                                      struct in6_addr **peer_pfx)
3991 {
3992         struct in6_addr *pfx = NULL;
3993
3994         *peer_pfx = NULL;
3995
3996         if (addr)
3997                 pfx = nla_data(addr);
3998
3999         if (local) {
4000                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4001                         *peer_pfx = pfx;
4002                 pfx = nla_data(local);
4003         }
4004
4005         return pfx;
4006 }
4007
4008 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4009         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4010         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4011         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4012         [IFA_FLAGS]             = { .len = sizeof(u32) },
4013 };
4014
4015 static int
4016 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4017 {
4018         struct net *net = sock_net(skb->sk);
4019         struct ifaddrmsg *ifm;
4020         struct nlattr *tb[IFA_MAX+1];
4021         struct in6_addr *pfx, *peer_pfx;
4022         u32 ifa_flags;
4023         int err;
4024
4025         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4026         if (err < 0)
4027                 return err;
4028
4029         ifm = nlmsg_data(nlh);
4030         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4031         if (!pfx)
4032                 return -EINVAL;
4033
4034         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4035
4036         /* We ignore other flags so far. */
4037         ifa_flags &= IFA_F_MANAGETEMPADDR;
4038
4039         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4040                               ifm->ifa_prefixlen);
4041 }
4042
4043 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4044                              u32 prefered_lft, u32 valid_lft)
4045 {
4046         u32 flags;
4047         clock_t expires;
4048         unsigned long timeout;
4049         bool was_managetempaddr;
4050         bool had_prefixroute;
4051
4052         ASSERT_RTNL();
4053
4054         if (!valid_lft || (prefered_lft > valid_lft))
4055                 return -EINVAL;
4056
4057         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4058             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4059                 return -EINVAL;
4060
4061         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4062         if (addrconf_finite_timeout(timeout)) {
4063                 expires = jiffies_to_clock_t(timeout * HZ);
4064                 valid_lft = timeout;
4065                 flags = RTF_EXPIRES;
4066         } else {
4067                 expires = 0;
4068                 flags = 0;
4069                 ifa_flags |= IFA_F_PERMANENT;
4070         }
4071
4072         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4073         if (addrconf_finite_timeout(timeout)) {
4074                 if (timeout == 0)
4075                         ifa_flags |= IFA_F_DEPRECATED;
4076                 prefered_lft = timeout;
4077         }
4078
4079         spin_lock_bh(&ifp->lock);
4080         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4081         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4082                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4083         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4084                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4085                         IFA_F_NOPREFIXROUTE);
4086         ifp->flags |= ifa_flags;
4087         ifp->tstamp = jiffies;
4088         ifp->valid_lft = valid_lft;
4089         ifp->prefered_lft = prefered_lft;
4090
4091         spin_unlock_bh(&ifp->lock);
4092         if (!(ifp->flags&IFA_F_TENTATIVE))
4093                 ipv6_ifa_notify(0, ifp);
4094
4095         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4096                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4097                                       expires, flags);
4098         } else if (had_prefixroute) {
4099                 enum cleanup_prefix_rt_t action;
4100                 unsigned long rt_expires;
4101
4102                 write_lock_bh(&ifp->idev->lock);
4103                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4104                 write_unlock_bh(&ifp->idev->lock);
4105
4106                 if (action != CLEANUP_PREFIX_RT_NOP) {
4107                         cleanup_prefix_route(ifp, rt_expires,
4108                                 action == CLEANUP_PREFIX_RT_DEL);
4109                 }
4110         }
4111
4112         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4113                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4114                         valid_lft = prefered_lft = 0;
4115                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4116                                  !was_managetempaddr, jiffies);
4117         }
4118
4119         addrconf_verify_rtnl();
4120
4121         return 0;
4122 }
4123
4124 static int
4125 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4126 {
4127         struct net *net = sock_net(skb->sk);
4128         struct ifaddrmsg *ifm;
4129         struct nlattr *tb[IFA_MAX+1];
4130         struct in6_addr *pfx, *peer_pfx;
4131         struct inet6_ifaddr *ifa;
4132         struct net_device *dev;
4133         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4134         u32 ifa_flags;
4135         int err;
4136
4137         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4138         if (err < 0)
4139                 return err;
4140
4141         ifm = nlmsg_data(nlh);
4142         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4143         if (!pfx)
4144                 return -EINVAL;
4145
4146         if (tb[IFA_CACHEINFO]) {
4147                 struct ifa_cacheinfo *ci;
4148
4149                 ci = nla_data(tb[IFA_CACHEINFO]);
4150                 valid_lft = ci->ifa_valid;
4151                 preferred_lft = ci->ifa_prefered;
4152         } else {
4153                 preferred_lft = INFINITY_LIFE_TIME;
4154                 valid_lft = INFINITY_LIFE_TIME;
4155         }
4156
4157         dev =  __dev_get_by_index(net, ifm->ifa_index);
4158         if (!dev)
4159                 return -ENODEV;
4160
4161         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4162
4163         /* We ignore other flags so far. */
4164         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4165                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4166
4167         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4168         if (!ifa) {
4169                 /*
4170                  * It would be best to check for !NLM_F_CREATE here but
4171                  * userspace already relies on not having to provide this.
4172                  */
4173                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4174                                       ifm->ifa_prefixlen, ifa_flags,
4175                                       preferred_lft, valid_lft);
4176         }
4177
4178         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4179             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4180                 err = -EEXIST;
4181         else
4182                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4183
4184         in6_ifa_put(ifa);
4185
4186         return err;
4187 }
4188
4189 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4190                           u8 scope, int ifindex)
4191 {
4192         struct ifaddrmsg *ifm;
4193
4194         ifm = nlmsg_data(nlh);
4195         ifm->ifa_family = AF_INET6;
4196         ifm->ifa_prefixlen = prefixlen;
4197         ifm->ifa_flags = flags;
4198         ifm->ifa_scope = scope;
4199         ifm->ifa_index = ifindex;
4200 }
4201
4202 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4203                          unsigned long tstamp, u32 preferred, u32 valid)
4204 {
4205         struct ifa_cacheinfo ci;
4206
4207         ci.cstamp = cstamp_delta(cstamp);
4208         ci.tstamp = cstamp_delta(tstamp);
4209         ci.ifa_prefered = preferred;
4210         ci.ifa_valid = valid;
4211
4212         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4213 }
4214
4215 static inline int rt_scope(int ifa_scope)
4216 {
4217         if (ifa_scope & IFA_HOST)
4218                 return RT_SCOPE_HOST;
4219         else if (ifa_scope & IFA_LINK)
4220                 return RT_SCOPE_LINK;
4221         else if (ifa_scope & IFA_SITE)
4222                 return RT_SCOPE_SITE;
4223         else
4224                 return RT_SCOPE_UNIVERSE;
4225 }
4226
4227 static inline int inet6_ifaddr_msgsize(void)
4228 {
4229         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4230                + nla_total_size(16) /* IFA_LOCAL */
4231                + nla_total_size(16) /* IFA_ADDRESS */
4232                + nla_total_size(sizeof(struct ifa_cacheinfo))
4233                + nla_total_size(4)  /* IFA_FLAGS */;
4234 }
4235
4236 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4237                              u32 portid, u32 seq, int event, unsigned int flags)
4238 {
4239         struct nlmsghdr  *nlh;
4240         u32 preferred, valid;
4241
4242         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4243         if (!nlh)
4244                 return -EMSGSIZE;
4245
4246         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4247                       ifa->idev->dev->ifindex);
4248
4249         if (!((ifa->flags&IFA_F_PERMANENT) &&
4250               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4251                 preferred = ifa->prefered_lft;
4252                 valid = ifa->valid_lft;
4253                 if (preferred != INFINITY_LIFE_TIME) {
4254                         long tval = (jiffies - ifa->tstamp)/HZ;
4255                         if (preferred > tval)
4256                                 preferred -= tval;
4257                         else
4258                                 preferred = 0;
4259                         if (valid != INFINITY_LIFE_TIME) {
4260                                 if (valid > tval)
4261                                         valid -= tval;
4262                                 else
4263                                         valid = 0;
4264                         }
4265                 }
4266         } else {
4267                 preferred = INFINITY_LIFE_TIME;
4268                 valid = INFINITY_LIFE_TIME;
4269         }
4270
4271         if (!ipv6_addr_any(&ifa->peer_addr)) {
4272                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4273                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4274                         goto error;
4275         } else
4276                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4277                         goto error;
4278
4279         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4280                 goto error;
4281
4282         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4283                 goto error;
4284
4285         nlmsg_end(skb, nlh);
4286         return 0;
4287
4288 error:
4289         nlmsg_cancel(skb, nlh);
4290         return -EMSGSIZE;
4291 }
4292
4293 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4294                                 u32 portid, u32 seq, int event, u16 flags)
4295 {
4296         struct nlmsghdr  *nlh;
4297         u8 scope = RT_SCOPE_UNIVERSE;
4298         int ifindex = ifmca->idev->dev->ifindex;
4299
4300         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4301                 scope = RT_SCOPE_SITE;
4302
4303         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4304         if (!nlh)
4305                 return -EMSGSIZE;
4306
4307         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4308         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4309             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4310                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4311                 nlmsg_cancel(skb, nlh);
4312                 return -EMSGSIZE;
4313         }
4314
4315         nlmsg_end(skb, nlh);
4316         return 0;
4317 }
4318
4319 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4320                                 u32 portid, u32 seq, int event, unsigned int flags)
4321 {
4322         struct nlmsghdr  *nlh;
4323         u8 scope = RT_SCOPE_UNIVERSE;
4324         int ifindex = ifaca->aca_idev->dev->ifindex;
4325
4326         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4327                 scope = RT_SCOPE_SITE;
4328
4329         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4330         if (!nlh)
4331                 return -EMSGSIZE;
4332
4333         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4334         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4335             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4336                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4337                 nlmsg_cancel(skb, nlh);
4338                 return -EMSGSIZE;
4339         }
4340
4341         nlmsg_end(skb, nlh);
4342         return 0;
4343 }
4344
4345 enum addr_type_t {
4346         UNICAST_ADDR,
4347         MULTICAST_ADDR,
4348         ANYCAST_ADDR,
4349 };
4350
4351 /* called with rcu_read_lock() */
4352 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4353                           struct netlink_callback *cb, enum addr_type_t type,
4354                           int s_ip_idx, int *p_ip_idx)
4355 {
4356         struct ifmcaddr6 *ifmca;
4357         struct ifacaddr6 *ifaca;
4358         int err = 1;
4359         int ip_idx = *p_ip_idx;
4360
4361         read_lock_bh(&idev->lock);
4362         switch (type) {
4363         case UNICAST_ADDR: {
4364                 struct inet6_ifaddr *ifa;
4365
4366                 /* unicast address incl. temp addr */
4367                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4368                         if (++ip_idx < s_ip_idx)
4369                                 continue;
4370                         err = inet6_fill_ifaddr(skb, ifa,
4371                                                 NETLINK_CB(cb->skb).portid,
4372                                                 cb->nlh->nlmsg_seq,
4373                                                 RTM_NEWADDR,
4374                                                 NLM_F_MULTI);
4375                         if (err < 0)
4376                                 break;
4377                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4378                 }
4379                 break;
4380         }
4381         case MULTICAST_ADDR:
4382                 /* multicast address */
4383                 for (ifmca = idev->mc_list; ifmca;
4384                      ifmca = ifmca->next, ip_idx++) {
4385                         if (ip_idx < s_ip_idx)
4386                                 continue;
4387                         err = inet6_fill_ifmcaddr(skb, ifmca,
4388                                                   NETLINK_CB(cb->skb).portid,
4389                                                   cb->nlh->nlmsg_seq,
4390                                                   RTM_GETMULTICAST,
4391                                                   NLM_F_MULTI);
4392                         if (err < 0)
4393                                 break;
4394                 }
4395                 break;
4396         case ANYCAST_ADDR:
4397                 /* anycast address */
4398                 for (ifaca = idev->ac_list; ifaca;
4399                      ifaca = ifaca->aca_next, ip_idx++) {
4400                         if (ip_idx < s_ip_idx)
4401                                 continue;
4402                         err = inet6_fill_ifacaddr(skb, ifaca,
4403                                                   NETLINK_CB(cb->skb).portid,
4404                                                   cb->nlh->nlmsg_seq,
4405                                                   RTM_GETANYCAST,
4406                                                   NLM_F_MULTI);
4407                         if (err < 0)
4408                                 break;
4409                 }
4410                 break;
4411         default:
4412                 break;
4413         }
4414         read_unlock_bh(&idev->lock);
4415         *p_ip_idx = ip_idx;
4416         return err;
4417 }
4418
4419 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4420                            enum addr_type_t type)
4421 {
4422         struct net *net = sock_net(skb->sk);
4423         int h, s_h;
4424         int idx, ip_idx;
4425         int s_idx, s_ip_idx;
4426         struct net_device *dev;
4427         struct inet6_dev *idev;
4428         struct hlist_head *head;
4429
4430         s_h = cb->args[0];
4431         s_idx = idx = cb->args[1];
4432         s_ip_idx = ip_idx = cb->args[2];
4433
4434         rcu_read_lock();
4435         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4436         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4437                 idx = 0;
4438                 head = &net->dev_index_head[h];
4439                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4440                         if (idx < s_idx)
4441                                 goto cont;
4442                         if (h > s_h || idx > s_idx)
4443                                 s_ip_idx = 0;
4444                         ip_idx = 0;
4445                         idev = __in6_dev_get(dev);
4446                         if (!idev)
4447                                 goto cont;
4448
4449                         if (in6_dump_addrs(idev, skb, cb, type,
4450                                            s_ip_idx, &ip_idx) < 0)
4451                                 goto done;
4452 cont:
4453                         idx++;
4454                 }
4455         }
4456 done:
4457         rcu_read_unlock();
4458         cb->args[0] = h;
4459         cb->args[1] = idx;
4460         cb->args[2] = ip_idx;
4461
4462         return skb->len;
4463 }
4464
4465 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4466 {
4467         enum addr_type_t type = UNICAST_ADDR;
4468
4469         return inet6_dump_addr(skb, cb, type);
4470 }
4471
4472 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4473 {
4474         enum addr_type_t type = MULTICAST_ADDR;
4475
4476         return inet6_dump_addr(skb, cb, type);
4477 }
4478
4479
4480 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4481 {
4482         enum addr_type_t type = ANYCAST_ADDR;
4483
4484         return inet6_dump_addr(skb, cb, type);
4485 }
4486
4487 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4488 {
4489         struct net *net = sock_net(in_skb->sk);
4490         struct ifaddrmsg *ifm;
4491         struct nlattr *tb[IFA_MAX+1];
4492         struct in6_addr *addr = NULL, *peer;
4493         struct net_device *dev = NULL;
4494         struct inet6_ifaddr *ifa;
4495         struct sk_buff *skb;
4496         int err;
4497
4498         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4499         if (err < 0)
4500                 goto errout;
4501
4502         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4503         if (!addr) {
4504                 err = -EINVAL;
4505                 goto errout;
4506         }
4507
4508         ifm = nlmsg_data(nlh);
4509         if (ifm->ifa_index)
4510                 dev = __dev_get_by_index(net, ifm->ifa_index);
4511
4512         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4513         if (!ifa) {
4514                 err = -EADDRNOTAVAIL;
4515                 goto errout;
4516         }
4517
4518         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4519         if (!skb) {
4520                 err = -ENOBUFS;
4521                 goto errout_ifa;
4522         }
4523
4524         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4525                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4526         if (err < 0) {
4527                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4528                 WARN_ON(err == -EMSGSIZE);
4529                 kfree_skb(skb);
4530                 goto errout_ifa;
4531         }
4532         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4533 errout_ifa:
4534         in6_ifa_put(ifa);
4535 errout:
4536         return err;
4537 }
4538
4539 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4540 {
4541         struct sk_buff *skb;
4542         struct net *net = dev_net(ifa->idev->dev);
4543         int err = -ENOBUFS;
4544
4545         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4546         if (!skb)
4547                 goto errout;
4548
4549         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4550         if (err < 0) {
4551                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4552                 WARN_ON(err == -EMSGSIZE);
4553                 kfree_skb(skb);
4554                 goto errout;
4555         }
4556         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4557         return;
4558 errout:
4559         if (err < 0)
4560                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4561 }
4562
4563 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4564                                 __s32 *array, int bytes)
4565 {
4566         BUG_ON(bytes < (DEVCONF_MAX * 4));
4567
4568         memset(array, 0, bytes);
4569         array[DEVCONF_FORWARDING] = cnf->forwarding;
4570         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4571         array[DEVCONF_MTU6] = cnf->mtu6;
4572         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4573         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4574         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4575         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4576         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4577         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4578                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4579         array[DEVCONF_RTR_SOLICIT_DELAY] =
4580                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4581         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4582         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4583                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4584         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4585                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4586         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4587         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4588         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4589         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4590         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4591         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4592         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4593         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4594         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4595 #ifdef CONFIG_IPV6_ROUTER_PREF
4596         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4597         array[DEVCONF_RTR_PROBE_INTERVAL] =
4598                 jiffies_to_msecs(cnf->rtr_probe_interval);
4599 #ifdef CONFIG_IPV6_ROUTE_INFO
4600         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4601 #endif
4602 #endif
4603         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4604         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4605 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4606         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4607         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4608 #endif
4609 #ifdef CONFIG_IPV6_MROUTE
4610         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4611 #endif
4612         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4613         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4614         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4615         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4616         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4617         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4618         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4619         /* we omit DEVCONF_STABLE_SECRET for now */
4620         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4621 }
4622
4623 static inline size_t inet6_ifla6_size(void)
4624 {
4625         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4626              + nla_total_size(sizeof(struct ifla_cacheinfo))
4627              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4628              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4629              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4630              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4631 }
4632
4633 static inline size_t inet6_if_nlmsg_size(void)
4634 {
4635         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4636                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4637                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4638                + nla_total_size(4) /* IFLA_MTU */
4639                + nla_total_size(4) /* IFLA_LINK */
4640                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4641 }
4642
4643 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4644                                       int items, int bytes)
4645 {
4646         int i;
4647         int pad = bytes - sizeof(u64) * items;
4648         BUG_ON(pad < 0);
4649
4650         /* Use put_unaligned() because stats may not be aligned for u64. */
4651         put_unaligned(items, &stats[0]);
4652         for (i = 1; i < items; i++)
4653                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4654
4655         memset(&stats[items], 0, pad);
4656 }
4657
4658 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4659                                       int items, int bytes, size_t syncpoff)
4660 {
4661         int i;
4662         int pad = bytes - sizeof(u64) * items;
4663         BUG_ON(pad < 0);
4664
4665         /* Use put_unaligned() because stats may not be aligned for u64. */
4666         put_unaligned(items, &stats[0]);
4667         for (i = 1; i < items; i++)
4668                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4669
4670         memset(&stats[items], 0, pad);
4671 }
4672
4673 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4674                              int bytes)
4675 {
4676         switch (attrtype) {
4677         case IFLA_INET6_STATS:
4678                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4679                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4680                 break;
4681         case IFLA_INET6_ICMP6STATS:
4682                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4683                 break;
4684         }
4685 }
4686
4687 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4688 {
4689         struct nlattr *nla;
4690         struct ifla_cacheinfo ci;
4691
4692         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4693                 goto nla_put_failure;
4694         ci.max_reasm_len = IPV6_MAXPLEN;
4695         ci.tstamp = cstamp_delta(idev->tstamp);
4696         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4697         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4698         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4699                 goto nla_put_failure;
4700         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4701         if (!nla)
4702                 goto nla_put_failure;
4703         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4704
4705         /* XXX - MC not implemented */
4706
4707         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4708         if (!nla)
4709                 goto nla_put_failure;
4710         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4711
4712         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4713         if (!nla)
4714                 goto nla_put_failure;
4715         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4716
4717         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4718         if (!nla)
4719                 goto nla_put_failure;
4720
4721         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4722                 goto nla_put_failure;
4723
4724         read_lock_bh(&idev->lock);
4725         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4726         read_unlock_bh(&idev->lock);
4727
4728         return 0;
4729
4730 nla_put_failure:
4731         return -EMSGSIZE;
4732 }
4733
4734 static size_t inet6_get_link_af_size(const struct net_device *dev)
4735 {
4736         if (!__in6_dev_get(dev))
4737                 return 0;
4738
4739         return inet6_ifla6_size();
4740 }
4741
4742 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4743 {
4744         struct inet6_dev *idev = __in6_dev_get(dev);
4745
4746         if (!idev)
4747                 return -ENODATA;
4748
4749         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4750                 return -EMSGSIZE;
4751
4752         return 0;
4753 }
4754
4755 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4756 {
4757         struct inet6_ifaddr *ifp;
4758         struct net_device *dev = idev->dev;
4759         bool update_rs = false;
4760         struct in6_addr ll_addr;
4761
4762         ASSERT_RTNL();
4763
4764         if (!token)
4765                 return -EINVAL;
4766         if (ipv6_addr_any(token))
4767                 return -EINVAL;
4768         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4769                 return -EINVAL;
4770         if (!ipv6_accept_ra(idev))
4771                 return -EINVAL;
4772         if (idev->cnf.rtr_solicits <= 0)
4773                 return -EINVAL;
4774
4775         write_lock_bh(&idev->lock);
4776
4777         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4778         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4779
4780         write_unlock_bh(&idev->lock);
4781
4782         if (!idev->dead && (idev->if_flags & IF_READY) &&
4783             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4784                              IFA_F_OPTIMISTIC)) {
4785
4786                 /* If we're not ready, then normal ifup will take care
4787                  * of this. Otherwise, we need to request our rs here.
4788                  */
4789                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4790                 update_rs = true;
4791         }
4792
4793         write_lock_bh(&idev->lock);
4794
4795         if (update_rs) {
4796                 idev->if_flags |= IF_RS_SENT;
4797                 idev->rs_probes = 1;
4798                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4799         }
4800
4801         /* Well, that's kinda nasty ... */
4802         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4803                 spin_lock(&ifp->lock);
4804                 if (ifp->tokenized) {
4805                         ifp->valid_lft = 0;
4806                         ifp->prefered_lft = 0;
4807                 }
4808                 spin_unlock(&ifp->lock);
4809         }
4810
4811         write_unlock_bh(&idev->lock);
4812         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4813         addrconf_verify_rtnl();
4814         return 0;
4815 }
4816
4817 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4818         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4819         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4820 };
4821
4822 static int inet6_validate_link_af(const struct net_device *dev,
4823                                   const struct nlattr *nla)
4824 {
4825         struct nlattr *tb[IFLA_INET6_MAX + 1];
4826
4827         if (dev && !__in6_dev_get(dev))
4828                 return -EAFNOSUPPORT;
4829
4830         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4831 }
4832
4833 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4834 {
4835         int err = -EINVAL;
4836         struct inet6_dev *idev = __in6_dev_get(dev);
4837         struct nlattr *tb[IFLA_INET6_MAX + 1];
4838
4839         if (!idev)
4840                 return -EAFNOSUPPORT;
4841
4842         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4843                 BUG();
4844
4845         if (tb[IFLA_INET6_TOKEN]) {
4846                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4847                 if (err)
4848                         return err;
4849         }
4850
4851         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4852                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4853
4854                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4855                     mode != IN6_ADDR_GEN_MODE_NONE &&
4856                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4857                         return -EINVAL;
4858
4859                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4860                     !idev->cnf.stable_secret.initialized &&
4861                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4862                         return -EINVAL;
4863
4864                 idev->addr_gen_mode = mode;
4865                 err = 0;
4866         }
4867
4868         return err;
4869 }
4870
4871 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4872                              u32 portid, u32 seq, int event, unsigned int flags)
4873 {
4874         struct net_device *dev = idev->dev;
4875         struct ifinfomsg *hdr;
4876         struct nlmsghdr *nlh;
4877         void *protoinfo;
4878
4879         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4880         if (!nlh)
4881                 return -EMSGSIZE;
4882
4883         hdr = nlmsg_data(nlh);
4884         hdr->ifi_family = AF_INET6;
4885         hdr->__ifi_pad = 0;
4886         hdr->ifi_type = dev->type;
4887         hdr->ifi_index = dev->ifindex;
4888         hdr->ifi_flags = dev_get_flags(dev);
4889         hdr->ifi_change = 0;
4890
4891         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4892             (dev->addr_len &&
4893              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4894             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4895             (dev->ifindex != dev_get_iflink(dev) &&
4896              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4897                 goto nla_put_failure;
4898         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4899         if (!protoinfo)
4900                 goto nla_put_failure;
4901
4902         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4903                 goto nla_put_failure;
4904
4905         nla_nest_end(skb, protoinfo);
4906         nlmsg_end(skb, nlh);
4907         return 0;
4908
4909 nla_put_failure:
4910         nlmsg_cancel(skb, nlh);
4911         return -EMSGSIZE;
4912 }
4913
4914 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4915 {
4916         struct net *net = sock_net(skb->sk);
4917         int h, s_h;
4918         int idx = 0, s_idx;
4919         struct net_device *dev;
4920         struct inet6_dev *idev;
4921         struct hlist_head *head;
4922
4923         s_h = cb->args[0];
4924         s_idx = cb->args[1];
4925
4926         rcu_read_lock();
4927         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4928                 idx = 0;
4929                 head = &net->dev_index_head[h];
4930                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4931                         if (idx < s_idx)
4932                                 goto cont;
4933                         idev = __in6_dev_get(dev);
4934                         if (!idev)
4935                                 goto cont;
4936                         if (inet6_fill_ifinfo(skb, idev,
4937                                               NETLINK_CB(cb->skb).portid,
4938                                               cb->nlh->nlmsg_seq,
4939                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
4940                                 goto out;
4941 cont:
4942                         idx++;
4943                 }
4944         }
4945 out:
4946         rcu_read_unlock();
4947         cb->args[1] = idx;
4948         cb->args[0] = h;
4949
4950         return skb->len;
4951 }
4952
4953 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4954 {
4955         struct sk_buff *skb;
4956         struct net *net = dev_net(idev->dev);
4957         int err = -ENOBUFS;
4958
4959         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4960         if (!skb)
4961                 goto errout;
4962
4963         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4964         if (err < 0) {
4965                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4966                 WARN_ON(err == -EMSGSIZE);
4967                 kfree_skb(skb);
4968                 goto errout;
4969         }
4970         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4971         return;
4972 errout:
4973         if (err < 0)
4974                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4975 }
4976
4977 static inline size_t inet6_prefix_nlmsg_size(void)
4978 {
4979         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4980                + nla_total_size(sizeof(struct in6_addr))
4981                + nla_total_size(sizeof(struct prefix_cacheinfo));
4982 }
4983
4984 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4985                              struct prefix_info *pinfo, u32 portid, u32 seq,
4986                              int event, unsigned int flags)
4987 {
4988         struct prefixmsg *pmsg;
4989         struct nlmsghdr *nlh;
4990         struct prefix_cacheinfo ci;
4991
4992         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4993         if (!nlh)
4994                 return -EMSGSIZE;
4995
4996         pmsg = nlmsg_data(nlh);
4997         pmsg->prefix_family = AF_INET6;
4998         pmsg->prefix_pad1 = 0;
4999         pmsg->prefix_pad2 = 0;
5000         pmsg->prefix_ifindex = idev->dev->ifindex;
5001         pmsg->prefix_len = pinfo->prefix_len;
5002         pmsg->prefix_type = pinfo->type;
5003         pmsg->prefix_pad3 = 0;
5004         pmsg->prefix_flags = 0;
5005         if (pinfo->onlink)
5006                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5007         if (pinfo->autoconf)
5008                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5009
5010         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5011                 goto nla_put_failure;
5012         ci.preferred_time = ntohl(pinfo->prefered);
5013         ci.valid_time = ntohl(pinfo->valid);
5014         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5015                 goto nla_put_failure;
5016         nlmsg_end(skb, nlh);
5017         return 0;
5018
5019 nla_put_failure:
5020         nlmsg_cancel(skb, nlh);
5021         return -EMSGSIZE;
5022 }
5023
5024 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5025                          struct prefix_info *pinfo)
5026 {
5027         struct sk_buff *skb;
5028         struct net *net = dev_net(idev->dev);
5029         int err = -ENOBUFS;
5030
5031         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5032         if (!skb)
5033                 goto errout;
5034
5035         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5036         if (err < 0) {
5037                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5038                 WARN_ON(err == -EMSGSIZE);
5039                 kfree_skb(skb);
5040                 goto errout;
5041         }
5042         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5043         return;
5044 errout:
5045         if (err < 0)
5046                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5047 }
5048
5049 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5050 {
5051         struct net *net = dev_net(ifp->idev->dev);
5052
5053         if (event)
5054                 ASSERT_RTNL();
5055
5056         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5057
5058         switch (event) {
5059         case RTM_NEWADDR:
5060                 /*
5061                  * If the address was optimistic
5062                  * we inserted the route at the start of
5063                  * our DAD process, so we don't need
5064                  * to do it again
5065                  */
5066                 if (!(ifp->rt->rt6i_node))
5067                         ip6_ins_rt(ifp->rt);
5068                 if (ifp->idev->cnf.forwarding)
5069                         addrconf_join_anycast(ifp);
5070                 if (!ipv6_addr_any(&ifp->peer_addr))
5071                         addrconf_prefix_route(&ifp->peer_addr, 128,
5072                                               ifp->idev->dev, 0, 0);
5073                 break;
5074         case RTM_DELADDR:
5075                 if (ifp->idev->cnf.forwarding)
5076                         addrconf_leave_anycast(ifp);
5077                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5078                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5079                         struct rt6_info *rt;
5080
5081                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5082                                                        ifp->idev->dev, 0, 0);
5083                         if (rt && ip6_del_rt(rt))
5084                                 dst_free(&rt->dst);
5085                 }
5086                 dst_hold(&ifp->rt->dst);
5087
5088                 if (ip6_del_rt(ifp->rt))
5089                         dst_free(&ifp->rt->dst);
5090
5091                 rt_genid_bump_ipv6(net);
5092                 break;
5093         }
5094         atomic_inc(&net->ipv6.dev_addr_genid);
5095 }
5096
5097 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5098 {
5099         rcu_read_lock_bh();
5100         if (likely(ifp->idev->dead == 0))
5101                 __ipv6_ifa_notify(event, ifp);
5102         rcu_read_unlock_bh();
5103 }
5104
5105 #ifdef CONFIG_SYSCTL
5106
5107 static
5108 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5109                            void __user *buffer, size_t *lenp, loff_t *ppos)
5110 {
5111         int *valp = ctl->data;
5112         int val = *valp;
5113         loff_t pos = *ppos;
5114         struct ctl_table lctl;
5115         int ret;
5116
5117         /*
5118          * ctl->data points to idev->cnf.forwarding, we should
5119          * not modify it until we get the rtnl lock.
5120          */
5121         lctl = *ctl;
5122         lctl.data = &val;
5123
5124         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5125
5126         if (write)
5127                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5128         if (ret)
5129                 *ppos = pos;
5130         return ret;
5131 }
5132
5133 static
5134 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5135                         void __user *buffer, size_t *lenp, loff_t *ppos)
5136 {
5137         struct inet6_dev *idev = ctl->extra1;
5138         int min_mtu = IPV6_MIN_MTU;
5139         struct ctl_table lctl;
5140
5141         lctl = *ctl;
5142         lctl.extra1 = &min_mtu;
5143         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5144
5145         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5146 }
5147
5148 static void dev_disable_change(struct inet6_dev *idev)
5149 {
5150         struct netdev_notifier_info info;
5151
5152         if (!idev || !idev->dev)
5153                 return;
5154
5155         netdev_notifier_info_init(&info, idev->dev);
5156         if (idev->cnf.disable_ipv6)
5157                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5158         else
5159                 addrconf_notify(NULL, NETDEV_UP, &info);
5160 }
5161
5162 static void addrconf_disable_change(struct net *net, __s32 newf)
5163 {
5164         struct net_device *dev;
5165         struct inet6_dev *idev;
5166
5167         rcu_read_lock();
5168         for_each_netdev_rcu(net, dev) {
5169                 idev = __in6_dev_get(dev);
5170                 if (idev) {
5171                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5172                         idev->cnf.disable_ipv6 = newf;
5173                         if (changed)
5174                                 dev_disable_change(idev);
5175                 }
5176         }
5177         rcu_read_unlock();
5178 }
5179
5180 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5181 {
5182         struct net *net;
5183         int old;
5184
5185         if (!rtnl_trylock())
5186                 return restart_syscall();
5187
5188         net = (struct net *)table->extra2;
5189         old = *p;
5190         *p = newf;
5191
5192         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5193                 rtnl_unlock();
5194                 return 0;
5195         }
5196
5197         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5198                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5199                 addrconf_disable_change(net, newf);
5200         } else if ((!newf) ^ (!old))
5201                 dev_disable_change((struct inet6_dev *)table->extra1);
5202
5203         rtnl_unlock();
5204         return 0;
5205 }
5206
5207 static
5208 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5209                             void __user *buffer, size_t *lenp, loff_t *ppos)
5210 {
5211         int *valp = ctl->data;
5212         int val = *valp;
5213         loff_t pos = *ppos;
5214         struct ctl_table lctl;
5215         int ret;
5216
5217         /*
5218          * ctl->data points to idev->cnf.disable_ipv6, we should
5219          * not modify it until we get the rtnl lock.
5220          */
5221         lctl = *ctl;
5222         lctl.data = &val;
5223
5224         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5225
5226         if (write)
5227                 ret = addrconf_disable_ipv6(ctl, valp, val);
5228         if (ret)
5229                 *ppos = pos;
5230         return ret;
5231 }
5232
5233 static
5234 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5235                               void __user *buffer, size_t *lenp, loff_t *ppos)
5236 {
5237         int *valp = ctl->data;
5238         int ret;
5239         int old, new;
5240
5241         old = *valp;
5242         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5243         new = *valp;
5244
5245         if (write && old != new) {
5246                 struct net *net = ctl->extra2;
5247
5248                 if (!rtnl_trylock())
5249                         return restart_syscall();
5250
5251                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5252                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5253                                                      NETCONFA_IFINDEX_DEFAULT,
5254                                                      net->ipv6.devconf_dflt);
5255                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5256                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5257                                                      NETCONFA_IFINDEX_ALL,
5258                                                      net->ipv6.devconf_all);
5259                 else {
5260                         struct inet6_dev *idev = ctl->extra1;
5261
5262                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5263                                                      idev->dev->ifindex,
5264                                                      &idev->cnf);
5265                 }
5266                 rtnl_unlock();
5267         }
5268
5269         return ret;
5270 }
5271
5272 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5273                                          void __user *buffer, size_t *lenp,
5274                                          loff_t *ppos)
5275 {
5276         int err;
5277         struct in6_addr addr;
5278         char str[IPV6_MAX_STRLEN];
5279         struct ctl_table lctl = *ctl;
5280         struct net *net = ctl->extra2;
5281         struct ipv6_stable_secret *secret = ctl->data;
5282
5283         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5284                 return -EIO;
5285
5286         lctl.maxlen = IPV6_MAX_STRLEN;
5287         lctl.data = str;
5288
5289         if (!rtnl_trylock())
5290                 return restart_syscall();
5291
5292         if (!write && !secret->initialized) {
5293                 err = -EIO;
5294                 goto out;
5295         }
5296
5297         if (!write) {
5298                 err = snprintf(str, sizeof(str), "%pI6",
5299                                &secret->secret);
5300                 if (err >= sizeof(str)) {
5301                         err = -EIO;
5302                         goto out;
5303                 }
5304         }
5305
5306         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5307         if (err || !write)
5308                 goto out;
5309
5310         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5311                 err = -EIO;
5312                 goto out;
5313         }
5314
5315         secret->initialized = true;
5316         secret->secret = addr;
5317
5318         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5319                 struct net_device *dev;
5320
5321                 for_each_netdev(net, dev) {
5322                         struct inet6_dev *idev = __in6_dev_get(dev);
5323
5324                         if (idev) {
5325                                 idev->addr_gen_mode =
5326                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5327                         }
5328                 }
5329         } else {
5330                 struct inet6_dev *idev = ctl->extra1;
5331
5332                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5333         }
5334
5335 out:
5336         rtnl_unlock();
5337
5338         return err;
5339 }
5340
5341 static struct addrconf_sysctl_table
5342 {
5343         struct ctl_table_header *sysctl_header;
5344         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5345 } addrconf_sysctl __read_mostly = {
5346         .sysctl_header = NULL,
5347         .addrconf_vars = {
5348                 {
5349                         .procname       = "forwarding",
5350                         .data           = &ipv6_devconf.forwarding,
5351                         .maxlen         = sizeof(int),
5352                         .mode           = 0644,
5353                         .proc_handler   = addrconf_sysctl_forward,
5354                 },
5355                 {
5356                         .procname       = "hop_limit",
5357                         .data           = &ipv6_devconf.hop_limit,
5358                         .maxlen         = sizeof(int),
5359                         .mode           = 0644,
5360                         .proc_handler   = proc_dointvec,
5361                 },
5362                 {
5363                         .procname       = "mtu",
5364                         .data           = &ipv6_devconf.mtu6,
5365                         .maxlen         = sizeof(int),
5366                         .mode           = 0644,
5367                         .proc_handler   = addrconf_sysctl_mtu,
5368                 },
5369                 {
5370                         .procname       = "accept_ra",
5371                         .data           = &ipv6_devconf.accept_ra,
5372                         .maxlen         = sizeof(int),
5373                         .mode           = 0644,
5374                         .proc_handler   = proc_dointvec,
5375                 },
5376                 {
5377                         .procname       = "accept_redirects",
5378                         .data           = &ipv6_devconf.accept_redirects,
5379                         .maxlen         = sizeof(int),
5380                         .mode           = 0644,
5381                         .proc_handler   = proc_dointvec,
5382                 },
5383                 {
5384                         .procname       = "autoconf",
5385                         .data           = &ipv6_devconf.autoconf,
5386                         .maxlen         = sizeof(int),
5387                         .mode           = 0644,
5388                         .proc_handler   = proc_dointvec,
5389                 },
5390                 {
5391                         .procname       = "dad_transmits",
5392                         .data           = &ipv6_devconf.dad_transmits,
5393                         .maxlen         = sizeof(int),
5394                         .mode           = 0644,
5395                         .proc_handler   = proc_dointvec,
5396                 },
5397                 {
5398                         .procname       = "router_solicitations",
5399                         .data           = &ipv6_devconf.rtr_solicits,
5400                         .maxlen         = sizeof(int),
5401                         .mode           = 0644,
5402                         .proc_handler   = proc_dointvec,
5403                 },
5404                 {
5405                         .procname       = "router_solicitation_interval",
5406                         .data           = &ipv6_devconf.rtr_solicit_interval,
5407                         .maxlen         = sizeof(int),
5408                         .mode           = 0644,
5409                         .proc_handler   = proc_dointvec_jiffies,
5410                 },
5411                 {
5412                         .procname       = "router_solicitation_delay",
5413                         .data           = &ipv6_devconf.rtr_solicit_delay,
5414                         .maxlen         = sizeof(int),
5415                         .mode           = 0644,
5416                         .proc_handler   = proc_dointvec_jiffies,
5417                 },
5418                 {
5419                         .procname       = "force_mld_version",
5420                         .data           = &ipv6_devconf.force_mld_version,
5421                         .maxlen         = sizeof(int),
5422                         .mode           = 0644,
5423                         .proc_handler   = proc_dointvec,
5424                 },
5425                 {
5426                         .procname       = "mldv1_unsolicited_report_interval",
5427                         .data           =
5428                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5429                         .maxlen         = sizeof(int),
5430                         .mode           = 0644,
5431                         .proc_handler   = proc_dointvec_ms_jiffies,
5432                 },
5433                 {
5434                         .procname       = "mldv2_unsolicited_report_interval",
5435                         .data           =
5436                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5437                         .maxlen         = sizeof(int),
5438                         .mode           = 0644,
5439                         .proc_handler   = proc_dointvec_ms_jiffies,
5440                 },
5441                 {
5442                         .procname       = "use_tempaddr",
5443                         .data           = &ipv6_devconf.use_tempaddr,
5444                         .maxlen         = sizeof(int),
5445                         .mode           = 0644,
5446                         .proc_handler   = proc_dointvec,
5447                 },
5448                 {
5449                         .procname       = "temp_valid_lft",
5450                         .data           = &ipv6_devconf.temp_valid_lft,
5451                         .maxlen         = sizeof(int),
5452                         .mode           = 0644,
5453                         .proc_handler   = proc_dointvec,
5454                 },
5455                 {
5456                         .procname       = "temp_prefered_lft",
5457                         .data           = &ipv6_devconf.temp_prefered_lft,
5458                         .maxlen         = sizeof(int),
5459                         .mode           = 0644,
5460                         .proc_handler   = proc_dointvec,
5461                 },
5462                 {
5463                         .procname       = "regen_max_retry",
5464                         .data           = &ipv6_devconf.regen_max_retry,
5465                         .maxlen         = sizeof(int),
5466                         .mode           = 0644,
5467                         .proc_handler   = proc_dointvec,
5468                 },
5469                 {
5470                         .procname       = "max_desync_factor",
5471                         .data           = &ipv6_devconf.max_desync_factor,
5472                         .maxlen         = sizeof(int),
5473                         .mode           = 0644,
5474                         .proc_handler   = proc_dointvec,
5475                 },
5476                 {
5477                         .procname       = "max_addresses",
5478                         .data           = &ipv6_devconf.max_addresses,
5479                         .maxlen         = sizeof(int),
5480                         .mode           = 0644,
5481                         .proc_handler   = proc_dointvec,
5482                 },
5483                 {
5484                         .procname       = "accept_ra_defrtr",
5485                         .data           = &ipv6_devconf.accept_ra_defrtr,
5486                         .maxlen         = sizeof(int),
5487                         .mode           = 0644,
5488                         .proc_handler   = proc_dointvec,
5489                 },
5490                 {
5491                         .procname       = "accept_ra_min_hop_limit",
5492                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5493                         .maxlen         = sizeof(int),
5494                         .mode           = 0644,
5495                         .proc_handler   = proc_dointvec,
5496                 },
5497                 {
5498                         .procname       = "accept_ra_pinfo",
5499                         .data           = &ipv6_devconf.accept_ra_pinfo,
5500                         .maxlen         = sizeof(int),
5501                         .mode           = 0644,
5502                         .proc_handler   = proc_dointvec,
5503                 },
5504 #ifdef CONFIG_IPV6_ROUTER_PREF
5505                 {
5506                         .procname       = "accept_ra_rtr_pref",
5507                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5508                         .maxlen         = sizeof(int),
5509                         .mode           = 0644,
5510                         .proc_handler   = proc_dointvec,
5511                 },
5512                 {
5513                         .procname       = "router_probe_interval",
5514                         .data           = &ipv6_devconf.rtr_probe_interval,
5515                         .maxlen         = sizeof(int),
5516                         .mode           = 0644,
5517                         .proc_handler   = proc_dointvec_jiffies,
5518                 },
5519 #ifdef CONFIG_IPV6_ROUTE_INFO
5520                 {
5521                         .procname       = "accept_ra_rt_info_max_plen",
5522                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5523                         .maxlen         = sizeof(int),
5524                         .mode           = 0644,
5525                         .proc_handler   = proc_dointvec,
5526                 },
5527 #endif
5528 #endif
5529                 {
5530                         .procname       = "proxy_ndp",
5531                         .data           = &ipv6_devconf.proxy_ndp,
5532                         .maxlen         = sizeof(int),
5533                         .mode           = 0644,
5534                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5535                 },
5536                 {
5537                         .procname       = "accept_source_route",
5538                         .data           = &ipv6_devconf.accept_source_route,
5539                         .maxlen         = sizeof(int),
5540                         .mode           = 0644,
5541                         .proc_handler   = proc_dointvec,
5542                 },
5543 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5544                 {
5545                         .procname       = "optimistic_dad",
5546                         .data           = &ipv6_devconf.optimistic_dad,
5547                         .maxlen         = sizeof(int),
5548                         .mode           = 0644,
5549                         .proc_handler   = proc_dointvec,
5550
5551                 },
5552                 {
5553                         .procname       = "use_optimistic",
5554                         .data           = &ipv6_devconf.use_optimistic,
5555                         .maxlen         = sizeof(int),
5556                         .mode           = 0644,
5557                         .proc_handler   = proc_dointvec,
5558
5559                 },
5560 #endif
5561 #ifdef CONFIG_IPV6_MROUTE
5562                 {
5563                         .procname       = "mc_forwarding",
5564                         .data           = &ipv6_devconf.mc_forwarding,
5565                         .maxlen         = sizeof(int),
5566                         .mode           = 0444,
5567                         .proc_handler   = proc_dointvec,
5568                 },
5569 #endif
5570                 {
5571                         .procname       = "disable_ipv6",
5572                         .data           = &ipv6_devconf.disable_ipv6,
5573                         .maxlen         = sizeof(int),
5574                         .mode           = 0644,
5575                         .proc_handler   = addrconf_sysctl_disable,
5576                 },
5577                 {
5578                         .procname       = "accept_dad",
5579                         .data           = &ipv6_devconf.accept_dad,
5580                         .maxlen         = sizeof(int),
5581                         .mode           = 0644,
5582                         .proc_handler   = proc_dointvec,
5583                 },
5584                 {
5585                         .procname       = "force_tllao",
5586                         .data           = &ipv6_devconf.force_tllao,
5587                         .maxlen         = sizeof(int),
5588                         .mode           = 0644,
5589                         .proc_handler   = proc_dointvec
5590                 },
5591                 {
5592                         .procname       = "ndisc_notify",
5593                         .data           = &ipv6_devconf.ndisc_notify,
5594                         .maxlen         = sizeof(int),
5595                         .mode           = 0644,
5596                         .proc_handler   = proc_dointvec
5597                 },
5598                 {
5599                         .procname       = "suppress_frag_ndisc",
5600                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5601                         .maxlen         = sizeof(int),
5602                         .mode           = 0644,
5603                         .proc_handler   = proc_dointvec
5604                 },
5605                 {
5606                         .procname       = "accept_ra_from_local",
5607                         .data           = &ipv6_devconf.accept_ra_from_local,
5608                         .maxlen         = sizeof(int),
5609                         .mode           = 0644,
5610                         .proc_handler   = proc_dointvec,
5611                 },
5612                 {
5613                         .procname       = "accept_ra_mtu",
5614                         .data           = &ipv6_devconf.accept_ra_mtu,
5615                         .maxlen         = sizeof(int),
5616                         .mode           = 0644,
5617                         .proc_handler   = proc_dointvec,
5618                 },
5619                 {
5620                         .procname       = "stable_secret",
5621                         .data           = &ipv6_devconf.stable_secret,
5622                         .maxlen         = IPV6_MAX_STRLEN,
5623                         .mode           = 0600,
5624                         .proc_handler   = addrconf_sysctl_stable_secret,
5625                 },
5626                 {
5627                         .procname       = "use_oif_addrs_only",
5628                         .data           = &ipv6_devconf.use_oif_addrs_only,
5629                         .maxlen         = sizeof(int),
5630                         .mode           = 0644,
5631                         .proc_handler   = proc_dointvec,
5632
5633                 },
5634                 {
5635                         /* sentinel */
5636                 }
5637         },
5638 };
5639
5640 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5641                 struct inet6_dev *idev, struct ipv6_devconf *p)
5642 {
5643         int i;
5644         struct addrconf_sysctl_table *t;
5645         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5646
5647         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5648         if (!t)
5649                 goto out;
5650
5651         for (i = 0; t->addrconf_vars[i].data; i++) {
5652                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5653                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5654                 t->addrconf_vars[i].extra2 = net;
5655         }
5656
5657         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5658
5659         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5660         if (!t->sysctl_header)
5661                 goto free;
5662
5663         p->sysctl = t;
5664         return 0;
5665
5666 free:
5667         kfree(t);
5668 out:
5669         return -ENOBUFS;
5670 }
5671
5672 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5673 {
5674         struct addrconf_sysctl_table *t;
5675
5676         if (!p->sysctl)
5677                 return;
5678
5679         t = p->sysctl;
5680         p->sysctl = NULL;
5681         unregister_net_sysctl_table(t->sysctl_header);
5682         kfree(t);
5683 }
5684
5685 static int addrconf_sysctl_register(struct inet6_dev *idev)
5686 {
5687         int err;
5688
5689         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5690                 return -EINVAL;
5691
5692         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5693                                     &ndisc_ifinfo_sysctl_change);
5694         if (err)
5695                 return err;
5696         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5697                                          idev, &idev->cnf);
5698         if (err)
5699                 neigh_sysctl_unregister(idev->nd_parms);
5700
5701         return err;
5702 }
5703
5704 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5705 {
5706         __addrconf_sysctl_unregister(&idev->cnf);
5707         neigh_sysctl_unregister(idev->nd_parms);
5708 }
5709
5710
5711 #endif
5712
5713 static int __net_init addrconf_init_net(struct net *net)
5714 {
5715         int err = -ENOMEM;
5716         struct ipv6_devconf *all, *dflt;
5717
5718         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5719         if (!all)
5720                 goto err_alloc_all;
5721
5722         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5723         if (!dflt)
5724                 goto err_alloc_dflt;
5725
5726         /* these will be inherited by all namespaces */
5727         dflt->autoconf = ipv6_defaults.autoconf;
5728         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5729
5730         dflt->stable_secret.initialized = false;
5731         all->stable_secret.initialized = false;
5732
5733         net->ipv6.devconf_all = all;
5734         net->ipv6.devconf_dflt = dflt;
5735
5736 #ifdef CONFIG_SYSCTL
5737         err = __addrconf_sysctl_register(net, "all", NULL, all);
5738         if (err < 0)
5739                 goto err_reg_all;
5740
5741         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5742         if (err < 0)
5743                 goto err_reg_dflt;
5744 #endif
5745         return 0;
5746
5747 #ifdef CONFIG_SYSCTL
5748 err_reg_dflt:
5749         __addrconf_sysctl_unregister(all);
5750 err_reg_all:
5751         kfree(dflt);
5752 #endif
5753 err_alloc_dflt:
5754         kfree(all);
5755 err_alloc_all:
5756         return err;
5757 }
5758
5759 static void __net_exit addrconf_exit_net(struct net *net)
5760 {
5761 #ifdef CONFIG_SYSCTL
5762         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5763         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5764 #endif
5765         kfree(net->ipv6.devconf_dflt);
5766         kfree(net->ipv6.devconf_all);
5767 }
5768
5769 static struct pernet_operations addrconf_ops = {
5770         .init = addrconf_init_net,
5771         .exit = addrconf_exit_net,
5772 };
5773
5774 static struct rtnl_af_ops inet6_ops __read_mostly = {
5775         .family           = AF_INET6,
5776         .fill_link_af     = inet6_fill_link_af,
5777         .get_link_af_size = inet6_get_link_af_size,
5778         .validate_link_af = inet6_validate_link_af,
5779         .set_link_af      = inet6_set_link_af,
5780 };
5781
5782 /*
5783  *      Init / cleanup code
5784  */
5785
5786 int __init addrconf_init(void)
5787 {
5788         struct inet6_dev *idev;
5789         int i, err;
5790
5791         err = ipv6_addr_label_init();
5792         if (err < 0) {
5793                 pr_crit("%s: cannot initialize default policy table: %d\n",
5794                         __func__, err);
5795                 goto out;
5796         }
5797
5798         err = register_pernet_subsys(&addrconf_ops);
5799         if (err < 0)
5800                 goto out_addrlabel;
5801
5802         addrconf_wq = create_workqueue("ipv6_addrconf");
5803         if (!addrconf_wq) {
5804                 err = -ENOMEM;
5805                 goto out_nowq;
5806         }
5807
5808         /* The addrconf netdev notifier requires that loopback_dev
5809          * has it's ipv6 private information allocated and setup
5810          * before it can bring up and give link-local addresses
5811          * to other devices which are up.
5812          *
5813          * Unfortunately, loopback_dev is not necessarily the first
5814          * entry in the global dev_base list of net devices.  In fact,
5815          * it is likely to be the very last entry on that list.
5816          * So this causes the notifier registry below to try and
5817          * give link-local addresses to all devices besides loopback_dev
5818          * first, then loopback_dev, which cases all the non-loopback_dev
5819          * devices to fail to get a link-local address.
5820          *
5821          * So, as a temporary fix, allocate the ipv6 structure for
5822          * loopback_dev first by hand.
5823          * Longer term, all of the dependencies ipv6 has upon the loopback
5824          * device and it being up should be removed.
5825          */
5826         rtnl_lock();
5827         idev = ipv6_add_dev(init_net.loopback_dev);
5828         rtnl_unlock();
5829         if (IS_ERR(idev)) {
5830                 err = PTR_ERR(idev);
5831                 goto errlo;
5832         }
5833
5834         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5835                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5836
5837         register_netdevice_notifier(&ipv6_dev_notf);
5838
5839         addrconf_verify();
5840
5841         rtnl_af_register(&inet6_ops);
5842
5843         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5844                               NULL);
5845         if (err < 0)
5846                 goto errout;
5847
5848         /* Only the first call to __rtnl_register can fail */
5849         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5850         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5851         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5852                         inet6_dump_ifaddr, NULL);
5853         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5854                         inet6_dump_ifmcaddr, NULL);
5855         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5856                         inet6_dump_ifacaddr, NULL);
5857         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5858                         inet6_netconf_dump_devconf, NULL);
5859
5860         ipv6_addr_label_rtnl_register();
5861
5862         return 0;
5863 errout:
5864         rtnl_af_unregister(&inet6_ops);
5865         unregister_netdevice_notifier(&ipv6_dev_notf);
5866 errlo:
5867         destroy_workqueue(addrconf_wq);
5868 out_nowq:
5869         unregister_pernet_subsys(&addrconf_ops);
5870 out_addrlabel:
5871         ipv6_addr_label_cleanup();
5872 out:
5873         return err;
5874 }
5875
5876 void addrconf_cleanup(void)
5877 {
5878         struct net_device *dev;
5879         int i;
5880
5881         unregister_netdevice_notifier(&ipv6_dev_notf);
5882         unregister_pernet_subsys(&addrconf_ops);
5883         ipv6_addr_label_cleanup();
5884
5885         rtnl_lock();
5886
5887         __rtnl_af_unregister(&inet6_ops);
5888
5889         /* clean dev list */
5890         for_each_netdev(&init_net, dev) {
5891                 if (__in6_dev_get(dev) == NULL)
5892                         continue;
5893                 addrconf_ifdown(dev, 1);
5894         }
5895         addrconf_ifdown(init_net.loopback_dev, 2);
5896
5897         /*
5898          *      Check hash table.
5899          */
5900         spin_lock_bh(&addrconf_hash_lock);
5901         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5902                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5903         spin_unlock_bh(&addrconf_hash_lock);
5904         cancel_delayed_work(&addr_chk_work);
5905         rtnl_unlock();
5906
5907         destroy_workqueue(addrconf_wq);
5908 }